Abstract
Using nationally representative longitudinal data collected in South Korea from 2006 to 2013, this study evaluates the associations between nonstandard employment and various health outcomes with a focus on gender differences. We also examine to what extent family status, that is, marital status and parenthood, conditions these associations for men and women. Our results reveal an important role of selection in that many of the significantly negative associations between nonstandard/self-employment and health outcomes disappear in fixed-effects models when time-invariant unobserved individual heterogeneity is taken into account. Such negative selection appears to be more pronounced for men than for women. We also find a significant association between nonstandard/self-employment and health, which differs by gender and particular health outcomes examined. As for the moderating role of family status, our results show that mothers do benefit from self-employment in terms of lower depressive symptoms.
The expansion of nonstandard employment (e.g., part-time, temporary, and contract work) is a consequence of broad economic and social changes that have been fueled by globalization, economic restructuring, and technological improvements (Kalleberg 2009). This rise in nonstandard work has been the subject of scholarly and public concern since the precarious nature and poor job quality of nonstandard jobs (e.g., low wages and lack of fringe benefits) may have negative consequences for workers and their families (e.g., Houseman and Osawa 2003; Kalleberg, Reskin, and Hudson 2000). Although a growing body of literature examined the impact of nonstandard employment on various outcomes (e.g., Ferber and Waldfogel 1998; Kalleberg et al. 2000), relatively little research has been done on the health implications of nonstandard employment (Benach and Muntaner 2007; Price and Burgard 2008). In particular, our limited understanding of gender differences in the impact of nonstandard work on health is unfortunate considering that health consequences of nonstandard work are theorized to vary by gender (Menéndez et al. 2007). Another limitation in the literature is its lack of attention to the role of family status in the relationship between nonstandard work and health even though different family responsibilities imposed on men and women are offered to explain motivations and reasons for taking nonstandard employment.
Using nationally representative longitudinal data collected in South Korea (Korean Welfare Panel Study, the KWPS hereafter), we aim to fill this gap by evaluating the association between nonstandard employment and health with a focus on gender differences. In light of the very rigid labor market segmentation between standard and nonstandard employment, South Korea is an ideal place for the observation of possible health disparity across different employment types. In addition, a high level of gender inequality within the family and at work indicates that South Korea is the very setting to evaluate hypotheses that the health consequences of nonstandard employment may vary by family circumstances, which have different implications for men and women. Analytically, our study advances current scholarship on nonstandard work and health in that we compare results from two model specifications based on different assumptions about selection, including simple ordinary least squares (OLS) models and fixed-effects models. To foreshadow, we document the very important role of selection, which explains the substantially negative association between nonstandard work and various health outcomes for both men and women. These findings contribute to the literature on health implications of nonstandard work in that most previous research was based on cross-sectional data analyses and, thus, failed to deal with endogeneity effectively (for a review, see Park et al. 2016). The findings of our study further indicate that research evidence, both past and future, needs to be evaluated with close attention to the potential role of selection in the observed relationship between nonstandard work and health outcomes.
Gender, Nonstandard Employment, and Health
Nonstandard employment refers to various employment arrangements in contrast to standard employment, which is characterized as full-time and permanent employment, typically on a fixed schedule, involving a direct relationship between the employer and the employee (Kalleberg 2000). Nonstandard employment is also called contingent, nontraditional, irregular, or precarious work. Examples of nonstandard employment include part-time work, day labor, on-call work, temporary and contract work, and self-employment (Houseman and Osawa 2003; Kalleberg 2000). Nonstandard employment is often referred to as “bad” because many nonstandard employment arrangements pay low wages, lack fringe benefits and union protection, and are consequently unstable (Houseman and Osawa 2003; Kalleberg et al. 2000).
Considering the bad job quality of nonstandard work, it is not surprising that nonstandard workers fare worse than standard workers in terms of health outcomes. According to studies conducted in the United States and Europe, nonstandard work is associated with poor mental health (e.g., depressive symptoms), higher mortality risk, and physical health issues such as coronary heart disease, occupational injuries, and poorer self-rated health (for a detailed review, see Benach et al. 2014; Price and Burgard 2008). Several factors are found to link nonstandard work and bad health outcomes. First, fewer economic resources and income instability may have a negative impact on health (e.g., Ferrie et al. 2008; Kalleberg et al. 2000). Second, nonstandard workers are more likely to be exposed to dangerous working conditions while they are less protected by unions and labor laws (Quinlan 1999; Sverke, Hellgren, and Näswall 2002). Socio-psychological factors associated with nonstandard work may also have negative health implications. For instance, lack of control over their work (and personal lives), job dissatisfaction, and employment instability and strain make nonstandard workers susceptible to constant and/or intense levels of stress. Exposure to stress, prolonged or immediate, is found to have negative health consequences, such as the malfunction of the immune system, which increases the risk of severe illness (e.g., McEwen 2002).
One important characteristic of nonstandard work is that women are disproportionately represented. The feminization of nonstandard work is observed in many developed countries: Women’s share of part-time employment, the most prevalent nonstandard job, ranges from 65 percent (e.g., the United States) to 80 percent (e.g., the United Kingdom and Japan) to 90 percent (e.g., Germany and France) (Blossfeld and Hakim 1997; Kalleberg 2000). Women’s concentration in nonstandard employment is often attributed to a desire for flexibility, which may help mothers balance work and family obligations (Hakim 1997). Research shows that women tend to be drawn to jobs with flexible schedules and often choose part-time work voluntarily due to family responsibilities (Cassier 2003; Price and Burgard 2008). One might assume that, to the extent that women’s nonstandard work eases women’s dual responsibilities of work and family, nonstandard work may be beneficial to women’s health (e.g., Hakim 1995, 1997).
