1 Introduction

In this paper we evaluate the effect of the introduction of divorce in Chile in 2004 on a woman’s decision regarding when to have a first child. The evolution of Chilean society and the historical trends of the regulation of marriage in Chile provide a rich context to study a topic where the economic literature has yet to form a consensus. In 1884 the regulation and registration of marriages passed from the Catholic Church to the State. However, the influence of the Church was not eliminated, especially with regard to the termination of marriages. Couples were not allowed to divorce until 2004, when new legislation introduced the concept of “divorce a vincula matrimonii” or complete divorce, under which either spouse can request a complete termination of the marriage. The new legislation also introduced monetary compensation for a spouse who gives up professional development or advancement for the good of the household.

The Chilean divorce law (henceforth, DL) standardizes the process through which a marriage is terminated and thereby removes the uncertainty and potential negotiation costs the couple would incur if they decide to obtain the divorce. Becker (1974) suggests that the outcome of a divorce is related to specific investments made during the marriage, and lists children as one such type of investment. A married couple receives utility from children, but if the time cost of raising a child falls mainly on the woman, her career prospects are reduced. If this cost outweighs the benefit to the woman herself, having children reduces her net utility, and she would opt not to do so. If children are potentially Pareto-improving, the man might offer to transfer resources to the woman in order to compensate her for the costs she bears, so that having the child would become individually rational for both parties. While such a transfer may be fairly easy to implement for couples who anticipate being able to cooperate indefinitely, others may be unable to make such an arrangement unless they could count on an external enforcement mechanism. The system in place in Chile before the divorce law was introduced could not provide that enforcement, and many separated men were able to avoid making spousal support payments. Thus, the new divorce law in effect raises fertility because it rectifies a market failure by making some desirable implicit contracts feasible.

The increase in fertility may actually have little to do with the ability to divorce itself, but instead is driven by the fact that the divorce law now enforces implicit contracts concerning alimony. It may also account for an increase in non-marital fertility, assuming that the courts now make a greater effort to enforce child support payments more generally, which suggests that the divorce reform is more closely associated with decreases in the mean age of first-time mothers than with decreases in the mean age of first-time brides. Moreover, the effect on the fertility may be greater for women with higher potential incomes, because they have more to lose if their career is derailed.Footnote 1

The goal of this paper is to determine the degree to which the DL affects a woman’s decision regarding when to have her first child. Using the Social Security Survey (Encuesta de Previsión Social–EPS) Panel 2002–2009, we find that the DL has a positive effect on the hazard rate of having a first child for more educated women, controlling for the woman’s age and length of marriage, the negative trend observed on fertility rates in Chile since the mid-1960s, and other socioeconomic characteristics. We also find that a higher potential income level is associated with an increase in the hazard rate for first birth.

This paper is structured as follows. Section 2 presents the context in which the Chilean DL was introduced, and our motivation to focus on first birth. The third section consists of a review of the literature which focuses on divorce laws and fertility decisions. In the fourth section we present the data, describe the empirical strategy and discuss the results of the estimation. Finally, in the fifth section we present our conclusions.

2 Background and motivation

Historically religion has played an important role in marital legislation in Latin America, where, in most countries, divorce was prohibited throughout the twentieth century (Rossetti 1993). Attempts to implement divorce in Chile can be traced back more than a century.Footnote 2 The most recent was in November 1995 and was finally approved by the Chilean Congress in March 2004 and implemented in November of the same year.

Before 2004 Chilean marital law was based on the 1884 civil code, which permitted dissolution of marriage through annulment only. To obtain an annulment, spouses must argue before a judge that some aspect of the marriage was improper, for example, that one of the witnesses was under 18 years of age, or that the name of the spouse was misspelled on the marriage certificate. The 1884 civil code also allowed for the separation of the couple but this separation did not dissolve the marriage and marital obligations continued.Footnote 3 The DL suspended the succession rights, introduced a monetary compensation given to the spouse who sacrifices personal development for the good of the household, and most importantly, introduced a standard procedure to terminate a marriage, thereby removing the uncertainty and potential negotiation costs that seeking the annulment implied.Footnote 4 Moreover, the DL allows unilateral and no-fault divorce, requiring in both cases a 1-year separation period. However, if one of the spouses files for divorce on the grounds of a fault, no separation period is required.

