Outsourcing Household Production: Effects of Foreign Domestic Helpers on Native Labor Supply in Hong Kong

Outsourcing Household Production: Effects of Foreign Domestic Helpers on Native Labor Supply in Hong Kong Patricia Cortes Jessica Pan University of Chicago Graduate School of Business October 31, 2008 Abstract This paper estimates the causal impact of hiring a foreign domestic helper on the labor force participation and occupational choice decisions of native females in Hong Kong. We present cross country evidence comparing Hong Kong and Taiwan that is consistent with the hypothesis that the foreign worker program led to differentially increasing female labor force participation among females with young children. Using cross-sectional data, our IV estimates show that hiring a maid increases the likelihood of working by approximately 20 percentage points. Women with domestic helpers are also more likely to be engaged in occupations that are less flexible and more time intensive. Taken together, these two pieces of evidence suggest that part of the increase in female employment in Hong Kong over the last two decades can be attributed to the foreign domestic helper program, which provided native women with a means of outsourcing household production. We are grateful to David Autor, Marianne Bertrand, Divya Mathur and seminar participants at the University of Hong Kong 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. 1

1 Introduction In the past decade, there has been a surge in the number of low-skilled workers from developing countries such as the Philippines, Indonesia, Thailand and Sri Lanka migrating to the new rich countries as domestic helpers. In Hong Kong, the proportion of households hiring at least one foreign domestic helper increased from less than 2% in 1986 to close to 8% in 2006 (table 1). Among households with young children, more than one in three hired at least one foreign domestic helper. Similarly, 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). This period has also seen a concurrent rise in the labor force participation rates of married women in many of the newly industrialized countries that have adopted some of the most liberal policies towards these foreign domestic helpers. There is a long line of research that documents the impact of low-skilled immigrants on native wage and displacement outcomes in host countries. However, the recent influx of low-skilled temporary migrants to the household sector has received considerably less attention. A recent paper by Kremer and Watt (2008) points out that the economic implications of the temporary migration of private household workers can differ substantially from that of conventional low-skilled migrants. Since these temporary domestic helpers generally substitute for household production, they tend to induce greater labor market participation among highly skilled individuals, especially women. In this paper, we build on the analysis by Kremer and Watt (2008) by assessing whether the expansion of the foreign domestic helper program in Hong Kong and the concurrent increase in female labor force participation is causally related. Hong Kong is a particularly relevant case study as the number of foreign maids has increased enormously in the last two decades as a result of a policy change in the early 1980s that enabled the systematic importation of these domestic migrants. Despite the prevalence of these programs in many countries, there are few empirical studies of the labor market implications of the influx of temporary low-skilled domestic migrant workers to developed countries. A recent paper by Cortes and Tessada (2007) documents that the decrease in prices of domestic services, as a result of the growth in low-skilled immigrants, has a relatively large impact on the time-use of high-skilled American women. To our knowledge, only two papers have studied similar questions in the Hong Kong context. Suen (1993) and Chan (2006) both provide some evidence that hiring a live-in domestic worker is associated with a higher likelihood of female labor force participation. Neither paper addresses the causal question of whether hiring a domestic helper has an impact on female labor supply. The main contribution of this paper is to seriously tackle the causality issue and, more broadly, to evaluate the extent to which the outsourcing of household production has impacted the labor force participation of native females in Hong Kong. We present two main pieces of evidence that, when taken together, suggest that 2

the widespread availability of foreign domestic helpers in Hong Kong led to an increase in female labor force participation and enabled women to take on more demanding occupations. Our first strategy exploits the variation in program adoption in Hong Kong and Taiwan and differences in the demand for household services by households with relatively younger or older children to form difference-in-difference estimates of the impact of the foreign domestic helper policy on female employment rates across the two countries. For our second strategy, we take a micro approach, utilizing pooled cross-sectional data from the 2001-2006 Hong Kong population census to estimate whether hiring a foreign domestic helper impacts the likelihood of female employment and occupational choice. To account for the endogeneity of foreign domestic helper hire, we propose instrumenting foreign domestic helper hire with the number of rooms in a household, controlling for proxies of household wealth. This instrument is motivated by the fact that most Hong Kong households are relatively space constrained and the intuition behind this instrument is that conditional on household wealth, the number of rooms is likely to have an impact on the propensity of maid hire, but should be uncorrelated to a woman s unobserved propensity of work. Thus, if we are able to control adequately for determinants of housing choice, this may serve as a valid instrument. Using the time-series difference-in-difference approach we find that, on the aggregate, the foreign worker program in Hong Kong is associated with a 10 percentage point increase in employment of females with a young child, compared to females with a relatively older child over the period from 1976-2006. Consistent with the view that high-skilled 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 more educated females. Perhaps interestingly, our micro-level estimates are consistently smaller than our macro-level estimates. In particular, our instrumental variable estimates indicate that hiring a domestic helper is associated with a 20-30 percentage point increase in the likelihood of employment (or an average increase in labor force participation of about 2-6 percentage points). We also find that women with domestic helpers are also significantly more likely to be employed in occupations with higher average work hours and that are less female-intensive. To the extent that concerns about the endogeneity of our instrument would imply that these individual level estimates are upward-biased, the discrepancy between our micro and macro level estimates suggests the presence of a social multiplier. Our empirical estimates are consistent with the idea that social interactions may be important in understanding the demand for employment among females. We conjecture that the introduction of foreign domestic helpers may have fundamentally altered the role of the woman in the family. By enabling more women with young children to enter the workforce, this policy may have reduced both the economic and cultural barriers for later cohorts of women to enter the workforce, leading to an even larger increase in the demand for female employment over time. In the presence of these positive social interactions, the foreign worker program may have provided the initial stimulus 3

for the rapid growth in female labor force participation that we observe in Hong Kong. This is particularly important from a policy point of view as it suggests that increasing the access to affordable and flexible childcare options can have potentially large effects on increasing female labor force participation and female representation in traditionally male-dominated occupations. The availability of inexpensive domestic help might have an impact on fertility and education choices of women. We leave the study of these potentially important and interesting effects for future work. We do acknolewdge that if either of these forces are at work, we could have compositional changes in the population under consideration. However, we expect the compositional effects to be of second order for the interpretation of our results. 2 Data Description We use the 5% sample data from the 1996-2006 Hong Kong Population census and by-census and the 1% sample data from 1976-1991. We will also make use of the General Household Survey (GHS) from 1985-2007. For the cross-country analysis, we will combine the Hong Kong data with the 1978-2007 Taiwan Manpower Utilization Survey (MUS). The Hong Kong population census and by-census are conducted every five years while the GHS is conducted quarterly. Both data sets provide a range of demographic information on all members of an enumerated household. The presence of live-in foreign domestic maids in a household is inferred from a variable that indicates whether the relation of the respondent to the household head is that of a live-in foreign domestic helper, chauffeur or gardener and the nationality of the respondent. While it is very likely that almost all workers in this category are foreign domestic helpers, to ensure that this is the case, we only include the female workers in this category. The Taiwan Manpower Utilization Survey (MUS) is a household survey conducted yearly by Taiwan s Directorate-General of Budget, Accounting, and Statistics (DGBAS). It provides labor force information for a representative sample of about 60,000 individuals over the age of 15. The sample covers the civilian population in Taiwan and excludes foreign workers who do not have citizenship. 3 Background 3.1 Foreign domestic helper policy in Hong Kong The foreign domestic helper (FDH) program was first introduced in Hong Kong in the early 1980s. To provide a sense of how rapidly this program has expanded, table 1 presents the number FDHs in Hong Kong from 1981-2006 as a proportion of the total labor force and domestic households. 4

