Gender Inequality Reinforced: The Impact of a Child's Health Shock on Parents' Labor Market Trajectories

Abstract

This article employs a couple-level framework to examine how a child's severe illness affects within-family gender inequality. We study parental labor market responses to a child's cancer diagnosis by exploiting an event-study methodology and rich individual-level administrative data on hospitalizations and labor market variables for the total population in Finland. We find that a child's cancer negatively affects the mother's and the father's labor income. The effect is considerably larger for women, increasing gender inequality beyond the well-documented motherhood penalty. We test three potential moderators explaining the more negative outcomes among mothers: (1) breadwinner status, (2) adherence to traditional gender roles and conservative values, and (3) the child's care needs. We find that mothers who are the main breadwinner experience a smaller reduction in their household income contribution than other mothers. Additionally, working in a gender-typical industry and a child's augmented care needs reinforce mothers' gendered responses. These findings contribute to the literature by providing new insights into gender roles when a child falls ill and demonstrating the effects of child health on gender inequality in two-parent households.

Introduction

Despite advances in gender equality both in the labor market and at home, women continue to suffer from poorer labor market outcomes and contribute more to unpaid work within the household (Bianchi et al. 2000; Blau and Kahn 2017; Goldin 2014; Sevilla-Sanz et al. 2010). The literature on the motherhood penalty has shown greater income effects of childbirth for mothers than for fathers, as well as increases in within-couple gender inequality (Kleven, Landais, Posch et al. 2019; Musick et al. 2020).

However, less evidence exists on other situations in which couples renegotiate their division of labor in the household. In this article, we examine the effects of an unanticipated health shock of a child—namely, cancer—on parents' labor market participation and sharing of the responsibility of caring for a sick child. Our study combines research strands on the motherhood penalty, family spillover effects of health shocks, and gender inequality in care and the labor market.

A serious illness can have spillover effects, affecting not just the person who is ill (Fadlon and Nielsen 2021; García-Gómez et al. 2013; Jeon and Pohl 2017). For example, family members may need to provide special care at home when a person's health conditions limit normal functioning. Furthermore, the health shock of a loved one can cause stress and anxiety and change how family members value work and leisure time. Spillover effects are expected to be gendered, given that women traditionally carry the greatest burden of informal care (Bracke et al. 2008).

Causal evidence on spillover effects of children's health shocks is scarce. However, recent studies exploiting Nordic administrative data suggest that children's health shocks have a larger impact on the employment and mental health of mothers relative to fathers (Adhvaryu et al. 2023; Breivik and Costa-Ramón 2022; Eriksen et al. 2021; Gunnsteinsson and Steingrimsdottir 2019). Parental responses to a child's ill health can be considerable, raising the question of whether a child's health shock reinforces gendered behaviors. Contributing to this emerging literature, we test three potential moderators explaining the more negative outcomes we expect to observe among mothers: (1) breadwinner status, (2) adherence to traditional gender roles and conservative values, and (3) the child's care needs.

Rather than merely comparing the impact of childhood illness on the mother's and father's employment and income, as previous studies have done, we draw on important economic and sociological theories on the household division of labor and gender roles and norms. Moreover, following Musick et al. (2020), we analyze the impact on the mothers' share of the couple's labor income to study within-couple inequality as opposed to individual trajectories.

We contribute to research on gender inequality and the motherhood penalty (e.g., Budig and England 2001; Kleven, Landais, Posch et al. 2019; Musick et al. 2020) by extending the empirical focus from childbirth to a child's health shock. Although parenthood has been acknowledged as a turning point in the gendered household division of labor, less is known about the disruptions of a child's health conditions on domestic work arrangements. Furthermore, our study is an important addition to the literature on health spillover effects. To our knowledge, our study is the only one to apply a couple framework to study health spillovers except for Riekhoff and Vaalavuo (2021).

We use unique individual-level panel data from Finland that allow us to investigate the relatively rare event of childhood cancer without concerns about bias from nonparticipation or attrition. We use the event-study framework to assess labor market outcomes before and after the cancer diagnosis. We demonstrate that compared with many other health conditions among children, cancers are particularly suitable for causal inference in this framework because they are largely unrelated to parental socioeconomic characteristics or the child's prior health.

Our results show that mothers suffer the harshest economic penalty from children's health shocks. The magnitude of the effect is remarkable, even larger than when a woman herself has cancer (compared with results on breast cancer survivors in Vaalavuo (2021)). We contribute to the literature on gender inequality by demonstrating that children's severe health shocks aggravate within-couple gender inequality related to parenthood. This is particularly the case for households where the mother's contribution to the household's income was low before the shock; where the mother works in a female-dominated industry, indicating adherence to more traditional gender roles in general; and where childcare needs are likely to increase more because of the child's young age and cancer severity. Our results suggest that care needs arising from a child's health conditions lead to additional penalties beyond the post-childbirth motherhood penalty. Our results demonstrate that gender plays an important role in work and care decisions and that couples do not rely solely on economic optimization.

Theoretical Framework

Linked Lives and Spillover Effects of Health Shocks

Life course research emphasizes interdependence in decision-making in social relationships. According to Elder's (1987) linked lives principle, an individual's life course events should not be analyzed in isolation because people share the impacts of each event (Elder 1987; Settersten 2015). This perspective is relevant for understanding the interdependence of family members' well-being, including health spillovers, and the household division of labor.

