This is our 11th blog post for our Job Market Paper Series blog for 2024-2025.
Gabriela Deschamps is a PhD candidate at the London School of Economics. She is a development economist with interests in gender inequality and family economics. You can find her JMP here.
Intimate Partner Violence (IPV) against women is a major public health concern, with 27% of women worldwide having experienced physical or sexual violence by an intimate partner (WHO, 2021). The consequences extend beyond the victims, as children raised in violent households often face long-term adverse effects (Aizer, 2011; Wolfe et al., 2003). The birth of the first child creates big changes in household dynamics, including greater economic dependence of women on their partners (Kleven et al., 2023). Therefore, understanding the relationship between motherhood and IPV is crucial. Recent studies suggest mothers are at higher risk of experiencing IPV than non-mothers (e.g Britto et al., 2024; Massenkoff et al., 2024), but the mechanisms behind this remain largely unexplained.
In my job market paper, I propose two alternative explanations. The first is that having a baby causes changes in men’s violent behavior. The second is that motherhood makes women more likely to stay with a violent partner, leading mothers to be at higher risk of violence even if men’s violent behavior does not change after childbirth. A challenge here is that changes in violent behavior and separation appear observationally similar in most IPV data sources, as they often capture only current IPV. Both a woman who leaves a violent partner and a woman not experiencing violence would report no current violence. To disentangle the mechanisms, I propose a model and an empirical test.
Game-theoretic model
I propose a two-period model where men decide whether to exert violence, and women decide to stay or leave the relationship. Some men are abusive, meaning they derive positive utility from exerting violence. Before the game starts, women do not know whether their partner is abusive, but they can learn through his actions. At the beginning of the first period, some couples exogenously have a child while the rest wait until the second period. I focus on two main cases:
- Direct channel: When the for violence is not too high, only abusive fathers exert violence, while abusive non-fathers wait until they become fathers to exert violence. Fathers are thus more likely to exert violence than non-fathers. Mothers with low dissolution costs will leave; the rest will stay. Consequently, mothers will be more likely to be separated and more likely to experience violence compared to non-mothers.
- Selection channel: When the satisfaction for violence is high, all abusive men exert violence irrespective of whether they are fathers or not. However, mothers will be strictly less likely to leave. Consequently, mothers will be more likely to experience violence compared to non-mothers.
Both channels predict mothers will be more likely to experience violence compared to non-mothers, but they differ in predictions about men’s violent behavior and women’s separation behavior.
From the model to the data
As mentioned, usual IPV data is a proxy for whether women currently experience violence, which is not useful to disentangle these channels. The solution is to use a sample of both partnered and separated women with violence history for current and past partners. This data lets us identify if women who are now separated previously experienced violence with their partner (when they were partnered), even if they are no longer exposed to the violence. I use two waves of ENDIREH (2011 and 2016), a nationally representative survey of household dynamics in Mexico, which includes all civil statuses, fertility history, relationship history, and violence history with current and past partners. In particular, women are asked whether their (ex-)partner exerted violence in the last 12 months and at any point in the relationship. The first is a proxy for current violence and a function of both violence and separation behavior. The second is a proxy for men’s violent behavior as both separated and partnered women with an abusive (ex-)partner will respond positively if their partner exerted violence.
I test the model’s predictions using women who recently began their first cohabiting relationship. Consistent with both channels, I find motherhood is associated with a higher probability of currently experiencing violence. However, mothers and non-mothers are equally likely to have experienced violence with their first partner and are less likely to be separated. These results provide suggestive evidence that motherhood increases IPV risk through the selection channel.
Using pharmacy openings as a shock to fertility
To further test the mechanisms of the model, I use pharmacy openings in Mexico as an instrument for fertility. In Mexico, women typically have children soon after beginning a relationship; the median time between the start of a relationship and the birth of the first child is just two years. Additionally, 45% of Mexican women aged 15–24 did not use contraceptives during their first sexual experience (ENADID, 2015). A significant factor behind this is limited knowledge about where to access contraceptives, cited by 18.8% of respondents. For women who do use contraceptives, pharmacies play a crucial role, particularly for those who are not yet mothers. Nearly 80% of non-mothers who use contraceptives obtain them from pharmacies.
I use multiple survey waves (2006, 2011, 2016), the confidential location of the women, and the universe of Mexican pharmacies to identify those who gained access to a pharmacy between 2011 and 2016. I estimate the effect of a pharmacy opening on fertility, violence, and separation in a difference-in-difference framework, using variation across time and access to pharmacies.
First, I show that a pharmacy opening is effective at reducing the probability of becoming a mother. I then test the three predictions of the model. The first prediction is the effect of motherhood on current violence. I use a binary outcome variable indicating whether the woman experienced physical violence in the last twelve months. Consistent with both channels, a pharmacy opening reduces the probability of currently experiencing physical violence.
The second prediction is the effect of motherhood on men’s violent behavior. I use women’s relationship history and violence history to identify whether the partner from five years before the survey (before the pharmacy opening) was physically violent at any point in the relationship. The direct channel predicts a negative coefficient in this variable since fewer men become violent due to the pharmacy opening. The selection channel predicts a coefficient of zero, as having a child does not change men’s propensity to exert violence. Consistent with the selection channel, I find women in the treatment and control groups are equally likely to have experienced violence with their partner before the pharmacy opening.
Lastly, to test women’s separation behavior, I use an outcome variable of whether a woman had a breakup in the last five years. Under the direct channel, a pharmacy opening would result in a decrease in the probability of breakup, as fewer men become violent, leading to lower separation rates. The selection channel predicts a positive coefficient, as delaying pregnancy allows women time to learn if their partner is violent and leave. Consistent with the selection channel, I find that pharmacy access increases the probability of breakup.
I also show that treatment and control groups followed parallel trends before the pharmacy openings and perform two placebo exercises. First, I use recent mothers prior to the pharmacy opening as a placebo group. These women had access to all services that a pharmacy provides, but the timing of the opening was after they became mothers. For this group, I find no effect on fertility, violence, or separation. To rule out that other unobserved factors correlated with a pharmacy opening affect mothers and non-mothers differently, I estimate the effect of other business openings (e.g., bakery, hardware store) on non-mothers. Only a pharmacy opening, not other business openings, has the expected effects on fertility, violence, and separation.
Implications for Policy
I find that motherhood increases the risk of experiencing violence mainly through the selection channel: men are violent regardless of fatherhood, but motherhood increases separation costs for women. My results suggest that policies which strengthen women’s control over their fertility can reduce the incidence of IPV by enabling them to leave violent relationships before they are locked in by motherhood.
Feature image shared by the author from unsplash.com