Over the twenty year period, the size of the FDH population has increased twenty-five-fold, with foreign domestic helpers comprising 6.8% of the total labor force in 2003 and almost one in ten households employing a domestic helper. These statistics suggest that the outsourcing of household production to temporary migrant workers has become a distinctive facet of Hong Kong s labor market and social landscape. Compared to other receiving countries, Hong Kong has a relatively liberal policy toward these foreign workers. 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. 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 fulfil the requisite conditions set out in the standard contract (Chan, 2006). The main restriction is 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. Most of the foreign workers are drawn from the Philippines, Indonesia and Thailand. Figure 1 shows the stock of foreign domestic workers in Hong Kong by nationality. In the early 1990s, the FDH market was dominated by Filipinos, but their market share started eroding toward the end of the 1990s as lower-cost Indonesian maids entered into the domestic helper market. By 2006, there were roughly equal numbers of Filipinos and Indonesians working in Hong Kong as domestic helpers. Figure 2 plots the proportion of households with foreign and local domestic helpers from 1976-2006. In 1976, the vast majority of domestic helpers were local domestic helpers, but their share rapidly declined with the introduction of the foreign worker program in the early 1980s. Table 2 presents some descriptive statistics on the background characteristics of these foreign domestic helpers. On average, Filipino domestic helpers are older, tend to have more work experience and are more educated than their Indonesian counterparts. This could, in part, account for the relatively higher wage received by Filipino domestic helpers. Strikingly, 20% of the Filipino maids have completed a college degree, compared to 12% of Hong Kong-ers and a mere 3% of Indonesians. While Filipinos have an English speaking advantage, interestingly, close to 90% of Indonesian maids report speaking Cantonese. 3.2 Link to female labor force participation rates? To motivate the link between labor force participation of women and the employment of foreign domestic helpers, figure 3 displays the proportion of married females aged 20-55 with a live-in foreign domestic helper by the age range of their children. Consistent with the notion that households with younger children have a greater demand for household production, we observe that the consumption of these maid services has increased more rapidly for families with younger children. By 2006, more than 30% of households with at least one child aged less than five years had a maid. 5

Figure 4 displays the labor force participation rates of Hong Kong women from 1976-2006. Similar to trends observed in other industrialized countries, the employment rate of women has increased considerably over the last two decades. Most of this increase in employment rates came from the marked increase in workforce participation of married women as compared to unmarried women. These trends are of similar magnitude to those observed in the United States and Europe. Figure 5 further breaks down the increase in employment rates for females over the time period by the age range of children in the household. We plot the female employment rates of married females with no children, at least one child aged between 0 to 2, at least one child aged between 3 to 5, 6 to 12 and 13 to 18. From 1976-1986, the increase in employment rates appear to be relatively similar for all five groups (with a somewhat larger increase for women with no children), but this pattern changes markedly post-1986. In particular, we observe that while the employment rates for females without children and those with at least one child aged 13-18 remained relatively constant, the employment rates of those with children aged 0 to 5 increased markedly over the period from 1986-2006, such that by 2001, the employment rate of females with very young children (0 to 2) was actually on par with that of females with older children (13 to 18). Some of these trends reflect compositional changes, since the group with relatively young children are the younger cohorts of women who have higher female labor force participation rates than the older cohorts. Nevertheless, a comparison of figures 3 and 5 suggests that female employment grew fastest for households with the largest increase in consumption of foreign maid services. This is consistent with the notion that these foreign maids are a substitute for a woman s time in the household, potentially freeing up more native women to participate in the labor force. 3.3 Comparison to Taiwan While figures 3-5 provide some suggestive evidence of a possible link between the adoption of the foreign domestic helper policy and aggregate trends in female LFP in Hong Kong, they provide little evidence that the two phenomena are causally related. The ideal quasi-experiment would probably involve comparing female employment rates in regions that introduced a foreign domestic helper scheme to regions that did not, assuming that regions exogenously decide whether or not the implement such a scheme. Unfortunately, Hong Kong is a relatively small country and the policy was implemented at a national level, hence we cannot exploit geographical variation within Hong Kong. Looking outside Hong Kong, however, suggests that we can use Taiwan as a possible control group given the close proximity as well as economic and cultural similarity of the two countries. It is worth pointing out that while Taiwan does have a foreign domestic helper program, the magnitude and scope of the program is far smaller than that of Hong Kong s. Table 1 shows the size of the foreign domestic helper/caretaker program in Taiwan and Hong Kong. In 2001, foreign domestic helpers comprised approximately 1.1% of the labor force in Taiwan compared to 5.3% in 6

Hong Kong. Moreover, the requirements for hiring a foreign worker in Taiwan are fairly restrictive. 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. The official foreign domestic helper scheme began in 1992, and at it s peak in 1996, accounted for approximately 13,000 foreign workers. This program has since been scaled down 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 care 1. This is in sharp contrast to the program in Hong Kong, where household income is the only eligibility requirement. We will exploit differences in the ease of engaging a foreign domestic helper between Hong Kong and Taiwan as well as differences in the probability of domestic maid hire by child age structure to examine the impact of foreign domestic helper policies on female labor force participation rates. As shown in Figure 5, households with younger children are more likely to hire foreign domestic helpers and this group has also seen the most rapid increase in domestic helper hire. This is not surprising since we would expect households with younger children to be more sensitive to the price and availability of such services. The cross-country comparison allows us to use Taiwan as a control group to difference out group-specific trends in employment, while the comparison of females with older (at least one child aged 6 to 18) versus females with younger children (at least one child aged 0 to 5) allows us to difference out country-specific trends in female labor force participation over the time-period that affect women in both groups. As one might expect the female labor force participation rates of females with an older child and a younger child to change differentially across time, even within a country, we will compare differences in the growth rates of employment across the groups. We interpret the difference in the growth of female LFP of these two groups, adjusting for composition changes, as providing a measure of the impact of the foriegn worker policy on LFP rates in Hong Kong. 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 18 has evolved quite differently across the two countries post-1986. 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 post- 1986, such that by 2006, it actually exceeded that of women with older children. This is in stark 1 http://dhsc.evta.gov.tw/eng/applicant.html. 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 will 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. 7

contrast to Taiwan, where the growth in employment rates across the two groups of women remained virtually parallel over the entire sample time period from 1978-2006. In figure 7, we separately graph the trends in LFP rates of women in the two countries by education level. Perhaps strikingly, most of the catching-up in LFP of younger women in Hong Kong can be attributed to trends in the LFP of higher educated women. The employment trends of lower-educated females across the two groups of women appear to be mostly stable across our sample time period. This evidence is 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. Nevertheless, these graphs 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 females with older/younger children across time, these effects may confound the aggregate trends that we observe in the graph. In the next section, we will provide formal econometric evidence that adjusts for such composition effects as well as a series of placebo tests that suggest that the aggregate trends depicted in figures 6 and 7 are indeed consistent with a causal interpretation. 4 Formal econometric evidence 4.1 Cross-country approach: Hong Kong vs. Taiwan We estimate the regression analogue of figures 6 and 7, adjusting for relevant individual covariates such as age and education: Y ijgt = γ j + λ t + α g + β 1 D jgt + β 2 D jt + β 3 D gt + β 4 D jg + δx jgt + ɛ ijgt where i is the individual, j is the country, g is the group (whether female has older or younger child), t is the time period. Vector X jgt are individual-level controls for age and education. D represents the relevant indicator variables; D jgt = 1[HK = 1, Y oungkid = 1, t 1985, D jt = 1[HK = 1, t 1985], D gt = 1[Y oungkid = 1, t 1985] and D jg = 1[HK = 1, Y oungkid = 1]. A positive and significant β 1 will indicate that the difference in the growth in employment rates for women with younger versus older children is significantly larger in Hong Kong with the introduction of the foreign worker program as compared to Taiwan. If our identification assumptions are satisfied, we can interpret this as an estimate of the causal effect of the introduction of foreign domestic helpers on female labor force participation in Hong Kong. First, we present results that test the parallel trend assumption. The first column of table 3 shows that in the period prior to the FDH program, 1976-1981, the difference in female employment rates 8