On the other hand, some contend that women’s concentration in nonstandard work reflects structural constraints and a gendered labor market—both of which limit women’s standard full-time employment opportunities (e.g., Walsh 1999; Yu 2002). Indeed, there is evidence that women and other minority groups with few socioeconomic resources tend to be employed in nonstandard jobs (Hipple 2001) and that women are disproportionately represented in nonstandard jobs that are insecure, hazardous, and low-paid (Menéndez et al. 2007). In this regard, women’s overrepresentation in nonstandard work is a reflection of the double play of inequality that women face in both the family and the labor market. In this scenario, nonstandard employment is expected to lead to bad health outcomes for women, probably more than for men (e.g., Benach et al. 2014; Kalleberg et al. 2000).
Nonstandard Employment and Health: The Potential Role of Family Status
Since one’s family circumstances affect one’s motivations for taking nonstandard jobs, the relationship between nonstandard work and health may be conditioned by family status such as marital status and parenthood (Hakim 1997; Menéndez et al. 2007). A woman with young children, for instance, may prefer to work reduced hours or with a flexible schedule even if the job pays low wages and offers no benefits, to secure the time to care for her children while avoiding greater health risks, such as stress and fatigue. On the contrary, for a man with children, switching to nonstandard employment may not be necessarily beneficial to his health if it signals his inability to perform as a provider for the family (Christiansen and Palkovitz 2001).
Considering the role of family status is also important in light of evidence on interrelationships between gender, marriage, and health. In the family literature, it has long been debated whether health benefits of marriage can be applied equally to both genders (e.g., Bernard [1972] 1982; for a review see Waite and Gallagher 2000; Wood, Goesling, and Avellar 2007). While recent evidence suggests that both men and women do appear to reap health benefits from marriage, the magnitudes of these benefits often differ by gender. Furthermore, results often vary depending on the particular health outcome examined: Men tend to benefit from physical health monitoring from a spouse while women show improved mental health outcomes (e.g., depression) (Waite and Gallagher 2000; Wood et al. 2007).
Therefore, an important question arises from prior research: whether nonstandard work is equally beneficial or harmful for both men and women, regardless of their family circumstances (e.g., marital and parenting status). Unfortunately, very few studies have explicitly examined gender and family status in terms of the relationship between nonstandard work and health. One exception is the study by Richard H. Price and Sarah A. Burgard (2008). They find that women with childcare responsibilities who choose part-time work voluntarily have lower depressive symptoms. Evidence from studies on the health effect of nonstandard work schedules, which are correlated with many nonstandard jobs, shows that nonstandard work schedules reduce psychological distress among married parents in one study (Liu et al. 2011) and among mothers in particular in another study (Lozano, Hamplová, and Bourdais 2016). But there is also evidence that nonstandard work schedules increase parenting stress as opposed to standard work schedules (Joshi and Bogen 2007). In yet another study on Japanese women’s health (Lim and Raymo 2016), however, nonstandard employment (e.g., part-time work) is associated with better health compared with full-time, standard employment. The same study further documents that marriage is beneficial to women’s health due in part to married women’s lower labor force participation than never-married women who tend to work in full-time, standard jobs. While suggestive, existing evidence is inconsistent and leaves little room to evaluate the extent to which family status affects the relationship between nonstandard work and various health outcomes.
Nonstandard Employment and Health in South Korea
South Korea is an ideal setting to evaluate gender differences and the moderating role of family status in the association between nonstandard employment and health. First of all, rigid labor market segmentation and discriminatory treatment against nonstandard workers indicate that the negative health implications of nonstandard work, if any, may be pronounced in a society such as South Korea. In South Korea, labor statistics distinguish waged workers, which include standard and nonstandard workers, from nonwaged workers (i.e., self-employed workers). Among waged workers, nonstandard workers include those with temporary work, part-time work, day labor, or on-call work (Ministry of Employment and Labor 2017). 1 The proportion of nonstandard workers has rapidly increased after the 1997 Asian Economic Crisis, which resulted in unprecedented high rates of unemployment (unemployment rates increased to 6.8 percent in 1998 from 2.5 percent in 1997). The Economic Crisis also led to market-oriented reforms and structural adjustments, and forced displaced or laid-off workers into nonstandard jobs and self-employment in the low-skill service domain (Y.-S. Kim 2016). According to statistics from 2016, more than one out of five workers in South Korea were classified to be in self-employment, and one-third of waged workers were employed in some form of nonstandard jobs (Ministry of Employment and Labor 2017). 2 Such levels of nonstandard work and self-employment are among the highest in Organisation for Economic Co-operation and Development (OECD) countries (OECD 2015).
In South Korea, such an increase in nonstandard/self-employment has very negative implications for workers’ well-being given rigid labor market segmentation between standard and nonstandard employment. The working conditions of nonstandard workers are much inferior to those of standard workers: Studies document that nonstandard workers receive lower wages and fewer fringe benefits (e.g., unemployment insurance) (e.g., Sohn 2011). More importantly, most nonstandard workers lack the mobility to move to standard work positions because, in South Korea (and Japan as well), whether one has (non)standard employment reflects one’s labor market status, which is closely related to both current and future labor market trajectories (Y.-S. Kim 2016). This characteristic is in clear contrast to other countries where nonstandard work is often perceived as a transitory status to serve as a bridge to a standard job (Cassier 2003; Y.-S. Kim 2016).
The distinctive gender context in South Korea further suggests that South Korea is a great setting for the evaluation of competing hypotheses regarding gender differences in health implications of nonstandard work. According to recent data, about one-third of male employees in waged employment hold nonstandard jobs (36.7 percent) while more than half of females (54.5 percent) are employed in nonstandard jobs in mid-2010 (Y.-S. Kim 2016). Besides, job quality of nonstandard workers is inferior for women compared with job quality for men. For instance, male nonstandard workers earn 53.1 percent of male standard workers, but female nonstandard workers earn only 35.8 percent (Y.-S. Kim 2016). The concentration of women in nonstandard work and their poor job quality are a reflection of gender discrimination against females in the labor market and of a culture that emphasizes traditional gender roles. In spite of an increase in female education, women’s labor force participation is still being interrupted by marriage and childbirth, and there exist strong normative expectations of women taking on most of the responsibilities in housework and childrearing (e.g., Brinton 2001; Lee and Hirata 2001). In this context, both theoretical arguments about the relationship between nonstandard work and women’s health—that is, a negative one resulting from gender discrimination and bad job quality and a positive (or neutral) one resulting from the flexibility to balance work and family obligations—are very relevant.
