A. Analysis Sample

This study utilises two nationally representative household survey datasets: the Korean Labor and Income Panel Study (KLIPS) for Korea and the China Family Panel Studies (CFPS) for China. Both datasets provide rich demographic, economic, and social information, making them well suited for analysing changes in marriage patterns across different socioeconomic groups over time.

Korean Labor and Income Panel Study (KLIPS)

The KLIPS is an annual longitudinal survey that has been conducted by the Korea Labor Institute since 1998. It collects detailed panel data on individuals and households residing in urban areas across Korea using a two-stage stratified cluster sampling method. First, 1,000 enumeration districts in cities nationwide are selected, after which five households are randomly sampled from each district. The survey is now in its 26th wave (completed in 2023).

This study uses data from 2003 to 2023 (21 waves) to examine trends over the past two decades. Our analysis focuses on individuals aged 25 to 45, provided that information on key variables for the main analysis — in this case, marital status, age, gender, education, current residential city information, and parental education — is available.

China Family Panel Studies (CFPS)

The CFPS is a biennial longitudinal survey launched in 2010 by the Institute of Social Science Survey at Peking University. It is one of the most comprehensive panel datasets available in China, covering over 16,000 households across 25 provinces and municipalities. As of May of 2025, seven waves have been released, including the 2010 baseline and follow-ups in 2012, 2014, 2016, 2018, 2020 and 2022. The CFPS employs a probability-proportionate-to- size-sampling strategy, stratified by administrative unit (province, county/district, village/neighbourhood/community) and socioeconomic status, with households randomly selected within the sampled communities. For our analysis, we use CFPS waves from 2010 to 2022, restricting the sample to individuals aged 20 to 35 with complete data on marital status, age, gender, education, current household registration (hukou) status, residential area, and parental education.

B. Descriptive Statistics

Columns 1 and 2 of Table 1 present summary statistics for the Korean analysis sample, which includes individuals aged 25 to 45 in each survey wave, conducted between 2003 and 2023. At the time of the survey, 38% of Korean men and 22% of Korean women were never married. The average age of the respondents is 33. Among males, 40% have completed at least a four-year university degree, compared to 33% of females. The average annual income for Korean men in the sample is 33.72 million KRW, while women earn approximately 60% of that amount. Although there are substantial gender disparities in individual-level characteristics, such as educational attainment and income, parental education, specifically, father’s education, shows no notable gender differences. Approximately 20% of the sample resides in Seoul.

TABLE 1

SUMMARY STATISTICS

Note: This table reports summary statistics of each country’s baseline analysis sample. The Korean sample includes individuals aged 25 to 45. The Chinese sample includes individuals aged 20 to 40.

Columns 3 and 4 summarise the characteristics of the Chinese sample, which consists of individuals aged 20 to 40 across the survey waves from 2010 to 2022. As in Korea, the proportion of never-married individuals is higher among men than women: 32% of Chinese men and 20% of women report never having been married. The average age of Chinese respondents is 30 years. More than half of the sample has attained a middle school education or below, while 31-33% have completed high school. Income data are available for individuals who were employed at the time of the survey. On average, men earn 33,498 yuan annually, while women earn about 60% of that amount - similar to the gender income gap observed in the Korean data. Parental education levels, particularly fathers’ education, do not exhibit significant gender differences. While half of the respondents live in urban areas, only 23-24% hold an urban hukou.

To assess the representativeness of our samples, we compare key characteristics with external data. For China, Table A1 compares our 2016 CFPS sample with the 2015 Population Census.3 Compared to the census data, the CFPS sample has slightly fewer never-married individuals (0.33 vs. 0.36), less education, and a higher share of urban residents. For Korea, Table A2 compares our sample with the Korea Welfare Panel Study (KoWePS), a national panel survey that oversamples low-income households and includes rural areas; the comparison is restricted to urban residents for consistency.4 Compared to our sample, KoWePS respondents have lower education levels, lower incomes, and higher marriage rates. Overall, the Chinese sample appears reasonably representative, while the Korean sample’s representativeness warrants further consideration.

