Beyond generational discourse: An age-period-cohort analysis of South Koreans’ attitudes toward reunification with North Korea

ABSTRACT

This study conducts age-period-cohort analyses of South Koreans’ attitudes toward reunification with North Korea, a hostile polity that shares a national identity, using the Korea Institute for National Unification survey data accumulated since 2014. It is found that the attitudes toward unification, in general, vary with age. Older respondents tend to show a greater need for unification, a weaker orientation toward peaceful coexistence, and a greater willingness to bear the unification tax burden. Unlike findings of the prior studies, cohort (generation) effects are limitedly found. While the youngest generation has the least interest in North Korea, the middle-aged generations who used to see North Korea through a lens of its own context have a relatively stronger interest in North Korea. In contrast, period effects outweigh cohort effects on South Koreans’ attitudes toward reunification and North Korea. These findings imply that the application of the “generational discourse of reunification with North Korea” in previous studies and the media may be overly exaggerated.

KEYWORDS:

• Age-period-cohort analysis
• Public attitudes
• Generational effect
• Unification
• South Korea

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/03623319.2024.2334888.

Notes

¹ Three scholarly journals, whose focus encompasses Unification/North Korean studies, rank among the top 10 academic journals within the disciplines of political science and international relations in South Korea, when evaluated based on the Korean Citation Index (KCI), as of 2022.

² We cannot find a significant annual variation in the number of respondents by cohort as shown in Table A2.

³ The accumulated datasets also include some nuanced items regarding attitudes toward unification. For robustness checks, this paper also conducts an APC analysis of additional items (unification as an immediate goal and nationalistic unification) conducted at least five consecutive years (see Table A2).

⁴ According to Raudenbush and Bryk (2002), the reasons for centering variables in multilevel regression models are to obtain meaningful intercepts and estimates of explanatory variables, and numerical stability of the estimates. Of these, the first two reasons are particularly critical. Centering a variable allows for a more meaningful interpretation of the intercept, as it represents the predicted value when all predictors are set to their mean values. In addition, centering the explanatory variables can lead to more stable and accurate estimates of their coefficients. This is because the influence of extreme values is reduced when the variables are centered, which can improve the numerical stability of the estimates (31–35). The use of theoretically meaningful values, such as the population mean of an explanatory variable, is also viable option for centering. Certain datasets being analyzed in this paper include respondents aged 19 and older or those aged 18 and older, which was caused by a revision to the election law lowering the voting age to 18 eligible voters (see Table A1), with the proportion of such respondents being very small. Centering the variable at age 20 can help mitigate the influence of extreme values, especially those associated with aged 18 and 19 respondents, thereby improving the numerical stability of the estimates.

⁵ We include the age-squared variables in the analysis to see if the age effects are non-linear but did not observe statistically significant results. Therefore, the study does not include them in the models. As generational effects showed a non-linear (U- or upside-down U-shaped) curve in some instances, it can be inferred that several results, originally understood as the age effects, may have been influenced by cohort effects.

⁶ We anticipate mitigating issues stemming from the skewed distribution of household income by applying logarithmic transformation to the household income.

⁷ The effectiveness of the dummy variables are demonstrated in group-level means of self-placed ideology measured on an 11-point scale (0 = liberal, 5 = moderate, 10 = conservative) scale. The group-level means of Daegu/North Gyeongsang (Conservative region 1), Busan/Ulsan/South Gyeongsang (Conservative region 2), and Honam (Liberal region) are 5.86 (SE = 0.06), 5.37 (SE = 0.06), and 4.36 (SE = 0.07), respectively.

⁸ Questionnaires utilizing a four-point scale may simplify the dichotomization of responses, but they could face criticism for compelling respondents to select one of two sides on a given issue.

⁹ To test the joint cohort and period effects, we employ the methodological framework of ANOVA. As a likelihood ratio test that contrasts two distinct models, this approach is not influenced by the possible emergence of statistically significant outcomes that could be attributable to random noises.Thus, it is not necessary to apply a correction for multiple Chi-squared comparisons in a traditional ANOVA.

¹⁰ The Unification Consciousness Survey conducted by the Institute for Peace and Unification Studies Seoul National University (IPUS) which was launched in 2007 provides relevant information about attitudes toward unification and North Korea policy in the previous period. The survey results do not show any dramatic changes in attitudes toward unification before the KINU survey was launched. Pro-unification attitudes were the lowest in 2015, and such a trend maintained until the survey conducted in 2017 when the liberal president was elected (B. S. Kim et al., 2022). See Figure A1 for details.

¹¹ As a reviewer suggested, this paper conducts an APC analysis of additional items regarding perceptions of North Korea conducted at least five consecutive years (see Table A3).

¹² The term, immanent approach, refers to a method of analysis that seeks to understand a phenomenon, system, or entity from within its own context, history, and set of conditions, rather than imposing external models or perceptions upon it (Koh, 2019). When applied to understanding North Korea, an immanent approach would suggest an effort to understand North Korea based on its own terms rather than imposing external standards. This could include, for instance, a deep understanding of the Juche Ideology (North Korea’s official state ideology that emphasizes self-reliance), the dynamic leadership structure, and the nation’s historical experiences.