Keywords

1 Introduction

This research examines the effects on divorce of the number of children, as well as having young children, in the pronatalist society of Israel. By analyzing the relation between children and divorce in this context, we can address two theoretical questions that have remained open in the literature. First, does the number of children have a consistent linear effect on divorce? Second, to what extent does the relation between children and divorce depend on the couple’s ethnic and socioeconomic group?

Israel has the highest fertility rate among OECD countries today: an average of 3.08 children per woman compared to 1.68 in the OECD (2017). While fertility rates across Europe, though varied, are generally low, and in some countries have been below population replacement level for more than 20 years (Esping-Andersen and Billari 2015; Eurostat 2017), the fertility rate in Israel has been high and rather stable for the past two decades (ICBS 2018). Indeed, the rate in Israel during the period under study (2003–2015) was high even compared to the time when most OECD countries had high fertility rates (Billari and Kohler 2004; OECD 2018). The large proportion of Israeli families with many children enables us to look for a change in the direction of the relation between children and divorce after the number of children exceeds a certain threshold. On the one hand, the more children there are, the more the couple has already passed the selection stage of one child (Jalovaara 2013), suggesting that marriages with many children are more stable. On the other hand, a large number of children can place considerable pressure on parents, suggesting a destabilizing effect. This issue is especially acute as Israeli family policies are neither particularly supportive nor generous (Gal 2010; Herbst 2013; Herbst and Kaplan 2016; Herbst-Debby 2019).

Notwithstanding the high aggregate level of fertility in Israel, there is diversity in regard to this factor across ethnic and religious groups (Okun 2013). Among Israeli-Palestinian women, fertility has declined sharply in the last few decades. For Muslims – the largest group among the Palestinians – the rate dropped from 9.2 children per woman in the 1960s to 3.4 in 2017 (ICBS 2018). In comparison, the average fertility rate of Jewish women declined from 3.4 in the 1960s to 2.6 in 1990s and then returned to 3.2 in 2017 (ICBS 2018; Levy 2016). This variation in fertility and family patterns between the two ethnic groups – which has also been found with regard to migration in other countries (Furtado et al. 2013; Hewitt 2008) – intersects with socioeconomic inequality, as Jews are economically better off than Palestinians (Swirski et al. 2017). Moreover, earnings inequality in Israel is among the highest compared to the OECD countries: cost of living is high (Rosenhek and Shalev 2013), as are poverty rates of children (Endeweld et al. 2018). Given the diverging destinies thesis (McLanahan 2004; McLanahan and Jacobsen 2014) and the stratified patterns of divorce perspective (Härkönen and Dronkers 2006; Kaplan and Herbst 2015), it is important to examine how relations between children and divorce vary across both ethnic and socioeconomic groups. Accordingly, we will compare Jews and Palestinians, controlling for economic inequality between these two groups.Footnote 1 This examination will enable us to learn whether and how the effect of children on divorce is contextual, a subject that has received limited attention (for exceptions, see Chan and Halpin 2002, 2008; Hart et al. 2017; Jalovaara 2013).

To address the above issues, we use a unique dataset created specifically for this research that merged administrative data from the tax authorities with the National Insurance Institute (NII) database. We focus on those women who married in 2003, following them until 2015. These data include information on ethnicity, as well as detailed time-varied information on several dimensions of children, income and marital status, enabling us to control for the endogeneity between children and divorce (Svarer and Verner 2008).

The research examines two main questions: How are two aspects of having children (young children at home and number of children) related to the likelihood of divorce in Israeli society? Does the relations between children and divorce vary when comparing Israeli-Jews to Israeli-Palestinians, as well as household income levels?

2 Pronatalism in Israeli Society

The majority of scholars consider Israel a family-centered society (Fogiel-Bijaoui 2002; Hashiloni-Dolev 2018; Izraeli 1988). This assertion is founded on behavioral, cultural and policy criteria. At the behavioral level, Israel is characterized by relatively high marriage and birth rates, as well as low divorce rates, compared to most other Western societies (Fogiel-Bijaoui 2002; OECD 2018).With regard to divorce, rates doubled from the early 1970s to the late 1990s, reaching a certain level of stabilization in the last decade (ICBS 2018). According to the NII data used for this research, the rate of families of divorce with children dropped from 8.6% in 2003 to 8.3% in 2015.