The DL has contributed to the observed trends of first births and first marriages in Chile in the last decade. The left panel in Fig. 1 shows a clear change in the trend of first births after the implementation of the DL in late 2004. Between 1997 and 2004 the proportion of first births decreased from 1.45 to 1.16 %, but increased from 1.21 % in 2005 to 1.31 % in 2008, before decreasing to 1.22 % in 2012. On the other hand, the right panel in Fig. 1 shows a persistent decline in the proportion of first marriages before 2005.Footnote 5 Between 1997 and 2005 marriage records show that the proportion of single women who got married with respect to the total female population decreased from one to 0.63 %. In 2006 this figure increased to 0.67 %, and decreased to 0.61 % in 2012.

Fig. 1
figure 1

Proportion of first births and first marriages for women in Chile 1997–2012, with respect to total female population (%). Source: National Institute of Statistics, Chile

Full size image

The decreasing trend in first marriages is in line with the increasing preference towards cohabitating relationships: according to Census information, the share of the population (15 or over) in a cohabitating relationship increased from 5.7 % in 1992 to 16.2 % in 2012. The decline in first marriages significantly decreases the contribution to first births by married couples between 1999 and 2006 (from 36.5 to about 22 %), according to administrative records (Larrañaga 2006). Palma (2006) suggests that 50 % of out-of-wedlock births are to women in a cohabitating relationship; separated women and partnerless women contribute 10 and 40 %, respectively. Therefore, given that the DL clarifies the procedure to terminate a marriage, cohabitating couples may decide to get married and have a first child, which may explain in part the relative stability of the proportion of first marriages and the positive trend in first births after 2005.

Another important factor in understanding women’s first birth decisions and the contribution of married and cohabiting women is the trend in average age at first marriage for women and the average age at which they have their first child. The trends for both are positive, suggesting that Chilean women have become less willing to sacrifice certain personal development for marrying and starting a family (see Fig. 2 below). This trend is not surprising given the increased participation of women in the labor market and significant increases in female educational attainment (see Larrañaga 2004). Moreover, the overall fertility rate has declined since the mid-1960s (Economic Commission for Latin America and Caribbean, ECLAC, and INE 2002). However, after the DL was approved, administrative data from INE suggests a decrease in the average age at which women give birth to their first child: in the period 1997–2005 this average increased from 22.73 to 23.39, but fell to 23.03 in 2006 and 23.14 in 2007. Although it is still too early to conclude that there is a change in the trend, it appears that the average age at first pregnancy has declined after the DL.

Fig. 2
figure 2

Average age at first birth and first marriage for women in Chile 1997–2008. Source: National Institute of Statistics, Chile

Full size image

Moreover, a simple comparison of both trends in Fig. 2 suggests that more Chilean women have the first child before getting married for the first time.Footnote 6 This trend is consistent with the increasing preference towards cohabitation mentioned before. Given the DL, and considering that 45 % of Chileans agree with the idea that cohabitating couples should get married when they have a child (35 % disagree and the rest are indifferent), according to the 2008 Bicentennial Survey (see Salinas 2008), it is possible that more cohabitating women have the first child and marry afterwards.

The evidence presented in Figs. 1 and 2 suggests that the DL had an effect on first births for both married women and cohabitating women. Married women may delay the first birth to see if the relationship succeeds, while in the case of cohabitating women the effect of the DL on first birth may be related to the decision of whether to get married. However, the available data do not suggest a substantial change in the average age of women at first marriage after the passage of the DL.

