Parenthood and Sickness Absence

(1)

Parenthood and Sickness Absence

A comparative analysis between opposite-sex and same-sex couples

Author: Hanna Tegmyr Supervisor: Per Johansson

Abstract

Earlier studies have concluded that entering parenthood induces higher sickness absence among birthmothers as compared to their partners in opposite-sex couples. Increased household commitment, causing birthmothers to reduce their labor supply after entering parenthood is suggested to cause this increase of the female sickness absence in opposite-sex couples. Some researchers argue that the labor supply decisions in opposite-sex couples are based on comparative advantages. Other researchers argue that gender identities are of greater importance in explaining the labor supply decisions. The aim with this study is to investigate whether gender composition matters for how parenthood affects the within-couple gap in sickness absence. The within-couple income and sickness absence gap in opposite-sex and female same-sex couples entering parenthood are studied. The effect of parenthood on the within-couple income gap and sickness absence gap are estimated using ordinary least squares in a difference-in-differences setting. Analogous to other studies, I find that birthmothers’ sickness absence generally increases in opposite-sex couples after couples have their first child. Additionally, I find that parenthood does not induce a generally higher sickness absence among birthmothers or partners in same-sex couples. Eligible sharing of household commitment, allowing a relatively higher labor market attachment among both birthmothers and partners in same-sex couples appears to be the main explanation to why same-sex couples do not experience a higher level of sickness absence.

(2)

1 Introduction

The level female sickness absence in Sweden has increased since the 1980s as compared to that of men. During this same period there has been an influx of women to the labor market. Policy changes during the 1970s and 1980s such as abolishment of joint taxation together with a more generous social insurance and publicly provided childcare facilitated the possibility of entering the labor force among women with young children. Other European countries, in which the female labor supply has increased during the last decades, have also experienced an increased level of female sickness absence (Angelov et al., 2011; Bratberg et al., 2002). This has led researchers to believe that worse relative health among women may not be the explanation of the development of female sickness absence.

Women spend in general more hours on unpaid household work than men. Some researchers argue that the increase of female sickness absence is due to deteriorated health caused by women’s dual responsibility, conventionally referred to as the double burden (see e.g. Bratberg et al., 2002). When combining labor market work and household work, the dual role risks increasing stress, which is suggested to affect individual’s health negatively. Since women in general take on the dual responsibility they also face a higher risk of becoming sick absent (Bratberg et al. 2002; SIA, 2014:14). The unequal gender gap in time spent on home production among opposite-sex couples widens as couples get their first child (Boye, 2008; Sanchez and Thomson, 1997; Baxter et al. 2008). Once couples have their first child women increase the amount of hours spent in housework whereas the hours of men do not change. Moreover, entering parenthood has implications for female labor supply. Earlier studies conclude that women decrease their labor market supply after having a child whereas men are unaffected by parenthood (see e.g. Kennerberg, 2007). The increased household commitment among women leads to a reduced attachment to the labor market and is according to Angelov et al. (2013:9) the main explanation behind the increased female sickness absence after entering parenthood.

(4)

within-couple specialization since spouses in opposite-sex couples act in accordance with specific behaviors prescribed to individuals defined as “man” or “woman”.

In this study I aim to test how well comparative advantages, in terms of earnings potential, can help explain why entering parenthood risk increasing the female sickness absence (in terms of days of utilized sickness benefit and/or disability benefits). Opposite-sex and female same-sex couples will be studied in order to detect potential behavioral differences that could be explained by gender identities. The main question I aim to answer is: does gender composition matter for how parenthood affects the within-couple sickness absence? The analysis is divided into two parts. First I will study the within-couple income gap, which will function as a proxy for selection of which spouse commit to greater household commitment when entering parenthood. A reduction of labor market income after the child is born signals greater household commitment, which could explain a potential higher sickness absence among these spouses. In the second part I will analyze the effect of parenthood on the within-couple sickness absence.

To my knowledge this is the first study analyzing parenthood’s effect on the within-couple sickness absence in both opposite-sex and female same-sex couples. The study is implemented on couples having their first child sometime between the years 2001-2010. The data is collected from the Swedish Social Insurance Agency (SIA) and consists of 84,837 opposite-sex couples and 107 female same-sex couples. The causal effect of parenthood on the within-couple income and sickness absence gap is estimated using ordinary least squares in a difference-in-differences setting.

(5)

2 Background

In this section, a theoretical background will be presented as well as earlier studies on labor supply and sickness absence. The studies are divided according to different topics in order to facilitate for the reader. A summary is presented in the end of the section.

2.1 Theoretical Background

Having a child requires time spent nursing and raising the child and couples need to allocate their time accordingly. In general women in opposite-sex couples reallocate part of their time from the labor market to the household, whereas partners stay in the labor market. Couples having a child thus face certain monetary costs, in terms of forgone earnings when nursing the child. How this latter monetary cost is shared within a couple is according to Angelov et al. (forthcoming) to a great extent dependent on comparative advantages, in terms of earnings potential. In accordance with Angelov et al. (forthcoming), I will assume that the labor supply decisions within a couple are based on a bargaining process, where comparative advantages partly determine the spouses’ bargaining power. According to the theory of comparative advantages, introduced by Becker (1981), spouses in a couple allocate their time in the market in which they are more productive i.e. according to their comparative advantages. In order to maximize the total income and household production of the family, the spouse with the highest earnings potential specializes within the market sector. Consequently, the (unequal) gender gap in labor market supply among opposite-sex couples is a consequence of a rational choice made by the couple (Van Dongen, 2009).

(6)

sector. This gender neutrality also implies that the theory is applicable to couples regardless of gender composition.

There are however studies criticizing the theory of comparative advantages based on findings contradicting the theory’s gender neutrality. Some of these studies suggest instead that gender identities play an important part in how opposite-sex couples share household commitment (Tichenor, 2005; Akerlof and Kranton, 2000). Tichenor (2005) analyzes gender identities within opposite-sex marriages in the U.S. where the wives earn more or have a higher status occupation than their husbands. Tichenor describes a situation of negotiating identity and power within the marriage. Although women are the primary earners they perform the majority of childcare and domestic work. Akerlof and Kranton (2000) analyze the importance of gender identities in different economic outcomes. In their study they analyze how individuals act in order to affirm their identity as a “man” or a “woman” and how not acting according to ones identity can induce distress in others and in oneself. According to Van Dongen (2009), couples implementing more egalitarian gender identities will divide their labor supply more equally. The studies by Tichenor (2005) and Akerlof and Kranton (2000) present an alternative to comparative advantages as explanation to household specialization. These studies deal with gender identities in opposite-sex couples. There is a lack in the literature regarding the identities of same-sex couples. Given that same-sex couples do not consist of the categories “man” and “woman”, one can assume that same-sex couples easier deviate from gender identities, which could result in less specialization within same-sex couples.