C. Estimation Model

This study explores the factors associated with recent marriage rate declines in Korea and China, focusing on whether the recent trends are linked to socioeconomic status (SES) indicators, such as individual education levels, incomes and parental backgrounds. The main analysis estimates the association between SES and the likelihood of being never married within each age cohort in each country and assesses how the influence of these SES factors has changed over time based on the following equation:

Here, the dependent variable equals one if individual i is never married in survey year t, and zero otherwise. Survey year fixed effects, β_t, capture temporal changes in never-married rates and are the main coefficients of interest. Note that these coefficients capture ‘growth in the NVMR’, not level differences across SES groups5 Regional fixed effects, δ_c, control for time-invariant geographic differences in marriage patterns: metropolitan cities and provinces in Korea, and counties in China.6 The vectorit X_it includes individual and household characteristics, varying by country. In Korea, these include fixed effects for age and education; in China, they additionally include indicators for urban hukou (household registration status) and urban residence.

The analysis employs an age-fixed cohort framework, meaning that the coefficients of survey year dummies capture changes in the proportion of never-married individuals within age groups, net of age effects. We expect β_t to be positive, indicating a rising NVMR over time. The omitted (reference) survey years are 2003 for Korea and 2010 for China.

A. Trends in Never-married Rates among Men

Figure 4 presents the results for men. Panel (Ka) shows the estimated NVMR growth among Korean men aged 25-45 between 2003 and 2023, controlling for age, education and region fixed effects. The vertical axis represents the change in the NVMR relative to the reference year, 2003. Specifically, the values indicate how much higher (or lower) the NVMR is in each year compared to 2003 for each SES group,

FIGURE 4.

NEVER-MARRIED RATE TRENDS AMONG MEN

Note: This figure presents the estimates of βt for each year in Eq. (1), which captures changes in the proportion of never-married males relative to the reference year. For the Korean sample, the reference year is 2003, and for the Chinese sample, 2010 is omitted to be the reference. We divide the sample according to individual and paternal education levels: (1) LE-LF: less-educated individuals with less-educated fathers; (2) LE-HF: less-educated individuals with more-educated fathers; (3) HE-LF: more-educated individuals with less-educated fathers; and (4) HE-HF: more-educated individuals with more-educated fathers. More details are available in the text.

Two patterns stand out. First, the highest-SES group (HE-HF) consistently exhibits the lowest NVMR growth, while the other three groups show substantially larger increases. This pattern is consistent with prior research indicating that higher socioeconomic status is associated with earlier marriage (Lim, 2021). Second, the gap between the HE-HF group and the lower-SES groups has widened, particularly since around 2010. A plausible explanation is the rapid escalation of living costs during that period, such as housing prices, which likely placed a disproportionate burden on individuals with lower SES levels.8

To examine the role of financial barriers, we compare outcomes between Seoul — a highly urbanised area — and other urban regions. Panels (Kb) and (Kc) present the NVMR growth for residents inside and outside of Seoul, respectively. A notable distinction emerges between the two regions. In Seoul, the groups with less-educated fathers (LE-LF and HE-LF) exhibit the steepest increase in the NVMR in recent years, while the highest-SES group (HE-HF) consistently shows the lowest NVMR growth in Seoul. The NVMR increased by more than 30% points for the LE-LF group by 2023, compared to only about 20%p for the HE-HF group. This pattern contrasts with non-Seoul regions, where NVMR growth among the four groups has been converging over the last several years. Another important observation is that NVMR growth was lowest for the lowest SES group until the mid-2010s in non-Seoul regions, while this group quickly caught up with others after the late 2010s.

The Chinese sample exhibits a similar pattern, as shown in Panels (Ca)-(Cc) of Figure 4. Since 2018, the gap in NVMR growth across SES groups has widened, with less-educated men experiencing the fastest increases. Panel (Ca) indicates that by 2022, their NVMR had risen by nearly 30%, while the increase is less than 15% for HE-HF men. The pattern is more pronounced among men living in urban areas, as shown in Panel (Cb). In rural areas, less-educated men also experience the largest growth in the NVMR, but group differences are smaller and the divergence begins slightly later than in urban areas, also coinciding with the pandemic period.

