A Study on the Effects of Gendered Social Norms on the Tradeoff Between Paid and Unpaid Work in Korea

Abstract

In this study, Korean time-use survey data for coupled households is analyzed to show that unpaid work time is endogenous in its relationship with paid work time because the views of traditional gender roles affect gender disparity in unpaid work time. The data not only includes time allocation between husbands and wives but also their views of traditional gender roles within their households, and this information can represent gendered social norms that can potentially explain the distribution of unpaid work between husbands and wives. The control function model is estimated to identify the tradeoff between unpaid work time and paid work time by solving the endogeneity problem. The results of this study show that wives’ unpaid work is likely to be affected by gendered social norms and that the effect can be larger for those having children. In addition, only in the case of wives, unpaid work time is found to be negatively associated with whether to work full-time, showing that wives’ burden of unpaid work could prevent them from working full time. The results indicate that it is crucial to recognize the need to change gendered social norms to address an asymmetric division of unpaid work between husbands and wives.

Introduction

Women’s unpaid household work is invisible and undervalued but women contribute to GDP through the informal economy. Reducing women’s unpaid work could therefore increase national income as well as the income of women. Korea has been conducting national time-use surveys every 5 years since 1999 and publishing statistics for the household production satellite account. Calculations based on the data estimated that women’s unpaid work amounted to 25.5% of GDP in 2019.

According to the ILO (Addati et al., 2018), women do 3.2 times more unpaid work than men, ranging from 1.7 times in the Americas, to up to 4.7 times in the Arab states. In Asia and the Pacific, women conduct 4.1 times more. Korean men perform only 17.2 percent of the total volume of unpaid work while their Northern European counterparts perform over 40 percent. Meanwhile, the Economist’s annual glass-ceiling index 2022 shows that women had a 15.6 point lower labor force participation rate than men on the OECD average. The gender gap in the participation rate is 18.8 points in Korea, which is not only much larger than the average but also the second largest after Turkey.

In other words, Korean women seem to spend more time on unpaid work and less publicly on paid work compared to global averages. In the same vein, the 2022 glass-ceiling index shows that “four Nordic countries—Sweden, Iceland, Finland and Norway—top the index as the best places for working women” and “Japan and South Korea, where women must still choose between a family or a career, fill the bottom two places” (The Economist, Mar 7th, 2022).

This asymmetric division of unpaid and paid work between men and women in Korea should have been mitigated given that the education level of women has improved and publicly subsidized formal childcare has been expanded during the last decade. For example, according to the 2022 glass-ceiling index, the percentage of GMAT exams taken by women is around the OECD average while the net child-care costs are much lower than the average.

In fact, this gender gap in the labor market could have deleterious effects on the efficiency of the Korean economy. The country’s total fertility rate hit a record low of 0.84 in 2020 and for the first time, the government cited the gender gap in the labor market as one of the causes of the low birth rate in “The 4th Basic Plan on Low Birth Rates in an Aging Society (2021–2025).” Women in their 20s and 30s tend to prioritize their careers over child-rearing compared to the older generation, and men also prefer dual incomes. However, the majority of the care work is still undertaken mostly by women, which can lead to women interrupting their careers after childbirth or delaying their childbirth.1

In Korea, the gender gap in unpaid and paid work is likely to widen further because of gendered social norms which associate men and women with different ideal physical attributes and prescribed behaviors. For example, a gendered social norm, that regards unpaid work as a female prerogative, can make it difficult for working-age females to participate in the labor market. Given that women's roles were confined to domestic spheres in traditional Korean society, it would not be easy to change perceptions of the traditional role of women in parallel with the trajectory of Korea’s economic development. The 2021 Social Survey (Statistics Korea, 2021) clearly demonstrates that men tend to prioritize their work and women tend to prioritize their family although the share of people prioritizing their families over work has increased during the last decade. Compared to 2011, the share of people prioritizing their families over work rose from 11.5 to 48.2%. The share increased from 8.2 to 16.2% for men and from 16.4 to 21.1% for women, which shows that a gender gap in the proportion still remains.

The contribution of this study is its focus on a gendered social norm, that is, “men should work in the labor market and women should work at home”. This study aims to show how the level of approval of the traditional gender role affects the allocation of unpaid work time between husband and wife in Korea. To put it in another way, unpaid work time is assumed to be endogenous in its relationship with paid work time. Hence, the main purpose of this study is to investigate the tradeoff between time spent on unpaid work and time spent on paid work by husband and wife and analyze how the gendered social norm affects this tradeoff in Korea.

Background

Significant gender gaps in labor market outcomes, such as wages, hours of work, and occupational choices, exist in all countries. Altonji and Blank (1999) focus on two factors, differences in human capital accumulation and discrimination, as the main sources of these gender gaps. While these two factors have been studied extensively since that article was published, psychological and socio-psychological factors are also commonly discussed as possible explanations for gender differences in labor market outcomes (Blau & Kahn, 2017). One of the less-studied factors is a social norm about what actions are appropriate for men and women to undertake.

In this regard, Akerlof and Kranton (2000) propose a theoretical model where one’s identity enters the utility function and apply the identity model to both the decision to participate in the labor force and the allocation of unpaid work between spouses. The proposed utility function is:

$$U_j=U_j\left(a_j,,,a_{-j},,,I_j\right).$$

$j$’s utility depends on $j$’s actions, $a_j$, and others’ actions, $a_{-j}$, and $j$’s identity,$I_j$. $I_j$ depends on the social status of $j$’s assigned social category, the extent to which $j$’s characteristics match the ideal of $j$’s assigned category, and the extent to which $j$’s own and others’ actions correspond to the social norms of $j$’s assigned category.

