In this section, we present an analysis of how the life satisfaction of different parental sub-groups is associated with parental leave in the Netherlands. Firstly, we examined how the Dutch parental leave scheme may shape the relationship. Secondly, we explored the influence of socio-demographic characteristics on the estimate impact of parental leave on life satisfaction.
Variation in Parental Leave Scheme
The estimated coefficient of our linear fixed-effects model by length of parental leave is shown in Table 9, panel a. Taking parental leave is related to an average life satisfaction increase of about 0.3 points for parents who are on parental leave for one month up to one and a half years. A similar significant coefficient size is found for parents during parental leave of between six months to eighteen months. Parents on parental leave for more than one and a half years, however, no longer enjoyed the benefits of this leave. This result may be explain by the process of hedonic adaptation (Brickman and Campbell, 1971) which suggest that individuals return to baseline levels of happiness following a change in life circumstance24.A possible explanation is that after two years parents get used to their part-time work arrangement and their extra free time no longer increases their life satisfaction. Additionally, Adema et al. (2015) find that across European countries negative wage and slower career opportunity progression largely follow long periods of leave from work, e.g. one or two years or more. As a consequence, when the parental leave arrangement is spread over an extended period it may generate negative work outcomes for career progression and wages that offset the work-life balance benefits of taking parental leave.
Estimated effects of parental leave weekly hours on life satisfaction are presented in Table 9, panel b. The size of estimated coefficient does not differ significantly by the number of weekly working hours of parental leave. A reduction of working hours may impact life satisfaction in two contrasting ways. Firstly, taking more hours per week off may help an individual to balance life and work in a better way, leading to an increase in life satisfaction. Secondly, in contrast, reducing an individual weekly working hours can induce lower earnings, reduce their capacity to deal with work demands, restrict their career opportunities, and encourage negative Parental leave and life satisfaction judgments from co-workers (Garnero, 2016). Hence, taking beyond a certain amount of leave per week may have negative work outcomes for an individual: reduc-ing weekly working hours may exceed the work-life balance benefits of parental leave.
Caregiving and Health Outcomes
Informal care and psychological health outcomes are linked because caregiving implies perceived overload due to the difficulty of combining leisure time, family duties, work demands and care tasks, and because the decline in health status of the care recipient affects one’s emotions negatively. In this regard, in a review of the literature, Schulz et al. (1990) indicated that caregivers tend to show an above-average level of psychiatric symptoms. Additionally, Bom et al. (2018) summarised different studies showing that caregiving resulted in higher prevalence of depressive feelings and lower mental health scores. Estimates of the physical health effects of informal care were more ambiguous. Caregiving required physically demanding duties to be carried out over a long duration, thus it might lead to an unhealthy life style, stress and lower psychological health, possibly inducing hypertension and cardiovascular diseases (Pinquart and Sörensen, 2007).
Informal caregivers are not all equal when it comes to health issues. An extended literature highlighted how the impact of caregiving on physical and mental health varied depending on specific socio-demographic characteristics. The negative health effect of caregiving was larger for married individuals Bom et al. (2018) and working female caregivers Kenny et al. (2014). According to Llacer et al. (2002), spousal caregivers had a lower socio-economic status, poorer health and a lower level of well-being than child caregivers; however, child caregivers were significantly more burdened.
The patterns of care also matters. For instance, providing informal care to close family members induced a larger subjective burden than caring for non-family members (García-Castro et al., 2019; Kramer, 1997; Bom and Stöckel, 2021). Bom and Stöckel (2021) explored whether the health impact of providing informal care differs between the United Kingdom and the Netherlands. They found that for both countries, individuals providing more than 20 hours of informal care per week, and those who faced a double burden of care and full-time employment experienced the most severe negative health effects. They highlighted, however, some differences between the two countries with Dutch low intensity caregivers experiencing smaller negative mental health effects than British’s. In turn, women in the Netherlands experienced a larger mental health burden. Pinquart and Sörensen (2007) found that both care recipients’ behavioral problems, e.g., disruptive and aggressive behavior, and the time spent on caregiving, placed a burden on the caregiver and increased symptoms of depression, with aggressive behavioral problems being particularly important when caring for people with dementia. Additionally, the authors pointed out that the most severe physical impairments were more likely to occur for older male caregivers in charge of dementia patients, while women bore higher psychological costs due to a higher perceived care burden.