for women with older/younger children across Hong Kong and Taiwan is slightly negative, although not significantly different from zero. This suggests that, if anything, pre-existing trends will bias our results slightly downward. Column 2 adds in controls for age and education and this has little effect on the interpretation of the estimates. In the third and fourth column, we consider the period from 1976-1985 and obtain virtually similar estimates. The main results are presented in panel (A1) of table 4. The first column of each time period is the raw, unadjusted, difference. The second column adjusts for an individual s age and education. The results indicate that the labor force participation rates of females with young children in Hong Kong started increasing differentially post-1985. Controlling for age and education, the diff-in-diff estimates indicate a 2.8 percentage point increase in LFP females with young children in Hong Kong relative to Taiwan from 1976-1990, a 6.9 percentage point increase from 1976-1995, an increase of 10 percentage points from 1976-2000 and an increase of 10.1 percentage points from 1976-2006. These estimates are large and economically significant. Moreover, looking across the two columns for each time period, it is reassuring that the inclusion of age and education covariates has little impact on the raw estimates, suggesting that our estimates do not merely reflect composition shifts in our treatment and control groups. 4.1.1 Do these effects vary by the education level of the female? Next, we test whether the increase in female labor force participation rates varies by the educational attainment of the women. We define low education as those with less than the equivalent of Grade 9 2 qualification in Hong Kong and Taiwan. Panels A2 and A3 in table 4 indicate that while the employment rates of lower educated females with young children in Hong Kong increase by 1.9 percentage points from 1976-2006 (non-significant at 10% level), more educated females experienced an 11 percentage point increase in employment over the same time period. Similar to the results for the overall sample, the placebo diff-in-diffs are supportive of the parallel trend assumption (results available on request). This suggests that most of the differential increase in LFP rates of married females with young children in Hong Kong was due to an increase in employment rates of more educated females. 4.1.2 Robustness check - adjusting for standard errors Since we only have two countries and two groups, there is a limit to which we are able to cluster our standard errors to allow for serial-correlation in our observations. To account for this issue, we also estimate the model above using group averages instead of micro data. For our raw estimates, 2 This is equivalent to Form 3 in Hong Kong and Junior High School in Taiwan. 9

this is equivalent to running weighted least squares on group-average outcomes using group size as weights. In order to include micro covariates into our group-level regressions,we follow the two-step procedure as suggested in Angrist and Pischke (2008). In the first step, we construct covariateadjusted group effects by regressing the individual level outcomes on a full set of group dummies (in our case 18 time periods x 2 countries x 2 groups = 92 groups), adjusting for age and education. In the second step, we regress the estimated group effects as the outcome variable in our differencein-difference set-up, weighting by the group size. These results are presented in the panel B1-B3 of table 4. The coefficients are virtually identical to those obtained using the micro-data, and although the standard errors are about 2-3 times larger, the coefficients from 1976-1995 onwards continue to be significant at the 10% level or greater for the overall sample. 4.2 Cross-sectional instrumental variables approach Next, we make use of the 2001 and 2006 Hong Kong census microdata to estimate the impact of hiring a foreign domestic helper on the decision to enter the labor force and occupational choice. The key outcomes that we are interested in are employment status and occupational choice. We use the following regression specification: y it = α + βf DH it + γx it + η t + θ k + ɛ it where i is an individual and t indicates if the individual was sampled in the 2001 or 2006 census. The individual covariates, X it include age, age-squared, dummies for educational attainment and spouse s educational attainment, household size and indicators for the presence of children aged 0 to 2, 3 to 5, 6 to 12 and 13 to 18 as well as the presence of a live-in parent aged above 60 years. In some specifications, we will include either log spouse income or the number of bathrooms as a proxy for household wealth. θ k is a set of fixed effects for the housing district and the type of residential quarters. We also include a dummy variable, η t, that indicates whether the individual was sampled in 2001 or 2006. Our sample comprises of married females aged 20-55 who are employed, unemployed or home-makers. We exclude individuals who are not seeking work, students, retirees, and the institutionalized population. In the presence of omitted variables, the regression specification above says little about causality between FDH hire and labor market outcomes. To the extent that women in households that hire maids have an unobserved propensity for market work, simple OLS estimates will be upward biased. As evident from the summary statistics (appendix table 1), hiring a domestic helper is concentrated among women who are more educated and have relatively high personal monthly earnings, suggesting that selection bias may account for part of the positive correlation between hiring a domestic helper and labor force outcomes. Unobserved household wealth will have the 10

effect of biasing the OLS estimates downward since a household s permanent income is positively correlated to foreign maid hire, but tends to reduce the likelihood of female employment. At the same time, it is also plausible that households that decide to hire maids may have an unobserved need for more housework. While some of these household differences can be taken into account through the introduction of relevant household composition controls such as the age structure of kids and the presence of live-in elderly parents, the presence of such unobservables could, in principle, lead to a downward bias in the observed OLS estimate. Hence, while our a priori expectation is for there to be an upward bias, it is not clear that this would necessarily be the case. Any hope of establishing a causal relationship depends crucially on the ability to identify some sources of exogenous variation that might plausibly affect the propensity of maid hire and at the same time is uncorrelated with the individual s propensity to work. 4.2.1 Proposed instrument The main instrument that we use is the number of rooms in the household. The motivation behind this instrument is the fact that space limitations in Hong Kong coupled with restrictions on lodging for domestic workers (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) imply that all else equal, a household living in a house with more rooms is more likely to hire a domestic worker. Hence, assuming we are able to control adequately for household wealth, we would not expect the number of rooms in the house to be correlated to an individual s unobserved work propensity. At the same time, the number of rooms in a household is likely to increase the likelihood of hiring a foreign domestic helper due to reduced space constraints. There are a number of concerns in using the number of rooms as an instrument. We discuss each of these concerns in detail. 4.2.2 Is the instrument endogenous? Issue 1: Moving concerns The first 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 first stage 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. While we may not be able to fully dispel this concern, we can explicitly test whether households that have moved in the last five years are more likely to hire domestic helpers. In table 5, we regress 11

