Jan 2016
The relationship between the
age of giving birth and job
satisfaction of women
Bachelor Thesis
MICHELLE LAHAIJE (10364439)
Statement of Originality:
This document is written by Student Michelle Lahaije who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Introduction
owadays much more women are higher educated than in the past. The gender gap is decreasing and women are attaining better jobs. All of this has an influence on the job satisfaction of women. Women rate their job satisfaction higher than men do (Clark, 1997, p.342). One thing that will always disturb the career of a woman is giving birth to a child. This could influence her career and her job satisfaction. Lots of research has been done on the influence of having children on a woman’s career, but not whether the age of giving birth to her first child has an impact on this. The successfulness of someone’s career is hard to measure, because everyone has their own view on successfulness. Heslin (2002) in his study explains that subjective measurements are more applicable to measure the successfulness of someone’s career than objective measurements, such as salary. Therefore, I will use job satisfaction as a measurement for the success of a woman’s career. The research question for this thesis will be: Is the job satisfaction of women in the age between 47 and 56 affected by the age of giving birth to her first child? I think this research question is interesting because not much research has been done on this specific subject and much more women in their early twenties are wondering if having children at a younger age could disturb the chance of having a successful career.
This thesis is built on existing literature which has investigated the relationship between job satisfaction and gender, the family gap, the relationship between income and job satisfaction, the relationship between hours worked and job satisfaction and the effect of motherhood timing. Clark (1997) investigates the difference in job satisfaction between men and women. Overall he concludes that men report a lower job satisfaction than women do, whilst evidence shows that women’s jobs are worse in terms of employment conditions, such as hiring, firing and job content. Clark (1997) separates the job satisfaction into three categories: job satisfaction with pay, job satisfaction with the work itself and the overall job satisfaction. In all three cases the job satisfaction of men is significantly lower than that of women, whereas the biggest difference between men and women is in pay satisfaction. Additionally, Clark (1997) researches the effect of numerous variables on the employment of men and women. Variables that are included in this regression are the number of children in the household and the age of the children in the household. A negative relationship is recorded for all the different numbers of children in the household and for the different ages of the children, with one exception. If the child is in the age between 12 and 15 there is a positive influence on the employment of the parent. In my thesis I will not use the employment, but the job satisfaction
as dependent variable. I assume children are also influencing the job satisfaction of women and I will implement a new variable to the regression, the age at first birth.
Waldfogel (1998) investigates the ‘family gap’ in pay for women with children. She proves that over time the gender gap is decreasing whilst the family gap is increasing. The family gap is the difference in wage between women with and without children. This could be explained by the fact that maternity leave is not covered in the United States (Waldfogel, 1998, p. 141). She finds evidence in other countries, like Belgium, which are covering the maternity leave for women and Belgian women are more likely to return to their original firm and have a higher pay than women who are not getting maternity coverage, ceteris paribus. Apart from the increase in pay for women, Waldfogel (1998, p. 138) thinks that if women with children get more retention, the work experience and job tenure increases, which has a positive effect on maintaining a good job match. This could explain a higher job satisfaction. The sample in my research is from the United States, which implies that the women are not receiving maternity coverage (Waldfogel, 1998, p.141). Therefore, I expect that women with children have a lower job satisfaction than women without children.
The effect of income and education on job satisfaction is investigated by Clark and Oswald (1996). They find a negative relationship for both income and education on job satisfaction. This negative relationship with education can be explained by the fact that the aspiration on a certain job gets higher when the education is higher. In other words, the aspiration targets are higher. This will cause a lower job satisfaction (Clark and Oswald, 1996, p. 361). They show that the income effect on job satisfaction is U-shaped. I will include income as a quadratic polynomial and education in my regression to measure the job satisfaction. Booth and Ours (2008) investigated the relationship between part time work and hours satisfaction, job satisfaction and life satisfaction. They distinguishthese results between men and women. Booth and Ours (2008) conclude that women are more satisfied with their job and hours of work than men are, which is in line with the results from Clark (1997). They also find that if the family income increases, the job satisfaction of a woman decreases and that women are more satisfied if they are working less than 15 hours per week (Booth and Ours, 2008, p. 88).