(7)

2.2 Female sickness absence in Sweden

In the beginning of the 1980s the gender difference in sickness absence increased in Sweden. Figure 2.2.1 shows the average daily amount of sickness benefit granted to men and women. As seen in the figure the female sickness absent increased compared to that of men.

Figure 2.2.1: The average daily amount of sick leave days utilized per person and year in Sweden, divided on men and women, aged 16-65. Source: Angelov et al. (2013:9)

Angelov et al. (2011:2) identify several variables of importance when explaining the female sickness absence in Sweden from the 1980s onward. Higher level of education and to some extent age and type of occupation are of biggest importance. A larger share of women reports having occupations which are more psychological than physical demanding. Angelov et al. do not find evidence of the increased sickness absence being solely due to a growing share of women experiencing worse health, although it plausibly explains part of the increase. The authors also conclude that gender segregation in the labor market do not considerably impact the sickness absence.

(8)

to entering parenthood, the difference in sickness absence between men and women is small. After entering parenthood, however, the within couple gap in sickness absence increases. When the child turns two years old the female sickness absence is twice as high as that of men. The gender gap in sickness absence stays high up to fifteen years after the first child is born1.

According to Swedish law, parents are allowed to work 75 percent of their regular working hours until their child turns 8 years old. In addition, they are eligible parental allowance. Women mainly take advantage of the possibility of reducing working hours after entering parenthood. As pointed out by Angelov et al. (2013:9), this generous parental leave system may change the intensive-margin labor supply among individuals when entering parenthood. A lower labor supply is suggested to lower the threshold for using sickness benefits, since the cost of being absent is reduced. Additionally, a lower presence at the work place might yield less possibilities of career advancement, which in turn can affect work incentives. Angelov et al. (2013:9) study the income trajectories of women entering parenthood and find that higher income trajectories (i.e. higher labor market attachment) reduce the effect of parenthood on the within-couples gap in sickness absence. The authors conclude that unequal gender division of family responsibilities causes women to reduce their labor supply after entering parenthood. This in turn lowers the attachment to the labor market and is, according to the authors, the main explanation to the increased gender gap in sickness absence. Higher attachment to the labor market lowers the effect of parenthood on women’s sickness absence. Angelov et al. (2013:9) find no evidence of long-run female health deterioration in comparison with their spouses after having the first child. An additional important finding of Angelov et al. (2913:9) is that a lower labor supply induces higher levels of sickness absence rather than the other way round.

Similar to Angelov et al. (2013:9), the SIA (2014:14) acknowledges a higher risk among spouses in opposite-sex couples that commit relatively more to household work, to become sick absent two years after having a child. The SIA suggests that the sickness absence is correlated with both parents increasing their labor supply after parental leave, which increases the double burden (in terms of combining labor market work and household commitment) of the spouse committing relatively more to

1_{Angelov et al. control for having a second child during the observed period, but the result stays close}

(9)

household work, and accordingly the risk of becoming sick absent. Bratberg et al. (2002) analyze if combining a lifestyle of family and career induces higher levels of sickness absence among women in Norway. Some women experience a conflict between combining work and family roles. This conflict of roles can contribute to increased stress, which is suggested to negatively affect women’s health (Bratberg et al., 2002). Bratberg et al. find supportive evidence of the double burden affecting the sickness absence among women.

2.3 Household specialization

As mentioned in section 1.1, the (unequal) gender gap in labor supply among opposite-sex couples is suggested to be resulting from rational behavior. In this section, a more detailed review of earlier studies analyzing explanations for division of labor supply in both opposite-sex and same-sex couples is presented.

Angelov et al. (2013:3) analyze how differences in family responsibility arising from parenthood can help explain the within-couple income and wage gap. The analysis is based on register data where opposite-sex couples are studied between the years 1986 and 2008 in Sweden. The authors conclude that the effect of parenthood on the within-couple income and wage gap depends on how the individual income and wage growth would have developed in absence of the child.

(10)

explanation for the gender division of household work, but that comparative advantages play an important role.

The SIA has analyzed what variables are important determinants for equally shared family responsibility in terms of childcare in Sweden (SIA, 2013). According to the SIA families tend to be more willing to accept the income loss of the mother when she is the primary earner in order for her to stay home with the child. Couples are however less willing to sacrifice the income of the father when he is the primary earner. This finding suggests that gender is of greater importance than economics in explaining household specialization associated with parenthood.

Goldberg and Perry-Jenkins (2007) test the hypothesis that same-sex couples in contrast to opposite-sex couples are more successful in negotiating a more equal division of labor supply when entering parenthood. In the study only same-sex couples that biologically become first time parents in the U.S. are included. Three month after the child was born the study showed that biological mothers had reduced their hours of work (some women had reduced labor supply prior to entering parenthood). Moreover they engaged in childcare tasks to a greater extent, particularly when working fewer hours. Non-biological mothers returned to work within a couple of weeks after entering parenthood due to “financial considerations”. As regards household work the couples succeeded in equally sharing these duties after the transition to parenthood. Ahmed et al. (2011) describe the within household income gap in different sets of couples in Sweden. They find that female same-sex couples earn less than opposite-sex couples. However the earnings differentials within female same-sex couples are smaller compared to that of opposite-sex couples. According to Ahmed et al. (2011) this implies that there is less of specialization among female same-sex couples than in opposite-sex couples.

2.4 Parenthood and labor supply decisions

(11)

longer than men, which plausibly correlates with the fact that more women went from working full time to part time.

Labor supply Women pre Women post Men pre Men post Short part-time 10 % 14 % 3 % 2 %

Long part-time 15 % 33 % 3 % 3 %

Full time 75 % 47 % 94 % 93 %

Table 2.4.1: Changes of labor supply after transition to parenthood (Westerlund et al., 2005). “Pre” refers to the year prior to having a child and “post” refers to the two years following childbirth. Full time is equivalent to working 100 percent. Long part-time is equivalent to working 75 – 99 percent and short part-time equals 1 – 74 percent. Four percent of the men and women were no longer employed in 2002.

