Voting the past: Political inertia and retrospective loyalty among South Korea's elderly

Voting habit and political inertia

Beyond affective identification, voting behavior can become routinized through repetition. Political science and psychological research show that repeated actions, such as supporting the same party across elections, can evolve into automatic behaviors that require minimal cognitive engagement (Gerber et al., 2003; Goerres, 2007). Once established, such habits are difficult to break. Changing vote choice involves psychological and informational costs, including evaluating unfamiliar alternatives, managing social expectations, and confronting inconsistencies with prior behavior.

This process gives rise to what Achen and Bartels (2016) and Dinas (2012) call political inertia: a behavioral tendency to maintain the status quo, not necessarily because of ideological conviction but because of the ease and comfort of routine. For many older voters, vote choice reflects a durable behavioral script that becomes increasingly resistant to change. This inertia helps explain why electoral decisions in later life remain stable even in the face of shifting political contexts or competing policy appeals.

Retrospective evaluations and long-term perspectives

Another source of electoral consistency in later life is retrospective voting, defined as the tendency to base political judgments on past performance rather than future promises. According to Key (1966), voters are not passive recipients of campaign messaging but active evaluators of incumbent competence. For elderly voters, this evaluative logic is particularly relevant: their long-term political memory enables them to draw on a wide range of past experiences when assessing parties or candidates (Canes-Wrone et al., 2002; Persson et al., 1997).

These accumulated judgments often function as cognitive shortcuts. Rather than reassessing preferences based on current platforms or isolated events, older voters rely on long-standing evaluations of which parties have governed effectively (Norris, 1987). As Fiorina (1981) notes, party identification may be viewed as a “running tally” of retrospective evaluations. Once these judgments solidify, they become resistant to updating (Seixas and Lourenço, 2024). Even dramatic political changes may fail to override a history of favorable assessments. Older voters therefore tend to vote the past, prioritizing continuity and stability over new or speculative political alternatives.

Generational imprinting and political socialization

From a life-course perspective, enduring political behavior in old age is often rooted in generational imprinting. Political socialization research has long shown that core political attitudes, such as partisanship, turnout, and ideological orientation, are largely shaped during adolescence and early adulthood, when individuals are most impressionable (Jennings and Niemi, 1981; Sears and Valentino, 1997). These formative experiences tend to persist throughout life, shaping political behavior decades after initial exposure.

Cohorts socialized during moments of historical significance, such as war, authoritarianism, or democratization, develop distinct political orientations that reflect those contexts (Inglehart, 1977; Mannheim, 1952). These generational imprints remain stable across the life course and often override subsequent political developments. Numerous studies have confirmed that political participation (Franklin, 2004; Grasso, 2014), attitudes (Down and Wilson, 2013; Neundorf, 2010), party preferences (Dinas, 2014; Dassonneville, 2013), and alignment (van der Brug and Rekker, 2021; Wagner and Kritzinger, 2012) are strongly shaped by early political environments.

As individuals age, cognitive reinforcement and social identity further entrench these early dispositions. Consequently, older cohorts tend to remain loyal to the parties and ideological traditions associated with their formative political years, even when these traditions lose salience or relevance. This produces cohort stability, whereby generational differences in political preferences persist over time and contribute more to long-term electoral change than shifts within cohorts themselves (Jennings et al., 2009; van der Brug and Kritzinger, 2025). In this view, the limited responsiveness of elderly voters to contemporary appeals reflects not ignorance or disengagement but the enduring influence of historically situated political worldviews.

Ideological constraint, symbolic identification, and political stability in later life

Ideological constraint, understood as the degree to which policy preferences are organized by a coherent ideological framework, provides one possible account of political consistency in later life (Converse, 1964). Individuals with higher levels of constraint tend to hold policy positions across domains, such as welfare and national security, that align with broader ideological orientations including conservatism or liberalism (Hare, 2022; Lönnqvist et al., 2025; Tyler and Iyengar, 2022). Where present, such internal coherence can contribute to predictable political behavior and reduce sensitivity to short-term issue framing.

Ideological constraint has been shown to correlate with political sophistication and sustained engagement (Feldman, 1988; Zaller, 1992), characteristics that may be more common among older voters with long histories of political participation. At the same time, a substantial body of research cautions against assuming that ideological self-identification in the mass public reliably reflects structured belief systems. Fiorina and Abrams (2008) and Mason (2018) argue that many citizens adopt ideological labels without consistently organizing their policy preferences around them. Similarly, Kinder and Kalmoe (2017) suggest that ideological identities often function less as cognitive frameworks for policy reasoning than as symbolic markers of social and political affiliation.