The trends in first marriages and first births may seem surprising considering that Chilean society is very traditional—the Chilean Marital Law of 1884 aimed to protect this tradition. Considering that Chile has reached a high level of development and economic growth in the last 20 years, the negative trends shown in Fig. 1 are less surprising. In fact, decreasing marriage and fertility rates are also observed in many western European countries. For instance, between 1998 and 2002 the largest declines in the birth rate are observed in Portugal (19.7 %), Malta (18.4 %), Germany (16 %), Luxemburg (13.5 %), and the Netherlands (11 %). Other Eurozone countries, which receive larger flows of migration or have child-benefit programs and a relatively more flexible post-natal legislation, show a positive or relatively lower decreasing trend in birth rates in the same period: for instance, Spain (28.9 %), Greece (15.6 %), France (0 %) and Italy (−2.6 %).Footnote 7 Since Chile does not have a child benefit program as many European countries do and post-natal leave was extended to 3 months only in 2011,Footnote 8 we would not expect the effect of the DL to be confounded by any other child-related legislation change.Footnote 9

It is also important to consider unplanned pregnancies in Chile (see Diaz, year unknown). Unfortunately, administrative records on pregnancies do not distinguish between planned and unplanned pregnancies. As far as under-age pregnancies are concerned, Molina et al. (2004) reports that most of the under-age pregnancies are to eventual high-school dropouts. The conclusions on first births presented in this section do not change when women with first conceptions at age 19 or younger are included in the analysis.

3 Literature review

The US literature on divorce law has examined the introduction of no-fault and unilateral divorce in the 1970s.Footnote 10 No-fault divorce means a divorce granted without a “fault” by one of the spouses (for instance, adultery, physical or mental abuse, and abandonment), while unilateral divorce means that the consent of both spouses is not required. These new laws passed at different times in different states,Footnote 11 and much of the literature has taken advantage of this variation to identify the effects of these laws. The literature on divorce laws and fertility decisions is more scarce.

Alesina and Giuliano (2007) use unilateral divorce to analyze the relationship between divorce and fertility. They claim that more flexible divorce laws have two possible effects. The first effect is that the number of children born to married couples decreases, and people have more children out of wedlock. The second effect is that the cost of entering the ‘wrong’ marriage is lower, because it is easier to dissolve the relationship. Alesina and Giuliano say that the literature on divorce law has focused on the first effect and has not given sufficient attention to the second effect.

To analyze both effects, Alesina and Giuliano use birth certificates from the National Vital Statistics of the US to construct different measures of fertility at the state level. Using a panel of states for the period 1968–1999, they recover the impact of the introduction of unilateral divorce on state fertility rates. Their results show that the introduction of no-fault divorce is associated with a decline of 3 % in the fertility rate for those states that introduce the new legislation. Alesina and Giuliano also construct a panel of state-age records based on Census data from 1960 to 1990. Their dependent variable is the number of children ever born to women residents aged 15–44 in states that adopt unilateral divorce. Their results corroborate the negative effect of no-fault divorce on fertility rates. They conclude that as divorce becomes easier, couples feel less locked in, and women who are considering having children or are already pregnant are more willing to “try” marriage. Therefore out-of-wedlock fertility declines and marriage rates increase.

Drewianka (2008) tests the effect of no-fault and unilateral divorce law in the United States on US marriage divorce, fertility and legitimacy rates. As far as the latter two rates are concerned, he finds that no-fault divorce lowers birth rates by about 2–3 %, with the effect concentrated on married couples, whereas unilateral divorce increases marital fertility but decreases non-marital birth rates. Moreover, the negative effect of unilateral divorce on non-marital fertility becomes stronger as time passes.

A study by Ekert-Jaffe and Grossbard (2008) relates directly to the monetary compensation introduced in the DL. Ekert-Jaffe and Grossbard develop a model of fertility in which it is assumed that a woman makes decisions regarding whether to have marital fertility, non-marital in-couple fertility, or out-of-couple fertility. The less to be gained from marital or in-couple fertility, respectively, the more likely a woman is to choose non-marital or out-of-couple fertility. The authors demonstrate that laws regulating divorce settlements influence the benefits of in-couple fertility relative to out-of-couple fertility. Using a sample of close to 20 different legal regimes, Ekert-Jaffe and Grossbard find that the likelihood of an unpartnered birth is lower when a country or province has community property laws guiding division of assets in case of partnership dissolution that are more favorable to women than the alternatives. The most convincing case is that of New Zealand, where there were large decreases in out-of-couple fertility after community property law favorable to women was introduced.

