Outsourcing Household Production: Foreign Domestic Workers and Native Labor Supply in Hong Kong Patricia Cortésy

Jessica Y. Panz

August, 2011

Abstract This paper explores how the availability of a¤ordable live-in help provided by foreign domestic workers (FDWs) in Hong-Kong a¤ected native women’s labor supply and welfare. First, we exploit di¤erences in the FDW program between Hong-Kong and Taiwan. Second, we utilize cross-sectional variation in the cost of a FDW to estimate a model of labor force participation and FDW hire. FDWs increased the participation of mothers with a young child (relative to older children) by 10-14 percentage points and have generated a monthly consumer surplus of US$130-200. By reducing childcare costs through immigration, this is a market-based alternative to childcare subsidies.

We are grateful to David Autor, Marianne Bertrand, David Card, Matthew Gentzkow, Chris Hansen, Divya Mathur, Jesse Shapiro, Wing Suen, and seminar participants at the University of Hong Kong, Booth School of Business, SOLE, Princeton, Atlanta Fed, UC Berkeley, Boston College, Boston University SGM, Dartmouth, NBER Summer Institute and the 3rd Asia Joint Workshop in Economics for numerous helpful comments and suggestions. We are also grateful to the Hong Kong Census and Statistics Department for providing the data and their invaluable assistance. y School of Management, Boston University. Email: [email protected] z National University of Singapore. Email: [email protected]

1

1

Introduction

In the past decade, there has been a surge in the number of low-skilled female workers from developing countries such as the Philippines, Indonesia, Thailand, and Sri Lanka migrating to newly industrialized countries as domestic helpers. These women enter the host countries under explicit programs that grant temporary visas permitting them to work as private household workers but restrict them to this sector. These movements generate large labor flows for both the source and the host countries. For example, each year, close to 100,000 Filipinas migrate to work as domestic helpers and caregivers. In Singapore, by 2000, there were approximately 100,000 migrant domestic helpers in the workforce, amounting to one foreign maid in eight households (Yeoh et al, 1999). In Hong Kong, the proportion of households hiring at least one foreign domestic worker (FDW) increased from less than 2% in 1986 to close to 8% in 2006. Among households with young children, more than one in three hired at least one FDW. Foreign domestic workers are also very common in Bahrain, Kuwait and Saudi Arabia.1 Temporary domestic worker programs have also been present in the West for some time, albeit at a much smaller scale. Since the mid 1980s, the US and UK have Au Pair programs that allow temporary domestic workers, usually students, to work in households.2 Canada has a Foreign Livein Caregiver Program and Israel has a special visa program for foreign caregivers.3 Notably, even though the US does not have a formal foreign domestic helper program, it has been estimated that 35% of female illegal immigrants report that their first job was working in a private household (Cortes, 2004). Furthermore, low-skilled immigrants in the US are also regarded as providing much more flexible household services at lower prices than those provided by native workers and companies.4 In the light of the “nannygate” controversies in the Clinton administration in the early 1990s, formal guest worker programs have also been discussed as part of immigration reform in the United States.5 1

For example, FDWs comprise 19.9% of the labor force in Kuwait in 1995, 10% of the labor force in Bahrain in 2001 and 8.9% of the labor force in Saudi Arabia (Kremer and Watt, 2008). Note that in the Gulf states, social norms restrict female labor force participation. Therefore, the time freed up in the household by FDWs may result in increased leisure rather than labor market participation. 2 In the US, students between the ages of 18 to 26 are allowed to enter the country as an Au Pair for one year under a J1 visa. The IRS estimates that there are around 12,000 Au Pairs in the US in any given year. In 2000, the UK admitted 12,900 people on Au Pair visas (Anderson, 2001). 3 These programs have remained relatively small due to various restrictions placed on the number and types of caregivers that can enter these countries, the kinds of work that Au Pairs or caregivers are expected to perform and the relatively complicating procedures involved in hiring caregivers under the present schemes. 4 A particular incident that stands out is the “nannygate” controversy in the early 1990s when Clinton administration nominees Zoe Baird and Kimba Wood were withdrawn over allegations that they hired illegal domestic help. Baird had supposedly placed an advertisement in three local newspapers seeking a legal nanny but had not received a single response (Crittenden, 2001). 5 For example, in 2006, Senator Arlen Specter introduced a bill to create a Guest Program, where workers would

2

The economic implications of the temporary migration of private household workers can differ substantially from that of conventional low-skilled migration. As these temporary domestic helpers generally substitute for household production, they potentially influence the labor supply and fertility decisions of women, particularly the middle and highly skilled (Kremer and Watt, 2008, Cortes and Tessada, 2011, and Farre et al., 2011). Moreover, foreign domestic workers are not permitted to work in other occupations, thereby limiting their effect on the labor market outcomes of similarly skilled natives.6 This paper exploits a policy change in the late 1970s that enabled the systematic importation of foreign domestic workers into the Hong Kong labor market to investigate the effects of the availability of affordable and flexible full-time, live-in domestic help on native Hong Kong women’s labor supply decisions and the welfare consequences of this policy. This is a unique setting to analyze the labor market effects of FDW programs thus providing implications for debates surrounding such temporary worker policies in other countries. This study also adds a different dimension to the growing literature that studies the link between childcare costs and maternal labor supply. Most of the recent work in this area assess the effects of government policies that reduce the cost of childcare through childcare subsidies (e.g. Baker et al., 2008, Lefbvre and Merrigan, 2008), admission rules to public schools (Gelbach, 2002) and the introduction of public kindergartens (Cascio, 2009).7 The FDW program in Hong Kong allows us to consider the effects of an alternative, market-based intervention that reduces the price of childcare through the increased availability of domestic helpers. This is an important difference worth emphasizing as, unlike typical interventions that reduce childcare costs through subsidies, the FDW program is based on a market mechanism. By allowing FDWs to enter the labor market, the government does not incur direct costs of providing such services.8 Moreover, native women choose, based on the prevailing market prices, whether or not to purchase domestic services. To our knowledge, this paper is the first empirical study that attempts to causally establish the labor supply effects of this alternative market-based approach to reducing childcare costs through the systematic importation of foreign domestic helpers.9 not have the right to become permanent residents or citizens. 6 Another economic implication, suggested in Kremer and Watt (2008) is that by allowing high-skilled native women to increase market labor supply, this type of immigration increases the wages of low-skilled natives and provides a fiscal benefit by correcting tax distortions toward home production. 7 Earlier work exploring this question use variation in household expenditures in day care to estimate the price of child care (for example, Blau and Robins, 1988, Connelly, 1992; for a summary see Blau and Currie, 2003). These earlier studies face identification problems as the price of childcare tends to be endogenous to the employment decision. These studies find a very large range of elasticities of labor supply, stretching from 0 to -1.26 (Blau, 2003). 8 For example, the childcare subsidy program initiated in Quebec in 1997 to offer subsidized child care ($5 per child per day) for all children under 4 years of age cost the government approximately 1.6 billion dollars in 2006 (for approximately 200,000 day-care places) (Lefbvre and Merrigan, 2008). 9 Suen (1993) and Chan (2006) provide some evidence that hiring a live-in domestic worker is associated with a

3

Furthermore, day-care subsidies only reflect a fraction of the total cost that married women face in child-rearing and household production. Most childcare centers do not admit children below the age of three and are open only during limited hours. Live-in foreign domestic helpers, on the other hand, not only provide essentially all-day childcare, but also substitute for other household tasks such as cooking and cleaning. In some respects, one can regard FDWs as providing one of the most complete forms of outsourcing household production. Therefore, this is a unique setting to consider the responsiveness of maternal labor supply to the availability of an affordable and flexible alternative to their time in the household and how much this option is valued. One might plausibly expect the FDW policy to have a larger female labor supply response as compared to that of government mandated childcare subsidies. Our empirical strategy is based on two complementary approaches - the macro approach exploits differences in the availability and relative cost of hiring an FDW in Hong Kong and Taiwan over time and within-country variation in the demand for childcare services. The micro approach uses cross-sectional variation at the individual level in the cost of hiring a FDW to calibrate a structural model of female labor supply and the decision to hire a domestic helper. Given that both approaches have their limitations and that neither constitutes an ideal natural experiment, comparison of the estimates from the two approaches provides a test of the robustness and validity of our results. The macro approach utilizes variation in the availability of FDWs in Hong Kong and Taiwan over time as well as differences in the demand for household production across mothers with older versus younger children. This triple-difference procedure compares the growth in employment rates of mothers with a younger child (youngest child aged 0 to 5) to mothers with an older child (youngest child aged 6 to 17) in Hong Kong and Taiwan over the period from 1978 to 2006. This approach allows us to separately identify the impact of the availability of foreign domestic workers on female labor supply from effects that might arise from jurisdiction-specific labor market shocks that may differentially affect female employment trends in Hong Kong and Taiwan (captured by the comparison of mothers with children of different ages) and unobserved differences in the demand for mothers of older versus younger children (captured by the cross-country comparison).10 Analyzing trends in economic and demographic outcomes as well as comparing the structure of childcare markets in both countries, we provide evidence that suggests that Taiwan is a reasonable control group. Our triple-difference estimates indicate that, on the aggregate, the foreign worker program in Hong Kong is associated with a 8 to 12 percentage point increase in employment of females with a young child, compared to females with a relatively older child from 1978 to 2006. higher likelihood of female labor force participation in Hong Kong, but neither study addresses causality concerns. 10 While the focus of this paper is on the effects of FDW on maternal labor supply primarily through its the effect on childcare costs, we acknowledge that that FDWs may also be hired to take care of elderly parents. We consider some implications of this on female labor supply decisions but find little evidence that labor supply decisions were affected for women with elderly parents due to the availability of FDWs.

4

Consistent with the view that natives with a higher opportunity cost of time are more likely to purchase such domestic services and supply more labor, we find that this increase is almost entirely driven by the increase in employment rates of middle and highly skilled females. The micro approach utilizes pooled cross-sectional data from the 2001 and 2006 Hong Kong population census to estimate a model of female labor supply, where the decision to participate in the labor force and the decision to hire a FDW are modeled jointly. One advantage of the structural model is that because the model is derived from utility maximization, the estimates can be interpreted as structural determinants of the demand for outsourcing household production. In particular, we can estimate the degree of complementarity between the two decisions and thus infer the extent to which foreign domestic helpers substitute for native women’s time spent in household production. Finally, we can also use the estimated parameters to calculate welfare effects of the availability of FDWs and simulate counterfactual labor supply decisions in the absence of the program. We use a multinomial probit model to study how women choose between the four possible labor force participation-FDW states. To separately identify the degree of complementarity between the two decisions from correlation in unobservables, we propose using the number of rooms in a house as an instrument. The identification assumption here is that the number of rooms affects the utility derived from hiring a FDW but does not directly affect the utility from participating in the labor force. This exclusion restriction is motivated by the fact that most Hong Kong households are relatively space constrained and that conditional on household wealth, the number of rooms should be uncorrelated to a woman’s unobserved propensity to work. A natural concern with this instrument is that unobserved preferences for work may lead a woman to simultaneously choose more rooms and decide to work. While we cannot provide conclusive evidence that the number of rooms are “as good as randomly assigned”, a number of placebo tests on the reduced form of the effect of the number of rooms on the labor force participation decisions for groups of women with very low demand for domestic help such as unmarried women, married women with no children and low-income households provide some assurance that the instrument is not merely spuriously picking up the effects of unobserved preferences for work, household wealth or income effects. We also consider a sample of households who reside in government subsidized sale flats, where the particular nature of the housing allocation scheme is such that households living in these flats face restrictions on the choice of their flat and cannot move out of their existing apartments freely. Finally, as suggestive evidence that the validity of the instrument is not compromised by households moving into houses with a larger number of rooms when they they wish to hire a domestic helper, we study whether households that have moved in the last five years are more likely to hire a FDW. We also present results restricting the sample to households that have not moved in the last five years. We find evidence of strong complementarity between the labor force participation and the decision 5

to hire a FDW: reductions in the relative wage of FDWs significantly increase the probability that a woman decides to join the labor force. This complementarity is especially strong for mothers of very young children implying a significant degree of substitution between the mother’s time and the domestic worker’s time in caring for the child. Our welfare estimates indicate that mothers of very young children, women with high education level and women with high unearned income have benefited most from the availability of FDWs. The average monthly consumer surplus for the whole population of mothers aged 25 to 54 is between 473 and 728 HKD (62 to 94 US$), and between 980 to 1500 HKD (130 to 200 US$) for mothers of young children. The average consumer surplus for mothers with a high education level is more than double that for mothers with a medium education level, and more than ten times larger than that for low educated mothers. To check that our magnitudes are reasonable, we compare the differential willingness to pay between two women identical in all observable and unobservable characteristics but in the age of the youngest child to the difference between the minimum wage of a FDW and the wage of a native unskilled worker, and they are roughly similar. To compare the estimates from our micro and macro approaches, we use the structural estimates to simulate the optimal labor supply decisions of women assuming that they faced the 1981 relative cost of hiring a FDW instead of the 2001 relative cost. For ease of comparison with the macro triple difference estimates, we perform the simulation separately for women with a younger or older child. The simulated micro difference-in-difference estimate of the effect of the FDW program on female labor supply is between 12 to 13 percentage points. While each approach has its limitations, the similarity between the macro and micro estimates, despite the use of different sources of variation, suggests that our estimates of the effect of the FDW program on female labor supply are robust and reliable. While this paper studies the FDW program in the Hong Kong context, the implications extend to other countries. Firstly, this paper highlights the potential benefits of such temporary migratory flows that extend beyond the traditional focus on adverse wage implications of conventional low-skilled immigration. Secondly, our findings are applicable to other East Asian countries such as Singapore, Taiwan, Korea and Japan. As these countries face relatively similar cultural, demographic and economic conditions, this study highlights the potential role of FDW policies in easing household constraints and encouraging women to enter the labor force. Furthermore, to the extent that the social norms of Hong Kong natives with regard to work are similar to western norms (Podmore and Chaney, 1974) and women in Hong Kong experience similar work-family tradeoffs as women in Western countries, the findings can be applied to recent policy debates over foreign domestic worker policies in other developed countries. Nevertheless, a full consideration of the costs and benefits of temporary domestic helper programs will have to address important ethical and

6

political economy issues that are likely to be country-specific and beyond the scope of this paper.11 These findings also have important implications for understanding the sources of persistent gender differences in employment and for policies that seek to encourage female participation in the labor force. That we find large female labor supply responses as a result of the decrease in relative costs of hiring a foreign domestic helper suggests that at least part of the gender difference in labor market outcomes can be attributed to the lack of affordable and flexible childcare. This finding is in line with previous studies that have demonstrated that maternal labor supply is sensitive to the availability and price of childcare. Our estimates imply an elasticity of the labor force participation of mothers of young children (0 to 5) to the cost of hiring a FDW of -0.74, which is about twice as large as estimates far from well identified elasticities for the US and Canada, which gravitate around -0.35 (Baker et al, 2008). As mentioned previously, this larger elasticity is possibly due to the substantially more flexible household services offered by FDWs as compared to childcare centers. Finally, our results also suggest that an immigration policy that permits temporary foreign domestic workers can have important policy implications for encouraging skilled women to enter the labor market and to bridge the gender gap. The rest of the paper proceeds as follows. The next section describes the foreign domestic worker program in Hong Kong and the data. The time-series cross-country analysis is presented in Section 3. Section 4 develops and estimates the structural model using cross-sectional data and Section 5 concludes.

2

Data and Background

2.1

Data Description

Hong Kong Data. The data for Hong Kong is from the 1976 to 2006 Population Census and ByCensus and the 1985 to 2006 General Household Survey (GHS). The Hong Kong population census and by-census are conducted every five years while the GHS is conducted quarterly.12 Both the 11

For example, many see the limited rights accorded to these workers (they cannot become citizens) as a fundamental violation of American values. Moreover, their working conditions (having to live with their employers) might be conducive to abuse. Finally, past experience with large-scale temporary worker programs in the US such as the Bracero Program that imported large numbers of Mexican agricultural workers in the post-war years has demonstrated that without strong enforcement, temporary worker programs could translate into greater illegal immigration, thus imposing real costs onto society. 12 The 1976, 1981, 1986 and 1991 Census data are 1% samples while the 1996, 2001 and 2006 Census data are 5% samples. The GHS data from 1985-1992 are from the first quarter, while the 1993-2006 data are from all four quarters. The Hong Kong Census and GHS data were accessed through a confidentiality agreement with the Hong Kong Census and Statistics Department.

