[22nd General Election Research Series] Differences in Perceptions of Early Voting, Election Day Voting, and Election Fairness: Analysis of Post-Election Public Opinion Survey for the 22nd National Assembly Election

I. Introduction

The 22nd National Assembly election on April 10, 2024, concluded with a landslide victory for the opposition and a crushing defeat for the ruling party. Two days after the election, on April 12, as nearly all media outlets were engrossed in analyzing the election results and the future political landscape, a court ruling regarding the 2022 local elections was announced. The Supreme Court dismissed the election invalidation lawsuit concerning the Gyeonggi Province gubernatorial election held on June 1, 2022. This confirmed the court's judgment that the election was not fraudulent. During the election, the race between Kim Dong-yeon of the Democratic Party and Kim Eun-hye of the People Power Party was extremely close until the final vote count. The result was that Kim Dong-yeon won by 8,913 votes, a margin of 0.15%. For conservative voters who went to bed late at night after watching broadcasts showing Kim Eun-hye leading by a significant margin midway through the vote count, the outcome must have been hard to believe. Exit polls by the three major broadcasting networks predicted Kim Eun-hye would lead by 0.6% just before the vote count concluded. It was not until the early morning of the day after the election that Kim Dong-yeon took the lead. After the counting of ballots from Election Day was completed, and the early voting ballots were opened, a surge of votes for Kim Dong-yeon occurred. The gap between the two candidates began to narrow, and the lead was reversed around 5:32 AM (Oh Yeon-seo 2024). Subsequently, a conservative civic group filed an election invalidation lawsuit alleging 'election fraud,' and the ruling on that case was issued. The results from the early voting ballots that morning may have influenced the initiation of the lawsuit.

Just as controversies over referee fairness are common in sports, there have always been disputes, large and small, regarding the fairness of the Election Commission (EC), which oversees elections. However, the issue of the EC's overall bias or manipulation began to be seriously raised in the 21st National Assembly election in 2020. At that time, the ruling party secured over 180 seats, achieving an unexpected landslide victory. After the election, conspiracy theories about organized election fraud were actively propagated on online communities and by conservative YouTubers. These claims gained momentum as conservative politicians, including former opposition leader Hwang Kyo-ahn, Min Kyung-wook, and Kim So-yeon, joined the discourse. The controversy over the appointment of EC members at the time, as well as incidents and accidents due to the EC's inadequate management of regional election commissions, fueled these conspiracy theories. The issue escalated to election invalidation lawsuits and street protests, and these claims were legally settled when the Supreme Court dismissed the election invalidation lawsuit on July 28, 2022. Disputes over fairness, claims of election fraud, and legal battles after elections may have become a constant. Suspicion regarding the fairness of elections and the EC has not abated to this day.

In the 22nd general election, the early voting rate reached a record high of 32.28%. In this election, politicians from major parties did not officially raise concerns about election fairness or early voting. Leaders of both the ruling and opposition parties cast early votes and publicly encouraged participation. Early voting has become an accepted choice rather than a point of contention, and its participation has expanded beyond age and political affiliation. However, various opinion polls have shown that a significant number of voters do not trust the EC's management of early voting, and that suspicions of election fraud persist.

This study analyzes voters' trust in election methods and their assessment of election fairness using post-election survey data from the 22nd National Assembly election. Specifically, this study poses four main questions: First, how much do voters trust the EC's management of early voting? Does a difference in trust between early voting and Election Day voting still emerge in the post-election survey for the 22nd election, as shown in previous public surveys? Second, are voters' trust in early voting and their trust in Election Day voting related? If so, what are the characteristics of this relationship? Can this relationship be verified through statistical analysis? Third, how fair was this election perceived to be? What factors are associated with this judgment of fairness? Specifically, does the degree of agreement with the claims of election fraud in the 21st National Assembly election influence the assessment of fairness in the 22nd National Assembly election? Fourth, what are the reasons for distrusting the EC's election management?