The effect of motherhood timing on the career path of a woman is investigated by Miller (2011). She concludes that there is an increase in earnings of 9.6% when the motherhood is delayed by a year (Miller, 2011, p. 1083). The mean age women give birth to their first child in the sample is 26 years old, where Miller (2011) excluded all teen pregnancies. She presents five different cases to illustrate the effect of a career interruption. These different figures drawn
from these cases are called a ‘mommy track’. In the first case when there is no growth in earnings over the years, no additional costs are associated with motherhood. However, when the mother has a reduction in experience from the career interruption, the timing of the child should influence her wage over time. In Millers paper the age at first birth is related to her earnings and wage. In my thesis the age at first birth is used to measure the job satisfaction instead of earnings and wage. I think this is a better measurement to rate the successfulness of her career. As described before by Heslin (2002), subjective measurements are better to measure a successful career than objective measurements.
The set-up of this paper is as follows. In the first section the data are described. Section 2 describes the method and the regression used to test the job satisfaction of women. In section 3 the results are given and in section 4 conclusions are drawn.
Data
The database I will use for my research is NLSY79, the National Longitudinal Survey of Youth. The sample used in this database are American youth born between 1957 and 1964. They are interviewed in 25 rounds. Round 1 in 1979 and the last round in 2012. The number of individuals interviewed is 12686, of which 6283 (50%) are female. In the first interview the respondents are between 14 and 22 years old and in the last interview in 2012 they are around 47 to 56 years old. The information gathered from this cohort are: labour market behaviour, educational experience, family background, armed services, high school performance, family life, health issues and assets and income. The samplesused in the NSLY are a cross-sectional sample, a set of supplemental samples and a military sample, whereas the individuals are randomly chosen in selected areas in the United States. Together these samples represent the entire youth population in the US.
For my research I will use the dataset of the final round, which took place in 2012. The respondents are between 47 and 56 years old. They should have had enough time to make career and they are probably not planning to have children at this age. The sample consists of 1880 women who are working for pay. Job satisfaction is the independent variable in this thesis and is rated by the respondents as follows: (1) Dislike it very much, (2) Dislike it somewhat, (3) Like it fairly well, (4) Like it very much. The median job satisfaction of all women in this
sample is like it fairly well. In table I below the women who reported ‘Like it very much’ are summarized. 1
Table I.
The t-statistics on difference in women with and without children is 9.15, which implies that this difference is significant at a confidence level of 1%.
One more thing that is tested in this thesis is whether the age of giving birth to her first child has an impact on the job satisfaction. In table II the age of giving birth is divided into three groups. From each group the percentage women reporting the highest job satisfaction are presented. The mean age of giving birth to her first child is 23 years old. Therefore, I separate the groups at age 23. The first group is when she gives birth between 13 and 18 years old, when she is in high school. The second group is from 18 to 23 years old and the third group is above 23 years old.
Table II.
1 In the dataset of NLSY the job satisfaction is reported the other way around, like it very much was reported with number 1. To interpret the result more easily I changed the order of the job satisfaction rating.
Total observations Women report Like it very much Percentage
Overall 1880 914 48.62 %
With children 1595 770 48.70 %
Without children 303 144 48.16 %
Total observations Women report Like it very much Percentage
Overall 1581 770 48.70%
13 < Age R at 1st Birth ≤ 18 303 150 49.50%
18 < Age R at 1st Birth ≤ 23 551 260 47.19%
The percentages are more divergent than in table I and they imply that the relationship between job satisfaction and the age of giving birth is not linear. The t-statistics on the difference of the groups are all different from zero except one. The difference between giving birth to a child at an age younger than 18 and older than 23 is not significant.