The importance that Israeli culture attaches to family life in general and childbirth in particular can be attributed to various factors: the Jewish faith, collective Holocaust trauma, ongoing security concerns and the “demographic problem” of a growing Israeli-Palestinian population (Fogiel-Bijaoui 2002; Izraeli 1988; Peres and Katz 1991).

Israel’s high fertility rates go hand in hand with its pronatalist policy, which, since the state’s establishment, has encouraged (mainly Jewish) women to have many children (Berkovitch 1997). Indeed, Israel has historically had a pronatal welfare policy for Jewish women; specifically, the enduring Israeli-Palestinian conflict and the ongoing growth of the Palestinian minority has led to the construction of motherhood as a national mission for Jews (Yuval-Davis 1996) and the framing of the conflict as a “demographic threat” to them (Berkovitch 1997). In the early years of the Israeli state, a special grant was given to women who had ten children or more (Stoler-Liss 2003). The state thus views the Jewish mother’s womb as a means of coping with existential fear (Yuval-Davis 1996). Jewish mothers have played an inherent role in Israeli nation-building, and the welfare benefits accorded them are meant to help achieve that goal. Ajzenstadt and Gal (2001) argued that Israel’s welfare policy has been marked by a concern for national-demographic goals, and less with women as individuals who require social benefits. This claim is reflected in the fact that Israel has the world’s highest number of fertility clinics proportionate to population size (Hashiloni-Dolev 2006). Single Israeli women have the same entitlements and access to assisted fertility as do married women, which is by no means self-evident in developed countries (Hacker 2012). Israeli nationalist discourse underlies the Single-Parent Family Law, enacted in 1992 during peak immigration from the former Soviet Union, which, among other things, expanded income support allowances for poor single mothers, offering mothers compensation for childcare duties (Helman 2011). The central justification for the legislation was that these mothers take part in the Zionist project either as immigrants or as women raising its future generations (Herbst and Benjamin 2012).

However, Israel’s pronatalist policy is a double-edged sword. While welfare policy and the courts have codified the right to bear offspring and women are encouraged to have many children, the state has not ensured the family’s economic wherewithal to raise them (Renan Barzilay 2012). This aspect of policy has worsened since the 2003 welfare reform (Herbst 2013), which instituted cutbacks anchored in a neoliberal ideology that delegitimized welfare recipients and disadvantaged segments of society (Ajzenstadt 2009; Herbst 2013). The reforms reduced child support (payment assurance) allowances (given by the state in the case of non-paying fathers) and cut income support allowances by 25–35% and universal child allowances by 40–50% (NII 2004). In fact, despite the high fertility rates, Israel has among the lowest level of child benefits relative to the OECD average in terms of both payments per capita and the average wage (about 1.6% of average wage compared to 4% in the OECD; Bendelac 2017). These reforms have significantly pushed up poverty rates, especially for families with children (Endeweld et al. 2018). The current research examines relations between children and divorce within this context.

3 The Relation Between Children and Divorce

Several theoretical links between childbearing and divorce have been suggested in the literature, with a distinction between those who see children as a stabilizing mechanism and those who see them as destabilizing. The dominant approach on the subject – the economic approach – suggests that children reduce the likelihood of divorce as they constitute “union-specific capital” (Becker et al. 1977). Children are a marital-specific investment and hence become less valuable to parents when the marriage dissolves (Becker et al. 1977). Marital-specific capital increases marital gains in the current marriage more than outside utility, lowering the probability of divorce (Chiappori et al. 2016). Another theoretical approach in this direction, of social psychology (Brines and Joyner 1999), views children as a joint product that increases the partners’ commitment to the union.

In contrast to this line of reasoning, several explanations suggest that children may decrease parental satisfaction and thus destabilize marriages and encourage divorce (Twenge et al. 2003). For instance, the role conflict model explains how the reorganization of social roles toward the traditional family pattern when children are born may lead to marital dissatisfaction. The parental role adds to other roles, such as professional roles, which create tensions and conflicts within couples. In addition, the restriction of freedom model suggests that the presence of children in the household limits the freedom of parents, thereby increasing dissatisfaction, especially among mothers of young children (Twenge et al. 2003), given the ideology of intensive motherhood (Hays 1996). Lastly, the financial cost model suggests that since raising children is expensive, this adds financial pressure on the couple, increasing the likelihood of divorce (Twenge et al. 2003).