Overall, these findings suggest that lower-SES men in both Korea and China are experiencing the steepest decline in marriage rates, and the speed has accelerated since the mid-2010s. The pattern is far more pronounced in Seoul in urban China.9 This disproportionate acceleration may indicate a widening gap in marriage opportunities due to financial constraints, effectively eroding the historical pattern in which lower-SES groups married earlier and more universally.10

B. Trends in Never-married Rates among Women

Figure 5 presents the year-specific estimates of βt for women. Panels (Ka) and (Ca) depict the NVMR growth across the four SES groups over time. Overall, NVMR growth speeds are diverging across SES groups in both Korea and China among women, but with different patterns. For Korea, the highest-SES group (HE-HF) consistently exhibits the highest NVMR growth, while lower-SES groups experienced the most rapid growth in the NVMR among Chinese women.

FIGURE 5.

NEVER-MARRIED RATE TRENDS AMONG WOMEN

Note: This figure presents the estimates of βt for each year in Eq. (1), which captures changes in the proportion of never-married females relative to the reference year. For the Korean sample, the reference year is 2003, and for the Chinese sample, 2010 is omitted to be the reference. We divide the sample according to individual and paternal education levels: (1) LE-LF: less-educated individuals with less-educated fathers; (2) LE-HF: less-educated individuals with more-educated fathers; (3) HE-LF: more-educated individuals with less-educated fathers; and (4) HE-HF: more-educated individuals with more-educated fathers. Further details are available in the text.

Panels (Kb)-(Cc) further clarify these differences. The rise in NVMR growth among HE-HF women occurs outside of Seoul. In contrast, the lowest-SES women experienced the fastest growth in never-married rates inside Seoul, consistent with the patterns for Chinese women and for Korean and Chinese men in Figure 4. The NVMR for the lowest-SES women in Seoul increased by almost 25 percentage points over the past decade (from about 0.1 to more than 0.3). The rise is comparable to that for the lowest-SES Chinese women in urban regions (from nearly 0 in 2012 to 0.3 in 2022). On the other hand, outside of Seoul, the HE-HF group exhibits the fastest growth, while that of the LE-LF groups is relatively slow. This suggests that multiple factors are at play in shaping marriage patterns among women, with financial barriers being more salient in highly urbanised areas.

These findings are notable because the rise in never-married men is often attributed to a marriage squeeze (Huang, 2014; Ebenstein, 2010) from imbalanced sex ratios in East Asia, but this logic does not apply to women. For low-SES women in urban areas, delayed marriage may result from migration (Lu and Tao, 2015), which is applicable to the Chinese case. Another potential mechanism is the growing financial burden of marriage, likely to explain the case of Korean women in Seoul and Chinese women in urban areas. Indeed, existing studies find that the major factors for marriage in Korea include rising housing costs and the high financial costs associated with having children (Jones, 2019; Lee et al., 2021). Cultural shifts towards individualism and evolving gender norms may also play a role, although these mechanisms cannot be directly tested in this study.

In China, while education can serve as a proxy for socioeconomic status, an individual’s household registration (hukou) status and residential area (urban vs. rural) are also critical determinants of SES and access to resources (Hung, 2022).11 Accordingly, we divide the Chinese sample into three groups: (1) rural residents with a rural hukou, (2) urban residents with a rural hukou (i.e., migrants), and (3) urban residents with an urban hukou, considered the highest SES group.12

Figure 6 presents the results. Panel (a) shows that the NVMR among Chinese men has increased at similar rates across all three groups. Among women, growth patterns are more heterogeneous: rural residents with a rural hukou show the slowest increase, and urban residents with an urban hukou and migrants (urban residents with a rural hukou) display the fastest growth. As noted in prior research (Koh et al., 2025), many women with a rural hukou migrate to urban areas seeking better marriage prospects. The rapid rise in the NVMR among female migrants possibly suggests growing challenges in finding marriage partners or mismatches within the urban marriage or labour market.