When it comes to gender identity and labor force participation, a gendered social norm prescribing that “men should work in the labor market and women should work at home” could explain the gender gap in labor force participation. If women’s identity is enhanced by unpaid work at home, they will have lower labor force attachment than men. For example, women are traditionally considered the main providers of child care within the household and are more likely to be stay-at-home wives or to work part-time because of negative social attitudes toward working mothers. This can be a source of child penalty, the effect of having children on gender gaps. The child penalty effect remains significant in many studies (Goldin, 2014; Kleven & Landais, 2017; Kleven et al., 2019).

The gender identity model can explain why there is an asymmetric division of unpaid work between husband and wife, which can cause gender gaps in labor force participation. However, theories based on comparative advantage (for example Becker, 1965) do not predict this asymmetry. As husband and wife have the same utility function, that in decreasing paid work time and unpaid work time and increasing quantity of a household public good obtained from their joint work (Van Klaveren et al., 2008), more time will be spent on paid work, and less spent on housework, whether by the husband or the wife. On the other hand, in the gender identity model, the husband loses identity when he does unpaid work and when his wife earns more than half of the household income (Akerlof & Kranton, 2000). Hence, the utilities of husband and wife are equalized when the wife undertakes more unpaid work than her husband.

The relevance of the gender identity model for explaining women’s labor market outcomes has been tested in a number of empirical papers. The notion that the husband should be the breadwinner and the wife be the homemaker (Fortin, 2005, 2008), the men’s view of the gender role (Charles et al., 2009), the traditional gender role divide (Booth & van Ours, 2009), and family values (Gousse et al., 2017), all these factors have been found to relate to women’s labor market outcomes. However, it is not clear exactly through which channels gendered social norms affect women’s labor market participation.

Bertrand et al. (2015) analyzed how the behavioral prescription that “a man should earn more than his wife” affects social and economic outcomes, and found that the gender gap in home production is larger in couples where the wife earns more than her husband. Since unpaid work is considered a dependent variable in this study, the effect of unpaid work, which is easily influenced by gendered social norms, on labor market outcomes does not seem to be addressed. Could gender inequality in unpaid work be a missing link in the analysis of gender gaps in labor market outcomes (Ferrant et al., 2014)?

Based on the above discussion, my research hypothesis is that there exists endogeneity between unpaid work and paid work because gendered social norms affect the tradeoffs for husband and wife, which can imply that the gender identity model is relevant in explaining an asymmetric division of unpaid work between husband and wife.

Data and Empirical Strategy

Data

The empirical data on three key variables of this study—respondents’ paid work time, unpaid work time, and their view of gender roles—is derived from the Korean time use survey data in 2019, which is the most recent version. The 2019 Korean time use survey covers 12,435 households across the country and the respondents are asked to report what they did each 10-min interval of the previous 2 days. For the empirical data, a person’s daily average time on weekdays is used. Also, the survey includes data on the respondents’ socio-economic characteristics and household composition.

Time-use data have a great advantage over conventional household surveys for analyzing unpaid work (Esquivel et al., 2008; Dong & An, 2015). From the Korean time use data, I can get a daily average of each respondent’s time (in minutes) spent on both paid and unpaid work. According to the classification of activities in time use data, paid work time is defined as the amount of time spent on “employment”. Unpaid work time is defined as the amount of time spent on unpaid care work for household members, which includes “caring for a household: preparing food, cleaning, arrangement, shopping, etc.”, “family and household member care: child care, etc.”, and “traveling related to the managing of a household and caring”.

The definition of unpaid work is based on the scope of care activities formulated by the ILO (Addati et al., 2018). Care activities comprise two broad kinds. First, those that consist of direct, face-to-face, personal care activities, such as feeding a baby, nursing a sick partner, helping an older person take a bath, carrying out health check-ups, or teaching young children. Second, those involving indirect care activities, which do not necessarily entail face-to-face personal care, such as cleaning, cooking, doing the laundry, and other household maintenance tasks, that provide the preconditions for personal caregiving.

The Korean time-use data not only includes respondents’ time allocation but also their views of gender roles within their households. The level of approval on the traditional gender role, “men should work in the labor market and women work at home”, is measured on a 4-point ordinal scale (strongly agree, agree, disagree, and strongly disagree). This information can represent gendered social norms and stereotypes that can potentially explain the distribution of responsibilities for care and housework between husbands and wives. Considering the purpose of the study, married couples living together are selected for analysis.

Descriptive Statistics of three key variables are given in Table 1. On average, wives spend more than 60% of total work time on unpaid work while husbands spend only less than 30%. In Korea, women’s employment rate is lower than men’s while part-time employment rate is higher among women.2 Therefore, this seems to indicate a tradeoff between paid and unpaid work.

Table 1.

