The gender wage gap: evidence from South Korea

My analysis is based on the KLIPS, conducted each year by the Korea Labor Institute from 1998 to 2020. It consists of the household dataset from the household questionnaire and the individual dataset from individual questionnaires. The initial data comprises labor activities and outcomes among 5,000 urban households and all households’ members aged ≥15.

The share of the population living in urban areas was 87.5% and 91.8% in 1998 and 2020, respectively (Source: Ministry of Land, Infrastructure and Transport in South Korea). Most Koreans reside in urban areas.

The goal of this paper is to provide explanations for the gender wage gap in South Korea from 1998 to 2020. I restrict my sample to full-time workers between the ages of 20 and 59 and those who are nonfarm, nonmilitary, wage, and salaried workers.

The sample excludes the following industries: agriculture and forestry (KSIC 01-02), fishing (KSIC 05), public administration and defense (KSIC 76), activities of household as employers (KSIC 95), and activities of extraterritorial organizations and bodies (KSIC 99) in the 8th Korean Standard Industrial Classification (KSIC).

Table 1 presents descriptive statistics for full-time workers by gender and year. From the table, the average age, as a proxy for work experience, has increased for both full-time male and female workers. Among full-time female workers, the average age was about 33.6 years in 1998 and about 6.8 years higher in 2020. There was a less dramatic increase among full-time male workers. The gender gap in the average age has decreased from about 3.0 years to 1.7 years. In the case of having obtained a bachelor's or advanced degree, while male workers were more likely than female workers to have a bachelor's or advanced degree in 1998 and 2020, female workers have narrowed the gender gap in having obtained a bachelor's or advanced degree over the time. The improvement in females’ incidence of having obtained an advanced degree is much more pronounced. The gender gap for the advanced degree has decreased from 3 percentage points to 1 percentage point. The average working hours per month for both male and female workers decreased in 1998 and 2020. The average hours worked per month for male workers fell by approximately 41.5 h and those for female workers decreased by approximately 36.57 h over the period. The gender difference in working hours decreased from about 11.6 h to about 6.7 h. Regarding tenure at the current workplace, the average tenure for both full-time male and female workers increased, being about 2.7 years higher for male workers and about 3 years higher for female workers over the period. In 1998, full-time male workers stayed longer than full-time female workers in one workplace by about 2.3 years. By 2020, the gender gap for tenure had decreased slightly to about 1.9 years. The union coverage rate for full-time male and female workers declined significantly over this time. Among full-time male workers, the union coverage rate was approximately 0.63 in 1998, and was 0.5 lower in 2020. There was a similar decrease among full-time female workers. The gender gap in the union coverage rate decreased during the period. The gender difference in the union coverage rate fell from 0.04 in 1998 to 0.02 in 2020. As for the incidence of professional jobs, females were more likely than males to have a professional job over the period.

Professional jobs include science professionals, computer-related professionals, engineering science professionals, health and medical professionals, teaching professionals, administration, business and finance professionals, legal, social welfare and religion professionals, and culture, arts, and broadcasting professionals in the 5th KSCO.

In the case of marital status and having children, full-time male workers were more likely than female workers to be married or have children over the period. For example, in the KLIPS, while 75% of full-time male workers were married and 72% had children, 52% of full-time female workers were married and 55% had children in 1998. This reflects the low labor participation of married females and females with children. Over the time, the marriage age increased and the total fertility rate decreased in South Korea. The proportion of married males and the proportion of males with children among full-time male workers narrowed significantly over time.

The average real monthly wage for males and females increased over the period.

There is no top coding in the KLIPS.

Figure 2 compares trends in the labor force participation rate for male and female workers from 1998 to 2020 in South Korea and includes the OECD average. The male labor force participation rate decreased from about 94% to about 88%, while the female rate rose from about 56% to 67% over the period in South Korea. Despite this increase in the female labor force participation rate, the gender gap in the participation rates is still large: about 21 percentage points in 2020. The labor force participation rate for the OECD average male workers was 93% in 1998 and about 90% in 2020, while the same rate for female workers was about 68% in 1998 and about 73% in 2020. The gender gap in labor force participation rate narrowed from about 25 percentage points in 1998 to about 17 percentage points in 2020. While the level of the male labor force participation rate in South Korea is similar to the OECD average, the level of the female labor force participation rate in South Korea is lower than the OECD average. The difference in the female labor force participation rate between South Korea and the OECD average fell from about 12 percentage points in 1998 to about 6 percentage points in 2020.