In a recent paper Bellido and Marcen (2014) examine the effect of no-fault unilateral divorce law reforms on fertility for the case of Europe. They find that the negative effects of divorce law reform on marital fertility rates are transitory. On the other hand, the negative effects on non-marital fertility are permanent. Thus, in both studies by Drewianka and by Bellido and Marcen there are strong negative effects on non-marital birth rates, and mixed or transitory effects on marital birth rates.

We conclude that the literature on divorce law and fertility decisions is scarce and focuses on the response of fertility rates. While rates are important, there are other important aspects of fertility to consider. The overall fertility rate in Chile has decreased from 2.09 to 1.9 births per women between 2003 and 2010,Footnote 12 and it is possible that the introduction of the DL in 2004 may partially explain this trend. However, the monetary compensation provisions of the DL may have substantially different effects for different people, calling for a micro-level analysis.Footnote 13 In the next section we present the data and the empirical strategy, and discuss our results.

4 Data and estimation

According to the Vital Statistics Report 2001–2008 provided by INE, Chilean fertility rates are declining. At the same time, the data suggest that the DL has shortened the time to first births for women with more than 12 years of schooling in the period 2005–2008.Footnote 14 To quantify the conditional probability of first births controlling for the socioeconomic characteristics of the woman, we estimate a hazard model for first births.

The estimation of the hazard of first births raises several issues. In duration data it is not always possible to observe the occurrence of the event of interest. In the case of first births, some women have a first child during their sample period, while others do not. Biases are introduced if we discard the childless women from the analysis. At the same time, these women cannot be treated as women who give birth to a child. Relevant work on applying hazard analysis to fertility has been developed by Newman and McCulloch (1984). The authors explain that hazard analysis corrects for censoring bias without generating selectivity bias. Kravdal (2001) studies the duration of second and third birth intervals in Norway. Li and Choe (1997) study second birth intervals in China, where the one-child policy creates disincentives for couples to have a second child. The present study uses a semi-parametric Cox proportional hazard model and includes the DL information to examine whether this policy has an effect on women’s hazard of having a first child.

4.1 Data: the social security survey (EPS)

The 2002–2009 waves of the Social Security Survey (Encuesta de Previsión Social–EPS), Social Security Subsecretary (2002), provide 8126 birth histories.Footnote 15 For the main analysis we use all women who turned 18 in 1980 or later so that the birth histories comprise no more than 27 years starting at age 18, because we believe that the decisions of younger women are less likely to be determined by potential education or income.Footnote 16 Accordingly, our results should be interpreted as not having conceived their first child at age 18 or before. A woman’s birth history is terminated when she conceives her first child (2820 cases), when she turns 45 (1046 women), or when she is no longer observed in the sample.

The socioeconomic information for the woman is time-invariant and sample means are presented in Table 1. However, the indicator for whether the DL has been passed is a time-varying indicator variable. The birth history of a woman conceiving her first child before November 2004 (when the DL was approved), is unaffected by the DL. The DL only affects the birth histories of women conceiving their first child after this date. Therefore, the DL divides the duration of the spell into two intervals.

Table 1 Table of means–time—invariant covariates
Full size table

4.2 The DL and women’s expected level of education

As mentioned before, the DL standardizes the procedure to terminate a marriage and the possible monetary compensation may create incentives for a woman to have her first child earlier, as she would be entitled to compensation in case of a future divorce that would make it easier to live alone as a single parent. These incentives are likely to be greater for women with higher levels of education. To recover the effect of the DL we estimate the following equation:

$$\ln \,h\left( t \right) = \gamma_{0} T\left( t \right) + \gamma_{1} T80\left( t \right) + \gamma_{2} M\left( t \right) + \beta_{w} W + \beta_{E} E + \delta DL\left( t \right) + \theta E \cdot DL\left( t \right),$$
(1)

where ln h(t) is the log-hazard of first birth at time t, T(t) is a piecewise-linear spline representing the woman’s age, T80(t) is a piecewise-linear spline that provides a time trend with 1980 as the origin, M(t) is a piecewise-linear spline representing the number of years since the woman first married or started cohabitating (t m ),Footnote 17 Ws a vector of time-invariant socioeconomic characteristics,Footnote 18 Es a vector of potential levels of education, and DL(t) is a time-varying indicator for whether the DL is in effect.