7

Census and the GHS provide a range of demographic and labor force information on all members of an enumerated household. Live-in domestic helpers are defined as females who are present in the household at the time of enumeration and report that their relationship to the household head is that of a live-in domestic helper, chauffeur or gardener. In the Census data, we are further able to classify whether the live-in domestic workers are foreign or local based on their reported place of birth. Foreign domestic workers are defined as those who are born outside Hong Kong, mainland China or Macao. Taiwan Data. Our source of Taiwanese data is the 1978-2006 Manpower Utilization Survey (MUS).13 This is a household-level survey that provides labor force information for a representative sample of Taiwanese individuals over the age of 15. The sample covers the civilian population in Taiwan and excludes foreign workers who do not have citizenship. For the cross-country macro approach, we combine the 1976-1981 Hong Kong Census, 1985-2006 Hong Kong GHS and the 1978-2006 Taiwan Manpower Utilization Survey. For the micro approach, we use data from the 5% sample of the 2001 and 2006 Hong Kong Population Census and By-Census. We are not able to use earlier Censuses or the GHS as these datasets do not have information on the number of rooms in the household, which is needed to identify the structural model. Our main variable of interest is female labor force participation. Individuals are defined as participating in the labor force if they are employed, or unemployed and actively seeking work. It is worth pointing out that in both the Hong Kong and Taiwan data, unpaid workers in family enterprises are also classified as employed. They are distinct from home-makers who are classified as outside the labor force. The wage data is obtained from answers to the question “What are your monthly earnings in your main job?”. Full-time employees are defined as those who report working 35 plus hours in the previous week. For both estimation strategies, the native sample is restricted to married females between the ages 25 to 54 who have at least one child aged 0 to 17. This includes all resident females (whether or not they were born in Hong Kong or Taiwan, respectively) but does not include females who are live-in domestic helpers.

2.2

Foreign Domestic Helper Policy in Hong Kong

The foreign domestic helper program was first introduced in Hong Kong in the mid-1970s. In 1974, the Hong Kong government opened a legal immigration channel that permitted domestic workers from other Asian countries to work in Hong Kong. Prior to this, only unskilled workers from Mainland China were allowed to enter the country. By 1976, the government had established a Foreign Domestic Workers Service Section within the Department of Labour to control and administer matters relating to foreign domestic workers in Hong Kong (Oishi, 2005). At the same 13

The data can be obtained from the Survey Research Data Archive (http://srda.sinica.edu.tw/).

8

time, concurrent developments in the Philippines ensured a ready supply of domestic helpers. The Labor Code of 1974 initiated by President Marcos marked the beginning of Philippines’ formal labor migration program. Coupled with the demand for domestic help among highly skilled natives in Hong Kong, these developments led to a rapid rise in the number of FDWs in Hong Kong starting in the late 1970s and early 1980s. Compared to other receiving countries, Hong Kong has a relatively liberal policy toward these foreign workers.14 The government does not impose a quota on the number of foreign domestic helpers in Hong Kong and employers are free to hire these workers so long as they fulfill the requisite conditions set out in the standard contract. The main restrictions are that the FDW has to work and reside in the employer’s residence and that households have to meet an income criteria in order to hire a foreign maid. In 2008, this was set at the median household income (HKD$15,000) or the equivalent in assets. These workers are entitled to a minimum wage and are protected under the Employment Ordinance and the Standard Contract for the Employment of a Foreign Domestic Helper. Figure 1 shows the evolution of the minimum monthly wage for FDWs (there is no minimum wage for natives) and the average monthly wages for native married women who were full time employees (defined as those working 35-plus hours per week).15 As observed, the relative price of FDWs decreased monotonically until 2001. Since then it has stayed relatively stable, with a slight increase from 2000 to 2005. Figure 2 shows the evolution of the relative wage of native married women to the minimum wage for FDWs separately by the education level of native women. Low education is defined as having at most primary education, medium education as having more than a primary education but less than a college degree, and high education as having a college degree or a graduate degree. This classification applies to both Hong Kong and Taiwan. To provide a sense of how rapidly this program has expanded, Figure 3 presents the share of native women with a FDW by education level and the presence of children of different ages. We consider three types of women: those with youngest child aged 0 to 5, 6 to 17 and those with no children. Several observations are worth mentioning. First, although women at the top education level are most likely to hire FDWs, the share hiring at least one FDW has stayed relatively constant since the mid 1990s. Interestingly, women with a medium level of education had the most continuous 14

Women from mainland China are not eligible to enter Hong Kong as FDWs. This is due to administrative difficulties in monitoring as mainland Chinese are indistinguishable from locals and concerns that Hong Kong residents will bring in their family members from the Mainland on the pretext of hiring them as domestic helpers (Chiu, 2004). 15 The minimum wage is binding. For example, based on the 2001 Hong Kong Census, 45 percent of households with a FDW paid exactly the minimum wage (HK$3670) and the average was only slightly higher, at HK$3757. Historical information on the minimum wage for domestic helpers is from the Labour Bureau (http://www.legco.gov.hk/yr0203/english/panels/mp/papers/mp0328cb2-1515-1e.pdf). The minimum wage remained at 3670 HK$ from 1999 to 2003. In 2003, the minimum wage was reduced to 3270 HK$ but the government instituted a 400 HK$ Employees’ Retraining Levy thereby essentially maintaining the minimum wage at 3670 HK$. In 2005, the total minimum wage (inclusive of the Retraining Levy) was raised to 3720 HK$.

9

growth in demand for FDWs over time. For unskilled women the share hiring FDWs has increased slightly, but continues to be significantly smaller than that of the other two groups. Second, the demand for domestic help comes almost exclusively from households with children. This is true even among highly skilled women. This might be explained by a number of factors such as the cost in terms of privacy loss of having a non-family member living in the house and the high cost of space in Hong Kong. A couple of statistics show the amount of substitution of household work that takes place in households that decide to hire a FDW. In the time-use supplement of the 2001 General Household Survey, 35 percent of FDWs report doing 100 percent of the household work, and at least 70 percent report doing more than 70 percent of the household work. By law, FDWs are only allowed one free day a week - this explains why 80 percent of FDWs report working more than 50 hours a week. In the 1970s, almost all of the domestic workers were mainland Chinese who worked for upper class families. By 1981, foreigners, almost all Filipinas, accounted for 34% of all live-in domestic help. By the early 1990s, foreigners accounted for 96% of all live-in domestic help, with Filipinas alone accounting for 88% of all live-in domestic helpers. The market share of Filipinas started eroding toward the end of the 1990s as lower-cost Indonesian maids entered the domestic helper market. By 2006, there were roughly equal numbers of Filipinas and Indonesians working in Hong Kong as domestic helpers. Interestingly, FDWs tend not to be drawn from the lower tail of the education distribution of their home countries. Filipinas, in particular, are very educated; 20% of them have completed a college degree, compared to 12% of Hong Kongers and a mere 3% of Indonesians. Most Filipinas also speak English. Indonesians, in contrast, compensate for their lack of English speaking abilities by learning the local language, with close to 90% of Indonesians reporting an ability to speak Cantonese. In sum, the FDW program in Hong Kong has provided native women with a relatively inexpensive and reliable source of housekeeping and childcare services.

3

Macro Approach: Hong Kong versus Taiwan

As shown in Figure 1, there has been substantial time-series variation in the relative cost of hiring a domestic worker in Hong Kong over the past couple of decades. To estimate the effect of the FDW program on female labor supply, the ideal quasi-experiment would involve comparing female employment rates in regions that introduced a FDW program to regions that did not, assuming that regions exogenously decide whether or not to implement such a scheme. As Hong Kong is a relatively small country and the FDW policy was implemented at a national level, we cannot exploit geographic variation within Hong Kong. Looking outside Hong Kong, however, suggests that we

10

can use Taiwan as a possible control group given the close proximity as well as the economic and cultural similarity of the two countries.16 The identification strategy that underlies the macro approach exploits differences in the ease of engaging a foreign domestic helper in Hong Kong and Taiwan. The official foreign domestic helper scheme in Taiwan began in 1992 when the Council of Labor Affairs (CLA) granted work permits to domestic caretakers. When the program was first introduced, foreign domestic workers could only be employed to take care of the severely ill or disabled. Subsequently, a limited number of quotas were released for the employment of domestic helpers to care for children under the age of 12 or elderly members above the age of 70 (Lan, 2003). Currently, there are two main programs through which foreign nationals can work as domestic helpers in Taiwanese households - the foreign domestic helper scheme and the foreign caretaker scheme. In recent years, the program has been scaled down even further and only permits special applications for foreign investors and families requiring special child or elderly care. The bulk of foreign domestic workers to Taiwan have since entered under the foreign caretaker scheme. This scheme, however, requires applicants to demonstrate that the person under their care has a medical condition that requires 24 hours care17 . This is in sharp contrast to the program in Hong Kong, where household income is the only eligibility requirement. As a result of these restrictions, the magnitude and scope of the FDW program in Taiwan are far smaller than that of Hong Kong’s. In 2001, FDWs comprised approximately 1.1% of the labor force in Taiwan compared to 5.3% in Hong Kong.18 This is reflected in tabulations from the Female Employment and Marriage Survey which is similar to the Manpower Utilization Survey but includes additional questions about women’s family and career situations. In 2000, 72% of children aged 0 to 3 had their parents as primary caregivers, 21% were taken care of by grandparents, 6.5% by nannies, 0.33% by daycare and a mere 0.2 percent by FDWs.19 16

Ideally, we would have liked to use Taipei as the control group, as Taiwan is a much larger country with a significant fraction of the population living in smaller cities and rural areas. Unfortunately, this is not feasible as doing so would drastically reduce the sample size. We discuss some results from restricting the focus to the larger urban cities in Taiwan in Section 3.3.2. 17 Note that it is common for households to forge medical documents in order to hire a foreign caretaker to perform domestic or childcare duties at home. For our purposes, we do not draw a distinction between foreign caretakers and foreign domestic helpers in Taiwan. It is likely that the total stock of foreign caretakers and foreign domestic helpers is an upper bound for the number of foreigners performing domestic childcare duties in households in Taiwan. 18 The Taiwan estimate is from the Yearbook of Labor Statistics published by the Council of Labor Affairs (http://statdb.cla.gov.tw/html/year/d13010.htm). Foreign domestic helpers in Taiwan are defined as the sum of nursing workers and home-maids. The Hong Kong statistic is based on our tabulations from the Hong Kong census. 19 The exact question asked in the Female Employment and Marriage Survey is “Who is/was the primary caregiver of your youngest child when he/she was less than 3 years old?”.

11

3.1

Identification

The basic idea of the macro approach is to exploit differences in the scope of program adoption in Hong Kong and Taiwan and to examine how female labor force participation changed differentially over time across both countries. Hong Kong experienced rapid economic growth during the period in which the FDW program was established. As the economy grew, wages and unearned incomes of women rose; therefore, changes in the observed quantities of foreign domestic helpers and the labor force participation of women cannot be fully attributed to the creation of the FDW program. That is, part of the increase in in female LFP in Hong Kong over this time period is due to demand shifts, unrelated to the FDW program. Furthermore, we do not have a clear before-and-after policy change within Hong Kong as the FDW program was adopted gradually in Hong Kong starting from the late 1970s. Nevertheless, as Taiwan experienced a similar growth path and shares important economic and cultural similarities to Hong Kong (this is further elaborated in Section 3.2), the comparison of female LFP rates across the two countries allows us to address these empirical challenges. Assuming that Taiwanese women experienced similar labor demand shifts as their counterparts in Hong Kong, the differential change in female labor force participation over time across the two countries can be attributed to the differential increase in female labor supply induced by the availability of FDWs. A natural concern with the simple difference-in-difference comparison of female LFP rates between Hong Kong and Taiwan over time is that it is likely to pick up country-specific shocks that are unrelated to the FDW program.20 To alleviate these concerns, we introduce a third difference by comparing the labor force participation rates of mothers with older children (youngest child aged 6 to 17) to mothers with younger children (youngest child aged 0 to 5). As the age structure of the youngest child is a reasonably good proxy for the demand for household services, mothers with younger children are likely to have a greater response to declines in the relative prices of domestic help as compared to mothers of older children. As shown in Figure 3, among native women with high education, starting from the early 1980s, just a few years after FDWs were first allowed into Hong Kong, mothers with youngest child aged 0 to 5 were about 20 to 30 percentage points more likely to hire a FDW than women whose youngest child was 6 to 17 years old. Among native women with medium education, demand for FDWs only started increasing in the early 1990s. By the late 1990s, those with youngest children aged 0 to 5 were about 2 times more likely to hire a FDW than women with older children. Among mothers with low education, the employment of FDWs remains very low for women with and without young children over the entire time period.21 20

For example, Hong Kong and Taiwan have experienced large demographic and economic changes over the past two decades which may have differentially affected female LFP rates across the two countries (Lui and Suen, 2005, Vere, 2005). Section 3.2 provides some suggestive evidence that observed compositional changes e.g. in terms of age, education, presence of young child and unearned income have evolved quite similarly over time for both countries. 21 For Figure 3, of the 158,148 married women with youngest child aged 0 to 5 in our sample, 10% have high education, 70% have mid education and 20% have low education. Among the 290,070 women with youngest child

12

The third difference, by child age structure, allows us to introduce country and country-year fixed effects to account for differences in underlying trends in female employment across countries and unobserved country-specific demand shocks. Similarly, differential trends and unobserved shocks that affect mothers of younger versus older children can also be captured in the model. This triple-difference strategy implies that differential trends in demographic and economic conditions in Hong Kong and Taiwan that are unrelated to the FDW program will only bias our estimates if they have differential effects across women with younger versus older children within each country. We interpret the difference in the growth of female labor force participation of these two groups, adjusting for composition changes, as providing a measure of the impact of the foreign worker policy on female labor supply in Hong Kong. The key identifying assumption is that in the absence of the FDW program, the relative growth of female LFP rates across mothers with younger and older children in Hong Kong would be similar to that in Taiwan.

3.2

Taiwan as a Control Group

Before turning to the main graphical analysis and empirical estimates of the macro approach, this section motivates the use of Taiwan as a control group by providing a brief description and comparison of the economic and demographic trends as well as the structure of the childcare market in Taiwan and Hong Kong.

3.2.1

Economic and Demographic Trends in Hong Kong and Taiwan

As first evidence that Taiwan is a reasonable control group for Hong Kong, we show that the two countries experienced very similar trends in the main observed determinants of the labor supply for women. Figure 4 depicts the predicted labor force participation of mothers in Taiwan and Hong Kong from a cross-sectional model using as explanatory variables, age dummies, education dummies, husband’s income percentile dummies, a dummy for youngest child aged 0 to 5, and a dummy for Hong Kong. The model is estimated using all available years. As observed, the evolution of LFP due solely to compositional changes in the explanatory variables is remarkably similar across the two countries. Basic descriptive statistics of demographic and economic variables in Hong Kong and Taiwan are presented in Appendix Table A1. Although most of the differences between Hong Kong and Taiwan are statistically significant, the magnitude of the differences are relatively small considering that Hong Kong and Taiwan are two different countries. aged 6 to 17, 4% have high education, 58% have mid education and 38% have low education. Of the 240,905 women with no children aged 0 to 17, 10% have high education, 55% have mid education, while 35% have low education.