Another theoretical question is whether the relationship between children and divorce changes as the number of children increases. Due to the selection process and depending on the normative number of children, couples who have few children might have higher risks of divorce than both childless couples and couples with more children (Jalovaara 2013). Moreover, in keeping with the financial stress and role conflict models (Tweng et al. 2003), a large number of children could place more financial stress upon parents and generate greater role conflicts. In a different light, more children means that the couple have already passed the selection stage of one child (Jalovaara 2013), suggesting that marriages with many children would be more stable.

In sum, theoretically, children could be stabilizing or destabilizing mechanisms of marriage, and the pattern that emerges might be influenced by the number of children. It is of great interest to examine these opposing theoretical arguments in the Israeli context, given the centrality of having (many) children, together with the high cost of living and the relatively high percentage of mothers in the labor force (Stier 2010).

Empirical evidence is mixed regarding the direction of the effect of the number of children on divorce (Hart et al. 2017; Lyngstad and Jalovaara 2010; Svarer and Verner 2008). In the United States, the first child lowered the likelihood of divorce compared to childless women, whereas subsequent children had the opposite effect (Lillard and Waite 1993). In Denmark, too, additional births increased the likelihood of divorce (Svarer and Verner 2008). Similarly, Murphy (1985) found that couples in Britain with four or more children face significantly higher divorce risks. Other studies have concluded that additional children reduce the likelihood of divorce in Italy and Spain (Coppola and Di Cesare 2008; Vignoli and Ferro 2009) and in Russia (Chiappori et al. 2016). A study comparing marriage and cohabitation in Finland (Jalovaara 2013) also found a non-linear effect: having more than one child lowered separation rates compared to having no children, while having a single older child modestly increased the risk of marital separation. This non-linear effect may be related to selection, since given the two-child norm in Finland, couples who have little trust in the continuity of their marriage are less likely to have a second child (Jalovaara 2013).

Findings also suggest that the ages of the children are an essential predictor of divorce, as people are reluctant to separate when they have young children. The likelihood of divorce has been found to be lowest when children are young and to rise as they age, although findings vary somewhat from country to country (Andersson 1997; Bernardi and Martínez-Pastor 2011; Hart et al. 2017; Jalovaara 2001; Kalmijn and Poortman 2006; Todesco 2011; Twenge et al. 2003; Vignoli and Ferro 2009).

4 Relations Between Socioeconomic Position, Ethnicity, Children and Divorce

In addition to examining relations between several aspects of children and divorce in the general population, it is important to take a closer look at how these relations might vary across ethnic and socioeconomic hierarchies. This question is related to an open theoretical debate regarding the relation between divorce and inequality (Härkönen et al. 2017). Focusing on the ways children affect divorce, two main theoretical directions can be suggested.

First, McLanahan (2004) argues that the family changes associated with the second demographic transition (such as decline and delays in fertility and marriages, and an increase in divorce) have different implications for children, depending on the socioeconomic status of the mother. That is, for children born to mothers with a college education, the changes in family behavior are associated with gains in parental resources, while for children born to less educated mothers, the changes are associated with a loss of resources. McLanahan conceptualized this pattern as the diverging destinies thesis. This thesis gains support in recent studies of relations between household socioeconomic status (or women’s education) and divorce, as in some countries, including Israel, the higher the socioeconomic position, the lower the risks of divorce (e.g., Härkönen and Dronkers 2006; Kaplan and Herbst 2015; Matysiak et al. 2014). One explanation for this pattern is that couples with fewer financial resources suffer more stress in their relationship, resulting from economic pressures (Jalovaara 2003; Oppenheimer 1997). Based on the diverging destinies thesis, we can expect children to have a destabilizing effect on divorce among the lower socioeconomic group.

The second theoretical direction might be termed the heterogeneity thesis (Amato and Antony 2014; Bernardi and Boertien 2016, 2017; Erman and Härkönen 2017; Härkönen et al. 2017). This approach claims that “the socio-economic differences in family instability are less important in affecting inter-generational inequality than often thought” (Härkönen et al. 2017, 179). Accordingly, this stems from two different social processes – one related to class or socioeconomics (Bernardi and Boertien 2016, 2017) and the other to ethnicity (Erman and Härkönen 2017). Here, we focus on how children affect divorce. Based on this hypothesis, then, the destabilizing effect of children might be stronger among more affluent class groups. Moreover, in ethnic groups in which divorce is less common, and where women are more economically dependent on their spouses, children might become a stabilizing mechanism.

