Outsourcing for How Long? Factors and Duration of Employing Foreign Live-in Domestic Helpers

Needs, Resources, and Willingness: A Framework of Domestic Outsourcing

This study conceptualizes employing live-in domestic helpers as an option of domestic outsourcing to reconcile work-family conflicts. Domestic outsourcing refers to replacing domestic labor with the use of a wide range of paid domestic services, including meal preparation, dining out, use of childcare services, commercial laundry services, and the use of paid helpers for housekeeping (Sullivan & Gershuny, 2013). Past studies have suggested a wide range of relevant factors in flexible outsourcing (Baxter et al., 2009; Bittman et al., 1999; De Ruijter, 2004; Kornrich & Roberts, 2018; Pot et al., 2005; Treas & De Ruijter, 2008; Van der Lippe, Frey, & Tsvetkova, 2012; Windebank, 2010). Baxter et.al., (2009) conceptualized these factors of using flexible help into three domains: demand, resources, and attitudes. This study adopts a similar approach and conceptualizes the factors of employing live-in domestic help into three broad categories: need-based, resource-based and willingness-based factors.

Need-based factors refer to those that affect the intensity of work-family conflicts. Significant factors shown in past studies include indicators of high household demand, housework inequality that leads the wife to bear the entire burden, and high demands from work domains. Past studies have consistently shown the significance of child-related factors in domestic outsourcing (Cheung & Lui, 2021). Since the transition to parenthood significantly increases the housework burden, especially for women (Baxter et al., 2008; Craig et al., 2016; Kim & Cheung, 2019), past studies have found that couples with young children are more likely to outsource domestic labor (Cheung, 2014; Craig & Baxter, 2016; Sullivan & Gershuny, 2013). Meanwhile, as the demands from the household domain gradually diminish as the child ages, the expenditure on household services also declines (Brayfield & Hofferth, 1995). Other than child-related factors, past studies have also found that couples with women working long hours are more likely to use paid help (Baxter et al., 2009; De Ruijter, Treas, & Cohen, 2005; Van der Lippe et al., 2012). In addition, professional and managerial occupations are often associated with time- and strain-based work-family conflict. Couples working in professional and managerial positions are more likely to outsource domestic labor (Cohen, 1998; Raz-Yurovich & Marx, 2019). These factors suggest that one may seek paid help in resolving work-family conflicts. Following a similar logic, the availability of non-paid outside help, such as help from parents or other family members, to share the housework burden is a need-based factor because such help can relieve the housework burden from a couple.

Resource-based factors refer to those related to the ability to afford paid help. The most frequently examined resource-based factors include disposable income levels, wealth, and schooling (Baxter et al., 2009; Cheung & Lui, 2017; Kornrich & Roberts, 2018; Van der Lippe, Tijdens, & de Ruijter, 2004; Windebank, 2007). Past studies have consistently shown that families with greater economic resources are more likely to outsource domestic tasks. High-income couples are more likely to outsource than low-income couples (Bittman et al., 1999; Cohen, 1998; Gonalons-Pons, 2015; Gregson & Lowe, 2005; Spitze, 1999). In particular, some studies have shown that women’s economic resources are more closely related to the outsourcing of tasks traditionally defined as feminine, such as housekeeping and meal preparation, than are men’s economic resources (Cheung & Lui, 2017; Cohen, 1998; Sullivan & Gershuny, 2013). Due to the high-threshold nature of inflexible outsourcing, economically disadvantaged families would be very likely to face challenges in paying the wages and associated employment costs of live-in helpers. Therefore, only when the couples’ income level is substantially higher than the helper’s wage and the implicit economic costs are families able to afford the practice. Furthermore, due to high housing prices, the living space of ordinary families is very limited in East Asian cities. Employing a live-in helper implies an even more crowded household and greater invasion of privacy. Wealthy families who stay in large apartments can provide independent rooms for live-in helpers. Hence, household wealth (as indicated by house ownership and apartment size) is also positively related to the likelihood of employing a live-in domestic helper (Oropesa, 1993; Van der Lippe et al., 2012). In addition, resource-based factors are not limited to financial resources. Past studies have shown that women’s education level is positively related to the use of housekeeping services (Van der Lippe et al., 2004).