Figure 2

Figure 3 shows the female/male monthly wage ratio at the mean, 10th, 50th, and 90th percentiles for 1998 and 2020. Each value takes the exponential of the unconditional gender log wage gap at the mean, 10th, 50th, and 90th percentiles and obtains the female/male monthly wage ratio. For example, at the 10th percentile, this indicates a monthly wage gap between the male's wage at the 10th percentile over the male wage distribution and the female's wage at the 10th percentile over the female wage distribution. The female/male wage ratios at the mean, bottom, and middle of the wage distribution increased gradually over the period. For instance, the bottom of the wage distribution experienced a rise, with the ratio increasing from 71.4% to 75.0%. Females’ relative monthly wage at the middle of the wage distribution was 61.7% in 1998 and 73.3% in 2020. Compared to the female/male wage ratio at the bottom and middle of the wage distribution, the female/male wage ratio at the top of the wage distribution declined from 71.4% in 1998 to 65.5% in 2020. Table A1 in Appendix presents the shares of full-time male and female workers below the 10th percentile, below the 50th percentile, and above the 90th percentile of the total wage distribution that considers both male and female wages. For example, the log monthly wage at the 10th percentile was about 6.60 in 1998 and about 7.36 in 2020. The share of full-time male workers who earned below the 10th percentile was 26.72%, and that of full-time female workers was 73.28% in 1998. The share for males was 31.65% and that for females was 68.35% in 2020. The female share decreased over the period, but females accounted for a much larger share of total workers who earned below the 10th percentile. Conversely, the share of full-time female workers who earned above the 90th percentile was much lower than that of full-time male workers in 1998 and 2020. Table A2 in Appendix presents the shares of married male and female workers among full-time male and female workers by wage percentile. In 1998 and 2020, the shares of married female workers were higher than those of married male workers in the wage below the 10th percentile. For example, the ratio of married female workers to full-time female workers was 64.64% in 1998 and 67.05% in 2020. On the other hand, the shares of married male workers were higher than those of married female workers in the wage above the 90th percentile.

Figure 3

Figure 4 presents the female/male log monthly wage ratios by age cohort (20–29, 30–39, 40–49, and 50–59) for 1998 and 2020. Taken overall, the monthly wage ratios for all cohorts increased over time, which implies that the gender wage gaps narrowed across cohorts over the period. The largest wage ratio was for the youngest cohort of full-time workers, those aged 20–29 years, in 1998 and 2020. The gender wage gaps tended to be larger as workers aged in 2020. From Table A3 in Appendix, among full-time workers, the shares of female workers were lower than those of male workers, except for those in the age cohort 20–29.

Figure 4

Under the seniority-based wage system in South Korea, older workers are paid more than younger workers because older workers have longer tenure or more work experience than younger workers. This fact may explain why the gender wage gaps for older workers are larger than those for younger workers. Since female workers are more likely to leave the labor market for a short or long period due to marriage or childcare, tenure or work experience for female workers is lower than that for male workers of the same cohort. In Table 2, the shares of married males and females increased significantly for age cohort 30–39. For example, the ratio of married females to full-time female workers aged 30–39 years was 84.38% in 1998 and 62.12% in 2020. The shares for married males and females aged 30–39 years in 2020 declined significantly compared to those in 1998, which reflects the delay in the age of first marriage in South Korea,

In 1998, the male mean age at first marriage was 28.83 years and the female mean age at first marriage was 26.02 years. The age of first marriage gradually increases, and in 2020, the male mean age at first marriage was 33.35 years and the female mean age at first marriage was 31.08 years (Source: KOSIS).

Figure 5 presents female labor force participation rates by age cohorts for South Korea, Japan, and Italy, and includes the OECD average.

In Figure 1, Japan is an Asian country among OECD countries and has the higher gender wage gap. Italy has a lower gender wage gap, but it is one of the OECD countries with a low female labor force participation rate.

