Making an invisible care workforce visible: a survey of domestic workers in three cities in China

Domestic workers are a fast-growing segment of the labor force in China, but their work remains invisible. Using a respondent-driven sampling survey data collected in Beijing, Jinan, and Changsha during 2019 and 2020 (N = 2,20), this study provides socio-demographic and work profiles of domestic workers in these cities. Most domestic workers are rural women with junior-high-school or lower education. They are predominantly informal workers with no labor contract and have no access to labor rights protections and social security. A majority of domestic workers have done overtime work that is often underpaid or unpaid. Live-in domestic workers work excessive hours and earn significantly less per hour than their live-out counterparts. While the hourly wages of child and elder care workers are significantly lower than those of other domestic workers, the hourly pay of elder care workers is the lowest. Despite the working conditions, domestic workers have high levels of job satisfaction probably partly due to their low perceived value of paid domestic work. Policy implications of this study include the formulation of specialized laws to protect domestic workers’ labor rights and regulate family employers’ behaviors, improving informal domestic workers’ access to social security, change the social norms that devalue care work mainly performed by women, and collecting systematic statistics of paid domestic work.


Introduction
China has entered a stage of rapid population aging, low fertility, and small family size. Meanwhile, the bear of care responsibility by the individual family, the high labor force participation of women, and the rapid increase of family incomes have created a large demand for paid domestic services among urban families. These trends, together with large-scale rural-to-urban migration, have given rise to a fastgrowing domestic services industry and a care workforce of mostly women. The revenue produced by China's domestic services industry increased from CNY 203.4 billion in 2014 to CNY 878.2 billion in 2020; and the number of domestic workers is estimated to increase from 12.6 million in 2011 to over 30 million in 2018 (Qianzhan Industry Research Institute, 2021). 1 China still faces a shortage of domestic workers as care needs continue increasing under its current demographic changes (Mo, 2018).
Despite the large size of the domestic workforce, paid domestic labor remains virtually invisible in China. Paid domestic work can be rendered invisible and economically devalued through cultural, legal and spatial mechanisms (Hatton, 2017). Worldwide women are overrepresented in paid domestic work (ILO, 2021) which is considered unskilled and not economically valuable and as an extension of the unpaid care work that is mainly performed by women (England, 1992). Domestic workers are characterized as innately caring and doing what comes naturally (Folbre, 2001). Thus, the cultural mechanism of invisibility makes domestic labor obscure and devalued through hegemonic gender ideologies. Moreover, the legal mechanism of invisibility can make paid domestic labor devalued by excluding it from legal definitions of employment (Hatton, 2017). Domestic work was recognized as a formal occupation by China's Ministry of Labor in 1995. However, under the current legal framework, family employers of domestic workers are considered "consumers" of care services. Therefore, paid domestic labor is legally constructed as informal work that is not monitored and regulated by the state (Liu, 2017). Furthermore, paid domestic labor is rendered invisible and devalued through the spatial mechanism because it is performed in the private home that is constructed as a site in which real work does not take place (Hatton, 2017).
There is little systematic knowledge of paid domestic labor in China. Globally, the domestic workforce is largely invisible because a majority of domestic workers are women who are predominantly employed in an informal sector, and the size and work of which are underestimated in labor force statistics and national income accounts (Beneria, 1992;ILO, 2013ILO, , 2021. Several European countries have made efforts to collect survey data that estimate the size of the paid domestic labor and report on their working conditions and economic impact. Although these surveys provide useful data and sectoral analyses, they are of an ad hoc nature (Abrantes, 2014;Jokela, 2019;Suleman & Figueiredo, 2018). In developing countries, little 1 3 Making an invisible care workforce visible: a survey of domestic… attention has been paid to ensure that paid domestic work is accurately documented in national statistics. Care work is categorized as nurturant-and non-nurturant care. Nurturant care emphasizes emotional and relational interactions, including care for children, the elderly, sick and disabled. Non-nurturant care includes cooking, cleaning, driving, and other household maintenance activities (Duffy, 2005). Paid care work that requires nurturant skill suffer a wage penalty and is therefore more invisible than other types of care work (England et al., 2002). While domestic workers are a heterogeneous group, they are considered as part of a broadly defined "household service workers" in China's Censuses and most national surveys. Aside from a limited understanding gleaned from individual case studies, their profiles as workers remain largely unknown.
To fill in the knowledge gap, we conducted a research project on domestic workers in China. The study is based on a survey of domestic workers in three cities (i.e. Beijing, Jinan, Changsha) conducted during 2019 and 2020. We present sociodemographic and work profiles of domestic workers to shed light on policy formulations and data collection concerning paid domestic labor in China.

