The Impact of Grandparental Death on the Fertility of Adult Children

Abstract

The increasingly central role of vertical family kinship in Western societies underscores the potential value of intergenerational linkages that tie grandparents to the fertility of their adult children. Recent research has examined the changing demography of grandparenthood and the roles fulfilled by living grandparents, but the complex implications of grandparental death—a key feature of intergenerational linkages over the life course—have drawn less attention. In this article, we explore whether and how childbearing of adult women is affected by the death of grandparents—their own parent(s) or their spouse's parent(s). We develop a novel conceptual framework that presents the pathways of influence and considers the overall impact of grandparental death on childbearing of adult children. We then estimate fixed-effects models to identify causal relationships between grandparental death and childbearing, using linked micro-level census and population register data from Israel for the period 1986–2014. We find that grandparental death leads to a reduction of approximately 5 percentage points in the five-year probability of childbirth. The effects of grandparental death are negative across all parities examined and are broadly similar across grandparent's gender and kinship relation. Additional effects are identified, including how the impact of grandparental death varies with time since the previous birth as well as residential proximity prior to death. We explain how our findings regarding the effect of grandparental death offer insight into the role of living grandparents. Our results suggest that policy-makers concerned with low fertility should explore mechanisms that reinforce potential sources of support from grandparents.

Introduction

The increasing salience of vertical family kinship in Western societies (Fingerman et al. 2020) motivates the need for a better understanding of intergenerational linkages between grandparents and the fertility of their adult children. In recent years, an emerging scholarship has highlighted the changing demographics of grandparenthood across periods (Gisser and Ediev 2019; Margolis 2016; Margolis and Wright 2017; Seltzer and Bianchi 2013; Verdery et al. 2020; Wiemers and Bianchi 2015). In parallel, a separate literature has raised interest in the multiple ways that living grandparents can potentially influence family and fertility of their adult children in low-fertility settings (Aassve et al. 2012; Barber and Axinn 1998; Rutigliano 2020; Tanskanen and Rotkirch 2014). Whereas both narratives focus attention on the role of living grandparents, the impact of grandparental death itself—a key feature of intergenerational linkages—has drawn less research attention. In particular, it remains unclear whether and how the death of grandparents may alter the childbearing outcomes of adult children. Existing theoretical and empirical literature provides little clear guidance as to the anticipated impact of grandparental death despite the fact that this is an obvious feature of intergenerational linkages and may have important effects on fertility. In this article, we develop a novel theoretical model and exploit rich administrative data to assess the net impact of grandparental death on adult child fertility while also indicating the most likely pathways of influence. Our findings provide new demographic insights and contribute to a broader understanding of the role of grandparents in fertility behavior.

We pose a fundamental question regarding intergenerational demographic linkages: how does grandparental death impact reproductive outcomes for adult children? One way to answer this question is in terms of the overall net impact of grandparental death on fertility. A second perspective on this question focuses on identifying the primary channels through which grandparental death affects adult child fertility. Some of the channels may increase fertility, whereas others may depress the fertility of their adult children. For example, grandparental death could have negative impacts on childbearing through the loss of grandparental childcare services or in vivo transfers of financial resources to adult children. Alternatively, grandparental death may positively impact fertility of their adult children by relieving them of the burden of care for dependent grandparents (OECD 2019) or through the potentially substantial impact of bequests on household wealth (Hansen and Wiborg 2019; Korom 2018). Given the variety of possible influences, which often have opposing effects, the theoretical literature provides little clear guidance as to the anticipated net impact of grandparental death.

We detail a framework for thinking about the complex pathways through which grandparental death may influence adult child fertility, and we develop a number of research hypotheses stemming from our review of the literature. We then explore this question empirically using a unique set of data files produced by the Central Bureau of Statistics (CBS) of Israel that connect all individuals from the 2015 population registry of Israel with information from population censuses of 1972, 1983, 1995, and 2008. The data enable us to reconstruct reproductive histories and to use information about the occurrence and timing of grandparental death. The richness of our data allows us to utilize individual-level fixed-effects models that provide estimates of the causal impact of grandparental death on the fertility of their adult children. The combination of our unique data and empirical strategy allows us to shed light on a number of issues raised in prior literature (e.g., Dahlberg 2018). For example, we examine whether grandparental death effects vary by parity (Snopkowski and Sear 2016) or grandparental gender (Schaffnit and Sear 2014; Tanskanen et al. 2014; Waynforth 2011). In addition, we consider whether the impact of grandparental death on adult child fertility differs by residential proximity to the grandparent before death, and by age of the youngest child in the household. These analyses provide further insight regarding some of the most likely mechanisms that underlie our basic finding about the impact of grandparental death on adult child fertility.

Conceptual Framework

In this section, we offer a conceptual framework for thinking about the pathways through which the death of grandparents may affect adult child fertility.

