An analysis on the impact of false information on Taiwanese voters

Exit poll results from Taiwan’s 2020 presidential and legislative elections

Doublethink Lab
Doublethink Lab
11 min readMay 6, 2021

--

Authors:
Tseng Po-Yu (曾柏瑜), Doublethink Lab researcher
Chen Yun-Ju (陳韻如), Doublethink Lab researcher

Translator:
Lee Chafaud

This is the English version of 假訊息對選民的影響分析. Based on this study, we further conducted a behavioral experiment. The result will be published before the end of this year.

To better understand the impact of false information¹ on voters during Taiwan’s 2020 presidential and legislative elections, Doublethink Lab conducted an exit poll survey at 11 polling stations located in Taiwan’s six special municipalities (Taipei, New Taipei, Taoyuan, Taichung, Tainan, and Kaohsiung) on election day, January 11, 2020. Doublethink Lab sent 32 interns and six overseers to supervise and collected 892 valid questionnaires during the voting hours of 9 am to 4 pm. [Questionnaire, Responses]

To avoid bias based on the political preferences of voters at polling stations, we selected polling stations that supported both main political parties in a 50–50 split, based on the results of the party-list vote and presidential vote during the 2016 general election. We then selected 11 polling stations located in Taiwan’s six special municipalities (only one polling station was selected in Taoyuan).

Of the 892 valid questionnaires we collected, 80 percent of respondents said that the false information threat was “serious” while more than 70 percent of respondents thought their candidates had been targeted with false information. At least half of the respondents said the electoral prospects of their candidate were affected by false information, which indicates that the impact of false information during the election was widely felt by Taiwanese voters.

Figure 1: Do you think the threat posed by false information in Taiwan is serious?
Figure 2: Do you think your candidate has been attacked with false information?
Figure 3: Do you think false information affected the electoral prospects of your candidate?

In view of the false information spreading among voters, we selected the following four false statements based on the popularity of the story, the channel of dissemination and the political leanings of the survey respondents:

  1. Chinese spy Wang Liqiang (王立強) was only a pseudonym used for an interview; by declaring him a fraud, China has contradicted itself
  2. President Tsai Ing-wen’s (蔡英文) Ph.D. thesis is fake
  3. AIDS drug makers are the driving force behind Taiwan’s same-sex marriage legislation
  4. Han Kuo-yu (韓國瑜) received united front training from the CCP

The respondents were asked whether they had heard or seen these false statements, how frequently they had heard or seen the statements, how trustworthy they felt the statement was, and how they felt about the statements.

Survey results

1. Chinese spy Wang Liqiang was only a pseudonym used for an interview; by declaring him a fraud, China has contradicted itself

This false information question group is more complicated than others. First of all, the respondents had to have heard that the name of the Chinese spy is Wang Liqiang and heard of China’s claim that Wang is a fraud. They also need to have heard that Wang Liqiang was a pseudonym used during his interview with the Australian press and that China had “contradicted itself” by claiming he was a fraud. In total, there are three levels of information within the question.

However, even though information within the statement is complex, nearly 50 percent of respondents said they had heard of it. Of those respondents, almost half said they saw the statement frequently but only a quarter said they believed the statement. This shows that the majority of people who saw the statement did not consider it to be trustworthy. Based on correlation analysis between the frequency of the false statement appears and the credibility of the statement, we found that there is no obvious correlation between the frequency of the statement and its credibility.

In addition, we also found that the statement elicited negative emotional reactions of “disgust” or “anger” and a high proportion of respondents thought the story was “ridiculous”, but there was no significant correlation between the emotion elicited and the credibility of the information.

Figure 4: Have you heard this statement?
Figure 5: Respondents who have heard similar statements, with a follow-up question about frequency and subjective credibility
Figure 6: Emotions elicited from respondents to this information

2. President Tsai Ing-wen’s Ph.D. thesis is fake

As many as 80 percent of respondents have received information claiming that President Tsai Ing-wen plagiarized her Ph.D. dissertation, with 60 percent of respondents claiming they’ve seen the information frequently. Only 20 percent of respondents said they believed the statement was credible. There appears to be a weak correlation among respondents who have received this information and those who believe it. Compared to the false statement about Wang Liqiang in question group 1, the emotion elicited by this question group was more positive (“sympathy” for Tsai, “ridiculous” or mild “surprise”). However, there was no significant correlation between the emotions elicited by it and the credibility of the statement.