Health shocks principally affect one family member but can also alter loved ones' lives. Several studies have considered the spillover effects of health shocks, usually between spouses. For example, Jeon and Pohl (2017) and Anand et al. (2022) found adverse effects of severe illnesses on spouses' labor supply, and Riekhoff and Vaalavuo (2021) argued that the association between a health shock and the individual's employment trajectory also depends on the couple's characteristics.

Less evidence exists on the spillover effects of children's health shocks. However, when comparing different types of family spillovers for four distinct illnesses, Lavelle et al. (2014) suggested that having a child with cancer is associated with the greatest spillover effects. Although pediatric cancer survival rates have increased, the use of intensive treatments has also increased. Hospital medical treatment is complemented by extensive periods of informal care at home and parental care management. Surgeries, chemotherapy, and radiation treatment may lead to significant pain for affected children (Hickman et al. 2013), which is also associated with parents' own emotional distress (Pöder et al. 2010).

Child health shocks likely produce gendered responses among parents. Survey evidence from the United States suggests that mothers are more likely than fathers to manage children's health and care for them when they are sick (Kaiser Family Foundation 2018). A sick child's increased care needs usually requires parental time off from paid work. The motherhood penalty literature suggests that child health shocks could lead to forgone professional experience and work promotions (Kleven, Landais, and Søgaard 2019) and to employer discrimination (Correll et al. 2007; Ishizuka 2021), with potentially substantial longer term consequences for highly skilled women because of the higher returns on their human capital (e.g., England et al. 2016).

In addition to exploring individual employment and income trajectories, we examine how a child's cancer diagnosis influences the mother's share of a couple's total income. If childhood cancer affects the mother and father equally, their labor income will decrease by the same proportion, leaving the mother's share of labor income unchanged. Given the well-documented motherhood penalty in the labor market, we expect the mother's relative economic position in the couple to deteriorate after a child's illness (Hypothesis 1). However, we also expect some couple and child characteristics to moderate this impact, as discussed later.

Household-Level Division of Labor: Between Comparative Advantage and Gender Norms

A growing body of literature has shown that childbirth negatively affects mothers' earnings and employment trajectories and explains much of the observed gender inequality in the labor market (Budig and England 2001; Kleven, Landais, Posch et al. 2019; Musick et al. 2020; Sieppi and Pehkonen 2019). New evidence from a Nordic setting shows that childbirth affects fathers only in the most recent cohorts (Nylin et al. 2021).

A prominent explanation for the motherhood penalty stems from Becker's (1991) theory on the household division of labor, which assumes that one spouse specializes in paid work outside the home and the other specializes in unpaid housework based on the principle of comparative advantage. The female share of breadwinning has increased for childless women (from 20% to 50% in Europe, according to Klesment and Van Bavel 2017) but remains low for mothers with young children. Research on the motherhood penalty has demonstrated that the pre-birth constellation of household income may matter for gender inequality after childbirth. Swedish evidence suggests that income differences between men and women resulting from childbirth grow with the within-couple gender pay gap before childbirth (Angelov et al. 2015).

Regarding time dedicated to childcare, evidence from Germany (Kühhirt 2012) and Spain (Sevilla-Sanz et al. 2010) suggests that women's contribution to childcare remains unchanged over the distribution of their relative earnings, whereas evidence from the United States (Raley et al. 2012) suggests that increases in mother's share of earnings in the household shift childcare time from the mother to the father. However, reaching gender equality in domestic work may require women to outearn their spouses by a substantial margin (Siminski and Yetsenga 2021). To better understand parents' (gendered) responses to a child's illness, we examine whether pre-shock relative income within the couple moderates our basic results. Consistent with the theory of comparative advantage, we expect that the mother's better economic position relative to her spouse protects against the negative impacts on labor market outcomes (Hypothesis 2a).

Alternatively, traditional gender norms and parents' gendered perceptions of caregiving quality could dominate over economic considerations, supporting the “doing gender” hypothesis (e.g., Bittman et al. 2003). For example, Andresen and Nix (2022) compared labor supply responses between heterosexual parents, same-sex parents, and adopting parents using Norwegian administrative data; they found that gender norms, discrimination, and perceived gender differences in childcare quality are the most prominent reasons for the motherhood penalty. Therefore, it could also be expected that breadwinner status before the health shock does not moderate the result, highlighting the persistence of gender roles in domestic work (Hypothesis 2b).

We aim to provide explicit evidence on the moderating role of gender norms by examining parents' adherence to traditional gender roles and conservative values. Nontraditional gender attitudes and values are likely associated with fathers' greater involvement in childcare. Research has found that men's conceptions of the mother's primary role in care, male-breadwinner ideology, and gender equality orientation influence the father's parental leave use—even in the Nordic countries, where gender-egalitarian policies thrive (Duvander 2014; Lammi-Taskula 2008; Närvi and Salmi 2019). Still, women's gender role ideology might affect maternal gatekeeping, limiting the father's involvement (Kulik and Tsoref 2010).

However, most studies have focused on economic aspects and workplace factors that affect the opportunities and constraints of staying home with a child rather than values and attitudes related to gender equality (e.g., Bygren and Duvander 2006; Närvi and Salmi 2019). We expect mothers to be less severely affected by a child's cancer when they adhere to nontraditional gender roles at work and have less conservative values, regardless of their breadwinner status (Hypothesis 3).

