1. Empirical models and data
This paper investigates the relationship between foreign domestic workers and native women’s labor market outcomes in Taiwan. The Taiwan Manpower Utilization Survey (MUS) is used as the main database in order to proceed with the empirical analysis.
The Manpower Utilization Survey has been conducted every May by Taiwan’s Directorate-General of Budget, Accounting, and Statistics (DGBAS) since 1978. Each year the survey contains labor force information from about 60,000 individuals who are Taiwanese citizens above the age of 15, excluding those in military service or prison.
Two-stage stratified sampling is used to sample households for this survey. Sample units drawn in the first stage of sampling are TSUN/LIs, while those drawn in the second stage are households. Approximately 500 TSUN/LIs were drawn in the first stage of sampling and about 20,000 households were sampled in the second stage. The survey adopts a 2-10-2 sample rotation method, so each year almost half of the sample is the same as the previous year.
At the beginning of the questionnaire there are some questions about personal information, such as gender, age, and highest level of education. After the personal information, there are three parts in the questionnaire. The first part aims to understand the manpower utilization and job changing status of employees. It asks employees for their monthly income from their primary job, the number of hours per week usually worked, their duration in their present job, the number of times they have changed jobs, their previous places of work and type of work, their reason for leaving their previous job, the way they got their present job, and whether they expect to change jobs or to get an additional job in the meantime.
The second part aims to understand the manpower utilization and job changing status
‧
of unemployed people. It asks unemployed people about the kind of job that they wish to have and their expected monthly payment, whether they encounter any job opportunities when seeking for jobs, the reason why they decline a job, and the resources that they rely on while looking for a job.
The third part aims to understand the manpower utilization and job changing status of people who can work but are not in the labor force. It asks them about their work status from the last year and their reason for quitting the job, their experience of looking for jobs last year and the reason they stopped looking for a job, their willingness to work and expected payment in order to understand the status of the potential labor force. Moreover, the survey asks women who have spouses about the age of their children, in order to understand how the age of children affects the labor participation rate of women who have spouses.
In addition to MUS, the study draws data from the Department of Household Registration M.O.I and the Bureau of Employment and Vocational Training in order to get the population of local females and foreign domestic workers for counties and cities. The available duration of the data is from May 1978 to May 2012 in MUS while that from the Bureau of Employment and Vocational Training is available from August 2009 to April 2013. The study selects the overlapping period from May 2010 to May 2012 to conduct the analysis. The sample size is 173,166 in total.
The empirical model in this paper is based mainly on that of Barone and Mocetti (2011), who examined how female immigrants who served as domestic workers affected the local female labor supply in Italy. Moreover, I refer to 莊慧玲、林世昌(2006) who arranged most of the empirical studies over the past 50 years that relate to the female labor supply in Taiwan. They found that women’s labor force participation rate is significantly affected by their age, education level, and number and ages of children, while women’s
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
17
working hours were significantly affected by their age, education level, number of children, husband’s income and wage rates. Although they did not mention the effect of elders in the household on women’s employment decisions, I wonder about the impact that must exists given traditional social norms in Taiwan. So I refer to 賴意婷 (2011) establishing a variable ‘number of elders in the household’ in order to investigate the effect of elders in the household on women’s employment. I think that this newly-created variable is an ideal substitution for ‘whether there are elders in the household’, because the former is able to show the impact of elders in the household more accurately.
The study uses the above literature as references in order to sort out the dependent variables and independent variables in the model (Equation (4.1))
LMOi, =α+Si,l β+Xiγ+εi Equation (4.1)
LMO is the outcome variable, including whether women join the work force and the number of hours they work per week. i denotes individuals. S is the share of foreign domestic workers on the total local female population in the specific l county or city where woman i resides. X is the matrix of individual-level characters, including age, age square, highest level of education, income of her husband per month, number of children and number of elders in the household. Moreover, year and regional dummies are also included in X.
Since “husband income” and “number of children” are important independent variables for women’s labor market outcomes, the study only uses married women as the population.
The study applies a probit model to the first regression (whether women join the work force) and applies a tobit model to the second regression (number of hours they work per
‧
week). The panel data methods can be applied to this data structure. I do not use them here but may do so for future studies.
Foreign domestic workers contain foreign domestic helpers and foreign caretakers. In fact, the latter includes two kinds of workers- foreign domestic caretakers and rehabilitation center foreign caretakers. The study used the number of foreign caretakers but not foreign domestic caretakers to do the research because the Bureau of Employment and Vocational Training only released the data of foreign caretakers, and I think that the data is remain a good estimator because the number of foreign domestic caretakers accounted for about 95% of the number of foreign caretakers.
Figure 4.1 shows the share of foreign domestic workers on the total local female populations for counties and cities in Taiwan. The proportion of foreign domestic workers has increased in every county and city from 2010 to 2012 except in Taipei City and Hsinchu City.
This paper establishes three dummy variables in order to control for the effect of education level on women’s labor market outcomes. “Junior high or below” is defined as having a junior high or below education level, which is set as the reference group. The other two categories are “high school” and “college or above”.