Table 1 lists factors following grandparental death that will reduce fertility (left column) and those that are likely to increase fertility (right column). Factors that reduce fertility are those associated with the loss of supports that propped up fertility while grandparents were alive. These supports include childcare services as well as financial assistance that adult children receive from their parents (Albertini et al. 2007), which may increase fertility by making it easier to manage economic and time constraints (Barber and Axinn 1998; Harknett et al. 2014; Rutigliano 2020). The supportive roles of grandparents are made explicit in the cooperative breeding model, where the burden of childrearing is shifted to extended kin, including grandparents (Schaffnit and Sear 2014; Sear 2017; Snopkowski and Sear 2016). Although market-based or publicly supported childcare enables families to outsource part of their childcare (Aassve et al. 2012; Rindfuss et al. 2007), unpaid grandparental childcare may contribute to higher fertility by serving as a less costly and more trusted substitute (Raz-Yurovich 2014). In addition, grandparental childcare may also complement formal, paid childcare during days and hours when such settings do not operate (Hank and Buber 2009; Posadas and Vidal-Fernandez 2013).

Another source of support lost when grandparents die is in vivo financial transfers from grandparents to their adult children, which may encourage (further) bearing of grandchildren (Cox and Stark 2005) by easing financial strain caused by expenses associated with children. Despite the recognition of the importance of financial transfers from parents in establishing material well-being among young adults (Albertini et al. 2007; Cooney and Uhlenberg 1992; Semyonov and Lewin-Epstein 2001; Spilerman 2004), evidence on the relationship between such transfers and fertility is limited (but see Schaffnit and Sear 2017a, 2017b). The type of support that grandparents may be able to provide—whether through time support or financial support—may depend in part on the wealth and work force status of grandparents (Posadas and Vidal-Fernandez 2013).

The direct effect of grandparental death may also operate through its impact on various aspects of the physical and mental health and well-being of adult children. For example, grieving after parental death may lead to depression and other emotional and mental health issues, which may directly impact fertility intentions and actual fertility (Dahlberg 2018). We use the term “grieving” for brevity, but our reference is to a broad range of consequences stemming from parental death, such as loss of emotional support, social isolation of bereaved adult children, effects on interpersonal relationships with their spouses, and changes in values or life goals (Dahlberg 2018). Some empirical evidence suggests that parents' emotional support is associated with higher fertility among adult children (Schaffnit and Sear 2017b), but the ultimate impact on adult children of the emotional toll associated with parental death is ambiguous. Although some adults demonstrate declines in various dimensions of functioning following parental death, others show little decline or even improvements in functioning (Umberson and Chen 1994). Even if there is a decline in functioning, most distress is apparent only in the first few months after grandparental death (Umberson and Chen 1994), which may lead to very temporary impacts on childbearing.

Alongside the fertility-reducing effects, various other factors associated with grandparental death may operate to increase fertility. One positive effect may stem from inheritances, which are received after grandparental death. Bequests would presumably affect fertility via economic status, but empirical research on the relationship between bequests and fertility in Western contexts is sparse. Although inheritances have been shown to play a substantial role in driving socioeconomic differences among the younger generation (Elinder et al. 2018; Piketty 2014; Wagner et al. 2020), research indicates insignificant or negative relationships between household wealth and income and fertility in developed societies (e.g., Jones et al. 2010; Stulp and Barrett 2016).

Finally, Table 1 suggests that removal of burden of care for grandparents may also positively impact fertility. The burden of care may be nontrivial, particularly in cases where longevity translates into extended periods of dependency (McGrath et al. 2019). Research has shown that adult women may experience declining labor force participation in response to the need to care for elderly parents (Ettner 1995; Løken et al. 2017). From this perspective, the death of grandparents, although a potentially emotional life course episode, may reduce certain burdens and raise childbearing. Furthermore, adult children may anticipate the future financial and care needs of their possibly dependent parents by curtailing or delaying childbearing (Harknett et al. 2014). Thus, those adult children who do experience parental death during their childbearing years may respond by increasing fertility now that they no longer face the prospect of long-term care for dependent parents (Grundy and Henretta 2006; Harknett et al. 2014).

As the preceding discussion makes clear, some pathways focus on the loss of roles fulfilled or services required by living grandparents, whereas others are factors directly associated with the death of grandparents. It is the net combination of these factors that ultimately determines whether grandparental death increases or lowers fertility of their adult children. Thus, we have no a priori theoretical prediction about the net effect of grandparental death. The conceptual framework presented here offers a more encompassing approach than the previous literature given that it considers that multiple factors may have opposing influences.

Research Context

In this section, we provide an overview of relevant demographic and sociological factors in Israeli society, and we explain why Israel is well-suited for an examination of the impact of grandparental mortality on the fertility of adult children.

Israel is a heterogeneous society with a distinct and largely disadvantaged Arab minority, accounting for approximately 20% of the population over time. The complex and diverse Arab minority has experienced very different demographic trends, labor market patterns, and extended family relationships as well as historically constrained mobility and segregated residential patterns (Al-Haj 1986; Lewin-Epstein and Semyonov 1993; Okun and Friedlander 2005; Saabneh 2015; Yiftachel 1996). The research context discussed here is relevant primarily to the majority non-Arab (primarily Jewish) population, which is the focus of our study; a separate study on the Arab population would be a valued contribution.