Urgent Childcare Needs and Gendered Responses

Whereas fathers' childcare time has increased, mothers' childcare time has not decreased (Bianchi et al. 2000). Women play a central role in care, and men's time spent caregiving is often contingent on the women in their family (Bracke et al. 2008; Gerstel and Gallagher 2001). Notwithstanding advances in gender equality, mothers are considered primary caregivers.

Women are also likely to carry the main responsibility of urgent childcare needs. A rare study on urgent childcare needs found that in such situations, 78% of women and 26% of men take leave from work (Maume 2008). COVID-19 pandemic lockdowns also provided valuable insights into gendered responses to urgent childcare needs, with evidence suggesting that most of the additional childcare fell to mothers (Hupkau and Petrongolo 2020; Johnston et al. 2020; Sevilla and Smith 2020). The gender difference in childcare occurred even among mothers who were breadwinners (Andrew et al. 2020).

A child's cancer is likely to affect the family's care needs differentially depending on the child's age and the cancer severity. We expect a stronger negative impact on mothers when the child falling ill is young and has a severe cancer type, indicating a higher care burden (Hypothesis 4).

Mental Health Impacts

A broad literature has examined the associations between childhood illnesses and parents' mental health (Bruce 2006; Kazak et al. 1998; Pinquart 2019). Recent studies from the Nordic countries have also demonstrated that a child's illness exerts a psychological toll on the parents (e.g., Adhvaryu et al. 2023; Breivik and Costa-Ramón 2022; Eriksen et al. 2021), with especially acute mental health symptoms when the child dies (Pohlkamp et al. 2019).

Consequently, the mental health effect of a child's cancer could explain part of the negative labor market impact on the parents. In addition to time off to attend to the child's care needs, the parents may need to take additional leave to recover from resulting stress, depression, and anxiety. For example, using Swedish data, Hjelmstedt et al. (2017) found an increased probability of sick leave for up to six years following a child's cancer diagnosis. Moreover, evidence suggests that caregiving is associated with worse psychological well-being and parenting stress because family caregivers are expected to take up complex caregiving duties with little preparation or support (Cousino and Hazen 2013; Northouse et al. 2012).

We provide evidence on the mental health effects of a child's cancer. However, given the complex interrelationships among the child's cancer and the parents' mental health, care responsibilities, and labor market outcomes, we cannot robustly analyze the role of mental health in producing different reactions among parents.

Institutional Background

In Finland, despite traditionally high rates of women's labor market participation and the high value placed on gender equality, mothers take considerably longer parental leave than fathers. Nylin et al. (2021) argued that such post-childbirth behavior reinforces the idea of the mother as the main caretaker, with a long-lasting impact on gender equality within couples.

Finland has a comprehensive welfare system and a universal public health care system. Virtually all childhood cancer cases are treated with little direct costs to families. An annual ceiling of 683 euros (in 2021) applies for health care–related user fees, after which individuals pay for short-term inpatient care (22.50 euros per day). A ceiling of 580 euros is also established for drug expenses. Health-related travel costs are reimbursed through National Sickness Insurance. The private sector plays a negligible role in childhood cancer treatment. In general, private health care users can apply for reimbursement from the National Sickness Insurance, which covers approximately 30% of expenses.

Childhood cancer treatment is centralized in five university hospitals (Helsinki, Tampere, Turku, Oulu, and Kuopio), ensuring standardized treatment protocols. However, Tolkkinen et al. (2018) found that childhood cancer mortality in Finland was lower in families with more educated parents and in families in the highest quartile of the income distribution. Using private services could lead to the earlier detection of childhood cancers and thus explain the mortality gradient.

Parents who take time off work can apply for a special care allowance paid by the Social Insurance Institution of Finland. Parents whose child experiences a severe illness are eligible for the allowance for a maximum of 60 days. The amount is based on their previous year's earnings. The Social Insurance Institution also provides a disability allowance when the need for regular care, attention, or rehabilitation lasts more than six months. Parents with children of any age can apply for support for informal care, which municipalities pay, but receipt of this benefit requires a high care burden. The benefit amount and criteria vary between municipalities. All families with children also receive child allowances paid universally, and childcare is free for the poorest families and inexpensive for all others.

Data and Methods

Data

We used Finnish register data of the total population for 1993–2019. We linked individual information on family relations, income measures, labor market status, socioeconomic characteristics, region of residence, and causes of death from the registers of Statistics Finland. We also had access to data on public special health care from the Care Register for Health Care (HILMO) provided by the Finnish Institute for Health and Welfare. These data include information on 1987–2019 inpatient stays, 1998–2019 outpatient visits, and the ICD-10 codes for the primary diagnosis. Because administrative sources routinely collect the data nationwide, the only sources of attrition are emigration and mortality. In addition, we derived detailed information on prescribed medicine from the register of the Social Insurance Institution. The data include ATC codes (i.e., anatomical therapeutic chemical), allowing us to identify psychotropic drug purchases in a given year.

Using the HILMO data and ICD-10 codes for the diagnosis, we identified children with cancer (all ICD-10 codes in class C). We focused on children diagnosed with cancer for the first time during a hospital stay in 1997–2014 at ages 0–19 to ensure that we could observe the parental outcomes four years before and five years after the cancer diagnosis. Because the outpatient data have some deficiencies in cancer diagnoses (namely, false positives), we used inpatient data only for identifying childhood cancer and the families affected by it.