13

3.2.2

The Childcare Market in Taiwan and Hong Kong

Cultural Similarities and the Role of Grandparents Hong Kong and Taiwan share very similar cultural heritage and family values. The extended family system traditionally provides for many of the needs of family members, including childcare. Appendix Table A2 shows how the probability that a married woman (who is the household head or the spouse of the household head) lives with her mother or mother-in-law has evolved in the past decade in Hong Kong, Taiwan and the US.22 Two observations are worth noting. First, the share of married women living with the extended family is very high compared to the US. For example, married women are between 10 to 20 percentage points less likely to be the household head or the spouse of the household head in Hong Kong and Taiwan as compared to the US. Among those who are the household head or the spouse of the head, Hong Kong and Taiwanese women are about four times more likely to live with their mother or mother-in-law. Moreover, in Hong Kong and Taiwan, even among those who are not living with their extended family, it is still quite common for the parents to live close to their adult children and to be involved in childcare (Lee et al., 2000, Tam, 2001).23 Most important for our empirical exercise is the observation that for both countries, the probability of living with a mother or mother-in-law has stayed high and relatively constant over the past decade, in particular for women with a medium or high education.24 Market provision of childcare Nurseries for children younger than three are not very widespread in both countries. In Taiwan, the lack of FDWs has not been compensated by a larger supply of nurseries or childcare centers. As mentioned previously, using the Female Employment and Marriage Survey (FEMS), we found that in 2000, less than 0.5 percent of Taiwanese children aged 0 to 3 went to day care; most of them were taken care of by parents or grandparents (93 percent) or by nannies (6.5 percent).25 Similarly, Lin and Yang (2009) finds that in 1980, only 0.2% of Taiwanese children aged 0 to 3 went to creches or childcare facilities and in 2006, this number remained low at 0.33%. For Hong Kong, we were not able to locate a similar representative survey to provide directly comparable estimates of the care-giving situation. Turning to secondary sources, Tam (2001) interviewed 75 22

The time period (1993-2006) is restricted due to the availability of the variable describing the relationship of the household members to the household head in the Hong Kong GHS. This variable is not available prior to 1993. 23 The study by Lee et al. (2000) randomly sampled 2203 elderly Chinese aged 65 and over in Hong Kong in 1991 and found that of those with children, about 80 percent were living with their children. Of the remaining one-fifth, about a third were living in the same district as their children. 24 Due to restrictions on the codes for the reported relationship to the household head, we can only identify the presence of the mother or mother-in-law if the woman is the head or the spouse of the household head. We assume that if the women is neither, she is likely to be living in an extended household. 25 We were not able to use the 0 to 5 age range as the question asked in the FEMS survey was “who is/was the primary caregiver of your youngest child when he/she was less than three years old?”.

14

mothers of children aged 0 to 4 years who were working in a tertiary education institution in Hong Kong - approximately 50 percent of them reported that their major child care mode was FDWs, 31 percent were family members or relatives, 16 percent reported nursery and creches and 4 percent relied on childminders or nannies. Although this is not a representative sample of mothers in Hong Kong, it provides suggestive evidence of the differences in childcare options available to mothers in Taiwan and Hong Kong. Shifts in the demand for childcare As in many other countries, higher returns to education and the increase in the price of women’s skills have generated a substantial increase in the labor force participation of married women and mothers in Hong Kong and Taiwan. Figure 5 presents the evolution of the average monthly wage and the 10th percentile wage (in constant local units) for women in Hong Kong and Taiwan, and for Hong Kong, the minimum wage for FDWs. The purpose of this figure is three-fold. First, to show that average female wage has followed similar paths in the two countries. Second, assuming that maids and nannies are drawn from the bottom of the wage distribution, so have the prices of outsourcing household production in the absence of FDWs. Third, that the presence of FDWs has significantly reduced the relative price of outsourcing household production in Hong Kong as compared to Taiwan.

3.3 3.3.1

Triple Difference Estimates Graphical Evidence

Figure 6 provides graphical evidence that the trends in labor force participation of females with a younger child aged 0 to 5 and females with a relatively older child aged between 6 to 17 has evolved quite differently across the two countries starting in the late 1980s. In particular, while the change in employment of females with younger and older children was relatively similar prior to 1981 in both countries, labor force participation rates of women with younger children in Hong Kong accelerated starting in the late 1980s, such that by 2006, it actually exceeded that of women with older children. This is in stark contrast to Taiwan, where the growth in employment for the two groups of women remained virtually parallel over the entire sample time period from 1976 to 2006. In Figure 7, we separately graph the trends in LFP rates of women in the two countries by education level. All of the catching-up in LFP of younger women in Hong Kong can be attributed to trends in the LFP of higher and middle educated women. Highly educated mothers, whose youngest child is less than five, started participating in the labor market at almost the same rate as mothers of older children before 1986. Since then, they generally have higher participation levels compared to 15

mothers of older children. Highly educated mothers of young children in Taiwan, on the other hand, show very high levels of labor force participation, and although the gap between their LFP rates and that of mothers with older children is small, it has stayed permanently below. The catch-up in labor force participation of medium skilled mothers with young children in Hong Kong started in the early 1990s. They reach participation levels comparable to those of mothers with older children by the late 1990s, but diverge again at the end of our period of analysis. These employment patterns are broadly consistent with trends in the relative wages of native women as a fraction of the wages of FDWs as shown in Figure 2. Relative wages of high skilled women increased rapidly in the early 1980s, and by the mid 1980s, was about four times that of the minimum wage of FDWs. The relative wages of medium skilled women increased steadily over the time period, albeit at a slower pace than that of high skilled women. Convergence in LFP rates for medium skilled mothers with younger children occurred in the mid-1990s when relative wages were about three times that of the minimum wage of FDWs. The employment trends of lower-educated females in both Hong Kong and Taiwan appear to be mostly similar for both groups over the sample time period. Relative wages for this group of mothers increased relatively slowly over time but appears to be too low to justify the cost of hiring a FDW. Overall, these findings are consistent with the view that higher educated women are more likely to respond to changes in the price of domestic services due to their higher opportunity cost of household production. Nonetheless, these figures do not control for changes in the composition of both groups of women over time - to the extent that there may be differential changes in the composition of mothers with older or younger children across time, these effects may confound the aggregate trends that we observe in the graphs. In the next section, we will provide formal econometric evidence that adjusts for such composition effects.

3.3.2

Formal Econometric Evidence

We estimate the regression analogue of Figures 6 and 7, adjusting for relevant individual covariates such as age and education:

Yijgt = γjt + λtg + τjg + βt Djgt + δXijgt + ijgt

(1)

where i is the individual, j is the country, g is the group (whether female has older or younger child) and t is the time period. The time period, t, can take the values of a dummy for 19761984, 1985-1987, 1989-1993, 1994-1998, 1999-2002 and 2003-2006. Note that the Taiwanese data only starts in 1978 while the closest census year available for Hong Kong is 1976. For most of our cross-country analysis, we will compare 1976 LFP rates in Hong Kong with 1978 LFP rates

16

in Taiwan.26 Vector Xjgt are individual-level controls for age and education. D represents the relevant indicator variables; Djgt = 1[HK = 1, Y oungchild = 1, period = t]. All the specifications for the macro approach control for year, country and child-age fixed effects as well as country*year and child-age*year fixed effects. This approach enables us to identify the causal effect of the expansion of the foreign domestic helper program on female LFP rates in Hong Kong provided that there are there are no differential shocks affecting mothers of younger versus older children across Hong Kong and Taiwan. To further control for possible demand shocks that may have differential effects across older vs. younger children in both countries due to the fact that the presence of a young child is likely to be correlated with mother’s age and education and that the returns to these characteristics may be changing over time, we also estimate specifications that allow the effects of mother’s age and education to vary by year by including age*year and education*year fixed effects. Finally, to more adequately control for secular changes in composition of mothers across groups, for example, due to delays in childbearing over time, we also estimate specifications that include a full set of interactions between age*education*year. This will capture cohort-specific shocks to labor demand and allow us to compare women with younger versus older children within cohort and education levels across Taiwan and Hong Kong, making it less likely that differences in composition or the returns to characteristics such as age and education across mothers with older vs. younger children over time are confounding our estimates.27 The main results are presented in panel A of Table 1. The first column of each time period is the raw, unadjusted, difference obtained from the coefficient on Djgt . The second column adjusts for an individual’s age and education and the third column include age and education fixed effects interacted with year and age*education*year. In the fourth column, we also allow the effects of education to vary by country by including a full set of interactions for education*year*hk and age*education*year*hk. Standard errors are clustered at the country-year level.28 Across the four specifications, the coefficient estimates are very similar. The first row suggests that, relative to mothers of older children, mothers of younger children participated less in HK than in Taiwan. The results for the most flexible specification indicate that relative to 1976-1984 (the base period), the gap between HK and Taiwan started closing in 1985-1987; by 1989-1993 there was no difference 26

In some figures, employment data for Hong Kong in 1976 are labeled as 1978 for ease of comparison with the 1978 employment data in Taiwan. 27 We thank a referee for making this suggestion. 28 It is likely that the error terms are not only correlated within country*year groups (which we address by clustering by country*year) but also across time generating serial correlation issues. A common approach is to cluster at a higher level. In our case it will imply clustering at the country level, but that will leave us with only two clusters. We increase the number of clusters by assuming independence across education groups within countries and clustering at the country*education group. Standard errors clustered at this level are presented in square parenthesis in panel A of Table 1.

17

between the two countries, and by the mid 1990s-early 2000s the relative participation of mothers of young children vs. older children in HK was 7 to 9 percent higher than the relative participation of mothers of young children in Taiwan. In 2003-2006 the difference decreased slightly to 3 to 6 percentage points. Summarizing, between 1976 and 2006 the gap in LFP rates between mothers of younger and older children decreased by between 7.6 to 11.6 percentage points more in HK than in Taiwan. Panel B of Table 1 analyzes whether the evolution of female labor force participation rates varies by the educational attainment of the women. Highly educated mothers of young children were the first to significantly increase their labor force participation. By the mid 1980s (a few years after the first FDWs came to HK), their relative labor supply had achieved its maximum. On the other hand, medium educated mothers of young children only started to significantly increase their relative LFP in the early 1990s, in line with the secular decline in the relative price of hiring a FDW. As predicted by the model in the next section, declines in the relative cost of the FDWs prompted more and more women to start hiring them and participating in the labor force. Finally, we observe very small effects on mothers with the lowest education levels, at least until the last period. This is consistent with the fact that the potential wage of most mothers in this group is below what it needs to be to justify hiring an FDW. Hence, changes in the prices of domestic help had little effect on their labor force participation.29 In the later two periods from 1999 to 2006, the estimated program effects for mothers with low education are larger and statistically significant. This is not entirely surprising as the relative wage of low educated mothers has been steadily increasing from 1980 to 2000 and was twice that of the FDWs by around the late 1990s, making hiring an FDW an option for some low-educated mothers as well. As mentioned previously, ideally, we would have liked to compare LFP trends in urban Taiwan and Hong Kong, as urban and rural areas in Taiwan may be quite different. However, doing so would result in a large reduction in the sample size, making it difficult to investigate trends in LFP rates separately by education groups.30 We estimate the main regression specification for the subset of Taiwanese women residing in the seven largest urban cities in Taiwan.31 The results are shown in Appendix Table A3 - the estimates are quite similar to the first four columns of Table 1. This suggests that our estimates are not primarily driven by potentially dissimilar LFP trends in rural Taiwan. 29

Note that unskilled women are much more likely to belong to lower income households, and thus might not be eligible to hire a FDW due to the minimum household income requirement. 30 The urban sample is about 20% of the overall Taiwan sample. Furthermore, the number of Taiwanese women in the urban sample with low and high education is very small in the later and earlier years of the sample period, respectively. 31 The seven largest urban cities include Taipei, Kaohsiung, Taichung, Tainan, Chiayi, Keelung and Hsinchu.

18

3.3.3

Potential Confounds

One limitation of this approach is that if underlying trends or unobserved shocks to female LFP vary across Hong Kong and Taiwan by child age, this may confound our triple difference estimates. The graphical and empirical analysis by mother’s education level provide some suggestive evidence that the patterns we document are not entirely driven by the presence of such confounds. The household production model and income restrictions on hiring a FDW suggest that mothers with low income are less likely to respond to the availability of FDWs, at least in the earlier time period of our study. This suggests that low educated mothers can be used as a “placebo” group to test our identification assumptions. A casual inspection of the trends in female LFP over time for mothers of younger vs. older children in Hong Kong and Taiwan for low educated mothers in Figure 7 indicates that within each country, the trends in female LFP are virtually parallel. This is further corroborated by the empirical estimates in Table 1. Hence, while it is possible that differential economic shocks and other sources of differential trends in female labor supply by age of children in Hong Kong and Taiwan may confound the estimates, such a shock will have to affect only highly skilled mothers and not mothers with low education levels. Moreover, as mentioned previously, the magnitude and timing of the program estimates for each education group matches the trends in the relative wages of FDWs very well (see Figure 2), suggesting that potential confounds would have to fit these employment and relative wage patterns across education groups as well. Are the estimates driven by the comparison with Taiwan? Finally, one concern with the use of Taiwan as a comparison group is that unobserved labor demand shocks due to business cycle fluctuations, differences in underlying trends in female LFP or differential trends in wage inequality across the two countries may make it hard to draw any inference about policy impacts. Depending on one’s priors, this concern is made more acute by the fact that Hong Kong and Taiwan are different countries and are likely to be subject to differential economic shocks. For example, one might be concerned that wages are much more unequal in Hong Kong than in Taiwan (see Figure 5) and that our estimates could be picking up some of these underlying differences across the two countries. The use of a within-country comparison group, namely, mothers of older versus younger children, helps to ameliorate this concern to some extent. In Appendix Table A4, we test whether the estimates are driven primarily by the comparison with Taiwan by looking at a difference-in-difference specification within Hong Kong alone that compares the growth in female labor supply for mothers with older versus younger children. These overall estimates are quite similar in magnitude to that of the triple difference approach for the overall sample of women and

19

exhibit largely similar patterns by mother’s education level. This suggests a somewhat limited role for unobserved shocks that differentially affect Hong Kong and Taiwan to fully account for the time-series patterns in female labor supply that we document. As a robustness check, we also explore female LFP trends in the United States. To address the concern that our estimates may be driven by differential trends in wage inequality in Hong Kong and Taiwan, looking at female LFP trends in the US may be instructive as, like Hong Kong, the US also experienced a large increase in inequality since the early 1980s. Moreover, since the US had no foreign worker program, we can consider the US as an alternative control group. Using the March CPS, Appendix Figure A1 shows the overall LFP rates of mothers by age of youngest child in the US and separately by education level. Strikingly, both the overall patterns as well as patterns by education group are very similar to those found for Taiwan - trends in LFP rates of mothers are virtually parallel for mothers of younger versus older children over the period from 1978 to 2006.

4

Micro Approach: A Structural Model of Female Labor Supply

In this section, we develop and estimate a structural model of labor force participation and the decision to hire a FDW using pooled cross-sectional data from the 2001 and 2006 Hong Kong population census. The advantage of deriving our empirical specification from an economic model of behavior is that we can interpret the estimated coefficients as meaningful parameters related to women’s labor supply and perform welfare calculations. We can also simulate the counterfactual supposing no foreign domestic program to estimate the microeconomic effects of the program on female labor supply. This will enable us to compare the estimates from the structural model to the macro difference-in-difference estimates.

4.1

Model

We consider a static model of utility maximization, where we assume that fertility and education are exogenously determined. Women maximize a utility function that depends on the consumption of a market good and leisure, subject to budget and time constraints. There are two discrete choice variables: labor force participation (LFP) and the decision to hire a foreign domestic worker (FDW).32 The woman’s problem is: 32

We do not model or estimate the effect on intensive margin of labor supply because the number of hours worked is not reported in the Census. Note, however, that in Hong Kong part-time jobs are very rare, only 10 percent of the working women in our sample report working less than 35 hours per week. Our wage measure refers to monthly employment earnings.

20

M ax

U (c, l)

st.

(2)

I + LF P ∗ w = c + F DW ∗ wn l = T − d − LF P ∗ h f (d, F DW ) = R where I is unearned income, w is the market wage for the woman, wn is the monetary cost of hiring a FDW, T is total time, d is the woman’s own time devoted to household work, and h is the fixed number of working hours. R is the fixed amount of household work and childcare needed to keep a household running smoothly; it is produced using the woman’s time and the F DW. We do not impose any constraint on the degree of substitution between the woman’s time and that of the F DW . R and f will depend on the household composition of the family, in particular, the presence of children younger than 5. We assume the utility function is additively separable in consumption:33

U (c, l) = U1 (c) + U2 (l; F DW )

(3)

We assume that the marginal utility of leisure depends directly on the presence of a F DW at home. Evidence suggests that the presence of a F DW , keeping time-use constant, has potential negative effects on privacy and family life quality in general (Lan, 2003). Without this assumption, it is difficult to explain why very rich women with no children rarely hire a F DW . We also allow the function U to vary by observable characteristics of the woman xi , that might include, for example, the education level of the woman and the age composition of her children. The woman faces four mutually exclusive alternatives denoted by j: j = 0 if LF P = 0 and F DW = 0, j = (p)articipate if LF P = 1 and F DW = 0, j = f (d)w if LF P = 0 and F DW = 1, and j = pd if LF P = 1 and F DW = 1. Normalizing U0 = 0, we approximate the utility attached to each alternative relative j = 0 using the following reduced form empirical model: 33

We need separability for the validity of the identification assumption. Our instrument, the number of rooms, enters in the utility equation of hiring a FDW, but not in the utility equation of participating in the labor force (the instrument is discussed in Section 4.3). Note, however, that it is a strong assumption and studies have found evidence against it (Browning and Meghir (1991), Shaw (1989), and Ziliak and Kniesner (2005)). Ziliak and Kniesner (2005), in particular, find that hours worked and consumption are complements, and that assuming separability leads to underestimation of the wage elasticity of labor supply.