II. Was the 22nd General Election an election where early voting had a greater impact than the 21st?

In the 21st election, the Democratic Party (including its satellite party, the Democratic Party of Citizens) secured 180 seats. Including the 3 seats of the pro-government Open Democratic Party, the total reached 183 seats. In contrast, the conservative opposition party, the United Future Party (including the Future Korea Party), won only 103 seats. In this election, the winners in 37 constituencies were reversed the following morning. A pattern emerged in several constituencies where conservative party candidates were leading during the counting of Election Day ballots, but the Democratic Party candidates took the lead once the early voting ballots were opened. The conspiracy theories surrounding the 21st general election stem from distrust in the results where the early voting ballot count overwhelmingly favored the Democratic Party.

In the 22nd general election, the results from early voting ballots also significantly favored the Democratic Party. The number of constituencies where the outcome was determined by early voting results in the 22nd election was 52, an increase of 15 compared to the 21st election. In 34 constituencies in Seoul, Gyeonggi, and Incheon, the winners were changed by the early voting results. Table 1 compares the number of constituencies where the outcome was determined by the early voting margin between the 21st and 22nd elections. Table 2 shows the vote difference between the Democratic Party and the People Power Party (United Future Party in the 21st election) for both early voting and overall voting.

<Table 1> Number of Constituencies Where Outcome Was Determined by Early Voting Margin in the 21st and 22nd General Elections

Note 1. All 52 constituencies (22nd) and 37 constituencies (21st) where the outcome was determined by the early voting margin were won by the Democratic Party over the People Power Party (22nd) / United Future Party (21st).

[Source] Converted from graphical data by Kim Do-hyung (2024).

<Table 2> Vote Difference in Early Voting and Overall Voting for Ruling and Opposition Parties in the 22nd and 21st General Elections

[Source] Converted from graphical data by Kim Do-hyung (2024).

III. Differences in Trust Levels Between Early Voting and Election Day Voting

Of the 1,528 respondents in this survey, 88.3% answered that they voted in the 22nd National Assembly election. Those who did not vote accounted for 11.7%. Among the 1,350 respondents who voted, 57.1% chose early voting (including absentee voting, mail-in voting, etc.), while 42.9% voted on Election Day, April 10th.

Interestingly, among respondents who chose early voting and those who voted on Election Day, there were no significant differences in demographic factors such as gender, age, region of residence, education level, income, or employment status, at least according to the survey data. However, there is a statistically significant difference in the choice between early voting and Election Day voting based on the respondent's party affiliation. More than 65% of respondents supporting the Democratic Party and the Jo Guk Innovation Party chose early voting (Democratic Party 65.8%, Jo Guk Innovation Party 65.6%). For those supporting the People Power Party, even though early voters constitute a high proportion of all respondents in this survey, the proportion of those who chose Election Day voting (52.6%) was slightly higher than those who chose early voting (47.4%).

Slightly less than 60% of all respondents (59.8%) trust the Election Commission's overall management of the 22nd general election. Combining those who answered 'do not trust at all' (8.2%) and 'do not trust much' (25.7%), 33.9% of the total did not trust the EC's vote management. The gap in trust between early voting and Election Day voting is also clearly evident in this election. Only 52.8% of respondents expressed trust in early voting, while trust in Election Day voting was 65.1%, a difference of about 12.3 percentage points. Conversely, those who 'do not trust much' (28.5%) or 'do not trust at all' (12.3%) regarding early voting amounted to 40.8% of the total. Those who did not trust the management of the Election Day voting were 28.5%. The early voting system was introduced to provide convenience in terms of time and place for voters, ultimately aiming to increase voter turnout. However, separate from the issue of 'convenience,' there is a low level of trust in the EC's management of early voting. This has been a common phenomenon in past elections.