Kennerberg (2007) analyzed the change in labor market situations among opposite-sex couples as compared to childless opposite-opposite-sex couples. Her main finding supports the results found by Westerlund et al. (2005). According to Kennerberg women reduce their working hours after entering parenthood, the working hours among men are however not affected. Kennerberg claims that these findings can be attributed to the unequal sharing of parental leave, where women utilize the majority of parental allowance. An interesting founding in the study of Angelov et al. (2013:3) is that the effect of parenthood on the within-couple income gap is different from the effect on the within-couple wage gap. There seems to be no short run effect of parenthood on women’s wage growth. Angelov et al. do however observe a long run effect, which is explained being due to mothers’ lower level of labor supply as compared to that of fathers. The study of Angelov et al. (2013:3) indicates that mothers to a greater extent start working part-time after entering parenthood. The study also estimates a long-term effect of parenthood on the within-couple income gap. Accordingly, it appears as if women who start working part-time after entering parenthood continue this during several years.

2.5 Summary

(12)

market in which they are most productive. According to this theory women reducing their labor supply are less productive in the labor market and hence allocate their time within home production where they are more productive. The theory of gender identities, instead suggests that the division of household commitment is a matter of negotiating identities and accordingly gender is of more importance than productivity. I intend to contribute to additional insight of how well comparative advantages can help explain couples’ labor supply decisions that could risk higher female sickness absence in Sweden. When studying opposite-sex couples exclusively it can be difficult to separate behavior that is induced by gender identities. Therefore, the within-couple sickness absence gap of opposite-sex and same-sex couples will be studied. If couples act solely according to their comparative advantages we should expect the two different populations to behave in a similar way when entering parenthood, depending on which spouse has higher earnings potential. If couples do not act according to their comparative advantages, this could be an indication of couples negotiating identities.

According to the theory of comparative advantages spouses in a couple allocate their time according to where they are relatively more productive. I will assume that the spouse with highest expected labor market income two years prior to parenthood is relatively more productive in the labor market. This spouse should accordingly specialize in the market sector after entering parenthood, indicating that the labor supply after entering parenthood should not be affected substantially. Instead the spouses with relatively lower expected labor market income should spend more hours at home, reducing their labor supply. Consequently one would expect the sickness absence among the secondary earners to increase as they increase household commitment and lose attachment to the labor market.

3 Social Insurance in Sweden

(13)

3.1 Sickness Insurance

The compensation to absent employees due to illness goes through three stages. There is no compensation during the first day of absence in the public and private sector.2 From day 1 to 14, the employer compensates the employee with a sick pay, based on 80 percent of the wage. If the employee is sick for more than seven days a doctor’s certificate needs to be submitted to the employer. From day 15 and onwards the SIA decides whether individuals are eligible sickness benefit.3 The sickness benefit paid is based on 80 percent of the yearly labor market income. This compensation is paid maximum 364 days during a period of 450 days (15 months). Throughout this period the SIA reevaluate the individual’s health and his/her eligibility of sickness benefit.4 After the first 364 days the compensation falls to 75 percent of the yearly income. The maximum amount of continued days for which an individual can be compensated is 550 days. The total amount of days an individual is eligible sickness benefit is thus 914 days.5

The SIA offers a particular type of health insurance for individuals suffering from a long-term illness, injury or disability. Employed individuals between the ages 19 and 29 are eligible temporary disability benefits, if the work capacity is reduced by at least 25 percent over more than one year. Between the ages 30 to 64 individuals are eligible for disability benefits, given that their work capacity is reduced by at least 25 percent during the foreseeable future.6 Henceforth, sickness benefit and disability benefits (both temporary and permanent) will be referred to as sickness benefits.

The SIA employs an income cap of a maximum amount at which the sickness benefits will be based. This cap is 7.5 times the price base amount, which is currently a yearly earning of 333,700 SEK7. The minimum amount earned is 10,700 SEK.8 Table 3.1.1 shows the level of sickness benefits the insured is eligible depending on the level of absence from work.

2_{Self-employed workers choose between 1, 14, 30, 60 or 90 days of illness without compensation from}

the SIA.

3_{http://www.forsakringskassan.se/arbetsgivare/sjukfranvaro/sjuk_1_14/om_sjuklon/} 4_{Before 2008 there was no time limit for the use of sickness benefit.}

5_{https://www.forsakringskassan.se/wps/portal/sjukvard/sjukskrivning_och_sjukpenning/tidsbegransnin}

g_av_sjukpenning

6_{https://www.forsakringskassan.se/wps/portal/arbetsgivare/sjukfranvaro/sjuk_langre_14/sjukersattning}

_och_aktivitetsersattning

7_{The Swedish price base amount is 44 500 SEK, which is valid from 2015-01-01.}

http://www.scb.se/sv_/Hitta-statistik/Statistik-efter-amne/Priser-och-konsumtion/Konsumentprisindex/Konsumentprisindex-KPI/33772/33779/Behallare-for-Press/375807/

8

(14)

Table 3.1.1: Eligibility of sickness benefits in Sweden.

3.2 Parental Insurance

In Sweden, parents are granted 480 days of parental allowance for each child9, approximately equal to 16 months. Of these 480 days, the parental allowance of 390 days is based on previous labor market earnings. 60 days are earmarked for each parent. The remaining days can however be used in whatever way the parents find optimal.10 Parents are permitted parental allowance until the child turns eight years old or ends first grade.11

If a parent is unemployed or has not been employed during 240 days before the

child is born the parental allowance is based on a daily base amount of 225 SEK.12

Since this study aims to analyze the within couple gender difference in income and sickness absence, these parents will be excluded from the sample I use.

The SIA also offers temporary parental allowance eligible to parents if the child becomes sick or if the parent currently nursing the child becomes too sick to nurse the child. The temporary parental allowance is granted until the child is 12 years old and covers 120 days per child and year. 13

Similar to the sickness insurance, the parental allowance is based on 80 percent of previous labor market earnings. The income cap used in the parental insurance is based on 10 times the price base amount, which is currently a yearly income of 445,000 SEK. The minimum amount earned is 10,700 SEK. The income cap used for temporary parental allowance however is 7.5 times the price base amount and thus the

9_{Parents are eligible 660 days of parental leave when having twins. Twins are however not included in}

this study.

10_{https://www.forsakringskassan.se/wps/portal/privatpers/foralder/barnet_fott/foraldrapenning/om_fora}

ldrapenning?

11_{From January 1}st_{2015 the Swedish government has changed the child’s age limit of utilizing}

parental insurance. This does however not affect my sample since they are born before at latest in year 2010.