21

Up = β1 + β2 xi + β3 ln w + ε2

(4)

Ud = δ1 + δ2 xi + ε3 Upd = Up + Ud + complementarity/substitution = (β1 + δ1 + π1 ) + (δ2 + β2 + π2 )xi + β3 ∗ ln w + ε2 + ε3 We interpret ε2 and ε3 as unobserved components of the utility from hiring a FDW and working in the labor market. β1 proxies for the disutility from working in the market, δ1 for the utility of the extra leisure available to the woman from hiring the FDW net of the monetary and psychic costs of hiring her (we cannot separately identify the benefits from the costs because there is no variation in the FDW’s minimum wage across households), and through β2 and δ2 , we allow these parameters to vary with observable characteristics of the woman (denoted by xi ). The utility of both working and hiring a FDW (Upd ) is defined as the sum of the utilities of each action separately modified by potential interactions (positive or negative) between the two, which we model with the π 0 s. Specifically, the complementarity/substitution effect is modeled as π1 +π2 xi , where π1 is the constant effect across all women and π2 allows the interaction effect to vary across different groups of women. A positive interaction coefficient implies that the decisions of whether to hire a FDW and of working in the market are complements, and therefore, changes to variables that affect the utility of hiring a FDW, for example a reduction in the minimum wage of foreign domestic helpers, will not only induce women to outsource more household production, but also to join the labor force. This complementarity is closely related to the degree of substitution between a woman’s time and that of the FDW in caring for children and doing household chores. Given that most people in Hong Kong work full-time, if a woman participates in the labor market, hires a FDW and has a young kid, most of the childcare is likely to be provided by the FDW without the mother being present (unlike the case when a FDW is hired but the mother stays at home). The complementarity might also come from a woman’s ability to work in higher paying jobs because of the flexibility in working hours allowed by hiring the FDW. Note that we allow the interaction π2 to vary by observable characteristics of the woman. In particular, for the reasons stated above, we might expect the complementarity term to be especially large for women with very young children. The interaction terms are key in predicting the effect of the foreign domestic worker program on the labor supply of women and to study which groups of women are most likely to change their participation decisions as a result of the program. We restrict the δ 0 s, β 0 s, and π 0 s to be constant across individuals and assume the x0 s to be independent from the error terms. The variance-covariance matrix of the error terms takes the following form: 22

"

ε2 ε3

#

" ∼N

0,

1 σ . 1

#!

As the above expression suggests, we are modeling our discrete choice maximization problem as a multivariate probit. Note that we are imposing an additional restriction beyond what is needed to set the scale of utility by assuming there is no unobserved complementarity component (ε2 + ε3 = ε4 ).

4.2

Identification of the Complementarity between LFP and FDW Hire

We have already normalized the model to account for the fact that the level and scale of utility are irrelevant. While a normalized multivariate probit is formally identified as long as the model includes at least one variable that varies at the individual level (Heckman and Sedlacek, 1985), in the absence of exclusion restrictions, the model is extremely fragile (Keane, 1992). Without an exclusion restriction, there is no variation in the data that allows us to discern if women who work and hire a FDW choose this alternative because of the complementarity between the two choices or because of a strong correlation between the unobserved determinants of the choices’ utilities. Identification relies solely on functional form assumptions. To solve the identification issue we use the number of rooms, specifically having 4 or more rooms, as an instrument. We argue that this variable is likely to have a direct effect on the utility from hiring a FDW but arguably does not enter the utility from working in the market directly. Intuitively, the idea behind how this exclusion restriction identifies the complementarity term is the following: Suppose there are two identical households that only differ in the number of rooms in the house. Complementarity between LFP and hiring a FDW implies that the woman living in the house with an additional room will be more likely to work in the market. In the absence of an interaction between the two choices, women in both households should be equally likely to participate in the labor force. We will discuss the validity of the exclusion restriction in detail in the next section. Before turning to the discussion of the instrument, in Table 2 we present the summary statistics for the sample that will be used in the empirical analysis in this section.34 As mentioned before, the sample is drawn from the 5% sample of the 2001 and 2006 Hong Kong Census. We restrict our sample to married mothers aged 25 to 54; the lower limit is set such that most women would have completed their education by that age. 34

Note that the share of mothers with a FDW is lower than in Figure 3. The reason is that the sample used in the micro approach includes only women that live in places with 3 or 4 rooms. See Table A5 for the distribution of number of rooms across the population.

23

4.3

Proposed Instrument

The instrument that we use is the number of rooms in the household, specifically having 4 or more rooms. We would have preferred to use the number of bedrooms, but the Census question does not distinguish between bedrooms and other types of rooms.35 The motivation behind this instrument is that space limitations in Hong Kong coupled with restrictions on lodging for domestic workers imply that all else equal, a household living in a house with more rooms is more likely to hire a domestic worker.36 Hence, assuming we are able to control adequately for household wealth and other characteristics of the household, we would not expect the number of rooms in the house to be correlated to an individual’s unobserved work propensity. There are a number of concerns with using the number of rooms as an instrument. We discuss each of these concerns in detail.

4.3.1

Issue 1: Endogeneity of the Number of Rooms

A primary concern with using the number of rooms as an instrument is it may (1) be correlated with an individual’s unobserved preferences for work and (2) proxy for household wealth. Appendix Table 5 shows basic demographic and economic characteristics by number of rooms of the households. It is clear that there are statistically significant differences between households who live in places with different number of rooms. Larger, richer, and more educated families live in places with more rooms. While the model conditions on all available observable household characteristics, the concern that arises is that the instrument may nonetheless be correlated with unobserved determinants of female labor supply. To address the concern that our instrument is merely proxying for an individual’s unobserved propensity to work or unobserved household wealth, we run “placebo” tests of the reduced form of having 4 or more rooms on labor force participation for subsets of households that have a very low probability of maid hire, such as married households without children and low income households who are not eligible to hire a foreign maid. The idea here is that if our instrument merely proxies for an individual’s unobserved preferences for work, we would expect to find a significant positive relationship between the number of rooms and the employment status of females of these households, regardless of their low demand for domestic services. However, if the reduced form relationship is only significant in the sample of households that have a relatively high demand for domestic help (our main sample), then this suggests that the number of rooms affects employment decisions of females through its impact on maid hire, as opposed to merely proxying for some unobserved 35

The exact question in the Census is: “Number of rooms in the accomodation, excluding kitchens and bathroom/ toilets/cocklofts/bedspaces.” The possible options range from zero rooms to six rooms or more. 36 For example, it is stated in the employment contract that they cannot sleep in the kitchen or share a room with an adult of the opposite sex. It is very common, however, for the FDW to share a room and even the bed with at least one of the children (Chan, 2005).

24

variables that might be correlated with the individual’s propensity to work. Further, to circumvent the issue that omitted household wealth might understate the degree of complementarity between the two decisions, we restrict our sample to households that have 3 or 4 rooms, and define our instrument as a dummy for having 4 rooms.37 We do this for two reasons. First, given that we are looking at families with at least one child and assuming one room is the couple’s bedroom, the second is the child’s and the third is the living or dining room, going from 3 to 4 rooms will surely relax the space constraint. Second, even after controlling for husband’s wage, owning very large houses (5, 6, or more rooms) or very small ones (1 and 2 rooms) are likely to be a good proxy for unearned income, especially given how unusual they are. More specifically, for our placebo tests we estimate the following linear regression:

LF Pi = α + δI(N umber of rooms = 4)i + γXi + θdq + ijt

(5)

and its probit counterpart, where i is a mother, θdq are fixed effects for Hong Kong districts (d) and quarter types (q).38 The covariates (Xi ) in this regression include those in the structural model: age and age squared of the mother, education dummies, number of children, dummy for child five or younger, dummy for person 65 or older in the household, log of husband’s wage and household size We run the above specification for three groups of women: (1) married women with children (our main sample), (2) married women with no children, (3) low-educated mothers whose husband earns less than 10,000 HK dollars per month (the bottom quartile of the wage distribution). The share of households with an FDW for the three groups is 14 percent, 2.6 percent, and 1.5 percent, respectively. Table 3 presents the results: the coefficient on the number of rooms for the sample of married women with children is positive and highly significant. By contrast, the coefficient is close to zero and not significant for married women with no children, and positive, but half the size and not statistically significant, for low educated mothers with low unearned income. The lack of a significant reduced form relationship between our instrument and the employment probability of females in these households that have a relatively low demand for maid services is reassuring and reinforces the validity of our instrument. 37

Our main results and placebo tests are robust to alternative definitions of the sample and of the instrument. Estimations using (1) original definition of the instrument but sample extended to include households living in places with 5 rooms and (2) instrument defined as having 5 or more rooms and sample restricted to places with 4 or 5 rooms are presented in Tables A6 and A9 of the Appendix. 38 There are 30 districts and 7 types of quarters identified in the Census questionnaire. Results are robust to excluding these fixed effects from the placebo test estimation. Fixed effects for district and quarter type could not be incorporated into the multivariate probit model (structural model) given the large number of dummies that would need to be included in the estimation.

25

One limitation of these placebo tests is that we cannot reject the possibility that the number of rooms may have different effects on female labor supply decisions unrelated to FDW hire that vary systematically across different types of households. Nevertheless, given that each of these groups considered (married women without children and low educated women) represent a non-trivial segment of the female working population, this possibility appears unlikely.

4.3.2

Issue 2: Moving Concerns

The second concern is that exogeneity of this instrument implicitly requires that individuals either face prohibitively high moving costs or that some frictions in the housing market limit the ease of moving. Since we only have cross-sectional data, it is not clear whether the observed relationship between the number of rooms and the probability of maid hire reflect space constraints or households moving to a larger place when they decide to hire a maid. Such actions by the household may lead to endogeneity in the choice of the number of rooms in the household. We follow two strategies to address this concern. First, we present models with the sample restricted to women living in subsidized sale flats. Due to limited space and the high costs of housing, almost half of the population in Hong Kong resides in some form of government housing while the remainder lives in private housing. In 2005, 29.1% of households were tenants in government provided housing while another 15.8% owned subsidized flats through the Home Ownership Scheme (HOS) (Census and Statistics Dept, 2006).39 For individuals residing in government subsidized housing, mobility and choice are rather limited due to various restrictions imposed on the resale and allocation of flats (Lui and Suen, 2010). The HOS allows citizens in just the right income bands to buy their flats at the cost of construction and get the land element for free.40 This normally means discounts of between 30 percent and 50 percent on private flat prices. Each year the government offers a given number of flats for sale. The flats are in very large projects constructed by the government for this purpose. Each household can apply for a flat in only one project and does not have to state its preferred number of rooms or size of the flat. Applications usually outnumber supply by huge numbers; for example, in 1994, 112,345 families applied to buy the 8,168 flats offered. Allocation is based on a computer ballot that determines the flat selection sequence. After the third year of occupancy, HOS flat owners may sell their flats in the open market, but only after paying a premium to the Housing Authority, which is equal to the value of the subsidy at prevailing market prices. Additionally, if they sell they become ineligible to apply for subsidized housing in the future. Referring to the allocation and resale rules of the HOS, the South China Morning Post states “Thus, most of the buyers are 39 40

In our sample, the share living in subsidized sale flats is higher at 24 percent. The upper limit has always been above the minimum income required to hire a FDW.

26

forced to stay where they are, irrespective of where they would like to be, which they never had much choice of anyway when they bought their homes.” Second, using the entire sample, we can explicitly test whether households that have moved in the last five years are more likely to hire domestic helpers. In results presented in Appendix Table A7, we find that having moved in the past 5 years is uncorrelated with a higher probability of having a domestic worker. We also perform this test on a subset of households whose only child is five or younger. Given that the probability of maid hire increases substantially when a household has young children, this group of households is likely to be first-time employers of foreign maids. Hence, looking at their moving behavior in the previous five years provides a test of whether households move in anticipation of hiring a foreign maid. The results for this subsample of households with small children are also not statistically significant and very small. We interpret these results as suggesting that families are not moving in large numbers to accommodate a foreign domestic worker. To complement the evidence in Table A7, we will also present estimates of our labor supply models using a subsample of households that did not move in the last five years. We will compare the estimates obtained from the full sample to the sample of non-movers to see if this is indeed a large concern.

4.3.3

Fertility

Before turning to the structural estimates, we will briefly address the implications of assuming that fertility is exogenous in our model. In a standard model of fertility and labor supply behavior, both decisions are jointly determined and depend on the mother’s earnings and the costs of raising children. Given the joint nature of the labor supply and fertility decision, ideally, we would like to implement a dynamic model that estimates the effects of the FDW program on married women’s decisions to participate in the labor force and have to have children.41 Due to the lack of panel data for Hong Kong, we are unable to endogenize fertility in a dynamic model. Even if we settled for a static model that explains completed fertility and labor force participation, we would face two major empirical challenges. The first is that we would have to limit our sample to women old enough to have completed their fertility, but young enough such that all her children are still living at home e.g. women between the ages of 40 to 44. Second, we would need to find another reasonable instrument to identify the fertility equation. For these reasons, we conclude that endogenizing fertility in our structural model is not viable. Nevertheless, we will discuss the ramifications of endogenous fertility on our structural estimates. This discussion is based on a sketch of a simple static model by Blau and Robins (1989). The reduction in the cost of childcare induced by the availability of FDWs increases the relative wages 41

Moffitt (1984) provides a discussion of static and dynamic models of fertility and labor supply.

27

of mothers and at the same time reduces the cost of having children. Therefore, assuming that substitution effects dominate, the direct effect of the FDW policy would be to increase female labor supply and fertility. However, the increase in labor supply would indirectly affect fertility since the increase in labor supply itself tends to increase the cost of childcare as part of this cost increases in the number of hours the mother works. Similarly, if lower childcare costs raises fertility, this would indirectly reduce the labor supply of mothers as they incur the costs of bearing and raising additional children. Without further assumptions, the net effects of childcare costs on fertility and labor supply cannot be signed. What this discussion implies is that the indirect effects of a reduction of childcare costs on either labor supply (through increased fertility) or fertility (through reduced labor supply), would tend to offset the direct effects. Therefore, our structural model would tend to lead to an underestimate of the effect of childcare costs on labor supply as fertility is not explicitly taken into account. In other words, our estimates are capturing the potential reduction in labor supply due to the offsetting effects of fertility choices in response to the change in childcare costs.

4.4

Structural Estimates

In this section we discuss how we estimate the model (4) and present the results. We first focus on the estimation of the effect of changes in wages on labor force participation (parameter β3 of our model) and then turn to the estimation of the whole model.