The remaining data gathered from the NLSY 79 database are: Highest degree ever received, income from wages and salary, working hours per week, the chance of getting a promotion within the next 2 years, having a partner and if the respondents own their house. The highest degree ever received is scaled from 0 to 8 ((0) None, (1) High school diploma, (2) Associate / Junior College AA, (3) Bachelor of Arts Degree BA, (4) Bachelor of Science Degree BS, (5) Master’s Degree, (6) Doctoral Degree PhD, (7) Professional Degree, (8) Other). The median score of the highest education received in my sample is a High School Diploma. The income from wages and salaries is only from their primary job and is expressed in US dollars per year. I dropped the observations if income was below $900 and above $175,000. These outliers are unlikely to be correctly recorded by the respondents. The average income of my sample is $43,027 per year. The standard deviation from income is $29,015 per year. The working hours are per week and only from their primary job. These women work 38 hours per week on average, with a standard deviation of 9 hours per week. The question asked to the women about promotion is: Do you believe promotion is possible in the next 2 years? 47% of all women answered yes to this question. Approximately 58% of the sample has a partner and 70% of the women own their house. This could influence job satisfaction because house owners are less satisfied than renters because it may be harder for them to give up a job they dislike (Clark, 1997, p. 349).
Methodology
In this section I will explain the methodology and the regression I use to answer my research question. Before investigating whether the age of the mother when giving birth to her first child is influencing her job satisfaction, I investigate if having children at all influences her job satisfaction. The first hypothesis is:
H0: The job satisfaction of a woman is not related to the fact that she has (a) child/children.
I think that women with children will report a lower job satisfaction than women without children. Women without children do not have the interruption of giving birth to a child in the middle of her career and they can spend more time on their career than women with children. The second hypothesis is:
H0: The age of giving birth to a first child is not related to the job satisfaction of the mother.
This hypothesis is answering my research question. I expect a negative relationship between the age of giving birth and the job satisfaction. When a woman gives birth to a child when she is younger she will have the interruption in an earlier stage in her career. When she gets back in the labour market she has enough time to continue her career and catch up the possible negative consequences from her time out. For the firm it is less costly to have a female employee dropping out at a younger age than at an older age, assuming her salary increases over time. The method I use to test these hypotheses is ordered probit. This method suits better with my data than OLS, because my dependent variable, job satisfaction, is a categorical variable. This model is estimated using maximum likelihood. The job satisfaction is classified in different (ordered) categories ((1) Dislike it very much, (2) Dislike it somewhat, (3) Like it fairly well, (4) Like it very much). The latent continuous variable, y*, is the job satisfaction. In this model the y* is a linear combinations of predictors plus the standard error:
"#∗= &
#'(#+ +#
Alpha indicates the boundaries between the different categories. For very low y* < -# women dislike their job very much, for -#> y* > -/ they dislike their job somewhat, for -/> y* > -0 it improves to liking their job fairly well and if y* > -0 they like their job very much. The regression parameters, ( and the threshold parameters -#, -/ and -0 are obtained by maximizing the log likelihood with 234 = Pr "3 = 7 .
The likelihood is defined as follows:
Pr "3 = 7 = F α:− <3'( − = -4>#− <3'(
The ordered probit model is standard normally distributed with = ∙ = Φ ∙ (Cameron and Trividi, 2010, pp. 526 – 527). The outcomesof this model, the betas, can only be interpreted by its sign, not how much the independent variable influences the dependent variable. When the beta is positive, it means that an increase in <3, increases the probability for that individual to report a high job satisfaction and decreases the probability that she is reporting a low job satisfaction. For example, when measuring the effect of income on the job satisfaction. The outcome of ( corresponding to the income is positive and significant different from zero. This
means that when the income from individual one is higher than from individual two, individual one is more likely to report a high job satisfaction, like it very much, than individual two. The exact difference from a change in income and its effect on the job satisfaction cannot be interpreted from these outcomes.