In terms of national statistics, Israel has been undergoing many of the same social and economic processes occurring in other developed societies and is ranked by the United Nations Human Development Index as among the “very high human development countries” in terms of income, education, and health (United Nations 2018). Its income (GDP per capita of approximately $33,132 in 2017) puts it just below Spain and just above the Czech Republic (United Nations 2018). In terms of education, Israel is ranked near the top among OECD countries in terms of proportions of persons aged 25–64 with tertiary-level education (46% among persons aged 25–64 in 2011) (OECD 2013).

In order to understand how grandparental mortality may affect fertility, it is important to understand the role of public and private supports for fertility in Israel. Israel is classified as having a conservative welfare regime in which individuals rely largely on their families for their welfare (Esping-Andersen 1999; Stier et al. 2001). In general, Jews in Israel are family-oriented, with a strong emphasis on marriage, marital fertility, and high overall fertility (Lavee and Katz 2003; Toren 2003). Societal norms for high fertility among Jews stand alongside strong expectations that women will work in paid labor and contribute to household income (Kraus 2002).

Consistent with its Human Development Index ranking, Israel is a low-mortality setting, with life expectancy topping age 84 among women and 80 among men in recent years (Israel Central Bureau of Statistics [ICBS] 2019). Mortality has improved in recent decades, with overall life expectancy in 2013 ranked among the top 10 countries worldwide (Weinreb 2016). As in most other developed societies, most of the improvement in life expectancy in recent decades has been at older ages (Friedlander et al. 1995; Manor et al. 2000).

Fertility levels among Jews in Israel are very high, with the modal completed family size at three, parity progression from two to three children at 67% (Okun 2013), and total fertility rate (TFR) levels above 3 in recent years (ICBS 2019). Despite higher overall childbearing levels than in other rich societies, mean age at first birth among Jews in Israel (around 25 for cohorts born in the mid-1960s) lies roughly in the middle of the range for European societies and has been increasing among more recent cohorts (Okun 2013). These shifts in childbearing are important because, other things being equal, they raise the odds that adult children enter childbearing around the time that their own parents may be at increasing risk of death. In contrast, improving mortality among older Jewish Israelis might counterbalance this effect of increasing age at first birth. These aggregate indicators conceal variations both within and between key Jewish subpopulations in Israel. For example, very high fertility levels have been documented among Ultra-Orthodox Jews. In contrast, some of the lowest levels of fertility are experienced among the large group of Jews who identify themselves as secular, or nonreligious (Okun 2013).

Smooha (2005) emphasized the importance of Israel's familial support system, citing the continuing intensity of relationships between parents and adult children in Israel, with frequent contact being extremely common. Parents provide significant financial and caregiving aid to their adult children (Fogiel-Bijaoui and Rutlinger-Reiner 2013). For example, young adults in Israel are financially dependent on their parents for substantial assistance in purchasing of a home or other transfers of wealth (Semyonov and Lewin-Epstein 2001; Spilerman 2004). The ability of grandparents to support their adult children financially may change with the death of a grandparent, and research suggests that widowhood has larger negative financial impacts for surviving grandmothers than for surviving grandfathers (Ahn 2004; Lewin and Stier 2003).

Israel stands out in terms of the extent of grandparental assistance in childcare (Okun 2016; Sharel and Yakir 2016; Spilerman 2004). In fact, Israel Social Survey data from 2009 show that among all women in Israel aged 25–39 who have at least one child, 71% report receiving grandparental assistance in childcare (Okun 2016). In European contexts, data from the Generations and Gender Programme indicate that analogous proportions are lower and generally range from 30% to 60%, with the exception of the Netherlands (Okun 2016; Thomese and Liefbroer 2013). Interestingly, Israel and the Netherlands are both geographically small and densely populated countries. Furthermore, in Israel, the Jewish population is geographically concentrated in the center of the country, lending itself to grandparental caregiving and close intergenerational ties given that adult children tend to live near their parents. In addition, during most of the period under consideration, costs of childcare for the majority of preschool children in Israel were high; thus, grandparental assistance may be particularly relevant to the context at hand (Hasson and Dagan Buzaglio 2019).

Both net financial and time assistance from grandparents appear to remain strongly positive when grandparents are in their 50s, 60s, and 70s, turning negative only in their 80s (Litwin et al 2008). Moreover, calculations based on data from the 2009 Israel Social Survey show that adult children who receive financial assistance or assistance with childcare from their parents have higher ideal family sizes (Okun 2016; authors' calculations).

Bequests also play a potentially important role if the flow of resources from deceased parents alters the fertility behavior of adult children. Evidence from Israel suggests that inheritances play an impressive role in building household wealth. One study on Israelis over age 50 noted that about 28% of household wealth originates from inheritances (Semyonov and Lewin-Epstein 2011). However, evidence points to a weak relationship between economic status and fertility in Israel (Cohen et al. 2013).