With regard to Israeli society, earnings inequality is one of the highest among OECD countries (Rosenhek and Shalev 2013), as is the poverty rate of children (Endeweld et al. 2018); and ethnicity is related to economic and employment status (Swirski et al. 2017). Moreover, there are differences between Israeli-Jews and Israeli-Palestinians, not only regarding fertility and family patterns (Sabbah-Karkaby and Stier 2017), but also because the economic dependence of Palestinian women on their husbands is higher than that of Jewish women, as the former’s employment rate is about half that of the latter (Stier and Herzberg 2013). Under these conditions, responsibility for the children might prevent Palestinian women from breaking up the marital relationship despite the strains. Therefore, the Israeli context is an excellent case for examining the conflicting theses regarding the ways divorce relates to inequality.

Only a handful of studies have examined variations in the relation between children and divorce across ethnic or socioeconomic groups. For instance, a study in Britain found that couples with children are at a substantially higher risk of divorce than similar childless couples, but the destabilizing effect of children is most pronounced for low-income households (Chan and Halpin 2002). Chan and Halpin (2008) also found divorce to be associated with premarital birth, and premarital birth to be strongly associated with low educational attainment. They found that those with less education are more likely to divorce than university graduates. Moreover, they are much more likely to have premarital births, which means that their children are more likely to have a destabilizing effect on the marriage. These results clearly echo some of the concerns for the “diverging destinies” of US children (McLanahan 2004; McLanahan and Jacobsen 2014).

5 Data and Methods

The study used a unique dataset created specifically for this research. This dataset merged administrative data from the tax authorities, including information about employment income and pensions, with the National Insurance Institute (NII) database, which includes sociodemographic information (e.g., marital status, number of children, age of children), as well as information about social benefits (e.g., disability benefits).

The administrative data are for all Israeli women who married in 2003. We took a random sample of 25% of these cases and followed them until 2015. From the year of marriage (2003), we created person-year files, to which each woman contributed an observation for every year she was married and living in the countryFootnote 2 (i.e., until one year after she divorced or until 2015). Our sample totals 20,473 women, yielding 127,725 records for all years covered.

Using this dataset accorded three major advantages. Firstly, as it covers 25% of the entire 2003 marriage cohort, our data include a very large number of cases that fit our research questions, aimed at comparing the effect of children on divorce across ethnic and income levels, given that divorce rates in Israel are relatively low. Secondly, the data provide longitudinal panel information on our main research variables, namely age of children, number of children, household income level and marital status. Thirdly, in contrast to cross-sectional data, our micro-level panel data enabled us to compare the effect of a time-varying variable, such as number of children, on an event, such as divorce.

The dependent variable is the likelihood of each woman to divorce at time t, provided she was previously married (divorced = 1; married = 0).Footnote 3 During the period of study, 8% of the women in our sample reported divorcing.

We used two time-varying measures to assess the effect of children on likelihood of divorce. The first is a variable measuring whether the women has children up to age three (coded “1”) or not (coded “0”) in the household. The second is a measure of the number of children, divided into six groups: no children (the reference group), one child, two children, three children, four children and five or more (due to the small number of families with more than three children in our relatively young population, some models include only four categories: no children, one child, two children and three or more).

To estimate the effect of socioeconomic position on divorce, we used a measure of the couple’s annual co-earnings from work.Footnote 4 Based on that, we divided the sample into six groups: no earnings (W0),Footnote 5 low earnings (W1: 50,000 NIS per year, the equivalent of one minimum monthly wage); low–middle earnings (W2: 50,000–100,000 NIS per year); middle earnings (W3: 100,000–150,000 NIS per year – the reference category); middle–high earnings (W4: 150,000–300,000 NIS per year); and high earnings (W5: over 300,000 NIS per year).Footnote 6 In the partial regressions (by income groups), we used only four (rather than six) earnings levels: W0, W1, W2 + W3, and W4 + W5, reflecting no earnings, low earnings, middle earnings, and high earnings, respectively.

