Election polling news provides helpful information for voters to make a political decision during election campaigns (Rothenberg, 2014). Despite numerous criticisms, such as misleading voters’ estimation of the public’s mood and attitudes toward candidates and reducing their desire to vote, election polling news is perceived to influence political behaviors and the outcome of elections (Perryman et al., 2020). The perceived effects of election polling news have been studied from the perspective of third-person perception (TPP), which explains people’s tendency to perceive other voters would be influenced by polling results while they themselves are less influenced (Wei et al., 2011).

A plethora of studies across various media contents and communication contexts have confirmed the TPP since Davison’s seminal study of third-person effect, which addresses TPP and the consequences of TPP (Davison, 1983). TPP is likely to be greater when media content is not perceived to be personally beneficial or socially desirable (Comstock & Scharrer, 2005). When a media message is perceived as socially desirable, individuals are likely to show less TPP or the first-person perception (FPP) that they will be more influenced by the message than others will (Duck et al., 1995; Lim et al., 2020).

Research shows the perception that polling news would exert an influence on the self and on other voters may have an actual influence on the perceivers’ behaviors (Wei et al., 2011). Further, emotions may play a role in the process in which perceived effects of election polling news influence voters’ attitudes and behavior in that emotions generated by cognitive appraisals of events are intimately linked to subsequent cognition and action (Frijda, 2010). Previous research demonstrates that indignation mediates perception of media bias and willingness to engage in discursive activities on the perceived media bias (Hwang et al., 2008). In the current study, we investigate how the third- and first-person perceptions of election polling news each influence voters’ political behaviors through emotions in the context of the 2022 South Korean local elections. Given that previous research on the behavioral effects of media perceptions has focused on the third-person perception of negative media effects, the current study on the first-person perception of positive effects of election polling news and emotions should contribute to expanding the third- and first-person effects research.

Literature Review

Some mediated messages, such as direct-to-consumer pharmaceutical advertising, are perceived to have both positive and negative effects on people (Huh et al., 2004). Polling news is perceived to be useful, but at the same time it is thought to have undesirable effects. Election polling news is often criticized as horse-race journalism (Bradburn & Sudman, 1988), and voters tend to believe that election polling news is biased against their preferred candidate (Perryman et al., 2020). In that voters may perceive polling news as having both beneficial and harmful effects on others and the self, the TPP of negative effects of polling news and the FPP of positive effects of polling news may affect their attitudes and behavior together. Indeed, previous research indicates that TPP is a significant predictor of corrective actions, such as support for restrictions on media content (Wei et al., 2011). Based on previous research, we hypothesize that the TPP of election polling news is positively related to support for restrictions on election polling news.

Hypothesis 1 (H1): Third-person perception of election polling news is positively related to support for restrictions on election polling news.

Previous research also demonstrates that when the effects of media content are perceived to be desirable as compared to when they are perceived to be undesirable, people are more likely to estimate a greater influence of the media content on the self than on others (Golan & Day, 2008). Lim et al. (2020) report that the FPP becomes larger as the message desirability of public service announcement increases. Those who believe election polling news is informative and beneficial may perceive a greater influence of the positive effect on the self than on others and believe that they become more knowledgeable after being exposed to polling news, which will lead to increased confidence in their choice of candidate (Kaid, 2001). Thus, a positive relationship between FPP and reinforcement of candidate support is hypothesized.

Hypothesis 2 (H2): First-person perception of election polling news is positively related to reinforcement of support for the preferred candidate.

Emotions may mediate the associations between the TPP and FPP of election polling news and behavioral consequences of the perceptions. According to the cognitive appraisal theory of emotion, the relationship between cognitive appraisals of an event and emotions depends on individuals’ evaluation of what kind of benefit or harm is involved and how important the event is for their personal well-being (Frijda et al., 1989; Lazarus, 1991). When exposed to election polling news, a person may evaluate the news in terms of its implications for their well-being, which triggers certain emotions. In the current study, we focus on anxiety and pride as the emotions elicited by the TPP of negative effects of election polling news and the FPP of positive effects of election polling news.

