Effect of Fertility on Female Labour Force Participation in the United Kingdom

Abstract

This article aims to estimate the causal effect of fertility on the labour force participation of British women. In order to account for the possible endogeneity of fertility variables, the sex composition of the two previous children of a mother has been used as an instrument for fertility. The Two-stage Residual Inclusion (2SRI) estimation method is used on a relatively new British data set—‘Understanding Society’. The results suggest that fertility is indeed endogenous to the labour force participation decisions of women in the sample, and that not accounting for the endogeneity of the fertility variable leads to an exaggeration of the negative effect of fertility on female labour force participation. Important policy recommendations are drawn from the results.

JEL Classification: J13, J22

Keywords

Female Labour Force Participation Fertility Endogeneity Instrumental Variables Two-stage Residual Inclusion

1. INTRODUCTION

Since 1970, most developed nations have witnessed an impressive rise in female education levels and female labour force participation and a decline in fertility. However, even though the gender wage gap has also narrowed over the years in these countries, women still earn less and provide lower labour supply than men. The primary reason for this is that women are responsible for childbearing and a significant part of childrearing, and female labour force participation is largely dependent on the number of children, especially young children. Most studies have found that a mother’s labour force participation is significantly negatively affected by the number of children. The magnitude of this negative correlation, however, varies across countries, ethnicity, age, religion, marital status, education levels, etc.

Estimating the effect of fertility on female labour market outcomes is not a straightforward exercise. A woman’s labour force participation decision and her fertility decision could be driven by unobserved variables—such as personal preference regarding career and children. Also, it is possible that her labour force participation decision and fertility decision are determined together—such as, a woman planning her family, keeping her present income and future career outcomes in mind. This may lead to what is known as the ‘endogeneity’ of fertility variables. If these are not accounted for, then the estimated result cannot be read as the causal effect of fertility on female labour force participation.

Many studies have used instrumental variable estimation to account for the endogeneity of female fertility variables, and the instruments used have been varied, such as, twin births, infertility, her education level and length of marriage. It may be argued that some of these instruments are not ‘good’ or valid instruments, in the sense that they may be correlated with both fertility and labour market outcomes (Nakamura and Nakamura, 1992). For example, a woman’s religious beliefs, her marriage duration and her education level may affect both her fertility and labour supply decisions. Furthermore, instruments, such as, twin births and infertility, can only be applied to very large data sets due to their relatively rare occurrence.

In this context, Angrist and Evans (1998) made an extremely valuable contribution to the literature by using the sex composition of the first two children of a woman as an instrument for the birth of her third child. Their reasoning comes from the well-researched and long-understood finding in demography literature that parents in developed countries prefer to have a ‘balanced family’: a woman whose first two children are of the same sex, is more likely to have a third child, than a woman with a son and a daughter. Thus, the sex composition of the first two children acts as an exogenous ‘shock’ to a woman’s fertility. The sex composition of the children is also random and has no impact on a woman’s labour market behaviour, other than through its effect on fertility.

This empirical research aims to estimate the causal effect of an additional child¹ on British women’s labour force participation. In order to account for the possible endogeneity of fertility variables, the Angrist and Evans (1998) methodology² has been followed—the sex composition of a woman’s first two children has been used as an ‘instrument’ for her decision to have another child. The two-stage residual inclusion (2SRI) estimation method (Terza et al., 2008) has been used, along with a standard probit estimation for comparison (shown in the Appendix 4). The data for the analysis comes from the relatively new British longitudinal study, Understanding Society.

This research makes certain important contributions to the existing literature. To the best of my knowledge, till recent times,³ the Angrist and Evans (1998) methodology has been used only once before using a British data set. Iacovou (2001) estimated the impact of fertility on British women’s labour supply and hours of work using conventional linear instrumental variables estimation. This article extends the literature by using a different econometric estimation technique on British data. Also the data set used in this analysis is relatively new and, to the best of my knowledge, has not been used previously for this particular purpose.

The remainder of the article is organised as follows: Section 2 provides an account of the existing literature on the topic at hand. Section 3 gives an overview of the female labour force participation and fertility statistics in the UK, in order to give a clear understanding of the research. Section 4 discusses the theoretical framework of my research. Section 5 discusses the econometric framework that has been used in my analysis. Section 6 discusses the data and its sources, and provides relevant descriptive statistics. Section 7 presents the results of my research, and all the sensitivity analyses and section 8 discusses policy recommendations that are drawn from my results and concludes my research. References and the appendices follow Section 8.