Study objects and sites
This study uses data from the "Survey of Domestic Workers' Employment Status and Needs" that was conducted during 2019 and 2020 in Beijing, Jinan, and Changsha by researchers from the School of Sociology at the Beijing Normal University. We defined domestic work as work performed in or for a household or households, within an informal or formal employment relationship, and on an occupational basis following the ILO framework (2021) while taking into consideration Chinese social contexts. Eligible respondents in our survey included domestic workers who performed nurturant care work (i.e. maternal-, child-, and elder-care workers) 2 or non-nurturant work (i.e. housekeepers and cleaners) in or for private household(s) as their main source of earnings. Respondents were eligible regardless of their specific working arrangements (i.e. whether they were live-in or live-out; worked on an hourly, daily, monthly, or other bases; and whether they were employed by a household or through a service provider).
We selected Beijing, Jinan, and Changsha as our survey sites because we expected to find considerable differences in the characteristics of the care workforces in these cities. While Beijing is a first-tier metropolitan city, Jinan and Changsha are secondtier provincial capitals. There are significant differences in the levels of socio-economic development, the demographic profiles, and the labor market among them. Beijing's levels of economic development, the size of the aging population, and overall population size are all among the highest in the country. In 2018, Beijing had a resident population of 21.54 million including 178,000 newborn babies and 3.64 million older adults aged 60 years and over; the average per capita income of urban residents was CNY 68,000 in 2019 (Beijing Municipal Bureau of Statistics, 2019). It is estimated that at least one-third of households in Beijing had demand for domestic services in 2018, particularly for elder and child care (Huayan Industry Research Institute, 2019). Jinan, the capital of Shandong Province in the eastern region of China, had a resident population of 7.46 million in 2018, among which 95,000 were newborn babies and 1.54 million were older adults aged 60 years and over; the average per capita income of urban residents exceeded CNY 50,000 in 2019 (Jinan Provincial Bureau of Statistics, 2019). Changsha, the capital of Hunan Province in the southern region of China, was home to a resident population of 8.39 million with about 104,000 newborn babies and 1.5 million older adults aged 60 years and over in 2019; the average per capita income of urban residents was CNY 55,211 in 2019 (Hunan Provincial Bureau of Statistics, 2020).

Data collection
We conducted the survey using a respondent-driven sampling (RDS) method. RDS has proven to be an important recruitment and analysis tool for sampling populations that have no sampling frames and are socially connected. RDS, unlike a traditional snowball sampling method, can yield a form of probability sample through multiple waves of peer-to-peer recruitment and statistical adjustments for bias that mainly result from differential social network sizes of the recruits and differential recruitment (Heckathorn & Cameron, 2017). Based on its assumptions, RDS can provide unbiased estimates for the structure of a population if certain characteristics of recruits reach a stable status. RDS has been widely used for studies of hard-toreach populations, such as drug users, sex workers, and illegal immigrants. It is also applicable to the study of China's rural-to-urban migrant workers, a population that is large and has an obscure sampling frame (Liu et al., 2015;Qiu et al., 2012). Given that domestic workers are networked socially (Liu, 2017;Xu & Palmer, 2011) and the lack of a sampling frame for a population working in private households, RDS is an appropriate sampling method for our study.
Details about the design and implementation of the RDS for this survey are shown in Table 1. The RDS recruitment process began with 26 to 27 seeds (i.e. the first round of respondents) in each city. The seeds were not randomly selected but they were diverse with respect to work type, working arrangements, and working duration. We started the first round of the survey with a few well-networked seeds. The survey was conducted using face-to-face interviews that averaged about 45 min each. After the interview, each seed was given a predetermined number of coupons that they can use to recruit peer domestic workers in their social networks (i.e. recruits of wave 1). The recruits of wave 1 then completed the same interview process and recruited respondents of wave 2. The interviews and referral chains continued until they failed or reached the maximum wave of recruitment set for this study. Due to time constraint of the field work, the desired sample size is not reached in all the three cities.
To encourage recruitment, we provided dual incentives to those who agreed to participate: cash or a gift equivalent to CNY 40-50 as the primary incentive for participation in the survey; and cash or a gift equivalent to CNY 10-15 as a secondary incentive for each respondent they successfully recruited. Domestic workers' busy and inflexible work schedules posed a challenge for our survey. To reduce travel time for respondents and facilitate recruitment, interviews were conducted at places and times that were convenient for respondents. Interviews frequently took place in public places near respondents' workplaces or residences.
The survey collected information on domestic workers' socio-demographic characteristics, employment status and working conditions, labor and social protections, occupational training, attitudes and perceptions, social support, and health conditions.