Using personal identifiers, we linked children to their biological parents. We included only cohabiting parents living in Finland just before the cancer diagnosis. Recent studies have identified differences in the motherhood penalty between heterosexual and same-sex couples (Andresen and Nix 2022). However, we focused on heterosexual couples because of the rarity of childhood cancer combined with the low numbers of same-sex couples in the sample. We also excluded families with two or more children diagnosed with cancer and parents with data not available for the complete 10-year follow-up and missing values on the mother's share of labor income variable.

Matched Analytic Sample

We constructed a counterfactual of what would have happened in the absence of childhood cancer via matching. We matched the treated individuals with six unique controls (cases without a childhood cancer diagnosis) according to birth order (of parents' shared children), the parents' birth year, and the child's birth year. The choice of the number of matches is discretionary. As a robustness check, we tested how the number of matches (from one to six) affects results (see Figure S8; all figures and tables designated with an “S” are in the online appendix). Increasing the number of matches slightly increased precision from one to three matches but not much after that. Apart from the parents' birth year, which could vary by one year, background characteristics were matched exactly. We focused on couples and therefore restricted the matching to be based on children with two biological parents who lived together at the end of the year preceding the cancer diagnosis. We implemented matching using the Stata command calipmatch.

Matching ensured that the family structure was largely similar for the treated and the controls in terms of ages. We identified 2,476 mothers and 2,463 fathers of 2,476 children diagnosed with cancer using the criteria described earlier. These families were matched with 14,326 mothers, 14,250 fathers, and 14,326 children from unaffected families. Because we excluded parents who were not present throughout the follow-up or had missing household labor income values, the final sample does not lead to exactly six unique matches for every treated individual. Table 1 describes the study sample and compares our treatment and control groups in the period preceding the index year. The index year refers to the year of the child's cancer diagnosis for the treated individuals and their matched controls whose children had not experienced cancer. The index year for the control group can be described as the placebo diagnosis year. Key variables are described in more detail in the following section.

The treatment and the control groups are almost identical on the matched characteristics. The unmatched background characteristics—such as the child's gender, the parents' employment status, the mother's share of labor income, urban status, and moderating variables—are also well balanced (Table 1).

Main Outcome Variables and Moderators

We included several dependent variables. First, we measured each parent's annual income from labor (before taxes and social contributions) deflated to the 2019 price level using the Harmonized Index of Consumer Prices. To address outliers at the top of the income distribution, we top-coded at the 99th percentile. Additionally, we examined labor supply changes using parents' employment status at the end of the calendar year.

Second, to examine couple-level dynamics, we analyzed the impact on the mother's share of a couple's total labor income. This measure has been used as a proxy for the household power balance to study motherhood effects (Musick et al. 2020) and as a predictor of divorce (Bittman et al. 2003; Schwartz and Gonalons-Pons 2016).

Third, we examined the effects on parents' mental health. We created a dummy variable for every year indicating a parent's mental health treatment (0 = no psychotropic drug use, 1 = any psychotropic drug use) by using ATC codes N03 (antileptics), N05 (psycholeptics), N06A (antidepressants), N06B (psychostimulants), and N06C (combination of psycholeptics and psychoanaleptics). Besides psychotropic medication, we also separately analyzed antidepressant use by considering annual purchases of psychotropic drugs provided by the Social Insurance Institution of Finland. Next, we created a dummy variable indicating inpatient or outpatient psychiatric admission with a psychiatric diagnosis (relying on ICD-10 codes in the F class).

After analyzing the main effects on individual income, employment, the mother's share of the couple's income, and mental health, we investigated the role of potential moderators. In these models, our outcome variable was the mother's share of the couple's income. First, to examine income trajectories from the perspective of comparative advantage within the household, we used the mother's breadwinner status as a moderator of the impact—that is, the mother's share of the couple's total labor income in the year preceding the cancer diagnosis. We also constructed this variable using income information from the four years preceding the cancer diagnosis to avoid the potential bias from temporary fluctuations in income.

We categorized couples into one of three breadwinner groups: (1) the mother's share of the couple's total labor income was less than 40% (secondary earner), (2) the parents had roughly equal income shares of 40% to 60% of total earnings (equal earners), and (3) the mother's share of the couples' total labor income was more than 60% (main breadwinner). Previous research has used this type of discretization of the mother's income share to study breadwinner status heterogeneity in marriage and cohabitation dissolution (Kalmijn et al. 2007) and the gendered impacts of spousal residential mobility (McKinnish 2008).

Second, to study the moderating role of adherence to traditional gender roles or conservative values, we used two variables as proxies, given that administrative data do not provide this information directly. We built a variable indicating the gender balance in the mother's occupational sector. We characterized sectors composed of more than 70% women or men as female- or male-dominated sectors, respectively. Using this information, we regard mothers as acting consistent with traditional gender roles when they work in a female-dominated sector. To assess conservative values, we used information on 2008 voting behavior in the municipality of the child's birth. We used the share of votes received by the Christian Democrats (the most conservative party in Finland) in the municipal elections as a proxy for parents' conservative values (below or above the median share of Christian Democratic votes).