It is also noteworthy that trust in early voting differs depending on the respondent's party affiliation. There was no significant difference in trust in the EC's management of Election Day voting between supporters of the Democratic Party and the People Power Party. 67% of Democratic Party supporters and 64.8% of People Power Party supporters expressed trust in the management of Election Day voting. However, the assessment of early voting management varies significantly by political affiliation. While 60.6% of Democratic Party supporters expressed trust, this figure was only 43.2% for People Power Party supporters. 35.2% of Democratic Party supporters and 52.4% of People Power Party supporters did not trust the EC's management of early voting. Table 3 summarizes this information.

<Table 3> Distribution of Trust in EC's Early Voting and Election Day Management as Shown in the Post-Election Survey for the 22nd General Election(Unit: %)

1) All Respondents

2) Breakdown by Respondent's Party Affiliation[1]

We will examine the results from this survey more closely using simple descriptive statistics. First, respondents are categorized into three groups: those who voted on Election Day, those who voted early, and those who did not vote. Then, for each group, we calculate the mean and standard deviation of responses to the questions about trust in Election Day management and early voting management, as well as the difference between these means.

Before examining the results summarized in Table 4, we explain the method and input used to measure the relevant items. Respondents' trust in the EC's management of Election Day voting (or early voting) is measured on an ordinal 4-point scale: 'Trust very much (1)', 'Trust somewhat (2)', 'Do not trust much (3)', 'Do not trust at all (4)'. Higher values for each indicator signify lower trust in the respective item. The difference between trust in Election Day voting and trust in early voting, the last item in Table 4, indicates that when the value is small (-), respondents' trust in Election Day voting is greater than their trust in early voting. Conversely, when the value is large (+), trust in Election Day voting is less than trust in early voting.

There are three main characteristics evident in Table 4. First, the difference in trust between Election Day voting and early voting generally appears regardless of how respondents participated in the 22nd National Assembly election. Whether respondents voted early, voted on Election Day, or did not vote, the average trust score for Election Day voting is lower than the average trust score for early voting. Second, the average trust score was lowest among the group of non-voters. This applies to both early voting and Election Day voting. Although there is a significant difference between the proportion of non-voters in the post-election survey and the actual proportion of non-voters in the election,[2]the 178 respondents who answered they did not vote in this survey, approximately 11.8% of the total 1528 respondents, cannot be overlooked. The implication that a considerable number of non-voters do not place much trust in election management, if generalized from the results of this survey, cannot be ignored.

Third, the difference in the average trust scores for Election Day voting and early voting is largest among the group of voters who voted on Election Day. The difference in trust for Election Day voting versus early voting among the Election Day voting group is -0.32, while the difference for the early voting group is -0.16. By simple calculation, there is approximately a twofold difference. Of course, caution is needed when comparing the simple difference of values measured on a 4-point scale for two items. However, a relative comparison of the magnitude of these values is not unreasonable.

How should these results be interpreted? The Election Day voting group (compared to the early voting group or the non-voting group) trusts Election Day voting more than early voting. If these results influence the behavior of these groups, this could extend beyond statistical implications to become a policy issue. A simple interpretation leads to the proposition that 'a considerable number of voters believe that Election Day voting is managed better than early voting, and this belief leads these individuals to vote on Election Day (rather than early voting).

Furthermore, even among the early voting group, the average trust score for Election Day voting is higher than the average trust score for early voting. Among the 776 respondents who reported voting early in this survey, 69 individuals responded that they do not trust early voting much but trust Election Day voting somewhat. This represents 4.5% of all respondents and 8.8% of early voting participants. If generalized, this suggests that approximately 8.8 out of every 100 people believe that Election Day voting is managed better than early voting, even though they themselves voted early.

<Table 4> Difference in Trust Between Election Day Voting and Early Voting

f2e34dd1a320d063

1. Confidence in same-day voting and early voting is measured on a 4-point scale: Very Confident (1) - Somewhat Confident (2) - Not Very Confident (3) - Not at All Confident (4). A higher average confidence score in Table 4 indicates a lower level of confidence in that voting method. 2. Respondents who answered 'Don't Know' are treated as missing values.