The adverse impact of caregiving, however, could be softened by the use of psychological resources such as mastery, coping strategies, social support, and having sufficient financial resources. Jansson et al. (1997) demonstrated that informal caregivers meeting other caregivers in the same situation increased their spirit of community, their knowledge of caregiving and their ability to handle their personal situations. Lin et al. (2013) provided evidence that the correlation between caregivers’ duties and the caregivers’ level of depression was weaker when participants had a high level of feedback from others or had a good parent-child relationship. In another study, García-Castro et al. (2019) found that caregivers experiencing the greatest burden were those who perceived a decreasing leisure time and who were under high financial stress. They also found that personality traits such as hope, zest, social intelligence and love mediate the relationship between perceived stress and care burden.
The process of caring may generate negative feelings like stress because it is phys-ically and mentally demanding. It creates a perceived burden that varies depending on the other duties that caregivers have to discharge, and on the psychological, financial and external resources they have at their disposal. Yet, some studies high-lighted the positive effects of providing care (Ashworth and Baker, 2000; Grünwald et al., 2021). Caregivers could derive positive utility from the process of caring it-self, through an increase in self-esteem or by developing an affinity with the care recipient. Cohen et al. (2002) found that caregiving was associated with positive aspects such as companionship and a sense of it being fulfilling and rewarding.
Caregiving and Life Satisfaction
Few researchers analysed the effect of providing informal care on subjective well-being. Most of the studies dealing with this topic focused on health out-comes. Since the seminal article of the American economist Easterlin (1974), the economists’ theoretical debate on utility has shifted from an objective approach based on the concept of decision utility to an acceptance of a subjective approach. In this context, economists consider that subjective well-being can be used as a proxy for measuring subjective utility. According to the four- fold quality-of-life matrix developed by Veenhoven (2000), both concepts concern the inner qualities of individuals. Subjective well-being implies inner appreciation of life, while health is an individual objective condition for achieving well-being. However, these are different conceptions of quality of life. The former implies a self-appraisal of one’s overall life while the latter focuses on the degree to which one’s life meets the explicit normative standards of what defines a “good life”. Thus, subjective well-being reflects one’s past experiences, cognitive appreciation of life, and overall feelings of pleasure and pain. Moreover, the development of measures of social progress and well-being that go “beyond GDP” has seen a boom in recent decades. New measures of GDP have been proposed in policy circles, such as the better Life Initiative in 2011 (Durand, 2015): this framework measures well-being by considering 11 dimensions covering both current material conditions and quality of life including a measure of satisfaction with life.
Among the few studies focusing on how informal care could affect subjective well-being, mixed evidences were found. Collecting data on informal caregivers in Sweden, Borg and Hallberg (2006) determined that a high frequency of care-giving decreased life satisfaction, while no significant difference existed between less-frequent caregivers and non-caregivers. Using panel data from the Household, Income and Labour Dynamics in Australia survey (HILDA), Leigh (2010) studied the effect of informal care for an elderly or disabled person on labor market outcomes, including life satisfaction. The author found that informal caregivers had a lower level of life satisfaction than non-carers, although this effect became insignificant when individual fixed-effects were taken into account. Weatherly et al. (2014) used eleven waves of the HILDA to estimate the impact of informal caregiving on self-reported well-being. They applied a fixed-effect ordered logit and found that providing informal care had a negative effect on subjective well-being.