Figure 5

In this section, I decompose the wage gap for full-time male and female workers into the explained portion, accounted for by measuring labor market characteristics, and the unexplained portion. I use the Oaxaca–Blinder decomposition of log monthly wages in 1998 and 2020. The Oaxaca–Blinder decomposition measures both the explained and unexplained components of the mean wage gap by estimating the separate male and female wage equations for individuals (e.g., Blinder, 1973; Oaxaca, 1973). The following equations present the Oaxaca–Blinder decomposition: (1) Wm=Xmβm+εm {{W}_{m}}={{X}_{m}}{{\beta }_{m}}+{{\varepsilon }_{m}} (2) Wf=Xfβf+εf {{W}_{f}}={{X}_{f}}{{\beta }_{f}}+{{\varepsilon }_{f}} (3) W¯m−W¯f=bmX¯m−bfX¯f=bm(X¯m−X¯f)+X¯f(bm−bf) {{\bar{W}}_{m}}-{{\bar{W}}_{f}}={{b}_{m}}{{\bar{X}}_{m}}-{{b}_{f}}{{\bar{X}}_{f}}={{b}_{m}}\left( {{{\bar{X}}}_{m}}-{{{\bar{X}}}_{f}} \right)+{{\bar{X}}_{f}}\left( {{b}_{m}}-{{b}_{f}} \right) where W is the log monthly wages and m and f denote male and female, respectively; X_m and X_f are vectors containing the explanatory variables and a constant, and the explanatory variables include age, age squared, level of educational attainment (middle school, high school, associate degree, bachelor's degree, and advanced degree), tenure at the current workplace, hours of work, union status, the size of the establishment, industry, and occupation

I categorize the size of the establishment into 10 sectors: 1–4 workers, 5–9 workers, 10–29 workers, 30–49 workers, 50–69 workers, 70–99 workers, 100–299 workers, 300–499 workers, 500–999 workers, and over 1,000 workers. The base group is the establishment with 1–4 workers.

Table 3 displays the results of the Oaxaca–Blinder decomposition for 1998 and 2020. First, the gender differences in age substantially contributed 0.122 log points and accounted for 29.1% of the gender wage gap in 1998. Compared to the effects of age in 1998, the much smaller gender differences in age contribute 0.017 log points or 5.1% of the gender wage gap in 2020. Age is used as a proxy for work experience because actual work experience is not available in the KLIPS. Age may overstate the years that workers actually participate in the labor market, when in fact workers may leave the market because of schooling, training, or child-care. The differences between actual and potential work experience are relatively larger for female workers. Therefore, overstating work experience may cause an underestimation of its impact on the gender wage gaps. Second, gender differences in education accounted for 3.9% and 1.3% of the gender wage gaps in 1998 and 2020, respectively. This may reflect the improvement in females’ education level during this period. Table 1 shows the decrease in the gender gap in cases of obtaining the bachelor's and advanced degrees – in particular, the improvement in females’ having obtained an advanced degree is much more significant. Third, the gender differences in tenure with the current workplace accounted for a substantial share of the gender wage gaps in 1998 and 2020, thus playing an important role in explaining the wage gaps. In South Korea's labor market, the seniority-based wage system is affected by tenure; this system also determines workers’ wages. Female workers are much more likely than male workers to leave the labor market as they get married or have children. In Figure 6, the difference in labor force participation rates between males and females in 1998 and 2020 increased significantly when they married or cared for their children.

In South Korea, males and females got married in their late 20s in 1998 and in their early 30s in 2020, on average.

Figure 6

The total gender wage gap in South Korea declined from 0.418 log points in 1998 to 0.338 log points in 2020. While the explained gender wage gap decreased from 0.202 log points to 0.111 points, the unexplained gap increased from 0.216 log points to 0.226 log points. The unexplained factors accounted for a larger share of the gender wage gap in both 1998 and 2020—51.7% and 67.0%, respectively—increasing the relative importance of these unexplained factors. Based on this decomposition analysis, a substantial share of the total gender wage gap cannot be explained by measured labor market characteristics.