The system of elderly care in Israel relies heavily on private home caregivers, with about 80% of elderly receiving support from the state; however, there is also a strong emphasis on not entirely “replacing the family” (Asiskovitch 2013). Because adult children remain involved, the burden of care may in some cases be substantial, particularly among more vulnerable groups, such as immigrants (Remennick 2012). However, given evidence that resources and support appear to flow primarily from the elderly to their adult children (Litwin et al. 2008), relief from burden of care for dependent parents is not likely a dominant feature of net intergenerational relations.

As mentioned earlier, the potential impact of grandparental death on the fertility of adult children may be related to grieving. Grandparental death may affect the physical and mental health and well-being of the adult children (Umberson and Chen 1994), as well as modify values, life goals, and the importance of social relationships (Scharlach and Fredriksen 1993). In the Israeli context, although considerable research has focused on the study of bereavement associated with military losses and terrorism (Rubin 1992; Rubin et al. 1999), the mental and emotional impact on adult children of the loss of parents has received little research attention. The overall impression based on research conducted in other contexts is that adult children's grieving and mourning for the loss of elderly parents may be strong but probably does not play a dominant role in driving adult child reproductive behaviors.

Research Hypotheses

Our reading of the literature both in the Israeli context and more broadly as reviewed earlier in this paper leads us to focus on the loss of childcare and the loss of in vivo transfers as the most likely driving forces behind changes in fertility following grandparental death (see Table 1). The literature suggests that an adult child's loss of a parent, while leading to distress, may have both positive and negative effects on childbearing (Schaffnit and Sear 2017b; Umberson and Chen 1994). Furthermore, the literature suggests that the relief from the (potential) burden of care for an elderly grandparent is unlikely on its own to impact adult child fertility (Harknett et al. 2014). Similarly, although inheritances may be substantial in many cases, they may have already been incorporated into adult reproductive planning prior to the death of the elderly parents. In summary, the loss of childcare and in vivo transfers are most likely to be the dominant influences. Consequently, we hypothesize that the death of a grandparent will be associated with a decline in fertility because we expect that the negative impacts of grandparental death dominate positive ones.

We also explore whether the dominant negative effect would likely be similar across grandparents. As mentioned earlier, surviving widows' ability to continue supporting adult children financially would likely be reduced more than that of surviving widowers. Thus, the loss of in vivo financial transfers may matter more in the case of the death of a grandfather. In contrast, given that grandmothers tend to provide more childcare than grandfathers (Litwin et al. 2008; Zagheni and Zannella 2013), the loss of childcare may matter more in the case of the death of a grandmother. On balance, we predict a negative effect both for grandmaternal as well as grandpaternal death. Although previous research has suggested that in some cultural contexts, wives' parents may be more involved in childcare than husbands' parents (Thomese and Liefbroer 2013), there is no clear reason to believe this asymmetry also holds for in vivo financial transfers, which may serve as partial substitutes for childcare time. Thus, we do not expect that the net effects of grandparental death will differ sharply between husbands' and wives' parents.

In order to explore which mechanisms underlie the hypothesized decline in fertility following grandparental death, we consider how the impact of death varies for families under different conditions. We expect that grandparental childcare services matter more when the grandchildren are younger. Accordingly, we hypothesize that the loss of a grandparent would affect future fertility more negatively within the first few years following the birth of the youngest child. In contrast, we have no clear reason to predict that the fertility effect of grandparental death via grieving of adult children or in vivo financial transfers would strongly depend on the age of the children at home.

In addition, we exploit variation in the residential proximity of adult children and their parents to test whether the effects of grandparental death are stronger when the residential distance is smaller. The logic underlying this test is that provision of childcare services is inherently dependent on residential proximity.

Data and Methodology

We build on data collected and managed by the Central Bureau of Statistics of Israel (CBS). For our analysis, several specific data files were prepared and made available by the CBS. One key feature is individual-level identification numbers, which allow for the linkage of individuals from the 2015 population registry of Israel with variables obtained from the four census years. Population registration data were linked to 100% samples of the population censuses of 1972, 1983, and 1995. Because of changes in census methodology in the 2008 census, the linkage for that year was limited to a roughly 17% sample for which information on residence was collected.

Several major advantages are inherent in the use of these combined registry-census data files for examining Israeli society: (1) coverage is nearly universal for the population we examine and includes nearly three decades, from 1986 to 2014; (2) matching through the population register allows us to link individuals to their parents, spouses, spouses' parents, and children; (3) the extended time frame and the quality of registration data allow for linkage across three generations, including reconstruction of reproductive histories; (4) accurate information on living status and dates of death of parents of adult children, as well as parents of adult children's spouses, allows for examination of periods before and after grandparental death; and (5) repeated rounds of census data on residential location allow us to assess geographic proximity to kin.

