Partisan Division and Sorting of Korean Voters: A Comparison of the 2012 and 2022 Presidential Elections

1. Introduction

Is the political polarization of Korean voters intensifying? How does political polarization affect the voting choices of Korean voters? What are the implications of voter political polarization for the development of Korean democracy? This study is an attempt to answer these three questions.

Voter political polarization can be divided into two dimensions: ‘ideological polarization’ and ‘affective polarization,’ with the former having constructive effects on democratic development and the latter having destructive effects on democratic development. Voter political polarization can be understood, on the one hand, as a phenomenon of ‘partisan divergence,’ where the heterogeneity between camps increases, and on the other hand, as a phenomenon of ‘partisan sorting,’ where the homogeneity within camps increases. The study posits that voter voting choices based on ideological partisan divergence or sorting can lead to ‘democratic development,’ while voting choices based on affective partisan divergence or sorting can lead to ‘democratic backsliding.’

This study empirically examines theoretical hypotheses regarding the impact of voter political polarization on democratic development and voting choices, using panel survey data from the 2012 and 2022 South Korean presidential elections. The second section develops conceptual discussions on ideological and affective polarization, and partisan divergence and sorting, and presents theoretical hypotheses on the impact of voter political polarization on democratic development and voting choices. The third section describes the progress of ideological and affective partisan divergence and sorting among South Korean voters in 2012 and 2022, reports statistical analyses on the impact of voter political polarization on voting choices, and summarizes the implications for the development of South Korean democracy. The findings reveal that over the past decade, the political factors influencing the voting choices of South Korean voters have shifted from ideological partisan sorting to affective partisan sorting, and the study concludes by summarizing the theoretical discussions and empirical findings.

2. Theoretical Hypotheses: Voter Political Polarization, Quality of Democracy, and Voting Choices

This section presents theoretical hypotheses regarding the impact of voter political polarization on democratic development at the macro level and on voting choices at the micro level. Voter political polarization can be divided into two dimensions: ideological polarization and affective polarization (Mason, 2018). First, ideological polarization refers to the phenomenon where voters divided into two camps move in opposite directions in a one-dimensional space, with one extreme representing agreement with progressive values and the other extreme representing agreement with conservative values (Levendusky and Pope, 2011). Second, affective polarization refers to the phenomenon where voters divided into two camps move in opposite directions in a one-dimensional space, with one extreme representing antipathy towards progressive parties and the other extreme representing antipathy towards conservative parties (Druckman et al., Forthcoming).

The impact of voter political polarization on the quality of democracy may vary depending on its nature (Stavrakakis, 2018). First, the advancement of ideological polarization can arise as a consequence of a political process where party competition provides voters with clear choices and makes party policy differences discernible, potentially contributing to democratic development (McCoy and Somer, 2019). Second, the advancement of affective polarization can arise as a consequence of a political process where party competition is perceived by voters as a battlefield between hostile camps, leading them to view the opposing camp as an existential threat, potentially leading to effects that undermine democratic development (Somer, McCoy, and Luke, 2021).

[Table 1] Ideological Polarization, Affective Polarization, and Quality of Democracy

f2e34dd1a320d063

f2e34dd1a320d063

f2e34dd1a320d063

[Table 1] categorizes the relationship between voter ideological and affective polarization and the quality of democracy. (1) If the level of ideological polarization is high and the level of affective polarization is low, the effects of the former can offset the effects of the latter, contributing to ‘democratic development.’ (2) If the level of ideological polarization is low and the level of affective polarization is low, the effects of the former and the latter can mutually offset each other, leading to a risk of ‘democratic stagnation.’ (3) If the level of ideological polarization is low and the level of affective polarization is high, the effects of the former can be offset by the effects of the latter, leading to a risk of ‘democratic backsliding.’ (4) If the level of ideological polarization is high and the level of affective polarization is high, the effects of the former and the latter can mutually reinforce each other, leading to a risk of ‘democratic breakdown.’

There are two ways to understand and capture political polarization: understanding it as a phenomenon of partisan divergence and understanding it as a phenomenon of partisan sorting (Lelkes, 2016). First, partisan divergence refers to the phenomenon where heterogeneity between two camps, divided along ideological or affective lines, increases. Ideological partisan divergence is the phenomenon where the policy differences between the camp of voters who agree with progressive values and support progressive parties, and the camp of voters who agree with conservative values and support conservative parties, widen. Affective partisan divergence is the phenomenon where the emotional differences between the camp of voters who are antipathetic towards conservative parties and support progressive parties, and the camp of voters who are antipathetic towards progressive parties and support conservative parties, widen (Fiorina, 2017). Second, partisan sorting refers to the phenomenon where homogeneity within two camps, divided along ideological or affective lines, increases. Ideological partisan sorting is the phenomenon where the proportion of voters who agree with progressive (conservative) values increases within the camp of voters who support progressive (conservative) parties. Affective partisan sorting is the phenomenon where the proportion of voters who are antipathetic towards conservative (progressive) parties increases within the camp of voters who support progressive (conservative) parties (Levendusky, 2009).

[Table 2] Partisan Divergence, Partisan Sorting, and Voting Choices

f2e34dd1a320d063

f2e34dd1a320d063

f2e34dd1a320d063

[Table 2] categorizes the relationship between voter partisan divergence and sorting, and voting choices. (1) If the degree of partisan divergence is high and the degree of partisan sorting is low, the probability of voters engaging in partisan voting increases due to the effect of the former. (2) If the degree of partisan divergence is low and the degree of partisan sorting is low, the effects of the former and the latter mutually offset each other, reducing the probability of voters engaging in partisan voting. (3) If the degree of partisan divergence is low and the degree of partisan sorting is high, the probability of voters engaging in partisan voting increases due to the effect of the latter. (4) If the degree of partisan divergence is high and the degree of partisan sorting is high, the effects of the former and the latter mutually reinforce each other, increasing the probability of voters engaging in partisan voting.