4.4.1

Identification of Wage Effects

Given the lack of wage data for non-participants, in order to identify wage effects, we follow a two-step procedure based on Wooldridge (2002). In the first step, we estimate a Heckman selection model that allows us to construct the predicted wage for both participants and non-participants. In the second step, we include the predicted wage in the multivariate probit model. To identify the selection equation, we use three instruments commonly used in previous literature to estimate models of female labor supply (see Mroz, 1987 for a summary). These instruments include a dummy for having a child aged 0 to 5, a dummy for a person aged 65 or older in the household and husband’s wage. Conceptually, these variables are related to the productivity of the woman in household production, the amount of household work required and to the woman’s unearned income, which are important determinants of labor force participation in a simple time-use model, but a priori should not relate directly to the woman’s potential wage in the labor market, at least conditional on the individual and household characteristics in the model. As the predicted wage variable is included in a model that estimates the LFP decision in the second

28

step, in order to separately identify wage effects from the participation decision, we will need an instrument that belongs to the wage equation but is excluded from the participation equation. The instrument we use is a dummy variable that indicates if the woman speaks English - the rationale is that speaking English is likely to affect wages as it is considered a valuable skill in the Hong Kong labor market, but is unlikely to have a direct effect on an individual’s productivity in household production, conditional on other individual and household characteristics.42 Results from the Heckman selection model are presented in Table A8 of the Appendix. The coefficients on our instruments all have the expected signs and are highly statistically significant. It is worth mentioning that the main results of the paper do not depend on using the Heckman selection model to identify wage effects. The estimated effects of the FDW program on the labor force participation of Hong Kong women are very similar if we use a reduced form approach and include demographic determinants of market wage, instead of predicted wages, in the utility equations.43

4.4.2

Multivariable Probit Model

We estimate the multivariate probit model (4) using the simulated maximum likelihood (SML) implemented by the Geweke-Hajivassiliou-Keane algorithm.44 Table 4 presents results for three samples: (A) mothers aged 25 to 54 who live in homes with 3 or 4 rooms, (B) subset of (A) who live in subsidized sale flats, and (C) subset of women that reported not having moved in the past 5 years. We include the following economic and demographic variables in all equations: dummy for youngest child aged 0 to 5, number of children, dummy for a person aged 65 or older living in the household, household size, dummies for the education level of the woman, the woman’s age and age squared, and the log of husband’s income.45 The last column of each panel shows the estimate of the complementarity in the LFP and FDW decisions and its statistical significance. Our instrument is omitted from the equation of participating in the labor market and not hiring a FDW. Predicted wage is excluded from the equation of hiring a FDW and not participating in the labor market. Given that the utility equations are relative to the option of not participating in the labor market 42

Mroz (1987) uses age and education interactions as extra identifying instruments while Blundell et al. (1998) use interactions between cohort, education and year to identify the wage equation separately from the hours equation. In our case, these instruments did not work as well as we have a cross-section, and these instruments are meant to exploit variation through time in the returns to education (wage dispersion) and tax schedules. Wooldridge (2002) proposes membership in a union as a possible variable that affects wages but not the decision to participate in the labor. 43 Results available upon request. 44 This is estimated using the STATA command, mvprobit. 45 One concern with the 25 to 54 age range is that the number of children in a household could be an underestimate if children have moved out of the household (this is likely to affect older women in the sample). Results from the multivariate probit model are similar if we consider mothers in the 25-44 age range.

29

and not hiring a FDW, all but the predicted wage coefficient in the labor force participation equation (Column (1)) can be interpreted as differential effects of demographic and economic variables on the disutility of working. Therefore, our estimates suggest that women with more children, younger children, with either high or low education levels, from smaller households, and with richer husbands tend to have a higher disutility of working in the market. As expected, the utility of working outside of the household is increasing in the market wage. The magnitude of the coefficient suggests that a 10 percent increase in the wage increases the marginal probability of participating in the labor force (and not hiring a FDW) by 0.5 percent.46 The constant in the utility equation for hiring a FDW and not participating in the labor market (Column (2)) suggests that this alternative is only attractive for women with very wealthy husbands, with high education, and many children. For these women, the availability of FDWs has not changed their labor force participation decisions, but has allowed them to enjoy more leisure time or spend more quality time with their children. Our instrument is positive and statistically significant in the FDW equation. Column (3) in the panels present the coefficients in the equation of the utility of both hiring a FDW and participating in the labor market relative to staying at home and not hiring a FDW, and in Column (4) we compute the implied complementarity effect and its significance. The coefficient on predicted wage is positive and statistically significant as expected in Column (3), but its magnitude is much larger than that for the option of working in the market and not hiring a FDW in Column (1). We can think of two potential explanations. First, having a very high wage actually allows you to hire a FDW, given the restriction on the income level of the household. Second, women with a higher predicted wage are likely to work in occupations that value flexibility (lawyers and doctors, for example), and therefore, hiring a FDW increases the returns to working. The magnitude of the predicted wage coefficient suggests that a 10 percent increase in the predicted wage increases the marginal probability of choosing this option by 2.1 percent. Estimates from Column (4) also suggest that complementarity is especially large for women with very young kids and with several children. The strong interaction effect for those with a young child suggests that women view FDWs as a very good substitute for their time spent in childcare. Note that in Hong Kong most people work full-time, hence the child will spend many hours of her day with the FDW. One possibility as to why we observe a smaller complementarity effect for women with a high education (controlling for market wage) is that their time devoted to raising children may be less substitutable with that of the FDW’s. While the focus thus far has been on the effect of FDWs on female labor supply decisions due to 46

To estimate this elasticity, we computed, for each observation, the predicted marginal probabilities of each of the options before and after an increase in the predicted wage of 10 percent. We report the sample average increase in the marginal probability of working in the labor force and not hiring a FDW.

30

reductions in childcare costs, it is possible that FDWs are also hired to take care of elderly parents. The positive and (marginally) significant coefficient on the dummy for a person in the household older than 65 in the FDW utility equation suggests that FDWs are indeed hired to help care for the elderly. However, the coefficient is positive and non-significant in the LFP equation, indicating that the presence of elderly parents, unlike young children, does not prevent women from entering the labor market. Moreover, the implied complementarity effects for households with an older person aged 65+ is small and non-significant, suggesting that while FDWs are hired to care for the elderly, they do not appear to significantly affect women’s labor supply decisions.47 Panels (B) and (C) of Table 4 reproduce the exercise reported in panel (A) but restricting the sample to women for whom the number of rooms is less likely to have been endogenously chosen. Results are very similar for the sample of non-movers, both in direction and magnitude of the coefficients. For the sample of women living in subsidized sale flats all the coefficients are in the same direction as for the whole sample and are relatively similar in magnitude, only less precisely estimated, given the significantly smaller number of observations. We estimate a negative correlation coefficient between the unobserved determinants of labor force participation and the decision to hire a FDW, suggesting that a naive estimation will underestimate the extent to which the availability of FDWs has changed the labor supply decisions of women. In specifications not shown here, we ran linear models of labor force participation on having hired a FDW. The OLS coefficient is positive but significantly smaller than the 2SLS specification that uses the number of rooms as an instrument. Thus far, we have only discussed the direction of the effects of the observables on the utility of the different alternatives. To translate the magnitudes of the coefficients to willingness to pay in HK dollars, we need a way of converting utils to dollars. To do so, we use the coefficient of the variable having an extra room, combined with an estimate of the cost of getting the extra space to transform the interpretation of the coefficients from utils to dollars.48 Although this is clearly an imprecise approach, it allows us to check whether our estimates are sensible. Based on Census data on monthly mortgage payments, the monthly cost of having an extra room is between 2000 to 2500 Hong Kong dollars (260 to 325 US dollars). From the structural estimates in Table 4, having an extra room increases the utility from hiring a FDW by 0.25 to 0.3 utils, therefore a util corresponds roughly to 6500 to 10000 HK dollars. Applying these numbers to the difference between the utility levels of two women identical in all observable and unobservable characteristics except in the age of the youngest child, we find that women with a child aged 0 to 5 are willing to pay between 2700 to 47

Additionally, the large coefficient on the dummy for an older person 65+ in the FDW equation might result from the possibility that having an elderly person in the household is highly correlated with wealth/income, perhaps even more so than the age of the youngest child. 48 We have constrained the dummy for having 4 rooms to be equal in the two equations that it enters. Estimations of the model without imposing this restriction strongly suggest it is reasonable to do so.

31

6000 HK dollars (351 to 780 US dollars) more than women with an older child to have the option of hiring a FDW at the current prices.49 Is this a reasonable number? Suppose that the alternative is to hire a female low-skilled native; the average full time wage for a low education native is about 3800 HK dollars higher than the minimum wage for the FDW. This number is in the range of our estimate of the willingness to pay. Using the util to dollar conversion we can also compute the consumer surplus derived from the availability of FDWs, which we define as:

CS = M ax[0, M ax(Ud , Upd ) − M ax(Up, 0)] Table 5 presents our estimates. Our model implies that because of the program, mothers of children aged 0 to 17 in Hong Kong enjoy an average monthly consumer surplus of between 473 to 728 HKD (62 to 94 US$). As observed, mothers of younger children have an average consumer surplus twice as large as the consumer surplus for mothers of school age children. The results also suggest that the program has disproportionately benefited highly educated women.

4.5

Comparing Macro and Micro Estimates

To compare the macro and micro estimates, we use the structural model to simulate the program effect on the labor force participation of mothers. We simulate the optimal labor supply choices when the relative price of hiring a FDW is set at the early 1980s level.50 We do this separately for mothers of younger and older children to construct a difference-in-difference estimator comparable to the macro estimates. Table 6 presents the results. Our micro difference-in-difference estimates suggest that the decline in the relative price of hiring a FDW from its level in 1980s to its level today resulted in an increase in the relative labor force participation of younger vs. older children of 12.4 to 13.1 percentage points. As with the macro approach, the labor supply change comes mostly from mothers with medium to high levels of education. While we acknowledge that both the macro and micro approaches have important limitations and are based on relatively strong identification assumptions, the similarity of the estimates suggests that our estimates of the effect of the FDW program on female labor supply are fairly robust and reliable. This similarity is especially noteworthy given that the macro and micro approaches use 49

Chosen to guarantee a positive willingness to pay. More specifically, what we do is the following: we predict the individual level utilities for each alternative at the 2001 minimum wage using our estimates from Table 4 and random draws from a multivariate standardized normal distribution with correlation coefficient σ b. We take these predicted utilities and subtract the difference in utils between the 1981 and 2001 cost of hiring a FDW. We calculate how many women will have chosen to work at the 2001 minimum wage level and how many at the much higher 1981 level. 50

32

very different sources of variation for identification. The numbers in Table 6 are also useful to calculate the elasticity of labor force participation with respect to the price of FDWs implied by our structural model. As observed in the table, a 75 percent decrease in the relative price of hiring a FDW (from its 1981 level to its 2001 level) increased the probability of mothers of young children participating in the labor force by 56 percent, an implied elasticity of -0.74. Our estimates imply an elasticity larger in absolute value but not too far from well identified elasticities for the US and Canada, which gravitate around -0.35 (Baker et al, 2008). Note that the elasticities are not perfectly comparable as the services offered by FDWs differ from regular child care centers in the US and Canada. Daycare centers do not perform other domestic tasks (e.g. cooking and cleaning) and are only open for limited hours. This could be a possible reason for the higher elasticity that we obtain.

5

Conclusion

The outsourcing of household production to temporary foreign domestic helpers is a distinctive feature of many newly industrialized nations. Moreover, this form of migration is also becoming increasingly prevalent in some developed countries as a result of demographic changes and increasing demand for household services as women seek to enter the labor market. In this paper, we find that temporary foreign domestic helper policies significantly increased female labor force participation rates in Hong Kong especially for mothers of young children. Reducedform estimates from comparing labor force participation rates over time in Hong Kong versus Taiwan and simulations from a structural model of female labor supply imply that the program raised the labor force participation of mothers of young children in Hong Kong by between 8 to 13 percentage points relative to mothers of older children. These labor supply effects are concentrated among medium and highly skilled women, consistent with the fact that these women face higher opportunity costs of household production. Moreover, the program has increased the welfare of women in Hong Kong significantly, especially for mothers of younger children and mothers with high levels of education. The results from the structural model suggest that mothers regard FDWs as a good substitute for their time spent in household production. The influx of domestic migrant workers is likely to have different economic implications on the host country labor market as compared to conventional low-skilled migrants. Since these workers substitute for household production, they free up native women to take up employment in the labor market and potentially allow them to enter more demanding occupations. That we observe such large effects on labor supply decisions in response to the decrease in childcare costs as a result of the FDW program also suggests that at least part of the differences in labor market outcomes of

33

men and women can be attributed to constraints that women face in juggling their dual roles in the household and labor market. These results suggest that FDW programs can have important policy ramifications for encouraging women to enter the labor market and to bridge the gender gap. Nevertheless, such a policy is likely to raise important ethical and political economy considerations that may outweigh the potential benefits. A full discussion of the viability of such temporary foreign domestic worker programs outside the Hong Kong context is likely to be country-specific and is beyond the scope of this paper.

References [1] Anderson, Bridget (2001), “Reproductive Labor and Migration”.Paper given at the Sixth Metropolis Conference, Rotterdam, 26-30. WPTC-02-01. [2] Baker, Michael, Jonathan Gruber and Kevin Milligan (2008), “Universal Child Care, Maternal Labor Supply, and Family Well-Being”. Journal of Political Economy. Vol. 116(4), August, pp. 709-745. [3] Blau, David (2003), “Childcare Subsidy Programs,” in Robert Moffitt (ed.) Means-Tested Transfer Programs in the U.S. Chicago: University of Chicago Press. [4] Blau, David and Janet Currie (2006), “Preschool, Day Care, and After-School Care: Who’s Minding the Kids?” Handbook of the Economics of Education, Elsevier. [5] Blau, David and Philip Robins (1988), “Child-Care Costs and Family Labor Supply”. The Review of Economics and Statistics. Vol. 70, pp. 374-381. [6] Blundell, Richard, Alan Duncan and Costas Meghir (1998), “Estimating Labor Supply Responses Using Tax Reforms.” Econometrica Vol. 66, No. 4, pp. 827-861. [7] Browning, Martin, and Costas Meghir (1991), “The Effects of Male and Female Labor Supply on Commodity Demands.” Econometrica Vol. 59, No. 4, pp. 925-951. [8] Cascio, Elizabeth (2009), “Maternal Labor Supply and the Introduction of Kindergartens into American Public Schools.” The Journal of Human Resources, Vol. 44, No. 1, pp. 140-170. [9] Chan, Annie (2005), “Live-in Foreign Domestic Workers and their Impact on Hong Kong’s Middle Class Families.” Journal of Family and Economic Issues. Vol. 26, No. 4, pp. 509-528. [10] Chan, Annie (2006), “The Effects of Full-Time Domestic Workers on Married Women’s Economic Activity Status in Hong Kong, 1981-2001”. International Sociology. Vol. 21, No. 1, pp.133-159. 34

[11] Chiu, Stephen W.K. (2004), “Recent Trends in Migration Movements and Policies in Asia: Hong Kong Region Report”. Paper prepared for the “Workshop on International Migration and Labour Markets in Asia,” Japan Institute of Labour. [12] Connelly, Rachel (1992), “The Effects of Child Care Costs on Married Women’s Labor Force Participation”. Review of Economics and Statistics. Vol. 74, pp. 83-90. [13] Cortes, Patricia and Jose Tessada (2011), “Low-Skilled Immigration and the Labor Supply of Highly Skilled Women”. American Economic Journal: Applied Economics. Vol. 3, No. 3, pp. 88-123. [14] Crittenden, Anna (2001), “The Price of Motherhood: Why is the Most Important Job in the World is Still the Least Valued”. Henry Holt and Company, LLC. New York, New York. [15] Farre, Lindia, Libertad Gonzalez, and Francesc Ortega (2011), “Immigration, Family Responsibilities and the Labor Supply of Skilled Native Women”. The B.E. Journal of Economic Analysis & Policy. Vol. 11: Issue 1 (Contributions), Article 34. [16] Gelbach, Jonah B. (2002), “Public Schooling for Young Children and Maternal Labor Supply”. American Economic Review 92(1) pp. 307-22. [17] Heckman, J., and Sedlacek, G. (1985), “Heterogeneity, Aggregation, and Market Wage Functions: An Empirical Model of Self-Selection in the Labor Market”, Journal of Political Economy, 93, 1077-1125. [18] Keane, Michael P. 1992. “A Note on Identification in the Multinomial Probit Model.” Journal of Business and Economic Statistics, 10(2): 193–200. [19] Kremer, Michael and Stanley Watt (2008), “The Globalization of Household Production.” mimeo, Harvard University. [20] Lan, Pei-Chia (2003), “Negotiating Social Boundaries and Private Zones: The Micropolitics of Employing Migrant Domestic Workers.” Social Problems, Vol. 50, No. 4, pp. 525-549. [21] Lee, Rance P.P., Jik-Joen Lee, Elena S.H. Yu, Shong-Gong Sun and William T. Liu (2000) “Living Arrangements and Elderly Care: The Case of Hong Kong,” in William T. Liu and Hal Kendig (eds.) Who Should Care for the Elderly? An East-West Value Divide. Singapore University Press. [22] Lefebvre, Pierre and Philip Merrigan (2008). “Child-Care Policy and the Labor Supply of Mothers with Young Children: A Natural Experiment from Canada.” Journal of Labor Economics, Vol. 25, No. 3, pp. 519-548.