2. LITERATURE REVIEW

Most studies find that labour force participation of women is negatively correlated with the number of children. For European countries, Joshi and Davies (1992) find that mothers with two children lose approximately 10 years of employment time in Germany, 8 years in the UK, 2 years in Sweden, etc. Earlier literature often assumed fertility to be an exogenous variable in the female labour supply and earnings functions, such as, Gronau (1973), Heckman (1974) and Heckman and Willis (1977). Later literature obviously recognised fertility to be endogenous, such as, Schultz (1990), Goldin (1994) and Xie (1997), and used several different methods to account for the endogeneity problem, the most popular method being instrumental variables (IV) estimation.

‘Twin Births’, a popular instrumental variable in this subject area, has been used by Rosenzweig and Wolpin (1980), Gangadharan et al. (1996), Jacobsen et al. (1999), etc. Twins result in an unanticipated additional child, and this provides an exogenous variation to the family size. However, since twin births are relatively rare occurrences,⁴ it is not a very effective instrument for small data sets.

As an excellent addition to the economic literature, Angrist and Evans (1998) used the sex composition of a woman’s first two children as an instrument for fertility. This strategy is based on the finding in demography literature that parents in developed countries have a preference for a mixed sex composition for their children. Ben-Porath and Welch (1976) and Nancy Williamson (1983) provide extensive research on the sibling sex preferences in developed countries.

Angrist and Evans (1998) opines that if the first two children in a family belong to the same sex, that family has a greater probability of having a third child, than a family with a mixed sex composition for the first two children. Thus, sibling sex composition is a valid instrument because it is correlated with fertility variables, but has no correlation with female labour market behaviour. Using US census data, they confirm that children do indeed reduce the labour supply of women, though the effect is nil for highly educated women and women with high-income husbands.

The Angrist and Evans (1998) methodology has been implemented in several later research for developed and developing nations alike: Iacovou (2001) for the United Kingdom, Chun and Oh (2002) for Korea, Ebenstein (2007) for Taiwan, Daouli et al. (2009) for Greece, Maurin and Moschion (2009) for France, Hirvonen (2009) for Sweden, Cruces and Galiani (2007) for Argentina and Mexico and so on.

3. OVERVIEW OF FEMALE LABOUR FORCE PARTICIPATION AND FERTILITY IN THE UK

3.1 UK Female Labour Force Participation

The UK is one of the many developed countries in the world, along with the Scandinavian countries and the US, where the rate of female labour force participation began to rise significantly after the Second World War. In fact, particularly between 1980 and 2000, female labour force participation increased manifold, mainly due to changing attitudes towards employing women, increased education and opportunities for women, improvements in childcare options, low unemployment levels for women, etc. Table 1 shows the trends in female employment rates in several OECD countries from 1975 to 1997.

Of course, labour force participation rates for women vary according to age group, education levels, marital status, number of children in the household, etc.

Table 1
Trends in Female Employment Rates 1974–97 (OECD Countries)

Countries 1974 1980 1997

Sweden 64.9 76.0 73.2

United Kingdom 54.3 58.3 67.2

United States 52.3 61.5 71.3

Portugal 51.2 54.3 65.1

Germany 50.6 52.8 61.8

France 50.6 56.0 59.9

Norway 50.0 64.2 75.5

Canada 48.5 58.4 68.5

Belgium 42.4 45.3 57.4

Ireland 34.2 35.8 50.4

Italy 33.7 39.2 44.1

The Netherlands 29.7 36.3 61.8

Spain 27.6 32.2 47.1

Countries	1974	1980	1997
Sweden	64.9	76.0	73.2
United Kingdom	54.3	58.3	67.2
United States	52.3	61.5	71.3
Portugal	51.2	54.3	65.1
Germany	50.6	52.8	61.8
France	50.6	56.0	59.9
Norway	50.0	64.2	75.5
Canada	48.5	58.4	68.5
Belgium	42.4	45.3	57.4
Ireland	34.2	35.8	50.4
Italy	33.7	39.2	44.1
The Netherlands	29.7	36.3	61.8
Spain	27.6	32.2	47.1

Source: Labour Force Statistics, OECD (1999).

As Table 2 shows, there have been substantial changes in female labour force participation from 1984 to 2002. Labour supply of married women increased from 62 per cent in 1984 to 74 per cent in 2002, while the labour supply of single women reduced from 76 per cent to 68 per cent in the same time-frame. Married women with no children have had a higher labour force participation rate than married women with children. But, married women with only one dependent child have had a higher participation rate compared to married women with two or more children. Also as expected with increasing education levels for women, women with degrees have a greater participation rate, than women without any degrees.