Summarizing the above discussion, the following observable implications can be derived. (1) As the degree of ideological partisan divergence increases, the probability of partisan voting among voters will increase, leading to a positive impact on democratic development. (2) As the degree of affective partisan divergence increases, the probability of partisan voting among voters will increase, leading to a negative impact on democratic development. (3) As the degree of ideological partisan sorting increases, the probability of partisan voting among voters will increase, leading to a positive impact on democratic development. (4) As the degree of affective partisan sorting increases, the probability of partisan voting among voters will increase, leading to a negative impact on democratic development.

3. Empirical Verification: Comparison of Panel Survey Results from the 2012 and 2022 South Korean Presidential Elections

This section empirically verifies the theoretical hypotheses regarding the impact of voter political polarization on democratic development and voting choices, using panel survey data from the 2012 and 2022 South Korean presidential elections.[1]

[Figure 1] South Korean Presidential Election Results, 1963-2022[2]

Source: National Election Commission Election Statistics Systemhttp://info.nec.go.kr/(Accessed: April 24, 2022).

[Figure 1] is a scatter plot of presidential election results from 1963 to 2022, with ‘two-party contestation’ on the horizontal axis and ‘two-party mobilization’ on the vertical axis. ‘Two-party contestation’ is defined as the inverse of the vote share difference between the top two parties, and ‘election mobilization’ is defined as the sum of the vote shares of the top two parties. Observation points located to the right of the upper value of ‘two-party contestation,’ indicated by the solid line connecting top and bottom, represent ‘hyper-contestation’ two-party competition. Observation points located above the upper value of ‘two-party mobilization,’ indicated by the solid line connecting left and right, represent ‘hyper-mobilization’ two-party competition. Compared to the 2022 election, which recorded ‘two-party contestation’ of 0.007 and ‘election mobilization’ of 0.955, comparable observation points include the 2012 election, which recorded ‘two-party contestation’ of 0.035 and ‘election mobilization’ of 0.996, and the 2002 election, which recorded ‘two-party contestation’ of 0.023 and ‘election mobilization’ of 0.946. Below, the panel survey data from the 2012 and 2022 presidential elections are compared among these three observation points.[3]

① Partisan Divergence Among South Korean Voters

Ideological or affective partisan divergence is measured in the following three ways. First, kernel density estimates of the voter ideological or affective distribution are plotted. This visualizes the voter ideological or affective distribution, which, while not highly precise, allows for intuitive judgment. Second, the bimodality coefficient of the voter ideological or affective distribution is calculated.[4]

A bimodality coefficient of 1 indicates a completely bimodal distribution, while 0 indicates a completely unimodal distribution. If the bimodality coefficient exceeds 0.55, the voter ideological or affective distribution is considered polarized. Third, the average ideological difference between the camp of voters who agree with progressive values and support progressive parties and the camp of voters who agree with conservative values and support conservative parties, or the average affective difference between the camp of voters who are antipathetic towards conservative parties and support progressive parties and the camp of voters who are antipathetic towards progressive parties and support conservative parties, is measured. The degree of ideological or affective partisan divergence is assessed by the score difference between the two camps.

[Figure 2] compares the ideological and affective partisan divergence of South Korean voters in 2012 and 2022 using kernel density estimation. On the horizontal axis of ideological partisan divergence on the left, 0 represents the maximum agreement with progressive values, and 10 represents the maximum agreement with conservative values. On the horizontal axis of affective partisan divergence on the right, 0 represents the maximum antipathy towards conservative parties, and 10 represents the maximum antipathy towards progressive parties. The score, obtained by subtracting the progressive party favorability score (0-10) from the conservative party favorability score (0-10) (resulting in a ‘partisan affect score’ of -10 to 10), has been transformed to a 0-10 scale.

[Figure 2] Partisan Divergence of South Korean Voters in 2012 and 2022: Kernel Density Estimation[5]

Source: Ideological partisan divergence: East Asia Institute 2012 General Election Presidential Election Panel 7th Survey Question 1 Background and East Asia Institute 2022 Presidential Election Panel 2nd Survey Question 6 Background. Affective partisan divergence: East Asia Institute 2012 General Election Presidential Election Panel 7th Survey Questions 6-1-3 and 6-1-4, and East Asia Institute 2022 Presidential Election Panel 2nd Survey Questions 9-1 and 9-2. 2012 datahttps://kossda.snu.ac.kr/ (accessed April 24, 2022).

In terms of ideological partisan division, compared to 2012, the proportion of liberal voters slightly increased in 2022, while centrist voters decreased and conservative voters slightly decreased. Regarding emotional partisan division, compared to 2012, anti-conservative voters slightly increased in 2022, neutral emotion voters significantly decreased, and anti-liberal voters slightly increased. While it can be confirmed that both emotional and ideological partisan divisions have progressed over the past decade, it is undeniable that both dimensions of partisan division are closer to a unimodal distribution than a bimodal distribution. Visually, the ideological and emotional partisan divisions of Korean voters appear to be far from polarization.

[Table 3] Partisan Division of Korean Voters in 2012 and 2022: Bimodal Coefficient[6]

f2e34dd1a320d063

f2e34dd1a320d063

f2e34dd1a320d063

[Table 3] compares the ideological and emotional partisan divisions of Korean voters in 2012 and 2022 by calculating the bimodal coefficient. The ideological partisan division bimodal coefficient increased by 0.082 from 0.336 in 2012 to 0.418 in 2022, and the emotional partisan division bimodal coefficient increased by 0.027 from 0.345 in 2012 to 0.372 in 2022. However, neither reached the bimodal distribution threshold (閾値) of 0.55. Statistically, the ideological and emotional partisan divisions of Korean voters appear to be far from polarization.