12_{The amount varies depending on birth year of the child. If the child is born between 2006-2014, the}

daily base amount is 180 SEK. If the child is born before 2006, the amount is 45 SEK a day.

13_{https://www.forsakringskassan.se/wps/portal/privatpers/foralder/barnet_sjukt/om_vab} Level of absence from work due to illness Level of sickness benefit/(temporary)

disability benefits

25 % – 49 % 1/4

50 % – 74 % 1/2

75 % – 99 % 3/4

(15)

same as the sickness benefit. Table 3.2.1 shows the eligibility of parental allowance depending on the level of absence from work.

Level of absence from work due to childcare

Level of parental/temporary parental allowance eligible 0 % – 12 % 0 12.5 % – 24 % 1/8 25 % – 49 % 1/4 50 % – 74 % 1/2 75 % – 99 % 3/4 100 % 1

Table 3.2.1: Eligibility of parental insurance in Sweden.

4 Identification strategy and econometric method

The aim with this paper is to estimate the effect of gender composition on the within-couple gap in sickness absence associated with entering parenthood. In this section I present the method used as well as the necessary identification strategy needed to reach a valid result. Initially, the regression used to estimate earnings potential will be presented.

4.1 Estimating earnings potential

Earnings potential is not equivalent to labor market income but, as the expression suggests, the potential labor market income of an individual. In this study, expected labor market incomes will function as an indicator of earnings potential. Regression (1) is used to estimate the expected yearly incomes of each individual, based on a given age and education level.

𝑦_! = 𝛼_!+ 𝛽_!𝑎𝑔𝑒_! + 𝛽_!𝑎𝑔𝑒_!!_{+ 𝛽}

!𝑒𝑑𝑢!+ 𝑒!

The parameters are estimated using ordinary least squares (OLS) with robust standard errors. Age, age-squared and years of education are used as explanation variables of the expected income. Labor market income increases with age and years of education. The relationship between age and income is however not linear but levels off at a

(16)

higher age, and thus age-squared is added to the regression.14If the estimated expected income (𝑦_!) deviates from the true income (𝑦_!), this is captured by the residual 𝑒_!. If 𝑒_! is positive, the individual earn more than his/hers expected income given his/hers age and years of education. If the residual is negative, the opposite is true. The expected incomes two years prior to childbirth will be used to define which of the spouses in a couple has higher earnings potential. Two years before childbirth, the incomes of the birthmothers-to-be do not risk being affected by pregnancy.

4.2 Identification strategy

Analyzing the effect of parenthood on labor supply and sickness absence is challenging. Parenthood is plausibly correlated with health and economic stability. Unobservable individual characteristics and how the individuals react given these characteristics could affect timing and likelihood of having a child. Spouses in a couple moreover plausibly influence each other.

In order to control for unobserved individual characteristics that could affect timing and likelihood of having a child, I will follow the strategy used by Angelov et al. (2013:9) and analyze the within-couple income and sickness absence gap over time. The approach used is similar to a difference-in-differences setting where the treatment must be exogenous, i.e. not depending on prospective information correlated with the outcomes in the absence of the treatment. The treatment is the timing of having a child. Here the identifying assumption means that the timing of when to have a child should not be determined on prospective information on gender differences in my outcomes, in the absence of entering parenthood. Examples of this would be that couples decide to have a child when the spouse with highest earnings potential is being promised a raise, or if the spouse with relatively lower earnings potential is being diagnosed with an illness. These scenarios could produce biased results. The expected within couple income and sickness absence gap should be constant for the two populations studied in absence of parenthood. Opposite-sex and same-sex couples should have the same pre-trend evolvement of the income and

14_{Similar to age, education can be argued to evolve nonlinearly with income. I included a variable}

(17)

sickness absence gap. Both individuals in a couple are affected by the treatment but the effect is allowed to vary between the spouses.

I will control for differences in age, education and income that could potentially affect timing and likelihood of having a child.

4.3 Empirical approach

In this paper opposite-sex and female same-sex couples giving birth to their first child during the observed period will be included. The parents to a child are tracked in the data and matched. In order to study comparable opposite-sex and same-sex couples, the same-sex couples are limited to female same-sex couples giving birth to the child. The couples will function as their own control group by adding couple fixed effects in the regression model.

In the first part of the analysis I will study within couple labor supply decisions by comparing the within couple income gap prior to, during and after entering parenthood. The labor market income functions as a proxy for selection of which spouse commit to greater family responsibility when entering parenthood. A reduction of labor market income after the child is born signals greater family responsibilities.15 In the second part of my analysis I will study the within-couple sickness absence, defined as number of days of utilized sickness benefits.

4.4 Estimating the income gap

When studying the within couple income gap of opposite-sex couples, earlier studies usually subtract the woman’s income from the man’s. In this empirical model the spouses are referred to 1 and 2 where the birthmothers in the opposite-sex and same-sex couples will be referred to as spouse 2 while their partners will be referred to as spouse 1. Regression (2) estimates the within-couple income gap.

15_{I choose to study the changes of labor market income instead of utilized parental allowance, as}

records on parental allowance do not correctly mirror individuals’ absenteeism from work. During the child’s first year parents are not required to utilize any of their eligible days of parental allowance they have however by law the right to absent from work in order to take care of the child.

(18)

2 Δ𝑦!" = 𝛼 + 𝛼!𝑆𝑎𝑚𝑒𝑆𝑒𝑥 + 𝛽!!!𝑄!∆𝑖𝑛𝑐𝑜𝑚𝑒!!! ! !!! + 𝑆𝑎𝑚𝑒𝑆𝑒𝑥 ∗ 𝛽!!!𝑄!∆𝑖𝑛𝑐𝑜𝑚𝑒!!! ! !!! + 𝜌!" ! !!! 1 𝑡 = 𝑗 + 𝑆𝑎𝑚𝑒𝑆𝑒𝑥 ∗ 𝜆_!" ! !!! 1 𝑡 = 𝑗 + 𝜃_! + 𝜕_!+ 𝑥_!!!! _{+ 𝑢} !"

The yearly labor market incomes (y) are logged so the outcome is to be interpreted as the percentage change of the within-couple income gap measured by Δ𝑦_!" =

𝑙𝑛𝑦!− 𝑙𝑛𝑦! !" . i= 1, …,n index couples, t index time measured in years. j index time since birth measured in years, where j = 0 equals the year in which the child was born. The base year in the regression is one year prior to childbirth

The variable SameSex takes the value one if a couple consists of two same-sex spouses and zero otherwise. 𝛼_! thus measures the average pre-child difference in the income gap between opposite-sex and same-sex couples belonging to the first income-gap quintile. 𝜃_! captures calendar year fixed effects and 𝜕_! indicates the child’s birth month.