We focus on woman-years of exposure contributed by parous Jewish women of reproductive age during the years 1986 through 2014.¹ As noted earlier, the non-Jewish, primarily Arab population of Israel, which over the period accounts for approximately 19% of the population of women of reproductive age in Israel, has distinct residential and family patterns and is not included in this analysis. Moreover, our sample includes only cases in which both the woman and her spouse were born in Israel given that the foreign-born are much more likely to have parents who did not live in Israel, thus making it impossible to determine their living status in the population registry. At roughly the midpoint of our observation period (2000–2001), the large majority of Jewish women (72%) of reproductive age were native-born (ICBS 2002). We restrict our analysis to women whose children are all linked to the same father in order to reduce the effect of union dissolution on fertility patterns. Given that marital dissolution is fairly low among Jews in Israel, relative to other developed societies, with roughly 8% to 14% of couples divorcing within 10 years of marriage among the relevant marriage cohorts (ICBS 2009, 2019), this is not a severe restriction.² It was also necessary to exclude same-sex partners because of the nature of our administrative data, but this has little impact on our sample size. Our final samples, which vary by parity and by whether we include controls for geographic proximity, range from 238,000 to 3.1 million depending on the models (see upcoming Table 4 for sample sizes).

Mortality of grandparents is our key explanatory variable. Available data sources provide information on year of birth only for persons who ever lived in Israel; information on year of death is available only for persons who died in Israel.³ Information on the dates of death in the registration file allows us to determine accurately whether a grandparental death occurred before or after the year of birth of a child.

Our units of analysis are parity-specific couple-years of exposure, with each couple potentially providing multiple years of exposure at various parities. The dependent variable is a binary indicator for whether a woman experiences a birth of parity j in a given year, among women whose parity in the previous year is j – 1. Each woman at parity j – 1 contributes exposure to the analysis of progression to parity j until birth j occurs, or until the she attains age 45. We explore the impact of grandparental death at both lower and higher parities with separate analyses of parity progression from 1 to 2, 2 to 3, 3 to 4, and 4 to 5 births.

Our primary explanatory variables are time-varying dummy variables for each of the four grandparents that indicate whether each grandparent is dead. An alternative specification includes an aggregated time-varying dummy variable that indicates whether at least one of the grandparents is dead. In order to eliminate variation caused by deaths of grandparents in the distant past, we condition our sample on having all four grandparents alive at the beginning of exposure to each parity progression.⁴ Thus, the death variable represents a change in living status of grandparents who were alive at the birth of the previous child.

Our statistical models incorporate additional explanatory variables that interact with grandparental mortality. These include a time-varying dummy variable to indicate whether the youngest child is at least age 6. This variable allows us to test whether grandparental death has greater impact on fertility when parents are caring for small children. The second variable is an indicator of whether grandparents and adult children reside in the same locality. This information is derived from census questions on current residence and allows us to evaluate the importance of close or distant residency as a moderating factor for the impact of grandparental death on childbearing patterns.

Fertility and mortality are age-dependent, so we control for age in all models using time-varying dummy variables for the age of each woman and her spouse (each in five-year age groups). We also control for the time-varying age of the living grandparents using dummy variables (in five-year age groups). Thus, the effect of grandparental death is relative to having a living grandparent in the reference age group (60–64 years old). We also include single-year dummy variables for calendar year (to capture period effects) as well as single-year dummy variables to capture duration since birth (j – 1).⁵

Our principle modeling strategy is linear fixed effects (FE), with repeated observations for women over time. Like discrete-time event-history models, the linear FE models use panel data to analyze the parity-specific probability of childbearing over the years of exposure for each woman. In addition, the FE framework—unlike a more standard event-history model—controls for unobserved time-invariant heterogeneity. This is important because failure to control for unobserved heterogeneity in factors such as family health, which may be correlated with both fertility and longevity of grandparents, may lead to bias in estimated coefficients on grandparental death (Schaffnit and Sear 2014). In contrast to other nonlinear FE models that also control for time-invariant unobservable characteristics, the linear FE models are computationally straightforward.⁶ In addition, the linear FE models offer a simple interpretation of the consequences of grandparental death in terms of percentage point change in the annual probability of childbearing in periods subsequent to grandparental death.

One concern with panel data is the possibility of serial correlation for individual women across time. This type of within-panel serial correlation may lead to biased estimates of standard errors. In models not presented here, we examined the sensitivity of our substantive findings to the estimation of robust standard errors (Wooldridge 2013). We found that the statistical significance of the estimated effects in the models with the robust standard errors were not reduced, leaving our substantive conclusions unchanged (results available upon request).

We first estimate models of parity progression from the second to the third birth. This is arguably the most sociologically interesting transition in the Israeli context, where modal completed family size is three. Initially, we present models that include four time-varying dummy variables indicating that a grandparent is dead during the exposure year —one for each grandparent—enabling us to consider differences in the impact of death by type of grandparent. Later, for ease of interpretation, the grandparental death variables are aggregated into a single dummy variable, indicating whether one or more grandparents are dead during the exposure year.

Following our in-depth analyses of parity-progression to the third birth, we consider progressions to the second, fourth, and fifth birth separately. Later models combine exposure to all parity progressions, controlling for current parity. By combining exposure at all parities, we can more easily explore interactions between grandparental death and (1) the age of the youngest child and (2) the proximity of residence.