Both Bookwala (2009) and Chen et al. (2019) focused on female informal caregivers and provided different results. Based on a US sample of adult daughters and sons, the former found no significant effect of parental care on caregivers’ life satisfaction. On the contrary, the latter, using three waves of the China Health and Nutrition Survey (CHNS), showed that informal care significantly reduced the subjective well-being of female caregivers using the Instrumental Variable (IV) ordered probit model. The caregiving effect on subjective well-being of female caregivers was more significant for rural caregivers than for urban’s. Additionally, the authors identified two channels of “wealth” and “health” through which informal care lowered subjective well-being. mOther studies focused on the monetary evaluation of informal caregiving. Van Den Berg and Ferrer-i Carbonell (2007) examined the compensating variation necessary to keep the same level of well-being among Dutch informal caregivers. They estimated that an extra hour of informal care was worth about nine to ten euros, falling to about eight to nine euros if the care recipient was a family member and to about seven to nine euros if not. MacDonald (2019) used the well-being valuation approach to estimate and monetize the well-being impact of informal care provision on caregivers. They found that permanent income would have to increase by approximately 102k pounds per year on average to compensate for well-being losses due to informal caregiving.
Sensitivity Analysis: Selection Bias and Propensity Score Matching
The observed effect of informal caregiving on life satisfaction might result from the self-selection of individuals into the provision of informal care. In other words, can specific personal characteristics predispose individuals to self-selection into informal care provision? More precisely, the “selection in” caregiving refers to people deciding to become caregiver (Do et al., 2015). We know from the literature that the individuals who become caregivers and keep on providing support over years are more often women, poorer and have lower opportunity costs (for a review, see Bauer and Sousa-Poza (2015)). Under these conditions, the level of life satisfaction that underemployed older women without children and working few hours would have reported if they had not provided informal care remains unclear. Another criteria inducing self-selection into caregiving might be mental health (Coe and Van Houtven, 2009). We might reasonably wonder whether health status determine who will provide care inside a family (Schulz et al., 1990). Is the unhealthiest child, compared with her siblings, less likely to care for her parents? Thereupon, the selection of caregivers with respect to health is increasing with age, as health deteriorates over time (Easterlin, 2003), meaning that age is also a determinant of the selection into caregiving duties. On top of that, we might wonder how life satisfaction, which is worsening with the decline in health, impacts selection into caregiving. For instance, we might worry that people who are least satisfied with their life have lower propensities to become caregivers, or that individuals need a given degree of satisfaction with their own life before diving into caregiving activities (Coe and Van Houtven, 2009).
Propensity score matching reduces this selection bias by comparing the happi-ness of informal caregivers to that of non-caregivers (Rosenbaum and Rubin, 1983; Caliendo and Kopeinig, 2008) who are as similar as possible in all other respects. This methodology has recently been applied in other happiness studies (Binder and Coad, 2013; Nikolova and Graham, 2014; Tiefenbach and Kohlbacher, 2015; Hessels et al., 2018; Arampatzi et al., 2018) and could be compared to a randomized control trial in which two groups of individuals are randomly assigned to the treatment under study or to a control group.
The treatment is the informal care provision. The effect of the treatment is referred as the Average Treatment effect (ATE) and it can be defined as the difference between informal caregivers and non-informal caregivers regarding their expected life satisfaction. For the purpose of our present research, we use the Nearest-Neighbour Matching estimator, which is often used in propensity score matching (Becker and Ichino, 2002). We chose this matching estimator because there are many comparable untreated respondents in the common sample (Caliendo and Kopeinig, 2008), that is to say, many respondents that do not provide care. To apply the Nearest-Neighbor Matching method, of which the minimum matching request between observations is 1, respondents are matched on the following characteristics: gender, age category, mean of life satisfaction, occupational status, number of children, log of standardized net household income, weekly working hours and year, allowing us to match individuals with similar characteristics within a year. We also correct for a large-sample bias that exists when matching on more than one continuous covariate (Abadie and Imbens, 2006, 2011), namely log of standardized net household income and weekly working hours. The first model displayed in Table 2.6 shows the difference between the treated and the untreated based on the matching criteria mentioned above and considering the common sample. The significance of the difference means that there is a clear and negative effect of the treatment.