[Table 4] Partisan Division of Korean Voters in 2012 and 2022: Mean Difference[7]

f2e34dd1a320d063

f2e34dd1a320d063

f2e34dd1a320d063

Source: Party Identification: East Asia Institute (EAI) 2012 General Election and Presidential Election Panel Survey, 6th Wave, Question 7, and East Asia Institute (EAI) 2022 Presidential Election Panel Survey, 1st Wave, Question 9. The rest is the same as [Figure 2]. 2012 datahttps://kossda.snu.ac.kr/ (Accessed: April 24, 2022).

[Table 4] compares the ideological and affective partisan divisions of South Korean voters in 2012 and 2022 by examining the average differences in ideology and affect between voters supporting progressive parties and voters supporting conservative parties. First, the average ideology of voters supporting progressive parties shifted 0.7 points to the left, from 4.5 in 2012 to 3.8 in 2022, while the average ideology of voters supporting conservative parties shifted 0.1 points to the left, from 7.0 in 2012 to 6.9 in 2022. The average ideological difference between voters supporting progressive parties and voters supporting conservative parties increased from 2.5 in 2012 to 3.1 in 2022, an increase of 0.6. Second, the average affect of voters supporting progressive parties shifted 1.0 points to the left, from 3.4 in 2012 to 2.4 in 2022, while the average affect of voters supporting conservative parties shifted 0.9 points to the right, from 6.8 in 2012 to 7.7 in 2022. The average affective difference between voters supporting progressive parties and voters supporting conservative parties increased from 3.4 in 2012 to 5.3 in 2022, an increase of 1.9. Third, the proportion of independent voters decreased by 1.5 percentage points from 35.8% in 2012 to 34.3% in 2022 for ideological partisan division, and by 1.4 percentage points from 35.3% in 2012 to 33.9% in 2022 for affective partisan division. However, due to the presence of independent voters constituting over 30% of the total electorate, the distribution of voter party support does not approximate a bimodal distribution. This is why it may be premature to characterize the increase in the average ideological or affective difference between voters supporting progressive parties and voters supporting conservative parties as a polarization phenomenon.[8]

② Partisan Alignment of South Korean Voters

Ideological or affective partisan alignment is measured in three ways. First, kernel density estimation of the ideological or affective distribution of voters in the progressive party support camp or the conservative party support camp is visualized. This is a visualization of the ideological or affective distribution of camp voters that allows for intuitive judgment, although its precision is not high. Second, the overlapping coefficient of the ideological or affective distribution of camp voters is calculated.[9]

An overlapping coefficient of 1 indicates complete overlap between the two camp distributions, while 0 indicates complete separation. Third, the ideological or affective composition of the voter camp supporting progressive parties is compared with the ideological or affective composition of the voter camp supporting conservative parties. The degree of ideological or affective partisan alignment is gauged by the increase or decrease in the proportion of voters who agree with progressive values or have negative feelings towards conservative parties within the progressive party support camp, and the proportion of voters who agree with conservative values or have negative feelings towards progressive parties within the conservative party support camp.

[Figure 3] visualizes and compares the ideological and affective partisan alignment of South Korean voters in 2012 and 2022 using kernel density estimation. On the horizontal axis of ideological partisan alignment at the top, 0 represents the maximum agreement with progressive values, and 10 represents the maximum agreement with conservative values. On the horizontal axis of affective partisan alignment at the bottom, 0 represents the maximum negative feelings towards conservative parties, and 10 represents the maximum negative feelings towards progressive parties.

Regarding ideological partisan alignment, compared to 2012, the proportion of voters with progressive tendencies appears to have increased in the composition of voters supporting progressive parties in 2022, while the proportions of centrist and conservative voters appear to have decreased. In the composition of voters supporting conservative parties, the proportion of voters with conservative tendencies does not appear to have changed significantly, while the proportion of centrist voters appears to have increased and the proportion of voters with progressive tendencies appears to have decreased. Despite changes in the composition of camp voters, a significant overlap between the distribution of voters supporting progressive parties and the distribution of voters supporting conservative parties can be observed, suggesting that it is difficult to claim that ideological partisan alignment has significantly advanced over the past decade.

Regarding affective partisan division, compared to 2012, the proportion of voters with negative feelings towards conservative parties appears to have significantly increased in the composition of voters supporting progressive parties in 2022, while the proportions of affectively neutral voters and voters with negative feelings towards progressive parties appear to have significantly decreased. In the composition of voters supporting conservative parties, the proportion of voters with negative feelings towards progressive parties appears to have significantly increased, while the proportions of affectively neutral voters and voters with negative feelings towards conservative parties appear to have significantly decreased. Due to changes in the composition of camp voters, a reduction in the degree of overlap between the distribution of voters supporting progressive parties and the distribution of voters supporting conservative parties can be observed, suggesting that it is reasonable to claim that affective partisan alignment has significantly advanced over the past decade. While the ideological partisan alignment of South Korean voters observed visually appears to be far from a polarization phenomenon, the affective partisan alignment of South Korean voters can be assessed as approaching a polarization phenomenon.

[Figure 3] Partisan Alignment of South Korean Voters in 2012 and 2022: Kernel Density Estimation[10]

[Table 5] compares the ideological and affective partisan alignment of South Korean voters in 2012 and 2022 by calculating the overlapping coefficient. The overlapping coefficient for ideological partisan alignment decreased by 0.096 from 0.534 in 2012 to 0.438 in 2022, but it is difficult to deny that the degree of overlap between the distributions of camp voters is relatively large, making it questionable to call this result a polarization phenomenon. The overlapping coefficient for affective partisan alignment decreased by 0.149 from 0.245 in 2012 to 0.096 in 2022, and the degree of overlap between the distributions of camp voters has become relatively very small, making it reasonable to call this result a polarization phenomenon. While the ideological partisan alignment of South Korean voters confirmed through descriptive statistics appears to be far from a polarization phenomenon, it is reasonable to assess that affective partisan alignment has effectively approached a polarization phenomenon.