Summary statistics are shown in Table 2. These computations are made at the level of the woman and are based on the exposure she contributes at all parities combined.⁷ Women are potentially followed over the entire period of 29 years, but average exposure per woman is reduced to 8.96 years because of the timing of entrance into and out of childbearing ages, censoring, and dropping out of subsequent parity progressions following grandparental death. As indicated in column 1, about 3% to 11% of women experienced any particular grandparental death during the exposure years included in our sample. A woman's mother's death is least likely to be experienced, and a spouse's father's death is most likely. Overall, about 23% of women experience any grandparental deaths during exposure. Column 2 presents the average proportion of exposure per woman in which one or more of the four grandparents is dead, separately for each grandparent and also for any of the grandparents. On average, women experience about 1% to 4% of their exposure with any specific grandparent being dead, and 8% with any of the grandparents being dead.

Results

We present the impact of grandparental death on annual probabilities of parity progression from the second to the third birth in Table 3.⁸ The first column shows results from FE models in which the living status of each of the four grandparents is modeled separately.⁹ The second column of Table 3 presents estimated effects from FE models in which we aggregate the living status variables across the grandparents into a single binary indicator.

Model 1 indicates that the death of each grandparent reduces the annual probability of having the third birth, with reductions ranging from roughly 1 to 2 percentage points per year. The pattern of effects is broadly similar across grandparents, suggesting neutrality by gender and kinship, and we estimate Model 2 with the aggregated indicator representing the death of any of the four grandparents. R-squared fit measures are identical across models, indicating no loss of explanatory power resulting from the aggregation of the individual grandparental coefficients by gender and kinship relation. The estimated effect on this aggregated binary variable in the second model shows that the annual birth probability is about 1.8 percentage points lower if at least one grandparent died since the previous birth than if all grandparents were alive in the reference age category.

Next we explore how the impact of grandparental death on parity progression probabilities varies across the initial parity at the start of the interval (initial parities vary from 1 to 4) (Figure 1). Grandparental death is aggregated into a single binary variable, and our results are presented graphically (see Table 4 for main coefficients and standard errors).¹⁰ The estimated effects of grandparental death can be interpreted as the (negative) contribution of grandparental death to the parity-specific annual probability of parity progression (as discussed earlier for the results in Table 3). An alternative way to interpret the impact of grandparental death is in terms of the cumulative effects of one or more deaths over the course of five years of exposure to parity progression. These cumulative effects are functions of the estimated effects of grandparental death on annual probabilities of parity progression, based on the main results of Table 4. In addition, effects on cumulative probabilities depend on differences in the level of annual birth probabilities across parities (see the text in the online appendix for further details on calculation of the cumulative effects).¹¹

The negative impact of grandparental death is clear and highly significant for each parity progression. The effects on annual probabilities of parity progression (the gray bars) vary from about 1 percentage point to 3 percentage points, with no obvious monotonic pattern across parities. When these estimates are accumulated over five years (the black bars), the pattern of effects is seen in a slightly different light because they are affected by parity-specific differences in progression probabilities. Overall, the cumulative negative effect varies from about 3 percentage points for those starting at parity 1 to nearly 9 percentage points for those beginning at parity 4. In the pooled model (which includes all parity progressions), the annual probability of parity progression is reduced by about 2 percentage points, and the cumulative probability over five years is reduced by about 4.5 percentage points. In summary, across each parity, grandparental death reduces the likelihood of parity progression, with substantial cumulative effects over several years.

These findings highlight the overall net effect of grandparental death. We extend these findings by introducing two interactions. The first, testing whether the impact of grandparental death varies by the age of the youngest child, involves the inclusion of an interaction variable between grandparental death and a time-varying indicator for whether the youngest child is at least 6 years old. Figure 2 shows the different effects of hypothetical grandparental deaths when the youngest child is under 6, and when the youngest child is 6 or older, both in comparison with the alternative scenario of no grandparental death (main results shown in Table 5).¹²

The effect of grandparental death when the youngest child is under 6 (annual probability of parity progression is reduced by 2.7 percentage points) is roughly twice as large as the effect when the youngest child is 6 or older: annual probabilities of parity progression are reduced by 2.7 percentage points and 1.4 percentage points, respectively, with the relevant interaction term highly significant (p < .001). The five-year cumulative effect implies that grandparental death reduces the cumulative probability of birth by just over 5 percentage points when the youngest child is under 6 but by just over 1 percentage point when the youngest child is 6 or older.¹³ Thus, Figure 2 shows that both the cumulative and the annual impact of grandparental death are larger when the youngest child is under 6.

Using information on the residential locations of adult children and their parents, we explore how residential proximity alters the impact of grandparental mortality on birth progression. Figure 3, which is based on estimated coefficients presented in Table 6, shows that grandparental death reduces the probability of a birth more when grandparents and adult children lived in the same locality prior to the death.¹⁴ When grandparents lived in a different locality, the annual probability of birth declines following grandparental death by about 1.5%. In contrast, when the death occurs to a grandparent who lived in the same locality, the effect on parity progression is stronger, at about 2.3%. The cumulative five-year probabilities are also reduced more if the grandparent lived in the same locality prior to death, with a 4% decline for those living in different localities and a 7% decline for those living in the same locality.