FIGURE 6.

NEVER-MARRIED RATE TRENDS AMONG CHINESE ACROSS REGIONS AND HUKOU GROUPS

Note: This figure presents the estimates of βt for each year in Eq. (1), which captures changes in the proportion of never-married females relative to the reference year. We divide the Chinese sample according to residential area (rural vs. urban regions) and household registration status (rural hukou vs. urban hukou).

A. Results for the Seoul Region in Korea

Given that earlier we noted that the trends are significantly different between Seoul and non-Seoul regions, here we focus only on individuals living in the Seoul region.14 To explore changes in socioeconomic differentials in marriage rates over time, we split the sample into two periods: the 2008-2015 survey years (‘Past’) and the 2016-2023 survey years (‘Recent’). Odd-numbered columns present results based on the earlier period, and even-numbered columns use ‘Recent’ observations. For Korea, individuals are aged between 25 and 45, and for China, between 20 and 40. Columns 1 and 2 include all individuals, while columns 3 to 6 restrict the sample to those with income information.

Table 2 Panel A — Seoul Men: Columns 1 and 2 in Panel A highlight three points: First, the LE-LF group among Korean men now has the highest NVMR, up from the second-highest in the past. Their NVMR is four percentage points above the reference group (HE-LF), while the other two groups are substantially lower (-0.027 and -0.055). Second, this ranking shift appears closely associated with individual income: the gap between LE-LF and HE-LF falls from 3.1%p to 2.0%p once individual income is controlled for, with similar reductions for LE-HF (3.1 to 0.4) and HE-HF (5.8 to 1.9). In short, income strongly explains SES differences in marriage rates. A one-unit increase in log income reduces the likelihood of never having been married by 16.3-18.3 percentage points, and the effect of log annual income has grown over time (by approximately two percentage points).

Table 2 Panel B — Seoul Women: For women, the HE-HF group consistently shows the highest NVMR, while LE-LF and LE-HF remain lower, and these differences are persistent and statistically significant across all specifications, ranging from 0.053 to 0.039 (columns 1 to 6). Importantly, the lowest-SES group once had higher marriage rates (i.e., lower NVMR: -0.024 to -0.044), but this tendency has nearly vanished (0.004), recalling the rapid NVMR increase in recent years, as shown in Figure 5. Unlike men, where LE-LF now have significantly lower marriage rates, female marriage rates are converging across SES groups, except for the highest-SES HE-HF group, whose NVMR continues to rise (2.7 to 4.5 percentage points). In other words, more education and having highly educated fathers are persistently associated with a higher NVMR among women. Income partially explains this difference, as shown by the change in the HE-HF coefficients when income is controlled for: from 0.045 in column 4 to 0.039 in column 6. Income has the opposite effect compared to men, and its positive effect on the NVMR is growing: the income coefficient is small and statistically insignificant in the past (column 5; 0.012), but it becomes larger and highly significant in recent years (column 6; 0.030). In contrast, the coefficients for the other two SES groups remain close to zero in columns 4 and 6, which indicates that, unlike men, income does not explain much of the SES differences in marriage rates among women.

TABLE 2

DIFFERENTIAL SOCIOECONOMIC STATUS EFFECTS OVER TIME: KOREA

Note: This table presents OLS estimates based on Eq. (2). The sample is restricted to individuals residing in Seoul and the surrounding Gyeonggi province. “Past” refers to survey years 2008-2015, and “Recent” to 2016-2023. LE-LF denotes less-educated individuals with less-educated fathers, LE-HF less-educated individuals with more-educated fathers, and HE-HF more-educated individuals with more-educated fathers. HE-LF serves as the reference group. Definitions of less- and more-educated are given in Section 4. Standard errors are clustered at the survey-year level. *p < 0.10, **p < 0.05, ***p < 0.01.