RDS sample assessment
The study recruited a total of 2173 respondents; 2120 completed valid interviews and 53 were invalid. We used RDSAT (Volz et al., 2012) to look at the equilibrium status, degree of homophily, differential recruitment activity, and bottlenecks for several characteristics-sex, age, hukou, marital status, education, work type, and working arrangement-in order to check potential biases of the sample. It turned out that distributions of sex, age, hukou, marital status, and education level reached an equilibrium between waves 1 and 5 in the three cities. While the distribution of work type and working arrangements reached an equilibrium between waves 3 and 7 in Beijing and Jinan, these variables did not converge in Changsha. We also identified a certain level of variations in recruitment homophily. A high degree of recruitment homophily was found among maternal care workers in all of the three cities. The variation of recruitment homophily among subgroups with respect to work type and Question for seed about network size The number of persons you know who are qualified for the survey and may join the survey through your recruitment working arrangements was much higher in Changsha than in Beijing and Jinan, but we did not find any obvious bottleneck in recruitment in any of the cities. The results of the RDS sample assessment indicated that the Jinan data were independent of characteristics of the seeds and therefore could be treated as a random sample of the population. The underrepresentation of the maternal care workers in the Beijing sample should not have substantial effects on the comparison of domestic workers by work type. However, it may cause some bias for the estimation of the characteristics of domestic workers in Beijing as a whole. The non-convergence of some key variables (i.e. work type and working arrangements) in the Changsha data suggests possible seed bias.

Weighting and statistical analysis
RDS estimators should be adjusted for differences in network size and recruitment homophily. In the analysis, we used the RDS-VH weight (Volz & Heckathorn, 2008) to adjust for the differences in network size. To adjust for the level of recruitment homophily, we applied the RDS-MOD approach (Selvaraj et al., 2016) to calculate sampling variances for estimates. The RDS sample assessment implied that the adjusted Jinan sample is representative of all domestic workers in Jinan, whereas estimation of domestic workers' characteristics in Beijing and Changsha as a whole based on the adjusted samples of these two cities should be interpreted with caution.
Because the samples in all three cities converged for most of the characteristics examined, seeds and their recruits were all included in the analysis (N = 2120). Given the difference in the labor market of domestic workers in the three surveyed cities, we conducted all analyses and presented the results stratified by cities. Significance of the differences in characteristics of domestic workers among the three cities were estimated using the Chi-square test for categorical variables and the ANOVA test for interval-level variables. We conducted the statistical analysis using Stata 16.1 (StataCorp, 2019) and used its svy package to calculate confidence intervals (CIs) for the estimates.

3
Making an invisible care workforce visible: a survey of domestic… Almost all of the domestic workers in Beijing were inter-provincial migrants, whereas the domestic workers in Changsha were overrepresented by intra-provincial migrants (63.8%) and in Jinan by local residents (83.5%) ( Table 2). The majority of married domestic workers in Beijing (76.5%) and Changsha (57.2%) lived apart from their spouses, compared with 17.1% living apart in Jinan.