Willingness-related factors refer to favorable norms and attitudes, and perceptual barriers to the hiring decision. Employing a live-in helper may generate issues of distrust and sexual or parenting jealousy (Constable, 2007; Groves & Lui, 2012; Lan, 2006). A resource-rich couple who are in need may not necessarily hire help because they may not be willing to outsource domestic labor to outsiders (Treas & Lui, 2013). Past studies have shown various perceptual barriers to outsourcing domestic help. De Ruijter and Van der Lippe (2009), for example, found that couples with a low level of generalized trust are less likely to outsource childcare and home maintenance. Similarly, Baxter et.al., (2009) found that negative attitudes toward outsourcing domestic labor were negatively related to the use of paid help. Some couples do not outsource because they assign positive values to household chores and seek to resist the commodification of private life (Windebank, 2010).

Willingness-based factors are not necessarily attitudes. For example, Spitze (1999) found that couples are more likely to use paid domestic help if the practice is commonly adopted in their social circle, suggesting that a positive social norm toward outsourcing is positively related to the use of help. For some couples, especially those without any prior experience or without information about the experience, uncertainty may hinder them from employing a live-in helper. Couples who grew up in families with a live-in domestic helper or are situated in a social network where hiring help is common may have more information about the experience and get help easily in recruiting and training a live-in helper. Thus, these couples may have a higher propensity to employ live-in helpers. In a similar vein, being situated in a social context where hiring help is very common means there will be more social infrastructure in supporting the practice (e.g., more agencies, more governmental, and other channels providing relevant information, and social norms favorable to using paid help); couples might therefore be more willing to use help. In times of economic downturn or uncertainty, however, couples might be less willing to employ live-in helpers.

Employing Live-in Domestic Helpers as an Inflexible Outsourcing Option

The general framework for domestic outsourcing is mainly developed in contexts where live-in helpers are not common such as Australia and the Netherlands. In these contexts, domestic outsourcing mostly involves flexible arrangements. Households with different needs use outsourcing to different extents (Killewald & Gough, 2010). Households usually purchase services separately, and purchase them only for the moment when they need that particular service (Raz-Yurovich, 2014). The thresholds for purchasing these services are usually very low, and there is no commitment to use the service for a fixed period. Employing foreign live-in domestic helpers differs from other domestic outsourcing options in constituting an all-in-one package, outsourcing multiple tasks simultaneously, and in representing an inflexible arrangement with a high threshold and a long-term commitment (Raz-Yurovich, 2014). Households employing foreign live-in helpers in Hong Kong outsource cleaning, cooking, and childcare tasks to helpers (Constable, 2007). Therefore, domestic helpers are also major caregivers to children within these households (Groves & Lui, 2012). Meanwhile, employing a live-in domestic helper in Hong Kong is an inflexible arrangement and has a high threshold. Couples have to spend months completing the entire recruitment process (from interviewing, through visa clearance, and up to the arrival of the helpers) and are committed to a 2-year contract for each recruitment. There are also explicit legal requirements for income level and apartment size.

Despite this inflexible and high-threshold nature, the number of households using live-in helpers is growing rapidly. Employing live-in helpers has become an indispensable part of family life for many households in the region. In Hong Kong, according to the data analyzed in this study, only around 2% of married couples were employing foreign live-in helpers in 1984, but that figure had risen to 19.4% in 2018 (as shown in Figure 1). Despite the rising importance of live-in helpers to families in many societies, there is a lack of empirical studies of the patterns and factors of employing live-in helpers. How well the framework of domestic outsourcing, which was developed based on studies of flexible outsourcing options, can be applied to understand the increasing prevalence of employing live-in helpers, which by nature is inflexible with a high threshold), remains unknown.OPEN IN VIEWER

Duration as a Neglected Aspect in Domestic Outsourcing

Most past studies on the use of paid domestic help rely on cross-sectional data and measure the percentage of the sampled families currently using domestic help. One of the major problems of these cross-sectional studies is their inability to study the time dimension of domestic outsourcing. In the literature, the duration of households outsourcing domestic labor and its factors is mostly unknown. It is unclear whether outsourcing domestic labor to the service economy is mostly a temporary arrangement or a regular practice that lasts for years. Neglecting the time dimension of domestic outsourcing is an important omission in the literature. The duration of outsourcing domestic labor might have profound implications on family dynamics and outcomes. For example, parent–child relations might not be significantly affected by the presence of a live-in helper for only a few months. However, the impact might be significant if the live-in helper was employed for many years. Similarly, the prolonged presence of a live-in helper may also alter the spousal relations and child development. Also, understanding the duration of the practice for the employers’ families may help us understand general social conditions that give rise to the helpers' exploitation and abusive experience. The information about the duration of the practice can also help policymakers predict future labor demands for foreign domestic helpers.