35

[23] Lin, Wan-I and Yang, Shin-Yi (2009). “From Successful Family Planning to the Lowest Low Fertility Levels: Taiwan’s Dilemma.” Asian Social Work and Policy Review, Vol. 3, pp. 95-112. [24] Lui, Hon-Kwong and Wing Suen (2005). “The Shrinking Earnings Premium for University Graduates in Hong Kong: The Effect of Quantity or Quality?” Contemporary Economic Policy, Vol. 23, No. 2, pp. 242-254. [25] Lui, Hon-Kwong and Wing Suen (2010), “The Effects of Public Housing on Internal Mobility in Hong Kong.” Journal of Housing Economics, Vol. 20, No. 1, pp.15-29. [26] Moffit, Robert (1984), “Profiles of Fertility, Labour Supply and Wages of Married Women: A Complete Life-Cycle Model.” The Review of Economic Studies, Vol. 51, No. 2, pp. 263-278. [27] Mroz, Thomas A. (1987), “The Sensitivity of an Empirical Model of Married Women’s Hours of Work to Economic and Statistical Assumptions.” Econometrica, Vol. 55, No. 4, pp. 765-799. [28] Oishi, Nana (2005), “Women in motion: Globalization, State Policies, and Labor Migration.” Stanford University Press, Stanford, California. [29] Podmore, Chaney and David Cheney (1974), “Family Norms in a Rapidly Industrializing Society: Hong Kong.” Journal of Marriage and Family. Vol 36, No. 2, pp. 400-407. [30] Shaw, Kathryn (1989), “Life-Cycle Labor Supply with Human Capital Accumulation.” International Economic Review, Vol 30, No. 2, pp. 431-456. [31] Suen, Wing (1993), “Market-procured Housework: The Demand for Domestic Servants and Female Labor Supply.” Labor Economics. Vol 1, pp. 289-302. [32] Tam, Vicky C.W. (1999), “Foreign domestic Helpers in Hong Kong and Their Role in Childcare Provision,” in Janet Henshall Momsen (ed.) Gender, Migration and Domestic Service. Routledge, London. [33] Tam, Vicky C.W. (2001) “A Family Ecological Analysis of Child Care Use in Hong Kong”. Children & Society, Vol 15, pp. 181-192. [34] Vere, James (2005) “Education, Development and Wage Inequality: The Case of Taiwan,” Economic Development and Cultural Change. Vol 53, No. 3, pp. 711-735. [35] Wooldridge, Jeffrey M (2002) “Econometric Analysis of Cross Section and Panel Data”. The MIT Press, Cambridge, Massachusetts. [36] Yeoh, Brenda S. A., Huang, Shirlena and Gonzalez III, Joaquin (1999), “Migrant Female Domestic Workers: Debating the Economic, Social and Political Impacts in Singapore.” International Migration Review, 33(1): 114-136 36

[37] Ziliak, James P., and Thomas J. Kniesnerm (2005), “The Effect of Income Taxation on Consumption and Labor Supply.” Journal of Labor Economics, 23(4): 769–796.

37

Data Appendix [Not for Publication] This appendix describes in more detail the samples used for the construction of the figures in our paper “Outsourcing Household Production: Demand for Foreign Domestic Helpers and Native Labor Supply”.

Sample Selection 1. Sample for Macro Approach: The main regression sample comprises of native married females between the ages 25 to 54 who have at least one child aged 0 to 17. This includes all resident females (whether or not they were born in Hong Kong or Taiwan, respectively) but does not include female who are live-in domestic helpers. Table DA-1 shows the breakdown of the sample from the full data and the sample inclusion criteria. Table 1: Sample Restrictions for Macro Approach Hong Kong Taiwan Native women aged 25-54 Including only: Married Married with at least one child aged 6 to 17 Non-missing information on LFP

Total

1,001,686

407,382

1,409,068

737,656 448,218 448,218

331,667 205,256 205,244

1,069,323 653,474 653,462

2. Sample Restrictions for Figures 1, 2 and 5: The difference between the main sample (described boave) and the sample used to construct figures 1, 2 and 5 is that women without kids are also included and the sample consists of those without missing observations on wages and the number of hours worked. The sample breakdown is shown in Table DA-2. Table 2: Sample Restrictions for Figures 1, 2 and 5 Hong Kong Taiwan Native women aged 25-54 Including only: Married Non-missing info on wages and hours Working at least 35-plus hours last week

Total

1,001,686

407,382

1,409,068

737,656 342,653 290,743

331,667 109,606 102,888

1,069,323 452,259 393,631

3. Breakdown of observations in each Education and Youngest Child Age Group in Figure 3 38

Table 3: No. of Observations in each Education and Youngest Child Age Group in Hong Kong Overall High Educ Mid Educ Low Educ Youngest Child 0-5 Proportion of sample

158,148

15,810 0.10

110,544 0.70

31,794 0.20

Youngest Child 6-17 Proportion of sample

290,070

12,115 0.04

166,952 0.58

111,003 0.38

No Children 0-17 Proportion of sample

240,905

23,806 0.10

133,566 0.55

83,533 0.35

4. Sample restriction for Micro Approach: The main model sample comprises of married females between the ages 25 to 54 who have at least one child aged 0 to 17 and for whom information about husband’s wage is available. Sample is restricted to households living in places with 3 or 4 rooms and at least one bathroom. Table DA-4 shows the sample breakdown from the full data. Table 4: Sample Restrictions for Micro Approach 2001 Census 2006 Census Women aged 25-54 Including only: Married Married with at least one child aged 0 to 17 Non-missing information on husband’s wage 3 or 4 Rooms

39

Total

80,243

85,216

165,459

57,913 35,554 30,777 20,065

57,691 33,455 27,948 19,302

115,604 69,009 58,725 39,367

0.9

16000

0.8

14000

0.7

12000

0.6

10000

0.5

8000

0.4

6000

0.3

4000

0.2

2000

0.1

0

Deflated Average Wage

Deflated FDW Minimum Wage

2005

2003

2001

1999

1997

1995

1993

1991

1989

1986

1984

1982

1980

1978

1976

0

Relative price of FDW

Figure 2. Relative Wage of Native Married Women to Min Wage of FDWs by Education Level in Hong Kong 10 9 8 7 6 5 4 3 2 1

High Education

Mid Education

2005

2003

2001

1999

1997

1995

1993

1991

1989

1986

1984

1982

1980

1978

1976

0

Low Education

Source: Hong Kong data - 1976, 1981 HK Census and 1985-2006 General Household Survey. Notes for Figures 1 and 2: Sample includes all married women aged 25 to 54 in Hong Kong who are not foreign domestic helpers. Reported wages are monthly employment earnings for full-time employees (defined as those working 35-plus hours per week). Wages are deflated using the CPI based on 2000 Hong Kong Dollars. Low education is defined as having at most primary education, medium education as having more than primary education but less than college and high education is defined as having attended college or more. Details on the sample used to construct the figure can be found in the data appendix.

Min Wage / Avg Wage

18000

Native Average Wage/Min Wage

Constant HK$

Figure 1. Trends in FDW Min Wage and Native Married Women's Average Wage in Hong Kong

Figure 3. Share of women with a FDW, by Education Level and Age of Youngest Child in Hong Kong High Education

2005

2003

2001

1999

1997

1995

1993

1991

1989

1986

1984

1982

1980

1978

1976

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Medium Education

1997

1999

2001

2003

2005

1997

1999

2001

2003

2005

1995

1993

1991

1989

1986

1984

1982

1980

1978

1976

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Low Education

Youngest Child 0‐5

1995

1993

1991

1989

1986

1984

1982

1980

1978

1976

0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Youngest Child 6‐17

No Children 0‐17 Source: Hong Kong data - 1976, 1981 HK Census and 1985-2006 General Household Survey. Notes: Sample includes all married women aged 25 to 54 in Hong Kong who are not foreign domestic helpers. Low education is defined as having at most primary education, medium education as having more than primary education but less than college and high education is defined as having attended college or more. Details on the sample used to construct the figure can be found in the data appendix.

Figure 4. Evolution of the LFP of Mothers Due to Changes in the Distribution of Observables in Hong Kong and Taiwan 1 0.9 0.8

Predicted LFP

0.7 0.6 0.5 0.4 0.3 0.2 0.1

Hong Kong

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

1991

1990

1989

1987

1986

1985

1981

1978

0

Taiwan

Source: Hong Kong data - 1976, 1981 HK Census and 1985-2006 General Household Survey. Taiwan data – 1978-2006 Manpower Utilization Survey. Notes: Sample includes all native married women aged 25 to 54 in Hong Kong and Taiwan. The figure shows the predicted labor force participation of mothers in Taiwan and Hong Kong from a cross-sectional model using the following variables as explanatory variables: age dummies, education dummies, husband’s income percentile dummies, a dummy for youngest child aged 0 to 5, and a dummy for Hong Kong. The model is estimated using all available years.

Figure 5. Evolution of Married Native Women’s Wages in Hong Kong and Taiwan Hong Kong (1976-2006) 18000 16000

Constant HK$

14000 12000 10000 8000 6000 4000 2000

Average Wage

10th Percentile

2006

2004

2002

2000

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

1978

1976

0

Minimum Wage FDW

Taiwan (1976-2006) 35000 30000

20000 15000 10000 5000

Average Wage

2006

2004

2002

2000

1998

1996

1994

1992

1990

1988

1986

1984

1982

1980

0 1978

Constant TW$

25000

10th Percentile

Source: Hong Kong data - 1976, 1981 HK Census and 1985-2006 General Household Survey. Taiwan data – 1978-2006 Manpower Utilization Survey. Notes: Sample includes all native married women aged 25 to 54 in Hong Kong and Taiwan. Reported wages are monthly employment earnings for full-time employees (defined as those working 35-plus hours per week). Wages are deflated based on 2000 local currency units. The exchange rate between the Hong Kong and Taiwan dollar fluctuated between 3.4 and 4.4 TW dollars per HK dollar over this period. Details on the sample used to construct the figure can be found in the data appendix.

Figure 6. Labor Force Participation of Native Mothers by Age of Youngest Child in Hong Kong and Taiwan

Youngest Child 0‐5

Youngest Child 6‐17

Youngest Child 0‐5

Youngest Child 6‐17

Source for Figures 6 and 7: Hong Kong data - 1976, 1981 HK Census and 1985-2006 General Household Survey. Taiwan data – 1978-2006 Manpower Utilization Survey. Notes for Figures 6 and 7: Sample includes all native married women aged 25 to 54 with at least one child aged 0 to 17. The sample used to construct the figures is identical to the regression sample in Table 1. Low education is defined as having at most primary education, medium education as having more than primary education but less than college and high education is defined as having attended college or more. The same classification applies to both Hong Kong and Taiwan.

2005

2003

2001

1999

1997

0 1995

0

1993

0.1 1991

0.1

1989

0.2

1986

0.2

1984

0.3

1982

0.3

2005

0.4

2003

0.4

2001

0.5

1999

0.5

1997

0.6

1995

0.6

1993

0.7

1991

0.7

1989

0.8

1986

0.8

1984

0.9

1982

0.9

1980

1

1978

1

1980

Taiwan

1978

Hong Kong

Youngest Child 0‐5

1

Youngest Child 6‐17

1978

1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0

Youngest Child 0‐5

1997 1999

1997 1999 2001 2003 2005

1997 1999 2001 2003 2005

Youngest Child 6‐17

2005

2003

2001

1995

0

1995

0 1995

0.2

1993

0.2 1993

0.4

1993

0.4 1991

0.6

1991

0.6

1991

0.8

1989

0.8 1989

1

1989

Low Education

1986

0 1986

0

1986

0.2

1984

0.2

1984

0.4

1984

0.4 1982

0.6

1982

0.6

1982

0.8

1980

0.8

1980

1

1980

Medium Education

1978

1978

2005

2003

2001

1999

1997

1995

1993

1991

1989

1986

1984

1982

1980

1

1978

2005

2003

2001

1999

1997

1995

1993

1991

1989

1986

1984

1982

1980

1978

1

2005

2003

2001

1999

1997

1995

1993

1991

1989

1986

1984

1982

1980

1978

Figure 7. Labor Force Participation of Native Mothers by Age of Youngest Child and Education Level in Hong Kong and Taiwan High Education

Table 1. Triple Difference Estimation of the Effect of FDWs in the LFP of Mothers of Young Children vs. Mothers of Older Children, Taiwan vs. Hong Kong : 1978-2006 (Age 25-54)

(1) HK*Child05 (Base period 78-84)

Dependent Variable: Labor Force Participation A. All Women B. By Education Level (2) (3) (4) Low Mid High

-0.035*** (0.007) [0.042]

-0.042*** (0.004) [0.019]

-0.053*** (0.005) [0.023]

-0.062*** (0.011) [0.034]

-0.063*** (0.008)

-0.035*** (0.009)

-0.208*** (0.034)

HK*Child05*period 85-87

0.017 (0.012) [0.014]

0.043*** (0.012) [0.020]

0.034** (0.013) [0.018]

0.030* (0.018) [0.030]

0.000 (0.014)

0.038*** (0.013)

0.268*** (0.069)

HK*Child05*period 89-93

0.034* (0.018) [0.028]

0.056*** (0.015) [0.031]

0.064*** (0.017) [0.017]

0.040*** (0.014) [0.030]

-0.010 (0.012)

0.063*** (0.019)

0.200*** (0.042)

HK*Child05*period 94-98

0.110*** (0.009) [0.040]

0.114*** (0.005) [0.035]

0.135*** (0.008) [0.016]

0.076*** (0.012) [0.038]

0.008 (0.011)

0.125*** (0.011)

0.240*** (0.037)

HK*Child05*period 99-02

0.126*** (0.007) [0.058]

0.112*** (0.004) [0.020]

0.134*** (0.006) [0.006]

0.129*** (0.012) [0.036]

0.014 (0.021)

0.123*** (0.009)

0.277*** (0.035)

HK*Child05*period 03-06

0.099*** (0.011) [0.068]

0.076*** (0.010) [0.015]

0.085*** (0.012) [0.013]

0.116*** (0.016) [0.031]

0.034*** (0.012)

0.067*** (0.017)

0.255*** (0.034)

Year FE Year*HK FE Child05*Year FE

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Controls: Age-group, Education Age*Education*Year Education*Year*HK Age*Education*Year*HK

No No No No

Controls No No No

Controls* Year Yes No No

Controls* Year Yes Yes Yes

653,462

653,462

653,462

653,462

Observations

Controls*Y Controls*Y Controls* ear ear Year 217,127

397,429

38,906

Source: Hong Kong - Census 1976 and 1981, General Household Survey 1985-2006; Taiwan Manpower Utilization Survey 1978-2006 Notes: 1. Each column corresponds to a separate linear probability model with female labor force participation as the dependent variable. The sample is restricted to married women aged 25-54 who have at least one child aged 0 to 17. HK is a dummy variable if the respondent is from Hong Kong. Child05 is a dummy variable if the youngest child is aged 0 to 5. p85-87, p89-93, p94-98, p99-02, p03-06 are dummy variables that denote the time-period considered. 2. Controls include age-group dummies (in five year intervals) and three dummies for the education level of the woman. Controls*Year indicate that all the controls are interacted with year dummies. Age*Education*Year indicate the full set of interactions for age-group, education-level and year. In column (4), the full set of intereactions for Education*Year*HK and Age*Education*Year*HK are also included. 3. Standard errors in parenthesis are clustered at the country-year level and standard errors in square brackets are clustered at the country-education level. Reported standard errors are robust to heteroskedesticity. The asterisks report the significance levels of estimates clustered at the country-year level ***significant at 1%, **5%, *10%.

Table 2. Descriptive Statistics of Mothers Included in the Micro Approach Sample Sample All

Subsidized H.

Non-movers

Low Edu

Mid Edu

High Edu

Mean

Std.

Mean

Std.