Figure 1 illustrates the proportion of respondents at the intersection of confidence in early voting and confidence in same-day voting, as revealed by this survey, using a heatmap. The total respondents are categorized into three groups: those who voted early, those who voted on election day, and those who did not vote. The horizontal axis represents the level of confidence in the management of early voting, and the vertical axis represents the level of confidence in the management of same-day voting. The numerical proportion of respondents at the intersection of these two axes is expressed through the shading of colors. A higher proportion of respondents results in a darker intersection area. If most respondents have the same level of confidence in both same-day and early voting, the area along the upward-sloping diagonal of the heatmap will be the darkest.

Two main characteristics can be observed from the heatmap in Figure 1. First, there were few instances where respondents showed extreme differences in their confidence scores for same-day and early voting. For example, there were no or very few respondents who expressed high confidence in same-day voting but no confidence at all in early voting. Not a single respondent expressed full confidence in early voting while simultaneously expressing complete distrust in same-day voting.[3] Second, the difference in respondents' confidence levels between same-day and early voting was often relative. For instance, a respondent might express some confidence in same-day voting but not much confidence in early voting. While there was a difference in the responses for individuals, it was incremental.

Third, in a significant number of cases, the confidence score for same-day voting was equal to or higher than that for early voting. Notably, within the group of same-day voters, the proportion of responses where confidence in same-day voting was higher than confidence in early voting was not small. For example, among the 574 respondents who reported voting on election day in the 22nd general election, 69 individuals stated they had some confidence in same-day voting but not much confidence in early voting. This represents 4.5% of all respondents and 12.0% of same-day voters. Conversely, 79 same-day voters (5.1% of all respondents) indicated they had 'somewhat confident' in both same-day and early voting, accounting for 13.7% of same-day voters. The proportions between these two groups were not significantly different.

Figure 1. Proportion of Respondents by Confidence in Early Voting and Same-Day Voting: Heatmap

Is the relationship between respondents' confidence in early voting and confidence in same-day voting statistically significant? First, respondents were divided into three groups: those who voted on election day, those who voted early, and those who did not vote. A chi-squared (χ²) test was conducted to examine the relationship between the level of confidence in early voting and the level of confidence in same-day voting within each group. The analysis revealed a statistically significant relationship between the two variables in all three groups (at a 95% confidence level). Respondents who expressed high confidence in the management of early voting tended to also express high confidence in same-day voting, and this trend was significant across all groups: same-day voters, early voters, and non-voters. Table 5 summarizes the results of the chi-squared test for the relationship between confidence in early voting and confidence in same-day voting.

Table 5. Relationship Between Confidence in Early Voting and Confidence in Same-Day Voting - Chi-Squared (χ²) Test

IV. Respondents' Evaluation of the Fairness of the 22nd General Election Management - Regression Analysis

This chapter examines, through a simple regression analysis, how respondents perceived the fairness of the 22nd General Election and what factors are associated with that perception.[4] The dependent variable in the model is the evaluation of the fairness of the 22nd general election, measured on a 4-point ordinal scale: 'Very Fair', 'Generally Fair', 'Somewhat Unfair', 'Very Unfair' in response to the question, 'How fair do you think the 21st General Election was?'

The key independent variable is the respondent's level of agreement with the claim that there was organized election fraud in the 21st General Election. Respondents answered this question on a 4-point ordinal scale: 'Strongly Agree', 'Somewhat Agree', 'Somewhat Disagree', 'Strongly Disagree'. The hypothesis is that the greater the agreement with the claim of organized fraud in the previous 21st election, the higher the likelihood that the respondent will perceive the subsequent 22nd General Election as unfair.

This survey asked about the extent of agreement with the claim of organized election fraud in the 21st general election held in 2020. Table 6 presents the results. Two somewhat surprising findings emerged from the responses to this question. First, approximately 39% of all respondents expressed agreement (either 'Strongly Agree' (8.9%) or 'Somewhat Agree' (30.1%)) with the claim of organized election fraud in the 21st general election. Second, over half of the People Power Party supporters (55.1%) agreed with this claim, which is approximately 18.9 percentage points higher than the corresponding figure for Democratic Party supporters (36.2%). The claim of organized election fraud in the 21st general election has been considered a settled matter, having been debated in academia and ruled upon by the courts as lacking factual basis. However, a significant portion of respondents still express agreement with these claims of organized election fraud.