Work characteristics and working conditions
Domestic workers who engaged in nurturant maternal-, child-, or elder-care work comprised 73.9% and 64.3% of this part of the labor force in Beijing and Jinan, while the corresponding figure was 47.4% in Changsha (Table 3). Domestic workers in Beijing and Changsha who were overrepresented by rural-to-urban migrants were also the main breadwinners for their families. In Beijing 71.9% and in Changsha 53.7% of domestic workers had contributed more than half of their households' annual income in the previous year, whereas the figure was 43.3% for domestic workers in Jinan. In spite of more than six and a half years of work experience on average in the domestic services industry, domestic workers' job stability was relatively low. Between 29.1% and 47.3% of domestic workers in the three cities changed family employers in the year prior to the survey, and among them a majority had changed employers for multiple times.
There were salient distinctions in the kinds of working arrangements and work time for domestic workers in the three cities and among different groups. As shown in Table 3, while a predominant majority of domestic workers in Beijing (82.8%) and Changsha (62.7%) lived in the employer's household, most domestic workers in Jinan lived out and worked on a daily (70.8%) or hourly basis (18.7%). As a result of the high proportion of live-in workers, domestic workers in Beijing worked an average of 12.4 h a day and 77.4 h a week, which were significantly higher than those of their counterparts in Jinan and Changsha. When we stratified working hours by working arrangements, we found that the average daily and weekly working hours of live-in domestic workers in all three cities were 12.4-13.5 h and 77.4-84 h, respectively. Further analysis (results not shown) indicated that the daily and weekly working hours of live-in workers were significantly longer than those of live-out workers in all cities. It is noted that domestic workers in Jinan had the fewest average daily (8.1 h) and weekly (45.3 h) working hours among the three cities due to the fact that a majority of them worked on a daily or hourly basis.
The overall wage levels of domestic workers in Beijing and Changsha were above the average wage levels of private sector service workers in these two cities, and there were significant differences in wages paid to domestic workers across cities and among groups with different work responsibilities. As shown in Fig. 1  1 3 Making an invisible care workforce visible: a survey of domestic… averages that were higher than the overall averages for urban private sector service workers in both cities. In contrast, the average annual income of domestic workers in Jinan was CNY 32,670 (95% CI 29,558-35,781) which is significantly lower than those of domestic workers in Beijing and Changsha (P < 0.001). The average annual income of domestic workers in Jinan was also lower than that of urban private sector service workers as a whole. 4 The findings may be because domestic workers in Jinan had worked fewer hours overall compared with their counterparts in the other two cities. The stratified analysis indicated significant group differences for domestic workers with different work responsibilities in their annual income in all three cities (P < 0.001). While maternal care workers (CNY 44,682-71,165) earned the highest average annual incomes, elder care workers (CNY 29,156-40,946) had relatively low average annual incomes in all of the three cities (Fig. 1). The average yearly earnings of elder care workers in Beijing, Jinan, and Changsha were 57.5%, 65.3%, and 59.4% of the average annual incomes of maternal care workers, respectively. Due to domestic workers' job instability and the drastic differences in their working hours, the hourly pay provided a better picture of domestic workers' wage levels than annual income. As indicated in Fig. 2, the pattern of significant group differences in the three cities remains when looking at the hourly pay (P < 0.001). Maternal care workers had the highest hourly pay rates, ranging from CNY 22.8-27.8 per hour in the three cities. These rates were higher than the minimum hourly wage standards of CNY 24, 19, and 17 per hour for part-time workers set by municipal governments in Beijing, Jinan, and Changsha, 5 respectively. The average hourly pay for a housekeeper/cleaner in Jinan (CNY 20.1, 95% CI 19.1-21.1) or Changsha (CNY 20.3, 95% CI 18.6-22.1) was also higher than local minimum hourly wage standards for part-time workers. In contrast, the hourly pay for child and elder care workers in all three cities was lower than the local minimum hourly wage standards for part-time workers. Eldercare workers had the lowest average hourly pay-CNY 14.7 (95% CI, 13.9-15.6), 14.5 (95% CI, 13.4-15.6), and 12.4 (95% CI, 11.4-13.4) per hour in Beijing, Jinan, and Changsha, respectively, amounts that were 61.3%, 75.9%, and 72.9% of the local minimum hourly wage standards for part-time workers in those cities.
Further analysis indicated (results not shown) that while the average annual income of live-in domestic workers was significantly higher than that of live-out workers in Jinan and Changsha, the average hourly pay of live-in workers was significantly lower than that of live-out ones in all three cities as a result of their excessive working hours (P < 0.001). The findings suggest that working conditions, in terms of the hourly wage, were worse for live-in workers relative to their live-out counterparts.