Table 2

Female Participation Rates (% of total population) by Category (1984–2002)

Years	Total	Marital Status		Dependent Children, Married				Qualifications
Years	Total	Married	Single	One	Two	Three +	None	Degree	None
1984	65	62	76	61	55	39	71	79	58
2002	72	74	68	77	74	55	79	88	48

Source: Bank of England Working Paper No. 221 (2004).

After 2002, female labour force employment, as a percentage of total female population aged 15–64, has largely remained static, with very small variations: in 2002, 68.5 per cent of the total female population was in paid employment, and in 2011, the figure is at 69.5 per cent.⁵ This largely suggests that female labour force employment seems to have reached a figure from which it is unlikely to budge in the near future. Additionally, the economic downturn post 2008 has led to the unemployment of several women, adversely affecting participation levels.

3.2 Fertility in the UK

According to the Office for National Statistics (ONS), the population in the UK is estimated to rise from approximately 63 million in 2010 to nearly 67.2 million in 2020, and then to nearly 70 million in mid-2027. The primary reasons for this are: more births than deaths, and net migration to the UK.

From 1965 to 2000, the total fertility rate (TFR)⁶ has more or less been on a declining trend in all four constituent countries of the UK. However, from the early twenty-first century onward, there has been an upturn in fertility rates, as shown in Figure 1.

Figure 1 shows a very sharp fall in the TFR in 2009 in all four constituent countries, which can be accounted for by the global recession.

Figure 1

Source: Office for National Statistics (ONS), Fertility Summary 2010.

According to the Office for National Statistics (ONS), over the past two decades, the average age of birth for British women has been steadily increasing. This could be caused by various factors: increased educational opportunities, changing attitudes towards employment and families, etc. Figure 2 suggests that in 2010, in Northern Ireland the average woman gave birth at the age of 30 years; in Scotland and England the corresponding age was approximately 29 years, while the average woman in Wales gave birth at approximately 28 years. A decade ago, in all four constituent countries, fertility rates were the highest for women in the age group 24–28 years (ONS, Fertility Summary, 2010). However, in 2010 and after that, fertility rates are highest for women in the age group of the early thirties.

4. THEORETICAL FRAMEWORK

In my analysis, I assume that the representative woman is a rational consumer, whose employment decision is determined on the basis of the utility she derives from working versus not working. Her labour supply is modelled keeping in mind the neo-classical (static) labour supply model.

Figure 2

Source: Office for National Statistics (ONS), Fertility Summary 2010.

The representative woman’s utility function is written as: U = U (L, C, Z),

where L is leisure time in hours, C is consumption and Z is the set of non-labour, social demographic factors. She maximises her utility subject to the following budget constraint: C = wH,

where w is the given market wage rate, and H stands for the hours of work. The model does not include any non-labour income.

The following relation is given between leisure and labour time: H = T – L,

where T is the total amount of available time in hours.

The solution to the utility maximisation problem gives the optimal level of labour supply as H₀ = H (w, Z). Hence, a woman’s supply of labour is determined by the wage rate and other socio-demographic factors. She has the freedom to choose her hours of work and the sector she works in.

A woman’s labour supply is likely to change in response to the birth of her child from the initial supply, but the direction of this change is ambiguous. Since a child consumes some part of total household income, a birth reduces the income of the household, ceteris paribus. This leads to an increase in the labour supply of either or both parents (the income effect). If both parents are working, additional childcare costs must be incurred, which leads to a fall in household income. A mother can respond by either increasing her labour supply to meet the childcare costs (the income effect) or by decreasing her labour supply (substitution effect) to reduce childcare costs. She can even stop supplying labour altogether. Hence, the effect of an additional child on a mother’s labour supply is likely to be ambiguous, and cannot be determined a priori.

5. ECONOMETRIC FRAMEWORK

Labour force participation is a binary outcome in the sense that a woman either works for pay or does not work for pay. For limited dependent variables, most econometricians increasingly prefer to use nonlinear regression models (instead of linear regression models) due to the greater suitability of nonlinear models to model limited dependent variables (Terza et al., 2008). The most popular models are the probit and logit models, which are estimated using maximum likelihood estimation.