Figure 7: Have you heard this statement?
Figure 8: Respondents who have heard similar statements, with a follow-up question about frequency and subjective credibility
Figure 9: Emotions elicited from respondents to this information

3. Drug manufacturers of AIDS treatments are the driving force for same-sex marriage legislation

Only 30 percent of respondents have received information in question group 3 — “drug manufacturers for AIDS treatments are the driving force for same-sex marriage legislation.” Compared with questions in groups 1 and 2, respondents answered that they infrequently saw information like this, and only 20 percent thought the information was credible. There appears to be a weak correlation between respondents who have seen the information and those that believe the statement.

There was a more polarized reaction to this information. The proportion of respondents who answered, “no feelings” was higher than other question groups. However, the emotions elicited from this information have a weak negative correlation. That is to say, the more negative emotions elicited from the information, the greater the chance they are to believe it.

Figure 10: Have you heard this statement?
Figure 11: Respondents who have heard similar narratives, with a follow-up question about frequency and subjective credibility
Figure 12: Emotions elicited from respondents to this information

4. Han Kuo-Yu received United Front training from the CCP

About 50 percent of respondents received the information “Han Kuo-yu received United Front training from the CCP.” Compared with the other question groups, respondents generally believed that they saw such information infrequently. Despite this, more than 40 percent of respondents said the information was credible. The frequency in which respondents saw this information was weakly correlated with whether they believed it; the correlation is also the highest among the four question groups.

The emotion elicited by this information tended to be negative, with a slightly higher proportion of respondents expressing “fear” at the information when compared to the other question groups. In addition, there was a moderate negative correlation between the emotions elicited by this information and the credibility of the information, which was also the highest among the four question groups. In other words, the more negative emotions elicited from the respondents, the more likely they were to believe the information.

Figure 13: Have you heard this statement?
Figure 14: Respondents who have heard similar narratives, with a follow-up question about frequency and subjective credibility
Figure 15: Emotions elicited from respondents to this information

Comprehensive Analysis

Comprehensive comparison of political leanings

Do people tend to believe information that fits their political position? The answer is yes. We further analyzed the respondents who believed the questions that had the highest credibility (ones who selected the highest score from 1 to 5) and analyzed their political leanings. It can be clearly seen that their reasons for believing the information in the question group are due to their political leanings. For example, questions 2 and 4 are related to candidates of both parties in the 2020 presidential election; we see that when the respondents have heard the information related to the questions, whether they judge the information to be true or not, they return to their subjective political leanings.

Figure 16: Political party preference for respondents who chose “5” or “highest credibility” in each question group; DPP and KMT are the two main political parties in Taiwan.

A correlation between the frequency information is seen, stirred emotions and credibility

In addition to the respondents’ own political leanings, the frequency with which the same information was seen and the emotions it elicited also affected the subjective credibility of the information. Based on the above four question groups, we calculated the Pearson correlation coefficient between the frequency a respondent saw the information and the subjective reliability of the information. In addition, we calculated the emotion of the respondents to the information by giving 1 point to positive emotions and -1 point to negative emotions. The Pearson correlation coefficient is calculated with the subjective credibility of the information and its correlation is shown in Table 1:

Table 1: Correlation coefficient between the credibility of each item group and the frequency it’s seen / emotions elicited

In terms of the Pearson correlation coefficient, it is generally believed that no correlation is found below 0.1, and a weak correlation is found between 0.1 and 0.3. Positive and negative numbers represent the direction of correlation, with positive numbers representing positive correlation and complex numbers representing negative correlation. In terms of question group 2 (Tsai Ing-wen’s fake Ph.D. thesis), group 3 (AIDS drug manufacturers pushing same-sex legislation), and group 4 (Han Kuo-yu received united front training), there is a positive weak correlation between the frequency the information was seen and the credibility of the information. That is to say, the more often respondents saw questions 2, 3, and 4, the more likely they were to believe the information. Such a relationship shows a weak correlation.

In question groups 3 and 4, there is a negative weak correlation between the emotions elicited and the credibility of the information, which means the more negative the perceptions of the information are, the more likely one is to believe it. Such a relationship shows a weak correlation and is higher than the correlation of how often the information is seen. As can be seen from the four emotion distribution charts (figures 6, 9, 12, and 15), the negative emotions of those who believed in the information are higher than the positive emotions, especially those in question groups 3 and 4, which is consistent with the results of the correlation coefficient. However, among those who didn’t believe the information, only group 4 showed more positive than negative emotions, while the other groups showed no significant difference in the distribution of positive and negative emotions. We speculate this is because each person’s subjective judgment of emotions is different. For example, in question group 2, people who did not believe that Tsai Ing-wen’s Ph.D. thesis is fake often replied that the information was “ridiculous” or “disgusting.” The spectrum of emotions behind this could be “how ridiculous/disgusting that someone is still spreading this information!” There is no obvious distinction between positive and negative emotions. If we want to analyze further, we will need to obtain more detailed information through interviews or other means.