The variable “husband’s income” refers to the spouse’s monthly income. Because the raw data of MUS did not pair husband and wife together and the questionnaire did not ask about the income of one’s spouse, the variable cannot be directly obtained. The study adopts a method from Chuang and Lin (2006) which uses the household ID number in order to match the sample of “housemaster” and “the spouse of housemaster” to pair them into couples. The variable is classified into six dummy variables and the range of each group is 20,000 NT dollars. Spousal incomes that are lower than 20,000 or even are deficit are classified as the reference group.
‧
The paper classifies “number of children” into three categories. There are “number of children under the age of 3”, “number of children from age 3 to 5”, and “number of children from age 6 to 17”. I refer to Cortes and Pan (2009) and Barone and Mocetti (2011) to set the boundary value.
All counties and cities are categorized into four regional dummies: north, east, south and midst. This classification is made according to the Department of Household Registration M.O.I. “East” is set as the reference group, which includes Taitung County and Hualien County. “North” includes New Taipei City, Taipei City, Yilan County, Taoyuan County, Hsinchu County, Keelung City and Hsinchu City. “Midst” includes Taichung City, Changhua County, Nantou County and Yunlin County. “South” includes Tainan City, Kaohsiung City, Chiayi County, Pingtung County, Penghu County and Chiayi City. The paper excludes Kinmen County and Lienchiang County from the analysis since the MUS does not publish data of native women for these two regions separately. It is worth mentioning that although Taiwan did conduct the city-county consolidation at the end of 2010, this did not affect the classification of the regional dummies. The detailed definitions of all the variables are listed in Table 4.1.1.
The sample size becomes much smaller due to pairing husbands and wives together. I dropped samples who were single or whose spouse did not participate in the survey. After keeping all the married women’s data, 36, 398 observations remained. Table 4.1.2 provides descriptive statistics for all the variables.
According to the descriptive statistics, 43% of women are participating in the labor market. The proportion of foreign domestic workers in counties or cities is about 1% on average. More than 50% of the population belongs to the lowest education level and nearly 20% of the population belongs to the highest education level. More than 44% of the female’s spouse’s incomes are lower than 20000 NT dollars, since lots of zeros are shown
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
20
in the raw data of income value. It is hard to know whether this is because they did not get paid or because they refused to answer the question. The number of samples was distributed equally in each year after the arrangement of the raw data. Only about 4% of the sample is drawn from the East of Taiwan, while that drawn from the other three regions, north, middle and south, does not differ too much from each other.
‧
Table 4.1.1 Definitions of All the Variables
Variables Definitions
Dependent variables
Labor force participation Dummy variable, equals 1 if a native woman participates in the labor market and 0 otherwise.
Weekly working hours Independent variables
The proportion of foreign domestic workers
The share of foreign domestic workers on the total local female population in the specific county or city.
Age
Husband’s monthly income (NT)
below 20000 The reference group.
20000~40000 Dummy variable.
Number of elders aged over 65 Year
2010 The reference group.
2011 Dummy variable.
‧
The proportion of foreign domestic
workers (%) 1.6513 0.4949 2.8194 0.9503
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
23
Figure 4.1 the Share of Foreign Domestic Workers on the Total Local Female Population for Counties and Cities in Taiwan Source: Taiwan’s Directorate-General of Budget, Accounting, and Statistics (DGBAS)
Population (%)
‧
participation for counties and cities. There is a positive correlation between the proportion of foreign domestic workers and local women’s labor force participation, statistically significant at the 1% level. When the proportion of foreign domestic workers increases by 1%, the model estimates a 3% increase in the labor force participation among women.The coefficient of age is positive and statistically significant at the 1% level while it is negative and statistically significant at the same level for age square. That means female labor force participation will initially rise with increasing age, but will start to decline when women get older.
Among women who have a high school education background, the model estimates an increase in labor force participation of 4.5% points, and a college-leveled education background raises the probability of being in labor force by 17% points. This implies that more education is associated with a higher probability of working.
The coefficients of the husband’s monthly income are positive. Among women whose husband’s income is between 20000 and 40000 NT dollars, this shows a large rise in labor force participation of 13% points at the 1% significance level, while there is no significant effect among women whose husband’s income is 100000 NT dollars or higher. This implies that a lower income of one’s husband is associated with a higher probability of the woman’s working.
For women who have one more child under the age of 3, the model estimates a large decline in labor force participation of 15% points, and having one more child aged 3 to 5
‧
reduces the probability of being in the labor force by 7% points. Women who have older children are more likely to work, but this is not a significant determinant of working.
The number of elderly people in the household is negatively related to women’s labor force participation, statistically significant at the 10% level. For women who live with one more elderly person aged 65 or older, the model estimates a decline in labor force participation of 1% point.
The positive correlation between female labor force participation and the service provided by foreign domestic workers accords with the prediction made in section III. The results of the other control variables of age, highest level of education, husband’s income and number of children are all consistent with most of the past empirical studies. And the result of number of elders in the household is consistent with the assumption which was mentioned earlier in the study.