Following the reasoning of Terza et al. (2008), instead of using a traditional linear instrumental variables (IV) estimation technique to account for the endogeneity problem in my research, I have used the Two-stage Residual Inclusion (2SRI) estimation method.⁸ Formally, the first-stage (reduced form) equation and the second-stage (outcome) equation of our analysis (both probit equations) are given as follows:

First-stage equation: _i(1)₀₁_i₂_iu_i

Second-stage equation: _i(2)₀₁_i₂₃ $\overset{}{u}$ _ie_i

Here, in equation (1), Thirdchild_i*_{is a latent (unobserved) variable that measures the i’th woman’s inclination towards having a third child; SameSex_i is a binary variable indicating whether the i’th woman’s first two children are of the same sex or not; A_i is the vector of other (socio-demographic) factors determining the i’th woman’s decision to have a third child; and u is the error of the first-stage regression. The variable that we do observe is Thirdchild_i, a binary variable indicating whether the i’th woman has a third child or not.}

Thus, _i(3)_i

where Thirdchild_i = 1 implies that the i’th woman has a third child, and Thirdchild_i = 0 implies that she does not.

Similarly, in equation (2), LFP_i* is a latent variable that measures the i’th woman’s inclination towards paid employment; Z_i is the vector of socio-demographic characteristics determining the i’th woman’s labour force participation decision; $\overset{}{u}$ is the predicted residual from the first-stage regression, and e is the error of the second stage regression. The variable that we do observe is LFP_i, which is a binary variable indicating the i’th woman’s labour force participation status.

_i(4)_i

where LFP_i = 1 implies the woman is in paid employment, and LFP_i = 0 implies that she in not.

A significant benefit of using the 2SRI method is that α₃, the coefficient of the predicted residual from the first stage regression, acts as a diagnostic tool to test for the endogeneity of the ‘Thirdchild’ variable. If α₃ is statistically significant, we reject the null hypothesis that the ‘Thirdchild’ variable is exogenous to the labour force participation decision, and we consider it to be endogenous. Otherwise, if α₃ is not statistically significant, then we cannot reject the null hypothesis that the ‘Thirdchild’ is exogenous and the outcome equation can be estimated using conventional nonlinear regression methods (Bollen et al., 1995).

The i’th woman’s labour force participation decision and decision to have a third child depend on a number of socio-demographic characteristics, which are included in the vectors Z_i and A_i, respectively.

Characteristics included in vector A_i are: (i) the i’th woman’s age and (ii) her age at her first childbirth. The age of the mother can be expected to have a negative effect on the decision to have a third child. The mother’s age at first birth is also likely to influence her decision to have a third child: for example, a woman who had her first child at 35 years of age is likely to make a different decision from a woman who had her first child at the age of 20 years.

Characteristics included in the vector are Z_i are: (i) the i’th woman’s age, (ii) her years of education, (iii) her marital status, (iv) the age of her youngest child in the household, (v) her health status, that is, whether she is in good health (i.e., health conducive to employment) or poor medical health, (vi) whether she has access to external childcare facilities and (vii) her (predicted) labour earnings.

The age of the youngest child in the household is likely to have a significant effect on the labour force participation decision of the mother, in the sense that the presence of a young child in the household will possibly lower a mother’s labour supply. Mothers with older children can be expected to have a greater labour supply than mothers with young/infant children. On the other hand, access to external childcare facilities is likely to have a positive effect on a mother’s labour supply. A positive relation can be expected between the mother’s years of education and her labour force participation.

Several studies, while trying to estimate the causal effect of fertility on female labour force participation, do not consider female labour earnings as an explanatory variable. My theoretical framework suggests that the wage rate, ‘w’, affects a woman’s labour supply, so female labour earnings is included as an explanatory variable in the analysis. However, in this context, there are two problems: (i) labour earnings are not observed for women who do not work, and (ii) labour earnings are possibly correlated with various latent variables that affect fertility and labour supply decisions together. To get around these problems, the Heckman (1979) sample selection procedure is used to assign predicted wages to non-working women, but using actual wages for working women and predicted wages for non-working women will give two different wage distributions, thus the study uses predicted wages for all women in the sample for the sake of simplicity.⁹

6. DATA AND DESCRIPTIVE STATISTICS

Iacovou (2001) estimated the causal effect of fertility on female labour force participation and hours of work, using two British data sets—the National Child Development Study and British Household Panel Survey. For this research, I have used a relatively new data set, ‘Understanding Society’—a British longitudinal survey designed by the Institute for Social and Economic Research (ISER). In it, 40,000 households across England, Scotland and Wales were surveyed as a representative sample of the population of the UK. This study uses data from the survey’s Wave 1 (surveyed in 2009–11) and parts of Wave 2 (surveyed in 2010–11).¹⁰ It may be a concern that utilising the Angrist and Evans (1998) strategy limits the sample to women with two or more children, but this is a necessary sacrifice considering the strategy gives strength to the analysis: sex composition is correlated with a mother’s fertility behaviour, but not her labour market behaviour.