Table 6. Proportion of Agreement with the Claim of Organized Election Fraud in the 2020 21st Election: Total Respondents

Variables reflecting respondents' socioeconomic background were also included in the model. Control variables included gender (Female=1, Male=0), age, education level, and homeownership (No Homeownership=1). Political interest, political ideology, and party affiliation (No Party Affiliation=1, Affiliated=0) were also included. Variables measuring respondents' perception of the ideological positions of the Democratic Party and the People Power Party, and the perceived difference between them (ideological difference), were also included.

Responses regarding whether respondents encountered fake news or disinformation during the campaign period of the 22nd general election were also included in the model. Respondents could indicate their experience by choosing from the following categories: 'Very Often' if they encountered fake news or disinformation more than 10 times during the election period; 'Often' if 5 to less than 10 times; 'Sometimes' if less than 5 times; and 'Never'.

Table 7 presents the results of the regression analysis. Three models, (1), (2), and (3), with different control variables, were estimated. This was done to verify how robustly the effect of the key independent variable, 'Agreement with the claim of organized fraud in the 21st election','Agreement with the claim of organized fraud in the 21st election'remained statistically significant. In all three estimated models, the variable 'Agreement with the claim of organized fraud in the 21st election' was statistically significant, and the direction and magnitude of the coefficient showed little change. Respondents who expressed less agreement with the claim of organized fraud in the 21st election were more likely to evaluate the 22nd General Election as fair. Conversely, respondents who agreed with the claim of organized fraud in the 21st election were more likely to evaluate the 22nd election as unfair. This suggests that voters (or a segment thereof) may be evaluating the fairness of the 22nd election as an extension of the controversy surrounding the 21st election.Degree of agreement with the claim of election fraud in the 21st general electionThe variables were statistically significant, and the direction and magnitude of the coefficients did not change significantly. The less respondents agreed with the claim that there was organized fraud in the 21st election, the more likely they were to evaluate that the 22nd National Assembly election was also conducted fairly. Respondents who agreed with the claim that there was organized fraud in the 21st election were more likely to evaluate that the 22nd election was also unfair. This can be interpreted as voters (or a portion of voters) evaluating the fairness of the 22nd election perhaps as an extension of the controversy surrounding the 21st election.

This statement might seem tautological. Voters who viewed the previous election as fraudulent would naturally assess the current election as unfair as well. Given the ongoing claims of organized fraud in the 21st general election and subsequent elections, and even court rulings, this perspective might be attributed to the group holding such beliefs itself. It is impossible to debunk all conspiracy theories, and their claims may stem from emotional rather than logical or evidence-based reasoning. It could be that they are unable to accept the election results as they are, particularly if their preferred candidate or party lost, or they may attribute unexpected outcomes to issues with newly introduced (or poorly understood) systems. There is also suspicion that these claims are linked to political or economic interests.

However, even if all these criticisms are acknowledged, the issue remains that many people still suspect organized fraud in the 21st election and express low confidence in early voting. As seen in Table 6, approximately 39% of all respondents in this survey agreed with the claim of organized election fraud in the 21st general election. The significant difference in these suspicions across political affiliations further complicates the issue. More than half of the respondents supporting the People Power Party (approximately 55.1%) agreed to some extent with the claim of organized fraud in the 21st election. This result persists even four years after the 21st general election, and despite three national elections (presidential, local, and the 22nd general election) having been held since then.

