Household’s Demand for Domestic Help:
The Effect of Married Women’s Employment
on Domestic Help in German Private Households
Thesis Supervisor: Prof. Erik Plug
Willemijn Jongbloed Bachelor Thesis
Abstract
During the last half of the twentieth century female labor force rose significantly. This has led to more dual-earnings households as well as a decrease of women’s time contribution in the household. Recently, the increase of women labour force and the growth of domestic labour are linked. In this paper the effect of married women’s employment on the decision to employ domestic help is researched. The results indicate that the employment of married women increases household’s demand for domestic help with 5.28% in richer households, compared to 2.66% in all households. However, the effect of domestic help on female employment is found to be stronger than the effect of female employment on domestic help: respectively 1.17%
Table of Contents
I.
Introduction
II.
Literature Review
III.
Methodology
IV.
Analysis
V.
Results
VI.
Conclusion
VII.
Discussion
Introduction
Data on the United States labor market show a growth of demand in low skilled and high skilled labor due to technological progress (Autor and Levy, 2003). Weinberg (2000) argues that technological advancements de-emphasizes physical skills and therefore benefits women’s employment relative to men. The increase of female labor force has led to an increase in dual earning households that have more disposable income than single earning households (Jacobs and Gerson, 2008). Moreover, dual earning households have less time available for housework. The link between the increase of female labour force and the growth of domestic labour is a recent phenomenon (Yeoh, 2012; Anderson, 2001). Treas and de Ruijter (2008) state that the increase of married women’s labor force participation is commonly linked to market substitutes for women’s time spend in the household. In addition, the income of a household is positively related to the demand for domestic help (Treas and de Ruijter, 2005).
Economic and sociologic literature has not paid much attention to the link between the female labor force and domestic help. Prior empirical research by Bittman at al. (1999) did observe a trend between the rise of female employment and the demand for domestic outsourcing in the last quarter of the twentieth century in Australia. However, the hypothesis that dual earning families outsource domestic work has not been tested yet. This omission is serious as the number of households where both men and women are employed is increasing. Moreover, identifying the impact of female labor on domestic help is challenging and difficult to estimate as the causal relation goes back and forth. Literature compares female and male earnings to expenditures on domestic help. Sociological literature focuses on the internal decision-making process of household’s time allocation and domestic help (Wang and Li, 2009; Stancanelli and Statton, 2010). Whereas, studies by De Ruijter et al. (2005), Treas and de Ruijter (2008), Bittman (1999) compare household’s expenditures on outsourcing specific tasks to earnings of both men and women. As prior research focused on the internal decision-making process of domestic outsourcing, this empirical research focuses on the external decision of domestic outsourcing. The aim of this thesis is to answer the following question: does the increase in female labor force affects household’s demand for domestic help?
The thesis uses micro-data from the German Socio-Economic Panel (G-SOEP) to examine the effect of married women’s employment on household’s demand for domestic help. The G-SOEP was used in a previous empirical study by Hank (1998) to model the joint decision between household’s supply of female labor and their demand for domestic help in 1994. Consistent with other literature the study by Hank examines the effect of household’s time allocation and the decision to hire domestic help and finds that both are complementary. Omitted in the study of Hank is the effect of female labor force on the demand for domestic help because the number of observations of domestic workers was too small at the time. Since 1994, the G-SOEP has expanded in number of observations (Appendix A). This paper adds to previous research by examining the external effect of female employment on household’s demand for domestic outsourcing.
Literature Review
This chapter will first provide an historical background of female labor force and domestic employment. Because, as noted in the introduction, since the 1990’s a parallel trend between female labor force and household’s employment of domestic help is observed. Secondly, whether female labor force and domestic help are causally linked is discussed in the ‘Empirical Studies’ part of this chapter. Interfering factors of the link between female labor force and domestic help -technological progress and household’s time allocation- will be discussed lastly. Historical Background
(I). Female Labor Force
Female labor force rose significantly in the last half of the twentieth century (Bianci and Milkie, 2010). Historical data show a U-shaped female labor participation in the United States in the twentieth century (Goldin, 1994; Appendix B). This trend of female labor force is common among most economically advanced countries (Jacobsen, 1999; Olivetti, 2013).From 1940 until 1990 women were more actively working outside the home environment. In 1947 31.5% of women were in the labour force and this has risen to 60% by 2000 (Blau and Kahn, 2007). The female labor force in Germany has increased from 42.5% to 52.5% from 1990 until 2000 (Appendix B). Female employment in Germany is relatively high compared to other European countries; 71.5% of German women in the age group of 20 to 64 are employed whereas the European average is 62.3% (German Census, 2012).
The share of married women in the labor force grew significantly faster than that of unmarried women (Goldin, 1994). Jacobsen (1999) found that marital status affects labor force participation. Moreover, the presence of children has contributed to the increase of female labor supply: married women with young children have increased their labor force participation more than married women with children over 18 years (Jacobsen, 1999). Also, the combined effect of marriage and children on the supply of labor is researched has been studied by Bianci and Milkie (2010): from the late 1990s on the labor force of single mother’s increased and the rate for married women decreased.
(II). Domestic Outsourcing
Over time, literature has stated different relative numbers of households that employ domestic workers. In 1994 4% of households outsource cleaning service in Australia (Bittman et al., 1999). According to Hank (1998) survey findings on domestic labor by Hatzhold (1986) concluded that 6% of all households employ full-time domestic labor and 30 percent employ part-time domestic labor (Hank, 1998). Stancelli and Stratton (2010) find that 7% of households employ domestic workers. Sullivan and Gershuny (2013) found that 13 % of UK households employ domestic help in the years 2000 and 2001. Since 1990 the domestic workforce has grown significantly. In 1995 the amount of labor in private household was estimated at 33.2 million and in 2010 this number had increased to 52.6 million. Worldwide accumulated national survey data estimates a growth of 19 million domestic workers (ILO, 2013). From the second half of the
twentieth century women from low economic developed countries immigrated to the richer countries. Especially the more privileged women outsource selected housework to migrant women from less developed countries (Yeoh, 2012).