It is also worth noting that being self-employed in South Korea—often resulting from restructuring of corporations and limited work opportunities for aged and female workers—might not be a voluntary, entrepreneurial decision as it often is in many western countries (G.-S. Kim and Cho 2009). In fact, a study comparing job values related to self-employment between the United States and South Korea based on the General Social Survey (in the United States) and the Korean General Social Survey (in South Korea) finds that job values such as job authority and interest in work are positively related to the selection of self-employment in the United States (Choi and Lee 2005), but no significant association was found in the South Korean data. This finding indicates that Korean workers do not necessarily enter into self-employment because they think highly of job values associated with self-employment.
It is also important to note that self-employed people are disproportionately working in nonprofessional and labor-intensive jobs even though self-employment can be found in a broad range of occupations. According to the Korean statistics bureau (B.S-Kim 2014), in 2014, the top four categories of self-employment were real estate/leasing (25.1 percent), retail sales (22.7 percent), lodging/restaurant (13.6 percent), and transportation (12.5 percent). A substantial proportion of workers in those industries are employed in low-skilled, manual jobs with poor financial rewards and low social standing (Ban 2012; B.-S. Kim 2014). Although there are some jobs in the self-employment sector with high social standing and income such as finance/insurance, professional/technology, and public health, those in these fields account for only 0.3, 1.9, and 1.3 percent, respectively, among all self-employed people (Ban 2012; B.-S. Kim 2014). Clearly, the average socioeconomic status of self-employed workers is inferior to that of standard workers. 3 All this evidence, along with the theoretical viewpoint that considers self-employment nonstandard (e.g., Houseman and Osawa 2003; Kalleberg 2000; Kalleberg et al. 2000), indicates that it is useful to examine health outcomes of self-employed individuals as a group (e.g., Ferber and Waldfogel 1998).
Several studies have examined health outcomes of South Korean nonstandard workers. Evidence of one study shows that nonstandard work, in contrast to full-time permanent work, is associated with the onset of severe depressive symptoms (measured by the Center for Epidemiological Studies Depression scale [CES-D]) for both male and female heads of households (Jang et al. 2015). Yet another study finds that nonstandard employment is associated with poorer mental health outcomes for women compared with men (I.-H. Kim et al. 2006). In addition, there is evidence that health disparity, measured by self-rated health, between standard employment and nonstandard employment (i.e., temporary/daily work) has declined among male workers but increased among female workers from the mid-1990s to the mid-2000s (I.-H. Kim et al. 2011). As for health implications of self-employment, studies find that self-employed individuals report to have health issues, including physical pains and fatigue (e.g., Y.-S. Kim 2014; H.-G. Kim and Lee 2016). Gender differences among the self-employed are also documented: Self-employed male workers have a higher prevalence of hypertension and diabetes mellitus than office workers (Chang et al. 2011), and female self-employed workers report higher depressive symptoms compared with their male counterparts (Y.-S. Kim and Ok 2013).
The Role of Selection
The characteristics of the labor market in South Korea suggest that it is important to consider the potential role of selection to fully understand the health implications of nonstandard/self-employment. As noted, displaced or laid-off workers from the formal wage economy tend to be sorted into nonstandard/self-employment, which implies that negative, often involuntary, selection may operate among workers in nonstandard/self-employment. There is indeed evidence that lower educated and older workers are concentrated among self-employed workers in comparison with standard workers (Ban 2012; Cho 2013; B.-S. Kim 2014). Such a negative selection is likely more pronounced for men than for women in South Korea where a rigid gender division of labor and a strong norm for male breadwinners exist (Bumpass and Choe 2004a; Menéndez et al. 2007).
All these discussions indicate that compositional characteristics—including both observed (e.g., human capital) and unobserved (e.g., motivations or preferences for work)—need to be taken into account to evaluate whether the association between nonstandard/self-employment and health is a reflection of causal impact of nonstandard work on health or of impact of selection. Considering the issue of selection is particularly important for understanding the moderating role of family circumstances in the relationship between nonstandard/self-employment and health. As discussed in the previous section, people may be selected into different family statuses (e.g., marriage and parenthood) depending on individual characteristics and resources such as income, personality, or health conditions (Waite and Gallagher 2000; Wood et al. 2007).
Unfortunately, the literature on health outcomes of nonstandard/self-employed workers in South Korea is limited in that it often relies on cross-sectional data (e.g., Chang et al. 2011; H.-G. Kim and Lee 2016; I.-H. Kim et al. 2006; Sohn 2011) and/or examines a very limited number of health indicators (e.g., Chin, Lee, and So 2011; Song and Kim 2012; Yoo, Cho, and Khang 2010). This study extends previous studies by examining three health outcomes, both physical and mental, including depressive symptoms, self-esteem, and self-rated health. Furthermore, we broaden the literature by employing statistical analyses in ways that explicitly account for a potential endogeneity because continued reliance on cross-sectional data makes it difficult to determine the causal link between nonstandard work and various health outcomes. To the best of our knowledge, no study has examined whether family status conditions the relationship between nonstandard/self-employment and multiple health indicators with longitudinal data and a statistical approach that deals with endogeneity as we do (see “Method” section for details about our analytical strategy). To foreshadow, we present results from (1) both simple OLS models that control only for observed compositional characteristics and (2) fixed-effects models that control for time-constant, unobserved individual heterogeneity (e.g., motivations for work/family, personality, and unmeasured health conditions). This analytic strategy will help us evaluate if and how nonstandard/self-employment is associated with health while endogeneity is taken into account. If selection, especially unobservable characteristics, contributes to the observed health (dis)advantages of nonstandard/self-employment, then health differences by employment status observed in the simple OLS models should be attenuated or eliminated in fixed-effects models.