Discussion

Our main results document an impressive and robust negative impact of grandparental death on parity progression of adult children. Our findings are consistent across each parity progression and with all parity progressions pooled together. This negative finding supports our first hypothesis that the negative impact of grandparental deaths on adult child fertility will outweigh the positive impact.

As indicated earlier, negative effects include the loss of grandparental childcare, loss of financial in vivo transfers, and grieving of adult children. The negative effects dominate the positive influences mentioned earlier, including inheritance and relief from burden of care for a dependent grandparent. Because our analysis builds on a fixed-effects analysis to exploit individual-level life-history data from administrative sources, the models provide causal evidence of a negative and substantial net impact of grandparental death on subsequent adult child fertility. Thus, our major finding fills a gap in the literature regarding the demographic implications of the changing contours of grandparenthood and vertical family relationships. The main finding points to the salience of grandparental roles as well as the possible impact of increasing grandparental longevity on childbearing patterns.

Our analyses do not allow us to distinguish directly between the various mechanisms that drive the negative effects of grandparental death on fertility. Nevertheless, our empirical results, together with our reading of the literature, lead us to put less weight on the grieving argument and more weight on the importance of the loss of caregiving and financial transfers that grandparents provide while they are alive. We find no evidence that following parental loss, adult children's grief-related responses are necessarily related to reduced or postponed childbearing; moreover, the impact of the death on fertility would not necessarily remain negative over a number of years (Pepper and Nettle 2013). In fact, the limited evidence that has been presented suggests that grieving the loss of a grandparent may be associated with accelerated childbearing (Dahlberg 2018). In contrast, the explanations for the negative effect of grandparental death on fertility focus on the loss of both caregiving and financial transfers that grandparents provide while they are alive. This is consistent with recent studies highlighting grandparents' large investments in childcare and in vivo financial transfers (Aassve et al. 2012; Aparicio-Fenoll and Vidal-Fernandez 2015; Hank and Buber 2009; Rutigliano 2020; Thomese and Liefbroer 2013).

In contrast with expectations from the literature in evolutionary sociology, we also hypothesized that all grandparental deaths—irrespective of kinship relation or grandparental gender—produce similar, negative impacts on fertility. In fact, our analyses do indicate that the effects of grandparental death on adult child fertility are broadly similar across grandparents. Given that grandmothers tend to provide more childcare than grandfathers (Litwin et al. 2008; Zagheni and Zannella 2013), our finding of gender neutrality suggests that grandparental childcare may not explain the whole story. Rather, the finding may be indicative of the additional importance of other factors, such as financial transfers from grandparents.

The gender and kin neutrality might also be understood in terms of extended family support networks. From this perspective, in-laws are linked through their joint interest in the well-being of their grandchildren. Our results are consistent with a model in which grandparents coordinate to help their adult children, although our data do not allow us to test such a model directly. Changes in grandparents' living status ripple through the social network and may compel surviving grandparents to take up slack when death strikes other grandparents in their network. In this sense, this social network provides another example of how family connections—even those involving in-laws—may matter (Ben-Porath 1980). Essentially, the four grandparents may cooperate to provide a form of nonmarket insurance for their adult children and shared grandchildren. The support provided by grandparents is easily observed and monitored by all involved—the type of conditions that make cooperation highly efficient (Arnott and Stiglitz 1991). These types of support systems are discussed in models of cooperative breeding, in which extended family members coordinate to support raising children (Sear 2017). These models of coordination may be particularly apt in contexts such as Israel, where marriage remains a strong institution that forms the basis for in-law cooperation (Okun 2013; Raz-Yurovich 2010).

The introduction of interaction models allows us to better understand the mechanisms underlying the negative effects of grandparental death. One hypothesis, which is supported by our findings, argues that grandparental death should have a larger negative impact on further childbearing when children in the household are young (under 6). Although we cannot rule out the importance of the loss of in vivo transfers and the possible role of grieving following grandparental death, the most compelling interpretation of the interaction model is in terms of the loss of grandparental childcare. Young grandchildren require time-intensive care, and grandparents may been seen as a preferable and more trustworthy option relative to formal and expensive childcare in the context of Israel.

Our second interaction tests a hypothesis that the death of grandparents living in the same locality causes a larger decline in parity progression relative to those living in a different place. This finding is also most directly consistent with impact of a loss of childcare as a primary underlying mechanism, strengthening the case that grandparental childcare assistance is a key—although not sole—factor in the childbearing of their adult children. Of course, residential proximity may also be associated with the greater loss of financial support and the intensity of grieving; however, the loss of childcare is more directly tied to geographical proximity (Pink 2018). Our support for this hypothesis is notable given Israel's small size and geographic concentration of its Jewish population. It is possible that the importance of living in the same locality would be even more important in larger and less densely populated Western contexts.