The International Labor Organization (ILO) conducted a study in 2013 on domestic workers worldwide. It estimated that the total number of domestic workers across the world is at least 52.6 million in 2010. Nevertheless, this number may be as high as 100 million as the estimation of domestic workers is based on official national household surveys and is therefore sensitive to measurement failures. The number of domestic workers in Germany was estimated to be 203,000 in 2009 (ILO, 2013), which is 0.5% of labor in the country. The registered
employment in private German households is 42,057 from which 5,707 are male and 37,350 are female (German Census Data). Other statistics record a number of 712,000 domestic workers. Variation between these figures could be due to estimation and measurement errors, the variation in assigning household tasks to domestic outsourcing or due to households not reporting the actual figures as many domestic workers are immigrants.
Empirical Studies
Empirical studies on the effect of employment on domestic outsourcing are scarce. Especially the external decision of households to demand domestic help is not well covered by literature. Nevertheless, some studies investigate the decision-making process of employing domestic help within a household. These studies highlight the gender-biased outsourcing decision of households by comparing men’s and women’s earnings on the expenditure for specific household tasks. Coltrane (2000) suggests that household’s decision for domestic help depends on the type of activities than men and women perform in the household. Gupta (2006,) allocates women’s time contribution in the household to cooking, cleaning, taking care of the children, doing dishes and laundry. Tasks that are typically the responsibilities of women are inflexible and have to be performed daily (Noonan, 2001). Tasks that men perform in the household are more flexible and recreational: mowing the grass and painting the house (Treas and de Ruijter, 2008; Schneider, 2012)
(I). The effect of earnings on the expenditure of domestic outsourcing
A woman with higher earnings reduces her share of housework (Killewald, 2011,). De Ruijter et al. (2005) study the earnings of men and women concerning the expenditure on domestic help regarding male and female household tasks. Although the contribution of both genders in the household is examined, the study lacks data on the time contribution of various tasks. As noted earlier, married women increased their labor force participation more than unmarried women. De Ruijter et al. (2005) did not find a link between marriage and household’s expenditure on domestic help. By comparing the expenditures of single men and single women on domestic help, the study tries to compare the effect of earnings of both genders to specific household tasks. The expenditures on domestic outsourcing of female-type tasks are higher for single men than single women. The expenditure on cleaning - a female-type task- is similar by men and women. Treas and de Ruijter (2008) studied the same relationship between earnings
and expenditure on domestic help, including the effect of marriage and the effect of women’s earnings on the outsourcing of gender-type tasks. Consistent with the study of De Ruijter et al. (2005) the study by Treas and de Ruijter (2008) did not find any effect of marriage on the overall household expenditure on male or female type tasks. Although the relationship between married men’s earnings and the outsourcing of female-type tasks in greater than for unmarried men. Contradicting with the outcome of De Ruijter et al. (2005) this study finds that the earnings of women in households are more positively related to the outsourcing of female type tasks than the earnings of men. Moreover, the higher the earnings of a woman the more a household spend on outsourcing male-type tasks. However, Treas and De Ruijter (2008) state that the earnings of women explain the expenditure on cleaning service more than the earnings of men. Moreover, De Ruijter et al. (2005) suggest that the flexibility of male-type tasks could lead to higher
expenditures on outsourcing these tasks. Gupta (2006) studied the relationship between the earnings of women and their time contribution in the household. The research finds that women’s earnings explain a larger part of the variation in time contribution in the household than men’s earnings. Moreover, the earnings of women are more positively related to the expenditure on domestic outsourcing than men’s earnings.
(II). The effect of Domestic Outsourcing on Female Labor Force: reversed causality.
An important interfering factor on the effect of female employment on domestic outsourcing is the reversed causality component: the employment of domestic help might facilitate job search for women. Empirical studies examining female labor force supply fail to find a relationship between household decision for outsourcing housework and female labor. The studies by Sullivan and Gershuny (2013) -who examine the substitution effect of domestic outsourcing on time allocation using UK survey data- Killewald (2011) with US data and de Ruijter and Lippe (2007) using Dutch data, all fail to find an effect of domestic workers on female labor force. Barone and Mocetti (2011) and Cortes and Tesseda (2011), focus on the substitution effect of immigrant women on time contributions by men and women within the household. Barone and Mocetti (2010) analyze the effect of immigration on hours of female labor supply. Immigration is linked to domestic outsourcing as female immigrants largely provide services in households. Survey data for Italian households from 2006 until 2008 illustrate that immigration has a positive impact on hours of female labor supply. The impact of immigration on female employment is not significant. Cortes and Tesseda (2011) examined the effect of immigration on the supply of female labor force. Time analysis with data of immigration for the 30 largest United States cities and female labor supply was conducted in 1980, 1990 and 2000. Women in high positions have increased their working hours due to the increase in immigration.
(III). Interfering Factors of the Link between Female Labor Force and Domestic Outsourcing
The increase in female labor force is linked to a decrease in time contribution of women in household (Artis and Pavalko, 2003). That is because housework is labor intensive (Bittman et al., 1999). Hereafter, other influence of the link between female labor and domestic help will be discussed. Firstly, technological progress have allowed women to allocate their time differently and could affect both female labor force as well as household’s domestic help. Secondly, the
allocation of housework responsibilities between men and women could influence both female labor force and domestic help (Cancedda, 2001,).
Coen-Pirani (2010) states that technological progress is the main driver of the increase in female labour force in the last half of the twentieth century. The introduction of home appliances and the quick decrease in price of these appliances allowed women to allocate their time differently (Cardia and Gomme, 2011). Data from the European Community Household Panel Survey of 2004 show that the incentive of women entering the labor market was the opportunity to reduce time spent on housework (Cancedda, 2001,). The share of married women of the increased labor force is greater than the share of unmarried women. Albanesi and Olivetti (2007) argue that technical and medical improvements related to motherhood allowed women to decrease time contribution in their maternal role allowing mothers to increase their labor force participation.