This symbolic dimension is likely to be particularly salient among elderly voters. For many in this cohort, ideological self-placement may be shaped by the social and generational meanings attached to political labels that emerged during periods of Cold War confrontation or democratic transition, rather than by systematically integrated policy views. In this sense, ideological identities operate as heuristics that stabilize political behavior even when issue-specific preferences are weakly articulated or internally inconsistent. As a result, continued support for a particular party among elderly voters may reflect the enduring social significance of ideological labels, rather than logical ideological consistency in the conventional sense.

The South Korean context: Authoritarian legacies and democratization

The South Korean case illustrates how historical legacies and socialization environments shape political behavior in later life. The current elderly generation came of age under authoritarian regimes that placed a strong emphasis on anti-communism, national security, and state-led economic development as central political values (Chang et al., 2007; Kang, 2008; Steinberg and Shin, 2006). These formative experiences continue to structure how older voters interpret political actors and evaluate perceived national threats, providing a durable frame of reference for political judgment (Han, 2025; Hong et al., 2023; Lee, 2023).

Within this context, conservative parties have long presented themselves as custodians of national stability and developmental achievement, drawing symbolic legitimacy from the legacy of leaders such as Park Chung-hee (Han, 2024). Appeals to national strength, economic growth, and a firm stance toward North Korea resonate with older voters not only for ideological reasons but because they activate historically grounded memories of order, security, and collective advancement. Importantly, these associations tend to persist even when conservative governments face democratic criticism or institutional crisis.

The consolidation of South Korea's party system following democratization further reinforced these alignments (Han, 2022). Older voters were increasingly confronted with a polarized electoral choice between a conservative establishment associated with continuity and order and a reformist opposition advocating political change. Many aligned with the former, and these early post-democratization choices have proven remarkably durable. This tendency appears particularly pronounced among cohorts whose formative political experiences predated the democratization movement and who therefore did not participate directly in it (Jung, 2022). Regional cleavages have further stabilized these patterns, with voters in the Gyeongsang region remaining disproportionately loyal to conservative parties linked to the Park Chung-hee legacy, while voters in Jeolla have consistently aligned with the liberal opposition associated with Kim Dae-jung (Kwon, 2004).

This dynamic was clearly visible in the 2012 presidential election. Park Geun-hye, the daughter of Park Chung-hee, mobilized nostalgia for the developmental state and secured strong support from elderly voters. For many in this cohort, her candidacy symbolized a return to stability and order, sentiments grounded more in historical memory than in contemporary policy evaluation (Hong et al., 2023). Over time, the repressive dimensions of authoritarian rule have tended to recede in collective recollection, while its perceived achievements have become more salient.

Recent evidence suggests that these historically rooted evaluations continue to shape political behavior in the present. Han (2025) shows that individuals who view martial law under past authoritarian regimes in a positive light are more likely to support President Yoon Suk Yeol's imposition of martial law, indicating that earlier political judgments can carry forward into contemporary crisis moments. This pattern is consistent with the notion that long-standing evaluations, rather than short-term policy considerations, guide political choice among older voters.

Collectively, these historical, psychological, and social mechanisms help explain why elderly voters in South Korea remain aligned with the conservative bloc even in the aftermath of democratic disruption. Habit formation, generational imprinting, symbolic identification, and regionally embedded partisanship shape the criteria through which past political performance is assessed. In this sense, retrospective voting among older voters reflects not episodic reactions to recent events but a cumulative evaluative logic anchored in earlier political experiences. Political stability in later life thus reflects a durable continuity with past commitments rather than ongoing updating in response to new information.

Research design

This study combines nationally representative panel data with an original survey experiment to examine how political orientation, retrospective voting, and exposure to policy information shape policy preferences and vote intentions among elderly voters. The integrated design enables a dual analysis of long-term ideological predispositions and short-term responsiveness to campaign cues in the context of a presidential election.

Two policy domains—elderly welfare and North Korea—serve as focal points for empirical analysis due to their persistent salience in structuring partisan alignments in South Korea. Welfare preferences capture distributive orientations, while attitudes toward North Korea reflect national security concerns and inter-Korean relations (Han, 2024). Within conservative discourse, support for welfare expansion is often framed as aligned with redistributive socialism, while engagement-oriented positions on North Korea are depicted as conciliatory or even sympathetic to the regime. These partisan framings reinforce ideological polarization, making both domains analytically useful for assessing how older voters process campaign messages.

The analysis proceeds in two stages. The first stage draws on both the Korean Welfare Panel Study and the original survey to test whether self-identified political orientation predicts support for increased welfare spending and a hardline stance on North Korea. The second stage, based exclusively on the experimental survey, introduces randomized policy treatments designed to mimic real-world campaign communication. Respondents were assigned to receive a candidate pledge either to expand elderly welfare or to strengthen national security through a tougher stance on North Korea.