The empirical study began with data on 50,994 men and women. After removing the following: men, women without children, women with less than two children, women under the age of 20 years and above 45 years, women whose youngest children were of 18 years or above,¹¹ and self-employed women, the sample had 4,764 women aged 20–45 who have two or more children. I also removed any observations with incomplete or missing data.

In the sample, 2,521 out of the 4,764 (52.92 per cent) women have positive hours of work, that is, are in paid employment.¹² Most of them (3,654 out of 4,764 women) live with a partner in the household—either a spouse or a non-married partner. Instead of dividing the sample into married and unmarried women, or focusing only on married women, the study chooses an alternate way to identifies ‘marital status’. The women have been divided into two groups, according to whether they live alone or with a partner in the household, as a large percentage of women in the United Kingdom have children outside marriage and live with their non-married partners, and restricting attention to only married women would severely limit the sample.

Of the 4,764 women, 2,148 had more than two children, while the rest (2,616 women) had only two children. Of the 2,148 women, 1,646 (76.63 per cent) had more than two children when their first two children belonged to the same sex, while only 23.37 per cent had more than two children even after having a mixed sex composition for their first two children. Thus, more women in my sample had gone on to have a third child when their first two children belonged to the same sex.

The variable ‘Same Sex’ must only be correlated with a mother’s fertility decision, not her labour force outcomes. In the sample, of the total 2,521 women who are in paid employment, 55.53 per cent had a same sex composition for their first two children, while 44.47 per cent had a mixed sex composition for their first two children. Thus, there does not seem to be any significant association between the probability of a mother’s working decision and the sex composition of her first two children. Hence, sex composition seems to be a valid instrument for my analysis.

7. RESULTS AND SENSITIVITY ANALYSIS

The two-stage residual inclusion (2SRI) estimation was conducted, treating the decision to have a third child as endogenous.¹³ Since the estimates of the coefficients of any probit model themselves cannot be meaningfully interpreted, marginal and partial effects (of both the first-stage and second-stage regressions) were computed. The first-stage probit regression’s relevant results are shown in Table 3.

We see that on average there is an increase in probability of having a third child by 33.4 percentage points, if the previous two children are of the same sex, ceteris paribus.¹⁴ The marginal and average partial effects of the second-stage regression are shown in Table 4.

Table 3
Marginal Effects and Average Partial Effects of the First-stage Probit Regression (effect on the probability of having a third child)

(1) (2)

Same sex (first two children are of the same sex) 0.359(0.014) 0.334(0.013)

Age of mother 0.0319861(0.002) 0.027(0.001)

Age of mother at her first childbirth −0.042(0.002) −0.035***(0.002)

	(1)	(2)
Same sex (first two children are of the same sex)	0.359***(0.014)	0.334***(0.013)
Age of mother	0.0319861(0.002)	0.027***(0.001)
Age of mother at her first childbirth	−0.042***(0.002)	−0.035***(0.002)

Notes: (1) shows the marginal effects and (2) the average partial effects of the given variables on the probability of having a third child. Standard errors are in parentheses.

*** denotes significance at the 1 per cent level, ** at the 5 per cent level and * at the 10 per cent level; heteroscedasticity robust standard errors have been used in all the regressions.

Table 4

Marginal Effects and Average Partial Effects of the Second-stage Probit Regression (effect on the probability of labour force participation)

	(1)	(2)
Third child	–0.029 (0.044)	−0.022 (0.033)
Age of mother	0.002(0.010)	0.001(0.007)
Years of education	−0.026(0.078)	−0.019(0.058)
Marital status	0.179***(0.019)	0.133***(0.015)
Age of youngest child	0.033***(0.026)	0.025***(0.002)
Health status of mother	0.313***(0.043)	0.232***(0.034)
External childcare access	0.460***(0.019)	0.342***(0.011)
Predicted earnings	0.051(0.046)	0.038(0.039)
Residual of first-stage regression	−0.086*(0.046)	−0.064*(0.035)
Observations	4,764	4,764

Notes: (1) shows second-stage marginal effects and (2) shows the average partial effects of the given variables on the probability of positive hours of work (labour force participation of the mother); standard errors are in parentheses; *** denotes significance at the 1 per cent level, ** at the 5 per cent level and * at the 10 per cent level; bootstrapped standard errors have been used in all the regressions.

