Thank God, It's a Boy?! The Heterogeneous Effects of Children's Gender on Domestic Violence - An Application of the Causal Forest Algorithm

Sammanfattning: This thesis aims to quantify heterogeneity in treatment effects of the gender of the first child on the mother's prob- ability of experiencing domestic violence. The identification strategy takes into account the potential endogeneity of gender preferences on the children's gender, as well as potential channels of a child's gender on domestic vio- lence. I use a large-scale data set from Colombia with close to 50,000 observations and rely on the causal forest algorithm introduced in Wager and Athey (2018) to quantify heterogeneous treatment effects. I identify mainly two different subgroups of women for which the size of the treatment effect differs around six percentage points. Women who are less likely to experience domestic violence when having a first-born son compared to a daughter are on average younger, have a larger age gap to their partner and are more likely to live in rural areas. Women who are more likely to experience domestic violence when having a first-born son compared to a daughter are on average older, are living in cities and more likely part of the upper social classes. The robustness checks provide evidence on, among others, reporting, attrition and sample bias.