Employment relationship and access to social security for domestic workers
Our study showed that domestic workers mostly engaged in informal employment. The finding is similar to that of a recent study of domestic workers in four cities in the Yangtze and Pearl River Delta areas (Worker's Daily, 2019). As shown in Table 4, while 52.9% of domestic workers in Changsha and 33.6% in Beijing were not affiliated with a private placement agency, the corresponding figure was significantly lower in Jinan (19.6%). Very few domestic workers, between 0.4% and 1.6% in the three cities had signed formal labor contracts with private placement agencies and were protected by China's Labor Law. A majority of domestic workers in Beijing (55.6%) and Jinan (69.2%) had signed a service agreement 6 with private placement agencies and/or family employers. While the remaining 30.4% of domestic workers in Jinan and 42.8% in Beijing did not have or did not know whether they had written contracts or not, this figure has reached 63.7% in Changsha.
At the same time, domestic workers in the three cities had limited access to social security benefits. As indicated in Table 4, domestic workers' participation rate in the urban employer-provided insurance was between 1.5% and 4.5% only. An overwhelming majority of domestic workers (between 87.6% and 97%) were enrolled in rural or urban residents' insurance. Between 1.4% and 8.9% of domestic workers in the three survey cities were not covered by any kind of social insurance scheme. The percentage of domestic workers who had no medical insurance coverage differed significantly among the three cities, with 8% of domestic workers uncovered in Beijing, 2.5% in Jinan, and 12.5% in Changsha. About one-fifth of domestic workers in Beijing and Changsha had no old-age pension insurance, while this figure was 14.9% in Jinan. In spite of financial support provided by the Jinan municipal government, only 27.8% of domestic workers in Jinan were covered by the commercial insurance for domestic services 7 ; this coverage rate was nonetheless slightly higher than those in Beijing (25.8%) and Changsha (16.5%).

The labor rights situation for domestic workers
The study showed that domestic workers were vulnerable to labor rights violations, particularly with respect to time worked and overtime compensation. More than one-fifth of domestic workers in Beijing (28%) and Changsha (20.2%) took no day off in the past month, compared with 4.9% in Jinan (Table 5). In Beijing, 20.3% of domestic workers enjoyed no time off during public holidays in the year prior to the survey. This figure was significantly higher in Beijing than it was in Jinan (8.3%) and Changsha (9.7%).
While more than half of domestic workers in Beijing and Changsha had worked overtime on some weekends during the previous year, only 9.3% and 12.3% received compensation higher than the regular rate (see Table 5); 5.4% of domestic workers in Beijing and 6.3% in Changsha received no pay for their overtime work, while 42.5% and 38.3% were underpaid at the regular rate. About two-thirds of domestic workers in Beijing (66.6%) and Changsha (62%) had worked overtime on public holidays during the year prior to the survey. Among them, 26.5% in Beijing and 21.4% in Changsha received compensation higher than the regular rate, whereas about onethird were underpaid at the regular rate, and between 6.5% and 7.9% received no compensation for the overtime work in these two cities. The percentage of domestic workers unpaid and underpaid for overtime in Jinan, though lower than it was in Beijing and Changsha, was nonetheless not low. Moreover, the proportion of domestic workers in Jinan who received compensation higher than the regular rate for the overtime work during weekends (5.8%) and public holidays (14.1%) was even lower than their counterparts in Beijing and Changsha. There were considerable proportions of domestic workers in Beijing (7.8%), Changsha (5.6%), and Jinan (4.4%) who had delayed wage payments during the year prior to the survey (Table 5).

Work-related attitudes and perceptions
Despite of the challenges in their working conditions, labor rights situation, and access to social security, domestic workers had an overall high level of job  Table 6, more than 72.5% of domestic workers reported being satisfied with their working conditions, and even higher proportions of them were satisfied with their relationship with family employers (between 76.5% and 86.4%) or reported an overall satisfaction with this job (between 76.3% and 84.8%). While a predominant majority (between 77.8% and 88.3%) of domestic workers in the three cities agreed that "domestic work can increase income", only a modest proportion of them-between 46.8% and 53.4%-were satisfied with the income from domestic work. The appearing contradiction between domestic workers' working conditions and their overall job satisfaction may be explained by their perceptions towards domestic work. As indicated in Table 6, between 56.7% and 59.9% domestic workers in the three cities agreed with the statement "I can't do other work, domestic work is relatively easy"; more than 80% of domestic workers believed that "domestic work is suitable for them". The findings may reflect that domestic workers had high levels of agreement with social norms that equate care work with women's work and a low perceived value of the work they had performed.
While 15.5% of migrant domestic workers in Beijing agreed that "domestic work is flexible and it allows me to take care of my family at the same time", the corresponding figures were 41.7% in Changsha and 62.3% in Jinan. The results revealed a salient pattern of work-family conflict among migrant domestic workers in Beijing. In contrast, the flexibility of informal domestic work may be an important factor that enables local rural women in Jinan to undertake this work. Paid domestic work allows these rural women in Jinan to reconcile work and family responsibilities but at the same time reinforces gendered division of household labor.