Method
Data
Data come from the KWPS, an annual longitudinal study based on a representative sample of 15,251 individuals from 7,072 households at the baseline. In this study, we use data from wave 1 (2006) to wave 8 (2013), that is, the most recent survey available at the time of conducting our analyses. Retention rates have been consistently high: 72.17 percent of households at the baseline survey still participated in interviews in wave 8. 4 We restrict the analytic sample to men and women of working ages (18 to 64) since our main interest is health disparity across different employment types. The prevalence of missing data ranges from 0.01 percent (employment type) to 9.28 percent (self-esteem) in the original sample, but there are very few cases missing for independent and control variables when individuals report health outcomes (dependent variables), that is, 0.46 percent of the sample. We, therefore, apply listwise deletion, and the final analytic sample comprises 31,891 to 31,937 person-year observations for men and 36,829 to 36,868 for women depending on the health outcome examined.
Measures
We measure respondents’ mental and physical health outcomes using three indicators, that is, depressive symptoms, self-esteem, and self-rated health (see appendix for the complete list of items for depressive symptoms and self-esteem). Using multiple health indicators is valuable, considering that physical health and mental health are often correlated (e.g., Ross, Mirowsky, and Goldsteen 1990), and relationships between gender, health, and marriage might differ by the dimension of health considered (e.g., Simon 2002).
Depressive symptoms are measured using 11 items from the CES-D scale. Specifically, respondents are asked to report how often they experience depressive symptoms (e.g., “I felt depressed,” “I felt lonely”), including “zero to one day per week” (1), “two to three days per week” (2), “four to five days per week” (3), and “six to seven days per week” (4). Two positively worded responses (e.g., “I felt that I was just as good as other people”) are reverse coded. Due to missing responses for some of the health indicators, we calculate the mean value of all valid responses (e.g., Lim and Raymo 2016). The level of depression is, thus, a mean response to the 11 items with an alpha reliability of 0.87 and a mean of 1.36 with higher scores indicating greater levels of depressive symptoms.
Self-esteem is measured with Rosenberg’s global self-esteem scale (10 items), which includes responses such as, “On the whole, I am satisfied with myself” and “I take a positive attitude toward myself.” Responses were coded on a 4-point scale, including “strongly agree” (4), “agree somewhat” (3), “disagree somewhat” (2), and “strongly disagree” (1). After reverse-coding five negatively worded items (e.g., “At times, I think I am no good at all” or “I feel I do not have much to be proud of”), the level of self-esteem is also calculated as a mean response to these 10 items with an alpha reliability of 0.80 and a mean of 3.09. Higher scores indicate greater levels of self-esteem.
Self-rated health is a categorical variable that measures respondents’ overall physical health conditions (Idler and Benyamini 1997). This measure is based on the question, “How would you rate your health in general?” with response options, “very poor” (1), “poor” (2), “acceptable” (3), “good” (4), and “very good” (5).
Employment type, an independent variable, is a categorical variable that includes standard employment (reference), nonstandard employment, self-employment, and nonemployment, following prior research (e.g., Houseman and Osawa 2003; Kalleberg 2000) and the classification of the Korean Statistics Bureau. Specifically, we define standard employment as full-time, regular jobs (without expected termination of contract) and nonstandard employment as part-time employment, temporary work, day labor, and other nonregular/short-term contract jobs. Self-employment is included as a separate category as discussed in the previous section. Nonemployment includes individuals who are unemployed or out of the labor force.
As for moderating variables, marital status is a categorical variable that includes married (reference), divorced (including widowed and separated), and never-married. Parenting status is a dichotomous variable, coded as 1 for respondents with at least one child and otherwise as 0. In light of very low fertility in South Korea (e.g., Total Fertility Rate was 1.23 in 2010; OECD 2016), we do not differentiate the number of children. In supplementary analyses, we estimate models by using a categorical variable for parity, but substantive conclusions are similar.
In all models, we include controls that might be related to labor force participation, family status, and health outcomes (e.g., Price and Burgard 2008), such as age (measured linearly), educational attainment (a categorical variable that includes high school degree or less, junior/two-year college, and college degree or more), and income (logged annual household income). We also control for health behaviors (i.e., alcohol consumption) and two measures of relationship satisfaction. Specifically, alcohol consumption is a categorical variable that measures a respondent’s frequency of drinking, broken down into never, 1 time per month, 2 to 4 times per month, 2 to −3 times per week, or more than 4 times per week. In addition, we include family relationship satisfaction and social relationship satisfaction, both of which identify whether a respondent is dissatisfied, neither dissatisfied nor satisfied, or satisfied. All covariates, except for gender, are time-varying measured at interview (see “Analytic Strategy”). 5
Analytic Strategy
As discussed above, people with different characteristics might be selected into certain employment arrangements, which may affect the observed association between nonstandard work and health. For this reason, we present results from (1) OLS regression models that control for observed compositional characteristics (e.g., income and education) and (2) fixed-effects models that control for unobserved individual heterogeneity that are constant over time (e.g., personality and unmeasured health conditions).
More specifically, we begin with summary results for simple OLS models that show the association between nonstandard/self-employment and various health outcomes, net of observed characteristics. 6 Thereafter, we present results from fixed-effects models: The baseline model describes the association between employment type and health outcome while adjusting for various compositional characteristics. Because the fixed-effects models deal with endogeneity and a potential for omitted variable bias, results from the baseline model should be interpreted as relationships between nonstandard/self-employment, net of selection, and either observable or time-constant, unobservable characteristics. The direct comparison of results from this baseline model in fixed-effects models with results from simple OLS models will help us evaluate how much selection accounts for the relationship between nonstandard work and health. In the subsequent models, we evaluate the moderating role of marital status (model 2) and parenthood (model 3) in the association between nonstandard employment and health.
