Power of the people or the expert? The influence of vox pop and expert statements on news-item evaluation, perceived public opinion, and personal opinion

News item evaluation

Increased competition, the battle for audience shares, and around-the-clock deadlines are often seen as factors that push broadcasters to continuously try to attract audiences (Paulussen and Ugille, 2008; Saltzis and Dickinson, 2008; Witschge and Nygren, 2009). One of the strategies to do so would be to make the news more personal and to focus on a human-interest approach (Hendriks Vettehen, Nuijten, and Beentjes, 2005; Hendriks Vettehen, Beentjes, Nuijten, and Peeters, 2010; Kleemans et al., 2017). Instead of using elite sources with more authority, the media would increasingly turn to ordinary citizens, who are easy for audiences to identify with, although elites still dominate the news for most issues. Despite the fact that a survey with Flemish journalists found that journalists think vox pops are used too frequently in the news (Beckers, 2018), several content analyses found vox pops were increasingly taking away airtime from traditional elites such as experts and politicians (Hendriks Vettehen et al., 2010; Kleemans et al., 2017). It thus seems that journalists regularly use vox pops in their news coverage. This is often explained by the assumption that they make news items more attractive to audiences (Brants and Van Praag, 2006; Brosius, 2003; D’Alessio, 2003; Hopmann, Elmelundpræstekær, Albæk, and De Vreese, 2009; Kleemans et al., 2017; Zillmann and Brosius, 2000). However, so far only assumptions exist about the perceived attraction of vox pops to audiences compared to other forms of public opinion information. It is not known whether this is indeed the case, and this study will be the first to compare the influence of vox pops exemplifying public opinion on news item attractiveness to an elite source, that is, an expert source presenting public opinion information.

Vox pops’ influence in the news is often ascribed to their apparent recognizability and authenticity. They would be perceived by audiences as trustworthy because they have the closest proximity to the audience as a representation of the general public in the news and provide news with some sort of legitimacy (Brosius, 2003; D’Alessio, 2003; Zillmann and Brosius, 2000). Vox pops draw their right to speak in television news from their authentic experiences and honest reactions because they have no direct involvement with a news issue (Carpentier, 2011; Lewis et al., 2005, p. 85). However, few studies have investigated whether vox pops are indeed perceived by audiences as being trustworthy. The first to systematically test this were Lefevere, Walgrave and De Swert (2012) in an experiment in which they compared vox pops and expert statements. They found that, against their expectations, news items containing expert statements were perceived as being equally trustworthy as news items containing vox pops. This might be explained by the fact that compared to other elite sources such as politicians or businesspeople, experts have fewer reasons to speak in the news with a certain agenda. There are therefore reasons to expect that the inclusion of both vox pops and expert statements increase a news item’s trustworthiness. As little evidence exists on the differences in trustworthiness between vox pop statements and public opinion information put forward by an expert and the spillover to the perceived trustworthiness of the news item, we formulate the following research question:

Perceived public opinion and personal opinion

For most people, news media are their main source of information to learn about public opinion on societal issues, so how journalists report on the public sentiment is important (Busselle and Shrum, 2003; Gunther, 1998; Moy and Scheufele, 2000). Vox pops are one of the most prevalent ways to refer to public opinion in television news, comprising four out of ten references to public opinion in UK and US news broadcasts (Lewis et al., 2005), and making up almost half of all types of citizen participation in Spanish television news (Bergillos, 2019). However, several other ways exist in which media can present public opinion in the news, such as public opinion polls and general inferences to public opinion (e. g., “the people are angry”).

Most research has studied the public opinion influence of vox pops by comparing them to base-rate information, which is usually based on more valid and reliable empirical data. Results of these studies have almost consistently indicated that (numerical and non-numerical) base-rate information has almost no impact whereas vox pops were found to influence people’s judgments and opinions (Brosius and Bathelt, 1994; Gibson and Zillmann, 1993, 1994; Zillmann and Brosius, 2000; Zillmann, Perkins, and Sundar, 1992). These effects were established for several news issues, (Brosius and Bathelt, 1994), even for morally loaded (Zerback and Fawzi, 2017) or politically important topics (i. e., elections; Daschmann, 2000). Moreover, in previous studies, vox pops were compared to other public opinion displays such as opinion polls (e. g., Daschmann, 2000) and general inferences (Zillmann, Gibson, Sundar, and Perkins, 1996). In these studies, vox pops were found to be influential whereas the other public opinion information had almost no impact. However, in most cases, the base-rate information was provided in a rather static manner (e. g., providing the information in a printed text without the use of quotes). Most studies did not ‘vivify’ the base-rate information.