[Table 5] Partisan Alignment of South Korean Voters in 2012 and 2022: Overlapping Coefficient[11]

f2e34dd1a320d063

f2e34dd1a320d063

f2e34dd1a320d063

[Table 6] Ideological and Affective Partisan Alignment of South Korean Voters in 2012 and 2022: Difference in Proportions[12]

f2e34dd1a320d063

f2e34dd1a320d063

f2e34dd1a320d063

[Table 6] compares the ideological and affective partisan alignment of South Korean voters in 2012 and 2022, examining the ideological or affective composition of the camp supporting progressive parties and the camp supporting conservative parties. The degree of ideological or affective partisan alignment is measured by the increase or decrease in the proportion of voters who agree with progressive values or have negative feelings toward conservative parties among supporters of progressive parties, and the proportion of voters who agree with conservative values or have negative feelings toward progressive parties among supporters of conservative parties.

Regarding ideological partisan alignment, first, the proportion of voters with a progressive orientation among supporters of progressive parties increased by 11.2 percentage points from 40.5% in 2012 to 51.7% in 2022. In contrast, the proportion of voters with a conservative orientation among supporters of conservative parties decreased by 3.8 percentage points from 70.9% in 2012 to 67.1% in 2022. Second, the proportion of voters with a conservative orientation among supporters of progressive parties decreased by 7.6 percentage points from 20.4% in 2012 to 12.8% in 2022. Meanwhile, the proportion of voters with a progressive orientation among supporters of conservative parties decreased by 3.0 percentage points from 5.5% in 2012 to 2.5% in 2022. Third, the proportion of centrist voters among supporters of progressive parties decreased by 3.6 percentage points from 39.1% in 2012 to 35.5% in 2022. Conversely, the proportion of centrist voters among supporters of conservative parties increased by 6.9 percentage points from 23.6% in 2012 to 30.5% in 2022. While there was a consistent upward trend in the intensity of ideological partisan alignment in the camp of progressive party supporters, with an increase in the proportion of progressive-oriented voters and a decrease in centrist and conservative-oriented voters, this was not the case for the camp of conservative party supporters, where the proportion of progressive-oriented voters decreased, but the proportion of centrist voters increased and the proportion of conservative-oriented voters decreased, thus not showing a consistent upward trend in ideological partisan alignment intensity.

Regarding affective partisan alignment, first, the proportion of voters with anti-conservative sentiment among supporters of progressive parties increased by 28.5 percentage points from 43.3% in 2012 to 71.8% in 2022. In contrast, the proportion of voters with anti-progressive sentiment among supporters of conservative parties increased by 28.4 percentage points from 44.6% in 2012 to 73.0% in 2022. Second, the proportion of voters with anti-progressive sentiment among supporters of progressive parties decreased by 2.1 percentage points from 2.4% in 2012 to 0.3% in 2022. Meanwhile, the proportion of voters with anti-conservative sentiment among supporters of conservative parties increased by 0.4 percentage points from 0.2% in 2012 to 0.6% in 2022. Third, the proportion of voters with neutral sentiment among supporters of progressive parties decreased by 26.5 percentage points from 54.4% in 2012 to 27.9% in 2022. Conversely, the proportion of voters with neutral sentiment among supporters of conservative parties decreased by 28.8 percentage points from 55.2% in 2012 to 26.4% in 2022. A consistent upward trend in the intensity of affective partisan alignment was observed in the camp of progressive party supporters, with a significant increase in the proportion of voters with anti-conservative sentiment and a significant decrease in the proportion of voters with neutral sentiment. Similarly, a consistent upward trend in the intensity of affective partisan alignment was observed in the camp of conservative party supporters, with a significant increase in the proportion of voters with anti-progressive sentiment and a significant decrease in the proportion of voters with neutral sentiment. Affective partisan alignment has significantly advanced over the past decade, and the results appear to be a clear manifestation of polarization.

In summary, over the past decade, partisan division and alignment among South Korean voters have advanced in both ideological and affective dimensions. Nevertheless, given that the distribution of voter party support deviates from a bimodal distribution, it is difficult to characterize either dimension as polarization. While the degree of ideological partisan alignment is difficult to label as polarization due to the considerable overlap between the distributions of voters in each camp, the degree of affective partisan alignment, where the overlap between the distributions of voters in each camp has nearly vanished, can reasonably be described as polarization.

③ Political Polarization and Voting Choices of South Korean Voters

Using the panel survey data from the 2012 and 2022 South Korean presidential elections as empirical subjects, this study statistically verified whether voter partisan alignment actually influenced voting choices. With voter voting choices as the dependent variable, four logistic multiple regression analyses were conducted: a model for selecting Moon Jae-in of the Democratic United Party in 2012, a model for selecting Park Geun-hye of the Saenuri Party in 2012, a model for selecting Lee Jae-myung of the Democratic Party in 2022, and a model for selecting Yoon Suk-yeol of the People Power Party in 2022. The dependent variable was assigned a value of 1 if the candidate was selected and 0 otherwise.[13]

For ideological partisan alignment, voters supporting progressive parties with an ideological orientation score between 0 and 4 were assigned a value of -4, those scoring 5 were assigned -3, and those scoring between 6 and 10 were assigned -2. Among unaffiliated voters, those scoring between 0 and 4 were assigned -1, those scoring 5 were assigned 0, and those scoring between 6 and 10 were assigned 1. For voters supporting conservative parties, those scoring between 0 and 4 were assigned 2, those scoring 5 were assigned 3, and those scoring between 6 and 10 were assigned 4.[14]A higher level of alignment between support for progressive parties and agreement with progressive values results in a larger negative value, while a higher level of alignment between support for conservative parties and agreement with conservative values results in a larger positive value.