We estimate models separately for men and women since the relationships between employment type and other covariates may differ by gender (e.g., Menéndez et al. 2007; Wood et al. 2007). OLS regression models are used, considering the nature of each health indicator and consistency across different model specifications. 7 In all models, robust standard errors are used to account for the nonindependence of repeated observations from the same individuals because clustered errors could result in a downward bias of the standard estimator for the variance (from the true variance) (Cameron and Miller 2015). 8
Results
Descriptive Statistics
Table 1 presents sample characteristics (means/percentages and standard deviations) for the entire sample by gender.
Sample Characteristics, by Gender.
p < .05. **p < .01. ***p < .001.
About one-third of respondents are not employed while 27.9 percent are working in standard employment, 21.3 percent in nonstandard work (e.g., part-time, temporary, dispatch workers), and 18.2 percent are self-employed. Combined, about 40 percent of our analytic sample from South Korea is holding nonstandard employment or is self-employed. The mean age of respondents in our analytic sample is 42.8: According to the supplementary analyses, those in standard employment are younger (39.5) than those in nonstandard employment (43.2) or self-employment (49.6). These age differences reflect the fact that displaced and/or aged workers are often forced to take nonstandard/self-employment (Cooke and Brown 2015).
Table 1 further reveals some differences in sample characteristics by gender. For instance, men are much more likely to work in standard employment than women. In fact, the proportion of women holding nonstandard jobs is higher than that for men even though women’s labor force participation rates are much lower (56 percent) than men’s (81 percent). These results are consistent with prior research documenting that women tend to be concentrated in nonstandard employment in the South Korean labor market (B.-S. Kim 2014). In addition, self-employment is more prevalent among men compared with women, presumably due to the fact that middle-aged (male) workers are often forced to enter self-employment faced with limited standard work opportunities after being laid off or after early retirement (e.g., Keum 2012).
Gender differences are also found with regard to measures for health outcome. Women report more depressive symptoms, but men have higher self-esteem (p < .001). The proportion of men rating their health good/very good is much higher. In addition, men are more highly educated (college degree or more) and have higher household income than women. However, few differences are observed across gender in marital status, parenthood, and satisfaction in family relationship and social life.
Results from Regression Analyses
Table 2 presents a summary of results from the simple cross-sectional models. All models control for demographic and socioeconomic characteristics that may affect employment type and health outcomes. 9 As noted, robust standard errors are used, but they are not presented due to space limitation (results available on request). We find clear health disadvantages of nonstandard work (e.g., part-time, temporary work) for all health outcomes examined, regardless of gender. One exception is that there are no statistically significant differences in self-rated health among women. In addition, self-employed males fare worse than their counterparts in standard employment in terms of depressive symptoms and self-rated health. It is worth noting that these negative associations are found, net of observed compositional characteristics such as socioeconomic resources and health behaviors. In other words, worse health outcomes of those in nonstandard/self-employment are not explained by their demographic characteristics, socioeconomic resources, social/family life satisfaction, and health behaviors. Not surprisingly, nonworking individuals have even worse health outcomes than standard workers.
Estimated Coefficients from OLS Models.
Note. Robust standard errors were used but not presented due to space limitation. Full tables are available on request. OLS = ordinary least squares.
p < .05. **p < .01. ***p < .001.
As for measures for marital status, divorced individuals in general fare worse than married people in terms of depressive symptoms and self-esteem. Divorced women also rate their physical health poorer compared with their married counterparts. Similarly, never-married singles, both men and women, have lower self-esteem, and single women report to have more depressive symptoms and worse physical health than married women. It is, however, interesting that single men have better self-rated health than married men. Also, parenthood is negatively associated with both indicators of mental health among men.
Tables 3 to 5 provide results from the fixed-effects models. Since fixed-effects models take into account unobserved time-invariant, individual heterogeneity, these results will help us understand the potential role of selection in the negative relationship between nonstandard/self-employment and health outcomes observed in the simple OLS models (Table 2). In particular, baseline models (model 1) in Tables 3 to 5 for each health outcome are comparable with models presented in Table 2.
Estimated Coefficients from Fixed-Effects Models, Depressive Symptoms.
Note. Robust standard errors were used.
p < .05. **p < .01. ***p < .001.
First, Table 3 presents results for depressive symptoms. Results from model 1 show that men’s nonstandard/self-employment and women’s nonstandard work are not related to depressive symptoms. In comparison with results from the simple OLS models (see the model for depressive symptoms presented in Table 2), the nonsignificant coefficients for nonstandard/self-employment for males suggest that men in nonstandard work and self-employment are different from standard workers (e.g., personality), which is associated with increased depressive symptoms. In other words, the higher risk of depressive symptoms disappears once endogeneity (unobserved individual heterogeneity) is taken into account in fixed-effects models. Similarly, the loss of statistical significance for nonstandard work among females in fixed-effects models suggests the potentially negative selection among women taking nonstandard jobs relative to women having standard jobs. Interestingly, women in self-employment still report more depressive symptoms than their counterparts in standard employment even after controlling for time-constant, unobserved individual characteristics. This result indicates that the association between self-employment and women’s depressive symptoms is not explained by compositional characteristics—both observed and unobserved—suggesting that self-employment may have a causal impact on women’s depressive symptoms in South Korea. In addition, nonemployed individuals, regardless of gender, are more likely to be depressed compared with those in standard employment.