Mean

Std.

Mean Std.

Mean

Std.

Mean

Std.

Has a FDW

0.13

0.33

0.10

0.30

0.11

0.31

0.03

0.16

0.20

0.40

0.43

0.50

Participates in the LF

0.54

0.50

0.55

0.50

0.54

0.50

0.44

0.50

0.64

0.48

0.74

0.44

Wage | working (HKD)

13177 11306

11556

8055.5

12636 10592

8287 6238 14785 10064 28269 18590

Age

39.35

6.00

39.73

5.86

40.81

5.57

40.38 6.08

38.41

5.76

37.47

5.48

Low-Education

0.51

0.50

0.55

0.50

0.54

0.50

Med-Education

0.42

0.49

0.42

0.49

0.41

0.49

High-Education

0.07

0.25

0.03

0.18

0.05

0.22

Hhld. Size

4.07

0.99

4.06

0.93

4.13

0.98

4.32

1.02

3.84

0.90

3.62

0.81

Dummy Child 0-5

0.32

0.46

0.28

0.45

0.22

0.42

0.21

0.41

0.39

0.49

0.62

0.49

Share of mothers of young child with FDW

0.24

0.43

0.23

0.42

0.25

0.43

0.06

0.24

0.30

0.46

0.51

0.50

Number of Children

1.81

0.79

1.81

0.77

1.89

0.80

2.06

0.84

1.58

0.65

1.43

0.596

Dummy for member 65+

0.11

0.32

0.13

0.33

0.12

0.32

0.11

0.32

0.12

0.32

0.09

0.29

10366

Husband's wage (HKD)

17357 15131

15403

Number of Obs.

39367

9606

16632 14022 13005 8893 19893 15908 34302 27131 23951

20088

16606

2673

Source: Hong Kong Census, 2001 and 2006. Notes: The sample is restricted to married women aged 25 to 54 with at least one child aged 0 to 17, who live in places with 3 or 4 rooms. Low education is defined as having at most primary education, medium education as having more than a primary education but less than a college degree, and high education as having a college degree or a graduate degree.

Table 3. Placebo Tests: Reduced form Dependent Variable: Labor Force Participation Married with Children 0-17 Married No Children 0-17 Low Edu, Husband's wage <10,000 (1) (2) (3) (4) (5) (6) OLS Probit - ME OLS Probit - ME OLS Probit - ME Dummy for 4 rooms

0.045*** (0.005)

0.048*** (0.006)

0.009 (0.006)

0.010 (0.007)

0.015 (0.014)

0.016 (0.015)

Demographic Controls

Yes

Yes

Yes

Yes

Yes

Yes

Year FE

Yes

Yes

Yes

Yes

Yes

Yes

Quarter Type FE

Yes

Yes

Yes

Yes

Yes

Yes

District FE

Yes

Yes

Yes

Yes

Yes

Yes

N. of Observations

39367

22756

7134

Source: Hong Kong Census, 2001 and 2006. Notes: 1. Each column corresponds to a separate regression with female labor force participation as the dependent variable. The sample is restricted to women aged 2554 with at least one child aged 0 to 17. The omitted category for the dummy for 4 rooms is staying in a place with 3 rooms. Demographic controls include age, age-squared, three dummies for educational attainment, household size, number of children, an indicator for the presence of children aged 0 to 5, an indicator for the presence of a live-in parent aged above 65 years and log spouse income. Additional controls include fixed effects for year (2001 or 2006), fixed effects for the type of living quarters and residential district. 2. Specification (6) in the text is estimated for three groups of women: (1) married women with children (main sample), (2) married women with no children and (3) low-educated mothers whose husband earns less than 10,000 HK dollars per month. 3. Probit-ME refers to the Marginal Effects obtained from the Probit model. Reported standard errors are robust to heteroskedesticity ***significant at 1%, **5%, *10%.

Table 4. Structural Model of Labor Supply and the Decision to hire a FDW A. All married women aged 25-54 (1) (2) (3) Interaction coeff. Only LFP Only DW Both (statistical sig.) Log(predicted wage) Youngest Child 0-5 High education Mid education Ln(husb. Wage) Age Age square Number of Children Household Size Dummy Person 65+ More than 3 rooms Constant

Correlation coefficient

No. Observations

0.144*** (0.035) -0.415*** 0.374*** (0.018) (0.046) 0.0582 0.361*** (0.057) (0.067) 0.119*** 0.192*** (0.029) (0.046) -0.373*** 0.440*** (0.012) (0.031) -0.0118 0.0326 (0.013) (0.033) 0.000187 -0.000418 (0.000) (0.000) -0.237*** 0.248*** (0.017) (0.055) 0.0916*** -0.155** (0.013) (0.050) 0.048 0.166* (0.027) (0.082) 0.329*** (0.015) 2.407*** -7.398*** (0.345) (0.696)

1.465*** (0.052) 0.494*** (0.023) -0.902*** (0.080) -0.231*** (0.045) 0.182*** (0.015) 0.0775*** (0.020) -0.00116*** (0.000) 0.325*** (0.031) -0.219*** (0.026) 0.301*** (0.045) 0.329*** (0.015) -17.22*** (0.540)

0.535 *** -1.321 *** -0.542 *** 0.115 *** 0.057 -0.001 0.314 *** -0.156 *** 0.087

-12.229 ***

-0.278*** (0.012) 39367

(1) Only LFP 0.340*** (0.086) -0.431*** (0.036) -0.171 (0.129) 0.00327 (0.057) -0.445*** (0.027) -0.0351 (0.026) 0.000455 (0.000) -0.269*** (0.037) 0.138*** (0.030) 0.0316 (0.056)

1.809* (0.852)

Sample B. Subsidized Sale Flats (2) (3) Interaction coeff. Only DW Both (statistical sig.)

(1) Only LFP

1.812*** (0.127) 0.549*** 0.586*** (0.111) (0.048) 0.21 -0.890*** (0.211) (0.171) 0.146 -0.208* (0.102) (0.087) 0.482*** 0.193*** (0.106) (0.038) 0.142 0.0836 (0.102) (0.045) -0.00178 -0.00116* (0.001) (0.001) 0.287 0.367*** (0.212) (0.070) -0.265 -0.270*** (0.192) (0.059) 0.394 0.306** (0.253) (0.096) 0.272*** 0.272*** (0.033) (0.033) -9.819*** -20.76*** (2.075) (1.303)

0.141** (0.048) -0.437*** 0.436*** (0.024) (0.062) 0.121 0.399*** (0.077) (0.094) 0.0975** 0.192** (0.038) (0.061) -0.390*** 0.462*** (0.015) (0.044) -0.000369 0.0546 (0.018) (0.050) 0.0000146 -0.000683 (0.000) (0.001) -0.239*** 0.333*** (0.023) (0.080) 0.105*** -0.275*** (0.019) (0.075) 0.0551 0.266* (0.036) (0.120) 0.333*** (0.021) 2.375*** -7.801*** (0.508) (1.083)

0.468 *** -0.929 *** -0.357 ** 0.156 -0.023 0.000 0.349 -0.143 -0.120

-12.750 ***

-0.302*** (0.031)

-0.310*** (0.016)

9606

23951

C. Nonmovers (2) (3) Interaction coeff. Only DW Both (statistical sig.) 1.591*** (0.073) 0.511*** (0.032) -1.040*** (0.110) -0.312*** (0.061) 0.176*** (0.021) 0.0725* (0.030) -0.00107** (0.000) 0.367*** (0.044) -0.264*** (0.039) 0.306*** (0.063) 0.333*** (0.021) -18.09*** (0.807)

0.512 *** -1.560 *** -0.602 *** 0.104 * 0.018 0.000 0.273 *** -0.094 -0.015

-12.664 ***

Source: Hong Kong Census, 2001 and 2006. Notes: 1. Each panel reports the estimates from a separate multivariate probit model (see text for description). In Panel (A), the sample is restricted to married women aged 25 to 54 with at least one child aged 0 to 17 who live in places with 3 or 4 rooms. The sample for Panel (B) is a subset of (A) who live in government subsidized sale flats. Panel (C) uses a subset of women in (A) who reported not having moved in the past 5 years. 2. Reported standard errors are robust to heteroskedesticity ***significant at 1%, **5%, *10%.

Table 5. Consumer Surplus Estimates from the FDW Program (in HKD) Mean

Std. Dev.

Max

All Women, age 25-54, with 3-4 rooms Consumer Surplus - lower bound Consumer Surplus - higher bound

473.03 727.74

1747.04 2687.76

22180.42 34123.71

Youngest Child, age 0-5 Consumer Surplus - lower bound Consumer Surplus - higher bound

984.04 1513.90

2512.85 3865.92

22180.42 34123.71

Youngest Child, age 6-17 Consumer Surplus - lower bound Consumer Surplus - higher bound

237.26 365.01

1170.93 1801.43

18505.95 28470.70

Low Education Consumer Surplus - lower bound Consumer Surplus - higher bound

78.72 121.11

616.77 948.88

19279.93 29661.43

Medium Education Consumer Surplus - lower bound Consumer Surplus - higher bound

731.75 1125.77

2118.85 3259.77

22180.42 34123.71

High Education Consumer Surplus - lower bound Consumer Surplus - higher bound

1829.06 2813.94

3283.72 5051.88

21024.67 32345.65

Source: Hong Kong Census, 2001 and 2006. Notes: 1. The consumer surplus estimates are based on simulations using the model estimated in Panel (A) of Table 4. The lower bound estimate of the consumer surplus uses the conversion 1 util=6500 HKD. The upper bound estimate of the consumer surplus uses the conversion 1 util=10,000 HKD. 2. Low education is defined as having at most primary education, medium education as having more than a primary education but less than a college degree, and high education as having a college degree or a graduate degree.

Table 6. Micro-approach Diffs-in-Diffs estimate of the LFP effect of the FDW program Upper bound

Lower Bound

Observed

Predicted wn at 2001 level

Predicted wn at 1981 level

Predicted wn at 1981 level

Child 0-5

0.547

0.530

0.340

0.348

Child 6-17

0.541

0.533

0.473

0.475

0.131

0.124

All

Diffs-in-Diffs Low Education Child 0-5

0.315

0.358

0.307

0.308

Child 6-17

0.473

0.462

0.446

0.446

0.035

0.034

Diffs-in-Diffs Medium Education Child 0-5

0.650

0.604

0.367

0.376

Child 6-17

0.627

0.627

0.514

0.516

0.123

0.116

Diffs-in-Diffs High Education Child 0-5

0.752

0.692

0.318

0.343

Child 6-17

0.732

0.685

0.488

0.501

0.177

0.166

Diffs-in-Diffs Source: Hong Kong Census, 2001 and 2006.

Notes: 1. The consumer surplus estimates are based on simulations using the model estimated in Panel (A) of Table 4. The lower bound estimate of the consumer surplus uses the conversion 1 util=6500 HKD. The upper bound estimate of the consumer surplus uses the conversion 1 util=10,000 HKD. 2. Low education is defined as having at most primary education, medium education as having more than a primary education but less than a college degree, and high education as having a college degree or a graduate degree.

Figure A1. Labor Force Participation of Mothers by Age of Youngest Child and Education Level in the United States (March CPS)

Youngest Child 6 to 17

2003

2005

2001

2003

2001

1999

1997

1995

1993

Youngest Child 0 to 5

2005

1999

1997

1995

1993

1991

1986

1984

1982

1980

1978

2005

2003

2001

0 1999

0 1997

0.2

1995

0.2

1993

0.4

1991

0.4

1989

0.6

1986

0.6

1984

0.8

1982

0.8

1980

1

1978

1

1989

Low Education

Medium Education

Youngest Child 0 to 5

1991

1986

1984

1982

1980

1978

2005

2003

2001

0 1999

0 1997

0.2

1995

0.2

1993

0.4

1991

0.4

1989

0.6

1986

0.6

1984

0.8

1982

0.8

1980

1

1978

1

1989

High Education

Overall

Youngest Child 6 to 17

Source: March Current Population Survey (1978 to 2006) Notes: Sample includes married women aged 25 to 54 with at least one child aged 0 to 17. Low Education is defined as having attained less than 12th grade, medium education is defined as more than 12th grade but less than a college degree, high education is defined as having a college degree or more.

APPENDIX Table A1. Descriptive Statistics for the Macro Approach: Married mothers aged 25-54 Period HK

1978-1981 difference (p-value) TW

HK

1985-1987 difference (p-value) TW

HK

1989-1993 difference (p-value) TW

Youngest child aged 0-5 Age 25-29 Age 30-34 Age 35-39 Age 40-44 Age 45-49 Age 50-54 Low Education Mid Education High Education

0.487 0.200 0.198 0.160 0.179 0.152 0.111 0.718 0.262 0.020

0.465 0.276 0.221 0.194 0.164 0.100 0.046 0.776 0.205 0.019

(0.000) (0.000) (0.000) (0.000) (0.002) (0.000) (0.000) (0.000) (0.000) (0.835)

0.458 0.173 0.269 0.248 0.142 0.103 0.065 0.561 0.413 0.026

0.457 0.254 0.293 0.215 0.134 0.076 0.027 0.611 0.359 0.031

(0.873) (0.000) (0.000) (0.000) (0.020) (0.000) (0.000) (0.000) (0.000) (0.004)

0.407 0.107 0.267 0.292 0.210 0.088 0.036 0.446 0.522 0.032

0.423 0.207 0.288 0.269 0.160 0.058 0.018 0.429 0.529 0.042

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.040) (0.000)

Labor Force Participation (LFP)

0.391

0.360

(0.000)

0.424

0.480

(0.000)

0.411

0.504

(0.000)

Number of Obs.

9348

17551

18151

29658

38760

47540

Period HK

1994-1998 difference (p-value) TW

HK

1999-2002 difference (p-value) TW

HK

2003-2006 difference (p-value) TW

Youngest child aged 0-5 Age 25-29 Age 30-34 Age 35-39 Age 40-44 Age 45-49 Age 50-54 Low Education Mid Education High Education

0.381 0.073 0.218 0.308 0.249 0.122 0.029 0.347 0.606 0.048

0.416 0.161 0.267 0.282 0.210 0.068 0.012 0.260 0.681 0.059

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

0.340 0.058 0.171 0.298 0.281 0.151 0.040 0.270 0.660 0.070

0.397 0.138 0.243 0.285 0.226 0.092 0.016 0.136 0.783 0.081

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

0.294 0.050 0.157 0.256 0.304 0.182 0.051 0.196 0.700 0.104

0.361 0.117 0.225 0.277 0.250 0.110 0.021 0.076 0.815 0.109

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.021)

Labor Force Participation (LFP)

0.443

0.552

(0.000)

0.488

0.581

(0.000)

0.532

0.609

(0.000)

144305

46476

127549

34003

110105

30016

Number of Obs.

Source: Hong Kong data - 1976, 1981 HK Census and 1985-2006 General Household Survey. Taiwan data – 1978-2006 Manpower Utilization Survey. Notes: 1. The sample is restricted to married women aged 25-54 who have at least one child aged 0 to 17. This sample is the same as the regression sample in Table 1 (total number of observations: 653, 492). Low education is defined as having at most primary education, medium education as having more than primary education but less than college and high education is defined as having attended college or more.

Table A2. Share of Married Women 25-54 who are Household Heads or the Spouse of Household Head Living with Mother-in-law or Mother

Taiwan Early period Late period 1993-2006 1993-1997 2003-2006

1993-2006

Hong Kong Early period Late period 1993-1997 2003-2006

USA 1990

2000

2007

All Education Levels

0.078

0.076

0.077

0.088

0.097

0.081

0.017

0.022

0.023

Low Education Mid Education High Education

0.070 0.083 0.078

0.073 0.079 0.070

0.061 0.082 0.079

0.065 0.100 0.082

0.079 0.111 0.075

0.050 0.091 0.087

0.024 0.017 0.015

0.031 0.022 0.020

0.031 0.024 0.021

0.774

0.779

0.769

0.860

0.869

0.857

0.976

0.968

0.971

Share of women who are household head or spouse of head

Source: Hong Kong 1993-2006 General Household Survey. Taiwan data – 1993-2006 Manpower Utilization Survey. USA data - 1990, 2000 Census and 2007 American Community Survey. Notes: The time period is restricted by the availability of the variable describing the relationship of household members to the household head in the Hong Kong General Household Survey. This variable is not available prior to 1993.