Another notable pattern is that the four socioeconomic indicators and income explain the variation in never-married rates better in recent years than in the past, as indicated by the increase in R-squared values across each pair of columns (e.g., from 0.331 in column 1 to 0.362 in column 2). Adding income improves the model fit for men in Panel A: R-squared increases from 0.277 to 0.319 between columns 3 and 5 and from 0.303 to 0.346 between columns 4 and 6, whereas R-squared changes very little for women in Panel B when income is included. This suggests that economic status plays a more central — and increasingly important — role in shaping marriage outcomes among men, whereas its explanatory power for women remains relatively limited.

B. Results for Urban China

Table 3 presents results for the Chinese sample. Because the sample only covers 2010 to 2022, we define 2010 as ‘Past’ and the 2020 and 2022 samples as ‘Recent’. We restrict the sample to those living in urban areas.

Table 3 Panel A — Chinese men in urban areas: In the past (column 1), all three SES groups had a higher NVMR relative to the reference group (HE-LF), with the HE-HF group showing the largest gap (0.133). By the recent period (column 2), these gaps have narrowed: LE-LF fell from 0.047 to 0.004 and HE-HF from 0.133 to 0.054. This convergence suggests that SES differentials in marriage rates have declined over time. Controlling for income in columns 5-6, the patterns remain similar, though income shows a weak negative association with the NVMR (-0.010 to -0.020), roughly doubling over time. Thus, income plays a modest but growing role in male marriage likelihood in urban China, similar to the Korean case.

Table 3 Panel B — Chinese men in urban areas: Higher-SES women (HE-HF) consistently exhibit a higher NVMR, though the gap decreased from 0.157 to 0.106. The LE-HF group also exhibits smaller gaps in recent years (0.117 to 0.082), while LE-LF shows no significant difference from the reference group. Including income (columns 5 to 6) does not alter these patterns; the income coefficient remains small and statistically insignificant (0.005 to 0.009). This indicates that, unlike men, income does not significantly explain marriage differentials among Chinese women, largely consistent with the pattern among Korean women.

TABLE 3

DIFFERENTIAL SOCIOECONOMIC STATUS EFFECTS OVER TIME: CHINA

Note: This table presents OLS estimates based on Eq. (2). The sample is restricted to individuals residing in urban regions. “Past” refers to the survey year 2010, and “Recent” to 2020-2022. LE-LF denotes less-educated individuals with less-educated fathers, LE-HF less-educated individuals with more-educated fathers, and HE-HF more-educated individuals with more-educated fathers. HE-LF serves as the reference group. Definitions of less- and more-educated are given in Section 4. Standard errors are clustered at the provincial level. *p < 0.10, **p < 0.05, ***p < 0.01.

C. Income Gradients in Never-married Rates Over Time in Both Countries

To examine temporal changes in the link between marriage likelihood and individual financial conditions further, we regress never-married status on individual income percentiles for each survey year separately, controlling for age and regional fixed effects.15 Figure 7 presents the estimated coefficients, with the never-married indicator scaled such that 100 corresponds to never-married and 0 otherwise. Thus, the y-axis reflects the association between income and the NVMR. The top panel reports results for highly urbanised areas — Seoul in Korea and urban regions in China — while the bottom panel shows estimates for less urbanised areas, non-Seoul regions in Korea and rural regions in China. ‘Past’ refers to survey years 2013 for Korea and 2010 for China, and ‘Recent’ refers to survey years 2023 for Korea and 2020 for China, a decade later.

FIGURE 7.

CORRELATION BETWEEN NEVER-MARRIED RATES AND INDIVIDUAL INCOME PERCENTILE

Note: This figure presents the estimated coefficients of income percentile on a binary indicator of being never married (scaled so that this never-married indicator equals 100 for never-married individuals and 0 otherwise). The estimation is based on a regression of never-married status on individual income percentiles for each survey year separately, controlling for age and regional fixed effects. Thus, the y-axis reflects the association between income and never-married rates. The top panel reports results for highly urbanised areas — Seoul in Korea and urban regions in China — while the bottom panel presents estimates for less urbanised areas, i.e., non-Seoul regions in Korea and rural regions in China. Past corresponds to survey years 2013 for Korea and 2010 for China, and Recent refers to survey years 2023 for Korea and 2020 for China, a decade later.