Literature links technological advancements which diminished women’s time contribution in the household to an increasing female labor supply (Artis and Pavalko, 2003; Bianchi et al., 2000; Coltrane, 2000). Women have decreased their time contribution in the household (Artis and Pavalko, 2003). Zick et al. (2008) state that until the 1960s the time that married women spend on housework has decreased with 14.1%. Although literature agrees that women decreased their time contribution in the household, the precise amount is ambiguous (Gardia and Gomme, 2011,). and women continue to spend more time doing housework than men (Killewald, 2011). Moreover, literature is not consistent on the amount of hourly decrease of women’s contribution (Gardia and Gomme, 2011). Ramey (2009) estimated a decline of time spend by women in the twentieth century to be 33% whereas Bryant (2006) argues that the decrease was 14%.
As technological progress allowed women to participate in the labor market, households were faced with a changing time-allocation of housework between men and women. Households in general have decreased their time contribution to housework (Bianchi et al., 2000). The trend of declining hours spend in the household has been observed since 1965. The difference between female and male time contribution in the household converged from 2000 until 2010 (Bianci and Milkie, 2010; Ramey, 2009). Coltrane (2000) argues that the employment of women explains a larger part of the change in housework than the employment of men. The employment of women negatively effects their time contribution in the household. Part-time employed women do not decrease their contribution in the household as much as full-time employed women. Noonan (2001) argues that time contributions of men and women in the household is more sensitive to female employment due to the nature of female tasks.
The hourly contributions to the household are higher for married women than unmarried women (Coltrane, 2000). What affects the observed difference in time allocation of men and women within households? Economic literature points to relative and absolute earnings of men and women as main causal factor for the observed change in unpaid household labor (Stancelli and Stratton, 2010). However, research does not agree about the effect of female earnings on the time contribution of men in the household. Although, Bianci et al. (2000) and
Treas and De Ruijter (2008) state that men did increase their share of housework, although not much. Married men perform less household work than unmarried men (Coltrane, 2000). However, Bianci (2000) found no link between marriage and time contribution in the household for men. Being employed has a slightly positive effect on hourly contributions in the household for men. Treas and de Ruijter (2008) state that men increased the time contribution in the household a little.
Methodology
This chapter provides a description of the data and the regression models used to analyze the effect of female employment on household’s domestic help. Household and personal survey data from the German Socio-Economic Data Panel (G-SOEP) is used to answer the research question. This chapter is organized as follows. First, the reason to use employment and not labor force in the empirical part of the thesis is explained; thereafter the concepts “domestic work” and “domestic workers” are explained. Second, details on the G-SOEP and variables will be explained. Third, the estimation technique will be discussed. Lastly, the models and method will be explained.
Concepts
(I). The Labor Force Participation Rate
The labor force participation rate is the relative number of people that are economically active over the people that are economically inactive (Psacharopoulos and Tzannatos, 1992). Economically active means employed persons and unemployed persons looking for work. The labor force participation rate is often ambiguous as ‘persons looking for work’ is hard to measure (Olivetti, 2013). This research aims to answer the question what the effect is of married women’s employment on household’s domestic employment. The labor force participation rate counts all women economically active; the employment rate counts all women that are currently employed. Hence, although literature on this topic has investigated ‘female labor force’, the labor force participation rate is not used in the empirical part of this thesis.
(II). Domestic Outsourcing
Domestic outsourcing refers to paid domestic help or paid employment in private
households (Sullivan and Gershuny, 2013). Domestic outsourcing is the substitution of unpaid housework for paid domestic labor (Bittman et al., 1999). Gultierrez-Rodriquez (2014) discusses domestic labor as affective labor including emotional and social orientation of work in private households. The International Labor Organization has come up with a definition for the terms ‘domestic work’ and ‘domestic workers’. Tasks performed by domestic workers vary across countries and are thus not specified in the definition (ILO, 2013). Article 1 of the Domestic Workers Convention 2011 defines domestic workers as following;
(a) the term “domestic work” means work performed in or for a household or households; (b) the term “domestic worker” means any person engaged in domestic work within an
employment relationship;
(c) a person who performs domestic work only occasionally or sporadically and not on an occupation basis is not a domestic worker.
According to the policy paper of the International Labor Organization (2009) the best way to count domestic workers is by an industry-based approach: the employed relationship with households. As noted earlier, the industry-based approach does not specify the tasks performed by domestic workers (ILO, 2013).
Data
(I). Data Set
The German Socio-Economic Data Panel (G-SOEP) is used for the empirical part of the thesis. During the years 1984-2012 the same 11,000 households are surveyed. The database includes personal and household survey data. This study uses the 2000-wave for year 1 regression. As noted in the introduction of the thesis, since the year 2000 the G-SOEP has increased in observations and therefore the year 2000 is the baseline year (Appendix A). Moreover, the 2000-wave of the G-SOEP contains the variable ‘cleaning lady, household help’. The 2001-wave is included to test for reversed causality.
(II). Conditions
As noted earlier, economic literature agrees that the share of married women in the labor force grew significantly faster than that of not married women (Goldin, 1994; Jacobsen, 1999). Although, the literature has failed to find an effect between marriage and household’s expenditure on domestic help, this study focusses on dual earning heterosexual households and all regressions are conditioned for married women (De Ruijter et al., 2005). To condition the regression for married women the dummy-variable ‘married’ is included and always 1 –yes- with every regression. Coltrane (2000) states that the employment of women explains a larger part of the change in housework than the employment of men. To condition the regressions for women only, the variable gender is included and always 1 –female- with every regression. The variable gender is retrieved from the personal database and includes 24,576 observations.
Variables
(I). Dependent Variable; Domestic Help
The dependent variable used for the analysis on household’s domestic help is ‘cleaning lady/household help’. This variable is a dummy and data are retrieved from the household dataset for the years 2000 and 2001. The number of observations is 13,248. Households answer yes or no on the survey question ‘do you occasionally or regularly employ household help?’ The variable is 1 for the answer yes. Consistent with the industry-based approach of measuring “domestic work” used by the ILO (2009) specific tasks performed in the household by domestic workers are not included in the data description.