To measure ethnicity, we used a dummy variable that indicates whether (=1) the woman is an Israeli-Palestinian (0 = Jewish).

We also controlled for variables that might affect the likelihood of divorce, consistent with studies in Israel and elsewhere (Cooke et al. 2013; Kaplan and Herbst 2015; Lewin 2006; Lyngstad and Jalovaara 2010; Raz-Yurovich 2012): marriage duration (in years); marriage duration squared; average number of working months; whether the woman resides in the periphery or in central Israel; age; whether the woman had been previously married; and a set of dummy variables that indicate whether (=1) or not (=0) the individual is an ultra-Orthodox Jew or immigrated to Israel in 1990 or afterwards. All independent variables were measured at t–1. Descriptive statistics for all variables are presented in Appendix A.

Fertility and divorce should be examined as two related processes, as fertility (number of children in our data) both affects the tendency to divorce and is affected by it. In order to overcome this loop of causality between the independent (number of children) and dependent (divorce) variables of a model, we estimated the effect of children on divorce by implementing a panel logit model with a dichotomous (0/1) dependent variable (divorced). The main regressors are: children up to age 3 and number of children. In order to examine whether the effect of children on divorce varies by ethnic group and income level, we estimated the regression model once for the total sample and then for the two ethnic groups (Palestinians and Jews), as well as for four groups of household income levels.

To check the robustness of this panel logit regression model, we applied two alternative models. The first used number of children as a continuous variable and number of children squared, to examine non-linear effects. We refer to this model in the Findings section (see also Appendix B). The second alternative was an IV model for panel data (where the first stage is OLS and the second stage is probit), under the assumption that one of the regressors – number of children – is endogenous, as couples who are less prone to split up are likely to have more children. Therefore, one might erroneously conclude that children stabilize marriage (Svarer and Verner 2008).

For this second model we defined two variables as (dummy) instruments: second marriage (assuming the number of children will be lower) and receipt of disability pension (assuming that people with disabilities will tend to have fewer children). Since these instruments are weak, we preferred to use the simple model outlined earlier. In the Findings section, we report the findings of this alternative model as well, as a robustness check.

6 Findings

Table 6.1 presents the results of the panel logit regression model. We start with our first research question, focusing on the coefficients of children aged 0–3 in the household and the number of children in the total sample (column 1). As can be seen, having young children (up to age 3) in the household decreases the likelihood of divorce. When the dummies of the number of children, in the sample as a whole, are compared to the dummy of couples without children (the reference category), all coefficients are positive and the pattern seems to be non-linear. One child increases the risk of divorce compared to childless couples, as do two and three children, but these risks decrease gradually as the number of children rises, up to three children. There is no statistically significant difference between couples with four children and those without children, but couples with five or more children reversed the chances of divorce, raising it again, perhaps because of the economic difficulties that this situation creates.

Table 6.1 Results (odds ratios) from regression, women, 2003–2015 (dependent variable: divorced)
Full size table

Regarding ethnicity, as expected, we found that Israeli-Palestinians had lower risks of divorce than Israeli-Jews. With regard to the effects of co-earnings, we found that couples with no income from work (W0), couples with low earnings (W1) and those with low-middle earnings (W2) tended to divorce more than couples with middle earnings (W3, the reference category). Having middle-high earnings (W4) decreased the odds of divorce compared to middle earnings, and very high earnings (W5) was in the same direction (i.e., negative), but non-significant. In other words, there is a clear class effect of the tendency to divorce. Being ultra-Orthodox Jews or immigrantsFootnote 7 decreased divorce risks, while number of annual working months, age and previous divorce increased it. As expected, marriage duration had a non-linear effect on divorce in the total sample.Footnote 8

With regard to the second research question, i.e., whether relations between children and divorce vary by ethnicity and household income level, findings of the partial regressions are presented in columns 2–7 of Table 6.1. We start with ethnicity (columns 2 and 3). While young children reduce divorce risks among both groups, the number of children has a differential effect. The results among Israeli-Jews (comprising about 80% of the sample) resemble those of the total sample, that is, there is a positive, but non-linear, relation of children to divorce. However, among Israeli-Palestinians, there is no significant difference between families without children and those having up to two children. The third and fourth child, however, reduce the chances of divorce compared to a family without children, unlike the (positive) pattern found among Jews.