The effect of parenthood on the within-couple income gap may vary depending on differences in spouses’ pre-birth labor market incomes. In order to capture this effect, the income gaps two years prior to childbirths are divided into quintiles, denoted 𝑄 . Interaction terms capturing same-sex couples belonging to different quintiles are also included.

(19)

3

Edu_{!" !} and Age_!!! – Age_{!!! !}. 𝑥_!!! _{also captures the difference in squared age and} information regarding if parents are born outside of Sweden.

The main parameters of interest are 𝜌! and 𝜆!. 𝜌! measures the effect of parenthood on the within-couple income gap during the child’s j:th year. 𝜆_! measures the effect of being a same-sex couple after entering parenthood. 𝜌_! thus measures the baseline effect of parenthood on opposite-sex couples.

4.5 Estimating the sickness absence gap

Regression (3) estimates the within-couple gap in sickness absence

Δ𝑠_!" = 𝛼_!"#+ 𝛼_!𝑆𝑎𝑚𝑒𝑆𝑒𝑥 + 𝛽_!!!𝑄_!∆𝑖𝑛𝑐𝑜𝑚𝑒_!!! ! !!! + 𝑆𝑎𝑚𝑒𝑆𝑒𝑥 ∗ 𝛽_!!!𝑄_!∆𝑖𝑛𝑐𝑜𝑚𝑒_!!! ! !!! + 𝜌_!! ! !!! 1 𝑡 = 𝑗 + 𝑆𝑎𝑚𝑒𝑆𝑒𝑥 ∗ 𝜆_!" ! !!! 1 𝑡 = 𝑗 + 𝜃_! + 𝜕_!+ 𝑥_!!!! _{+ 𝑢} !"

The sickness absence is measured by subtracting the partner’s days of sickness absence from birthmother’s days of sickness absence, 𝑠_!− 𝑠_{! !}. In this way the results are more easily interpreted since the birthmothers in general utilize more days of sickness benefits per year in pre-birth periods. 𝛼!"# measures pre-pregnancy differences in sickness absence i.e. 𝑠!− 𝑠! !!!. Similar to estimating the income gap 𝜃_! index calendar time fixed effects. Now 𝛼_! measures the average pre-birth difference in the sickness absence gap between opposite-sex and same-sex couples.

(20)

4.6 Same-sex couples and pregnancy

Female same-sex couples have a legal right to get inseminated as of 1st of July 2005 (RFSL, 2014). The insemination can only be implemented at one of Sweden’s university hospitals.16 The municipality, in which the hospital is located, determines the cost of insemination, and it ranges from a standard patient fee (approximately 300 SEK) up to 31,459 SEK.17 There is no minimum or maximum amount of time that female same-sex couples are bound to wait before realization of insemination. The time period instead varies depending on the availability of sperm donors (Swedish Association of Local Authorities and Regions, 2014).

An alternative to hospitalized insemination is home insemination. The insemination is free of charge but requires a man willing to donate his sperm. Some couples use a friend to access sperm; others find donors online, which could cost money. If couples implement home insemination, the partner not carrying the child has to adopt the child once it is born if she wants shared custody of the child.

When studying female same-sex couples, there are several challenges that need to be dealt with. The major challenge is that any of the spouses biologically can become pregnant. If same-sex couples chose the healthiest partner and/or the one with the highest labor market income to become pregnant, this could affect the results, since the identification assumptions are no longer valid. Another challenge is that same-sex couples are more likely to plan a pregnancy as compared to opposite-sex couples, threatening the assumption of treatment exogeneity. The planning consists of who will become pregnant and also what method to use in order to become pregnant. A third challenge, or rather a restriction, concerns the sample of same-sex couples. Even though home insemination is implemented among couples I cannot assume that my sample of same-sex couples is selected randomly. I do not have data on how the birth mothers became pregnant in same-sex couples. Clinical insemination, depending on municipality, can cost a significant amount of money. Thus I cannot reject that the sample of same-sex couples consists of a certain type of women having a certain economic stability.

16_{The hospitals are located in Stockholm, Göteborg, Lund/Malmö, Uppsala, Umeå, Örebro and}

Linköping.

17_{I have been contacting all the university hospitals in Sweden. The cost of 6 occasions of}

insemination or 3 IVF varies across the seven hospitals. Sahlgrenska and Akademiska charge a patient fee for this treatment. At Skånes universitetssjuk they charge 6072 for donated sperms and

(21)

Goldberg et al. (2012) analyzes what factors are important to consider among female same-sex couples when deciding what spouse will carry the child. Half of the couples in Goldberg’s study considered age, infertility, health and career advancement/job flexibility. The decisions made by the other half were based on one partner having a greater desire to experience giving birth to a child and/or to be related to the child genetically.

Of great importance for the assumptions made in this paper are decisions regarding age, health and career advancement. If the couples plan to have more than one child they might take turn, letting the oldest spouse carry the first child. The older spouse has plausibly reached a higher level at the income trajectory. Accordingly she might enjoy greater job flexibility, which facilitates home commitment once the child is born. The most rational decision from an economic perspective however could be argued to be that the spouse with relative lower earnings (indicating lower productivity) should give birth to the child since she will lose relatively less in earnings. This strategy allows the spouse with highest earnings potential to continue working and thus maximize family income. However, a job flexibility of the spouse with highest earning might facilitate for both spouses to continue working after childbirth. The most rational decision would then be to let the spouse with highest earnings potential become pregnant.

The most important health aspect to consider when choosing birthmother is the partner’s fertility and how able the partner is to carry and give birth to the child. If the partner is too ill to carry a child, this will plausibly be detected in the observed pre-birth periods. Accordingly this type of illness will be controlled for.

(22)

4.7 Subsequent births

There exists a possibility that birthmothers give birth to additional children during the observed period. Unfortunately my data do not contain information regarding if couples had additional children during the period studied. Angelov et al. (2013:9) discuss this matter when they analyze sickness absence within opposite-sex couples, but find no substantial effect of a second pregnancy on birthmothers’ sickness absence. This indicates that potential subsequent births among birthmothers in this study will not affect the estimated results of the within-couples sickness absence. Although, same-sex couples might take turns becoming pregnant, letting partners carry a second child. Angelov et al. (2013:9) show that the sickness absence among birthmothers increases substantially during the pregnancy. Partners in same-sex couples will reasonably experience a similar increase in the sickness absence due to pregnancy-related illness as the birthmothers in this study. Consequently the latter type of subsequent birth should be detected in the graphical results.