At the same time, it is worth noting that no differences are found across marital statuses. Changes for the coefficients for divorced and never-married individuals (the reference is married individuals) between two model specifications (Tables 2 and 3) imply that there is likely a negative selection for unmarried individuals. In fact, a recent study documents that selection (i.e., unobserved individual heterogeneity) helps explain the better health outcomes of married individuals relative to their unmarried counterparts in South Korea (Lim, Raymo, and Jeon 2017).
Model 2 evaluates the moderating role of marital status in the association between employment status and depressive symptoms. The coefficients for the interaction terms between employment and marital statuses (e.g., married, divorced) do not reach statistical significance at conventional levels. This lack of evidence for the moderating role of marital status in nonstandard/self-employment is largely due to the utilization of fixed-effects models. As discussed in model 1, individuals with undesirable characteristics (unobserved) appear to remain single or to get out of marriage. It is plausible that, therefore, there is little room for marital status to condition the relationship between nonstandard/self-employment and depressive symptoms once the role of endogeneity is taken into account in fixed-effects models. Indeed, divorced nonstandard workers, both men and women, are more depressed than married nonstandard workers in the simple OLS models, which supports this possibility (results not shown but available on request).
In addition, nonemployed divorced women have a higher risk of depressive symptoms relative to housewives (i.e., not-employed married women), presumably reflecting their unstable labor market status in the absence of spousal support. On the contrary, the significantly negative interaction term between nonemployment and single status among men suggests that not-employed married men report more depressive symptoms than not-employed single men. It is not surprising that nonemployment is associated with increased depressive symptoms for South Korean men and that this negative association might be more pronounced for married men given the strong emphasis on the man’s role as breadwinner (Bernard 1981; Bumpass and Choe 2004).
According to results from model 3, which adds the interaction term of employment and parenthood, mothers benefit from self-employment in terms of lower depressive symptoms compared with nonmothers in self-employment. Considering that parenthood is associated with higher depressive symptoms for both men and women (model 1), this finding indicates that flexible work schedules of self-employment are more compatible for women who also perform the role of mother (e.g., Y.-S. Kim 2014; Lee and Hirata 2001). In contrast, the relationship between nonstandard work and depressive symptoms does not vary by parenting status. This is consistent with prior research documenting that women with nonstandard jobs (e.g., part-time job at a large company) in South Korea are still subject to a rigid organizational culture such as long work and commute hours and frequent social gatherings, and may not enjoy the flexibility as much as self-employed women do (Brinton 2001; H.-G. Kim and Lee 2016). Nonstandard work or self-employment, however, does not benefit fathers’ mental health, presumably reflecting men’s limited involvement in housework and child care (Tsuya et al. 2005) and bad working conditions of nonstandard and self-employed workers (e.g., very long work hours) (Keum 2012).
Table 4 presents results for self-esteem. According to results from model 1, female nonstandard workers report lower levels of self-esteem than those in full-time, standard employment. This finding is in line with research evidence that nonstandard workers often perceive their work efforts not to be equally recognized or rewarded properly (Sohn 2011; Song and Kim 2012). The finding that nonstandard work is more detrimental to women’s self-esteem than men’s further reflects double discrimination that women face in South Korea (i.e., discrimination against women and nonstandard workers), which implies that women’s concentration in nonstandard employment is often the result of their inability to find standard jobs (e.g., B.-S. Kim 2014). It is interesting that such a negative association between nonstandard work and self-esteem is not found for men. Considering that male nonstandard workers have lower self-esteem than standard workers in the simple OLS models (model 2, Table 2), this result again indicates that male nonstandard workers are negatively selected in terms of characteristics (unobservable) that may lead to lower self-esteem.
Estimated Coefficients from Fixed-Effects Models, Self-Esteem.
Note. Robust standard errors were used.
p < .05. **p < .01. ***p < .001.
It is also worth noting that self-employment is not related to self-esteem, regardless of gender. We speculate that self-employed individuals, outside of a waged economy and corporate culture, both of which rigidly differentiate nonstandard workers from standard workers, are less subject to stressors with potentially negative implications for one’s self-esteem. Indeed, evidence shows that self-employed individuals have more autonomy than those in a waged economy (including both standard and nonstandard workers) (B.-S. Kim 2014). In addition, nonemployed men and women report lower levels of self-esteem relative to those in standard employment. Another interesting finding from model 1 is that marital status is not associated with one’s self-esteem, which holds true for both men and women. In comparison with clear disadvantages of unmarried individuals in terms of self-esteem observed in the simple OLS models (model 2, Table 2), this result implies that individuals with characteristics that are negatively associated with self-esteem (e.g., personality, interpersonal skills) are selected into never getting married or getting divorced.
Next, model 2 examines whether the relationships between employment types and self-esteem vary by marital status. We do not find evidence for the moderating role of marital status: This result is not surprising given the strong role of selection into nonstandard/self-employment and marriage observed in the previous models (e.g., Table 2 and model 1, Table 4). That is, the fact that selection operates in the process of labor force participation and union formation makes it hard for marital status to condition the relationship between nonstandard/self-employment and self-esteem once such selection (or endogeneity) is controlled for. Besides, nonemployed divorced men and nonemployed singles, both men and women, rate self-esteem higher than their married counterparts (who are not working either). It appears that nonemployment may have a more negative effect on married people’s self-esteem compared with the unmarried and that this negative impact is more pronounced for men presumably due to the persistent male-breadwinner norm (Bernard 1981; Bumpass and Choe 2004).
Results from model 3 show that there are no interactions between employment status and parenthood for self-esteem. According to our supplementary analysis (results not shown), there is weak evidence that fathers with nonstandard jobs have lower self-esteem than nonfathers with the same jobs in the simple OLS models (p < .10). This again indicates that once unobserved individual heterogeneity, which may be related to lower self-esteem of fathers with nonstandard jobs, is taken into account in the fixed-effects models, parenthood does not affect the relationship between nonstandard employment and self-esteem among men.