the probability of hiring a foreign maid on whether the household moved in the last 5 years. In the first column, without controlling for household wealth, we do find that having moved in the past 5 years is correlated with a higher probability of having a domestic worker. The coefficient becomes smaller and is no longer statistically significant after including the husband s income (column 2). Nevertheless, this is a weak test as we do not observe the full sequence of moves prior to the first instance of maid hire. In addition to this, in columns 3 and 4, we also perform this test on a subset of households with an oldest child aged less than five. Given that the probability of maid hire increases substantially when a household has young children, this group of households are 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 not statistically significant under either specification, although this arises mainly because of larger standard errors. We interpret these results as suggesting that families are not moving in large numbers to accommodate a foreign domestic worker; we do acknowledge, however, these tests cannot completely dispel concerns about the moving behavior of households. To complement the evidence in table 5, 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. Issue 2: Omitted household wealth Our instrumentation strategy is only valid if we are able to control adequately for determinants of household wealth that are correlated with the number of rooms. Since richer households tend to have a larger number of rooms and are also more likely to hire a foreign domestic helper, omitting household wealth would tend to lead to upward biased estimates of the first-stage. On the other hand, failing to include household wealth would tend to lead to a downward bias in the reduced form relationship since the probability of female employment is decreasing in permanent income. Hence, failing to account for differences in household wealth would tend to lead to a downward bias in our IV results. To circumvent this problem, we make use of two proxies for household wealth. The first is log spouse income and the second is the number of bathrooms in the house. Using log spouse income as a proxy for household wealth may be problematic if we expect log spouse income to be affected by hiring a domestic helper. Nevertheless, given that spouse income is a good indicator of household wealth, we will include specifications that use spouse income as a control, and interpret the results with the appropriate caution. The second proxy we will use is the number of bathrooms in the house. This variable also captures household wealth and is arguably more exogenous than spouse income. 12

4.2.3 First-stage estimates Before turning to the main labor force participation models, we first present and discuss the firststage estimates. We allow for two different functional forms of the instrument, one in which the number of rooms enters linearly and a more flexible specification that uses dummies. Table 6 shows the results and the columns differ in the groups of controls included. Column (1) includes basic controls only, i.e. demographic and education characteristics of the woman, her husband and the household. Columns (2) to (4) include alternative measures of household wealth, namely the log of the husband s wage and the number of bathrooms in the house. As expected having more rooms increase the probability of hiring a domestic helper. The linear estimates suggest that having an extra room increases the chances of having a domestic helper by between 3.5 percentage points to 6.8 percentage points. Note that the magnitude of the estimate is smaller when measures of wealth are included; this is expected given that number of rooms is likely to proxy for wealth, which is positively correlated with hiring a domestic worker. Panel B presents a very similar picture: having 3 or more rooms increases the probability of having a domestic worker and the effect is smaller when proxies for wealth are included. The likelihood increases particularly (between 12 and 21 percentage points) when households go from 5 to 6 rooms. Estimates change very little when the sample is restricted to families that have not moved in the 5 years before the Census (columns (5) and (6)). Given that panel B suggests that the effect of number of rooms on the probability of hiring a domestic worker is not linear, we will present estimates from models that use a set of dummies for the number of rooms as instruments. Results using the number of rooms entered linearly as an instrument are available on request. 4.3 Effects on labor force participation Table 7 presents the estimates from the OLS and IV labor force participation models, both for the sample that includes all women and for the sample restricted to households who have not moved in the past 5 years (non-movers). Columns (1) and (2) present the estimates when husband s income is not included in the regressions. The OLS estimate in this case is at least seven times larger than the IV; however, when husband s income or the number of bathrooms are included (Columns (3) to (6)), the IV estimate is only slightly lower than the OLS estimate. The large change in the IV estimate is explained by the fact that if we do not include a variable proxying for wealth, number of rooms is proxying for unearned income too (in addition to space constraints), and unearned income has an inverse relationship with labor force participation. Our OLS estimates are slightly larger than the IV estimates, albeit not in a statistically significant way. This suggests a small and positive omitted variable bias. The magnitude of our preferred IV coefficients implies that having 13

a foreign domestic worker at home has a large effect on the labor force participation; it increases participation by about 20 percentage points, a 32 percent increase over the baseline average level of 63 percent. It is reassuring that these results do not change when the sample is restricted to non-movers. For comparability with our cross-country estimates, we restrict the sample to women with at least one child aged less than 18. The estimates are slightly larger than those using the entire sample. Our preferred IV estimate suggests that hiring a foreign domestic helper increases participation of married females with at least one child by approximately 30 percentage points. Since approximately 20% of married women with children hire domestic helpers, this estimate corresponds to a 6 percentage point increase in average labor force participation rates. 4.4 Do the effects on labor force participation vary by education? A simple model of time-use suggests that women with high education will be most affected by the availability of domestic help because of the high opportunity cost of their time. To test this hypothesis we re-estimate the LFP model above separately for women with low, medium and high levels of education. Table 8 presents the results. The coefficients, although usually positive and statistically significant, do not present the increasing pattern as suggested by the theory; there seems to be no difference in the magnitude of the effects for groups of different educational levels. While this finding does not seem consistent with our time-series findings that higher-educated females have seen a disproportionate rise in labor force participation rates as a result of the influx of foreign domestic helpers, one has to bear in mind a number of caveats in comparing the crosssectional and time-series approach. A potential explanation is that there are other considerations in the decision to enter the labor force besides the potential wage that might differ by education level. For example, spending time with kids might bring higher returns if the mother is well educated. For low-skilled mothers, leaving the children with a more educated domestic worker might even improve her kid s education, providing an additional incentive to join the labor force. This is a form of omitted variable bias; in this case, the omitted variable is the returns to household production that can have heterogenous effects by educational level. By aggregating the data, the time-series approach is less subject to such omitted variable biases at the individual level. Furthermore, from an econometric point of view, given that highly educated females in Hong Kong had very high labor force participation rates to begin with in 2001 and 2006, there is a smaller margin for changing their behavior. Female labor force participation in the 1970s to the 1980s was rather low, even among educated females; hence, the time-series approach may be better able to pick up broad changes in LFP rates between the different education groups by exploiting the variation in women s labor force participation rates over time. 14

4.5 Effects on occupational choice Drawing on the literature on women s work and job flexibility, another potentially interesting margin of substitution is whether hiring a domestic helper allows women to take on less flexible jobs and more rewarding jobs. We classify a job in terms of two dimensions, industry and occupation. Our empirical strategy here makes use of variation in job flexibility across two-digit industry and occupation cells (we refer to each cell as a job ) to address the question of whether hiring a foreign domestic helper enables women to take on more demanding, less flexible and more rewarding jobs. Since the Census lacks direct information on the number of hours worked, we will make use of several proxies for job flexibility. The first is the share of males in a given job. The proportion female captures the idea that more female-intensive jobs are more likely to allow a greater degree of job flexibility which is amenable to a female worker s dual roles in the household and workplace. This is consistent with literature that documents that female-intensive occupations tend to have more flexible work arrangements, such as a greater proportion of part-time workers, lower average hours and lower average job tenure (Macpherson and Hirsch, 1995). The second variable that we use is the average men s salary in a particular job. This allows us to discern whether hiring a foreign domestic helper allows women to enter higher paying jobs. To obtain a more direct measure of how time-intensive and flexible various jobs are, we will make use of the 2002-2006 Taiwan Manpower Utilization Survey (MUS) to compute the average weekly hours worked by Taiwanese men and the proportion of men working less than 30 hours per week in each industryxoccupation cell. As one would expect the occupational demands in Hong Kong and Taiwan to be relatively similar, these variables should serve as reasonable proxies for job flexibility in Hong Kong. We use the following regression specification: y ijst = α + βf DH it + γx 1it + φx 2js + Λ j + η t + θ k + ɛ it where j refers to an occupation group and s refers to an industry group. The covariates used are the same as those used in the employment regressions, but in some specifications, we will also include controls for job education requirements, X 2js, as proxied by the share of workers in each job with low, medium or high education and fixed effects for selected occupations, Λ j. Thus, if the employment of foreign domestic helpers frees up a women s household responsibilities, we would expect β to be negative, indicating a greater propensity to participate in an occupation that is less female intensive. Similarly, one would expect hiring a foreign domestic helper to be associated with an increase in participation in jobs with longer average hours and less flexible work hours. We use two dependent variables, the share of female workers and the log of the average salary for 15