Testing the first hypothesis of this thesis, the main explanatory variable is whether the woman has a child/children, the dependent variable is job satisfaction and I control for: income as a quadratic polynomial, highest degree ever received, work hours per week, promotion, partner and if the individual owns her residence. When the outcome of the ( corresponding to children is positive and significantly different from zero, according to my expectation, the probability of a woman with children reporting the highest job satisfaction (like it very much) is less than the probability of a woman reporting the highest job satisfaction without children. The second hypothesis is about the age of the mother when giving birth to her first child, which is the main explanatory variable. The dependent variable is the job satisfaction and I control for the same variables as in my first regression, namely: income as a quadratic polynomial, highest degree ever received, work hours per week, promotion, partner and if the individual owns her residence. For example, when the outcome gives a negative ( for age of the woman at first birth. Then the probability of reporting a high job satisfaction is lower for women giving birth at an older age than women giving birth at a younger age.
Results
In this section the results are presented, I give an interpretation of the findings and the results are discussed. First of all, the results from the first regression are presented, concerning the effect of having children on job satisfaction. After that, the results of the second regression are published, regarding the age of the mother at first birth.
In table III the results are given from the first regression. In all the columns the results of the ordered probit method are presented, the difference between the columns is the number of control variables. First I included all the control variables to the regression, this is reported in the last column. In the first and second column I exclude the control variables which are not significant. All findings of the variable children are not significant. This indicates that having children is not related to the job satisfaction of a woman. Promotion is significantly different from zero at 1% confidence interval, and shows a positive sign. The interpretation of this is that a woman who believes promotion is possible in the next two years has a higher probability reporting liking her job very much than a woman who does not believe she will get a promotion
in two years’ time. The same holds for having a partner and being owner of a residence. Both variables indicate that if the woman has a partner or owns a residence she has a higher probability reporting a high job satisfaction. The results of promotion and having a partner are the same as Clark (1997, p.350) finds in his paper. The outcome for owning her own residence is not equal to the result in Clarks paper (1997, p. 350). He finds that when the individual rents the house he/she is more likely to report a higher job satisfaction. Clarks (1997) objective about this is: “Renters are more satisfied than homeowners, perhaps because it is easier for them to leave jobs that they dislike” (p. 349). The difference between Clark (1997) and this thesis is that the sample of this thesis only consists of women whereas Clark is looking at both men and women. It could be that men and women have a different opinion about this, further investigation has to be done to be able to draw a conclusion.
Table III.
Ordered probit regression of job satisfaction
Variables Ordered Probit Ordered Probit Ordered Probit
Children -0.0143 -0.0208 -0.0058 (0.0736) (0.0737) (0.0749) Income 0.0000 (0.0000) Income - squared 0.0000 (0.0000) Highest Degree 0.0024 (0.0171)
Hours per week 0.0043 -0.0060
(0.0031) (0.0034) Promotion 0.1923*** 0.1977*** 0.1940*** (0.0540) (0.0541) (0.0542) Partner 0.1135* 0.1065* 0.1012* (0.0583) (0.0585) (0.0586) Own residence 0.1138* 0.1152* 0.0923 (0.0625) (0.0625) (0.0645) Observations 1880 1880 1880
The betas from income, income squared, highest degree ever received and the work hours per week are not significant. This means that these three variables do not have an influence on the job satisfaction. The reason that work hours are not influencing the job satisfaction might be that it is not a choice for every woman to work more or less hours. Women working a lot of hours per week could do it because they really like their job or because they need the money. In other words, the motivation to work more hours per week could be different for individual women.
The reason why the outcome of children is not significant could be that it is a conscious decision not having children to live the dream of having a successful career. This makes the career aspiration higher, which makes her job less satisfying. Therefore, women without children could report a lower job satisfaction than I expected. Clark and Oswald (1996, p. 361) explain the same issue in their paper about the aspiration targets. The reason why the effect of children on job satisfaction is not significant could be due to omitted variables. One omitted variable could be the income or schooling of the partner. Leibowitz and Klerman (1995) find that the income of the partner has an effect on the employment of his wife, the same holds for the highest degree of the partner. All the women in their sample have children. Higher income opportunities for the partner, reduces the mother’s employment. This means that when the husband is earning less and his highest degree is a high school diploma, the wife is more likely to work. The reason then for her to work is to support her family, which could influence her job satisfaction. She might settle with a less attractive job to maintain her family. This is the opposite to what Booth and Ours (2008) find in their paper. They conclude that when the family income increases, the job satisfaction of the woman decreases.