Previous research indicates TPP of election polling news may elicit anxiety (Kim, 2016). In an anecdote introduced in the seminal study of third-person effect, American officers made the decision to withdraw African American troops in Iwo Jima during World War II because they were worried about the perceived influence of the Japanese leaflets on the African American troops (Davison, 1983). The emotional state of uneasiness from the perception that a message has undesirable influence on other people is identified as anxiety because anxiety is anticipatory and occurs under conditions of ambiguity (Lazarus, 1991). In that people can experience anxiety due to socio-political concerns (Eysenck, 1992), the voters’ perception that election polling news would have undesirable effects on other voters may cause anxiety. Also, anxiety can produce an overestimation of risk, which, in turn leads to risk averse and/or avoidance behavior (Huddy et al., 2007; Lazarus, 1991). In the context of the current study, those who are more anxious about the undesirable effects on others will be more likely to support restrictions on election polling news. Hence, the following hypotheses are put forth:

Hypothesis 3 (H3): Third-person perception of election polling news is positively related to anxiety.

Hypothesis 4 (H4): Anxiety predicts support for restrictions on election polling news.

Based on the literature reviewed for H3 and H4, a hypothesis predicting that anxiety mediates the relationship between TPP of election polling news and support for restrictions on election polling news is proposed:

Hypothesis 5 (H5): Third-person perception of election polling news is positively related to support for restrictions on election polling news through anxiety.

On the other hand, the FPP of election polling news is expected to elicit positive emotions, such as pride, which centers on ego-identity and its enhancement (Lazarus, 1991). Pride is usually placed with the family of happiness (Lazarus, 1991). The difference between pride and happiness is that the causal event associated with pride confirms or enhances personal worth (Lazarus, 1991). Pride arises from achievements that can be attributed to individuals’ abilities or efforts, and those who have a sense of pride in themselves tend to take certain actions to reproduce the positive emotional state (Lewis, 2000). Voters who perceive election polling news to have more positive effects on the self than on others would be more likely to feel proud of themselves for their ability to use the polling news and improved knowledge to make a political decision. Also, the action tendency of pride has to do with self-enhancement bolstering one’s sense of self-worth and directing one’s actions toward “behaviors that conform to social standards of worth or merit” (Mascolo & Fischer, 1995, p. 66). Proud individuals tend to try to reproduce the emotional state by taking a particular action (Lewis, 2000). Those who are proud of themselves would show reinforced confidence in their own choice of candidate and increase their support for the preferred candidate. Therefore, the following hypotheses are proposed.

Hypothesis 6 (H6): First-person perception of election polling news is positively related to pride.

Hypothesis 7 (H7): Pride predicts reinforced support for the preferred candidate.

The mediating role of pride in the association between FPP of election polling news and reinforced support for the preferred candidate is hypothesized:

Hypothesis 8 (H8): First-person perception of election polling news is positively related to reinforced support for the preferred candidate through pride.

Based on the theory and literature reviewed for hypotheses H1-H8, a model in which TPP and FPP of election polling news are directly and indirectly linked to support for restrictions on the election polling news and reinforcement of support for the preferred candidate through emotions is proposed (see Figure 1).

Figure 1
Figure 1.The Proposed Model

Note. TPP = third-person perception; FPP = first-person perception; Restriction = support for restrictions; Support = reinforcement of support for preferred candidate.

Method

To test the hypotheses, we conducted two web-based surveys among voters residing in Kyonggi-do, South Korea, before and after the 2022 Kyonggi-do gubernatorial election. Respondents were recruited from an online panel maintained by one of the leading online research firms in South Korea. The total panel size is over 1.2 million, and the composition of the panel in terms of demographics does not significantly differ from the composition of the population of South Korea. An e-mail asking for participation in the survey was sent to 5,253 voters in Kyonggi-do, and 710 of them participated in the pre-election survey conducted in May 2022 (response rate: 13.5%). They completed the questionnaire including items on perceived negative and positive effects of polling news, perceived credibility of news, emotions, election involvement, and demographics. After the election was held on June 1, 2022, another e-mail asking for participation in the post-election survey was sent to the participants in the pre-election survey. In the post-election survey, 559 voters completed the questionnaire including items on political behaviors (response rate: 78.7%). The cases of those who had never been exposed to polling news during the campaigns (n = 57) were excluded from the analyses. Forty-eight percent of the final sample (n = 502) were female, and the mean age was 43.87(SD = 12.63).