A few studies did manipulate the vividness of this base-rate information, but they did not use other forms of direct quotation. Brosius and Bathelt (1994), for instance, manipulated the vividness of the exemplars’ speech by using emotional quotes compared to monotonous and pallid quotes. Additionally, they compared more vivid, directly-quoted interviews with a journalist paraphrasing the information from the exemplars. While they did not find an effect of the vividness of the language used, they did find that respondents’ perceived public opinion and personal opinion were influenced more in the direction of the vivid direct interviews compared to the journalist paraphrasing the same exemplar statements. It thus seems that the fact that exemplars are directly quoted might play a role in their influence. Other studies also found directly-quoted exemplars of ordinary citizens to be more influential than similar indirect citations (Gibson and Zillmann, 1993, 1998). Consequently, it might be that base-rate information is influential when vivified, for instance, when explained by an expert.

Yet, there are other reasons to expect that vox pops are still more influential than vivified base-rate information on perceptions of public opinion and personal opinion. The effect of vox pops is often grounded in exemplification theory, which is based on heuristic information processing. The basic premise is that people do not have the time and/or motivation to process all information in the news attentively, and therefore often rely on simple heuristics, causing them to make cognitive shortcuts (Tversky and Kahneman, 1973). Of relevance for this study is what is called the “base-rate fallacy” (Bar-Hillel, 1980). When general information (e. g., statistics) and specific information (e. g., examples) are presented, people have the tendency to focus on the specific and neglect the general. They have difficulty interpreting the base-rate information and evaluating its diagnostic value (Tversky and Kahneman, 1971). So, people often do not base their decisions on concrete numbers but on specific cases such as vox pops. A related heuristic that applies to the context of comparing vox pops with expert statements is the representativeness heuristic. People are bad at interpreting samples and tend to generalize specific examples to the broader population. When doing so, people neglect other information about the general probability or occurrence of events.

Besides the representativeness heuristic, what is referred to as availability is often used to explain the almost consistent influence of exemplars over other information: People’s perceptions and judgements are based to a large degree on information that is easy to retrieve from memory (Tversky and Kahneman, 1973). In exemplification research, availability is often operationalized in terms of vividness: More vivid information would be easier to process and retrieve from memory when people are forming judgments. Several studies concluded that regardless of how clear or explicit the base-rate information was about the direction of public opinion, people followed the direction of the vox pop viewpoints. However, in most of these studies, base-rate information was kept constant and was presented in a static manner (e. g., Zillmann and Brosius, 2000). In our experiment, we take away the vividness advantage of the vox pops and present the base-rate information in an equally vivid manner.

Lefevere, De Swert, and Walgrave (2012) were the first (and to our knowledge the only researchers) to compare the influence of directly-quoted vox pops with directly-quoted experts and politicians on people’s personal opinion. They found respondents to be influenced least by the politicians’ quotes. Moreover, the vox pop (they called them “popular exemplars”) quotes were found to be more influential than the public opinion information given by the expert. Based on this empirical evidence and the different heuristics that are expected to come into play, we hypothesize that people do not take decisions based on quantified probabilities, even when they are vivified through an expert-quote, but based on specific cases such as vox pops. People will perceive the small sample of vox pops as an implicit representation of the public in the news and generalize the information they give to the entire population (Lewis et al., 2005; Myers, 2004). This is why we hypothesize the vox pop statements to be more influential than the expert statement, both for audience perceptions and opinions.

Stimulus materials

The stimulus news item covered a story about an ongoing discussion in the Flemish Parliament about investments in traffic infrastructure. The news anchor explains that due to budget cuts, the government will have to choose between the maintenance of regional highways or further investments in bicycle highways in the short term. After the introductory text by the news anchor, the voice-over further explains the choice the government will have to make, while footage of bicycle highways and regional highways is shown on-screen.