The vector E describes the woman’s potential education at age 18, and not her actual education level. The purpose of this time-invariant covariate is to recover the final level of education that an 18-year-old woman expects to eventually attain.Footnote 19 The education system in Chile allows the student to choose a professional track or an academic track once they complete primary school (8 years) and enter secondary school (4 years). The professional track is completed in a professional institute, while the academic track typically ends in a university degree. At age 18, an individual has completed primary and secondary school and already started post-secondary education in a professional institute or university.

For the 92 % of women in our sample who are no longer attending school when they are interviewed by the EPS, their potential education is the highest education level they report. The remaining women report attending school in at least one of the EPS interviews. The potential education of these women will differ from their level reported in the survey if they obtain a new degree at the last institution or in the last educational track reported.Footnote 20

The estimates are obtained using the aML software developed by Lillard and Panis (2003). Their software estimates a Cox proportional hazard model in which the duration baseline is estimated as a piecewise-linear spline in the log hazard, rather than being given a specific functional form such as Weibull. Table 2 presents the coefficients and marginal effects for the estimation of Eq. (1). In column (A), E is a dummy variable for all women with some university experience, i.e., the baseline category is for women with no university experience, while in column (B) the education indicator is for women with a university degree, i.e., the baseline category is for women with no university degree. Using the estimates in Table 2, 3 and 4 present the effect of DL on the hazard of first birth.

Table 2 Hazard of first birth—Eq. (1)—woman’s attained education
Full size table
Table 3 Effect of DL on first birth hazard—specification (1A)
Full size table
Table 4 Effect of DL on First birth hazard—specification (1B)
Full size table

The covariate of interest in Eq. (1) is DL(t), which is included both by itself and interacted with E. Therefore, the coefficient δ gives the effect of the DL for less educated women, while the sum of δ and θ gives the effect of the DL for more educated women.

The results in Table 3 shows that the hazard of first birth for more educated women increases by 61 % with the DL. The effect of the DL for women with lower education is insignificantly different from zero. When we add women who turn 18 between 1970 and 1980 to the sample, the magnitude of the effect of the DL decreases to 47 % and when we further include women who turn 18 between 1960 and 1970 the effect doesn’t change.Footnote 21 The magnitude of the estimates increases in Table 4. However, the conclusion remains unchanged: the DL increases the hazard for more educated women and has no significant effect for less educated women.Footnote 22

The coefficient β E recovers the effect of potential education on the hazard of having a first child for the period before the DL passes. The effect after the DL passes is given by β E  + θ. The top panel in Table 5 shows these estimates. Before the DL passes, women with more education have a lower hazard of having a first child. However, after the DL passes, the effect of education is insignificant. When we add women who turn 18 before 1980, the conclusion remains unchanged.

Table 5 Effect of potential attained education on first birth hazard
Full size table

In specification (A), the coefficient for the baseline hazard spline is significant between the woman’s 18th and 19th birthdays—the hazard rate increases by 26 %. Between the 24th and 35th birthdays, the baseline hazard spline is insignificantly different from zero. After the 35th birthday, the hazard decreases about 16.97 % every year. The estimates are relatively similar in both specifications. The second term, T80(t), captures the effect of the decrease in fertility rates observed in the administrative data. The estimates suggest a negative slope after 1990 in both specifications with a decrease of 3.2 % a year between 1990 and 2000. After 2000, the hazard decreases by about 19.97 % per year. When women who turn 18 in 1970 (1960) or later are included in the sample, we set the origin of the spline in 1970 (1960). The results for M(t) suggest that the hazard of having a first child increases by 23 % in the first year of the relationship. The increase is not as large for each year between the 1st and the fifth year of the relationship. The hazard starts decreasing after the fifth year, with no significant change after the tenth year. Similar results are observed in both specifications. Finally, the education attained by the woman’s mother increases the first birth hazard. This result is not surprising since we would expect a relationship between the woman’s education and the education of her parents.Footnote 23 , Footnote 24 In the next subsection we perform a robustness check on our results by replacing expected education level with potential income.