Descriptive Statistics of Three Key Variables (n = 11,040 people, 5520 households)

	Traditional gender role	# Of households	Husbands			Wives
			Paid work(min)	Unpaid work(min)	% of unpaid work	Paid work(min)	Unpaid work(min)	% of unpaid work
Total		5520	342.5	61.0	26.4	182.7	266.3	65.7
Husbands	Strongly disagree	932	374.3	78.8	24.7	238.3	248.8	55.9
	Disagree	2308	343.0	60.6	26.4	189.9	263.6	64.4
	Agree	1922	332.9	53.5	26.7	155.8	274.8	70.4
	Strongly agree	358	307.5	57.8	28.9	136.0	284.1	73.8
Wives	Strongly disagree	1859	380.5	67.0	22.4	234.6	257.0	57.2
	Disagree	2332	330.7	58.1	27.5	171.5	271.5	67.6
	Agree	1134	309.3	58.8	30.3	129.2	273.9	74.1
	Strongly agree	195	313.6	51.5	28.8	134.2	249.7	74.1
# of children under 10	0	4140	313.2	56.2	29.2	192.5	222.6	62.9
	1	752	427.5	70.5	17.5	154.3	372.2	73.3
	Over 2	628	433.6	81.3	18.5	152.1	428.0	74.9
Full-time working couples		918	445.0	53.8	12.3	385.0	162.4	30.4

The level of approval on the traditional gender role, “men should work in the labor market and women should work at home,” is surveyed on an ordinal scale, such as strongly agree, agree, disagree, and strongly disagree

However, Table 1 shows that this ratio tends to depend on the level of approval on the traditional gender role. Among wives, the higher the level of approval, the higher the share of unpaid work. Also, the wives’ ratio of unpaid work tends to increase as the husbands’ level of approval becomes higher. While wives seem to behave according to the gender norm, husbands appear to do the opposite, which shows that husbands are likely to be more “cooperative” when women ascribe to the gender norm. This can be explained in the identity model, where $I_j$ depends on the social status of $j$’s assigned social category and, to some extent an individual may also choose the category assignment. In terms of gender, the assignment differences between men and women have diminished over time, and prescribed behavior has changed as well. Table 1 shows that the wives’ ratio of agreeing with the traditional gender role, 24.1%, is much lower than the husbands’, 41.3%. Even if a wife indicates a positive attitude toward the traditional gender role, she wouldn’t gain in her identity as much as her husband would gain in his identity because she recognizes the traditional gender role is a controversial issue. Hence, equality of utility is likely to be restored when the husband undertakes more housework than he used to.

Also, the number of children aged under 10 is more likely to affect the wives’ ratio of unpaid work rather than the husbands,’ which could reveal the child penalty in the labor market. Even in the case of full-time working couples, an asymmetric division of unpaid work between husband and wife is observed. The wives’ ratio of unpaid work is more than twice the husbands’ despite the fact that both are working full time.

Empirical Strategy

A key motivation for the research question is to identify the tradeoff between paid work time ($y_1$) and unpaid work time ($y_2$) by solving the endogeneity problem. A linear structural form for paid work time ($y_1$) and a linear reduced form for unpaid work time ($y_2$), which is in turn an endogenous regressor of y₁, can be proposed as below.

$$y_1=\alpha y_2+x_1{\beta }_1+u_1,E(x_1^{\prime }{u}_1)=0,$$

$$y_2=x\gamma +u_2=x_1{\gamma }_1+x_2{\gamma }_2+u_2,E\left(x^{\prime }{u}_2\right)=0$$

All equations are estimated separately for wives and husbands. $x_1$ includes unity and is an n by $K_1$ subvector of ${x=(x}_1,x_2)$ and the rank condition holds if and only if $x_2\ne 0.$ The level of approval on the traditional gender role and the need for child care can be included in covariates $x_2$ and the endogeneity of $y_2$ is due to COR $(u_1,u_2)\equiv \rho \ne 0$.

The control function method solves the problem of endogenous explanatory variables in linear models and Heckman's two-stage least squares can be derived using this approach. As $u_1=\frac{\rho {\sigma }_1}{{\sigma }_2}{u}_2+\epsilon$, where the new error term $\epsilon \sim N\left(0,,,{\sigma }_{\epsilon }^2\right)$ is distributed independently of $u_2$, $y_1$ can be explained by exogenous regressors which are $y_2$, $x_1$, and $\widehat{u_2}$. This way of explicitly accounting for the part of the error term causing endogeneity is called the “control-function (CF) approach” and $\widehat{u_2}$ is a control function for the endogeneity of $y_2$ (Lee, 2010).

$$y_1=\alpha y_2+x_1{\beta }_1+\frac{\rho {\sigma }_1}{{\sigma }_2}\widehat{u_2}+\epsilon ,$$

$$E(\widehat{u_2}\epsilon )=0$$

Compared with the two-stage least squares approach, the control function method produces a heteroskedasticity-robust Hausman test of the null hypothesis ($H_0:\rho =0$), which means that $y_2$ is exogenous (Wooldridge, 2015). This test for no endogeneity ($H_0:\rho =0$) can be done with the t-value for $\widehat{u_2}$ ignoring the correction terms ($\frac{{\sigma }_1}{{\sigma }_2}$). Since using $\widehat{u_2}$ instead of $u_2$ affects the asymptotic distribution of $\epsilon$, the standard errors for the control function estimates are based on bootstrap replications. The linear reduced form for unpaid work time ($y_2$) can be estimated by OLS.

The linear structural form for paid work time ($y_1$) can be applied to the case where both husbands and wives participate in the labor market. Hence, the equation can be estimated by OLS and the purpose is to estimate the effect of unpaid work time ($y_2$) on paid work time ($y_1$), which is the intensive margins of labor supply for husbands and wives.

When it comes to the full-sample analysis including people who are not in the labor force, $y_1$ can be set as a dummy variable for measuring whether or not the respondent is working full-time.3 A significant number of Koreans tend to work part-time because they are unable to find a full-time job and may not be fully protected by the social security net. In this case, the $y_1$ equation is a binary response model, for which a Probit model is an appropriate specification. Compared to the OLS analysis involving working couples only, the purpose of the Probit analysis is to estimate the effects of unpaid work time ($y_2$) on whether to work full-time ($y_1$), which is the extensive margins of full-time labor supply for husbands and wives.

$$y_1=1\left[\alpha ,y_2,+,x_1,{\beta }_1,+,u_1,>,0\right],u_1\sim Normal(0,1)$$

$$P(y_1=1|y_2,x_1)=\mathrm{\Phi }(\alpha y_2+x_1{\beta }_1)$$

where $\mathrm{\Phi }\left(·\right)$ is the standard normal cumulative distribution function.