Alternative Definitions and Specifications
An alternative subjective well-being related outcome is happiness feelings. It differs from the life satisfaction question in the time period evaluated. The life satisfaction measure refers to an evaluation of the current life the individual is leading, while the happiness questions asks respondents to evaluate their life in the last month. More precisely, respondents answered the following question “How you felt over the last month? I felt happy”; the rating scale is from one, never happy, to six, continuously happy. Additionally, the question on life satisfaction involves cognitive appraisals based on aspirations, expectations and values, while the question on happiness is more reliant on the sensory system (Veenhoven, 2000). The results are displayed in Model (1) of Table 2.7 (see Table D.2 in Appendix D, for detailed results). Informal care provision leads to lower happiness levels, reducing it by 0.45 points. Although the overall negative result of caregiving does not change, current happiness score does not rely on its past realisation. It is also noteworthy that the Hansen test p-value is below the standard threshold of 0.1, thus casting doubt on instruments validity.
The care provision is alternatively measured with the intensive margin, referring to the number of hours of care provided within a week. The result displayed in Model (2) of Table 2.7 shows a negative and significant effect of the intensive margin as providing care for one more hour would decrease the life satisfaction level of 0.13 on a scale from zero to ten, confirming the findings of Pinquart and Sörensen (2007).
Previous works on system GMM highlighted the importance of the number of lags as instruments and how the results might be sensitive to it. Initially, instruments from the first lag and further for the lagged dependent variable and from the second lag and further for other endogenous regressors in the difference equation were used. In the level equation, both the lagged outcome and other endogenous regressors were instrumented with first-differences, except the first. There is a trade-off between the remaining serial correlation and the statistical power when using GMM as both are likely to decrease with older realisations while keeping a number of instruments quite low is still necessary. For the purpose of robustness tests, other sets of instruments for the lagged outcome and other endogenous regressors are implemented in the difference equation. The number of instruments is restricted from the first to the third lag of the lagged dependent variable in the difference equation in Model (3) of Table 2.7 in order to focus on the statistical power by keeping the most recent lags. In Model (4) of Table 2.7, we aimed at avoiding any remaining serial correlation by dropping the most recent lags of other endogenous regressors, at the cost of statistical power. Full estimates are given in Appendix D, see Table D.2. Consequently, the p-value of the AR(2) in Model (3) is quite low, slightly above the threshold of 0.1, however, the lagged outcome estimate is higher than the baseline estimate. On the contrary, the p-value of the AR(2) in Model (4) is much more higher than the AR(2) compared with the baseline specification but the lagged outcome turns out insignificant. In these two models, the number of instruments is well under the number of individuals. The negative and significant relationship between caregiving and life satisfaction holds regardless of the set of instruments.
Table of contents :
1 Parental Leave and Life Satisfaction
1.2 Data and Methodology
1.2.2 Econometric Model
1.3 Empirical Results
1.3.1 Baseline Estimates
1.3.2 Robustness Test
1.3.3 Heterogeneity Analysis
1.5.1 Appendix A: Details on our Data
1.5.2 Appendix B: Parameter Estimates Baseline model
1.5.3 Appendix C: Robustness Test
2 Informal Caregivers and Life Satisfaction
2.1.1 Caregiving and Health Outcomes
2.1.2 Caregiving and Life Satisfaction
2.2 Data and Summary Statistics
2.2.2 Descriptive Statistics
2.3 Empirical strategy
2.4 Empirical Results
2.4.1 Baseline Estimates
2.4.2 Heterogeneity Analysis
2.5 Robustness Tests
2.5.1 Sensitivity Analysis: Selection Bias and Propensity Score Matching
2.5.2 Alternative Definitions and Specifications
2.7.1 Appendix A: Merge Procedure
2.7.2 Appendix B: Definitions of Variables
2.7.3 Appendix C: Parameter Estimates Baseline Model
2.7.4 Appendix D: Robustness Test
3 Working hours mismatch and subjective well-being
3.2 Literature Review
3.3 Research Method
3.3.1 Data Description
3.3.2 Empirical Strategy
3.4.1 Working-time Mismatch and Subjective Well-being
3.5 Robustness Test
3.5.1 Heterogeneity Analysis
3.5.2 Sensitivity Analysis
3.7.1 Appendix A: Overview of Studies onWorking Hours Mismatch and Well-being
3.7.2 Appendix B: Merge Procedure
3.7.3 Appendix C: Definition of Variables
3.7.4 Appendix D: Detailed Results