Finally, Table 5 presents results for self-rated health. Results from the baseline model (model 1) show that self-employment is positively associated with women’s self-rated health (reference is standard employment). Results from a comparable model estimated from the OLS models (Table 2) show that male standard workers have better self-rated health than those in nonstandard/self-employment. Therefore, the disappearance of associations between nonstandard work and self-employment for men in fixed-effects models again indicates a negative (health) selection of individuals (with characteristics that may negatively affect self-rated health) into nonstandard/self-employment. However, the still significant association between self-employment and women’s self-rated health indicates that South Korean women do benefit from self-employment, presumably due to its flexibility such as reduced work hours and control over work schedules (e.g., Y.-S. Kim 2014; Lee and Hirata 2001). This finding is interesting in that women’s self-employment is positively associated with depressive symptoms (model 1, Table 3). These results suggest that health implications of self-employment on women may be complex in South Korea, which varies depending on the dimension of health outcome examined. In addition, nonemployed men rate their health poorer than their counterparts with standard employment.
Estimated Coefficients from Fixed-Effects Models, Self-Rated Health.
Note. Robust standard errors were used.
p < .05. **p < .01. ***p < .001.
Model 2 introduces interaction terms for employment type and marital status. In general, the relationships between various employment types and self-rated health do not differ across marital statuses. One exception is that nonemployed single men report better physical health than nonemployed married men. This result may result from demanding work conditions in South Korea (e.g., long work and commuting hours) and the possibility of negative selection of not-employed single men. For instance, single men with bad health conditions may be selected out of the labor force. Consequently, nonemployment is then positively associated with their perceptions of physical health. In model 3, we evaluate whether parenting status conditions the relationships between different employment types and self-rated health. However, none of the interaction terms are statistically significant at the conventional level (at p < .05).
Conclusion and Discussion
Using nationally representative longitudinal data collected in South Korea from 2006 to 2013, this study aims to evaluate the associations between nonstandard employment (both waged and nonwaged) and various health outcomes with a focus on gender differences. Since theoretical explanations for gender differences in nonstandard work emphasize varying family circumstances across gender, we also examine the extent to which family status, that is, marital status and parental status, conditions these associations among men and women. Analytically, we take advantage of longitudinal data updated annually to estimate models based on different assumptions on the role of selection. Specifically, we compare results from the simple OLS models that control for only observed compositional characteristics and fixed-effects models that control for time-invariant unobserved individual heterogeneity to delineate the potential role of selection from the association between nonstandard work and health. This is an important contribution since those in nonstandard/self-employment often differ from standard workers in terms of background characteristics (e.g., demographic and human capital characteristics) and motivations/preferences for work/family, which may in turn affect their employment, family status, and health conditions (Brinton 2001; Hakim 2000; Price and Burgard 2008). Considering that most previous studies employ a cross-sectional approach, our study findings will shed light on the extent to which selection/causality accounts for the association between nonstandard/self-employment and health outcomes.
Our results reveal an important role of selection (see Table 2 for results from the simple OLS models and Tables 3–5 from the fixed-effects models): When endogeneity is not considered (Table 2), nonstandard employment is consistently associated with negative health outcomes, including both mental (i.e., depressive symptoms and low self-esteem) and physical health (i.e., self-rated health). Also, self-employment is associated with a higher rate of depressive symptoms and poor self-rated health for males. These results are consistent with prior research on negative health implications of nonstandard work documented in the United States and Europe (e.g., Menéndez et al. 2007; Price and Burgard 2008), and in South Korea (e.g., I.-H. Kim et al. 2006; Song and Kim 2012). However, many of these significant associations between nonstandard/self-employment and health disappear in fixed-effects models. These results indicate that individuals with different characteristics (which are unobserved such as unmeasured health conditions and personality) are selected into different employment types and that such selection accounts for much of the negative health implications of nonstandard/self-employment estimated in models that do not take endogeneity into account. Changes between the simple OLS models and fixed-effects models further suggest that such negative selection into nonstandard/self-employment appears to be more pronounced for men than for women. This finding might reflect gender inequality and the specific labor market context in South Korea. For instance, a substantial proportion of self-employed males in South Korea are often middle-aged men who were either laid off or were forced to retire early involuntarily. In contrast, many women—especially those with young children—often take nonstandard jobs so as to balance work and family. Given the persistent male-breadwinner norm and the rigid labor market segmentation between standard and nonstandard employment, the selection process probably operates more strongly among men than women in that men with less desirable labor market skills are sorted into nonstandard/self-employment.
At the same time, it is important to recognize that nonstandard/self-employment does lead to worsened health outcomes even after any potential selection effect, on either observable or unobservable characteristics, is controlled for. For instance, we find that, for women, nonstandard work is negatively associated with self-esteem and that self-employment increases depressive symptoms. Consistent with this finding, prior research shows that female nonstandard workers are exposed to inferior work conditions and higher levels of work-related stress (which cause health problems) than male nonstandard workers (e.g., Park et al. 2007; Park et al. 2016 for a review). That said, nonstandard/self-employment may have a negative impact on women’s health beyond factors (both observed and unobserved) that select women into different employment statuses. It is interesting that such negative health implications of nonstandard/self-employment are not found among men in fixed-effects models. The fact that these stark gender differences are observed only in fixed-effects models again implies that South Korean men are more negatively selected into nonstandard/self-employment.