male workers in the industry-occupation cell in which the woman works. Results are presented in table 9; all specifications include the same controls as in our labor force participation models and additionally includes industry-occupation controls for the educational composition and dummies for selected occupations. The estimated coefficients move in the expected direction; hiring a domestic helper is associated with employment in a job with a higher share of males and higher average male earnings. The magnitude of our estimates suggest that having a domestic worker allows a woman to work in a job that pays males about 20 percentage points more and has about 10 percent more males. To obtain more direct information on job flexibility, we merge a number of industry-occupational characteristics computed from the 2002-2006 Taiwan Manpower Utilization Survey. Here, we consider two additional dependent variables; the average weekly hours men work, the proportion of men working less than 30 hours per week. In panels C and D, we find that females with domestic helpers are more likely to be engaged in jobs where Taiwanese men worked longer hours and in jobs with a lower share of men working less than 30 hours per week. 4.6 Omitted variables?: Placebo tests In this section, we directly address concerns that our IV results may be driven by omitted variables that are correlated with our instrument and female employment, leading to a spurious reduced form relationship. For example, if the number of rooms are chosen in anticipation of future labor force participation or is merely proxying for some unobserved propensity to work, this is likely to bias our IV results upwards. While we introduce a number of individual-level covariates to ameliorate this issue, it is impossible to directly verify the exogeneity assumption. A possible way to check if these omitted variable biases are driving the results is to conduct several placebo tests on a subset of households that have a very low probability of maid hire, such as low income households who are not eligible to hire a foreign maid, married households without children or single women. In particular, we will run the following regression, where the instrument, number of rooms, is introduced as a continuous variable or as separate dummies. F DH it = α + δrooms it + γx it + η t + θ k + ɛ ijt The income eligibility threshold for the hiring of foreign domestic helpers is the median household income, or 15,000 HKD. In 2001 and 2006, 1% of these low-income households had a live-in foreign domestic helper and 4% of married females without children hired domestic helpers, compared to the sample average of 13%. If our instrument merely proxies for an individual s unobserved propensity to work, we would expect to find a significant positive relationship between the number of rooms and the employment status of females in these households, regardless of their low demand 16

for domestic services. However, if δ is only significant in the sample of households that have a relatively high demand for domestic help, then this suggests that the number of rooms affects the employment decisions of females through its impact on maid hire, as opposed to merely proxying for some unobserved variables that might be correlated with the individual s propensity to work. In table 10, we report estimates for four groups of women: (1) our usual sample, (2) married women with children (3) women with household income less than 15,000 HKD a month (the minimum required to apply for a domestic helper) and (4) married households without children. The cleanest comparison group is the lower income sample, where only 1 percent of women have a domestic worker compared to 13 percent in the sample group and 19 percent in the sample of married women with children. In the sample of married women without children, even though about 4 percent have a domestic worker at home, we expect little effect of having a domestic worker (and therefore the number of rooms) since these women have substantially less household responsibilities. As expected, the reduced form estimates for the usual sample are large and statistically significant. Having 3 or more rooms increases labor force participation, and coefficients are larger when husband s income is included in the regressions. These results are even larger when we restrict the sample to married women with children. Conversely, these results do not hold when the sample is restricted to household s with an income lower than 15,000 or married women without children. As shown panel B, the coefficients are not statistically significant and do not follow an increasing pattern. The results from the placebo tests indicate that in contrast to the findings for the main sample of females or married females with children, the number of rooms in the household has little or no impact on the employment probability of lower income households and married women with no children. 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 further reinforces the validity of our instrument. We obtain similar results for placebo tests using industry-occupation characteristics as the dependent variable (results available on request). 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 western countries as a result of demographic changes and increasing demand as women seek to enter the labor market. In this paper, we find that temporary domestic helper policies has had a sizeable impact on increasing female labor force participation rates in Hong Kong. Using the number of rooms as an 17

instrument to address the endogeneity of foreign domestic helper hire, our cross-sectional estimates suggest that the hiring a domestic helper is associated with a 20-30 percentage point increase in the probability of employment for women. Back of the envelope calculations suggest that this number translates to approximately 28,000 women entering the labor force. In addition, we find that women with foreign domestic helpers are also able to work in jobs with a larger share of males, higher average wages and that are more time-intensive. Using a difference-in-difference approach that compares the trends in labor force participation in Hong Kong and Taiwan, we find that, on the aggregate, the opening of the Hong Kong labor market to such migrant domestic workers has led to a 10 percentage point increase in LFP rates of among females with young children, compared to those with relatively older children from 1976-2006. 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. This can have important policy ramifications for encouraging women to enter the labor market and to bridge the gender gap. Nevertheless, some of the newly industrialized nations that have depended on foreign domestic helpers as part of their developmental strategy to encourage women to enter the labor force are also forced to address some unintended effects of the outsourcing of household production. Among these are concerns of whether childrens educational production may have been adversely affected with the increasing dependence on domestic helpers. In addition, a new generation of youngsters brought up in the care of these domestic helpers have been criticized for becoming too dependent on their foreign caretakers, lacking the initiative and impetus to develop skills needed to care for themselves. These spillover effects due to the increasing reliance on domestic workers are important questions that we hope will be addressed in future work. 18

References [1] Angrist, Josh and Pischke, Jörn-Steffen. 2008. Mostly Harmless Econometrics: An Empiricist s Companion. [2] Chan, Annie. 2006. The Effects of Full-Time Domestic Workers on Married Women s Economic Activity Status in Hong Kong, 1981-2001. International Sociology. 21(1):133-159. [3] Chiu, Stephen W.K. 2006. Recent Trends in Migration Movements and Policies in Asia: Hong Kong Region Report. [4] Cortes, Patricia and Tessada, Jose. 2007. Cheap Maids and Nannies: How Low-skilled Immigration is Changing the Labor Supply of High-skilled American Women. Unpublished Draft. [5] Gronau, Reuben.1997. Leisure, Home Production, and Work - the Theory of Allocation of Time Revisited. Journal of Political Economy. 85(6): 1099-1123. [6] Macpherson, David and Hirsch, Barry. 1995. Wages and Gender Compostion: Why do Women s Jobs Pay Less?. Journal of Labor Economics. 13(3)426. [7] Kremer, Micheal and Watt, Stanley. 2008. The Globalization of Household Production. Unpublished Draft. [8] Suen, Wing. 1994. Market-procured Housework: The Demand for Domestic Servants and Female Labor Supply. Labor Economics. 1:289-302. [9] Yeoh, Brenda S. A., Huang, Shirlena and Gonzalez III, Joaquin. Migrant Female Domestic Workers: Debating the Economic, Social and Political Impacts in Singapore. International Migration Review. 33(1): 114-136. 19