For affective partisan alignment, voters supporting progressive parties with a partisan affect score between -10 and -4 were assigned a value of -4, those scoring between -3 and 3 were assigned -3, and those scoring between 4 and 10 were assigned -2. Among unaffiliated voters, those scoring between -10 and -4 were assigned -1, those scoring between -3 and 3 were assigned 0, and those scoring between 4 and 10 were assigned 1. For voters supporting conservative parties, those scoring between -10 and -4 were assigned 2, those scoring between -3 and 3 were assigned 3, and those scoring between 4 and 10 were assigned 4.[15]A higher level of alignment between support for progressive parties and anti-conservative sentiment results in a larger negative value, while a higher level of alignment between support for conservative parties and anti-progressive sentiment results in a larger positive value.

Control variables include the following. First, for the generation variable, voters under 30, voters in their 30s, voters in their 40s, voters in their 50s, and voters aged 60 and over were each treated as dummy variables, assigned a value of 1 if they belonged to that age group and 0 otherwise.[16] The reference category is voters aged 60 and over. Second, for the region variable, voters in the Seoul metropolitan area, voters in the Incheon and Gyeonggi area, voters in the Daejeon, Chungbuk, and Chungnam area (in 2012) or Daejeon, Sejong, and Chungcheong area (in 2022), voters in the Gwangju, Jeonbuk, and Jeonnam area, voters in the Daegu and Gyeongbuk area, voters in the Busan, Ulsan, and Gyeongnam area, and voters in the Gangwon and Jeju area were each treated as dummy variables, assigned a value of 1 if they belonged to that region and 0 otherwise.[17] The reference category is voters in the Daegu and Gyeongbuk area. Fourth, for the gender variable, males were assigned a value of 1 and females 0.[18]Fifth, for the education variable, those with a high school education or less were assigned a value of 1, and those with a college education or higher were assigned 0.[19] Sixth, for the household income variable, in 2012, incomes of 1 million won or less were assigned a value of 1, 1 million won to less than 2 million won were assigned 2, 2 million won to less than 4 million won were assigned 3, 4 million won to less than 5 million won were assigned 4, 5 million won to 7 million won or less were assigned 5, and over 7 million won were assigned 6. In 2022, incomes of less than 2 million won were assigned 1, 2 million won to less than 3 million won were assigned 2, 3 million won to less than 4 million won were assigned 3, 4 million won to less than 5 million won were assigned 4, 5 million won to less than 6 million won were assigned 5, 6 million won to less than 7 million won were assigned 6, and 7 million won or more were assigned 7.[20][Table 7] and [Table 8] summarize the descriptive statistics for the candidate selection models of the 2012 and 2022 South Korean presidential elections, respectively.

[Table 7] Summary of Descriptive Statistics for the 2012 South Korean Presidential Election Candidate Selection Model

f2e34dd1a320d063

f2e34dd1a320d063

f2e34dd1a320d063

N: 1,122.

[Table 8] Descriptive Statistics Summary of the 2022 South Korean Presidential Election Candidate Choice Model

f2e34dd1a320d063

f2e34dd1a320d063

f2e34dd1a320d063

Number of observations: 916.

[Figure 4] Regression Analysis Results of Candidate Choice Model for the 2012 South Korean Presidential Election[21]

[Figure 4] presents the regression analysis results of the candidate choice model for the 2012 South Korean Presidential Election. In the Moon Jae-in candidate choice model on the left, first, ideological partisan alignment had a statistically significant negative effect at the 0.001 level, and affective partisan alignment had a statistically significant negative effect at the 0.05 level. Interpreted as odds ratios, a one-unit increase in ideological score is associated with approximately a 40% decrease in the probability of voting for Moon Jae-in, and a one-unit increase in affective score is associated with approximately a 25% decrease in the probability of voting for Moon Jae-in. Second, compared to voters aged 60 and above, voters under 30 showed a statistically significant positive effect at the 0.01 level, and voters in their 30s and 40s showed a statistically significant positive effect at the 0.05 level. Interpreted as odds ratios, compared to voters aged 60 and above, voters under 30 are approximately 2.8 times more likely to vote for Moon Jae-in, and voters in their 30s and 40s are approximately 2.0 times more likely to vote for Moon Jae-in. The effect for voters in their 50s was not statistically significant. Third, compared to voters in the Daegu-Gyeongbuk region, voters in the Daejeon-Chungbuk-Chungnam region showed a statistically significant positive effect at the 0.05 level, and voters in the Gwangju-Jeonbuk-Jeonnam region showed a statistically significant positive effect at the 0.01 level. Interpreted as odds ratios, compared to voters in the Daegu-Gyeongbuk region, voters in the Daejeon-Chungbuk-Chungnam region are approximately 2.4 times more likely to vote for Moon Jae-in, and voters in the Gwangju-Jeonbuk-Jeonnam region are approximately 3.7 times more likely to vote for Moon Jae-in. The effects for voters in other regions were not statistically significant. The results for gender, education, and household income variables were not statistically significant.

In the Park Geun-hye candidate choice model on the right, first, ideological partisan alignment showed a statistically significant positive effect at the 0.001 level, but affective partisan alignment was not statistically significant. Interpreted as odds ratios, a one-unit increase in ideological score is associated with approximately a 1.7-fold increase in the probability of voting for Park Geun-hye. Second, compared to voters aged 60 and above, voters under 30 and voters in their 30s showed a statistically significant negative effect at the 0.001 level, and voters in their 40s showed a statistically significant negative effect at the 0.01 level. Interpreted as odds ratios, compared to voters aged 60 and above, the probability of voting for Park Geun-hye is approximately 78% lower for voters under 30, approximately 70% lower for voters in their 30s, and approximately 65% lower for voters in their 40s. The effect for voters in their 50s was not statistically significant. Third, compared to voters in the Daegu-Gyeongbuk region, voters in the Daejeon-Chungbuk-Chungnam region showed a statistically significant negative effect at the 0.05 level, and voters in the Gwangju-Jeonbuk-Jeonnam region showed a statistically significant negative effect at the 0.01 level. Interpreted as odds ratios, compared to voters in the Daegu-Gyeongbuk region, voters in the Daejeon-Chungbuk-Chungnam region are approximately 63% less likely to vote for Park Geun-hye, and voters in the Gwangju-Jeonbuk-Jeonnam region are approximately 74% less likely to vote for Park Geun-hye. The effects for voters in other regions were not statistically significant. The results for gender, education, and household income variables were not statistically significant.