In this section, I examine quantile decomposition of the gender wage gap at each point of the wage distribution. I focus on the unexplained gender wage gaps across the wage distribution, following the methodology of Chernozhukov et al. (2013). This methodology offers estimation and inference procedures regarding the entire counterfactual distribution of outcome (e.g., wages) and its functions, based on quantile regressions. It estimates the empirical distribution of the characteristics and uses quantile regressions for the conditional wage distribution. The empirical study decomposes the unconditional wage gaps between males and females into explained components and unexplained components at each percentile of the wage distribution. The explained gap is attributable to the distribution of the measured labor market characteristics, and the unexplained gap is attributable to the different wage functions conditional on the labor market characteristics. Let m denote the male workers and let f denote the female workers. The variable Y_i is log monthly wages and X_i is labor market characteristics. F_Ym|Xm(y|x) and F_Yf|Xf(y|x) are the conditional distributions of wages to male and female workers given their characteristics, respectively. F_Y[m|m] and F_Y[f|f] are the unconditional distribution functions of the wages for male and female workers, and F_Y[m|f] is the counterfactual wage distribution function for female workers if they received wages according to the male worker's wage function. F_Xm and F_Xf are the distribution of male and female characteristics, respectively. (4) FY[m|m](y)=∫FYm|Xm(y|x)dFXm(x) {{F}_{Y\left[ m|m \right]}}\left( y \right)=\int{{{F}_{Ym|Xm}}\left( y|x \right)d{{F}_{Xm}}\left( x \right)} (5) FY[f|f](y)=∫FYf|Xf(y|x)dFXf(x) {{F}_{Y\left[ f|f \right]}}\left( y \right)=\int{{{F}_{Yf|Xf}}\left( y|x \right)d{{F}_{Xf}}\left( x \right)} (6) FY[m|f](y)=∫FYm|Xm(y|x)dFXf(x) {{F}_{Y\left[ m|f \right]}}\left( y \right)=\int{{{F}_{Ym|Xm}}\left( y|x \right)d{{F}_{Xf}}\left( x \right)}

Eq. (6) is the counterfactual distribution, which is constructed by integrating the conditional distribution of male wages given the distribution of female characteristics. The assumption for the counterfactual distribution is that a change in the characteristics does not affect the parameters of the conditional distribution of wages given the characteristics. The counter-factual distribution reflects either a change in the conditional distribution of wages given the characteristics or a change in the distribution of characteristics. (7) FY[m|m]−FY[f|f]= {FY[m|m]−FY[m|f]}+{FY[m|f]−FY[f|f]} {{F}_{Y\left[ m|m \right]}}-{{F}_{Y\left[ f|f \right]}}=\,\left\{ {{F}_{Y\left[ m|m \right]}}-{{F}_{Y\left[ m|f \right]}} \right\}+\left\{ {{F}_{Y\left[ m|f \right]}}-{{F}_{Y\left[ f|f \right]}} \right\}

Eq. (7) presents the decomposition of the difference in the unconditional wage distribution between full-time male and female workers. The first term on the right-hand side of Eq. (7) indicates the effect of differences in characteristics. Hence, this term demonstrates the impact of differentials in the characteristic distribution between male and female workers, given the same wage schedule for males. The second term refers to the effect of the difference in the conditional distribution of wages. Hence, this term presents the impact of the difference in male and female workers’ wage schedules, given the distribution of female characteristics.

Figure 7 presents the quantile regression decompositions of the gender wage gaps in 1998 and 2020. The red dashed lines plot the explained wage gaps and the green dotted lines indicate the unexplained wage gaps. Figure 7A shows that the unexplained gender wage gap was much larger than the explained wage gap below the 10th percentile. The explained and unexplained portions took similar shares of the total gender wage gap, except at the bottom and top of the distribution. The unexplained gap sharply decreased at the top of the wage distribution. In 1998, the unexplained factors contributed substantially to the gender wage gap at the bottom of the wage distribution. This implies the possibility of the sticky floor effect, which means that female workers in low-paying jobs have low mobility and tend not to rise to higher wage levels unless or until they accumulate skills appropriate to such incomes.

Figure 7

Figure 7B shows that the magnitude of the unexplained wage gap over the wage distribution was higher than that of the explained gap. The unexplained factor contributed more to female workers’ wages, regardless of their positions in the labor market. The unexplained wage gap showed an increasing pattern over the distribution. This implies the possibility of the glass ceiling in high-level jobs in South Korea, which means that female workers are more likely to face barriers to entry or promotion to high-level jobs or that they are paid less than their male counterparts in high-level jobs. This may cause a relatively higher gender wage gap at the top of the distribution.

A comparable pattern is found for Sweden (Albrecht et al., 2003) and several other European countries, such as Finland, Italy, and Poland (Arulampalam et al., 2007; Christofides et al., 2013).

Comparing Figures 7A and 7B, we observe that the explained gender wage gap declined significantly in 2020 relative to 1998, except at the upper top of the wage distribution. This implies that full-time workers appear similar to their male counterparts in terms of the measured labor market characteristics in South Korea over this time. On the other hand, the unexplained gender wage gap in 2020 was higher than that in 1998 from the 57th percentile and above. The decline in the unexplained gap between 1998 and 2020 is smaller than that in the explained gap.