Table 1: FDH program in Hong Kong and Taiwan (1981-2006) Hong Kong Taiwan Year % of Labor Force 1981 0.30% - 1986 0.90% - 1991 2.00% - 1996 3.70% 0.32% 2001 5.30% 1.15% 2003 6.80% 1.20% Proportion of domestic households hiring at least one foreign domestic helper in Hong Kong Year All Sample* 1981 0.50% 0.10% 1986 1.50% 2.10% 1991 3.20% 4.70% 1996 5.80% 8.80% 2001 7.70% 11.10% 2006 7.90% 11.30% Source: Hong Kong: Summary Statistics of Population census/by-census, 1981-2001. 2003 Hong Kong data from Hong Kong Annual Digest of Statistics and Hong Kong Census and Statistics Bureau Taiwan: Yearbook of Labor Statistics, Council of Labor Affairs. Foreign worker data (http://statdb.cla.gov.tw/html/year/rptehidx13.htm) Labor force data (http://statdb.cla.gov.tw/html/mon/c2010.htm) Foreign domestic helpers in Taiwan are defined as the sum of nursing workers and home-maids. *Sample here refers to proportion of married households with female heads or household spouses aged 20-55 Table 2: Characteristics of Foreign Domestic Helpers by ationality in 2006 Filipino Indonesian Mean S.D Mean S.D Age 36.05 8.26 28.86 6.06 Age at first arrival 29.71 6.3 25.65 5.25 Married 0.47 0.5 0.43 0.5 Salary (2006 HK$) 3511.43 527.22 3411.37 447.9 Duration in HK 6.41 5.39 3.28 3.32 College 0.2 0.4 0.03 0.18 Speaks English 0.98 0.14 0.46 0.5 Speaks Cantonese 0.29 0.45 0.87 0.33 Notes: The exchange rate is approximately 1 USD to 7.8 HK$. 20

Table 3: Placebo Test - Difference-in-Differences Estimates in Pre-Period Time period: (1) (2) (3) (4) 1976-1981 1976-1985 Dep. Variable: Employment HK*(t>=1981)*(Child 0-5) -0.026-0.009-0.016 0.001 [0.023] [0.024] [0.021] [0.022] HK*(t>=1981) 0.077 0.07 0.04 0.029 [0.017]** [0.017]** [0.015]** [0.016] HK*(Child 0-5) 0-0.036 0-0.036 [0.016] [0.016]* [0.016] [0.016]* (t>=1981)*(child 0-5) -0.023-0.067-0.016-0.092 [0.013] [0.152] [0.011] [0.113] HK -0.009-0.005-0.009-0.005 [0.011] [0.011] [0.011] [0.011] (t>=1981) 0.008 0.012 0.058 0.068 [0.010] [0.129] [0.008]** [0.093] Child aged 0 to 5-0.094-0.191-0.094-0.191 [0.008]** [0.094]* [0.008]** [0.094]* Controls: Age NO YES NO YES Education NO YES NO YES Year dummies YES YES YES YES Observations 39320 39320 56646 56646 Notes: 1. Each column is a separate regression for the time period indicated in the top row of each table. 2. HK is a dummy variable that indicates if the individual is from Hong Kong. (t>=1981) refers to a dummy variable that indicates 1 for 1981 and all years post-1981. (Child aged 0 to 5) is an indicator variable that equals 1 if the individual has at least one child aged between 0 to 5. 3. Age controls include age, age interacted with a dummy variable indicating (t>=1981), age interacted with the presence of a child aged 0 to 5 and age interacted with both (t>=1981) and (Child aged 0 to 5). 4. Robust standard errors in brackets. * significant at 5%; ** significant at 1% 21

Time period: Table 4: Difference-in-Differences Estimates of Female Employment Growth Rates in Hong Kong vs. Taiwan (1) (2) (3) (4) 1976-1990 1976-1995 1976-2000 1976-2006 Method 1: Using micro-data with individual labor force participation as dependent variable Panel A1: Overall Sample HK*(t>=1985)*(Child 0-5) 0.002 0.028 0.051 0.069 0.089 0.1 0.094 0.101 [0.014] [0.014]** [0.012]*** [0.013]*** [0.012]*** [0.012]*** [0.012]*** [0.012]*** Observations 123453 123453 240512 240512 445753 445753 654105 654105 Panel A2: Low Education HK*(t>=1985)*(Child 0-5) -0.014-0.005 0 0.009 0.011 0.02 0.009 0.019 [0.015] [0.015] [0.014] [0.014] [0.013] [0.014] [0.013] [0.013] Observations 94509 94509 165037 165037 271251 271251 358495 358495 Panel A3: High Education HK*(t>=1985)*(Child 0-5) 0.029 0.066 0.069 0.094 0.1 0.116 0.098 0.111 [0.037] [0.038] [0.035]** [0.036]*** [0.034]*** [0.036]*** [0.034]*** [0.035]*** Observations 28944 28944 75475 75475 174502 174502 295610 295610 Method 2: Using group averages with average LFP rates in each country-year-group cell as dep. variable Panel B1: Overall Sample HK*(t>=1985)*(Child 0-5) 0.001 0.027 0.047 0.06 0.093 0.093 0.097 0.091 [0.042] [0.037] [0.039] [0.035]* [0.039]** [0.032]*** [0.040]** [0.032]*** Panel B2: Low Education HK*(t>=1985)*(Child 0-5) -0.014-0.007 0.003 0.008 0.011 0.013 0.007 0.008 [0.041] [0.042] [0.040] [0.042] [0.034] [0.035] [0.033] [0.034] Panel B3: High Education HK*(t>=1985)*(Child 0-5) 0.033 0.039 0.069 0.071 0.069 0.071 0.103 0.103 [0.061] [0.059] [0.055] [0.053] [0.055] [0.053] [0.052]* [0.050]** Observations 28 28 48 48 68 68 92 92 Controls: Age NO YES NO YES NO YES NO YES Education NO YES NO YES NO YES NO YES Year dummies YES YES YES YES YES YES YES YES Notes: 1.Each column is a separate regression for the time period and sample listed in the panel heading. 2. HK is a dummy variable that indicates if the individual is from Hong Kong. (t>=1985) refers to a dummy variable that indicates 1 for 1985 and all years post- 1985. (Child aged 0 to 5) is an indicator variable that equals 1 if the individual has at least one child aged between 0 to 5. 3. The dependent variable in the group-average regressions is the group average labor force participation in a given year. The unit of observation is country by year by presence of older or younger child; hence the total number of observations across the entire sample period is 23x2x2 = 92. All regressions are weighted by group size. 4. The grouped-data estimates are adjusted for individual-level covariates that include age, age-squared and dummies for low, medium and high education using a two-step procedure. In the first step, covariate-adjusted group means are obtained by regressing the micro-level outcomes on a full set of 92 group dummies, controlling for the individual level covariates. In step 2, the estimated group effects are used as dependent variables in the DID estimation procedure. 5. Robust standard errors in brackets. * significant at 10%; ** significant at 5%; ***significant at 1% 22

Table 5: The Effect of Having Recently Moved on the Probability of having a Foreign Domestic Worker Dependent Variable: Dummy for Domestic Worker Sample: All Sample: With oldest child 0-5 (1) (2) (3) (4) OLS OLS OLS OLS Dummy for Having Moved in past 5 yrs 0.004 0.002 0.011 0.006 [0.002]* [0.002] [0.010] [0.010] Controls: Basic YES YES YES YES Log of Husband's Income NO YES NO YES Dummies for number of bathrooms NO NO NO NO Observations 82500 9942 Notes: Basic controls include district, quarter type, and year fixed effects, woman's and husband's education dummies, age and age squared, household size, dummies for children 0-2, 3-5, 6-13, 14-18 and for household members aged 60-70 and 70+. 23