Third, to understand how childcare needs moderate the gendered response, we used cancer type, the child's five-year survival, and the child's age to measure the differential burden of care. We divided cancer types into acute lymphoblastic leukemia/lymphoblastic lymphoma (ALL/LBL), central nervous system (CNS) cancers, and other cancers. This categorization is commonly used to distinguish more severe childhood cancers (ALL and CNS) from all other childhood cancers. For instance, CNS treatments often involve high-dose cranial radiation therapy, and CNS cancer survivors may experience substantial cognitive declines (Mulhern et al. 2004).

Furthermore, cancer survival could be linked to more extreme contrasts than cancer type. This measure potentially reveals whether gendered responses arise from the bereavement process (death) or whether gendered responses also occur when a child survives. However, both can be associated with a high care burden. The child's age at cancer diagnosis is linked to baseline nurturing needs that are exacerbated by a cancer diagnosis. We examined three age groups: preschool age (0–6), primary school age (7–15), and secondary school age (16–19). The last threshold is also relevant because families with children affected by cancer may receive additional disability benefits until the child turns 16.

Research Design

To address the endogeneity between health and socioeconomic variables, research has increasingly used health shocks as a potential source of exogenous variation in health. Among adults, health shocks are often linked to lifestyle factors, but childhood cancer is not systematically associated with observed individual- or family-level characteristics (Cancer Research UK 2021). Because childhood cancer is unanticipated, concerns about anticipation effects or reverse causality are minor. Therefore, childhood cancer provides a solid empirical case for identifying the causal impact of a health shock on parental outcomes.

Figures S1–S6 illustrate the differences between families with a child hospitalized for the first time for more than four days with a specific diagnosis and families with similar family compositions whose children were never hospitalized (after childbirth). We demonstrate that childhood cancers are negatively related to household labor income (Figure S1), household taxable income (Figure S2), and parental education (Figures S5 and S6), but none of these correlation coefficients are statistically significant. Other health conditions typically have a stronger and statistically significant negative correlation with household labor income. The correlation of childhood cancer with the Apgar score (Figure S3) is very close to 0 and its correlation with birth weight is positive and statistically significant (Figure S4).

The motherhood penalty literature has recently started using an event-study framework focusing on sharp changes in labor outcomes between men and women around the transition to parenthood (e.g., Kleven, Landais, Posch et al. 2019; Musick et al. 2020). The approach typically examines pretreatment periods to construct a counterfactual prediction for posttreatment periods. We instead created an explicit control group from the population of untreated families. We employed a dynamic difference-in-differences design in an event-study framework in which affected families were matched with unaffected families. (See Figure S8 for extensive robustness checks with other estimation strategies and a related discussion.) Our estimated linear model can be written as follows:

where the dependent variable $Yi,t,r$ is the outcome of interest (employment, annual labor income, mother's share of the couple's labor income, or mental health) of parent i observed in period r after the child's cancer diagnosis year. I_r represents indicators relative to the index diagnosis year (actual cancer diagnosis for treated parents and placebo diagnosis for the matched controls).

The treatment variable is an indicator variable, $Ci$⁠, equal to 1 for parent i whose child was diagnosed with cancer at period r = 0 and 0 for a matched parent i whose child was not diagnosed with cancer. The parameter vector of interest, $δr$⁠, represents the differences in changes in labor market outcomes between the treatment and the control groups in relative time r. To interpret the income-related estimates in relative terms, we scale the absolute impact (in euros) with the predicted outcome from the parents in the control group (⁠$Y^r0$⁠) for each relative period r to form the estimates for the relative effect, $δrrel=δrY^r0$⁠, representing the percentage change in the outcome.

$X′it$ represent time-varying control variables for the parent and the child. We include only fixed effects for the mother's and child's age in $X′it$⁠. They adjust for age-related differences in parents' labor productivity and labor supply. Calendar year effects, $πt$⁠, and age fixed effects control for potential time-related differences in employment and income that might arise between the treatment and control groups—for instance, because of differences in age profiles and the macroeconomic situation.

The results regarding the mother's share of labor income included age fixed effects of only the mother. We also included fixed effects for the father's age as a robustness check. Doing so resulted in negligible changes in point estimates, and we therefore used only the fixed effects for the mother's age in these analyses. $θi$ represent individual fixed effects.

The underlying assumption in the difference-in-differences design is that the treatment and control groups would have followed similar trends in the absence of the cancer shock. We tested this assumption by examining the coefficients of interest before the shock (r  <  0) and reporting the p value of the hypothesis test of all pre-shock coefficients of interest being 0.

To test the moderating effect of (1) the mother's breadwinner status, (2) adherence to traditional gender roles and conservative values, and (3) the child's care needs, we used the triple-difference estimator to test the statistical significance of heterogeneity. To aid the interpretability of the results and enhance the detectability of subtle responses, we transformed the relative time variable into three categories. Our model for moderation analysis is as follows:

where the independent variable P_r represents the categorical variable based on relative time $r$⁠, taking a value of 0 in years before the index diagnosis (r  <  0), 1 in relative Periods 0–2 (short-term effect), and 2 in relative Periods 3–5 (long-term effect). $Mi$ is the moderator. The estimations were conducted using the mother's share of labor income both without and with subgroup-specific scaling.