Explanatory Variables

Based on the empirical strategy, $x_2$ are assumed to affect only unpaid work time ($y_2$). In the empirical data, $x_2$ measures the need for family care and the views on the traditional gender role, “men should work in the labor market and women should work at home”. These are variables that make unpaid work time ($y_2$) endogenous in its relationship with paid work time ($y_1$).

The level of approval on the traditional gender role is measured on an ordinal scale, such as strongly agree, agree, disagree, and strongly disagree. This information can represent gendered social norms and stereotypes that should be addressed to redistribute responsibilities for care and housework between husband and wife while it does not seem to indicate gendered social norms prevailing in society, which can affect paid work. That is why the views on the traditional gender role are assumed to affect paid work only through unpaid work in the empirical model.

Referring to the need for family care, we included ‘# family’, ‘65 older’, ‘Children4 10 and over’, and ‘Children under 10.’ The variable ‘# family’ is the number of family members, ‘65 older’ is a dummy variable indicating whether or not there are any household members over the age of 65, ‘Children 10 and over’ is a dummy variable for having children aged 10 and over, and ‘Children under 10’ is a dummy variable for having children under the age of 10. ‘Children 10 and over’ and ‘Children under 10’ are included to capture the need for child care and their estimates can be interpreted as the child penalty in the labor market.

Because the decision of having children and the experience of raising young children could be impacted by the views on the traditional gender role, the effect of the views on unpaid work time might depend on whether they have children or not. Households having no children are compared with those having children, which are limited to those households with children aged 10 and over5 in order to consider the experience of raising young children. It appears that the two groups are quite different with respect to their views on the traditional gender role (Table a7 in Appendix) and the % of time devoted to unpaid work according to the views (Table a8 in Appendix). Based on the results, not only ‘Children 10 and over’ but also the interaction terms with the views are included in $x_2$, which only affects unpaid work time ($y_2$).

Regarding the level of approval on the traditional gender role, “men should work in the labor market and women should work at home”, the 4-point ordinal scales are collapsed into binary responses, i.e., agree (including “agree” or “strongly agree”) vs. disagree (including “disagree” or “strongly disagree”). The variables ‘FAV_NORM_R’ and ‘FAV_NORM_S’ are defined as shown in Table 2. The level of approval is collapsed to binary responses for two reasons. First, the gender norm was self-reported by respondents who decide whether or not they agreed with the norm and rank the level of approval/disapproval. Unlike the former decision, each respondent, in the latter decision, may interpret the scale differently (Chevalier & Fielding, 2011). However, the empirical data doesn’t include anchoring vignettes linked to self-reported assessments of the norm for improving comparability across individuals. Second, there will be too many dummy variables that are transformed in Table 2 if the 4-point scale is applied, which can make empirical results harder to interpret.

Table 2.

Transformation of the Level of Approval on the Traditional Gender Role

		Respondent
		Husbands		Wives
		Agree	Disagree	Agree	Disagree
Spouse	Agree	FAV_NORM_R = 1	FAV_NORM_R = 0	FAV_NORM_R = 0	FAV_NORM_R = 1
		FAV_NORM_S = 1	FAV_NORM_S = 1	FAV_NORM_S = 0	FAV_NORM_S = 0
	Disagree	FAV_NORM_R = 1	FAV_NORM_R = 0	FAV_NORM_R = 0	FAV_NORM_R = 1
		FAV_NORM_S = 0	FAV_NORM_S = 0	FAV_NORM_S = 1	FAV_NORM_S = 1

‘FAV_NORM_R’ is a dummy variable for the respondent having a favorable view of the traditional gender norm concerning their own unpaid work time. For example, if the respondent is a wife and disagrees or strongly disagrees with the traditional gender role, then ‘FAV_NORM_R’ is 1. Thus, having a favorable view of the norm as a wife means that it decreases her unpaid work time. On the other hand, if the respondent is a husband and agrees or strongly agrees with the traditional gender role, then ‘FAV_NORM_R’ is 1. Thus, having a favorable view of the norm as a husband means that it decreases his unpaid work time. ‘FAV_NORM_S’ is a dummy for the respondent’s spouse having a favorable view of the traditional gender norm concerning the respondent’s unpaid work time. If the respondent is a wife and her husband disagrees or strongly disagrees with the traditional gender role, then ‘FAV_NORM_S’ is 1, as this decreases the wife’s unpaid work time. On the other hand, if the respondent is a husband and his wife agrees or strongly agrees with the traditional gender role, then again, ‘FAV_NORM_S’ is 1, as this should decrease the husband’s unpaid work time. This conversion is required since the effects of the level of approval on unpaid work time could depend on whether the respondent is a wife or husband. It would be hard to interpret the estimates if the level of approval is used as it is.

Information on the variables can be found in Tables 3 and 4. All explanatory variables besides the need for family care and the level of approval on the traditional gender role are also included in the $y_1$ equation. Variables ($x_1$) that are commonly included in both $y_1$ and $y_2$ equations are those that measure individual and household characteristics. An individual variable ‘Age’ is the age of respondent and ‘Junior college or above’ is a dummy variable of graduating from junior college or above. A household variable ‘Rural’ is a dummy for living in a rural area, ‘House size(m²)’ is the size of the house measured in square meters, ‘Own house’ is a dummy variable for owning a house, and ‘Fam. Inc’ measures respondent’s household income in ranges (Korean Won; $1≃KRW 1300). ‘House size (m²)’ and ‘Own house’ are included in $x_1$ as they can represent the amount of household assets as well as the need for housing maintenance.