The 2SRI second-stage estimation results suggest that when the endogeneity bias is accounted for, there is a negative relationship between having a third child and the labour force participation of the mother. On average, having a third child reduces the probability of female labour force participation by approximately 2.2 percentage points, but it is not statistically significant at any level. Furthermore, the coefficient of the first-stage residual is statistically significant at the 10 per cent level, implying that the decision to have a third child cannot be considered an exogenous variable in the labour force participation decision and hence the use of the 2SRI method is justified.

Marital status of the mother, the age of her youngest child, her health status and her access to external childcare facilities—all have significant positive effects on the probability of the mother’s labour force participation. For example, on average, being married or living with a partner increases a mother’s probability of labour force participation by 13.3 percentage points, ceteris paribus; while on average, the provision of or access to external childcare facilities increases the probability of a mother’s labour force participation by 34.2 percentage points, ceteris paribus.

A contradictory result is seen in the case of the mother’s years of education: on average, the probability of a mother’s labour force participation decreases by about 1.9 percentage points with each additional year of schooling, ceteris paribus. However, this result is not statistically significant at any level.

However, failing to account for the endogeneity bias gives us exaggeration of the negative effect of fertility on female labour force participation (detailed results shown in Appendix 4): without accounting for endogeneity, on average having a third child reduces the mother’s probability of labour force participation by 7.4 percentage points, ceteris paribus.

7.1 Sensitivity Analysis

Even though a developed country, such as the UK is known to have no son preference, it would still be worth checking if the preference to have a third child differs according to whether the mother has had two daughters or two sons. Thus, in the first-stage equation of the two-stage residual inclusion (2SRI) framework, dummy variables are used to indicate whether the first two same-sex children are girls or boys. The aim is to ascertain whether the probability of having a third child changes depending upon the gender of the two previous same-sex children. The results of the first-stage regression are shown in Table 5:

Table 5
Sensitivity Analysis (1) on Gender Preference (effect on the probability of having a third child)

	(1)	(2)
First two children are boys	0.367***(0.017)	0.316***(0.014)
First two children are girls	0.369***(0.017)	0.317***(0.014)
Age of the mother	0.032***(0.002)	0.027***(0.001)
Age of mother at her first birth	−0.042***(0.002)	−0.035***(0.001)
Observations	4764	4764

Notes: (1) shows the marginal effects and (2) shows the average partial effects of sex composition of the first two children on the probability of having a third child, analysing for gender preferences. Standard errors given in parentheses; *** denotes significance at the 1 per cent level, ** at the 5 per cent level and * at the 10 per cent level; heteroscedastic standard errors have been used in all the regressions.

There is no significant difference between the effect that two boys and two girls have on the probability of having more than two children. On an average, ceteris paribus, having two boys increases the probability of a mother having a third child by approximately 31.6 percentage points. Similarly, on average, having two girls increases the probability of the mother having a third child by 31.7 percentage points, ceteris paribus. This is expected considering the UK does not have any specifically recognised son preference.

The UK is a developed and multi-cultural country that attracts a large number of potential immigrants. It would be worth checking if the results derived previously would change if the mother is foreign-born, or whether the results would change for British women of different ethnic backgrounds. Hence, in the second-stage (outcome) equation of the 2SRI framework, I add dummy variables indicating the ethnicity of the mother, and whether she is born in the United Kingdom or not. The results are given in Table 6:

Table 6

Sensitivity Analysis (2): Ethnicity and Foreign Status of Mother (effect on the probability of labour force participation)

	(1)	(2)
Third child	−0.026(0.044)	−0.019(0.032)
Age of the mother	0.022**(0.011)	0.016**(0.008)
Years of education	0.129(0.079)	0.094(0.057)
Marital status	0.205***(0.022)	0.148***(0.016)
Age of youngest child	0.031***(0.003)	0.023***(0.002)
Health status of mother	0.307***(0.051)	0.222***(0.036)
External childcare access	0.429***(0.019)	0.311***(0.012)
Predicted earnings	−0.053(0.053)	−0.039(0.038)
Residual of first-stage Regression	−0.083*(0.047)	−0.060*(0.034)
Mother is foreign-born	−0.147***(0.025)	−0.107***(0.017)
Mother is of Black ethnicity	0.045(0.032)	0.033 (0.023)
Mother is of Asian origin	−0.108***(0.029)	−0.078***(0.021)
Observations	4,764	4,764

Notes: (1) shows the marginal effects and (2) the average partial effects of given variables on the probability of positive hours of work (labour force participation of the mother); standard errors are in parentheses; *** denotes significance at the 1 per cent level, ** at the 5 per cent level and * at the 10 per cent level; bootstrapped standard errors have been used in all the regressions.