Table 7. Evaluation of Fairness of the 22nd General Election - Ordinal Logit Regression Analysis

V. Conclusion: Why is the Election Commission's Vote Management Not Trusted?

Respondents were asked to provide reasons for not trusting the National Election Commission's (NEC) vote management, with seven options presented, including concerns about manipulation through early voting and allegations of political bias. Table 8 summarizes the responses regarding the reasons for not trusting the NEC's vote management. The most frequently cited reason by respondents who did not trust the NEC's vote management was the concern that 'early ballots may be vulnerable to manipulation, such as switching, while stored until election day.' Over half of all respondents (51.1%) selected this option. The next most common reason, selected by nearly half of all respondents (49%), was that 'the NEC ignored and did not address allegations regarding election procedures and processes.'

What implications can be drawn from these two points? It is possible that various incidents and accidents at the NEC and its regional offices concerning the inadequate storage of early ballots have negatively impacted public perception. For example, in the 2022 presidential election, the Bucheon City Election Commission in Gyeonggi Province faced issues when approximately 50,000 absentee early voting mail-in ballots were stored in a director's office without CCTV coverage (Kim Myung-jin 2022). The NEC argued this was a necessary measure for administrative convenience, and the central NEC explained it as a regional office error. However, such incidents and accidents can be sufficient to shake the public's confidence in the NEC's rigorous management of voting processes and ballot boxes. Furthermore, in addition to the NEC's inadequate management, there is also a criticism that the NEC has not sufficiently addressed concerns raised by civil society. Rather than viewing 'allegations regarding election processes and procedures' as indicative of 'election fraud proponents,' efforts for thorough explanations and broad sharing of those explanations are necessary.

Other concerns included the limitations of the NEC as a political organization and issues with absentee voting, which were cited less frequently. The claim that the increased possibility of manipulation due to the computerization of election management received the fewest responses. The respondents who agreed with this point, suggesting that the decline in trust in election management is an unavoidable consequence of computerization and informatization, were the fewest. This suggests that respondents are pointing to the results of actual mismanagement or errors by the NEC, rather than an unavoidable outcome of technological advancement.■

Table 8. Reasons for Not Trusting the NEC's Vote Management: 22nd General Election

References

Kim Do-hyeong. 2024. "'Early Voting' Influence Increased... Outcome Reversed in 52 Electoral Districts." Hankook Ilbo. April 22. (https://www.hankookilbo.com/News/Read/A2024042022010001915?did=NA, Accessed: April 22, 2024).

Kim Myung-jin. 2022. "Approximately 50,000 Early Voting Mail-in Ballots Stored in Director's Office Without CCTV." Chosun Ilbo. March 8. (https://www.chosun.com/politics/politics_general/2022/03/07/556XCFCOPVFHTN3PFCD3ZLUGLA/, Accessed: April 22, 2024).

Oh Yeon-seo. 2024. "'Last Minute Reversal' June 1st Gyeonggi Governor Election... Supreme Court: 'Not an Election Fraud'." Hankyoreh. April 15. https://www.hani.co.kr/arti/society/society_general/1136563.html, Accessed: April 22, 2024).

■ Author: Kim Jun-seok, Professor, Department of Political Science and International Relations, Dongguk University.

■ Managed and Edited by: Kim Sun-hee, Senior Researcher, EAI.

Inquiries: 02-2277-1683 (ext. 209), shkim@eai.or.kr

[1] Among the total 1,528 respondents, there were 503 supporters of the Democratic Party, 516 supporters of the People Power Party, 231 respondents with no party affiliation (or who did not disclose it), 120 supporters of the Cho Kuk Innovation Party, and 159 supporters of the Green Justice Party, New Future Party, Reform Party, and other parties.

[2] This is a common phenomenon in general public opinion surveys.

[3] Conversely, the case where respondents expressed extremely high confidence in same-day voting while having no confidence at all in early voting was rare but did appear in this survey. Such respondents who expressed extreme distrust solely in 'early voting' were not found among non-voters (0 individuals), but there were 3 such respondents among those who chose same-day voting and 4 among those who chose early voting (!). However, these numbers represent only 0.1% and 0.2% of the total 1,528 respondents, respectively.

[4]As this research paper was written with an emphasis on the speed of analysis immediately following the election, please understand that the model's configuration, analysis methods, and interpretations may undergo significant revisions in the future.