(II). Independent Variable; Female Employment
‘Gainful employment’ is the variable used for female employment. This is a dummy-variable and data are retrieved from the personal dataset for the years 2000 and 2001. The question in the personal survey is: ‘are you currently in paid employment?’ Five answers could be given to answer this question. The thesis focuses on female employment and therefore all answers covering ‘not gainful employed’ are coded as 0 and all other answers are coded as 1. A dummy variable for employment is chosen and not hours of employment, because the latter is not necessary for the research. The research question requires the effect not the amount of employment on the demand for domestic help. However, to test for this omission of hours of employment the dummy ‘gainful employed’ is split for full-time and part-time gainful employed.
Moreover, prior literature agrees that the hours spend on employment negatively affect women’s time contribution in the household. Part-time employed women do not decrease their
contribution in the household as much as full-time employed women (Coltrane, 2000).
(III). Control Variables
The following variables are chosen as control variables as the decision to employ household help depends on household characteristics (Wang and Li, 2009). Although the thesis does not focus on the internal decision-making process of households, which is previously done by other literature, the variables size of the house, children and partner’s earnings have been linked by prior research to household’s decision to employ domestic help (Treas and de Ruijter, 2008).
1. Age. Data on the variable age are retrieved from the personal dataset. The number of observations is 32,453.
2. Size of the House. Data are retrieved from the household dataset for the years 2000 and 2001. The ‘size of the house’-variable is a continuous variable reporting the square meters of the house for each household. The values smaller and equal to zero are dropped, as these make no sense. The number of observations without negative values is 13,248.
3. The income of the partner. Dropping the earnings of women and keeping the earnings for men generates the variable ‘earnings of the partner’. The income of a household is positively related to the demand for domestic help (Treas et al., 2005).. Crucial here is that labor supply of men is not affected by domestic work. The variable is transformed to a log-variable to deal with the fact that the distribution of earnings is not normal and the model that is estimated is assumed to be linear. In other words; the relationship between earnings of the partner and household’s domestic help is not linear and the log transformation is used in order to estimate a linear relationship. Data on earnings are retrieved from the personal dataset for the years 2000 and 2001. The number of observations after transformation is 9051. 4. Children under 16 present in a household. The variable is retrieved from the
household dataset for the years 2000 and 2001. Prior research has shown that the presence of children contributes to a changing female labor supply. Married women with young children have increased their labor force participation more than married women with children over 18 years (Jacobsen, 1999). Therefore, the presence of children under the age of 16 in the household could possibly correlate with women’s employment and omission would cause omitted variable bias. The variable is a dummy and is 1 for yes. The number of observations is 13,248.
Estimation Technique
(I). Ordinary Least Squares estimation; the Linear Probability Model
Ordinary Least Squares (OLS) estimation is used to test if female employment –the independent variable- is a significant contributor to household’s demand for domestic help. As the dependent variable ‘domestic help’ is binary and the model is estimated using OLS, the model is called a ‘linear probability model’ (LPM). The dependent variable ‘cleaning lady/domestic help’ can take on the value 1 when a household employs domestic help and 0 if not. Since the
estimated model is a linear probability model, two important properties of OLS-estimation are present; linear conditional unbiased estimators and consistency. Due to the property of unbiasedness, the estimated regression coefficient of an independent variable is the change in probability that a household employs domestic help -takes on the value 1- given a change in the independent variable: E(Y) = Pr(Y=1, given the value of the independent variable’s). An example of an estimation regression model is E(Yt₁=1) = b₀ + b₁E₁ + b₂X₁ + ѵ t₁. Most independent variables included in the model are binary. Therefore, the estimated coefficients express the change in probability, due to the independent variable expressing 1, on household’s employment of domestic help (Y=1). The R-square is a poor measure for the linear probability model.
(II). Benefits and Downfalls of OLS
The benefit of using OLS with a binary dependent variable is that the estimated coefficients are easy to interpret. Therefore, the linear probability model is used a not logit or probit estimation. The linear probability model has three downfalls. Firstly, the errors of the estimation are heteroskedastic as the residuals are related to the value of the independent variables. The variance of the conditional distribution of the residuals is not constant. Secondly, the OLS-assumption that the errors are normally distributed is violated. Thirdly, as the fitted model is assumed to be linear there is no bound on the predicted values. However, probabilities cannot exceed 1 and the effect on the probability that the dependent variable is 1 (D=1) of a given change in the independent variable must be nonlinear. In contrast, the linear probability model could lead to predicted probabilities that exceed 1 or drop below 0.
Analysis
The regressions are performed as follows; Firstly, with data for the year 2000 the effect of female employment on domestic help is estimated. Thereafter, control variables for the year 2000 are included in the regression to eliminate possible omitted variable bias in the baseline model. Omitted variable bias occurs whenever a variable is not included but is relevant in explaining the dependent variable or whenever the omitted variable is correlated with the independent variable female gainful employment. Thirdly, fixed effects models will be estimated. Fourthly, the effect of married women’s employment on household’s domestic help in richer households will be estimated. After the estimation of all the baseline models, reversed causality models will be estimated to test whether the causality runs from female employment to demand for domestic help or if employment of domestic helps facilitate job search for women.
Model Specification
Variable Description Measurement Y cleaning lady/domestic help Y is {0,1} and 1 is yes E gainful employment E is {0,1} and 1 is yes Eᶠ gainful employment full-time Eᶠ is {0,1} and 1 is yes Eᵖ gainful employment part-time Eᵖ is {0,1} and 1 is yes AGE the age of a person in the survey AGE is {25,65} S size of the house Continues variable C children below age of 16 present C is {0,1} and 1 is yes I the income of the partner Continues variable t time period of data t-₁ is year 2000, t₁ is 2001
Models
The models described hereafter are population regression models that will be estimated. To identify the impact of female employment on domestic help two problems must be
acknowledged; firstly, it might be that women who work are very different from women who do not work and these differences may be related to the demand for domestic work and secondly, the employment of domestic help might facilitate women’s job search . The first problem is an omitted variable problem and the second a reversed causality problem. To address these problems a multiple variable model is included to test for possible omitted variables in the base-line model and a reversed causality model will test if the causality runs from employment to demand for domestic help or the other way around. The following models are tested and will be explained more thorough in the separated sections hereafter;
(1). Yt= β₀ + β₁E t + β₂Xt + µ + ɛt-₁ baseline model (2). ∆Y = β₀ + β₁ ∆E + β₂ ∆X + ɛ fixed effects model (3). Yt-₁ = β₀ + β₁Et₁ + β₂Xt₁ + ɛt₁ reversed causality model
Y domestic work {0,1}, E gainful employment {0,1}, X are control variables, ∆Y= Yt₁- Yt₂ change in domestic help, ∆E= E₁- E₂ change in employment, ∆X = X₁ - X₂ change in control variables, t-₁ is the year 2000, t₁ is the year 2001.