Measures

Perceived negative and positive effects of polling news were measured on the self and others respectively, using six items with a 5-point scale (Wei et al., 2011). The TPP score was calculated by subtracting the score of the perceived negative effect of polling news on the self from the score of the perceived negative effect on others. The FPP score was calculated by subtracting the score of the perceived positive effect of polling news on others from the score of the perceived positive effect on the self. For anxiety and pride, measures using a 4-point scale from previous research were used (Kim, 2016). For support for restrictions, respondents’ attitude toward restrictions on election polling news was measured by four 5-point scale items (Wei et al., 2011). To measure reinforcement of support, respondents were asked to indicate how the weight of their support for the preferred candidate had changed since their participation in the pre-election survey until the election day, using a 11-point scale (Omoto & Snyder, 1995). Election involvement, perceived credibility of news, and demographics were measured and included in the analyses as control variables. The question items, scales, and descriptive statistics are available in the Appendix.

Results

To assess the proposed model and examine the hypotheses, a structural equation modeling (SEM) using AMOS 27 was performed. The estimation method used to test the fit of the predicted model was maximum likelihood. The model contained six variables: TPP, FPP, anxiety, pride, support for restrictions, and reinforcement of support. TPP and FPP were exogenous variables, and other variables were endogenous variables. Items to measure anxiety, pride, and support for restrictions were averaged into composite variables.

Paths from the latent constructs to their observed indices were fixed to 1.0, whereas the error variance of each composite was fixed to (1- α2) multiplied by its variance (Bollen, 1989). In order to control for demographic variables, election involvement, and perceived credibility of news, separate regression analysis was performed on each variable in the model. In each regression analysis, each of the variables in the model was entered as the dependent variable, and control variables were entered as predictors. The standardized residuals from the regression analyses were saved as the new variables in the model. Before testing the model, the bivariate relationships among the variables were observed. As indicated in Table 1, statistically significant correlations were found between TPP and anxiety, TPP and support for restrictions, anxiety and support for restrictions, FPP and pride, pride and reinforcement of support, and FPP and reinforcement of support.

Table 1.Zero-Order Correlations Among Exogenous and Endogenous Variables Used in Structural Equation Modeling
1 2 3 4 5
1. TPP -
2. FPP -.44*** -
3. Anxiety .15** .04 -
4. Pride -.07 .12** -.10* -
5. Restriction .19*** -.04 .14** .02 -
6. Support -.14** .26*** .09 .19*** .08

Note. TPP = third-person perception; FPP = first-person perception; Restriction = support for restrictions; Support = reinforcement of support for preferred candidate. *p < .05, **p < .01, ***p < .001, N = 502

The results of our testing of the hypothesized model showed good fit (χ2 (df = 6, N = 502) = 8.48, CFI = .98, GFI = .99, RMSEA = .03). According to Byrne (2001), a model that fits the data should manifest GFI and CFI ratings above .90, and an RMSEA rating less than .05 (see Figure 2).

Figure 2
Figure 2.The Third- and First-Person Effects of Election Polling News Through Emotions

χ2 (df = 6, N = 502) = 8.48, CFI = .98, GFI = .99, RMSEA = .03
Note. Unstandardized estimates are listed. TPP = third-person perception; FPP = first-person perception; Restriction = support for restrictions; Support = reinforcement of support for preferred candidate. Error terms are not presented. CFI = comparative fit index; GFI = goodness of fit index; RMSEA = root mean square error of approximation. *p < .05, **p < .01, *** p < .001

Then, the paths in the model were examined. The path from TPP to support for restrictions was statistically significant (β = .19, p < .01), supporting H1. The path from FPP to reinforcement of support was statistically significant (β = .31, p < .001). H2 was supported.