3
Making an invisible care workforce visible: a survey of domestic…

Concluding remarks
Based on the unique data collected in Beijing, Jinan, and Changsha between 2019 and 2020, this study provided socio-demographic and work profiles of domestic workers in these three cities. This segment of the labor force is growing fast, but the work of domestic workers remains invisible. Consistent with existing literature (Wang & Wu, 2016), our study showed that domestic workers in these three cities comprise a socially disadvantaged group -a majority of them are rural women with junior-high-school or lower educational attainment. Similar to the situation in many other countries (ILO, 2021), this study demonstrated that domestic workers were predominantly informal workers who had not signed labor contracts, and had no access to labor rights protections and social security. A considerable proportion of domestic workers enjoyed no time off during weekends and public holidays. A majority of domestic workers had performed overtime work that was often underpaid or unpaid, based on labor standards set by China's Labor Law. Over the past decade, the government has vigorously promoted the development of the domestic services industry. The main policy focuses have been to increase employment, satisfy the needs of urban households, and scale up the domestic services industry (Tong, 2017). Nonetheless, the employment of domestic workers remains largely informal. The informality of the paid domestic labor functions as a legal mechanism of invisibility that excludes most domestic workers from protections included in the Labor Law while failing to monitor and regulate the behaviors of family employers (Liu, 2017).
Under the circumstances, the boundary between daily work and daily life is obscure which renders live-in domestic workers particularly vulnerable through a spatial mechanism of invisibility (Hatton, 2017). Our study revealed that live-in domestic workers worked excessive hours and earned significantly lower hourly pay than did their live-out counterparts. Care work is a public good as the time and effort involved in caring for children, the sick, elderly, and disabled are essential for the well-being of individuals and the functioning of society (Folbre, 2006). Studies in developed countries show that poor working conditions of paid care workers have adverse impacts on both caregivers and the quality of care work (Folbre, 2012). Therefore, it is critically important for the government to formulate specialized laws and regulations to protect the labor rights and interests of domestic workers and to regulate the behavior of family employers to make them conform to the government set labor standards. As part of the government's broad policy initiative to extend social security coverage to informal workers, commitments are also needed to improve domestic workers' access to social security.
Our study demonstrated that the average hourly wages of child and elder care workers were significantly lower than those of domestic workers who do other types of work, and they were lower than the government-set minimum hourly wage standards for part-time workers in all three survey cities. Consistent with previous studies (Dong et al., 2016), the average hourly pay of eldercare workers was the lowest among domestic workers, suggesting that the labor devaluation is particularly salient for eldercare workers who are considered to perform "dirty work" (Wu, 2018).
Despite their working conditions, this study showed that domestic workers in general have high levels of job satisfaction. This may be partly due to their low perceived value of paid domestic labor as reflected by their high levels of acceptance of the gender norms that equate care work with women's work and naturalize skills of domestic workers. Given that gender ideologies functions as an important mechanism of invisibility (Hatton, 2017) for paid domestic work, more efforts should be put into promoting changes in social norms that devalue care work that is mainly performed by women.
While this study provides useful information of paid domestic labor in three Chinese cities, it cannot represent the overall situation of the paid care labor force in China. Care work makes indispensable contributions to the development of human and social capital, assets that are crucial to economic development. There is a consensus that more sophisticated measures of the care labor are needed to account for this invisible work mainly performed by women (Beneria, 1992;Folbre, 2006;Tong et al., 2015). This should be particularly so as national income accounts and labor force statistics has shifted away from the goal of measuring economic growth towards social welfare and human development (Beneria, 1992). Our study therefore provides support for the argument to collect systematic statistics of paid domestic work.