Like other Asian countries, South Korea has a long tradition of strict gender norms, such as a preference for sons over daughters, a male-dominated labor market, and a greater responsibility assigned to females for housework and childcare. Such stereotypical gender norms seem to disappear in a family and society when females enhance their education levels and become actively involved in the labor market. However, traditional gender norms persist in South Korea (Choi and Hwang, 2015; Hwang et al., 2019). Choi and Hwang (2015) demonstrated the significant son–daughter difference in parental treatment on children's time spent on housework and private education expenditure from 1998–2010 in South Korea. Sons are more likely than daughters to spend less time on housework and to receive greater support for education. This in turn reflects the son–daughter difference in parents’ expectations for their child's career. Parents of sons are more likely to expect them to choose professional or high-wage jobs. Such differences are evidence of persistent stereotypical gender norms, which also manifest in the labor market in the form of lower female participation and higher gender wage gaps compared to other developed countries. This may explain why employers discriminate against female workers, regardless of their position in the workplace, leading to the relatively large unexplained gap over most of the wage distribution in South Korea.

Traditional gender role norms decree that males are more likely to be the primary earners and females to have greater responsibility for housework and childcare. In particular, after getting married and having a child, female workers are more likely to leave the labor market or to invest less in their career so as to spend more time on their home duties (Becker, 1991; Albanesi and Olivetti, 2009; Kleven et al., 2019). Although females have improved their education, work experience, and commitment to the labor market over this period, getting married can still negatively affect their labor market outcomes (e.g., Korenman and Neumark, 1991; Antonovics and Town, 2004; Juhn and McCue., 2016). In South Korea, where traditional gender norms persist, females spend much more time than males performing home duties. Even if females work in the labor market like males, females are still much more responsible for housework (e.g., Hwang et al., 2019). Employers may expect married female workers to have less attachment to the labor force or to be more likely to leave their job. This may have negative effects on female employment and wages. Blau and Winkler (2017) suggested that mothers have fewer opportunities for firm-specific training and straggle to obtain the benefits that result from being matched with appropriate jobs. Thus, females may be employed in mismatched jobs or miss out on the wage premiums offered to male workers.

Figure 4 shows that the monthly wages for full-time female workers decreased relative to their male counterparts upon reaching their 30s. This may reflect the impact of marriage and childbirth, as males and females in South Korea tend to get married, on average, in their early 30s. First, I estimate the following statistical model to examine the relationship between labor market outcomes and marital status. To examine the effects of marriage independent of the child effects, my sample data include only full-time male and female workers without children. (8) Yit=β0+β1+Female+β2Married+β3Female*Married+β4Xit+ɛit {{Y}_{it}}={{\beta }_{0}}+{{\beta }_{1}}+Female+{{\beta }_{2}}Married+{{\beta }_{3}}Female*Married+{{\beta }_{4}}{{X}_{it}}+{{\varepsilon }_{it}} where i and t index an individual and time, respectively; Y_it represents log monthly wage, log hourly wage, log hours of work, and labor supply as the dependent variables; the Female and Married variables are indicators for female and marital status, respectively; the interaction term, Female * Married, is a dummy that equals 1 if the females are married; and X_it is a vector of the explanatory variables and year dummy, including a quadratic in age, education, tenure at the current workplace, hours of work, union, the size of the establishment, industry, and occupation.

Table 4 indicates that marriage may be a determinant of the gender wage gap. Specification (1) shows that average full-time female workers who are single are likely to receive a monthly wage 15.1 percentage points lower than the average full-time male worker who is single. Average full-time male workers who are married without children are likely to earn a monthly wage 14.1 percentage points greater than the average single full-time male worker. The relative loss due to marriage, the difference in differences of changes in monthly wage between males and females, is 8.9 percentage points of the average monthly wage. That is, marriage leads to a monthly wage that is 8.9 percentage points less than that for full-time female workers. In specification (2), there is a 15.8 percentage point mean difference in hourly wage between single male and female workers. There is a 17.3 percentage point mean gap in hourly wage between married and single males. Average full-time female workers are likely to earn 11.7 percentage points less due to marriage. In specification (3), average single female workers who work full time are likely to work 4.5 percentage points less than average single male workers. Average married male workers are 1.3 percentage points more likely to work than average single males. There is no significant effect of marriage on females’ working hours. Specification (4) shows that single females are 6.2 percentage points more likely to work than average single males. There is a 43.6 percentage point gap in labor supply between married and single males. Marriage leads to a 46.3 percentage point decrease in labor supply for females. Table 4 shows that marriage has negative effects on labor market outcomes for females.