The results reveal several patterns. First, for men in both countries, there is a clear negative correlation between income and never-married rates, indicating that higher income is associated with a lower likelihood of never having been married. For example, in Seoul in recent years, the income coefficient is approximately -0.6, suggesting that for every unit increase in income percentile, the probability of never having been married decreases by 0.6 percentage points. Second, this negative correlation has strengthened over time, particularly in highly urbanised areas. In Seoul, the income coefficient for men changed from approximately -0.45 in the past to -0.6 in recent years. This pattern is observed among urban Chinese men, where the coefficient moves from above -0.1 to below -0.2. Next, the correlation between income and the never-married rate is opposite among women in both countries in recent years. In Seoul, the income coefficient for women is around 0.2 in recent years, indicating that higher incomes are associated with a higher likelihood of never having been married. This positive correlation has also strengthened over time, as the coefficient was negative in the past. A similar pattern is observed among urban Chinese women, where the coefficient increases from far below 0.2 to close to 0.2. The bottom panel shows that these patterns also hold in less urbanised areas, but changes in the strength of the correlations are less pronounced for men. Overall, the results indicate that the association between individual income and marriage likelihood has intensified over time.

A. Macroeconomic Environments and Marriage Trends

The results presented above suggest that the recent rapidly growing NVMR among lower-SES individuals in both Korea and China may be linked to broader macroeconomic challenges that disproportionately affect these groups. One such challenge is the dramatic rise in housing prices, which has significantly increased the financial barriers to marriage and family formation in East Asia.

Figure 8 and Figure A1 confirm that housing affordability has worsened considerably in both countries over the past decade. Crucially, these price increases have disproportionately impacted lower-income households, as evidenced by the widening gap in the price-to-income ratio (PIR) across the income groups in Panel (b) of Figure 8. In 2006, low-income households needed approximately six years of income to afford a home; however, this number began to climb after 2010, reaching ten years by 2016. The PIR gap between low-income households and middle/high-income households has widened since 2010 as well, which suggests that overall housing price increases may disproportionally affect individuals of a lower socioeconomic status. China has also experienced similar trends over the last decade. Hu et al. (2022) analyse the PIRs for 336 cities in China, finding that the urban PIR had been stable until 2015 but surged rapidly afterwards until 2020, as shown in Figure A1.

FIGURE 8.

RECENT TRENDS IN THE HOUSING PRICE-TO-INCOME RATIO IN KOREA

Note: This figure displays the housing price-to-income ratio (PIR) in Korea, which is an indicator that measures housing affordability calculated by dividing median home prices by median annual household income. The data comes from the “Housing Condition Survey” of the Korean Ministry of Land, Infrastructure and Transport. ‘Seoul’ indicates the broader metropolitan area, including Seoul, Incheon and Gyeonggi province. Income groups in Panel (b) are categorised by monthly disposable household income: low income (deciles 1-4), mid-level income (deciles 5-7), and high income (deciles 8-10).

These macroeconomic challenges likely accelerated the NVMR increases among lower-SES individuals by exacerbating the financial burdens associated with marriage and child-rearing. For example, using China’s 2010 Population Census data, Qiu and Liu (2025) find that a ten percent increase in housing prices delays the age of first-time marriage by 0.91 months for the urban-hukou population. Similarly, Lee et al. (2021) find that housing prices have been a delaying factor in marriage in Korea. These findings suggest that rising housing costs create significant barriers to marriage, particularly for lower-SES individuals, who are more sensitive to financial constraints.