After the identical introduction, depending on the condition, interviews with different types of sources ensued (four vox pops or one expert), giving information about public opinion. The vox pop news clips contained vox pops that were diverse in terms of age and gender, making them as realistic as possible. Half of the interviewees were male, the other two female. Moreover, all vox pops came from different age groups and were displayed without a caption, as is also the case in the real news broadcasts. The expert was a 39-year-old man, accompanied with the caption “traffic expert” and filmed in front of a banner of the Flemish Traffic Foundation (Vlaamse Stichting Verkeerskunde, VSV).

The speaking sources either favored bicycle highways (“pro-bike”) or regional highways (“pro-car”). In the vox pop conditions, all four vox pops gave either pro-bike opinions (e. g., “I think they should prioritize weak road users, so bicyclists” or “I think the government should invest in bicycle highways first”), or pro-car (e. g., “On the news they are always talking about traffic jams. I think they need to take a look at the roads first” or “They should start paying attention to the Flemish roads. I think that is more important than a bicycle highway”). The expert talked about a survey which the VSV had conducted: “As VSV, we conducted a survey with a sample of Flemings to get to know their opinions about investments in safe road infrastructure. From the results it became clear that the Flemings would rather invest in bicycle highways [regional highways]. 64 % of Flemings indicated their preference for investments in additional bicycle highways [regional highways] over investments in regional roads [bicycle highways]. So, the government faces an important choice.”

Four vox pop statements were compared to one expert statement. This choice was made to keep the external validity of the news item as high as possible, in line with the study conducted by Lefevere, De Swert, and Walgrave (2012). In real news items, a single vox pop is rather rare, and it is also unlikely that results from more than one survey are presented. However, this choice for realistic news items decreases our internal validity. To minimize this confounding factor, we tried to keep the length of the news items similar across conditions. The news items containing the expert statements were on average longer (75 seconds) than the news items containing the vox pops (58 seconds). Differences existed due to variations in speaking speed and pauses between sentences, yet the actual scripts were as identical as possible.

The experiment uses a 2 (public opinion source: expert or vox pops) x 2 (viewpoint: pro-bike or pro-car) design. We chose to include only pro and contra viewpoint statements, as the study of Beckers, Walgrave, and Van den Bulck (2018) found that the large majority (73 %) of news items only presented one viewpoint through vox pops. In Table 1, an overview of the different conditions and the number of participants in each condition can be found.

Several mechanisms were used to control for actual exposure to the news clips. First, the time (in seconds) spent on the page displaying the news clip was measured and the button to go to the next page appeared after one minute. Second, the playback control buttons were disabled, making it impossible to skip forward or go back in the news item. Only participants who watched the entire news clip were included in the analyses.

Dependent variables

We measured news item attractiveness using the question: “We are interested in your opinion about the news item you just saw. Please indicate to what extent you thought the news item met the following characteristics: The news item was: ‘attractive’.” This question was followed by a five-point scale (going from 1 = totally disagree to 5 = totally agree). News item trustworthiness was measured using the same 5-point scale containing three items: ‘trustworthy’; ‘honest’ and ‘credible’ (based on: Hovland and Weiss, 1951; Ohanian, 1990). The scale was found to be reliable, Cronbach’s α = 0.88. Higher scores on this variable indicate a higher perceived trustworthiness of the news item.

Next, we measured participants’ personal opinion using the question: “We are interested in your own opinion. What do you prefer: investing in bicycle highways or investing in highways?”, followed by a 7-point scale (going from strongly in favor of investing in cycling highways [pro-bike] to strongly in favor of investing in regional highways [pro-car]). Higher scores on this variable indicate a personal opinion more pro-car. After this question about participants’ personal opinion, participants’ perceived public opinion was measured by asking: “Next, when you think of the Flemish population, what do you think is the preference of the majority of the Flemings; investing in bicycle highways or investing in regional highways?”, using the same 7-point scale. Higher scores on this variable indicate a perception of public opinion as being more pro-car. Across conditions, participants’ perceived public opinion is more pro-car (M = 4.49, SD = 1.59) than their personal opinion, which is more pro-bike (M = 2.96, SD = 1.79).