Juhn and McCue (2017) provided evidence for the impact of marriage on earning differences among full-time married workers in the US. Married full-time male workers earned more than single male workers across 5-year birth cohorts from 1936 to 1985. Married full-time female workers earned less than single female workers for the 1936–1945, 1946–1955, and 1956–65 birth cohorts, whereas they have earned more than single female workers since the 1966 birth cohort. Juhn and McCue (2016, 2017) suggested that marriage no longer has negative effects on the wages of female workers without children. Comparing my results with those in prior studies, it appears that the negative effect of marriage on average monthly wages, hourly wages, and labor supply for females without children in South Korea can be one explanation for the large gender wage gap.

To examine the dynamic paths of labor market outcomes between married full-time male and female workers, I use an event study around the years in which the workers got married. To trace the paths of labor market outcomes between males and females after marriage, I draw on a quasi-experiment based on the event study of Kleven et al. (2019). In my event study, I use an unbalanced panel of males and females who got married at any time after 2000, focusing on the period between the 2 years prior to getting married and the 2 years following marriage. To examine the impact of marriage independent of the child effects, I restrict my sample to married males and females without children. According to the Korean Statistical Information Service (KOSIS), the fertility rate for married females in South Korea was 96% in 2010 and 91.6% in 2020, which means that most married females gave birth. In addition, the first birth interval (time interval between marriage and first childbirth) was, on average, 2 years. Hence, I analyze only the effects of marriage on labor market outcomes for the period between the 2 years before getting married and the 2 years after.

The assumption of the event study in this paper is that getting married does not depend on unobserved expectations about labor market outcomes that workers would have experienced were they not getting married. The event (e.g., marriage) may cause two effects on labor market outcomes: pre-event effect and post-event effect. For example, the pre-marriage effect means that females may have less incentive to invest in their education and career in anticipation of getting married. The post-marriage effect is that females may change their job conditions in response to their marriage, such as moving to low-paying jobs or reducing their hours of work after getting married. The event study can identify only the post-event effect. Adda et al. (2017) and Kleven et al. (2019) suggested that the anticipation of motherhood does not have significant effects on labor market outcomes. The event is not determined by labor market outcomes, which means that the outcomes do not occur before getting married, whereas the significant changes in the outcomes occur just after getting married. Table 1 provides descriptive evidence that females improved their human capital investment over the time in South Korea. In my event graphs, I identify stable and identical pre-trends between males and females for the 2 years prior to the event. This implies that there is no significant pre-event effect.

The event study analyzes the dynamic paths of labor market outcomes, such as the monthly wage and labor supply around the event for full-time male and female workers. The estimation equation has the form: (9) Yistg=∑j≠−1αjg⋅I[j=t]+∑kβkg⋅I[k=ageis]+∑yγyg⋅I[y=s]+ηistg, Y_{ist}^{g}=\sum\limits_{j\ne -1}{\alpha _{j}^{g}\cdot \text{I}\left[ j=t \right]+}\sum\limits_{k}{\beta _{k}^{g}\cdot \text{I}\left[ k=ag{{e}_{is}} \right]+\sum\limits_{y}{\gamma _{y}^{g}\cdot \text{I}\left[ y=s \right]+\eta _{ist}^{g},}} where Yistg Y_{ist}^{g} is the labor market outcome variable for individual i of gender g in year s at event time t; t indicates all years relative to getting married for t = −2 to t = 2, and t = 0 denotes the year in which the individual got married. The equation is estimated separately for males and females and includes full sets of event time dummies, age dummies, and year dummies to control for different age profiles and business cycles. The coefficients in Eq. (9) are normalized to 0 for each gender in the year before getting married (t = −1). I specify Yistg Y_{ist}^{g} in levels because one of the variables (i.e., employment) produces 0 as an output – specifically, 1 if a person is employed and 0 if a person is unemployed. Thus, I compute the percentage effect (Ptg) \left( P_{t}^{g} \right) by dividing each coefficient through the predicted labor market outcome of gender g when omitting the event dummies: Ptg≡α^tgE[Y˜istg|t] P_{t}^{g}\equiv {}^{\hat{\alpha }_{t}^{g}}\!\!\diagup\!\!{}_{E\left[ \tilde{Y}_{ist}^{g}|t \right]}\; , where Y˜istg≡∑kβ^kg⋅I[k=ageis]+∑yγ^yg⋅I[y=s] \tilde{Y}_{ist}^{g}\equiv \sum\limits_{k}{\hat{\beta }_{k}^{g}\cdot \text{I}\left[ k=ag{{e}_{is}} \right]+}\sum\limits_{y}{\hat{\gamma }_{y}^{g}\cdot \text{I}\left[ y=s \right]} . The percentage effect indicates the effect of the marriage at event time (t) as a percentage of the counterfactual labor market outcome for males and females – that is, not getting married. The impact of marriage on labor market outcomes for females relative to males at event time t is: (10) Pt≡α^tm−α^tfE[Y˜istf|t] {{P}_{t}}\equiv \frac{\hat{\alpha }_{t}^{m}-\hat{\alpha }_{t}^{f}}{E\left[ \tilde{Y}_{ist}^{f}|t \right]} Eq. (10) measures the percentage by which marriage affects females’ labor market outcomes (e.g., wage, labor supply) compared to males, at event time (t). Figure 8 shows graphical evidence of the effects of marriage on labor market outcomes Ptm P_{t}^{m} and Ptf P_{t}^{f} , controlling for age and year fixed effects. Each graph plots the estimated coefficients from Eq. (9) and shows how the labor market outcomes of individuals who got married are different from those who remained unmarried over the period.