Table 6 suggests that, on average, being foreign born decreases the probability of labour force participation by 10.7 percentage points, ceteris paribus, a result that is highly significant. There could be several reasons for this, such as foreign attitudes towards female employment, foreign workers not getting adequate employment in the UK, etc. Being of Black ethnicity, on average, increases the probability of labour force participation by 3.3 percentage points, ceteris paribus, although the result is not statistically significant. On the other hand, being of Asian origin significantly reduces the probability of female labour force participation by 7.8 percentage points on average, ceteris paribus.

8. CONCLUSIONS

My results are in line with Iacovou (2001); both of us find that failing to account for endogeneity of fertility variables leads to an exaggerated negative effect of having more than two children (or a third child) on labour force participation. Failing to account for the endogeneity bias gives the result that, on average, having a third child reduces the mother’s probability of labour force participation by 7.4 percentage points, ceteris paribus. However, when accounting for the endogeneity bias, having a third child, on average, reduces the mother’s probability of labour force participation by 2.2 percentage points, ceteris paribus. Moreover, when endogeneity is accounted for, the negative effect is not statistically significant at any levels.

In my sample, having the first two children of the same sex has a significantly positive effect on the probability having a third child: this corroborates the well-known demographic finding that parents in developed nations prefer to have a ‘balanced family’. This study finds no particular differences in the gender preference for children—a mother is equally predicted to have a third child if her first two children are sons, than if her first two children are daughters, in stark contrast to Asian nations where there is an established preference for sons.

Additional explanatory variables, such as, the mother’s access to external childcare facilities, age of her youngest child, her marital and health status all have a positive impact on the probability of her labour force participation and these are significant at the 1 per cent level. Her age and predicted earnings also have positive effects on the probability of her labour force participation, though they are not significant at any level. Surprisingly, for the women in the sample, an additional year of education reduces the probability of labour force participation, ceteris paribus, though this result is not statistically significant. A possible explanation is that more educated women are capable of earning more in their early career years, so they can reduce their labour force participation in later years, with an increase in their family size.

The topic at hand has several important and relevant policy recommendations. It is obvious that several policies, such as subsidised childcare facilities, maternity leave provisions that allow women to return to work in a much easier manner after childbirth, childcare facilities in the workplace, etc., would reduce the negative effect that children tend to have on maternal labour force participation. Especially for single mothers, and women with lower education levels, these policy measures could lead to children having a less depressing impact on mothers’ labour force participation. In this context, studies have focused on the Nordic countries, where mothers have higher labour force participation than other European counties, questioning whether there is in fact, a ‘Nordic Model’ and whether other countries such as the UK should aspire to emulate it (Datta Gupta et al., 2008).

Footnotes

Acknowledgements

The author would like to thank Professor Wiji Arulampalam (Department of Economics, University of Warwick) for her suggestions, and the anonymous referee for very helpful comments.

APPENDICES

Notes

References

Amemiya

(1985). Advanced econometrics. Cambridge: Harvard University Press.

Angrist

J. D.

Evans

W. N.

(1998). Children and their parents’ labor supply: Evidence from exogenous variation in family size. The American Economic Review, 88(3), 450–77.

Asch

R. H.

Studd

(1995). Progress in Reproductive Medicine: Volume II. London, UK: CRC Press.

Ben-Porath

Welch

(1976). Do sex preferences really matter? The Quarterly Journal of Economics, 90(2), 285–307.

Bollen

K. A.

Guilkey

D. K.

Mroz

T. A.

(1995). Binary outcomes and endogenous explanatory variables: Tests and solutions with an application to the demand for contraceptive use in Tunisia. Demography, 32(1), 111–31.

Chun

(2002). An instrumental variable estimate of the effect of fertility on the labour force participation of married women. Applied Economics Letters, 9(10), 631–34.

Cristia

J. P.

(2008). The effect of a first child on female labor supply: Evidence from women seeking fertility services. Journal of Human Resources, 43(3), 487–510.

Cruces

Galiani

(2007). Fertility and female labor supply in Latin America: New causal evidence. Labour Economics, 14(3), 565–73.