TPP was positively associated with anxiety (β = .18, p <.001), supporting H3. Anxiety, in turn, was positively associated with support for restrictions (β = .15, p <.01). Hence, H4 was supported. These results suggest a mediating role of anxiety between TPP and support for restrictions. To formally test the indirect effect, bootstrapping with 5,000 samples was additionally conducted, using the PROCESS model (Hayes, 2018; Model 4). The estimates and bias-corrected 95% confidence intervals show that the mediating effect of anxiety is statistically significant (z = 1.81, 95% CI [.003, .059]). Hence, H5 was supported.

As the association between FPP and pride was statistically significant (β = .15, p < .01), H6 was supported. Pride was positively associated with reinforcement of support (β = .17, p < .001), supporting H7. However, the estimates and bias-corrected 95% confidence intervals show that the mediating effect of pride did not reach statistical significance (z = 1.36, 95% CI [-.002, .038]). H8 was not supported.

Discussion

We examined how the third- and first-person perceptions of election polling news were related to political behaviors by conducting two web-surveys before and after the Kyonggi-do gubernatorial election in South Korea. The results demonstrate that TPP of election polling news is directly and indirectly related to support for restrictions on election polling news through anxiety. FPP of election polling news is directly linked to reinforcement of support for the preferred candidate. These results suggest that voters’ perceptions about the effects of election polling news may have an actual influence on their political behaviors, and voters’ evaluation on the usefulness of election polling news may affect their attitudes toward candidates and voting behavior. Although previous research on the perceived media influence has focused on the perceived negative effects of socially undesirable media content, results of the current study highlight the impact of the perceived positive effects of election polling news.

The findings of the current study demonstrate anxiety is a powerful emotion that links TPP of polling news to support for restrictions on election polling news. The public has negative views about the role and influence of election polling news. Some argue that election polling news harms voters by misleading the public (Bradburn & Sudman, 1988), and others believe pollsters and news media are biased and attempt to influence election outcomes (Perryman et al., 2020). Results of the present study show that anxiety elicited from the perception that polling news misleads other voters’ understanding of the election while they themselves are less influenced may have a strong power to move voters to do something to avoid the presumed negative influence of the election polling news.

Emotional response is a significant factor that influences political information processing and participation (Wojcieszak et al., 2016). The results of the current study confirm that voters’ emotional response to the TPP of election polling news may significantly influence their political attitude in South Korea. Marcus et al. (2000) argue that both positive and negative emotions can affect political behaviors by triggering cognitive and behavioral strategies to deal with the emotions. For example, enthusiasm for candidate, party, or political issue can affect individuals’ political judgment and behaviors even without engaging rational reasoning (Marcus et al., 2000). However, the hypothesis that pride mediates FPP of election polling news and reinforcement of support for the preferred candidate was not supported in the current study. It is probable that the polling results released to the public at the time of data collection indicating it was a very close election were not perceived as positive enough to elicit strong pride among the participants who had been supporting either one of the two leading candidates. Testing the hypothesis with other positive emotions, such as enthusiasm, in diverse election environments will help us to understand the role of positive emotions in the first-person effect model.

There are other limitations that need to be considered in interpreting the findings. First, the results of the current study may be context-specific given that cultural, social, and political differences can affect media perceptions and their consequences (e.g., Wei et al., 2017). For example, in other countries, anger elicited from the TPP of negative effects of polling news might be the powerful emotion that motivates people to take action. It is necessary to test the hypotheses and model in more diverse contexts in future research. Another limitation is related to the sampling. Although the sample was not significantly different from the population of Kyonggi-do in terms of demographics, non-users of the Internet were not included in the study. The Internet penetration rate in South Korea is 93% as of 2021 (Ministry of Science and ICT, 2022). It is possible that the sample does not represent those who do not use the Internet, such as the elderly and those who are in a relatively low socio-economic status. In addition, those who volunteer for opt-in panels are likely to differ from other Internet users.

Despite these limitations, this study contributes to the perceived media effect research by examining the indirect effects of election polling news on individuals’ political behaviors through third- and first-person perceptions and discrete emotions as well as including the reinforcement effect in the investigation of the behavioral consequences of the FPP of election polling news. Further research on the indirect effects of media reports of poll results will help to establish more reasonable and effective regulations on election polling news.