4.8 Definition of earnings potential

When couples decide upon how to share household commitment after childbirth they plausibly consider short-time labor market incomes, and to some extent future incomes. Simply using pre-birth income levels as an indicator for earnings potential can therefore be problematic. Imagine for example a couple, where the observed labor market incomes two years prior to childbirth differ. One spouse is older than the other, but the other spouse has more years of education and accordingly entered the labor market later than the first spouse. In terms of observed incomes, the former spouse has a higher income than the latter. However, the estimated expected income of the latter spouse is higher, due to a higher education level. In order to maximize family income, the couple might let the latter spouse specialize within the market sector after entering parenthood. Thus, the expected incomes of spouses in a couple can be of greater importance in determining labor supply decisions. Accordingly I believe that letting the estimated expected income function as an indicator of earnings potential is a justified assumption.

(23)

5 Data

In this section I will describe the data used in the analysis, the restriction of the sample used and descriptive statistics.

5.1 Data

The data used in this study is from the Swedish Social Insurance Agency’s database MiDAS (MikroData för Analys av Socialförsäkringen/Micro Data for Analysis of the Social Insurance). MiDAS is a data set covering information about all individuals in Sweden that have been granted social benefits and is updated on a monthly basis. The sickness absence is defined as days of utilized sickness benefits. The data on sickness benefits are registered as number of days of utilized benefits per year. In addition to data on social insurance MiDAS also contains information collected from Statistics Sweden and the Swedish Tax Agency.

Apart from information on sickness absence and yearly labor market income18 the data also contains gender, age, years of education, profession, sector of employment, country of origin, marital status as well as records on individuals’ unemployment. In this study the sample is limited to include couples with positive earnings two years before childbirth.

I will not be able to observe individuals’ labor supply in terms of working hours. When individuals change their hours of work however this will result in a changed labor market income as shown in Angelov et al. (2013:3). Yearly labor market income will therefore function as a proxy for hours of work. The labor market income does not include taxable social security benefits.

The time period used is ranging from year 1999 until 2012. I will be able to track couples during two years prior to the birth of the first child as well as four years after the birth. Couples having their first child between the years 2001 – 2010 are included in this study. Parents to children born in 2010 can only be tracked two years after childbirth.

(24)

5.2 Restriction of couples

In order to better observe the impact of gender composition on the within-couple income and sickness absence gap, I want the couples included in this study to be as similar as possible when entering parenthood. Angelov et al. (2013:9) show that the female sickness absence increases the year before giving birth to the child. The higher level of sickness absence is explained as pregnancy-related illness and decreases once the child is born. The female sickness absence however increases again approximately in the child’s second year (Angelov et al., 2013:9). If one were to include parents where one spouse did not biologically give birth to a child, the effect of parenthood on the within-couple income and sickness absence gap might not transpire in the same manner, since part of the sickness absence could be due to birth-related injuries. In order to easier interpret the results, the populations are limited to include female same-sex couples and opposite sex couples giving birth to the child. Hence, adoptive parents are excluded.

5.3 Descriptive statistics

The couples used in this analysis are opposite-sex and female same-sex couples in partnership or marriage having their first child. I will study 84,944 couples of which 107 are same-sex couples. Parents are between 17 – 65 years old when having their first child.

In my analysis I use a time variable measuring years ranging from -2 to 4. Year 0 captures the year in which the child was born. The effect of parenthood on parents’ income and sickness absence will vary during year 0, depending on what month the child is born. Imagine for instance a couple having a child in January 1st of 2001. Their labor market incomes and days of sickness absence in 2001 could thus be affected by entering parenthood. A couple having their child in December 31st 2001 instead will not have been affected by parenthood in the same extent. Due to this I will also control for the month in which the child is born by including month dummies. By analyzing couples that enter parenthood in different years, labor supply and parenthood that could be affected by business cycles are controlled for.

(25)

child. I have also included a squared measure of the age and estimated that age gap since income does not evolve linearly with age. As regards individuals with education level labeled as missing or unknown two years prior to childbirth I have replaced the level of education with the first registered level of education during the observed years. The most likely explanation for not having a registered or missing education level is the person being an immigrant. It can take a couple of years before the Swedish authorities register information regarding education after the arrival to Sweden.

I have excluded couples where one or both spouses have a labor market income of zero or a missing observation of income two years prior to childbirth. I have also excluded couples were one or both spouses were registered as parents after the child turned one years old. One explanation to the latter limitation is that I do not want to include adoptive parents in the sample. Another explanation to the limitation of 365 days is that I want both spouses to have the opportunity of nursing the child once it is born.

When discussing underlying explanations of the female sickness absence, it is important to consider the child’s health. One can argue that if birthmothers nurse a child to a greater extent, they are also at risk of being infected by the child’s illnesses. As mentioned in 3.1, the SIA compensates workers that are absent from work due to illness from day 15. Therefore days of sickness absence between days 1-14 are not included in my analysis. Sickness absence caused by transient illnesses just mentioned will therefore be excluded. Sickness benefits and parental allowance can be used in different scales depending on how many hours an insured is absent from work (see 3.1, 3.2). In this study I am interested in studying the level of absenteeism due to illness and hence, a day of utilized sickness benefits will be reported as one day, regardless of scale. This approach is consistent with other studies (see e.g. SIA, 2014:14).

Self-employed workers have the possibility of deciding from what day the SIA should compensate their sickness absence. I could not access information regarding how many days the self-employed workers in my data go without compensation. Accordingly, the self-employed individuals had to be excluded from the sample.

(26)

Sweden (Statistics Sweden 2014:17). The birthmothers have generally slightly more years of education and lower labor market income. In same-sex couples, the average age of birthmothers is higher than that of partners. Partners in same-sex couples have generally more years of education and lower labor market income as compared to birthmothers. The average estimated expected incomes are also included in table 5.3.1. The average expected income of partners in opposite-sex couples is lower than the true average income, whereas the average expected income of birthmothers is higher than the observed average income. In same-sex couples, both spouses are in general expected to have higher labor market incomes. Partners in opposite-sex couples and birthmothers in same-sex couples have the highest expected income, which according to my assumption, means that these spouses have a relatively higher earnings potential. If couples act according to their comparative advantages, in terms of earnings potential, partners in opposite-sex couples and birthmothers in same-sex couples are expected to specialize within the market sector, while their spouses specialize within the household (i.e. reduce their labor supply). According to earlier studies, birthmothers in opposite-sex couples and partners in same-sex couples face a higher risk of being sick absent, due to a reduced labor market attachment.

