War, Abduction, and Bilateral Intimate Partner Violence: Evidence from Northern Uganda

The study involves 438 individuals (219 couples) between the ages of 19 and 55 , recruited with the help of local leaders from 3 parishes and 22 villages in the Kitgum district of Northern Uganda. Local leaders completed a full roster of couples in their village in September of 2024 where women were to be in a given age range. This gave a list of 348 couples from which 220 were selected for inclusion at random stratifying by the couple’s combined abduction statuses. 219 couples were found and eligible.

30 sessions comprised of 8 couples (16 individuals) were held across 3 parishes. Couples were randomly placed into a session with individuals from within their parish to ensure no one will travel a prohibitive distance. However, the parish level randomization ensure that individuals are less likely to know others in the group. Sessions were conducted at well-known gathering locations in each parish, one in the morning and one in the afternoon. Couples were identified and selection stratified based on 4 marriage types:

1. Neither FA: 50 couples where neither the man nor the woman was abducted.
2. Woman Only FA: 49 couples (50 if we can recover or redo one husband interview that was lost in CTO transfers) where the woman
was abducted, but the man was never abducted.
3. Man Only FA: 50 couples where the man was abducted, but the woman was never abducted.
4. Both FA: 70 couples where both individuals had some experience of abduction

This document pre-specifies the analysis of the data generated by the random selection of individuals to participate in survey and field games in Northern Uganda. This was written after data collection, but before any cleaning or analysis took place.

2024-09-16

2024-10-21

1. Relationship Formation:
o Partner Status: Abduction status of partner, war exposure of partner
o Partner selection (used to parameterize matching the model): Autonomy, age of first marriage, cowives, bride price paid, familial wealth and status
o Previous relationships

2. Household Violence:
o Violence towards spouse: Verbal, physical, physical intensity index
o Violence received from spouse: Verbal, physical, physical intensity index

1. Violence towards and from cowives: Verbal, physical
2. Violence towards children

Mechanism Analysis:
3. Psycho-Social: Cohen Stress Scale, PHQ-9 Depression Scale; Brief Self-Control Scale, CAGE Scale for alcohol dependency
4. Economic and financial outcomes: Education, job type, income, home and goods ownership, electricity and facilities
5. Behavior and belief elicitations: Patience, social norms
6. Game Decisions: Generosity, anti-social (taking) behavior, negative reciprocity

This project is divided into two parts. The applied portion of the paper leverages the random abduction of children as a source of variation to explore long-run relationship outcomes in their adult lives including intimate partner violence (IPV). This portion relies on novel survey data collected from randomly selected cohabitating couples.

The second portion goes further to assess potential mechanisms that may explain the relationship between abduction and household violence. Here I use lab-in-the-field games along with survey measures. The first game is a variation of the dictator game similar to Bauer et al. (2014) where individuals are asked to make five separate decisions about how much they would like to give to people from 5 categories: randomly selected Formerly Abducted Woman (WA), Never Abducted Woman (WNA), Formerly Abducted Man (MA), Never Abducted Man (MNA), and their spouse. These rounds are played in varied order and, if this round is selected for payment, recipients do not know who sent them to guard against potential spousal retaliation.

Individuals then play a strategy method variation of the "Power to Take" game (Bosman & van Winden, 2002). Players are paired asked to make choices about how much they would like to take from their opponent (0, small portion, large portion, or all) knowing that their opponent could choose to destroy some portion of their own endowment depending on how much is taken. They also make choices as the opponent being asked if they want to destroy none, half, or all of their endowment for each possible portion the opponent could have taken. Each participant plays both roles against 2 opponents: a random person from the other room (someone on the same session of the opposite gender) and their spouse. This allows for a deeper understanding of anti-social behavior and retaliation. The strategy method again protects participants as no one knows the actual choices of anyone else. This allows for them to make choices regarding their spouse without fear of spousal retaliation at home.

Random selection into the study and further into workshop groups was done in office by a computer. The randomization was stratified by marriage type.
Game ordering was predetermined for each session as individuals were randomly placed into sessions.
Payment was randomly determined by drawing numbered paper slips in the field.

Male/Female cohabitating couples were randomly selected into the study and then randomly placed into sessions based on their parish.

No

219 couples

438 individuals

1. Neither FA: 50 couples where neither the man nor the woman was abducted.
2. Woman Only FA: 49 couples (50 if we can recover or redo one husband interview that was lost in CTO transfers) where the woman
was abducted, but the man was never abducted.
3. Man Only FA: 50 couples where the man was abducted, but the woman was never abducted.
4. Both FA: 70 couples where both individuals had some experience of abduction