(1). Baseline Models
The baseline model tests the effect of gainful employment on household’s domestic help for the year 2000. The second baseline model tests gainfully employed full-time and gainful employed part-time separately to test whether the amount of hours of employment is significant in explaining household’s domestic help.
(1.1) Yt-₁= β₀ + β₁E t-₁ + β₂AGE t-₁ + ɛt-₁
(1.2) Yt-₁= β₀ + β₁Eᶠt-₁ + β₂ Eᵖt-₁ + β₃AGEt-₁ + ɛ t-₁
Y domestic work is {0,1}, E gainful employment is {0,1}, Eᶠ full-time is {0,1}, Eᵖ part-time is {0,1}, t-₁ is the year 2000
Multiple Variable Models
As noted previously, women who work might be different that women who do not and these differences might affect the demand for domestic help. Therefore, control variables -the variables size of house, number of children under 16 in the household and the income of the partner- are included in the model. This is a partial solution to the omitted variables problem as the true factors that affect the relationship between female employment and household’s domestic help are unknown and might be numerous.
(1.3) Y t-₁ = β₀ + β₁Et-₁ + β₂AGE t-₁ + β₃St-₁ + β₄Ct-₁ + β₅It-₁ + ɛt-₁
Y domestic work {0,1}, E gainful employment {0,1}, Eᶠ full-time {0,1}, Eᵖ part-time {0,1}, S size of the house is continuous, C children under the age of 16 present {0,1}, I income of the partner is continuous, t-₁ is the year 2000.
(2). Fixed Effects Model
The fixed effects model controls for omitted variables that are constant over time. The estimated model incorporates data from the years 2000 and 2001. Fixed effect model subtract factors that are time-invariant. The models (2.1) and (2.2) estimates the difference variables for year 2000 and 2001.
(2.1) ∆Y = β₀ + β₁ ∆E + ɛ (2.2) ∆Y = β₀ + β₁ ∆E + β₂∆X + ɛ
Y domestic work {0,1}, E gainful employment {0,1}, X are control variables, ∆Y= Yt₁- Yt₂ change in domestic help, ∆E= E₁- E₂ change in employment, ∆X = X₁ - X₂ change in control variables.
(3). Reversed Causality Model
The year 2001 is included in the model to test for reversed causality: whether the effect of female employment on domestic help is biggerthan the effect of domestic help on female employment. For comparability of the estimated coefficients, both models will test the effect of female employment on household’s domestic help.
(3.1) Yt₁ = β₀ + β₁Et-₁ + β₂AGEt-₁ + ɛt-₁
(3.2) Yt₁ = β₀ + β₁Et-₁+ β₂AGEt-₁ + β₃St-₁ + β₄ Ct-₁ + β₅ It-₁ + ɛ t-₁
(3.3) Yt-₁ = β₀ + β₁Et₁ + β₂AGEt₁ + ɛt₁
(3.4) Yt-₁ = β₀ + β₁Et₁ + β₂AGEt₁ + β₃St₁ + β₄Ct₁ + β₅I t₁ + ɛt₁
Y domestic work {0,1}, E gainful employment {0,1}, S size of the house is continuous, C children under the age of 16 present {0,1}, I income of the partner is continuous, t-₁ is the year 2000, t₁ is the year 2001.
Results
Descriptive Results
As shown in Table 1, in the year 2000 6.76% of all households employ domestic help. Moreover, the percentage of men employing domestic help is 5.42%, compared to 5.30% of women. Table 2 shows that 5.99% of the household’s in which the woman does not work employ domestic help. In accordance to the hypothesis, the percentage of household’s that employment domestic help is 7.17% when the woman does work. There is not much difference between full-time and part-full-time employed married women and ‘their’ demand for domestic help (Table 2). 7.24 % of the households with full-time employed married women have domestic help compared to 7.11% of the households for part-time employed women.
Table 1: Cleaning Lady/Household Help*
Dummy Frequency Percentage
0 11,775 93.24
1 854 6.76
*Data from the household survey, year 2000 Number of observations is 12,629
Table 2: Domestic Help and Gainful Employment for Married Women*
Employed Full-time Part-time
Total
No Yes No Yes No Yes Domestic Help No 2,147 1,528 2,957 718 2,865 810 3,675 Yes 137 118 199 56 193 62 225 Total 2,284 1,646 3156 774 3,058 872 3,930 %Yes 5.99 7.17 6.31 7.24 6.31 7.11 5.73
*Data on domestic help from the household survey and data on gainful employment from the personal survey, both for the year 2000
The amount of observations in the sample is 15,504. After controlling for married women the sample includes 5,571 observations. Table 3 shows the descriptive statistics of the sample used for the regressions. The mean of the variable domestic help is 5.83%. The mean of a binary variable is the proportion of observations that have code 1. Differently stated; 1 in 19 households employ -occasionally or regularly- domestic help. The majority of people that are employed work full-time.