Table 6: First Stage Dependent Variable: Dummy for having a Domestic Worker Panel A. Instrument: umber of Rooms Sample: All Instrument: (1) (2) (3) (4) (5) (6) Number of Rooms 0.068*** 0.061*** 0.037*** 0.035*** 0.055*** 0.034*** [0.001] [0.001] [0.001] [0.001] [0.002] [0.002] Controls: Basic YES YES YES YES YES YES Log of Husband's Income NO YES NO YES YES NO Dummies for number of bathrooms NO NO YES YES NO YES Panel B. Instruments: Dummies for number of rooms Sample: All Sample: Non-movers Sample: Non-movers Instruments: (1) (2) (3) (4) (5) (6) Rooms=2-0.016*** -0.016*** -0.014*** -0.014*** -0.012*** -0.011*** [0.003] [0.003] [0.003] [0.003] [0.003] [0.003] Rooms=3-0.002-0.005* 0-0.002 0 0.004 [0.003] [0.003] [0.003] [0.003] [0.003] [0.003] Rooms=4 0.054*** 0.045*** 0.040*** 0.034*** 0.044*** 0.038*** [0.003] [0.003] [0.003] [0.003] [0.004] [0.004] Rooms=5 0.136*** 0.119*** 0.087*** 0.078*** 0.110*** 0.081*** [0.005] [0.005] [0.005] [0.005] [0.006] [0.006] Rooms 6+ 0.345*** 0.314*** 0.211*** 0.198*** 0.316*** 0.209*** [0.008] [0.008] [0.009] [0.009] [0.011] [0.013] Controls: Basic YES YES YES YES YES YES Log of Husband's Income NO YES NO YES YES NO Dummies for number of bathrooms NO NO YES YES NO YES Observations 82500 47679 Notes: Basic controls include district, quarter type, and year fixed effects, woman's and husband's education dummies, age and age squared, household size, dummies for children 0-2, 3-5, 6-13, 14-18 and for household members aged 60-70 and 70+. 24

Sample: All Sample: Non-movers (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) OLS IV OLS IV OLS IV OLS IV IV IV Panel A: Overall Sample Domestic Worker 0.247*** 0.033 0.277*** 0.207*** 0.270*** 0.218*** 0.286*** 0.369*** 0.207*** 0.191*** [0.005] [0.021] [0.005] [0.023] [0.005] [0.042] [0.005] [0.045] [0.035] [0.063] N 82500 82500 82500 82500 82500 82500 82500 82500 47679 47679 R-squared 0.14 0.13 0.16 0.16 0.15 0.15 0.16 0.16 0.13 0.12 Panel B: With at least one child aged less than 18 Domestic Worker 0.286 0.092 0.316 0.293 0.304 0.298 0.322 0.475 0.263 0.259 [0.005]** [0.025]** [0.005]** [0.028]** [0.005]** [0.046]** [0.005]** [0.049]** [0.042]** [0.067]** N 52602 52602 52602 52602 52602 52602 52602 52602 30365 30365 Controls: Basic YES YES YES YES YES YES YES YES YES YES Log of Husband's Income NO NO YES YES NO NO YES YES YES NO Dummies for number of bathrooms NO NO NO NO YES YES YES YES NO YES 2. The instruments used are dummies for number of rooms Table 7: The Effect of Foreign Domestic Workers on Female Labor Supply Notes: 1. The mean labor force participation rates is 0.63 for the overall sample. Dependent Variable: Labor Force Participation 3. Basic controls include district, quarter type, and year fixed effects, woman's and husband's education dummies, age and age squared, household size, dummies for children 0-2, 3-5, 6-13, 14-18 and for household members aged 60-70 and 70+. Table 8: The Effect of Foreign Domestic Workers on Female Labor Supply by Education Sample: All Sample: Non-movers (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) OLS IV OLS IV OLS IV OLS IV IV IV Panel A: Low Education Domestic Worker 0.248-0.042 0.299 0.25 0.269 0.024 0.301 0.316 0.379 0.247 [0.015]** [0.066] [0.014]** [0.069]** [0.015]** [0.138] [0.015]** [0.143]* [0.102]** [0.204] N 38767 38767 38767 38767 38767 38767 38767 38767 26022 26022 Panel B: Medium Education Domestic Worker 0.252 0.027 0.278 0.185 0.273 0.213 0.287 0.358 0.176 0.177 [0.006]** [0.028] [0.006]** [0.031]** [0.006]** [0.052]** [0.006]** [0.056]** [0.043]** [0.071]* N 34089 34089 34089 34089 34089 34089 34089 34089 18060 18060 Panel C: High Education Domestic Worker 0.163 0.016 0.171 0.064 0.181 0.314 0.185 0.349 0.031 0.215 [0.010]** [0.042] [0.010]** [0.047] [0.010]** [0.074]** [0.010]** [0.078]** [0.071] [0.105]* N 9644 9644 9644 9644 9644 9644 9644 9644 3597 3597 Controls: Basic YES YES YES YES YES YES YES YES YES YES Log of Husband's Income NO NO YES YES NO NO YES YES YES NO Dummies for number of bathrooms NO NO NO NO YES YES YES YES NO YES Notes: 1. The mean labor force participation rates for each of the samples is 0.51 for the low educated, 0.71 for the medium educated and 0.82 for the high educated. 2. The instruments used are dummies for number of rooms. 3. Basic controls include district, quarter type, and year fixed effects, woman's and husband's education dummies, age and age squared, household size, dummies for children 0-2, 3-5, 6-13, 14-18 and for household members aged 60-70 and 70+. 25

Table 9: The Effect of Foreign Domestic Workers on Occupation Choice (1) (2) (3) (4) (5) (6) IV IV IV IV IV IV Panel A: Dep. Var - Share of men in occupation Dummy for Domestic Worker 0.166*** 0.125*** 0.089*** 0.095*** 0.078*** 0.072*** [0.014] [0.012] [0.010] [0.021] [0.017] [0.018] Observations 38039 38039 38039 38039 38039 38039 Panel B: Dep. Var - Log of average salary of men in occupation Dummy for Domestic Worker 0.667*** 0.237*** 0.175*** 0.168*** 0.128*** 0.119*** [0.036] [0.019] [0.016] [0.032] [0.026] [0.028] Observations 38039 38039 38039 38039 38039 38039 Panel C: Dep. Var - Average weekly hours worked by men in occupation Dummy for Domestic Worker 0.588 1.915 0.657 1.137 0.115 0.226 [0.209]** [0.188]** [0.153]** [0.323]** [0.279] [0.296] Observations 45783 45783 45783 45783 45783 45783 Panel D: Dep. Var - Proportion of men working less than 30 hours Dummy for Domestic Worker -0.025-0.019-0.011-0.017-0.012-0.011 [0.002]** [0.002]** [0.002]** [0.004]** [0.004]** [0.004]** Observations 45783 45783 45783 45783 45783 45783 Controls: Basic YES YES YES YES YES YES Log of Husband's Income YES YES YES NO NO YES Number of bathrooms dummies NO NO NO YES YES YES Occupation Educational Comp. NO YES YES YES YES YES Dummies for Doctors, lawyers, etc NO NO YES NO YES YES Notes: 1. The dependent variables in Panels A and B are average industry-occupation characteristics from the 2001 and 2006 Hong Kong Population Census. The dependent variables in Panels C and D are average occupation-industry characteristics constructed from the 2002-2006 Taiwan Manpower Utilization Survey. 2. The instruments used are dummies for number of rooms. 3. Basic controls include district, quarter type, and year fixed effects, woman's and husband's education dummies, age and age squared, household size, dummies for children 0-2, 3-5, 6-13, 14-18 and for household members aged 60-70 and 70+. 26