Table 3.

Descriptive Statistics of the Full Sample (n = 11,040)

		Husbands				Wives
		Mean	SD	Min	Max	Mean	SD	Min	Max
$y_2$	Unpaid work(min)	60.97	84.83	0	735	266.34	171.11	0	970
$y_1$	Whether to work full time	0.48	0.50	0	1	0.25	0.43	0	1
$x_2$	FAV_NORM_R	0.41	0.49	0	1	0.76	0.43	0	1
	FAV_NORM_RxChildren 10 and over	0.15	0.36	0	1	0.29	0.46	0	1
	FAV_NORM_S	0.24	0.43	0	1	0.59	0.49	0	1
	FAV_NORM_Sx Children 10 and over	0.07	0.26	0	1	0.21	0.41	0	1
	Children 10 and over	0.37	0.48	0	1	0.37	0.48	0	1
	Children under 10	0.25	0.43	0	1	0.25	0.43	0	1
	# family	3.04	1.05	2	9	3.04	1.05	2	9
	65 older	0.29	0.46	0	1	0.29	0.46	0	1
$x_1$	Age	54.51	13.72	21	94	51.60	13.11	19	90
	Junior college or above	0.50	0.50	0	1	0.42	0.49	0	1
	Rural	0.23	0.42	0	1	0.23	0.42	0	1
	House size (m2)	87.03	40.32	9	660	87.03	40.32	9	660
	Own house	0.75	0.43	0	1	0.75	0.43	0	1
	Fam. Inc
	2 ~ 3 Million Won	0.15	0.36	0	1	0.15	0.36	0	1
	3 ~ 4 Million Won	0.16	0.37	0	1	0.16	0.37	0	1
	4 ~ 5 Million Won	0.16	0.36	0	1	0.16	0.36	0	1
	5 ~ 6 Million Won	0.11	0.32	0	1	0.11	0.32	0	1
	6 million Won or more	0.24	0.43	0	1	0.24	0.43	0	1

Table 4.

Descriptive Statistics of Working Couples (n = 5316)

		Husbands				Wives
		Mean	SD	Min	Max	Mean	SD	Min	Max
$y_2$	Unpaid work(min)	48.76	69.11	0	735	189.72	130.16	0	820
$y_1$	Paid work(min)	409.52	163.12	0	980	331.02	170.20	0	960
$x_2$	FAV_NORM_R	0.36	0.48	0	1	0.82	0.39	0	1
	FAV_NORM_Rx Children 10 and over	0.15	0.36	0	1	0.36	0.48	0	1
	FAV_NORM_S	0.19	0.39	0	1	0.64	0.48	0	1
	FAV_NORM_Sx Children 10 and over	0.06	0.24	0	1	0.28	0.45	0	1
	Children 10 and over	0.43	0.50	0	1	0.43	0.50	0	1
	Children under 10	0.24	0.43	0	1	0.24	0.43	0	1
	# Family	3.15	1.07	2	8	3.15	1.07	2	8
	65 oLder	0.22	0.41	0	1	0.22	0.41	0	1
$x_1$	Age	52.13	12.24	21	86	49.28	11.68	22	86
	Junior college or above	0.31	0.46	0	1	0.26	0.44	0	1
	Rural	0.26	0.44	0	1	0.26	0.44	0	1
	House size (m2)	85.35	38.16	9	660	85.35	38.16	9	660
	Own house	0.75	0.44	0	1	0.75	0.44	0	1
	Fam. Inc
	2 ~ 3 Million Won	0.11	0.32	0	1	0.11	0.32	0	1
	3 ~ 4 Million Won	0.15	0.36	0	1	0.15	0.36	0	1
	4 ~ 5 Million Won	0.18	0.38	0	1	0.18	0.38	0	1
	5 ~ 6 Million Won	0.15	0.36	0	1	0.15	0.36	0	1
	6 Million Won or more	0.32	0.47	0	1	0.32	0.47	0	1
	Industry
	Manufacturing	0.20	0.40	0	1	0.12	0.32	0	1
	Construction	0.09	0.28	0	1	0.02	0.13	0	1
	Wholesale and retail trade	0.12	0.32	0	1	0.14	0.34	0	1
	Accommodation and food service activities	0.05	0.21	0	1	0.10	0.31	0	1
	Public administration and defense	0.07	0.25	0	1	0.03	0.18	0	1
	Education	0.04	0.21	0	1	0.14	0.34	0	1
	Human health and social work activities	0.03	0.16	0	1	0.15	0.35	0	1
	Occupation
	Managers	0.05	0.23	0	1	0.01	0.11	0	1
	Professionals and related Workers	0.14	0.34	0	1	0.22	0.41	0	1
	Clerks	0.17	0.38	0	1	0.16	0.37	0	1
	Service workers	0.06	0.24	0	1	0.19	0.39	0	1
	Sales workers	0.09	0.28	0	1	0.13	0.34	0	1
	Skilled agricultural, forestry, and fishery workers	0.13	0.34	0	1	0.11	0.31	0	1
	Craft and related trades workers	0.12	0.32	0	1	0.02	0.15	0	1
	Equipment, machine operating, and assembling workers	0.13	0.34	0	1	0.03	0.18	0	1

When analyzing working couples, industry and occupation classifications are added to the explanatory variables. When it comes to industry classification, all industries with 5% or more respondents are included except for ‘Agriculture, forestry and fishing’ which is likely to correlate with ‘Rural.’