Theoretically, the cross-sectional patterns between the different factors and outsourcing are the combined results of two different events: the differential risk of entrance and the differential risk of exit. The first event is related to the differential risks of starting the practice, while the second event is related to the differential risks of leaving the practice. Take the relationship between family resources and hiring help as an example. Assuming that resource-rich families have a higher tendency to use help but that all families who hire help retain the practice for the same duration, the prevalence of hiring help would appear to be higher for resource-rich families. However, assuming that all families have the same tendency to hire domestic help but that families with fewer resources cannot retain the practice for very long, the prevalence of hiring help would also appear to be higher for resourceful families. In addition, the lack of significant relationships between some factors and hiring help could be a consequence of conflicting effects of these variables on the start and the end of the practice. To use a hypothetical example, if the number of children has a positive effect on starting to use help (due to the extra housework burden), but families with more children are also likely to end the practice sooner (due to greater financial stress), the conflicting effects of the number of children in starting and leaving the practice may result in a misleading lack of significant association in a cross-sectional snapshot. The use of cross-sectional data in past studies prevents us from comparing the relative importance of the two events in creating the social patterns of using help. Knowing the relative importance of differential risks in experiencing the two events and their associated factors of the two events can help us to understand the nature of the social patterns of outsourcing domestic labor. To fill the above gaps in the literature, I examine the roles of need-based, resource-based and willingness-based factors in the transition into the practice and duration of the practice.

Sample and Data

The data for this study were taken from a representative household survey conducted in Hong Kong between November 2018 and July 2019. The survey covered married couples in Hong Kong with wives aged between 18 and 55. A two-stage stratified sample design was adopted, with the records in the frame of quarters first stratified by geographical area and type of quarters. For the second stage, eligible married couples were invited for interviews. One member of the married couple was randomly sampled for the interview. Face-to-face interviews with structured questionnaires were carried out by the interviewers. Of the 2943 valid addresses sampled, 2003 cases successfully completed the interviews, providing a response rate of 68.1%. The refusal rate and non-contact rate were 19.3% and 12.6%.

The questionnaire used in the survey measured a wide range of respondent and household characteristics, including resources, housework arrangements and family relations. Despite the nature of the cross-sectional survey, retrospective life histories of marriage, childbirth, living arrangements, employment, and the use of live-in domestic helpers were also collected. Employing a domestic helper requires a long recruitment process and couples have to complete many complicated procedures. Couples, therefore, have good records and memories of the timing of the start and the end of the practice. Hence, recording the event-history record with a retrospective design is feasible. There were a small number of cases with missing values, mainly for the income and employment variables. A listwise deletion strategy was adopted to handle the cases with missing values. The analytic sample size was 1980 (after 23 cases were excluded from analysis).

Analytical Approach

This study used discrete-time logit models (Allison, 2014) to examine the factors associated with (1) the duration from marriage to starting to employ domestic helpers (Entrance Model), and (2) the duration from starting to employ domestic helpers to the end of the practice (Exit Model). The durations were measured in the number of years. The units of analysis for the discrete-time logit models were person-year observations. For the first set of analyses (of the duration from marriage to starting to hire help), person-year data were constructed from the year of marriage until the year of hiring live-in help (if ever hired a helper) or the year of being interviewed (if never hired a live-in helper). There were 19,203 person-year observations made for the first set of analyses. For the second set of analyses (of the duration from the start of using help to the end of using help), person-year data were constructed from the year of entrance until the year of exiting the practice (if ever ended) or the year of being interviewed (if still hiring). There were 3242 person-year observations made for the second set of analyses. Note that there is a tiny proportion of cases (2%) for those who ever hired live-in help had more than one entrance and exit (Hiring the helper after the first childbirth, and discontinued the practice a few years later, and re-entrance of the practice after the second childbirth). To avoid complications, only the first entrance and exit are examined.

For each set of analyses, two discrete-time logit models were estimated. In the first model, I included time since marriage (in years) and marriage cohorts (a set of dummies) as the only covariates. This model examined the baseline hazard rate of transition for different cohorts. The purpose of the first model is to help understand the changes in the risk of experiencing events across marriage cohorts. The intercept and the coefficients of time since marriage and marriage cohorts were combined to produce the cumulative incidence rates for the transitions for different marriage cohorts. In the second model, I additionally included all sets of explanatory variables (need-based factors, resource-based factors and willingness-based factors) in the analysis to examine their net impacts on the hazard rate.