[Figure 5] Regression Analysis Results of Candidate Choice Model for the 2022 South Korean Presidential Election[22]

[Figure 5] presents the regression analysis results of the candidate choice model for the 2022 South Korean Presidential Election. In the Lee Jae-myung candidate choice model on the left, first, ideological partisan alignment was not statistically significant, and affective partisan alignment showed a statistically significant negative effect at the 0.001 level. Interpreted as odds ratios, a one-unit increase in affective score is associated with approximately a 58% decrease in the probability of voting for Lee Jae-myung. Second, compared to voters aged 60 and above, voters under 30 showed a statistically significant positive effect at the 0.05 level. Interpreted as odds ratios, compared to voters aged 60 and above, voters under 30 are approximately 2.1 times more likely to vote for Lee Jae-myung. The effects for voters in other age groups were not statistically significant. Third, compared to voters in the Daegu-Gyeongbuk region, voters in the Gwangju-Jeonbuk-Jeonnam region showed a statistically significant positive effect at the 0.05 level, and voters in the Gangwon-Jeju region showed a statistically significant positive effect at the 0.01 level. Interpreted as odds ratios, compared to voters in the Daegu-Gyeongbuk region, voters in the Gwangju-Jeonbuk-Jeonnam region are approximately 3.3 times more likely to vote for Lee Jae-myung, and voters in the Gangwon-Jeju region are approximately 5.4 times more likely to vote for Lee Jae-myung. The effects for voters in other regions were not statistically significant. Fourth, household income showed a statistically significant positive effect at the 0.05 level. Interpreted as odds ratios, a one-unit increase in household income score is associated with approximately a 10% increase in the probability of voting for Lee Jae-myung. The results for gender and education variables were not statistically significant.

In the Yoon Suk-yeol candidate choice model on the right, first, ideological partisan alignment was not statistically significant, and affective partisan alignment showed a statistically significant positive effect at the 0.001 level. Interpreted as odds ratios, a one-unit increase in affective score is associated with approximately a 2.9-fold increase in the probability of voting for Yoon Suk-yeol. Second, compared to voters aged 60 and above, voters under 30 showed a statistically significant negative effect at the 0.001 level. Interpreted as odds ratios, compared to voters aged 60 and above, voters under 30 are approximately 79% less likely to vote for Yoon Suk-yeol. The effects for voters in other age groups were not statistically significant. Third, compared to voters in the Daegu-Gyeongbuk region, no regional voters showed a statistically significant effect. The results for gender, education, and household income variables were not statistically significant.

In summary, while the 2012 South Korean presidential election was characterized by ideological partisan alignment influencing voter choice, the 2022 presidential election was influenced by affective partisan alignment. The magnitude of the effect of ideological partisan alignment on voting for Moon Jae-in was -0.508 in terms of the regression coefficient, and for Park Geun-hye it was 0.581. In contrast, the magnitude of the effect of affective partisan alignment on voting for Lee Jae-myung was -0.859, and for Yoon Suk-yeol it was 1.074. This indicates that not only has the magnitude of partisan alignment's influence on voter choice significantly increased over the past decade, but its nature has also shifted from 'partisan-ideological alignment' to 'partisan-affective alignment'.

The shift in the center of gravity of partisan factors determining voter candidate choice in South Korean presidential elections from ideological partisan alignment to affective partisan alignment has profound implications for the qualitative transformation of South Korean democracy. This is because an increase in the intensity of voters' affective partisan alignment is likely to have a negative impact on the qualitative development of South Korean democracy.

[Figure 6] illustrates the correlation between the degree of political polarization and the level of liberal democracy in 38 OECD member countries in 2021. The former uses a 'political polarization' score, an expert assessment of 'how much the society is polarized into hostile political camps,' ranging from a minimum of 0 to a maximum of 4. The latter uses the 'liberal democracy index'.[23]The hypothesis of this study, which posits that a low level of voter ideological polarization and a high level of affective polarization increase the likelihood of democratic regression by offsetting the constructive effect of the former on democracy with the destructive effect of the latter, was found to have empirical validity. A one-unit increase in the political polarization score has the effect of decreasing the liberal democracy index by 0.1. Indeed, among countries with a political polarization score of 3 or higher, excluding Italy, Mexico, Colombia, Turkey, Poland, and Hungary have experienced virtual democratic collapse, while the United States and Slovenia are experiencing democratic regression (Boese et al., 2022).

[Figure 6] Correlation between Political Polarization and Liberal Democracy in OECD Member Countries, 2021[24]

In conclusion, the 2022 presidential election, which confirmed that affective partisan alignment strongly influenced voters' candidate choices, suggests that the effects of future political polarization are more likely to be destructive than constructive for South Korean democracy. This is why attention should be paid to voter partisan alignment in relation to the qualitative transformation of South Korean democracy.

4. Conclusion

This study pointed out that voter political polarization has two dimensions: ideological polarization and affective polarization, and that the former has a constructive effect on the qualitative development of democracy, while the latter has a destructive effect. It proposed two approaches to understanding voter political polarization: one focusing on the increase in inter-group heterogeneity and the advancement of partisan division, and the other approaching voter political polarization by focusing on the increase in intra-group homogeneity and the advancement of partisan alignment.