6 Results

In this section the results of this study are presented, starting with some graphical results followed by the estimated effect of parenthood on the within couple income gap and sickness absence gap. The regression models are estimated with ordinary least squares, OLS, and standard errors are estimated by clustering at the couple level. Following the results, a heterogeneous analysis will be implemented. The average Age at t=0 partner Age at t=0 birthmother Years of education partner Years of education birthmother Income partner, t=-2 Expected income partner, t=-2 Income birthmother, t=-2 Expected income birthmother t=-2 Opposite-Sex couples 31.8 (4.9) 29.6 (4.1) 13.1 (2.2) 13.5 (2.1) 321,893 (224,789) 281,308 (73,572) 237,782 (134 375) 256,987 (71,993) Same-Sex couples 31.5 (3.9) 32.9 (3.8) 14.2 (2.2) 14.0 (1.9) 241,374 (142, 446) 290,782 (63,635) 293,619 (133 792) 310,298 (58,013)

(27)

gaps in income and sickness absence are measured one year before pregnancy, the year of childbirth as well as each year up to four years after childbirth.

6.1 Graphical results

The evolvement of average labor market income divided on spouses in opposite-sex and same-sex couples is shown in figure 6.1.1 (see next page). Confidence intervals of each group are shown in the graph. The income gap before parenthood is in general wider among opposite-sex than same-sex couples. In contrast to opposite-sex couples, the birthmothers in same-sex couples have the highest relative labor market income.

The evolvement of labor market income pre-birth appears to be somewhat constant within couples, suggesting a valid parallel trend assumption. A more accurate test of the parallel trend assumption will be presented in section 6.2.

The graph (including 95 percent confidence intervals displayed as bars) tells a similar story to what earlier studies have found analyzing parenthood’s effect on labor market income in opposite-sex couples. The partners earn in general more than birthmothers pre-birth and the within-couple income gap expands as couples have their first child. The partners are rather unaffected by parenthood and four years after the child is born they have reached a higher labor market income as compared to two years before having a child. Birthmothers however appear to be significantly affected by parenthood and have a lower labor market income up to four years after childbirth, which implies a lower labor supply.

(28)

Figure 6.1.1: The average labor market incomes in opposite-sex (OS) and same-sex couples (SS). Confidence intervals are displayed as bars.

Figure 6.1.2 (see next page) displays the average days of sickness benefits utilized per year during the years before and after parenthood. Again the evolvement of opposite-sex couples is similar to what is found in other studies. Birthmothers are more absent from work due to illness during the year before childbirth and the year in which the child is born, which is reasonably due to pregnancy-related illness. After the child is born the sickness absence decreases among birthmothers in both opposite-sex and same-sex couples. In accordance with earlier studies, the sickness absence among birthmothers in opposite-sex couples generally increases in year 2, and continues to be high up to four years after birth. The sickness absence of partners in opposite-sex couples does not appear to be affected by entering parenthood. The estimated means in opposite-sex couples are precisely estimated whereas the confidence intervals of means in same-sex couples are large. The amount of average days of sickness absence among birthmothers and partners in same-sex couples varies and the gap in sickness absence is difficult to interpret. What can however be observed is that birthmothers in same-sex couples, similar to birthmothers in opposite-sex couples, experience a higher sickness absence prior to childbirth and during the birth year. But unlike the latter group, the sickness absence among birthmothers in same-sex couples does not generally increase from year two onwards.

100000 150000 200000 250000 300000 350000 400000 450000 -‐2 -‐1 0 1 2 3 4 L abor m arke t i nc om e

Years before/after childbirth

Labor Market Income

(29)

Figure 6.1.2: The average days of utilized sickness benefits in opposite-sex (OS) and same-sex couples (SS). Confidence intervals are displayed as bars.

6.2 Main result of parenthood’s effect on the labor market income gap

The main results of parenthood’s effect on the labor market income gap are found in table 6.2.1. The results of each year should be interpreted as the change in the within-couple income gap compared to one year before childbirth. In order to obtain the change in the within-couple income gap of same-sex couples, one should use the point estimate of opposite-sex couples (e.g. Year 0) and add the point estimate of same-sex couples in an equivalent year (e.g. Same sex year 0).

Three specifications are used to estimate the effects of parenthood on the within-couple income gap. Column (1) displays the estimated effects using no controls. In column (2), calendar years controls are included and finally, in column (3) additional control variables are included.

Calendar year controls are included in order to control for economic shocks that could potentially affect the labor market income gaps during the period studied. Adding calendar year controls increases the estimated effect in opposite-sex couples while reducing that of same-sex couples. Including additional controls marginally reduces the size of the estimated effects for both populations. Overall, the results are robust; point estimates that are statistically significant at 1-percentage level in the first specification continue to be significant at the same level in the following two

-‐5 0 5 10 15 20 25 30 -‐2 -‐1 0 1 2 3 4 D ays of s ic kne ss be ne fi ts /ye ar

Years before/after childbirth

Sickness Absence

(30)

specifications. Also, point estimates possess the same sign in all three specifications. The estimates in column (3) will hereafter be discussed further.

During the birth year (year 0) the income gap significantly increases by

approximately 127 percent19 among opposite-sex couples. The change of the income

gap during this year is not significantly different between opposite-sex and same-sex couples. The increase of the within-couple income gap is expected given that birthmothers reasonably reduce labor supply during late pregnancies and once the child is born.

In opposite-sex couples the income gap increases by approximately 196 percent in year 1 as compared to the pre-birth income gap. The income gap continues to be significantly wider than pre-birth levels in opposite-sex couples up to four years after the child is born, where the biggest increase occur during the child’s first years. This result indicates that birthmothers in opposite-sex couples in general reduce labor supply after entering parenthood. In other words, the birthmothers in opposite-sex couples bear in general the cost – in terms of reduced labor market earnings – of having a child. This finding confirms the graphical results for opposite-sex couples in figure 6.1.1.