Panel A: (1) (2) (3) (4) (1) (2) (3) (4) Rooms=2 0.003 0.003 0.003 0.003 0.019 0.017 0.019 0.017 [0.010] [0.010] [0.010] [0.010] [0.014] [0.014] [0.014] [0.014] Rooms=3 0.022 0.028 0.022 0.028 0.048 0.054 0.048 0.054 [0.009]* [0.009]** [0.009]* [0.009]** [0.012]** [0.012]** [0.012]** [0.012]** Rooms=4 0.04 0.057 0.043 0.056 0.074 0.092 0.075 0.09 [0.010]** [0.010]** [0.010]** [0.010]** [0.013]** [0.013]** [0.013]** [0.013]** Rooms=5 0.044 0.076 0.054 0.075 0.084 0.12 0.091 0.115 [0.010]** [0.010]** [0.011]** [0.011]** [0.014]** [0.014]** [0.014]** [0.014]** Rooms 6+ 0.021 0.08 0.052 0.083 0.053 0.119 0.082 0.119 [0.012] [0.012]** [0.013]** [0.013]** [0.016]** [0.016]** [0.017]** [0.017]** Observations Panel B: Table 10: Placebo Test - Labor Force Participation Overall Sample (13% with foreign domestic helpers) Married With Children (19% with foreign domestic helpers) 82500 52798 Household Income <15000 Married Without Children (1% with foreign domestic helpers) (4% with foreign domestic helpers) Rooms=2-0.025-0.023-0.025-0.024-0.011-0.009-0.012-0.009 [0.015] [0.014] [0.015] [0.014] [0.015] [0.015] [0.015] [0.015] Rooms=3-0.013-0.006-0.013-0.006 0.009 0.014 0.009 0.014 [0.013] [0.013] [0.013] [0.013] [0.014] [0.014] [0.014] [0.014] Rooms=4-0.016-0.018-0.014-0.015 0.009 0.022 0.015 0.025 [0.016] [0.016] [0.016] [0.016] [0.015] [0.015] [0.015] [0.015] Rooms=5-0.026-0.028-0.017-0.017-0.007 0.019 0.011 0.028 [0.022] [0.022] [0.023] [0.022] [0.016] [0.016] [0.016] [0.016] Rooms 6+ -0.052-0.079-0.017-0.036-0.022 0.025 0.017 0.04 [0.039] [0.037]* [0.043] [0.041] [0.019] [0.019] [0.022] [0.022] Controls Basic YES YES YES YES YES YES YES YES Log of Husband's Income NO YES NO YES NO YES NO YES No. of bathrooms dummies NO NO YES YES NO NO YES YES Observations 16995 29702 Notes: Basic controls include district, quarter type, and year fixed effects, woman's and husband's education dummies, age and age squared, household size, dummies for children 0-2, 3-5, 6-13, 14-18 and for household members aged 60-70 and 70+. 27

Appendix Table 1: Descriptive Statistics - Married Women 18-60 years of Age Domestic Worker=1 Domestic Worker=0 Mean Std. Dev. Mean Std. Dev. Age 39.4 6.3 41.4 8.0 Primary School 0.01 0.09 0.09 0.29 Less than Form5 0.10 0.30 0.43 0.50 Form5 0.36 0.48 0.27 0.45 Post Secundary 0.23 0.42 0.11 0.31 College 0.23 0.42 0.07 0.26 Graduate 0.07 0.26 0.02 0.12 LFP 0.83 0.38 0.60 0.49 Working 0.82 0.38 0.59 0.49 Homemaker 0.17 0.38 0.40 0.49 Wage Working 27381 23145 13103 13005 Wage of Husband 41548 36190 18155 17973 Number of Rooms 4.56 1.08 3.38 1.06 Two or More Toilets 0.57 0.49 0.15 0.35 Household size 3.85 0.98 3.56 1.13 Dummy for Children 0.90 0.30 0.60 0.49 Dummy Youngest Child 0-2 0.25 0.43 0.07 0.25 Dummy Youngest Child 3-5 0.24 0.43 0.09 0.28 Dummy Youngest Child 6-13 0.35 0.48 0.28 0.45 Dummy Youngest Child 14-18 0.16 0.37 0.56 0.50 Dummy Female 60-70 0.03 0.18 0.02 0.14 Dummy Female 70+ 0.07 0.25 0.04 0.19 Dummy Male 60-70 0.03 0.16 0.05 0.23 Dummy Male 70+ 0.03 0.16 0.02 0.12 Observations 10855 10855 71645 71645 Appendix Table 2: umber of Rooms by FDH hire Domestic Worker=1 Domestic Worker=0 Freq. Percent Freq. Percent Rooms=0-1 14 0.13 3,067 4.28 Rooms=2 104 0.96 7,430 10.37 Rooms=3 1,959 18.05 32,663 45.59 Rooms=4 3,069 28.27 17,999 25.12 Rooms=5 3,122 28.76 8,347 11.65 Rooms=6+ 2,587 23.83 2,139 2.99 Observations 10855 71645 28

Figure 1: Stock of Foreign Domestic Helpers by Nationality (1990-2004) Source: 1991 and 1995 figures are unpublished information supplied by the Immigration Department. The 1992-1994 and 1997 to 2005 figures are from the Census and Statistics Department. Figure was plotted from tables in Chiu, 2006 Recent Trends in Migration Movements and Policies in Asia: Hong Kong Region Report. Figure 2: Proportion of households with a local or foreign domestic helper 29

Figure 3: Proportion of married females hiring a live-in foreign domestic helper by age of child Notes: The sample here consists of households with female household heads aged between 20-55. The proportion of households hiring a domestic helper in each year was computed as the proportion of households that had at least one or more domestic helpers in each subset of the population considered. Figure 4: Employment rates by gender and marital status over time Notes: The sample here consists of males and females aged between 20-55 who were household heads or spouses of household heads. Employment rates in each year were computed as the average participation rate of each subset of the population considered. 30

Figure 5: Employment rates of females aged 20-55 by age range of child Notes: The sample here consists of females aged between 20-55 who were household heads or spouses of household heads. Employment rates in each year were computed as the average participation rate of each subset of the population considered. Figure 6: Female labor force participation rates by presence of a younger or older child in Hong Kong and Taiwan Notes: The sample here consists of households with female household heads/ spouses of household heads aged 20-55 years who are married and have at least one child. Child aged 0 to 5 indicates the presence of at least one child aged 0 to 5 years. Child aged 5 to 18 indicates the presence of at least one child aged 5 to 18, but no children aged 0 to 5. For both countries, we exclude foreign domestic helpers. 31

Figure 7: Female labor force participation rates by presence of a younger or older child in Hong Kong and Taiwan, by education level Notes: The sample here consists of households with female household heads/ spouses of household heads aged 20-55 years who are married and have at least one child. Child aged 0 to 5 indicates the presence of at least one child aged 0 to 5 years. Child aged 5 to 18 indicates the presence of at least one child aged 5 to 18, but no children aged 0 to 5. The low education sample comprise of those with less than Grade 5 education. For both countries, we exclude foreign domestic helpers. 32