When comparing the communication channels for these four question groups, questions in group 3 or 4 were mostly transmitted through social media platforms like LINE or Facebook, while questions in groups 1 and 2 appeared more frequently in mainstream media. However, frequent discussion in mainstream media did not enhance the credibility of the information. The number of respondents who heard the information in question groups 3 and 4 was significantly less than in question groups 1 and 2.

Moreover, most of the respondents believed they frequently saw information from question groups 1 and 2, and even those who heard of the information from question groups 3 and 4 generally believed that the frequency these two statements were seen was not high. However, the correlation coefficient between frequency and credibility in question groups 3 and 4 is higher than that of question groups 1 and 2.

Table 2: Proportion who heard each question group

Based on the correlation coefficient between the frequency the information is seen and its credibility, as well as the correlation coefficient between the emotions elicited and its credibility, we can draw the following conclusions:

Table 3: Relationship between credibility and frequency seen / emotions elicited after distinguishing whether it was widely circulated

The finding is consistent with the psychology of rumors. DiFonzo and Bordia (2007) note that rumor spreading can help people deal with the anxiety and uncertainty of an external environment. When a rumor successfully elicits negative emotions in an individual and conforms to an existing cognitive framework, it can help the individual deal with the anxiety of an unknown situation, and also enhance its subjective credibility.

In addition, according to the research of Daniel A. Efron and Medha Raj (2019), whether the individual believes the information or not, repeated encounters with the same information will significantly increase their willingness to transmit the information. In other words, when the rumor is repeated enough times, its subjective credibility will increase. Even clarifying the rumor may promote the spread of rumors. However, this finding still needs further study.

Summary

Through this study, we can determine that only when information appeared in the form of “heard it through the grapevine” or “word of mouth” among voters of certain political affiliations, such as chat groups or social media, did the two studies align: the more people heard the information, the more likely they were to believe it. The more negative emotions were stirred, the more likely they were to believe it. But when the information was widely disseminated and discussed by mainstream media, voters were more likely to be influenced by other factors (existing political affiliations, more diverse opinions, etc.). Therefore the widespread dissemination of fact-checking initiatives may reduce the credibility of rumors, but the relationship needs to be verified by subsequent studies.

In addition, this study also echoes Wang Tai-li’s (2020) research on the impact of false information on the 2020 presidential election in Taiwan. Wang’s research notes that people’s habits of choosing news media are similar to their political preferences, and it is worth further studying the severity of the echo-chamber effect and degree of polarization in Taiwan.

[1] We use false information instead of dis/misinformation here because dis/misinformation indicates malicious or unintended actors while the intention of actors in this study is mixed.

Reference

Difonzo, N., & Bordia, P. (2007). Rumors influence: Toward a dynamic social impact theory of rumor. In A. R. Pratkanis (Ed.), Frontiers of social psychology. The science of social influence: Advances and future progress (p. 271–295). Psychology Press.

Rijdt, MSc Jasper te (2019) “Fake it till you make it”: An experiment of fake news perception by use of experts and support.

Wang, T.-L. (2020). Does Fake News Matter to Election Outcomes?: The Case Study of Taiwan’s 2018 Local Elections. Asian Journal for Public Opinion Research, 8(2), 67–104. https://doi.org/10.15206/ajpor.2020.8.2.67

Acknowledgments

The researchers are particularly grateful to professor Wang Tai-li (王泰俐) of National Taiwan University’s Graduate Institute of Journalism for their assistance in designing our questionnaire and structuring our research. This paper echoes the earlier conclusions of professor Wang’s research. We would also like to thank professor Austin Wang (王宏恩) of the University of Nevada’s department of political science and Who Governs TW editor Chen Fangyu (陳方隅) for their suggestions. Doublethink Lab expresses their sincere apologies and gratitude for any omissions in the report. Finally, a great thanks go to our friend, Lee Chafaud, for translating this report for us.

--

--

Doublethink Lab
Doublethink Lab

Doublethink Lab focuses on mapping the online information operation mechanisms as well as the surveillance technology exportation and digital authoritarianism.