Daouli

Demoussis

Giannakopoulos

(2009). Sibling-sex composition and its effects on fertility and labor supply of Greek mothers. Economics Letters, 102(3), 189–91.

10.

Datta Gupta

Smith

Verner

(2008). Child care and parental leave in the Nordic countries: A model to aspire to? Review of Economics of the Household, 6(1), 65–89.

11.

Ebenstein

(2007). The causal effect of fertility on female labor supply: Evidence from Taiwanese son preference. Retrieved from http://ssrn.com/abstract=988852 on 26 June, 2013.

12.

Fertility Summary. (2010). Office for National Statistics. United Kingdom.

13.

Gangadharan

Rosenbloom

Jacobson

Pearre III

J. W.

(1996). The effects of child-bearing on married women’s labor supply and earnings: Using twin births as a natural experiment, NBER Working Papers 5647. National Bureau of Economic Research, Inc.

14.

Goldin

(1994). Understanding the gender gap: An economic history of American women. NBER Book Series: Long-term Factors in Economic Development, Oxford University Press.

15.

Gronau

(1973). The intrafamily allocation of time: The value of the housewives’ time. The American Economic Review, 63(4), 634–51.

16.

Gutiérrez-Domenech

Bell

(2004). Female labour force participation in the United Kingdom: Evolving characteristics or changing behaviour? Bank of England Working Paper no. 221.

17.

Heckman

(1974). Shadow prices, market wages, and labor supply. Econometrica: Journal of the Econometric Society, 42(4), 679–94.

18.

Heckman

(1979). Sample selection bias as a specification error. Econometrica: Journal of the Econometric Society, 47(1), 153–161.

19.

Heckman

J. J.

Willis

R. J.

(1977). A beta-logistic model for the analysis of sequential labor force participation by married women. The Journal of Political Economy, 85, 27–58.

20.

Hirvonen

(2009). The effect of children on earnings using exogenous variation in family size: Swedish evidence. Swedish Institute for Social Research Working Paper.

21.

Iacovou

(2001). Fertility and female labour supply. Institute for Social and Economic Research Working Paper 2001–19. University of Essex.

22.

Jacobsen

J. P.

Pearce III

J. W.

Rosenbloom

J. L.

(1999). The effects of childbearing on married women’s labor supply and earnings: Using twin births as a natural experiment. Journal of Human Resources, 34(3), 449–74.

23.

Joshi

Davies

(1992). Day care in Europe and mothers’ forgone earnings. International Labour Review, 132(6), 561–79.

24.

Maurin

Moschion

(2009). The social multiplier and labor market participation of mothers. American Economic Journal: Applied Economics, 1(1), 251–72.

25.

Nakamura

(1992). The econometrics of female labor supply and children. Econometric Reviews, 11(1), 1–71.

26.

Rosenzweig

M. R.

Wolpin

K. I.

(1980). Testing the quantity-quality fertility model: The use of twins as a natural experiment. Econometrica: Journal of the Econometric Society, 48(1), 227.

27.

Schultz

T. P.

(1990). Testing the neoclassical model of family labor supply and fertility. Journal of Human Resources, 25(4), 599–634.

28.

Terza

J. V.

Basu

Rathouz

P. J.

(2008). Two-stage residual inclusion estimation: Addressing endogeneity in health econometric modeling. Journal of Health Economics, 27(3), 531–43.

29.

Terza

J. V.

Bradford

W. D.

Dismuke

C. E.

(2008). The use of linear instrumental variables methods in health services research and health economics: A cautionary note. Health Services Research, 43(3), 1102–20.

30.

The World Bank. (2011). World development indicators. Washington, DC: World Bank.

31.

Williamson

N. E.

(1983). Parental sex preferences and sex selection. In Bennett

N. G.

(Ed.), Sex selection of children (pp. 129–45). New York: Academic Press.

32.

Xie

(1997). Children and female labour supply behaviour. Applied Economics, 29(10), 1303–10.

Effect of Fertility on Female Labour Force Participation in the United Kingdom

Abstract

Keywords

1. INTRODUCTION

2. LITERATURE REVIEW

3. OVERVIEW OF FEMALE LABOUR FORCE PARTICIPATION AND FERTILITY IN THE UK

3.1 UK Female Labour Force Participation

7. RESULTS AND SENSITIVITY ANALYSIS

Table 5 Sensitivity Analysis (1) on Gender Preference (effect on the probability of having a third child)

Footnotes

Acknowledgements

APPENDICES

Notes

References

Table 5
Sensitivity Analysis (1) on Gender Preference (effect on the probability of having a third child)