This second stage includes both pre-treatment and post-treatment models. Pre-treatment models assess the independent effects of retrospective voting and issue preferences on vote intention. Post-treatment models evaluate whether exposure to candidate pledges moderates the relationship between retrospective vote history and current electoral choice. Taken together, this design allows for a comprehensive assessment of the extent to which elderly vote choice reflects entrenched behavioral patterns versus responsiveness to policy cues. An overview of the research design is presented in Figure 1.

OPEN IN VIEWER

Figure 1.

Overview of research design: Observational and experimental analyses.

Data

This study draws on two complementary data sources to examine the political behavior of elderly voters. The first is the Korean Welfare Panel Study, a nationally representative longitudinal survey administered by the Korea Institute for Health and Social Affairs. The panel targets adults aged 20 and older and is conducted annually through in-person interviews. For the purposes of this study, panel waves from 2006 to 2021 are selected based on their inclusion of attitudinal measures related to elderly welfare and national defense. To focus specifically on patterns of political behavior in later life, the analytic sample is restricted to respondents aged 70 and above.

The second data source is an original face-to-face survey developed in collaboration with Embrain, a Korean survey research firm. The target population consists of Korean citizens aged 70 and older. Trained interviewers administered the structured questionnaire in person, providing verbal instructions, confirming comprehension, and allowing respondents to complete the survey independently. To facilitate access to the target population, the survey was conducted primarily in locations with a high concentration of older adults, including welfare centers and senior citizen centers.

The survey included an embedded experimental component. Respondents were randomly assigned to one of two treatment conditions, each presenting a candidate pledge on either welfare expansion or national security. After receiving the treatment, participants were given sufficient time to consider the information before completing the post-treatment portion of the questionnaire. Fieldwork was conducted in April 2025, approximately two months prior to the presidential election. Further details on variable construction, measurement, and model specification are provided in the sections that follow. Supplementary information on the data and coding is included in Appendix A.

Political orientation, welfare preferences, and security attitudes

The first set of analyses investigates whether self-identified political orientation predicts policy preferences among voters aged 70 and older, using two data sources: the Korean Welfare Panel Study and an original face-to-face national survey. The analysis focuses on two policy domains central to partisan conflict in South Korea: redistribution through elderly welfare and national security.

In the panel analysis, the dependent variables are five-point ordinal scales indicating support for increased government spending on (1) elderly welfare and (2) national defense, with higher values denoting stronger support. The main independent variable is ideological self-placement, measured on a five-point scale from “very liberal” (1) to “very conservative” (5). To account for potential confounding, the models include controls for income, sex, education, marital status, and religion, all of which have been shown to shape policy attitudes in the Korean context (Han and Kwon, 2024). Fixed effects for region, individual, and year are also included to mitigate bias from unobserved heterogeneity and time-specific shocks.

The original survey includes comparable outcome measures: five-point items gauging support for (3) elderly welfare expansion and (4) a tougher stance on North Korea. Political orientation is measured on the same five-point ideological scale. Control variables include age, sex, education, income, homeownership, employment status, household composition, religion, region, and vote intention in the upcoming presidential election. Given the ordinal structure of the dependent variables, the models are estimated using ordered logistic regression.

Together, the two datasets provide complementary analytic leverage. The panel data enable within-individual estimation net of time-invariant characteristics, while the cross-sectional survey allows for richer sociodemographic controls. This dual approach facilitates a robust assessment of whether ideological orientation systematically shapes elderly policy preferences in the domains of redistribution and national security.

Table 1 presents the estimated coefficients for political orientation across all four models. In both the panel and original survey analyses, and across both policy domains of elderly welfare and national defense, ideological self-placement does not emerge as a statistically significant predictor of policy preferences. These null results are consistent across model specifications and remain robust after controlling for a comprehensive set of sociodemographic variables.

OPEN IN VIEWER

Table 1.

Political orientation of senior citizens, welfare support, and national security.

	Welfare panel data		Original data
	Welfare for senior citizens	National defense	Welfare for senior citizens	Tough stance on North Korea
	(1)	(2)	(3)	(4)
Political orientation	−0.036 (0.022)	0.037 (0.024)	0.03 (0.051)	0.047 (0.051)
Controls	√	√	√	√
Region FE	√	√	√	√
Individual FE	√	√
Year FE	√	√
N	3202	3180	1000	1000

Two-way fixed estimations for welfare panel data, ordinal logistic estimations for original data; controls not reported. See Table B1 in Appendix B for full results. * p < 0.1, ** p < 0.05, *** p < 0.01.