Furthermore, to illustrate the effect of childhood cancer relative to the motherhood penalty following childbirth, we conducted an additional analysis of the mother's income trajectories both after birth and after a childhood cancer diagnosis. In this analysis, we compared three groups of women: (1) women with a child with cancer, (2) women with a child without cancer, and (3) women who had fertility treatment in public health care that was unsuccessful in the first five years. The first two groups are counterparts in our main analyses: women with a childhood cancer shock and their matched counterparts. The last group provides a counterfactual for women with children in general. We compared the women in infertility treatment with women who had a child with cancer, similar to our comparison in the main analyses. However, because these women were negatively selected in terms of household income, we used education, mother tongue, and municipality of residence as additional matching variables. The expected birth year for women with unsuccessful fertility refers to the first fertility treatment recommended after one year of unsuccessful reproduction efforts.

Results

Impacts at the Individual and Couple Levels

Figure 1 presents results for the individual-level effects for mothers and fathers separately. Panels a and b display results for labor income and employment (scaled by the predicted outcomes of families in the control group) and the corresponding 95% confidence intervals following a childhood cancer diagnosis normalized at the year preceding the first diagnosis (r  =  −1), using unaffected families as a control group, as specified in Eq. (1).

Diverging trends in the periods preceding the index year would undermine the common trends assumption and suggest that the control group does not provide a plausible counterfactual for the treated group. However, in our case, there are no underlying trends, such as anticipatory behavior, that would undermine the causal interpretation of the estimates. A further look at the outcomes reveals that employment and labor income between the treatment and control groups are parallel before r = 0 (Figure S7).

A child's cancer diagnosis reduced mothers' annual labor income by approximately 14% in the diagnosis year and by 28% the following year. The reduction then stabilized at 6% relative to the control group five years after the cancer diagnosis (panel a). The evolution of the impact on labor income closely followed the changes in employment probability (panel b).

Fathers were also affected, albeit much more modestly. A child's cancer reduced annual labor income by 4% in the diagnosis year and by 6% the following year. The effects decreased thereafter, becoming statistically nonsignificant for the last three years of the follow-up. The point estimates for fathers' employment probability are considerably smaller and not statistically significant at the 5% level. The results confirm our Hypothesis 1 on gender differences: mothers experienced significant effects on income and employment, while fathers' income was negatively affected only in the short term and by a smaller magnitude.

A child's having cancer also increased parents' mental distress, as proxied by psychotropic medication use (panel c). One year after cancer diagnosis, the probability of psychotropic medication use increased by almost 6 percentage points for mothers (from the sample baseline of 8.5%) and by 3 percentage points for fathers (from the sample baseline of 5.9%). On average, childhood cancer was related to an increase in psychotropic medication use of 4.5 percentage points for mothers and of 1.6 percentage points for fathers during the entire follow-up. Mothers were also more affected in their probability of psychiatric admission and antidepressant medication use (Table S2). However, studying these mental health effects on labor market outcomes and gender differences therein is difficult because of the interlinkages among cancer severity, care needs, work leaves, and psychological turmoil.

Furthermore, panel d illustrates that childhood cancer had a negative impact on the mother's income share. At baseline, the mother's share was 39%, on average. Relative to unaffected mothers, the mother's relative income contribution dropped by almost 3 percentage points (7%) in the year of the cancer diagnosis and by 6 percentage points (16%) the following year. Although childhood cancer appears to affect the within-couple dynamics in income contribution in the short term, the results suggest that this impact is temporary, in contrast to the penalty related to childbirth.

Almost half of pediatric cancers occur at ages 0–6, when mothers are usually still affected by the motherhood penalty and have particularly low labor market participation relative to men. We conducted an additional analysis to investigate the effect of childhood cancer on mothers relative to the motherhood penalty due to childbirth. Figure 2 compares the predicted labor income trajectories of three groups: (1) women with a child with cancer, (2) women with a child without cancer, and (3) women undergoing fertility treatment in public health care unsuccessfully in the first five years.

In essence, the figure compares women in families with childhood cancer with two distinct alternative scenarios: (1) labor market performance in the absence of childbirth and (2) labor market performance after childbirth but no childhood cancer diagnosis. We see that the first childbirth produced a labor income deficit of more than 50% a year after childbirth relative to women with unsuccessful fertility treatments. Over time, the deficit relative to childless women narrowed, but mothers whose firstborns were diagnosed with cancer at age 2 suffered an extra penalty that was at its largest one year after a cancer diagnosis or three years after childbirth. Over time, the mother's labor market performance recovered from the child's cancer diagnosis but was still associated with an annual average deficit of 15% during the three years following the cancer diagnosis.

One concern in interpreting these results is potential differences in union dissolution patterns between the treatment and the control groups. For example, because childhood cancer causes parental stress and anxiety, it may also increase union instability. Because custody arrangements are gendered, union dissolution would increase the mother's childcare responsibilities and gender inequality in labor market outcomes. However, Figure S12 (panel a) shows that, if anything, a child's cancer had a minor positive effect on parents' union stability (i.e., cohabitation), especially when the child survived cancer (panel c). Furthermore, our analysis shows that a child's cancer diagnosis had no detectable impact on fertility in general (panel b) but a positive impact on fertility in the long term when a child died during the follow-up (panel d). Therefore, we conclude that the changes in parents' labor income or parents' relative income shares are not driven by union dissolution or fertility responses in general.

The Role of Potential Moderators

Next, we present the role of potential moderators for gendered responses, examining the differences in the impact of a child's cancer on the mother's labor income share by (1) the mother's breadwinner status, (2) adherence to traditional gender roles and conservative values, and (3) childcare needs.