Variables

Dependent variables. There are two time-varying dependent variables in this study: (1) start of employing live-in help, and (2) end of employing live-in help. The dependent variables were recorded as binary outcomes indicating whether the transition events occurred in a particular year: 0 indicates the transition event did not happen in that person-year observation, while one indicates the transition event did occur in that person-year observation. For the Entrance Model, of 19,203 person-year observations, there were 515 transition events. For the Exit Model, of 3242 person-year observations, there were 142 transition events.

Independent variables. The key independent variables were the same in the two transition models. In both models, I included time since marriage (in years) (Entrance Model) or time since hiring (in years) (Exit Model) to examine whether the hazard rate changed over the years. As an additional analysis, I also tested several polynomial specifications to examine the nonlinearity in the relationship between time and hazards. In both models, the second-order and above polynomial functions were not statistically significant and hence are not reported in the results section. As the sample includes couples married in different periods, I also include marriage cohorts (a series of dummy variables indicating that the couples married in the 1980s, 1990s, 2000s, or 2010s) in the models.

The explanatory variables were divided into three sets: need-based factors, resource-based factors, and willingness-related factors. In each set of factors, some measures are time-invariant, while others are time-varying. The need-based factors include having a child that year (Yes = 1), the number of children, and the age of the youngest child, which are indicative of the extent of the housework burden. The couples’ full-time work statuses and occupational statuses in professional/managerial positions were also included, as these indicate the time availability. As housework demand may change along with age, the ages of the couples were also included in the model. All of the above factors are time-varying measures. Lagged measures (t − 1) were used for the work-related variables to avoid the problem of reverse causality. In addition, co-residence statuses with couples’ parents in the marriage year (Yes = 1) were also examined as these may be an indication of outside help to relieve the housework burden and hence of a lower level of need for outsourcing the tasks. Due to data limitations, this measure was only taken at the time of marriage and is time-invariant.

Resource-based factors include the couples’ years of schooling and combined incomes at marriage (in HK$), the relative share of the couples’ income earned by the wife (as a percentage) and the level of household wealth (in percentiles). The couples’ incomes at marriage were measured in Hong Kong dollars. As the couples married in different periods, the income level was not directly comparable due to inflation; dollar amounts were therefore adjusted to 2019 levels by considering the official historical data of price indices. The household wealth level was calculated as a composite score of apartment size (number of rooms and toilets), homeownership and ownership of vehicles. The score was then ranked in descending order and recoded as a percentile.

The last set of factors is those that are willingness-related and includes couples’ histories of growing up in families with live-in helpers before marriage (0–2, indicating whether none, one or both of them grew up with live-in helpers) and the commonness of using live-in helpers in their social circles. The commonness of using live-in helpers was calculated as a composite score measuring how common it was to hire live-in helpers among couples’ relatives, friends, neighbors, and colleagues (on a scale of 1–5, from 1 = Not common at all to 5 = Very common). These items indicate how familiar the couples were with the experience of living with a helper at home and with social norms favorable to the practice. Familiarity with the experience and with favorable social norms will remove barriers to the hiring practice. The above items are time-invariant measures. In addition, the percentage of households in Hong Kong using live-in helpers at particular years was obtained from the official census data and was also included. The widespread nature of the practice in the territory indicates lower levels of information and transaction costs involved with the hiring practice, and hence implies a favorable social condition for hiring. The unemployment rate (in %) and inflation rate of the years (in %) were also obtained from official data sources and included in the model. A high unemployment rate and inflation rate, while controlling for the couples’ employment statuses and income levels, may indicate economic uncertainty and hence relate to a lower level of willingness to hire and a higher level of willingness to stop hiring. The percentage of households using live-in help, the unemployment rate and inflation rate are time-varying measures.

In general, the distributions of the variables (such as employment, income, education level, and age) in the analytic sample for the entrance model are similar to the married population in the official census data. The distribution of the variables in the analytic sample for the exit model differs from the analytic sample for the entrance model. Families who hired live-in helpers constitute a selected group of more resource-rich, have more housework demands, and are more familiar with the practice. Descriptive statistics of the variables for the analytic samples of the entrance and exit models are included in the Appendix.

Alternative Specifications for Robustness Check

Additional analyses are conducted to check for the robustness of the main findings. First, Cox proportional hazards models were estimated instead of discrete-time logit models. The strength, direction, and statistical significance of the variables between the models all remain the same. Only results from the discrete-time logit models are reported because logistic regression is more familiar to most researchers and more intuitive. For income variables, information for income at marriage and income at the current year are available. Linear interpolation between the two income points was conducted, and income at marriage was replaced by an interpolated income trend (as a time-varying variable) in the additional analysis. Again, the results remain the same with this alternative specification. Results with income at marriage are reported because this specification requires fewer assumptions.