In summation, the findings suggest that among older Korean adults, ideological self-identification is not reliably associated with preferences on either redistributive or security-related policies. This pattern points to a broader disjunction between symbolic political identity and concrete policy attitudes in later life. While many elderly voters are able to place themselves along an ideological spectrum, this self-categorization does not appear to translate into structured or predictable policy positions. These results raise important questions about the extent to which conventional ideological labels meaningfully guide issue preferences or inform electoral decisions.

Retrospective voting, policy treatments, and vote intention

The second set of analyses examines whether retrospective voting and exposure to candidate policy pledges influence vote intention among voters aged 70 and older. The dependent variable is binary, coded as 1 if the respondent intends to vote for a liberal candidate—defined as those affiliated with the Democratic Party of Korea or ideologically to its left—and 0 otherwise.

The pre-treatment models estimate two specifications. Model 1 uses the cumulative number of past liberal votes, across six recent presidential and legislative elections, as the primary independent variable, capturing retrospective partisanship. Model 2 incorporates two attitudinal predictors: support for increased elderly welfare spending and support for a hardline position on North Korea. Both models control for a consistent set of sociodemographic variables previously identified as relevant to vote choice in the South Korean context (Han and Kwon, 2023).

The post-treatment models leverage the randomized experimental design embedded in the original survey. Respondents were randomly assigned to one of two informational treatments simulating campaign messages. In the welfare treatment, participants were told that the liberal candidate supports expanding welfare benefits for seniors. In the security treatment, respondents were informed that the liberal candidate favors a tougher stance on North Korea, framed in terms of enhanced national security through closer alignment with the United States.

Models 3 and 4 test whether exposure to these policy cues moderates the effect of retrospective voting on electoral choice. Each model includes main effects for retrospective liberal voting and treatment assignment, along with an interaction term to assess whether campaign messages amplify or attenuate the influence of past voting behavior. As in the pre-treatment models, a comprehensive set of sociodemographic controls is included to account for potential confounders and ensure comparability across specifications.

Figure 2 reports odds ratios from four logistic regression models examining the determinants of vote intention among elderly respondents. Model 1 indicates a strong and statistically significant association between retrospective liberal voting and support for the liberal candidate. Substantively, each additional liberal vote cast in prior elections is associated with an increase of just over 50 percent in the odds of supporting the liberal candidate in the current election (OR = 1.53). By contrast, Model 2 shows that policy attitudes, including support for elderly welfare expansion and a hardline stance on North Korea, are not significantly related to vote intention. In both cases, the estimated effects are statistically indistinguishable from one, suggesting that issue preferences alone do not meaningfully predict electoral choice in this cohort.

OPEN IN VIEWER

Figure 2.

Retrospective voting, experimental treatments, and vote intentions in the coming election (odds ratio).

Models 3 and 4 assess whether this retrospective voting effect persists following exposure to policy treatments and whether it is moderated by candidate pledges. In both models, retrospective voting remains a strong and statistically significant predictor of liberal vote intention. Each additional prior liberal vote is associated with an increase of approximately 59 percent in Model 3 (OR = 1.59) and 68 percent in Model 4 (OR = 1.68) in the odds of supporting the liberal candidate. In contrast, the experimental treatments, which present policy pledges on welfare or national security, do not exhibit statistically significant main effects. The interaction terms between retrospective voting and treatment assignment are likewise not statistically significant, indicating that exposure to these policy cues does not alter the influence of accumulated past voting behavior.

To illustrate the substantive implications of the regression results, Figure 3 presents predicted probabilities of voting for the liberal candidate across the full range of retrospective liberal voting (0 to 6), disaggregated by treatment condition. In Figure 3(a), control group respondents exhibit a clear upward trajectory: predicted support for the liberal candidate increases from 0.21 among those with no prior liberal votes to 0.81 among those with consistent liberal voting histories. The pattern is nearly identical among those assigned to the welfare treatment, with probabilities ranging from 0.19 to 0.84. Figure 3(b) shows comparable trends in the security treatment condition: predicted probabilities rise from 0.19 to 0.84 in the control group, and from 0.21 to 0.81 in the treatment group.

OPEN IN VIEWER

Figure 3.

Predicted probability by treatments across liberal vote count.

The near-parallel slopes across all conditions reinforce the null results from the interaction terms in the regression models. Retrospective vote history remains the dominant predictor of vote intention, while exposure to candidate pledges on welfare or national security has no discernible effect on the probability of supporting the liberal candidate. These findings suggest that for elderly voters, long-term partisan behavior functions as a stable heuristic, largely immune to campaign messages. More broadly, the results underscore the limited persuasive capacity of issue-based treatments and highlight the enduring influence of retrospective partisanship in shaping electoral behavior in later life.