Results

OLS Estimates: Unpaid Work Time Equation

Table 5 presents the OLS estimates for the unpaid work time ($y_2$) equation. For the full sample, while husbands’ view, ‘FAV_NORM_R’, has a significant negative effect of − 13.48 min on their unpaid work time, their wives’ view, ‘FAV_NORM_S’, is not significant. The interaction terms with ‘Children 10 and over’ are not significant as well. This means that husbands’ unpaid work time seems to be affected by their own views rather than those of their wives and the effect of their views does not depend on whether or not they have children aged 10 and over.

Table 5.

OLS Estimates of Unpaid Work Time Equation ($y_2$)

	Variables	Full sample (n = 11,040)				Working couples (n = 5316)
		Husbands		Wives		Husbands		Wives
		Estimates	t-Values	Estimates	t-Values	Estimates	t-Values	Estimates	t-Values
	Intercept	68.45***	3.40	190.28***	5.22	112.28***	4.06	231.21***	5.71
$x_2$	FAV_NORM_R	− 13.48***	− 4.18	− 11.07a	− 1.95	− 12.92***	− 3.31	− 5.55	− 0.75
	FAV_NORM_Rx Children 10 and over	2.10	0.47	− 24.39*	− 2.38	− 0.12	− 0.02	6.41	0.53
	FAV_NORM_S	− 4.49	− 1.24	− 18.26***	− 3.40	4.70	1.05	− 12.50a	− 1.95
	FAV_NORM_Sx Children 10 and over	− 2.02	− 0.41	− 19.91*	− 2.27	− 3.39	− 0.54	− 12.25	− 1.26
	Children 10 and over	− 13.24***	− 3.33	23.81*	2.19	− 9.71*	− 2.15	− 4.03	− 0.31
	Children under 10	26.70***	6.36	152.12***	18.52	26.60***	5.75	93.54***	10.96
	# Family	1.50	0.84	26.37***	6.75	2.26	1.08	24.63***	5.65
	65 Older	6.78	1.56	− 18.24**	− 2.64	1.11	0.21	− 27.37***	− 3.49
$x_1$	Age	− 0.16	− 0.21	0.51	0.39	− 2.06a	− 1.88	− 2.95a	− 1.80
	Age2	0.01	0.82	0.00	− 0.16	0.02	1.58	0.04*	2.11
	Junior college or above	10.55***	4.17	17.56***	3.40	10.20**	3.02	16.97**	2.56
	Rural	− 0.96	− 0.35	− 11.32*	− 2.41	3.95	1.16	3.51	0.62
	House size (m2)	0.06a	1.90	0.09a	1.87	0.02	0.46	− 0.06	− 1.05
	Own house	− 3.13	− 1.16	2.26	0.44	− 0.84	− 0.27	− 3.97	− 0.67
	Fam. Inc
	2 ~ 3 Million Won	− 20.40***	− 4.38	− 9.98	− 1.42	− 8.38	− 1.28	− 33.84**	− 3.26
	3 ~ 4 Million Won	− 32.33***	− 6.58	− 34.23***	− 4.49	− 13.92*	− 2.11	− 37.42***	− 3.43
	4 ~ 5 Million Won	− 36.19***	− 7.25	− 56.60***	− 6.97	− 8.85	− 1.29	− 40.74***	− 3.53
	5 ~ 6 Million Won	− 31.20***	− 5.60	− 92.47***	− 10.50	− 2.66	− 0.35	− 64.67***	− 5.33
	6 Million Won or more	− 39.08***	− 7.66	− 101.4***	− 12.34	− 6.47	− 0.91	− 74.81***	− 6.44

The standard errors are robust to heteroskedasticity

Regarding working couples, the coefficients of industry and occupation classifications are omitted here

^ap < 0.1, *p < 0.05, **p < 0.01, ***p < 0.001

Meanwhile, all of the wives’ variables regarding the view, which are ‘FAV_NORM_R’, ‘FAV_NORM_S’, and the two interaction terms with ‘Children 10 and over’, have significant negative effects from − 11.07 to −24.39 min. This indicates that wives’ unpaid work time appears to be affected not only by their own views but also by their husbands’ views and the effects become larger when they have children aged 10 and over.

In the case of working couples, the results for husbands regarding the view are nearly the same as those from the full sample. For wives, only ‘FAV_NORM_S’ has a significant negative effect, which means that wives of working couples are more likely to be affected by their husbands’ views rather than their own views.

In both samples, the presence of young children, ‘Children under 10’, significantly increases unpaid work time for both husbands and wives, but the magnitude of the effect is much greater among wives, 152.12 and 93.54 min, than husbands, 26.7 and 26.6 min for each sample. The presence of young children appears to be the main factor that increases unpaid work time the most for both husbands and wives. Also, ‘65 older’ decreases only wives’ unpaid work time by 18.24 and 27.37 min for each sample, showing that household members over the age of 65 might act as informal caregivers rather than recipients of care. In contrast, ‘# family’ has positive significant effects on wives’ unpaid work time, which means that the greater the number of family members, the more time is spent by wives on unpaid work.

‘Junior college or above’ has significant positive effects across the board, so a higher level of education seems to increase unpaid work time significantly in both samples. This might be because people with higher education tend to invest more time in child education. In the empirical data, there is little difference in average unpaid work time by education level. However, looking at the child care time included in the unpaid work time, for people graduating junior college or above, it is revealed that husbands spend 2.8 times more and women 3.8 times more on average than their lower educated counterparts respectively.