Concluding Remarks

Our research contributes both theoretically and empirically to a better understanding of the relationship between grandparental death and adult child fertility. First, our conceptual contribution is to draw attention to the perspective of the adult child and to how the changing demography of grandparenthood affects the life course of adult children in the childbearing ages. This perspective complements previous demographic research that has highlighted changing grandparenthood in terms of its effects on older adults and their grandchildren (Margolis 2016). Prior research that has considered the implications of the death of a grandparent on adult children has tended to focus more on psychological impacts and physical health (Umberson and Chen 1994). Only recently has greater attention turned to grandparental death as a potentially important event affecting adult child fertility (Dahlberg 2018; Sear 2017). We build on these earlier contributions to develop a more comprehensive, conceptual framework that describes the various mechanisms by which the loss of a grandparent may impact adult child fertility while specifying the mechanisms that potentially increase or decrease childbearing.

Our second primary contributions are methodological and empirical. Previous literature has considered living grandparents and the roles they may play in the fertility of their adult children—for example, through grandparents' direct contributions to childcare. Although some of this literature suggests that grandparental support may increase adult child fertility, these studies may have been hampered by endogeneity concerns (Sear 2017). Whereas grandparental assistance may reduce childcare costs, improve work-family balance, reduce stress for parents, and perhaps also be a source of financial support, grandparental care and assistance may also be a response to the child-related financial, time, and emotional strain that parents experience. Endogeneity may partly explain why, despite some evidence supporting the case that grandparental assistance may play a role in the fertility of their adult children (e.g., Aassve et al. 2012; Sear and Coall 2011), not all evidence has pointed in the same direction (e.g., Kim 2017; Schaffnit and Sear 2017a, 2017b). Our approach shifts the focus to what happens when there is a death of a grandparent—an event that is largely exogenous to the childbearing patterns of adult children. Moreover, our data allow us to measure the individual-level impact of grandparental death on adult child fertility, net of fixed, unobserved factors that may affect both family health and reproduction. We are able to estimate the net causal effect on adult child fertility across multiple parities and separately by gender and kinship relationships of the grandparents. Our empirical results are suggestive of the key mechanisms driving the net decline in fertility: the contributions of living grandparents through childcare services and in vivo transfers. Thus, our results offer new support for arguments suggested in earlier studies (e.g., Aassve et al. 2012; Barber and Axinn 1998; Rutigliano 2020; Tanskanen and Rotkirch 2014).

Promoting higher fertility continues to be a challenging goal in many below-replacement countries. Solutions may be found in markets and public sector initiatives that reduce the burden of childrearing for parents and families. However, part of the answer in modern states may yet be found within families themselves (Sear 2017). Our study is focused on Israel, but other states that share little of the obvious demographic or cultural markers of Israel may learn from our findings. Facilitating the role of grandparents would be an obvious key step. For example, one recently proposed change in Israel would be to allow working grandparents to use paid sick leave to care for grandchildren, thereby freeing up their adult children and reducing parental costs of children (a 2017 proposed amendment to Knesset law on use of sick leave). This policy could have significant impact: data from the Israel Social Survey of 2009 shows that among grandparents aged 50–84, most (58%) grandparents are involved in caring for their grandchildren, even while many of them (44%) are still active in the labor force. Notably, labor force participation is higher among grandparents who provide childcare (52%) than among those that do not (32%).¹⁵

More generally, work-family policy—based on broader arguments to redistribute labor across generations (Vaupel and Loichinger 2006)—might build on possible substitutability in childcare provision across generations. Such policy could include reduced or flexible work hours for grandparents and grandparental leave following the birth of grandchildren. Other policies could be directed toward grandparents outside the labor force and might recognize that parents may not view paid childcare as a perfect substitute for parental or grandparental childcare because of issues of trust and accessibility (Raz-Yurovich 2014). Thus, grandparents who do not work but care for their grandchildren could be compensated by policies similar to those directed toward mothers who are not employed (Hakim 2003).

Another step might be to think more carefully about the geographic proximity of extended families. In may be that in many low-fertility countries that are struggling to increase levels of childbearing, practices such as suburbanization have contributed to distancing adult children from their parents, reducing the potential for grandparental involvement. One positive step might be city planning in new communities that ensures the availability of more diverse housing, thereby allowing families to remain proximate and supportive over time. Addressing issues of residential proximity between adult children and their parents might also be important for societies that are still experiencing substantial population growth but that are on a trajectory to near or below replacement level.

Even in societies that are not generally characterized by the “strong” families (Reher 1998) apparent in Israeli society, the potential importance of grandparents may be meaningful in understanding the behavior of many population subgroups. This might be the case, for example, among newly arrived immigrant groups, ethnic minorities, or less geographically mobile parts of the population. In these and other contexts, more attention to the role of intergenerational demographic ties should not be ignored as an important factor in understanding fertility change.

Acknowledgments

The authors contributed equally to all stages of this research. They benefitted from helpful comments from Seth Sanders and from feedback on earlier versions of this research from presentations given at AlpPop, European Population Conference, Israeli Sociological Society, the University of Maryland, and Duke University. They also acknowledge excellent research assistance from Dror Shvadron and assistance in the ICBS research room from David Gordon. Indirect support for construction of these data was made possible by the Israel Science Foundation (Grant No. 1509/14).

Notes

References