Contrary to our expectation, we do not find statistically significant differences in the impacts on the mother's labor income share by breadwinner status (Table 2). Secondary-earner mothers saw their income share drop by 4 percentage points in the short term and by almost 2 percentage points in the long term. The point estimates are fairly similar to those of equal earners (−4.1 and −1.3 percentage points, respectively) and breadwinners (−3.2 and −1.5 percentage points, respectively). However, the baseline shares differ for these groups, and the interpretation depends on whether the responses are examined at an absolute or relative scale. When using the relative scale, we find that the labor income share dropped more for mothers who were secondary earners (−14.5%) relative to equal-earner mothers (−8.7%) and breadwinner mothers (−4.3%). These results support the comparative advantage hypothesis (Hypothesis 2a).

On a more technical note, the breadwinner variable we used may be problematic because it is based on short-term income, which is prone to fluctuations. Especially in households with small children, the correlation between current income and lifetime income can be particularly low for women. Therefore, we also used an alternative measure of breadwinner status based on relative income averaged over four years before the index diagnosis. This measure reinforces the interpretation that breadwinner mothers would be less affected by a child's cancer diagnosis in the short term, even though the results are not statistically significant at the absolute scale. Furthermore, the results using modified breadwinner status when the extreme cases in relative income are omitted do not affect our general conclusions.

Second, we investigated the moderating role of traditional gender roles and conservative values with our two proxy variables: an indicator for living in a municipality where the vote share of the conservative party (Christian Democrats) is higher than the median and a variable for working in a female-dominated industry. We find no differences in the impacts of childhood cancer by our proxy for conservative values (Table 3). However, the short-term impact of childhood cancer on the mother's income share was stronger (−5.6 percentage points, or −12.4%) for women working in female-dominated industries (i.e., for mothers adhering to typical gender norms) than for women working in male-dominated industries (−1.5 percentage points, or −3.3%). The effect disappeared in the long term. Therefore, the results partially support Hypothesis 3 on gender roles and values as a moderator of gendered responses.

Third, we shift our focus to potential moderators linked to augmented childcare needs that could align behavior with traditional gender roles. Young age at cancer and cancer severity are factors associated with increased care needs. Additionally, we consider five-year child survival after the cancer diagnosis as an indicator of cancer severity. We find considerable heterogeneity in the child's age at cancer diagnosis (Table 4). Mothers of a child with cancer in adolescence did not experience a within-couple penalty, but mothers with younger children did. The negative effect was most pronounced for mothers with preschool children.

The age effect is linked to the severity effect: children of different ages are affected by different cancer types. In general, the same result on the moderating effect of higher care needs emerges when different cancer types are examined. Acute lymphoblastic leukemia and lymphoblastic lymphoma (ALL/LBL) caused the largest short-term drop in the mother's income contribution (−6.1 percentage points, or −15.4%). Central nervous system (CNS) cancers resulted in smaller short-term decreases (−3.8 percentage points, or −9.7%), as did other cancers (−3.6 percentage points, or −8.9%). In the long term, though, CNS cancers resulted in the largest deficit (−3.2 percentage points, or −8.2%), and ALL/LBL cancers did not statistically differ from other cancers.

Additional evidence shows that whereas CNS cancers had the highest probability of death (Table S3) and the highest disability benefits in the long term (Figure S10), ALL/LBL had the highest treatment intensity, especially in years immediately following the cancer diagnosis (Figure S11). These factors are reflected in the point estimates in Table 4.

Part of the heterogeneity related to cancer types might stem from survival differences. Although cancer-related mortality in childhood is relatively rare (16.5% in a five-year follow-up in our data) and standard errors are large, we find the effect to be larger when the child dies, especially in the long run. Overall, we find support for Hypothesis 4 of a more negative impact on mothers when the child falling ill is young and has a severe cancer type, indicating a higher care burden.

Accurately distinguishing the relative importance of the potential mechanisms driving gendered responses is difficult, given that many of these moderators correlate with one another. For instance, cancer severity, survival, the higher negative impact on gender inequality, and fertility responses are interlinked. Therefore, increased fertility might be one underlying mechanism behind heterogeneous responses in cancer severity, and the most severe child health shocks might induce an additional motherhood penalty.

Discussion and Conclusions

Despite remarkable advances toward greater gender equality, care responsibilities throughout the life cycle seem to fall on women's shoulders. The motherhood penalty—the drop in a mother's earnings after childbirth—has been well documented among Western countries, including Finland (e.g., Kleven, Landais, Posch et al. 2019; Musick et al. 2020; Nylin et al. 2021; Sieppi and Pehkonen 2019). However, we know much less about other situations in which parents and couples renegotiate the household division of labor. Therefore, in this article, we examined how a child's health shock affects within-couple inequality and parents' labor income and employment. As a significant contribution to the emerging literature on the topic, we tested three potential moderators of gendered responses: (1) the mother's breadwinner status, (2) adherence to traditional gender roles and conservative values, and (3) childcare needs.

Our study benefited from rich individual-level panel data, allowing us to investigate manifold repercussions of the relatively rare event of childhood cancer. The strong connection between socioeconomic status and health makes the causal inference between health and labor market outcomes challenging. We approached this obstacle by focusing on the unexpected health shock of childhood cancer. We demonstrated that families experiencing childhood cancer and other families are very similar to each other before the health shock, which is not necessarily true for other childhood health conditions. Thus, childhood cancers seem particularly suitable for assessing the spillover effects of family health shocks. The matching method, event-study approach and our focus on childhood cancer allow a plausible causal inference.