Table A3. Triple Difference Estimation of the Effect of FDWs in the LFP of Mothers of Young Children vs. Mothers of Older Children, Taiwan Cities vs. Hong Kong : 1978-2006 (Age 25-54) Dependent Variable: Labor Force Participation A. All Women (1) (2) (3) (4) HK*Child05 (Base period 78-84)

-0.068*** (0.019) [0.052]

-0.065*** (0.014) [0.026]

-0.082*** (0.016) [0.022]

-0.081*** (0.024) [0.060]

HK*Child05*period 85-87

-0.005 (0.020) [0.026]

0.019 (0.015) [0.024]

0.024 (0.017) [0.019]

0.020 (0.025) [0.034]

HK*Child05*period 89-93

0.013 (0.023) [0.040]

0.025 (0.018) [0.034]

0.045** (0.021) [0.016]

0.024 (0.026) [0.042]

HK*Child05*period 94-98

0.095*** (0.022) [0.057]

0.086*** (0.017) [0.040]

0.118*** (0.019) [0.013]

0.064** (0.027) [0.061]

HK*Child05*period 99-02

0.146*** (0.020) [0.069]

0.117*** (0.015) [0.034]

0.147*** (0.017) [0.013]

0.150*** (0.026) [0.067]

HK*Child05*period 03-06

0.111*** (0.022) [0.075]

0.076*** (0.018) [0.029]

0.097*** (0.020) [0.021]

0.120*** (0.029) [0.062]

Year FE Year*HK FE Child05*Year FE

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Controls: Age-group, Education Age*Education*Year Education*Year*HK Age*Education*Year*HK

No No No No

Controls No No No

Controls* Year Yes No No

Controls* Year Yes Yes Yes

513,435

513,435

513,435

513,435

Observations

Source: Hong Kong - Census 1976 and 1981, General Household Survey 1985-2006; Taiwan Manpower Utilization Survey 1978-2006 Notes: 1A. The Taiwanese sample is restricted to women who are residing in the following 7 cities - Taipei Municipality, Kaohsiung Municipality, Taichung City, Tainan City, Chiayi City, Keelung City and Hsinchu City. 1B. Each column corresponds to a separate linear probability model with female labor force participation as the dependent variable. The sample is restricted to married women aged 25-54 who have at least one child aged 0 to 17. HK is a dummy variable if the respondent is from Hong Kong. Child05 is a dummy variable if the youngest child is aged 0 to 5. p85-87, p8993, p94-98, p99-02, p03-06 are dummy variables that denote the time-period considered. 2. Controls include age-group dummies (in five year intervals) and three dummies for the education level of the woman Controls*Year indicate that all the controls are interacted with year dummies. Age*Education*Year indicate the full set of interactions for age-group, education-level and year. In column (4), the full set of intereactions for Education*Year*HK and Age*Education*Year*HK are also included. 3. Standard errors in parenthesis are clustered at the country-year-level and standard errors in square brackets are clustered at the country-education level. Reported standard errors are robust to heteroskedesticity. The asterisks report the significance levels of estimates clustered at the country-year level ***significant at 1%, **5%, *10%.

Table A4. Double Difference Estimation of the Effect of FDWs on the LFP of Mothers of Young Children vs. Mothers of Older Children in Hong Kong: 1978-2006 Dependent Variable: Labor Force Participation A. All Women B. By Education Level (1) (2) Low Mid High Child05 (Base period 78-84)

-0.137*** [0.022]

-0.139*** [0.023]

-0.149*** [0.021]

-0.117*** [0.039]

-0.074 [0.051]

Child05*period 85-87

0.009 [0.022]

0.023 [0.025]

-0.022 [0.024]

0.054 [0.040]

0.193* [0.100]

Child05*period 89-93

0.026 [0.024]

0.033 [0.024]

-0.021 [0.023]

0.051 [0.040]

0.122* [0.063]

Child05*period 94-98

0.080*** [0.022]

0.065** [0.024]

-0.032 [0.022]

0.084** [0.040]

0.070 [0.054]

Child05*period 99-02

0.117*** [0.022]

0.111*** [0.023]

-0.016 [0.022]

0.117*** [0.039]

0.113** [0.052]

Child05*period 03-06

0.093*** [0.023]

0.099*** [0.025]

-0.052** [0.023]

0.095** [0.041]

0.085 [0.051]

Yes

Yes

Yes

Yes

Yes

Controls: Age-group, Education Age*Education*Year

Controls No

Controls* Year Yes

Controls* Year -

Controls* Year -

Controls* Year -

Observations

448,218

448,218

142,797

277,496

27,925

Year FE

Source: 1976, 1981 HK Census and 1985-2006 General Household Survey. Notes: 1. Each column corresponds to a separate linear probability model with female labor force participation as the dependent variable. The sample is restricted to married women in Hong Kong aged 25-54 who have at least one child aged 0 to 17. Child05 is a dummy variable if the youngest child is aged 0 to 5. p85-87, p89-93, p94-98, p99-02, p03-06 are dummy variables that denote the time-period considered. 2. Controls include age-group dummies (in five year intervals), three dummies for the education level of the woman. Controls*Year indicate that all the controls are interacted with year dummies. In Column (2), the full set of interactions for age-group*education*year dummies are also included. 3. Standard errors in parenthesis are clustered at the year-level. Reported standard errors are robust to heteroskedesticity. ***significant at 1%, **5%, *10%.

Table A5. Descriptive Statistics by Number of Rooms in the Household 1

2

FDW

0.01 (0.09)

0.02 (0.16)

0.08 (0.27)

0.21 (0.42)

0.37 (0.51)

0.73 (0.68)

0.05 (0.22)

0.08 (0.27)

0.15 (0.36)

0.19 (0.40)

LFP

0.41 (0.49)

0.45 (0.50)

0.52 (0.50)

0.58 (0.49)

0.63 (0.48)

0.60 (0.49)

0.50 (0.50)

0.53 (0.50)

0.58 (0.49)

0.61 (0.49)

Number of Kids

1.70 (0.74)

1.69 (0.73)

1.78 (0.77)

1.87 (0.84)

1.81 (0.76)

1.91 (0.78)

1.60 (0.66)

1.76 (0.72)

1.88 (0.83)

1.86 (0.77)

High Education

0.013 (0.115)

0.023 (0.149)

0.044 (0.205)

0.108 (0.310)

0.19 (0.39)

0.36 (0.48)

0.026 (0.159)

0.026 (0.158)

0.048 (0.213)

0.046 (0.209)

Mid Education

0.247 (0.431)

0.301 (0.459)

0.386 (0.487)

0.481 (0.500)

0.54 (0.50)

0.49 (0.50)

0.386 (0.487)

0.389 (0.488)

0.461 (0.499)

0.525 (0.500)

Dummy for child under 6

0.32 (0.47)

0.33 (0.47)

0.30 (0.46)

0.34 (0.47)

0.37 (0.48)

0.38 (0.48)

0.28 (0.45)

0.26 (0.44)

0.32 (0.47)

0.36 (0.48)

Household size

3.84 (0.85)

3.85 (0.86)

3.99 (0.91)

4.20 (1.10)

4.14 (1.04)

4.38 (1.29)

3.68 (0.72)

3.93 (0.83)

4.29 (1.05)

4.32 (1.04)

10794.38 (7549.77)

11890.25 (8129.80)

0.02 (0.13)

0.10 (0.29)

0.25 (0.43)

0.24 (0.43)

0.22 (0.41)

0.04 (0.19)

Share of Sample*

0.04

0.08

0.42

0.25

0.14

0.06

N. of Observations

2305

4794

24591

14776

8416

3802

Household Income

Dummy for Subsidized Housing

3 4 (A) Full Sample

Number of Rooms 5 6+

14789.49 21629.89 32491.88 56041.53 (11248.44) (19239.81) (28702.46) (43928.88)

2

3 4 (B) Subsidized Sale Flats

5

13185.59 14506.64 16933.03 18451.93 (7616.65) (9286.58) (11831.81) (12171.39)

461

6058

3548

1841

Source: Hong Kong Census, 2001 and 2006. Notes: Sample in Panel (A) includes married women aged 25 to 54 with at least one child aged 0 to 17. Panel (B) is a subset of Panel (A) that includes only women who live in subsidized sale flats. Standard deviations are reported in parentheses.

Table A6. Placebo Tests: Robustness to Alternative Definitions of the Instrument Dependent Variable: Labor Force Participation Married with Children aged 0-17 OLS Instrument Obs. Instrument Obs. Instrument Obs.

0.045*** (0.005)

Married, No Children aged 0-17

Low Edu, Husband's wage <10,000

Probit ME* Probit - ME Probit - ME OLS OLS Instrument: Dummy for 4 rooms; Sample: 3 or 4 rooms 0.048*** 0.009 0.010 0.015 0.016 (0.006) (0.006) (0.007) (0.014) (0.015)

39367

22756

7134

Instrument: Dummy for 4 or more rooms; Sample: 3, 4 or 5 rooms 0.0514*** 0.0553*** 0.0073 0.0078 0.0246 0.0263 (0.005) (0.005) (0.006) (0.006) (0.013) (0.014) 47783 0.0288*** (0.007)

27026

7619

Instrument: Dummy for 5 rooms; Sample: 4 or 5 rooms 0.0307*** -0.001 -0.002 0.0538* 0.0567** (0.007) (0.008) (0.009) (0.026) (0.027)

23192

13111

2322

Demographic Controls

Yes

Yes

Yes

Yes

Yes

Yes

Year FE

Yes

Yes

Yes

Yes

Yes

Yes

Quarter Type FE

Yes

Yes

Yes

Yes

Yes

Yes

District FE

Yes

Yes

Yes

Yes

Yes

Yes

Source: Hong Kong Census, 2001 and 2006. Notes: 1. The sample is restricted to women aged 25-54 with at least one child aged 0 to 17. The omitted category for the first panel is 3 rooms, for the second panel is 3 rooms and for the third panel is 4 rooms. Demographic controls include age, agesquared, three dummies for educational attainment, household size, number of children, an indicator for the presence of children aged 0 to 5, an indicator for the presence of a live-in parent aged above 65 years and log spouse income. Additional controls include fixed effects for year (2001 or 2006), fixed effects for the type of living quarters and residential district. 2. Specification (6) in the text is estimated for three groups of women: (1) married women with children (main sample), (2) married women with no children and (3) low-educated mothers whose husband earns less than 10,000 HK dollars per month. 3. Probit-ME refers to the Marginal Effects obtained from the Probit model. Reported standard errors are robust to heteroskedesticity ***significant at 1%, **5%, *10%.

Table A7. Having Moved and the Likelihood of Hiring a FDW

OLS

Dep. Var: Dummy for FDW at Home All Only child aged 0-5 Probit - ME OLS Probit - ME

Moved in past 5 years

0.0001 (0.0051)

-0.0014 (0.0032)

0.0007 (0.0168)

0.0004 (0.0171)

Demographic controls

Yes

Yes

Yes

Yes

Quarter Type FE

Yes

Yes

Yes

Yes

District FE

Yes

Yes

Yes

Yes

Year FE

Yes

Yes

Yes

Yes

No. of Observations

29270

3282

Source: Hong Kong Census, 2001 and 2006. Notes: 1. The main sample is restricted to women aged 25 to 54 with at least one child aged 0 to 17. The last two columns restrict the sample to women with only one child aged 0-5. Moved in past 5 years is a dummy variable that indicates if the household reported moving at least once in the past five years (only 70% of the sample answered this question). Demographic controls include age, age-squared, three dummies for educational attainment and spouse educational attainment, household size and indicators for the presence of children aged 0 to 5 and 6 to 17, the presence of a live-in parent aged above 65 years and log spouse income. Additional controls include fixed effects for year (2001 or 2006), fixed effects for the type of living quarters and residential district. 2. Probit-ME refers to the Marginal Effects obtained from the Probit model. Reported standard errors are robust to heteroskedesticity ***significant at 1%, **5%, *10%.

Table A8. Heckman Selection Model (A) All women aged 25-54 Log Wage Eq. Selection Eq. Constant

7.077*** (0.162)

Youngest Child 0-5 Dummy Person 65+ Ln(husb. Wage) High education Mid education Age Age square Number of Children Household Size Speaks English No. Observations

1.113*** (0.018) 0.522** (0.011) 0.073*** (0.008) -0.0009*** (0.0001) -0.110*** (0.009) 0.002 (0.006) 0.471*** (0.011) 39367

1.819*** (0.261) -0.177*** (0.017) 0.124*** (0.025) -0.355*** (0.011) 0.783*** (0.031) 0.385*** (0.016) 0.078*** (0.012) -0.0010*** (0.0002) -0.193*** (0.015) 0.032*** (0.013) 0.450*** (0.016)

Sample (B) Subsidized Sale Flats Log Wage Eq. Selection Eq. 7.734*** (0.301)

0.959*** (0.041) 0.412*** (0.020) 0.053*** (0.015) -0.0008*** (0.0002) -0.101*** (0.017) 0.021* (0.012) 0.376*** (0.020) 9606

2.972*** (0.574) -0.122*** (0.037) 0.161*** (0.056) -0.395*** (0.026) 0.782*** (0.086) 0.320*** (0.032) 0.036 (0.026) -0.0005 (0.0003) -0.229*** (0.035) 0.085*** (0.030) 0.487*** (0.033)

(C) Nonmovers Log Wage Eq. Selection Eq. 7.458*** (0.239)

1.116*** (0.025) 0.514*** (0.014) 0.055*** (0.012) -0.007*** (0.0001) -0.124*** (0.012) 0.013 (0.009) 0.445*** (0.014)

2.282*** (0.389) -0.156*** (0.023) 0.129*** (0.034) -0.376*** (0.014) 0.860*** (0.045) 0.349*** (0.020) 0.065*** (0.018) -0.0009*** (0.0002) -0.199*** (0.021) 0.042** (0.018) 0.422*** (0.021)

23951

Source: Hong Kong Census, 2001 and 2006. Notes: 1. Each panel reports estimates from the Heckman selection model as described in the text. In Panel (A), the sample is restricted to married women aged 25 to 54 with at least one child aged 0 to 17 who live in places with 3 or 4 rooms. The sample for Panel (B) is a subset of (A) who live in government subsidized sale flats. Panel (C) uses a subset of women in (A) who reported not having moved in the past 5 years. 2. Reported standard errors are robust to heteroskedesticity ***significant at 1%, **5%, *10%.

Table A9. Micro-approach Diffs-in-Diffs estimate of the LFP effect of the FDW program: Robustness checks Upper bound Lower Bound Observed Predicted Predicted Predicted wn at 2001 level wn at 1981 level wn at 1981 level Baseline (from Table 6) Child 0-5

0.547

0.530

0.340

0.348

Child 6-17

0.541

0.533

0.473

0.475

0.131

0.124

Diffs-in-Diffs Variations in the instrument and sample: Instrument: Dummy for 4 rooms; Sample: 3, 4 or 5 rooms Child 0-5

0.571

0.546

0.319

0.325

Child 6-17

0.551

0.542

0.460

0.462

0.146

0.140

Diffs-in-Diffs Instrument: Dummy for 5 rooms; Sample: 4 or 5 rooms Child 0-5

0.636

0.602

0.403

0.435

Child 6-17

0.579

0.571

0.469

0.482

0.097

0.078

Diffs-in-Diffs Reduced form model Child 0-5

0.547

0.530

0.337

0.342

Child 6-17

0.541

0.532

0.472

0.473

0.133

0.128

Diffs-in-Diffs Source: Hong Kong Census, 2001 and 2006. Notes:

1. The baseline consumer surplus estimates are repeated from Table 6. The sample for the baseline estimates include married women aged 25 to 54 with at least one child aged 0 to 17 who live in places with 3 or 4 rooms. The remaining panels reports consumer surplus estimates based on variations in the instrument and sample. The last panel reports consumer surplus estimates from the multivariate probit model that uses a reduced form model for log wages instead of the heckman selection model (see text for details). The lower bound estimate of the consumer surplus uses the conversion 1 util=6500 HKD. The upper bound estimate of the consumer surplus uses the conversion 1 util=10,000 HKD. 2. Low education is defined as having at most primary education, medium education as having more than a primary education but less than a college degree, and high education as having a college degree or a graduate degree.