In addition, it is worth noting that health implications of self-employment for women depend on the particular health outcome examined. Our results show that women benefit from self-employment in terms of physical health (self-rated health). As some studies point out, self-employment, which is not affected by the rigid, hierarchical organizational culture that often requires long work/commute hours and socialization after work, probably provides better working conditions conducive to women’s physical health in South Korea (Brinton 2001; Lee and Hirata 2001). But men do not appear to benefit from self-employment, presumably resulting from the fact that self-employment is often an involuntary choice due to limited employment opportunities for full-time standard jobs, especially for older men (e.g., Keum 2012; B.-S. Kim 2014). At the same time, women’s self-employment is, however, associated with higher depressive symptoms (reference is standard employment). This seemingly contrasting result might reflect compositional characteristics of the analytic sample and the moderating role of parenting status in the association between self-employment and depressive symptoms. Our sample includes a higher proportion of nonmothers than mothers (see Table 1), and mothers in self-employment have lower depressive symptoms than childless women in self-employment (model 3, Table 3). If the associations (between self-employment and depressive symptoms) operate at the opposite directions depending on parenting status, which appears to be the case in our study, the observed association at the aggregate level might be misleading (e.g., Long and Freese 2006).
These findings indeed demonstrate the importance of considering potential moderators (e.g., parenthood) in examining health implications of nonstandard work, especially among female workers. As discussed in the “Results” section, individuals with different characteristics are selected into different family statuses as well as employment types. As a consequence, much of the moderating role of family circumstances, that is, marital and parenting status, in the relationship between nonstandard/self-employment observed in the simple OLS models disappears in fixed-effects models. 10 That is, different health implications of nonstandard/self-employment across various marital and parenting statuses observed in the simple OLS models reflect the role of selection and omitted variables, such as personality, motivations/preference for wok and life balance, and unobserved health conditions. Nevertheless, we want to point out that self-employment is associated with better health outcomes for mothers in terms of depressive symptoms. These findings imply that waged nonstandard work, such as part-time employment, does not necessarily provide South Korean mothers with flexibility, one of the hypotheses positing health benefits of nonstandard employment for women with children. In contrast, our study findings suggest that self-employed mothers, who are outside the authoritarian organizational culture, reap health benefits as opposed to those mothers holding waged employment in the formal labor market in South Korea (Brinton 2001; Lee and Hirata 2001). The finding that self-employment is not beneficial to fathers’ mental health also suggests that the purpose of pursuing nonstandard/self-employment might differ for men and women depending on their parenting status (Hakim 1995, 2000), which reflects a rigid gender division of labor and a strong emphasis on mothers’ role in children’s academic success in the South Korean society (Tsuya and Choe 2004).
Our specific findings for South Korea have broad implications for other industrial countries that have witnessed the growth of nonstandard work. Our study accounts for the role of selection into nonstandard/self-employment, which appears to explain a substantial proportion of the negative relationship between nonstandard/self-employment and health outcomes. It indicates that those with undesirable human capital and personal characteristics tend to be excluded from standard employment opportunities where the labor market is rigidly segmented between standard and nonstandard employment in a country like South Korea. In other words, the posited negative health effects of nonstandard work might be less pronounced in societies with a more flexible labor market, due in part to the relatively minimized role of selection into nonstandard employment. At the same time, we would also like to point out that contextual specificity needs to be considered when studying health implications of nonstandard work. Our results, for instance, document gender differences in the association between nonstandard/self-employment and health outcomes: Women’s health is more affected by nonstandard/self-employment than men’s even after various compositional characteristics and endogeneity are taken into account. It indicates that gender roles inside and outside the family as well as in the cultural context will lead to different health outcomes for those in nonstandard/self-employment. Therefore, future studies based on societies with varying socioeconomic and cultural contexts will help us better understand how nonstandard/self-employment affects men’s and women’s health in the same or in different ways across countries.
Although our study improves prior research by considering the role of selection and family status in health implications of nonstandard/self-employment, it does not identify exact mechanisms linking nonstandard/self-employment and health outcomes across gender. For example, nonstandard employment might be beneficial to mothers’ physical or mental health if it helps relieve the burden of combining work and family responsibilities by providing flexible schedules and reduced work hours (e.g., Lee and Hirata 2001). On the other hand, nonstandard employment might harm fathers’ mental health if it leads to lower household income and increased work-related stresses due to harsh working conditions (Keum 2012; Song and Kim 2012). Examining how transitions between standard and nonstandard employment are associated with changes in health would represent an important extension of the present study to more fully understand the causal impact of nonstandard work on health.
In spite of these limitations, our study is among the first few to explicitly examine the role of selection in the association between nonstandard employment and health by comparing results from different model specifications. Our study goes so far as to document how one’s family status conditions the association between nonstandard employment and various health outcomes, which also differ by gender. In light of the rise of nonstandard employment and our limited understanding of health effects of nonstandard work (Menéndez et al. 2007; Price and Burgard 2008), future research needs to expand our study by examining the relationships between various forms of nonstandard employment and health outcomes while taking into consideration potential selection issues and gender differences. In particular, we need more evidence from different countries because the role of contextual specificity—for example, the characteristics of specific labor markets, family norms, and relevant social policies—in shaping the relationship between nonstandard employment and health outcomes still remains largely unexplored.
Footnotes
Appendix
Depression and Self-Esteem Scales.
| CES-D Depression Scale |
|---|
| 1. I did not feel like eating. 2. I felt I was just as good as other people. <R> 3. I felt depressed. 4. I felt that everything I did was an effort. 5. My sleep was restless. 6. I felt lonely. 7. I enjoyed life. <R> 8. I felt that people dislike me. 9. I felt sad. 10. People were unfriendly. 11. I had trouble keeping my mind on what I was doing. |
| Rosenberg’s Global Self-Esteem Scale |
| 1. On the whole, I am satisfied with myself. 2. At times, I think I am no good at all. <R> 3. I am able to do things as well as most other people. 4. I certainly feel useless at times. <R> 5. I feel that I’m a person of worth, at least on an equal plane with others. 6. I wish I could have more respect for myself. <R> 7. I take a positive attitude toward myself. 8. All in all, I am inclined to feel that I am a failure. <R> 9. I feel I do not have much to be proud of. <R> 10. I feel that I have a number of good qualities. |
Note. CES-D = Center for Epidemiological Studies Depression scale. <R> denotes reverse coded answer.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.