The point estimates of opposite-sex couples are not too different from the ones found by Angelov et al. (2013). Although the estimates found in table 6.2.1 are persistently lower, by approximately ten percentage points. The biggest difference between the results can be found in year 1 where Angelov et al. (2013:3) found a significant increase of the income gap by 238 percent as compared to 165 percent in table 6.2.1. Angelov et al. (2013:3) studied opposite-sex couples having their first child between 1990-2002, whereas in this study the corresponding period is 2001-2010. As pointed out by Moberg (forthcoming), one could interpret the differences of the results as an indication of a more equal sharing of household commitment in opposite-sex couples. In 2002, the SIA earmarked a second month of the parental allowance to each parent. This led to an increased amount of utilized days of parental leave among fathers in opposite-sex couples (SIA, 2013), indicating a more equal sharing of household commitment. Although, the within-couple income gap in opposite-sex couples is, nevertheless, still consistently big.

(31)

Among same-sex couples the effect of parenthood on the income gap is significantly smaller compared to that of opposite sex couples. During year 0 there is no significant difference between opposite-sex and same-sex couples, indicating that gender composition does not matter for how the birth year’s income gap changes. The income gap increases in year 1 by approximately 104.8 percent (1.955-0.907). During years 2 and 3, the changes in income gap compared to pre-birth levels are smaller: 10.5 (1.165-1.060) and 40.8 percent (1.247-1.019). In year 4, there is no significant difference between opposite-sex and same-sex couples.

The variable Same sex, which takes the value 1 if couples are same-sex couples in pre-birth periods, is insignificant which indicates that there is no general difference between opposite-sex and same-sex couples during the pre-treatment period. Additionally there is no significant difference between the two populations with different magnitudes of income gaps two years prior to childbirth (see full table 6.2.2 in Appendix). This suggests that there is no difference in pre-treatment trends between opposite-sex and same-sex couples, which provides some suggestive evidence of the strict exogeneity model assumption.

(32)

Table 6.2.1: The yearly effect of parenthood on spousal income gap in opposite-sex and same-sex couples having their first child. The outcome variable is the within-couple difference in labor market income.

(1) (2) (3)

No controls Calendar year Full set of controls Year 0 1.275*** 1.280*** 1.265*** (0.008) (0.009) (0.009) Year 1 1.957*** 1.977*** 1.955*** (0.008) (0.010) (0.010) Year 2 1.123*** 1.173*** 1.165*** (0.008) (0.011) (0.011) Year 3 1.356*** 1.431*** 1.427*** (0.009) (0.013) (0.013) Year 4 1.006*** 1.119*** 1.110*** (0.009) (0.015) (0.015) Same sex year 0 -0.302 -0.268 -0.253

(0.236) (0.236) (0.234) Same sex year 1 -0.989*** -0.926*** -0.907***

(0.235) (0.235) (0.233) Same sex year 2 -1.156*** -1.065*** -1.060***

(0.236) (0.236) (0.234) Same sex year 3 -1.135*** -1.022*** -1.019***

(0.254) (0.254) (0.252) Same sex year 4 -0.576** -0.457 -0.449

(0.292) (0.292) (0.289) Same sex 0.065 0.094 -0.001 (0.311) (0.311) (0.307) Intercept -0.475*** -0.531*** -0.279*** (0.016) (0.027) (0.035) Calendar year Control variables No No Yes No Yes Yes R-squared 0.0863 0.0875 0.0950 Observations 961,093 961,093 943,099 Number of couples 84,865 84,865 83,649 *** p<0.01, ** p<0.05, * p<0.1

(33)

6.3 Main result of parenthood’s effect on the gap of sickness absence

According to the results found in 6.2, birthmothers bear in general the cost of having a child by reducing their labor supply. As reported by other empirical studies they face a higher risk of sickness absence.

The estimated effects of parenthood on the within-couple gap in sickness absence are presented in table 6.3.1. Similar to section 6.2, three specifications are used when estimating the effects of parenthood on the within-couple sickness absence gap. In column (1) no controls are used, in column (2) calendar year controls are added and finally additional control variables are included in column (3). The estimated effects change somewhat when adding calendar year controls, indicating that calendar years are important to control for. When adding additional control variables there is no considerable change in estimated effects apart from the intercept, which measures the mean gap in the sickness absence pre-birth. Similar to section 6.2, the estimated effects in column (3), table 6.3.1 will be discussed.

Table 6.3.1, column (3), shows that the sickness absence among birthmothers is in general significantly higher than the sickness absence of partners in both opposite-sex and same-opposite-sex couples during pre-birth periods (intercept). During the birth year (year 0), the sickness absence increases, which is plausibly due to pregnancy-related illness. During the first year after birth (year 1) the sickness absence gap decreases, which is expected given that the pregnancy has ended and that birthmothers are using parental allowance to a larger extent as compared to their partners.

The results found in table 6.3.1 are not directly comparable to the results found by Angelov et al. (2013:9)20, since they estimate the effect on monthly sickness absence. However, the trend in the sickness absence gap among opposite-sex couples found in table 6.3.1 is similar to the one found by Angelov et al. (2013:9). In year 2, birthmothers in opposite-sex couples utilize in general 1.8 days of sickness benefits more per year as compared to their partners. In year 3, birthmothers utilize 1.6 days more than their partners. Hence, the gap in opposite-sex couples increases as compared to pre-birth levels, which is consistent with the results found by Angelov et al. (2013:9). In year 4, the point estimate is small and positive, however not statistically significant, which implies that there is no significant difference in the sickness absence gap among opposite-sex couples in year 4 as compared to the pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre- pre-

20_{Angelov
et
al.
(2013:9)
studied
the
sickness
absence
of
opposite-‐sex
couples
having
their
first}

(34)

birth levels.

The evolvement of the sickness absence gap in same-sex couples after year 1 is different from that of opposite-sex couples. In year 2, the gap in sickness absence decreases by approximately 5.3 days (1.845-7.177) and in year 3 by 6.9 days (1.562-8.465) as compared to pre-birth levels. In year 4, the gap is also significantly decreasing. This indicates that after year 2, the within-couple difference in sickness absence is mowing towards a negative gap (where partners have a higher level of sickness absence)

The variable Same sex is insignificant. Thus, prior to childbirth, there is no general difference between the sickness absence gaps in opposite-sex and same-sex couples (see full table 6.3.2 in Appendix). This implies that there is no difference in pre-treatment trends between opposite-sex and same-sex couples, providing some suggestive evidence of the strict exogeneity model assumption.

The number of observed couples increases when estimating the gap in sickness absence. This is due to some opposite-sex couples being dropped when estimating the income gap because of years of missing values. No same-sex couples are however being dropped when estimating the income gap and since the sample of opposite-sex couples even after the drop is sufficiently big, I believe this will not be a problem with the inference.

Parenthood and Sickness Absence