B. Delaying Marriage versus Opting not to Marry

A question that arises when exploring the recent, rapidly growing NVMR is whether this trend reflects postponed marriages or individuals choosing permanent singlehood. To investigate this, we divide the sample into two age groups — younger (20-30 for China; 25-35 for Korea) and older (30-40 for China; 35-45 for Korea). We then estimate Eq. (1) separately for each age group and SES group. These results are presented in Figure 9, which shows the estimated coefficients of year dummies (β_t) from Eq. (1) for each year, capturing changes in the proportion of never-married individuals relative to the reference year (2013 for Korea; 2010 for China).

FIGURE 9.

NEVER-MARRIED RATE TRENDS BY AGE GROUP

Note: This figure presents the estimates of βt for each year, which captures changes in the proportion of never-married males relative to the reference year. For the Korean sample, the reference year is 2003, and for the Chinese sample, 2010 is omitted to be the reference.

The results indicate that in China (Panels a and b), recent declines in marriage rates are concentrated among individuals in their twenties, suggesting that the trend reflects delayed marriage. This pattern is especially pronounced among less-educated groups.16 A similar pattern is observed among Korean women.

In contrast, Korean men exhibit rapid increases in their NVMR across both younger and older cohorts. Among younger men (Panel c), the four SES groups do not diverge substantially.

However, the pattern among older men (Panel d) is markedly different. Korean men aged 35-45 show sizable increases in their NVMR, with the largest rise occurring among lower-SES groups. Panel (d) indicates that the LE-LF group experienced roughly a 15-percentage-point increase between 2013 and 2023, compared to less than five percentage points for the HE-HF group. This widening gap suggests that in Korea, the rising NVMR among lower-SES men reflects not only delayed marriage but an emerging pattern of long-term or permanent singlehood. This distinction is important: whereas the age profile of changes among Chinese and Korean women points primarily to delayed marriage, the trajectory for Korean men indicates deeper structural barriers to family formation, particularly for those with fewer socioeconomic resources.

C. Policy Implications

The findings of this study suggest two main policy intervention directions. First, because never-married rates have increased most rapidly among low-SES individuals in highly urbanised areas, policies should prioritise reducing the financial barriers to marriage and early family formation. Rising housing costs, income instability, and high expected child-rearing expenses appear to deter marriage disproportionately among these groups in recent years. Targeted support — such as housing assistance for young adults, preferential mortgage or rental schemes for first-time households, and programmes that improve income security for low-SES youth — may help ease these constraints. Reductions in anticipated family costs, including expanded public childcare and more accessible parental leave, may further lower the economic threshold associated with marriage.

Second, the faster pace of NVMR growth among lower-SES groups has led to convergence in overall NVMR levels, which suggests that delayed or forgone marriage is becoming a general phenomenon across all socioeconomic groups. Figure 10 shows that the NVMR gaps between high- and low-SES groups have narrowed in both Korea and China over the past decade, except for that of Korean men.17 This narrowing is driven by the lower-SES groups — who historically married earlier and more universally — experiencing the most rapid acceleration in singlehood, thereby eroding their previous advantage in family formation.

FIGURE 10.

CONVERGENCE IN NEVER-MARRIED RATES ACROSS SOCIOECONOMIC GROUPS

Note: This figure displays the proportion of never-married individuals according to socioeconomic status (SES) groups in Korea (Panels a and b) and China (Panels c and d). For the Korean sample, SES groups are defined based on individual education and paternal education, as in the previous analyses. For the Chinese sample, SES groups are defined based on residential and household registration locations. The y-axis represents the never-married rate, while the x-axis indicates the survey years.

This convergence implies that policies should not focus solely on targeted economic incentives for marriage but also on adapting institutional frameworks to reflect these broad structural shifts. Rather than attempting to reverse emerging attitudes toward autonomy, gender equality, or life-course diversity, governments could redesign existing systems to be less dependent on marital status — such as expanding eligibility for housing, taxation, and social insurance benefits to single-person and non-traditional households. Strengthening support for well-being and social inclusion regardless of marital status may be essential if attempting to align policy with this emerging demographic landscape.

Together, these approaches acknowledge both the financial constraints that disproportionately affect disadvantaged groups and the broader structural and cultural shifts that are reshaping family formation in contemporary East Asia.