4.3 The DL and women’s potential income

In this section we assume that an 18-year-old woman is able to foresee her potential income given her expected education level and her future labor experience and further assume that her decision of when to have her first child will be influenced by the income level she would expect to have in the future.

We predict the natural logarithm of the woman’s potential income level at age 25 based on the estimates from a Heckman two-step procedure using data from the EPS 2004. In the wage equation we regress the natural logarithm of hourly income in 2004 on the woman’s education, labor force experience and its square. In the selection equation we also include the education level of the woman’s mother.Footnote 25 Estimates from the wage equation are used to predict log hourly income at age 25 according to the woman’s potential education. The prediction does not include the estimate for the inverse of the Mill’s ratio, since the goal is to obtain a measure of the woman’s expected income, which does not take into account whether the woman chooses to work or not at that particular age.

We estimate a hazard model which includes the interaction of the DL indicator and the woman’s potential income:

$$\ln \,h\left( t \right) = \gamma_{0} T\left( t \right) + \gamma_{1} T80\left( t \right) + \gamma_{2} M\left( t \right) + \beta_{w} W + \beta_{I} I + \delta DL\left( t \right) + \theta I \cdot DL\left( t \right),$$
(2)

where I is the woman’s potential income at age 25. Based on the estimates of column (2) in Table 6, column (2) in Table 7 shows the estimated hazard elasticity of the woman’s potential income before (β I ) and after (β I  + θ) the DL. When the sample includes women who turned 18 in 1980 or later (first panel), the estimates suggest that a 1 % increase of the woman’s potential income is associated to a reduction of 0.82 % in the hazard of first birth before the DL, while for the period after the DL passed, the hazard elasticity is insignificantly different from zero.Footnote 26 These results do not change significantly when the woman’s potential income is predicted based on a Heckman 2-step procedure that uses the father’s education instead of the mother’s or uses both father’s and mother’s education.

Table 6 Hazard of first birth—introducing the woman’s potential income at age 25
Full size table
Table 7 Hazard of first births—interacting DL with woman’s potential income at age 25
Full size table

We conclude that the hazard of having a first child increases for women with higher potential income level after the DL was passed. This result is consistent with the positive effect of the DL on the hazard for more educated women from Eq. (1), and does not change when the sample includes women who turn 18 in 1970 or later (second panel), or who turn 18 in 1960 (third panel) or later: before the DL the reduction in the hazard is 0.64 and 0.58 %, respectively, while the effect after the DL is insignificantly different from zero in both cases.

To further investigate these results, we estimate Eq. (2) adding the woman’s expected education level (E):

$$\ln h\left( t \right) = \gamma_{0} T\left( t \right) + \gamma_{1} T80\left( t \right) + \gamma_{2} M\left( t \right) + \beta_{w} W + \beta_{E} E + \beta_{I} I + \delta DL\left( t \right) + \theta I \cdot DL\left( t \right).$$
(3)

Column (3A) of Table 7 shows the hazard elasticity before and after the DL for Eq. (3) when the education variable reflects university experience.Footnote 27 Column (3B) considers the specification when the education indicator reflects university degree. The hazard elasticity in column (3A) for both before and after the DL do not differ significantly from the results shown in column (2) for any of the three samples. However, in column (3B), the decrease in the hazard is significantly larger than in column (2), both before and after the DL. These results suggest that the DL has increased the hazard of women with higher potential income given her education level. Moreover, women with higher levels of education are likely to have higher potential income, corroborating the results in Sect. 4.2.Footnote 28 , Footnote 29