In table IV the results are given from the second hypothesis. The age at first birth is not significant different from zero. The same holds for age at first birth squared. This variable is included in the regression because when looking at the percentage of women reporting a high job satisfaction between the different ages of first birth, it seems that the relationship is not linear. The percentage of women liking their job very much below 18 years old and older than 23 years old is almost equal, whereas the percentage highly satisfied with their job for women in the age between 18 and 23 is significant lower. The age of first birth does not have an influence on the job satisfaction. The same holds for income and highest degree ever received. The number of working hours per week is only significant when all the control variables are added to the regression. The relationship between hours of work per week and the job satisfaction is negative. This implies that if a woman works more hours per week, the probability of reporting a high job satisfaction is lower than a woman working less hours per
week. Booth and Ours (2008) find the same result in their paper. The chance of promotion in 2 years, having a partner and owning a residence all have a positive relation with the job satisfaction, equal to the outcomes in the first regression.
Table IV.
Ordered probit regression of job satisfaction
Variables Ordered Probit Ordered Probit Ordered Probit
Age at 1st Birth 0.0211 0.0248 0.0219 (0.0372) (0.0373) (0.0375) Age at 1st Birth - squared 0.0004 0.0005 0.0005 (0.0007) (0.0007) (0.0007) Income 0.0000 (0.0000) Income - squared 0.0000 (0.0000) Highest Degree 0.01889 (0.0195)
Hours per week -0.0055 -0.0083**
(0.0034) (0.0038) Promotion 0.1736*** 0.1806*** 0.1807*** (0.0587) (0.0589) (0.0590) Partner 0.1296** 0.1195* 0.1077* (0.0643) (0.6452) (0.0648) Own residence 0.1210* 0.1229* 0.0996 (0.0691) (0.0692) (0.0704) Observations 1581 1581 1581
(*) Significant at 10%, (**) Significant at 5 %, (***) Significant at 1%.
The reason why the age of giving birth does not have an influence on the job satisfaction could be that having a child is not something every woman plans. Some women in the sample could have an unplanned child and perhaps some could not become pregnant first try. The motherhood timing might be different than they had planned and therefore the career they expect to have might be changed. Another thing that could explain the reason why I do not find a relationship between the age of first birth and job satisfaction could be due to omitted variables. Chipman and Morrison (2015) investigate how kin networks, environmental risks and reproductive risk taking could influence the individual preferences of having children and
at what age. They find that when the perceived neighbourhood risk is decreasing, the age of first birth is decreasing and when perceived school risk is decreasing, the age at first birth is increasing. Neighbourhood risk and school risk are measured by asking questions to the participants about how safe they feel, if there are bullies at school etc. The knowledge of safe sexual practice (KSSP) is positive related with the age of first birth (Chipman and Morrison, 2015, p. 701). The KSSP does say something about the education of a person and therefore about the job satisfaction as well. Neighbourhood risk and school risk could also influence the job satisfaction. Therefore, it could be that the age of birth does not have an influence on the job satisfaction because of these omitted variables.
Another reason why I do not find a relationship between the age of first birth and the job satisfaction might be that there is a difference between lower educated women and higher educated women. Herr (2015) investigates if the hours worked in the first year after giving birth are related to fertility timing. This relationship is different for high school graduated women and for college graduated women. College graduated women invest more in their career than high school graduates and therefore she finds different relationships between the high school and college graduates for first-birth timing on labour supply (Herr, 2015). The relationship between age at first birth might also be different for lower educated women and for higher educated women. To see if this might be the case I usethe same regression as in table IV, but now I split the sample into lower educated women (no degree, high school degree or junior college degree) and higher educated women (bachelor of arts degree BA, bachelor of science degree BS, master’s degree, doctoral degree PhD and professional degree) and I will not control for highest degree anymore. The results are reported in table V.