By empirically examining the impact of voter political polarization on democratic development and voting choices using panel survey data from the 2012 and 2022 South Korean presidential elections, the following findings were made. First, while South Korean voters' ideological and affective partisan divisions have advanced over the past decade, their level has not been sufficient to transform the distribution of voter party support into a bimodal distribution. Second, although South Korean voters' ideological partisan alignment has advanced over the past decade, its level has not been sufficient to eliminate the overlap between the camps of voters supporting progressive parties and those supporting conservative parties. South Korean voters' affective partisan alignment has advanced significantly over the past decade, reaching a level that has virtually eliminated the overlap between the camps of voters supporting progressive parties and those supporting conservative parties. The advancement of South Korean voter political polarization over the past decade can be evidenced by the phenomenon of affective partisan alignment.

In the 2012 presidential election, ideological partisan alignment strongly influenced voters' candidate choices. The higher the level of alignment between voters' progressive party support and their agreement with progressive values, the higher the probability of voting for the progressive party candidate. Conversely, the higher the level of alignment between voters' conservative party support and their agreement with conservative values, the higher the probability of voting for the conservative party candidate. In the 2022 presidential election, affective partisan alignment strongly influenced voters' candidate choices. The higher the level of alignment between voters' progressive party support and their antipathy towards conservative parties, the higher the probability of voting for the progressive party candidate. Conversely, the higher the level of alignment between voters' conservative party support and their antipathy towards progressive parties, the higher the probability of voting for the conservative party candidate. We found that over the past decade, the political factors influencing South Korean voters' choices have shifted from ideological partisan alignment to affective partisan alignment.

The results of the 2022 presidential election, where the effect of ideological partisan alignment declined and the effect of affective partisan alignment increased, strongly suggest that voter political polarization is likely to have a negative impact on the qualitative development of South Korean democracy. This is why researchers interested in the qualitative development of South Korean democracy need to continuously pay attention to the advancement of voter partisan alignment. ■

References

Boese, Vanessa A., Nazifa Alizada, Martin Lundstedt, Kelly Morrison, Natalia Natsika, Yuko Sato, Hugo Tai, and Staffan I. Lindberg. 2022. Autocratization Changing Nature? Democracy Report 2022. Gothenburg: Varieties of Democracy Institute (V-Dem).

Druckman, James N., Samara Klar, Yanna Krupnikov, Matthew Levendusky, and John Barry Ryan. Forthcoming. “(Mis)estimating Affective Polarization.” Journal of Politics.

Fiorina, Morris P. 2017. Unstable Majorities: Polarization, Party Sorting, and Political Stalemate. Stanford: Hoover Institute Press.

Freeman, Jonathan B. and Rick Dale. 2013. “Assessing Bimodality to Detect the Presence of a Dual Cognitive Process.” Behavior Research Methods 45(1):83-97.

Levendusky, Matthew S. and Jeremy C. Pope. 2011. “Red States vs. Blue States: Going Beyond the Mean.” Public Opinion Quarterly 75(2): 227-248.

Levendusky, Matthew. 2009. The Partisan Sort: How Liberals Became Democrats and Conservatives Became Republicans. Chicago: University of Chicago Press.

Mason, Lilliana. 2018. Uncivil Agreement: How Politics Became Our Identity. Chicago: University of Chicago Press.

McCoy, Jennifer and Murat Somer. 2019. Toward a Theory of Pernicious Polarization and How It Harms Democracies: Comparative Evidence and Possible Remedies.” The ANNALS of the American Academy of Political and Social Science 62(1): 234-271.

Schmid, Friedrich and Axel Schmidt. 2006. “Nonparametric Estimation of the Coefficient of Overlapping: Theory and Empirical Application.” Computational Statistics and Data Analysis 50: 1583-1596.

Somer, Murat, Jennifer L. McCoy, and Russell E. Luke. 2021. “Pernicious Polarization, Autocratization and Opposition Strategies.” Democratization 28(5): 929-948.

Stavrakakis, Yannis. 2018. “Paradoxes of Polarization: Democracy’s Inherent Division and the (Anti-) Populist Challenge.” American Behavioral Scientist 62(1): 43-58.

[1] The data used below are from the East Asia Institute (EAI) 2012 General Election and Presidential Election Panel Surveys (Waves 1-7) and the EAI 2022 Presidential Election Panel Survey (Waves 1-2). The former was conducted between March and December 2012, and the latter between January and March 2022. The 2012 data are provided by the Korean Social Science Data Archive (https://kossda.snu.ac.kr/, accessed April 24, 2022).https://kossda.snu.ac.kr/ accessed April 24, 2022).

[2] Election competitiveness is displayed with the value inverted.

[3] The EAI Presidential Election Panel Survey began in 2007, so data for the 2002 presidential election, which would be comparable, are not available.

[4] The Bimodality Coefficient (BC) is calculated as follows:

where s denotes skewness, k denotes excess kurtosis, and n denotes sample size (Freeman and Dale, 2013).

[5] Ideological partisan division: 0 represents the maximum agreement with progressive values, and 10 represents the maximum agreement with conservative values. Affective partisan division: This is derived by subtracting the progressive party favorability score (0-10) from the conservative party favorability score (0-10) and then converting the resulting partisan affective score (-10-10) to a 0-10 scale. 0 represents the maximum antipathy towards conservative parties, and 10 represents the maximum antipathy towards progressive parties.Ideological Partisan Divide: 0 indicates the maximum agreement with progressive values, and 10 indicates the maximum agreement with conservative values. Affective Partisan Divide: Affective partisan score (-10-10), obtained by subtracting the progressive party favorability score (0-10) from the conservative party favorability score (0-10), was rescaled to 0-10. 0 indicates the maximum antipathy toward conservative parties, and 10 indicates the maximum antipathy toward progressive parties.

[6] Same source as [Figure 2].