Table 3: Descriptive Statistics for Married Women
Variable Mean Standard Deviation*
95% Confidence Interval** Observations
Domestic help .058 .003 .052 .064 5571 Employment .451 .007 .437 .464 5571 Full-time .197 .005 .186 .207 5571 Part-time .254 .006 .243 .265 5571 Age 43.6 .144 43.3 43.9 5571 Income partner 10.0 .025 9.98 10.1 1624 Children .571 .007 .558 .584 5571 Size House 111 .558 110 112 5571 *Standard errors are heteroskedastic-robust
**Binomial exact confidence interval
Regression Results
As shown in Table 4, a gainful employed woman increases the probability that a household employs domestic help with 2.66% (Figure 1). Although this seems not much, just 5.73% of the households were married women are employed have domestic help. Therefore, an increase in probability of 2.66% is a relative increase of 46.42% for household’s in which women are employed. The estimator is statistically significant at the 1% level. The employment hours of women differ in effect on the probability for domestic employment. As seen in Table 6, full-time employment increases the probability of domestic employment with 3.27% and part-time with 2.19%. Both are significant different from zero at the 1% level. The coefficient on employment becomes statistically insignificant after including control variables. Interesting is that the amount of children under 16 present in the household, increases the probability of household’s domestic help with approximately 4%. Every increase in square meter of household size, increases the probability with 0.11%; if families buy a house 50 square meters bigger than their old, the probability that this household’s employs domestic help increases with 5.50%.
Due to the inclusion of partner’s earnings in the regression, sample includes 1624 observations. Moreover, the variables partner earnings and size of the house change the
intercept dramatically. Linear estimation cause a very negative slope intercept. The estimated result for the baseline model is shown, omitting size of the house and partner’s income, including children (Table 5). As shown in Table 5 the increase in probability due to women’s employment is higher in richer households than in all households; respectively 5.28% and 3.21%. Striking is the effect of children under the age of 16 in richer households (Table 5). The probability of domestic help increases with 22.02% due to the presence of children under the age of 16. As noted before, the presence of children in all households was estimated to increase the probability with 4%. The estimated coefficient for children is significantly different from zero at the 1% level (Table 5).
Table 4: Baseline-Regression Results, for married women
Domestic Help OLS OLS OLS OLS
Intercept .026 (.137) .026 (.014) -.395 (.064)** -.341 (.045)** Gainful employment .027 (.006)** .015 (.011) Fulltime .033 (.008)** .018 (.014) Part-time .022 (.008)** .012 (.012) Age .00 (.00) .000 (.00)** .003 (.00)** .003 (.00)** Income partner .018 (.005)** .018 (.005)** Children .039 (.013)** .040 (.013)** Size of house .001 (.00)** .001 (.00)** R-Square 0.0034 .004 .008 .078 Adjusted R-square 0.003 .003 .008 .075 Observations 5571 5571 1624 1624
All estimated results are for married women; married=1 and sex=1. Standard errors are in italics
*significant at the 5% level **significant at the 1% level
Figure 1: Fitted values for domestic help and employment
Table 5: Effect Rich Households, for married women
All estimated results are for married women; married=1 and sex=1 Standard errors are in italics
*significant at the 5% level **significant at the 1% level
.0 3 .0 4 .0 5 .0 6 .0 7 F it ted v a lu es 0 .2 .4 .6 .8 1 werk2000
Domestic help OLS OLS
Intercept -.078 (.020)** .009 (.126) Gainful Employed .032 (0.006)** .052 (.036) Age .002 (.000)** .003 (.002) Children .057 (.008)** .220 (.040)** R-Squared .013 .050 Adj R-Squared .012 .046 Observations 5575 663
As shown in Table 6, without time variation effects, the probability of household’s employment of domestic help increases with 5.68% due to the employment of women. Women’s employment on the probability that household’s employ domestic help is higher than in the baseline regression where it was estimated to affect the probability with 2.66%. The coefficient is significant different from zero at the 1 percent level. The probability is estimated to be 4.21% due to female employment including control variables. However, this estimated coefficient is not significant. The amount of observations of the latter estimated regression is small; 495.
Table 6: Fixed Effects Regression Results, for married women
Domestic Help FE FE Intercept -2.824 (.009)** -2.953 (.070)** Gainful employment .057 (.016)** .042 (.049) Income partner .032 (.024) Children .041 (.030) Size of house .002 (.00)** R-Square .003 .034 Adjusted R-square .002 .026 Observations 4772 495 All estimated results are for married women; married=1 and sex=1 Standard errors are in italics
*significant at the 5% level **significant at the 1% level
Table 7 presents the estimated results for the reversed causality models. The probability that a household employs domestic help a year later decreases with -10.01% due to female employment now is as show in table 7. The probability that a household employs domestic help now increases with 1.17% due to female labor force a year later. This is counter-intuitive as female employment was expected to increase the demand for domestic help. Thus, the expectation of future employment causes an increase in probability that household’s employ domestic help now. This reversed causality method was intended to test if the effect of female employment indeed increased the probability that household’s employ domestic help. However, it actually showed that domestic help is more explanatory in the decision for women to
participate in the labor force.
Table 7: Reversed Causality, for married women
Domestic help (t) RC RC Domestic help (t-₁) RC RC
Intercept t-₁ 3.030 (.030)** 3.716 (.145)** Intercept t .033 (.014) * -.3489 (.0676)** Employed t-₁ -.100 (.014)** -.071 (.025)** Employed t .012 (.006) .015 (.012) Age t-₁ -.003 (.00)** -.004 (.001)** Age t .00 (.00)* * .003 (.001)** Income Partner t-₁ -.038 (.012)** Income Partner t .029 (.006)** Children t-₁ -.012 (.028) Children t -.017 (.014) Size of House t-₁ -.002 (.00)** Size of House t .00 (.00) R-Squared .013 .064 R-Squared .234 .027 Adj. R-Squared .012 .061 Adj. R-Squared .001 .016 Observations 4772 1406 5575 1361
All estimated results are for married women; married=1 and sex=1 Standard errors are in italics *significant at the 5% level
Conclusion
The results of the regression indicate that the effect of female employment increases the probability of domestic help with 2.66%. Therefore, married women’s employment is expected to increase domestic help by 2.66% in ‘their’ households. Compared to the percentage of
households employing domestic help this is a relative increase of 46.42%. The partner’s income as well as children under the age of 16 present in a household increases the probability that a household’s employs domestic help. Although a 3.98% increase in probability of domestic help in the household due to children present and a 1.84% due to a unit increase in the income of the partner does not seem much, it is much compared to the amount of household’s that employ domestic help. The relative increase due to children is 69.46% and due to partner income is 32.11% in households with married women.