Sample and procedure

The study was completed by 749 participants belonging to a panel (University of Antwerp web-based electoral panel), which contains a diverse group of Flemish citizens in terms of gender, age, education, and political orientation. Of these 749 participants, 69.3 % are male, 30.7 % female. 75.5 % of the sample has finished higher education after secondary school (university or non-university), so male, highly-educated people are overrepresented. Respondents are on average 54 years old (SD = 14.11). Political leaning was measured using an 11-point left–right ideology self-placement scale with the following question: “In politics, the terms ‘left’ and ‘right’ are often used. Could you describe your own views on a scale from 0 to 10, in which 0 indicates ‘left’, 5 ‘center’ and 10 ‘right’?”. The mean score for this variable is 5.71 with a standard deviation of 2.54.

Respondents were assigned randomly to the different experimental conditions, and there are no differences between the conditions for the mentioned socio-demographics. There is also no difference between conditions in the left–right leaning of the participants. Participants were told that the research investigated the quality of Flemish television news. After watching the stimulus clip and completing the questionnaire with the dependent variables, participants were debriefed about the real goal of the experiment.

Manipulation check

Before initiating the experiment, a manipulation check was conducted to test whether the experimental manipulations succeeded. 62 people with diverse characteristics for age and gender participated in the manipulation check. First, respondents were questioned about the realism of the news item, using a 10-point scale going from totally unrealistic (1) to totally realistic (10). The news items were rated as satisfactorily realistic: M = 7.46, SD = 1.78. Respondents also had to indicate how many sources were interviewed in the news item. In the vox pop conditions, 30 people (96.8 %) correctly identified the number as “4”, one respondent answered “3”. In the expert conditions, 29 respondents (96.7 %) correctly answered “1”, one person incorrectly indicated “3”. All people in the pro-car conditions correctly indicated the viewpoint in the news item to be pro-car. In the pro-bike conditions, 97.7 % of the respondents correctly indicated the viewpoint to be pro-bike. People exposed to the expert conditions were asked whether they remembered the function of the expert. Six people did not answer the question, all other people (25) mentioned “VSV” or function descriptions such as “(traffic) expert” or “researcher”. The manipulations thus succeeded and were sufficient for further analysis.

News item evaluation

The results for news item attractiveness are displayed in Appendix A. Of the socio-demographic variables, age is found to be the only one influencing news-item attractiveness: Older participants perceived the news item as being more attractive compared to younger ones, F(1, 512) = 4.11, p < 0.05, η² = 0.008. None of the other socio-demographic variables has a significant influence on how attractive the news item is deemed by the participants. Participants’ political leaning also has no influence. There is a main effect of viewpoint on perceived news-item attractiveness. In general, respondents are more positive about the pro-bike news items (M = 3.27, SD = 0.82) than about the pro-car items (M = 3.04, SD = 0.87), probably because this is closer to their own viewpoints on the topic, F(1, 512) = 11.57, p < 0.001, η² = 0.022. H1 held that audiences would perceive a news item with vox pops as more attractive than a news item containing an expert. However, as also visualized in Figure 1, we find that respondents perceive the news items containing the expert-statement as being more attractive (M = 3.26, SD = 0.78) compared to the news item containing vox pops (M = 3.03, SD = 0.78), independent of the viewpoint that was given in the news item; F(1, 512) = 12.81, p < 0.001, η² = 0.024. So, against our expectations, news items containing vox pops are not perceived as being more attractive by the participants than news items containing expert statements providing statistical information. Our first hypothesis, consequently, has to be rejected.

Figure 1:

Perceived attractiveness of news items based on public opinion information and viewpoint (1 = not attractive at all; 5 = very attractive).

Regarding the perceived trustworthiness of the news item (Appendix A), we find no differences between different socio-demographic groups. We do find a main effect of viewpoint: People who saw a news item containing pro-bike viewpoints perceive the news items as being significantly more trustworthy (M = 3.46, SD = 0.87) than people exposed to a news item containing pro-car viewpoints (M = 3.32, SD = 0.92), F(1, 511) = 4.89); p < 0.05, η² = 0.009. However, the largest effect size is found for the nature of the public opinion information: F(1, 511) = 53.15); p < 0.001, η² = 0.094. When an expert discussed statistical public opinion information, participants rate the news item as significantly more trustworthy (M = 3.64, SD = 0.79) than when vox pops gave their point of view (M = 3.14, SD = 0.94). The results are visualized in Figure 2, where it can be seen that the expert news items receive consistently higher trustworthiness scores compared to the vox pop items.