[7]'No Party Affiliation': Refers to voters who responded 'no party to support' (2012 survey) or 'no party to favor' (2022 survey).

[8] Considering that in the United States, the proportions of Democratic, Republican, and independent voters were 46%, 39%, and 14% in 2012, 46%, 39%, and 15% in 2016, and 46%, 42%, and 12% in 2020, respectively (American National Election Studies), it can be confirmed that the proportion of voters with no party affiliation in South Korea is relatively large.https://electionstudies.org/resources/anes-guide/top-tables/?id=22 Accessed: April 24, 2022.

[9] The overlapping coefficient (OC) is calculated as follows.

f(x) represents the distribution of ideology or sentiment among supporters of conservative parties, and d(x) represents the distribution of ideology or sentiment among supporters of progressive parties (Schmid and Schmidt, 2006).

[10] Same source as [Table 4].

[11] Same source as [Figure 2].

[12] Same source as [Table 4]. The classification criteria are as follows: Progressive orientation: voters with an 'ideological orientation' score from 0 to 4; Moderate orientation: voters with an 'ideological orientation' score of 5; Conservative orientation: voters with an 'ideological orientation' score from 6 to 10. Anti-conservative sentiment: voters with a partisan sentiment score from -10 to -4; Neutral sentiment: voters with a partisan sentiment score from -3 to 3; Anti-progressive sentiment: voters with a partisan sentiment score from 4 to 10.

[13] East Asia Institute 2012 General Election and Presidential Election Panel Survey, 7th wave, question 1, and East Asia Institute 2022 Presidential Election Panel Survey, 2nd wave, question 2. 2012 datahttps://kossda.snu.ac.kr/ (Accessed: April 24, 2022.).

[14] East Asia Institute 2012 General Election and Presidential Election Panel Survey, 7th wave, background question 1, and East Asia Institute 2022 Presidential Election Panel Survey, 2nd wave, background question 6. 2012 datahttps://kossda.snu.ac.kr/(Accessed: April 24, 2022.).

[15] East Asia Institute 2012 General Election and Presidential Election Panel Survey, 7th wave, questions 6-1-3 and 6-1-4, and East Asia Institute 2022 Presidential Election Panel Survey, 2nd wave, questions 9-1 and 9-2. 2012 datahttps://kossda.snu.ac.kr/(Accessed: April 24, 2022.).

[16] East Asia Institute 2012 General Election and Presidential Election Panel Survey, 1st wave, background question 1, and East Asia Institute 2022 Presidential Election Panel Survey, 1st wave, background question 3. 2012 datahttps://kossda.snu.ac.kr/ (Accessed: April 24, 2022.).

[17] East Asia Institute 2012 General Election and Presidential Election Panel Survey, 1st wave, background question 3, and East Asia Institute 2022 Presidential Election Panel Survey, 1st wave, background question 1. 2012 datahttps://kossda.snu.ac.kr/ (Accessed: April 24, 2022.).

[18] East Asia Institute 2012 General Election and Presidential Election Panel Survey, 1st wave, background question 1, and East Asia Institute 2022 Presidential Election Panel Survey, 1st wave, gender determination result. 2012 datahttps://kossda.snu.ac.kr/ (Accessed: April 24, 2022.).

[19] East Asia Institute 2012 General Election and Presidential Election Panel Survey, 1st wave, background question 6, and East Asia Institute 2022 Presidential Election Panel Survey, 1st wave, background question 5. 2012 datahttps://kossda.snu.ac.kr/ (Accessed: April 24, 2022.).

[20] East Asia Institute 2012 General Election and Presidential Election Panel Survey, 1st wave, background question 11, and East Asia Institute 2022 Presidential Election Panel Survey, 1st wave, background question 7. 2012 datahttps://kossda.snu.ac.kr/ (Accessed: April 24, 2022.).

[21] Number of observations: 1,122; Log-likelihood for Moon Jae-in candidate selection model: -414.04; Log-likelihood for Park Geun-hye candidate selection model: -406.87.

[22] Number of observations: 916; Log-likelihood for Lee Jae-myung candidate selection model: -300.38; Log-likelihood for Yoon Suk-yeol candidate selection model: -296.06.

[23] V-Dem dataset version 12 (https://www.v-dem.net/vdemds.html Accessed: April 24, 2022.).

[24] V-Dem dataset version 12 (https://www.v-dem.net/vdemds.htmlAccessed: April 24, 2022.). The solid line represents the regression line, and the shaded area represents the 95% confidence interval.

■ Author: Kim Jeong_Associate Professor at the University of North Korean Studies. Ph.D. in Political Science from Yale University. Currently serves as a visiting professor at Yonsei University's Graduate School of International Studies, a regional coordinator for the Asia Democracy Network, and a policy advisor to the Ministry of National Defense and the Defense Intelligence Agency. Previously held positions as a visiting researcher at the University of Tokyo's Graduate School of Interdisciplinary Information Studies, a senior researcher at the East Asia Institute, and a principal researcher at Kyungnam University's Institute for Far Eastern Studies. Areas of interest include comparative political institutions, comparative political economy, inter-Korean relations, and East Asian international relations. Authored publications include “South Korean Democratization: A Comparative Empirical Appraisal” (2018), “The Legislative Production Capacity of a Democratic Constitutional State: The Case of South Korea's Divided Government” (2020), “A Working Parliament, A Talking Parliament, A Confrontational Parliament: Macro-Level Consequences and Micro-Level Foundations of Parliamentary Distrust” (2020), “Conditions for the Success of COVID-19 Quarantine Policies: A Comparative Study of the Korean Case” (2021), and “The Failure of the Constitution, The Failure of the Judiciary, The Failure of the President: Exercise Leadership to Rectify the Judiciary” (2022).

■ Editor: Jeon Ju-hyunEAI Researcher

    Inquiries: 02 2277 1683 (ext. 204) | jhjun@eai.or.kr