Consistent with prior literature -which indicated that the earnings of both men and women in the household are positively related to household’s expenditure on domestic outsourcing (Treas and de Ruijter, 2008)- this thesis has found an greater effect of married women’s employment on household’s domestic help in richer household’s. Married women’s employment increased household’s domestic help with 5.28% in richer households, compared to 3.21% in all households. In richer household does the presence of children under 16 increases household’s domestic help with 22.02%, compared to approximately 4% in all households. This might indicate that richer households more often employ domestic help, possibly in the form of an au pair. No relevant difference is found between full-time and part-time employment of women of household’s domestic help: both increase household’s domestic help almost equally. Hence, the employment hours of women does not so much affect household’s domestic help. Prior research indicated that it might be inappropriate to treat women’s time in a household as homogenous (Killewald, 2011). Moreover, prior literature agrees that the hours spend on employment negatively affect women’s time contribution in the household. Part-time employed women do not decrease their contribution in the household as much as full-time employed women (Coltrane, 2000). This thesis implies –by investigating the effect of women’s employment on domestic help- that women’s time contribution in the household is or can be substituted by domestic workers. The implied substitution effect of domestic labor could in fact also be complementary effect of women’s time contribution in the household.
The expectation of future employment causes an increase in probability that household’s employ domestic help now. The probability that a household employs domestic help a year later decreases with -10.01% due to female employment now, compared to a 1.17% increase in probability that a household employs domestic help now due to female labor force a year.
Without any time-varying effects the employment of women increases household’s domestic help with 5.68%.
Discussion
This thesis shows that the expected future employment for women increased the probability that households employ domestic help now. Prior literature examined the impact of immigration of low-skilled labor to developing countries on the labor supply in developed countries (Barone and Mocetti, 2001; Cortes and Tessada, 2011). Globalization could possibly lead to a shift of low-skilled labor to developed countries and high-skilled labor to developing countries. The most important allocation of housework is between unpaid and paid women. Since 2013 the G-SOEP has added a migration sample. It is the largest survey of immigrants and might lead to some interesting insights in macro-shifts of employment. Suggestions for further research would to explore the external effect of immigration on household’s employment decision.
Problematic is the amount of observations in the sample of the baseline regression after including control variables, which drops from 5571 to 1624. A small sample size -1624
observations- might not justify statements about the population.
Data on household’s employment of domestic workers might be inaccurate. Official statistics do not include undocumented domestic workers. The International Labor Conference (2010, 3) estimates that although the German Socio-Economic data panel for the year 2000 recognizes 40,000 domestic workers, the total amount could sum up to 1,1 million (Schupp, 2002). Also, most domestic workers work in multiple households a day, making the Official Statistics less accurate (Helen Lutz).
Appendices
Appendix ASource: The German Institute for Economic Research (DIW)
Appendix B
Figure 1: U-Shaped Female Labor Force (1890-2005)
Figure 2: German Labor Force
Source: Federal Reserve Economic Data
Figure: Labor Force Participation Rate for Men and Women
Appendix C
Table 1: Predicted Value Residuals
-1 0 1 2 3 Re s id ua ls -3.2 -3 -2.8 -2.6 -2.4 Fitted values
Bibliography
Albanesi, Stefania, and Olivetti, Claudia. 2007. ‘Gender Roles and Technological Progress’. Bureau
of Economic Research Working Paper.
Anderson, Bridget. 2001. ‘Just another Job? Paying for Domestic Work’. Gender and Development
Vol. 9 No. 1: 25-33.
Artis, Julie E., and Pavalko, Eliza K. 2003. ‘Explaining the Decline in Women’s Household Labor: Individual Change and Cohort Differences’. Journal of Marriage and Family Vol. 65 No. 3: 746-761.
Autor, D, Levy, F, Murnane. 2003. ‘Skill Content of Recent Technological Change an Empirical Exploration’. Quarterly Journal of Economics 118(4): 1279-1333.
Barone, Guglielmo, and Mocetti, Sauro. 2011. ‘With a Little Help from Abroad: The Effect of Low-Skilled Immigration on the Female Labour Supply’. Labor Economics 18: 664-675. Bianchi, Suzanne M., et al. 2000. ‘Is Anyone Doing the Housework? Trends in the Gender Division
of Household Labor’. Social Forces Vol. 79 No. 1: 191-228.
Bianci, Suzanne M., and Milkie, Melissa A. 2010. ‘Work and Family Research in the First Decade of the 21 Century’. Journal of Marriage and Family 72: 705-725.
Bianchi, Suzanne M., et al. 2012. ‘Housework: Who Did, Does or Will Do It, and How Much Does it Matter? Social Forces Vol. 91 No. 1: 55-63.
Bittman, Michael, et al. 1999. ‘The Changing Boundary Between Home and Market: Australian Trends in Outsourcing Domestic Labor’. Work, Employment and Society Vol. 13 Issue 2: 249-273.
Blau, Francine D., and Kahn, Lawrence M. 2012. ‘Changes in the Labor Supply Behavior of Married Women: 1980-2000’. Cornell University, Centre of Economic Studies and
Economic Research and Institute for the study of Labor.
Bryant, W. Keith. 2006. ‘Does Household Work Matter Anymore? Comparison of Household Production and the Distribution of Income in the United States in 1965-66 and 2003’. Cancedda, Alessandra. 2001. ‘Employment in the Household Services’. European Foundation for
the Improvement of Living and Working Conditions.
Cavalcanti, Tiago, and Tavares, Jose. 2008. ‘Assessing the “Engines of Liberation”: Home
Appliances and Female Labor Force Participation. The Review of Economics and Statistics
Vol. 90. No.1. The MIT Press: 81-88.