In an alternative model (see Appendix B), we estimated the effects of number of children by a continuous variable and number of children squared (to look for non-linear effects). Running this model separately for each group, we found the effects to be non-linear in both, but in opposite directions: while for Israeli-Jews, the linear term is positive and the squared term is negative, for Israeli-Palestinians, the linear term is negative and the squared term is positive. Taken together, the models indicate that the number of children increases divorce risks among Israeli-Jews, though this effect decreases in large families, while children stabilize the marriage among Israeli-Palestinians, but the stabilizing effect decreases in large families.

Interestingly, annual earnings also affect divorce differently in these two ethnic groups (Table 6.1). The pattern among Israeli-Jews resembles the one in the total sample, with divorce risks higher among couples with no earnings or low earnings. Among Palestinians, in contrast, the only difference found is that families without employment income tend to divorce more than the middle-earnings group (W3). On the whole, differences between income levels in the Israeli-Palestinian population are lower than for Jews, possibly because of a lower tendency to divorce and lower heterogeneity in earnings among the Palestinians (who, on average, earn much less than Israeli-Jews) (Sabbah-Karkaby and Stier 2017).

Turning to a comparison across couple’s earnings, columns 4–7 present the results of the partial regression in four earnings levels (rather than six levels, due to the small number of cases): W0, W1, W2 + W3, and W4 + W5, reflecting no earnings, low earnings, middle earnings, and high earnings, respectively. We also combined couples with three or more children due to the small number of cases of couples with four or five children (this is a relatively young population that married in 2003). First, as was found for both ethnic groups, children up to age 3 decreased the risks of divorce across all household income levels. However, the number of children seems to operate differently depending on income level. Among couples with no earnings from work (W0), i.e., at a low economic level, where divorce risks are high, the number of children is not significantly associated with divorce. Among households with low earnings (W1), one child increases the odds of divorce compared to couples with no children, but beyond that, there were no significant differences between couples with a higher number of children in their risks of divorce (possibly due to the small number of cases). In households with middle earnings (W2 + W3), one child and two children increase the risks of divorce, compared to childless couples, while the contribution of three and more children to divorce risks does not differ significantly from that of no children. Lastly, among couples with high earnings (W4 + W5), one child greatly increases the odds of divorce compared to no children, as do two children (albeit to a lower extent), and having three children or more continues the downward trend in odds of divorce, although not significantly.

A comparison of these models with the IV model (not shown, results can be obtain from the authors) reveals similar trends in the explanatory variables, although the IV model yields a smaller number of significant coefficients. This is so for both child variables: young children at home (decreasing divorce tendencies in both cases) and the number of children. On the whole, all models found an increase in income level to reduce the chances of divorce (when coefficients were significant).

7 Discussion

The study examined how young children at home and number of children might be related to marriage stability in a pronatalist society. While most studies on the effect of children on divorce have focused on countries with fertility levels below or near the replacement level, we explored whether the stabilizing effect of children on marriage holds in a society with the highest fertility rate among OECD countries (OECD 2017). Given the relatively high cost of living in Israel (Rosenhek and Shalev 2013), low state support for families with children (Bendelac 2017; Renan Barzilay 2012) and differences in fertility and divorce rates between Jews and Palestinians, the main ethnic division in Israeli society, we looked at these relations comparing the two ethnic groups as well as households with different earning levels. On the whole, we found that the effect of children on the likelihood to divorce is heterogeneous and depends upon the specific aspect of children examined, as well as the ethnic and income group focused upon.

Regarding ethnicity, after controlling for many predictors of divorce in the model, we found major differences between Israeli-Jews and Israeli-Palestinians in a number of respects. First, as expected from previous research (Kaplan and Herbst 2015), divorce risks are higher among the Jewish population. Second, children are related to divorce in distinct patterns. Children (especially the third and fourth child) in Israeli-Palestinian families, to which the pronatalist policy is less directed, stabilized the marriage. For Israeli-Jews, the opposite is true: children tended to destabilize the marriage. Compared to households without children, a household with one, two, three and five (or more) children increased the risk of divorce, though with decreasing effects from the first to the third child and a change in trend once there were five children. On the theoretical level, this non-linear effect can be explained in keeping with the financial stress and role conflict models (Tweng et al. 2003). It is possible that the hardships of having a second and third child are similar, because however the couple manages with two children, they can also manage with three (and perhaps four) children. Once they pass a certain threshold (a fourth child), however, the risk apparently gets considerably greater.