The results described so far are based on women’s characteristics at age 18, and do not take into consideration the information of a potential partner. We now explore the possibility that an 18-year-old woman considers the expected income of the future partner in her decision of when to have the first child.Footnote 30 The partner’s expected income affects the woman’s labor force participation decision. Therefore, we predict a new measure of the woman’s potential income at age 25 using the estimates from a Heckman two-step procedure that includes the partner’s expected income in the selection equation. When we estimate Eqs. (2) and (3) replacing I with the new income measure, the results corroborate the conclusions obtained in the previous subsection: before the DL the hazard elasticity is negative, and becomes insignificantly different from zero under the DL. This conclusion does not change when we include women who turned 18 before 1980 in the sample.

4.4 Summary of findings

The main result of our study is that the divorce law that passed in Chile in 2004 decreases the age at first birth for women in Chile with higher expected education levels, whether measured as university experience or university degree. The results are similar when potential income replaces expected education level in the estimation. There are no significant changes in age at first birth for women with lower levels of education. We feel that the important factor underlying these results is the monetary compensation introduced in the DL for divorced women who have sacrificed professional development or advancement for the good of the household. These sacrifices are likely to be greater for women with higher expected education levels and higher potential income.

Our results are consistent with those of Ekert-Jaffe and Grossbard (2008), Drewianka (2008) and Bellido and Marcen (2014) on the assumption that a lower age at first birth is likely to be correlated with higher fertility. Eckert-Jaffe and Grossbard find that unpartnered fertility decreases with the introduction of community property laws, while the latter two studies find that unilateral divorce law has strong negative effects on non-marital birth rates.

A woman who sacrifices professional development for the good of the family may subsequently find herself in a bad marriage that she cannot leave because it would be difficult to raise the child alone. Highly educated women and women with high potential incomes have the most to lose with this outcome. The monetary compensation in the Chilean DL that rewards this sacrifice in case of dissolution of the relationship makes investment in children less risky for these women and they have children earlier. The monetary compensation is an enforcement mechanism for the division of assets that is directly comparable to the community property laws favorable to women in the study by Eckert-Jaffe and Grossbard. Partnered fertility becomes more attractive and unpartnered fertility declines. More generally, the dissolution of a bad marriage is made easier, as with the cases of unilateral divorce studied by Drewianka and by Bellido and Marcen, marriage rates increase and non-marital fertility declines.

5 Conclusions

This study analyzes the effect of the new Chilean divorce law (DL) of 2004 on a woman’s decision of when to have her first child. The new DL introduces divorce in Chile and implements a monetary compensation regime, which benefits the spouse who gives up personal and professional development for the good of the household. The effect of the DL may depend critically on this compensation, since it creates an external enforcement mechanism for the transfer of marital resources to the woman in the case where the couple separates. Educated women and women with higher potential income stand to gain the most from this compensation.

Using the 2002–2009 EPS panel, we construct birth histories for women who turn 18 in 1980 or later (3688 women). The hazard model estimates show that, controlling for age, marital duration, education, declining fertility rates, region of residence when the woman was first interviewed and other socio-demographic characteristics, the DL increases the hazard of first births for women with post-secondary education by 61 % (110 % for women with a post-secondary degree). Moreover, the estimates show that before the DL the hazard elasticity with respect to potential income at age 25 is significantly negative but insignificantly different from zero after the DL. These results are in line with trends observed in the administrative data. Birth records provided by the National Institute of Statistics show that the contribution to first births of women who complete secondary education increased from 26 to 30 % in the period from 2004 to 2008. The trend is the same when births from women who are 19 years-old or younger are excluded from the analysis.

While there is a large literature on the effects of no-fault divorce in the United States in the 1970s, there is no consensus on the effects of introducing divorce for the first time on fertility decisions. In this sense, research on the effect of the DL on women’s fertility decisions is needed. The estimates we obtain suggest that the DL increased the hazard of having a first child for more educated women because their higher potential income loss while caring for the child is now offset by alimony if the marriage fails. It did not significantly change the hazard for less educated women.