The outcomes from table V are not significant, with an exception for promotion and hours per week. This means that the age of giving birth is not related to the job satisfaction for either lower educated women as for higher educated women. The difference between the average age of giving birth between lower educated women and higher educated women is big. The average age lower educated women give birth to their first child is 22.5 years old, while for higher educated women the average age of giving birth is 27 years old.
Table V.
Ordered probit regression of job satisfaction between lower educated women and higher educated women.
Ordered Probit
Variables All Lower educated women Higher educated women
Age at 1st Birth 0.0260 0.0175 0.0203 (0.0379) (0.0457) (0.0812) Age at 1st Birth - squared -0.0005 -0.0004 -0.0005 (0.0007) (0.0009) (0.0015) Income 0.0000 0.0000 0.0000 (0.0000) (0.0000) (0.0000) Income - squared 0.0000 0.0000 0.0000 (0.0000) (0.0000) (0.0000)
Hours per week -0.0080* -0.0051 -0.0148*
(0.0038) (0.0044) (0.0080) Promotion 0.1692*** 0.1648** 0.2111* (0.0594) (0.0680) (0.1277) Partner 0.0949 0.0772 0.1107 (0.0656) (0.0741) (0.1445) Own residence 0.0985 0.0962 0.1169 (0.0713) 0.0780 (0.1865) Observations 1544 1171 373
(*) Significant at 10%, (**) Significant at 5 %, (***) Significant at 1%.
The results from table III, IV and V can only be interpreted by their sign, but not how much the effect is on job satisfaction. To make this clearer I look at the marginal effects per variable on being in the highest category of job satisfaction, like it very much, while holding all other variables at their means. The outcome gives the probability of changes of liking the job very much as one variable increases, when the individual is an average person in any other respect. The results are reported in table VI and VII.
Table VI.
Marginal effect on job satisfaction per variable Dependent variable: Job satisfaction
Like it very much (4)
Children -0.0023
Income 0.0000
Highest degree 0.0010
Hours per week -0.0024
Promotion 0.0773
Partner 0.0403
Own residence 0.0368
Note: All other variables are fixed at their means.
Table VI gives the marginal effects per variable for the first regression. The outcome for the variable children in the first line of the table is -0.0023. This means that women with children are 0.23% less likely to be highly satisfied with their job than women without children, holding all other variables at their means. In other words, when a woman with an income of $43,027, a junior college diploma, working 38 hours per week, has 47% chance of getting a promotion in two years, having a partner and owning her house, is 0.23% less likely of reporting the highest job satisfaction when she has children than when she does not have children.
Table VII.
Marginal effect on job satisfaction per variable Dependent variable: Job satisfaction
Like it very much (4)
Age at 1st birth 0.0087
Income 0.0000
Highest degree 0.0075
Hours per week -0.0033
Promotion 0.0721
Partner 0.0429
Own residence 0.0397
In table VII the marginal effects are reported for the second regression. The marginal effect of the age at first birth on job satisfaction is 0.0087. This means that when the age of giving birth increases by 1 year, the probability of being highly satisfied increases by 0.87%, holding all other variables at their means. The average woman in this sample is a woman who give birth to a child at age 23, has an income of $41,495, her highest degree ever received is a high school degree, she works 38 hours per week, she believes that the chance of having a promotion in two years is 47%, she has a partner and she owns her residence.