Table 4.2.2 reports the results of the tobit estimates of a woman’s weekly working hours based on Equation (4.1), and the coefficients from the Table have been transformed into marginal effects. The tobit model would be applied if the dependent variable is constrained and there is a clustering of observations at the constraint. Because lots of values of weekly working hours are zero in our data, the current paper uses a tobit model to proceed with the analysis.
There is a positive correlation between the proportion of foreign domestic workers and hours worked by local women, statistically significant at the 1% level. When the proportion of foreign domestic workers increases by 1%, the model estimates a weekly one working hour increase among women.
The coefficient of age is positive and statistically significant at the 1%level, while it is negative and statistically significant at the same level for age square. Among women who have a high school education background, the model estimates an increase in working
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
26
hours by 1.8 hours per week, and a college-level background increases weekly working hours by 5.8 hours. This implies that higher education is associated with longer hours worked by native women.
Again, the impact of income is sizable and stronger for husbands with a lower income.
This implies that a lower husband’s income is associated with longer working hours for the woman.
Having a child under the age of 3 reduces the hours worked per week by more than 5 hours; the effect is statistically significant at the 1% level. Furthermore, having a child aged 3 to 5 reduces working hours by 2.5 hours per week, statistically significant at the 1%
level. The reduction is about 0.15 hours for children aged 6 to 17, but this is not a significant determinant of working.
The number of elders in the household is negatively related to women’s weekly working hours, statistically significant at the 1% level. For women who live with more than one elderly person aged 65 or older, the model estimates a decline in 0.7 hours worked per week.
The positive correlation between female working hours and the service provided by foreign domestic workers accords with the prediction made in section III. Age, highest level of education, husband’s income, number of children and number of elders all have significant and expected signs in this model. And the result of the number of elders in the household is also consistent with the assumption made in this study.
‧
Table 4.2.1 Foreign Domestic Workers and Labor Force Participation of Local Women Probit Model (marginal effects)
Variables Coefficients Standard Err.
The proportion of foreign domestic
workers (%) 0.0304*** 0.0081
Number of elders aged over 65 -0.0117* 0.0065 Year
Observations: 36, 398. *Significant at 10%. **Significant at 5%. ***Significant at 1%.
‧
Table 4.2.2 Foreign Domestic Workers and Working Hours of Local Women Tobit Model (marginal effects)
Variables Coefficients Standard Err.
The proportion of foreign domestic
workers (%) 1.0900*** 0.3075
Number of elders aged over 65 -0.7869*** 0.2523 Year
Constant -29.1706*** 2.4337
Observations: 36, 398. *Significant at 10%. **Significant at 5%. ***Significant at 1%.
‧
“foreign domestic helper scheme” into Taiwan. The aim of the two schemes was to lessen women’s burden from attending to the old and the young, and free up native women to join the labor market. However, whether the programs can raise women’s labor force participation rate had not been confirmed in the literature.
This study makes an initial attempt using an empirical method to analyze the connection between foreign domestic workers and women’s labor market outcomes in Taiwan. Probit and tobit models are applied to conduct the research. The study finds that the proportion of foreign domestic workers to the native female population has a strong positive correlation to native women’s labor market outcomes. When the proportion of foreign domestic workers increases by 1%, the model estimates an 3% increase in the labor force participation of women. And the correlation is statistically significant at the 1% level.
When the proportion of foreign domestic workers increases by 1%, the model estimates a weekly one working hour increase among women. The correlation between the two variables is also statistically significant at the 1% level.
The results of the research provide a newer point of view on the factors that are related to women’s labor market decisions. The existence of services provided by foreign domestic workers seems to have a strong positive relationship with women’s labor market outcomes in Taiwan. Moreover, the findings have certain implications for policies that encourage females to enter the labor market.
In the past years, there have been requests from citizens asking government to loosen the limitations placed on applying for foreign domestic workers. However, the government is concerned that native domestic workers would suffer from an unemployment shock because they request higher wage rates than foreign domestic workers. Given this, the
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
30
qualifications required to apply for a foreign domestic worker are still quite strict in Taiwan when compared to other Asian counties. Future studies can take related topics further by examining the positive impact on female labor market decisions and the negative impact on native domestic workers, analyzing the overall effect of the “foreign caretaker scheme” and the “foreign domestic helper scheme”, and thereby provide further suggestions about these programs to government.
‧
9. Guglilmo Barone and Sauro Mocetti (2011).With a little help from abroad: The effect of low-skilled immigration on the female labour supply. Labour Economics, 18(5), 664-675.
10. Patricia Cortes and Jose Tessada (2011). Low-Skilled Immigration and the Labor Supply of Highly Skilled Women. American Economic Journal, Applied Economics, 3(3), 88-123.
11. Patricia Cortes and Jessica Pan (2013). Outsourcing Household Production: Demand for Foreign Domestic Helpers and Native Labor Supply in Hong Kong. Journal of Labor Economics, 31(2), 327-371.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
32
12. Lídia Farré, Libertad González and Francesc Ortega (2011). Immigration, Family Responsibilities and the Labor Supply of Skilled Native Women. The B.E. Journal of
12. Lídia Farré, Libertad González and Francesc Ortega (2011). Immigration, Family Responsibilities and the Labor Supply of Skilled Native Women. The B.E. Journal of