Except for the husbands of working couples, ‘Fam. Inc’s have significant negative effects on all other occasions. For wives, the size of the effect increases as the income bracket increases. In other words, a higher level of household income reduces unpaid work time especially for wives regardless of their employment status. Given that income is a commonly used indicator of family economic well-being (Xiao, 2013), various services can be purchased in the market if the household income is sufficient. However, the opposite case is also conceivable. Let us suppose that an unemployed mother and her husband worry about their family’s economic well-being. Rao (2020) asks: “how do college-educated, heterosexual, married mothers experience involuntary unemployment?” and finds that “the experience of job loss is tempered for mothers as they derive a culturally valued identity from motherhood which also anchors their lives” (p. 299).

PROBIT/OLS Estimates: Whether to Work Full-Time/Paid Work Time Equation

Table 6 presents the estimates for $y_1$ equation: $y_1$ is whether to work full-time for the full sample and the amount of paid work time for the working couples, respectively. First of all, the Probit estimates for whether to work full-time can be interpreted as the effects of unpaid work time and other variables on the extensive margins of full-time labor supply. The t statistic on the control function ‘$\widehat{u_2}$’ is significant only for wives, which rejects the null that wives’ unpaid work time ($y_2$) is exogenous. To rephrase it, wives’ unpaid work time seems to be endogenous in its relationship with whether to work full time because their unpaid work time is affected by $x_2$ as can be seen in Table 5. Also, ‘Unpaid work’ of wives has a small but significant negative effect on whether they work full-time.

Table 6.

Estimates of Whether to Work Full-Time/Paid Work Time Equation ($y_1$)

	Full sample (n = 11,040) $y_1$: whether to work full-time (PROBIT)				Working couples (n = 5316) $y_1$: paid work(min) (OLS)
	Husbands		Wives		Husbands		Wives
Variables ($x_1$)	Estimates	t-Values	Estimates	t-Values	Estimates	t-Values	Estimates	t-Values
Intercept	0.431	0.908	− 0.212	− 0.496	287.7***	5.123	246.6***	5.513
Unpaid work(min)	− 0.002	− 1.150	− 0.004***	− 11.070	− 0.143	− 0.731	− 0.487***	− 9.397
$\widehat{u_2}$	− 0.001	− 0.570	− 0.001***	− 3.003	− 0.861***	− 4.257	− 0.376***	− 6.584
Age	0.000	0.016	0.021	1.276	7.942***	4.163	6.557***	4.249
Age2	0.000**	− 2.610	− 0.001**	− 3.200	− 0.101***	− 5.641	− 0.081***	− 5.285
Junior college or above	0.373***	7.225	0.173**	3.200	− 16.949*	− 2.461	− 10.008	− 1.451
Rural	− 0.185***	− 3.633	0.045	0.764	− 7.904	− 1.117	15.340*	2.486
House size (m2)	− 0.001**	− 2.315	− 0.001*	− 2.130	− 0.059	− 0.674	− 0.012	− 0.180
Own house	− 0.053	− 1.017	− 0.002	− 0.028	− 0.942	− 0.141	16.001**	2.587
Fam. Inc
2 ~ 3 Million won	0.517***	5.218	0.212a	1.800	15.482	1.107	17.342	1.420
3 ~ 4 Million won	0.790***	7.405	0.492***	4.425	25.111a	1.694	23.911a	1.934
4 ~ 5 Million won	0.915***	8.034	0.800***	6.976	15.672	1.036	27.473*	2.149
5 ~ 6 Million won	1.009***	8.795	1.031***	8.502	31.391*	2.090	47.547***	3.612
6 Million won or more	1.104***	9.537	1.075***	9.473	15.405	1.039	42.959**	3.216

The standard errors for the CF estimates are based on 1000 bootstrap replications

Regarding working couples, the coefficients of industry and occupation classifications are omitted here

^ap < 0.1, *p < 0.05, **p < 0.01, ***p < 0.001

As for the OLS estimates presenting the effects on the intensive margins of labor supply, the control function ‘$\widehat{u_2}$’ is significant both for husbands and wives. However, ‘Unpaid work’ has a significant effect only on wives’ paid work time. In a nutshell, only in the case of wives, unpaid work time is negatively associated with paid work time after controlling for the endogenous predictor and the same is also true for the decision on whether to work full time.

Among other variables, ‘Age’ seems to increase paid work time of the working couples while the higher education level, ‘Junior college or above’, increases the chance of working full time but decreases paid work time of husbands from the working couples. Household income, ‘Fam. Inc.’, tends to increase both the chance of working full time and the amount of paid work time, but the effects are more significant in the former equation for the full sample. Considering Table 5 showing that the income brackets have more significant negative effects on unpaid work time of the full sample, it seems that various services related to unpaid work can be purchased in the market as long as family economic well-being does not change much, which leads to a higher chance of working full time.

Conclusions

This study finds that women’s unpaid work is more likely to be affected by views on the traditional gender roles within their households and the effect can be larger for those having children. Hence, this gendered social norm seems to affect wives’ tradeoff between unpaid and paid work time in Korea. This finding suggests that the gender identity model should be relevant for explaining the asymmetric division of unpaid and paid work between husbands and wives.