Figure 8

Figure 8 shows the dynamic effects of marriage on labor market outcomes for full-time male and female workers. The patterns of labor market outcomes for full-time female workers are similar to the patterns for full-time male workers prior to the year in which they got married (t = −2 and t = −1). In Figures 8A–C, while the monthly wage, hourly wage, and hours of work for the full-time female workers fell slightly behind those for the full-time male workers after they got married, they are not statistically significant. Females’ wages and working hours do not significantly change in the first few years of marriage. In other words, if females remain in the labor market after getting married, their wages and hours of work do not decrease for the 2 years after getting married. Figure 8D plots the dynamic paths of labor supply across the event time. It indicates that female workers are more likely than males to leave the labor market after marriage. The labor supply for females shows a drop of about 22.4% in the year in which they get married (t = 0) relative to the years prior to marriage. The negative effect of marriage for females relative to males is 20.6% in the 2 years after getting married (t = 2), whereas the labor supply for males does not change, regardless of marriage. My findings imply that, in the short run, there is no significant difference between males and females in the effect that marriage produces on wages and working hours, but a significant number of females leave the labor market following marriage.

Next, I discuss how the presence of children can affect the labor market outcomes for full-time workers. As per the Korean parental leave and childcare leave system according to the Labor Standards Act in South Korea, an employer must provide a pregnant female with a total of 90 days of maternity leave. The female receives a payment equivalent to her ordinary wages for the period of maternity leave. In addition, the Equal Employment Opportunity and Work-Family Balance Assistance Act of 2008 introduced paternity leave. The Act stipulated 3 days of paternity leave, extended in 2012 to 5 days and in 2019 to 10 days. A male worker on paternity leave receives his normal pay for the duration of the leave. According to the Equal Employment Opportunity Act in South Korea, the introduction of childcare leave has been in effect since 1988.

The Equal Employment Opportunity Act was amended to the Equal Employment Opportunity and Work-Family Balance Assistance Act in 2007.

First, I analyze the following specifications to examine the relationship between labor market outcomes and the presence of children. (11) Yit=β0+β1Female+β2Children+β3Female*Children+β4Xit+ɛit {{Y}_{it}}={{\beta }_{0}}+{{\beta }_{1}}Female+{{\beta }_{2}}Children+{{\beta }_{3}}Female*Children+{{\beta }_{4}}{{X}_{it}}+{{\varepsilon }_{it}} where i and t index an individual and time, respectively; Y_it represents labor market outcomes as the dependent variable; the Female and Children variables are indicators for female and the presence of children, respectively; the interaction term is a dummy that equals 1 if the female has a child; and X_it is a vector of the explanatory variables and year dummy, as in Eq. (8).

Table 5 shows the average effects of the presence of children on wages, hours of work, and labor supply. In specification (1), there is an 18.0 percentage point mean difference in the monthly wage between males and females without children. Average full-time male workers with children are more likely to earn a monthly wage 16.0 percentage points higher than average full-time male workers without children. The relative loss due to the presence of children, the difference in differences of the changes in the monthly wage between males and females, is 23.8 percentage points of the average monthly wage, which means that full-time female workers who have children suffer negative effects on their monthly wage. In specifications (2)–(4), the presence of children negatively affects females’ hourly wage, working hours, and labor supply. The presence of children results in a decline of 4.1 percentage points in the hours of work for full-time female workers. Full-time female workers with children experience a relatively small drop in the average hours of work. Such a small decrease in hours of work reflects the fact that the hours of work for full-time workers in the Korean labor market can hardly be adjusted. In South Korea, childcare costs are expensive because of high expenditures on private childcare services and fewer public childcare programs.