We found a significant negative impact on all economic outcomes studied and a clear gendered impact, with mothers bearing most of the economic burden. These findings are in line with other studies from the Nordic countries (Adhvaryu et al. 2023; Breivik and Costa-Ramón 2022; Eriksen et al. 2021). Moreover, the mother's share of the couple's total income decreased by more than 4 percentage points in the short term (up to two years after the cancer diagnosis). The impact of a child's cancer on mothers' earnings is even larger than the previously documented impact of breast cancer on Finnish working-age breast cancer survivors (Vaalavuo 2021). This highlights the fact that spillover effects of health shocks on close family members should be considered in both economic evaluations and targeted psychosocial interventions.

The results regarding a mother's breadwinner status partially support the theory of comparative advantage: mothers who earn less than their spouse experienced a larger relative drop in earnings. These results are in line with Swedish evidence that the lower the mother's relative income contribution, the greater her motherhood penalty (Angelov et al. 2015). However, in households where mothers are breadwinners, men provide a low share of household income, on average (approximately 22%). Therefore, large absolute impacts for mothers relative to men may not affect the relative income shares greatly.

In general, we found that gender is still influential and that couples do not rely solely on economic optimization. To understand the moderating effect of gender roles and conservative values, we examined heterogeneous effects by the gender balance in the mother's occupation and conservative voting in the area of residence. Although conservative voting did not moderate the gendered responses, mothers working in female-dominated industries (i.e., mothers more likely to adhere to gender-typical roles) are significantly more affected by the child's cancer than mothers working in male-dominated industries. However, our proxies for parents' attitudes and values are coarse, and register data do not contain direct measures of these personal characteristics. Also, our measure of adherence to traditional gender roles based on the gender balance at the mother's occupation might indicate workplace characteristics affecting parental decision-making rather than the mother's attitudes. Previous studies have suggested that workplace characteristics influence fathers' parental leave take-up (e.g., Närvi and Salmi 2019).

As an important moderator of gendered responses, the impact on a mother is more negative when the child has augmented care needs because the child has a severe cancer type or the cancer diagnosis occurs at a very young age. However, breadwinner status, young age at diagnosis, and cancer severity are interlinked, and the relative importance of these moderators is hard to distinguish, partly because of statistical power issues.

Previous research has illustrated that a child's illness negatively affects parental mental well-being (e.g., Bruce 2006; Kazak et al. 1998; Pinquart 2019), and our data are consistent with this. Nevertheless, it is challenging to disentangle the causal direction between a child's cancer, parents' labor market responses, and their mental health after a cancer diagnosis. The conflict between paid work and caring for a sick child, the reduction in earnings, and intensive caretaking might enhance the negative psychological effects of a child's illness (e.g., Northouse et al. 2012). The role of mental health in producing different labor market outcomes provides a valuable avenue for future research.

Confirming findings from Breivik and Costa-Ramón (2022) and Adhvaryu et al. (2023), we found that the increased probability of psychiatric admissions and psychotropic drug use following cancer is larger and more persistent for women than men. This finding suggests that long-term psychiatric symptoms following a child's health shocks might be related to the long-term negative impacts on mothers' income.

Our results show the importance of acknowledging spouses' linked lives and analyzing spouses together rather than as isolated individuals (Riekhoff and Vaalavuo 2021; Settersten 2015). This strategy is likely to illuminate inequalities both between and within families. We provide evidence suggesting that the couples' joint decision-making in the described circumstances cannot be explained by straightforward economic optimization or gender norms and that many potentially important factors are interrelated.

Stronger negative effects on mothers could have long-term consequences for gender equality. According to the signaling theory, employers may perceive mothers as less committed to their jobs (Nylin et al. 2021). We illustrate that in addition to the motherhood penalty, other circumstances requiring more care at home can reinforce gender inequality. Our study adds to the literature on the couple perspective in studying parenthood-related gender inequality. Whereas parenthood is widely acknowledged as a critical point in men's and women's diverging career paths, the role of child health has been less explored in this context.

In the future, comparing the effectiveness of the welfare state's capacity to cover income losses due to illness in the family across countries would be an interesting endeavor, with important differences likely to emerge. The few previous studies on the topic have focused on the Nordic countries with their relatively similar welfare systems, gender ideologies, and men's and women's labor market participation. In addition to incurring productivity costs, a child's illness leads to a temporary reliance on social transfers. Thus, health shocks are associated with societal costs that are much higher than costs in the health care sector. A more comprehensive view of the costs of illness is warranted in debates on health policies and the economic evaluation of treatments (see also Al-Janabi et al. 2015). Institutional arrangements can also affect gender differences in the responses and overall opportunities and obstacles within couples and between couples of different socioeconomic statuses.

Acknowledgments

Maria Vaalavuo acknowledges support from the Kela (Social Insurance Institution of Finland) Research Foundation and SYLVA Foundation. Henri Salokangas acknowledges support from the Kela Research Foundation and the INVEST Research Flagship Centre. Ossi Tahvonen acknowledges support from the Kela Research Foundation and the Yrjö Jahnsson Foundation.

References