Figure 2:

Perceived trustworthiness of news items based on public opinion information and viewpoint (1 = not trustworthy at all; 5 = very trustworthy).

Perceived public opinion

Our second hypothesis states that vox pop statements influence participants’ perceived public opinion more than expert statements containing statistical information. The results of this analysis can be found in Appendix B. Here, we also do not find effects of the socio-demographic variables. As expected, the nature of the public opinion information (expert or vox pops) as such does not significantly affect perceptions of public opinion. We do find a strong main effect of viewpoint: Participants exposed to pro-car viewpoints perceive public opinion to be significantly more pro-car (M = 5.10, SD = 0.10) than participants exposed to pro-bike viewpoints (M = 3.81, SD = 0.10), regardless of the source giving the information; F(1, 519) = 118.510); p < 0.001, η² = 0.186. People are influenced in the direction of the viewpoint given, both in the vox pop and the expert conditions. When including the interaction terms in Model II, we do not find the hypothesized (H2) interaction effect of public opinion information (expert or vox pops) and viewpoint. There is no difference in perceptions after exposure to pro-bike viewpoints given by vox pops (M = 3.89, SD = 0.13) or an expert (M = 3.73, SD = 0.13). The same is true for respondents exposed to the pro-car viewpoint (M_{vox pops =} 5.10, SD_{vox pops =} 0.13; M_expert = 5.11, SD_expert = 0.13), F(1, 519) = 0.574); p > 0.05, η² = 0.001. It is thus not the case that vox pops implicitly representing public opinion are more influential than an expert giving explicit public opinion information. So, both the vox pop and expert accounts affect perceived public opinion but are found to be equally influential, as can be seen in Figure 3.

Figure 3:

Perceived public opinion based on public opinion information (vox pops or expert) and viewpoint (1 = strongly pro-bike; 7 = strongly pro-car).

Personal opinion

Our third and last hypothesis predicts that vox pops affect participants’ personal opinions more than expert statements providing statistical public opinion information. The table in Appendix C again shows that age, gender, and education do not exert an effect. However, while the left–right placement of participants did not matter in explaining perceptions of public opinion, we do find a main effect on personal opinion. The more left-wing participants are, the more pro-bike their personal opinion. This is the strongest predictor explaining differences in personal opinion: F(1, 519) = 31.04, p < 0.001, η² = 0.056. Similar to the results for perceived public opinion, we find a main effect of viewpoint on participants’ personal opinion, as can be seen in model I. Respondents exposed to pro-car viewpoints are significantly more pro-car after exposure to the news item (M = 3.19, SD = 0.11) than respondents exposed to pro-bike viewpoints (M = 2.85, SD = 0.12), F(1, 519) = 10.36, p < 0.001, η² = 0.02.

While we did not find an interaction effect of news source and viewpoint on perceived public opinion, we do find this interaction to exert a significant influence on people’s personal opinion, as can be seen in model II. This interaction is visualized in Figure 4. This graph shows that there is no difference between the expert conditions; people’s personal opinion does not differ between conditions containing the pro-car expert (M = 3.12, SD = 0. 15) or the pro-bike expert (M = 3.13, SD = 0.16). There is an effect of the vox pop viewpoints on participants’ personal opinion: people exposed to pro-car vox pops (M = 3.26, SD = 0.15) were significantly more pro-car than participants exposed to pro-bike vox pops (M = 2.57, SD = 0.16); F(1, 519) = 5.231, p < 0.05, η² = 0.01. Hence, people’s personal opinion is influenced more by vox pop statements than by the expert statements, especially for pro-bike statements, in line with our third hypothesis. It thus seems that it is not merely the viewpoint that matters, but who gives the viewpoint about public opinion and how matters as well. Vox pop statements have a stronger influence on people’s personal opinion than expert statements providing statistical information.

Figure 4:

Interaction effect of public opinion information (vox pops or expert) and viewpoint on personal opinion (1 = strongly pro-bike; 7 = strongly pro-car).