The different patterns of the effect of number of children on divorce might be explained by differences between these ethnic groups regarding level of conservatism, selection into different fertility patterns (higher among Israeli-Palestinians), the commonality of divorce and normative breadwinner models. As opposed to Israeli-Palestinians, most Israeli-Jewish couples are dual earners working long hours, even when children are young (Mandel and Birgier 2016; Stier 2010). This places a high burden on the shoulders of young Jewish parents, especially when there are many children, in accordance with the financial cost and role conflict models (Svarer and Verner 2008; Twenge et al. 2003). The situation is aggravated by Israel’s pronatalist policy, which encourages Jewish women to have many children, yet fails to support the family in raising them (Renan Barzilay 2012), especially if the couple have employment difficulties.Footnote 9 Combined with the country’s relatively high cost of living (Rosenhek and Shalev 2013), families with children thus become a locus of multidimensional pressures. However, this seems to work differently among Israeli-Palestinians. It is possible that, in this case, an increase in the number of children raises the dependency of the wife on her husband, thereby reducing the odds of divorce – a dependency which is also related to Palestinian women’s low labor force participation (Sabbah-Karkaby and Stier 2017). Moreover, following the logic of Erman and Härkönen (2017), where divorce is less common, children may strengthen the marital bond. If this group also has a low socioeconomic status, as is the case of Israeli-Palestinians, the relation between children and divorce might decrease economic inequality while at the same time reproducing gender inequality.

We also obtained heterogeneous findings when looking more closely at the relations between children and divorce across different levels of household earnings. As a whole, for those couples with low income levels (overlapping with the Israeli-Palestinian population), who are more likely to divorce than couples with high earnings, the number of children was not found to be significantly related to divorce. That is, families living in poverty are more likely to divorce than those earning high wages, but the number of children does not play a significant role in explaining divorce. On the other hand, among those with average and high incomes (overlapping more with the Jewish population), children increase the chances of divorce, albeit with a decreasing effect from the first child to each additional child.

Our findings are therefore mixed with regard to the broader question of how children and divorce are related to socioeconomic inequality. On the one hand, the results support the diverging destinies thesis, showing a clear class effect of earnings on divorce. On the other hand, taking the number of children into account, children were found to be destabilizing only among the more affluent groups, which somewhat contradicts the diverging destinies thesis and supports the heterogeneity thesis instead. In that sense, having children may weaken social inequality. Of course, more research and longitudinal studies are needed to test this interpretation.

It should be noted that, across all income levels and among both ethnic groups, young children at home increase marriage stability, supporting findings in earlier studies (Andersson 1997; Becker et al. 1977; Chiappori et al. 2016; Hart et al. 2017; Kalmijn and Poortman 2006). It seems, then, that in Israeli society, young children constitute “union-specific capital” (Becker et al. 1977) and hence reduce the likelihood of divorce for all groups, while the extent to which the number of children stabilizes the marriage depends on the couple’s ethnic and income group.

The study is not without limitations. Our administrative data did not include information on total income (beside salaries). Future research should examine the relation between children, income and divorce using additional income measures. Furthermore, our dataset unfortunately lacked a measure of religiosity (besides belonging to the ultra-Orthodox Jewish population). Given the importance of religiosity in explaining family patterns among both Israeli-Palestinians and Israeli-Jews (Okun 2013), this could be a fruitful direction in future studies. It would also be helpful to run separate analyses for ultra-Orthodox Jews, due to their high fertility and poverty rates. As mentioned, this was not possible in light of the small number of cases in our dataset. Additional research might also benefit from taking a comparative cross-country perspective, combining family policy research (Saxonberg 2013) with marital and divorce behaviors (Cooke et al. 2013; Härkönen and Dronkers 2006; Kaplan and Stier 2017).

Notwithstanding the above limitations, the findings of the current research point to the importance of examining relations between children and divorce across groups in the society, instead of assuming a linear and consistent trend. In that sense, the answer to the question of whether the more is really the merrier might depend, at least partly, on the ethnic group to which the couple belongs, as well as how much money they earn.