When it comes to unpaid work time, husbands seem to be more affected by their own views on traditional gender roles rather than those of their wives. Wives, meanwhile, are more likely to behave according to their spouses’ views, rather than their own. This raises the possibility that women could be more susceptible to gendered social norms. As a matter of fact, such a situation has been observed during the COVID-19 pandemic and related closures of daycare centers and schools. Jessen et al. (2021) finds in Germany that there has been a significant increase in the number of couples with the mother solely undertaking care work. Further, Alon et al. (2022) shows that the global recession triggered by the pandemic had a disproportionate impact on women’s employment, demonstrating larger employment declines among women with micro survey data gathered around six countries: the United States, Canada, Germany, the Netherlands, Spain, and the United Kingdom. These results establish that were it not for the essential care infrastructure, women’s employment could be prone to be swayed by gendered social norms even in western countries.

In addition, only in the case of wives, unpaid work time is found to be negatively associated with the chance of working full-time, showing that wives’ burden of unpaid work could prevent them from working full time. When the samples are restricted to working couples, there also exists a negative correlation between unpaid work time and paid work time. Parental leave is an employee benefit available in this matter, and it is supposed to incentivize labor market attachment for women before and after childbirth. However, this benefit can make the situation worse as long as the traditional gender role affects the allocation of unpaid work within the household. This is because male employees could fear they might be stigmatized by their employers and be disadvantaged at work by taking parental leave, which can make women more likely to take parental leave and spend more time on unpaid work than men.

The results presented in this paper show that it is critical to recognize the need to change gendered social norms to address an asymmetric division of unpaid work between husbands and wives. In the framework of the gender identity model in the Background section, policies may change favorable attitudes towards the traditional gender role into a more gender-equal one. This may make the asymmetric division of unpaid work between men and women more loathsome to household members, leading to lower values of their identity ($I_j$) and a deviation from the prescriptions of the traditional gender role. For example, government programs to improve work-life balance, such as parental leaves and flexible work arrangements, need to ensure that the programs are not used only by women. Also, we need to incentivize men to use the programs because there still exist gendered social norms that penalize men who choose to take parental leaves or flexible work arrangements.

To address this issue, the Korean government has implemented policies encouraging gender-equal use of parental leave over the past 10 years. Not only have they raised the income replacement rate during parental leave and but also offered financial incentives for men taking parental leave. The percentage of men on parental leaves has increased from 2% in 2010 to 21.2% in 2019. However, male parental leave is still not considered to be as common as female parental leave. Also, the duration of leave depends on gender. From September 2017 to August 2018, 52% of female parental leave users took leaves for the entire year and only 15% of leaves were less than 3 months. However, only 25% of male parental leave users took the entire year and 41% of leaves were less than 3 months (Park et al., 2020).

This study analyzes data consisting only of married couples living together, and it does not consider the fact that the decision of whom to marry is also endogenous and varies with the views on gender roles. Also, the child penalty could be detrimental to wives’ unpaid work time if society and schools continue to demand more from mothers than from fathers when it comes to child care. The empirical data do not have the relevant variables capturing gendered social norms prevailing in society and schools and therefore the study has limitations in this respect.

In future research, it would be interesting to test whether the gender gaps in the labor market are due to gendered social norms of society and education system that emphasizes the role of mothers in child-rearing. The comparison between different countries regarding gendered social norms and the structure of society and education may be helpful in this regard.

Appendix

(See Tables 7 and 8).

Table 7.

Views of the Traditional Gender Role (unit: %)

Traditional gender role		Full sample (n = 5520)	No children 10 and over (n = 3494)	Having children 10 and over (n = 2026)
Husbands	Strongly disagree	16.88	16.89	16.88
	Disagree	41.81	42.22	41.12
	Agree or strongly agree	41.30	40.90	42.00
Wives	Strongly disagree	33.68	31.34	37.71
	Disagree	42.25	42.10	42.50
	Agree or strongly agree	24.08	26.56	19.79

The level of approval on the traditional gender role, “men should work in the labor market and women should work at home,” is surveyed on an ordinal scale, such as strongly agree, agree, disagree, and strongly disagree

Pearson's chi-squared for the hypothesis that the view of the gender role and whether having children 10 and over are independent: for wives, chi2(2) = 39.8925 (Pr = 0.000) and for husbands: chi2(2) = 0.7502 (Pr = 0.687)

Table 8.

The % of Time Devoted to Unpaid Work According to the Views of the Traditional Gender Role

		Full sample (n = 5520)		No children 10 and over (n = 3494)		Having children 10 and over (n = 2026)
Traditional gender role		Husband	Wife	Husband	Wife	Husband	Wife
Total		26.4	65.7	32.2	68.7	16.3	60.4
Husbands	Strongly disagree	24.7	55.9	28.1	60.7	18.9	47.5
	Disagree	26.4	64.4	32.1	68.1	16.4	58.0
	Agree or strongly agree	27	70.9	34.1	72.7	15.1	68.0
Wives	Strongly disagree	22.4	57.2	26.5	61.2	16.4	51.4
	Disagree	27.5	67.6	33.6	70.3	16.9	63.0
	Agree or strongly agree	30.1	74.1	36.8	75.0	14.6	72.1

Full-time working couples		(n = 918)		(n = 486)		(n = 432)
		12.3	30.4	14.0	32.8	10.4	27.8

Funding

No funding was received for conducting this study.

Data Availability

The data that support the findings of this study are publicly available from microdata integrated service of Statistics Korea (refer to https://mdis.kostat.go.kr/eng/pageLink.do?link=mdisService).

Declarations

Conflict of interest

The author has no competing interests to declare that are relevant to the content of this article.

Ethical Approval

This study is a secondary data analysis that does not involve human participants or animals.

Consent for participants

Since my study involves only secondary analysis of anonymized data, participant consent was not required.

Consent for publication

Publication concent is not applicable since my study involves only secondary analysis of anonymized data.