Cardia, Emanuele, and Gomme, Paul. 2011. ‘The Household Revolution: Childcare, Housework, and Female Labor Force Participation’. Economic Research Centre CIREQ.
Coltrane, Scott. 2000. ‘Fatherhood and Marriage in the 21st Century’. National Forum No. 80:
25-28.
Cortes, Patricia, and Tessada, Jose. 2011. ‘Low-Skilled Immigration and the Labor Supply of Highly-Skilled Women’. American Economic Journal: Applied Economics 3: 88-123. Coen-Pirani et al. 2010. ‘The Effect of Household Appliances on Female Labor Force Participation
Fogli, Alessandra, and Veldhuis, Laura. 2011. ‘Nature or Nurture? Learning and the geography of Female Labor Force. Econometrica Vol. 79 No. 4: 1103-1138.
Goldin, Claudia. 1994. ‘The U-shaped Female Labor Supply Function in Economic Development and Economic History’. National Bureau of Economic Research, Working Paper No. 4707. Goldin, Claudia, and Olivetti, Claudia. 2013. ‘Shocking Labor Supply: A Reassessment of the Role
of World War II on Women’s Labor Supply’. American Economic Review.
Gorban, Debora, and Tizziani, Ania. 2014. ‘Inferiorization and Deference: The Construction of Social Hierarchies in the Context of Paid Domestic Labor’. Women Studies International
Forum.
Gultierrez-Rodriquez, Encarnacion. 2014. ‘Domestic Work-Affective Labor: On Feminization and the Coloniality of Labor’. Women Studies International Forum.
Gupta, Sanjiv. 2006. ‘Her Money, Her Time: Women’s Earnings and Their Housework Hours’.
Social Science Research 35: 975-999.
Hank, Karsten. 1998. ‘Household Labor Demand and Household Labor Supply: An Empirical Analysis of the Employment of Domestic Help in German Private Households and its Effect on Female Labor Force Participation’. Centre for Policy Research Syracuse
University.
Heimeshoff, Lisa-Marie, and Schwenken, Helen. 2010. ‘Domestic Workers Worldwide. Summary of Available Statistical Data and Estimates’. International Labor Domestics Workers
Network and International Labor Conference, 99th Session.
International Labor Organization (ILO). 2009. ‘Global and Regional Estimates on Domestic Workers’. Domestic Work Policy Brief 4.
International Labor Organization (ILO). 2013. ‘Domestic Workers Across the World: Global and Regional Statistics and the extend of Legal Protection’. International Labor Office, Geneva. Jacobs, Jerry A., and Gerson, Kathleen. 2008. ‘The Time Divide. Work, Family and Gender
Inequality’. Library of Congress Cataloging-in-Publication Data: 41-47.
Jacobsen, Joyce. 1999. ‘Labor Force Participation’. The Quarterly Review of Economics and Finance
39: 597-610.
Killewald, Alexandra. 2011. ‘Opting Out and Buying Out: Wives’ Earnings and Housework Time’.
Journal of Marriage and Family 73: 459-471.
Lutz, Helma. 2008. ‘When Home Becomes a Workplace: Domestic Work as an Ordinary Job in Germany’ . In Migration and Domestic Work: A European Perspective on a Global Theme, 1st ed., edited by Helam Lutz, 42-60. Ashgate Publishing Limited, 2008.
Noonan, Mary C. 2001. ‘The Impact of Domestic Work on Men’s and Women’s Labor Force Participation and Earnings’. Journal of Marriage and Family 63: 1134-1145. Olivetti, Claudia. 2013. ‘The Female Labor Force and Long-Run Development: The American
Experience in Comparative Perspective’. Boston University and National Bureau of
Economic Research Working Paper No. 19131.
Osnowitz, Debra. 2005. ‘Managing Time in Domestic Space: Home-Based Contractors and Household Work’. Gender and Society Vol. 19 No.1: 83-103.
Psacharoploulos, George, and Tzannatos, Zafiris. 1992. ‘Latin American Women’s Earnings and Participation in the Labor Force’. Policy Research Working Paper Series 856, The World
Bank.
Pissarides, Christopher, et al. 2003. ‘Women in the Labour Force: How well is Europe Doing?’
Oxford University Press.
Ramey, Valeria A. 2009. ‘Time Spend in Home Production in the Twentieth-Century United States: New Estimates from Old Data’. The Journal of Economic History Vol.69 No.1. de Ruijter, Esther, et al. 2005. ‘Outsourcing the Gender Factory: Living Arrangements and Service
Expenditures on Female and Male Tasks. Social Forces Vol. 84 No 1.
Stancanelli, Elena, and Stratton, Leslie Sundt. 2010. ‘Her Time, His Time, or the Maid’s Time: An Analysis of the Demand for Domestic Work’. Working Paper for the Forschungsinstitut zur
Zukunft der Arbeit No. 5253.
Schneider, Daniel. 2012. ‘Gender Deviance and Household Work: The Role of Occupation’.
American Journal of Sociology Vol. 117 No. 4: 1029-1072.
Sullivan, Oriel, and Gershuny, Jonathan. 2013. ‘Domestic Outsourcing and Multitasking: How Much do They Really Contribute?’ Social Science Research 42: 1311-1324.
Treas, Judith, and de Ruijter, Esther. 2008. ‘Earnings and Expenditures on Households Services in Married and Cohabiting Unions’. Journal of Marriage and Family 70: 796-805.
Wang, Donggen, and Li, Jiukun. 2009. ‘A Model of Household Time Allocation Taking into Consideration Of hiring Domestic Helpers’. Transportation Research Part B 43: 204-216. Yeoh, BSA., and Huang, S. 2012. ‘Home: Paid Domestic Labor’. National University of Hong Kong,
Singapore: 451-455.
Zick, Cathleen D., et al. 2001. ‘Mother’s employment, Parental Involvement, and the implications for Intermediate Child Outcomes’. Social Science Research Vol. 30: 25-49.
Zick, Cathleen D., et al. 2008. ‘Does Household Work Matter Anymore? Household production and the Distribution of Income in the United States in 1965-66 and 2003. Review of the