Although the Korean government has increased public spending on childcare and early childhood education over the last decade, public spending on family benefits—including child-related cash transfers, services for families with children, and financial support for families through the tax system—is still low compared to other OECD countries. South Korea spent 1.2% of its GDP on family benefits in 2018, which is much lower than the OECD average (2.1%) (Source: OECD Family Database).

I use the event study around the years in which full-time workers give birth to their first child to examine the dynamic trajectory of labor market outcomes. This event study focuses on an unbalanced panel of full-time workers who had their first child any time after 2000. Figure 9 shows the effects of childbirth on monthly and hourly wage, hours of work, and labor supply. Figure 9A shows that, while the monthly wages of the full-time male workers who had their first birth experience increased, the full-time female workers experienced a decrease in their wages. The monthly wage between males and females diverges after first childbirth; in particular, the monthly wage gaps from event time 7 to event time 10 are significant, and the full-time female workers fall behind about 33.4% relative to the full-time male workers 10 years after childbirth. The magnitude of the negative effects on the wage gap in South Korea is much larger than the gaps in Sweden and Denmark (Albrecht et al., 2018; Kleven et al., 2019). For example, the earning gap between males and females in Denmark is 19.4% 10 years after the first childbirth. Figure A1 in Appendix expands the event time to include 15 years after the first childbirth.

15 years after birth is the period during which children complete middle school in South Korea.

Figure 9

While my event study for child effects on full-time workers starts at the birth of the first child, it is possible to include the effects of the birth of a second or third child, which may cause a further decrease in wages and labor supply for female workers. According to the OECD database, the total fertility rate in South Korea gradually decreased over the period, from about 1.5 in 2002, to about 1.2 in 2010, and to about 0.8 in 2020.

It is the lowest total fertility rate among the OECD countries. The OECD average total fertility rate was about 1.7 in 2000 and about 1.6 in 2020 (Source: OECD.Stat.).

Further, I discuss the effects of childbirth on wages, hours of work, and labor supply depending on the wage distribution and occupation. Based on the monthly wage 1 year before the first childbirth, I classify the full-time workers into those whose monthly wages were less than the 50th percentile of the wage distribution in Figure 2 in Appendix, and those more than the 50th percentile in Figure 3 in Appendix. Figure 2 in Appendix shows that the child penalty for the monthly wage is 54.1% and that of working hours is 14.8% 10 years after the first childbirth. The supply of female labor from women workers whose wages were below the 50th percentile of the distribution shows an immediate drop in the year of childbirth and falls behind 60.4% relative to males in the same group at event time 10 (t = 10). In Figure 3 in Appendix, females whose wages were above the 50th percentile of the distribution have the negative effect of having child on the labor supply, except for wages and hours of work. The negative magnitude of the effects of having a child for females whose wages were above the 50th percentile of the distribution is smaller than that for those whose wages were below the 50th percentile of the distribution. In Figure A4 in Appendix, my event study focuses on full-time workers in professional occupations, and shows that females experience an immediate and significant decline in their labor supply just after giving birth. My findings suggest no negative effects of childbirth on the wages and hours of work for full-time female workers with relatively high wages or those in professional occupations. However, many of these females leave the labor market after they give birth, like those in Figure 9, and Figures A1 and A2 in Appendix.

The findings show that Korean females’ labor market outcomes are significantly and negatively affected by childbirth, while males’ labor market outcomes are not. Bertrand et al. (2021) classified South Korea as a high-sexism group, which means that South Korea is a country characterized by a gender-conservative society.

Bertrand et al. (2021) used 26 developed countries as their sample: Austria, Belgium, Canada, the Czech Republic, Denmark, Finland, France, Germany, Greece, Hong Kong, Hungary, Ireland, Italy, Japan, Korea, Netherlands, Norway, Poland, Portugal, Slovakia, Spain, Sweden, Switzerland, Taiwan, the UK, and the US. They used the gender-related statements from the Integrated Values Survey and the 2020 and 2012 International Social Survey Program to measure conservativeness of gender norms as sexism index. In Appendix Table 4 of Bertrand et al. (2021), Sweden, Denmark, and the US are classified as a low sexism group and Japan as a medium sexism group.