Conclusion
The research question of this thesis is: Is the job satisfaction of women in the age between 47 and 56 affected by the age of giving birth to her first child? The answer to this question is that the job satisfaction of a woman is not affected by the age of giving birth. To answer this research question I first look at whether having children influences the job satisfaction of women at all, which is not the case. Even when the sample is divided into lower educated women and higher educated women, the influence of the age at first birth on job satisfaction is not significant. The sample I useto test my hypotheses is from the NLSY79 database and consists of 1880 women, who are working for pay. The data and literature I usein this thesis is from the United States. The results are applicable to the United States, but not directly to the Netherlands, the country I live in. The women in my sample are on average 51 years old, in 2012. The average age of giving birth to the first child is 23, which means that on average the women in gave birth in 1984. In the Netherlands the age of the mother when giving birth to her first child is 26.5 years old in 1984 (CBS, 2016) and nowadays the age of giving birth to the first child is almost 29.5 years old in the Netherlands. This increase in the age of giving first birth suggests a change in the career path a woman takes these days. This might be interesting to investigate in a few years and see whether time and social culture has an influence on the relationship between the age at first birth and the job satisfaction. One more thing that might be interesting to research is if a firm makes a difference between hiring women with or without children. For example, when a firm is more likely to hire women without children, it is more likely that women will postpone the wish of having children until they have the job they want to have. But if the firm makes no difference between women with and without children the career path of a woman might look totally different. This could influence the job satisfaction as well.
References
Booth, A. L. &, Van Ours, J. C. (2008). Job satisfaction and family happiness: the part-time work puzzle. The Economic Journal, 118, pp. 77 – 99.
Cameron, A. C. &, Trivedi, P.K. (2010). Microeconomics using Stata (Revised Edition, pp. 491– 529). Texas: Stata press.
CBS (2016). Centraal Bureau voor Statistiek, Den haag/Heerlen. Retrieved on January 7th 2016, from:http://statline.cbs.nl/StatWeb/publication/?VW=T&DM=SLNL&PA=37422ne &LA=NL
Chipman, A. &, Morrison, E. (2015). Family Planning: Fertility and Parenting Ideals in Urban Adolescents. Archives of Sexual Behavior, 44(3), pp. 695 – 703.
Clark, A. E. (1996). Job satisfaction and gender: Why are women so happy at work? Labour Economics, 4, pp. 341 – 372.
Clark, A. E. &, Oswald, A. J. (1995). Satisfaction and comparison income. Journal of Public Economics, 61, pp. 359 – 381.
Herr, J.S. (2015). The Labor Supply Effects of Delayed First Birth. American Economic Review: Papers & Proceedings, 105(5), pp. 630 – 637.
Heslin, P. A. (2003). Self- and Other- Referent Criteria of Career Success. Journal of career assessment.
Leibowitz, A. &, Klerman, J. A. (1995). Explaining changes in married mothers’ employment over time. Demograph, 32(3), pp. 365 – 378.
Miller, A. R. (2011). The effects of motherhood timing on career path. Journal of population Economics, 24(3), pp. 1071 – 1100.
Waldfogel, J. (1998). Understanding the "Family Gap" in Pay for Women with Children. Journal of Economic Perspectives, 12(1), pp. 137 - 156.
Appendix
1. Regression 1, Summary data
Variable Obs Mean Median Stdev. Min Max
Job satisfaction 1880 3.3973 3 0.6711 1 4
Children 1880 0.8410 1 0.3658 0 1
Income 1880 43027 36704 29015 900 175000
Highest Degree 1880 2.1096 1 1.7629 0 8
Hours per week 1880 38.52 40 8.82 1 90
Promotion 1880 0.4665 0 0.4990 0 1
Partner 1880 0.5851 1 0.4928 0 1
Own residence 1880 0.7048 1 0.4563 0 1
2. Regression 2, Summary data
Variable Obs Mean Median Stdev. Min Max
Job satisfaction 1581 3.3984 3 0.6699 1 4
Age at 1st Birth 1581 23.69 23 5.68 13 45
Income 1581 41495 35000 28410 900 175000
Highest Degree 1581 1.9880 1 1.7027 0 8
Hours per week 1581 38.24 40 8.83 1 80
Promotion 1581 0.4769 0 0.4996 0 1
Partner 1581 0.6078 1 0.4884 0 1
