Silhouettes of quarreling parents and child
VoxEU Column Health Economics

Domestic violence and the mental health and wellbeing of victims and their children

Nearly a third of women worldwide report some form of physical or sexual violence by a partner in their lifetime, yet little is known about the long-term mental health effects on victims or their children. This column studies the costs associated with domestic violence in Norway, where data allow offenders to be linked to victims and their children over time. Domestic violence incidents documented by police are associated with marital dissolution, decreased financial resources, and lower test scores by children. These effects taper off over time for victims, but not for their children.

Domestic violence and poor mental health are twin problems confronting countries across the globe. Regarding domestic violence, the World Health Organization has compiled surveys from around the world, finding that almost one third of women report some form of physical or sexual violence by a partner in their lifetime. While slightly lower in high income countries, lifetime rates of intimate partner violence are still alarmingly high at 25% (WHO 2021).

Mental health issues are also widespread and have become more salient in recent years. New research following the COVID-19 outbreak from different countries reveals increases in mental health problems after lockdowns (e.g. Yamamura and Tsutsui 2020 for Japan, Hvide and Johnsen 2020 for Norway) and widening gender gaps in mental health (Rauh et al. 2021 for the US). Understanding the sources and consequences of increased mental health issues has become a key policy challenge in the aftermath of the pandemic. Mental health difficulties are also an important dimension to study because they pose an obstacle to educational and labour market success (e.g. Currie et al. 2010, Moser et al. 2021).

Challenges to understanding the link between domestic violence and mental health

In recent research, we explore the connection between domestic violence and the mental health of victims and their children (Bhuller et al. 2022). Figuring out the health and wellbeing consequences of domestic violence is a difficult question due to two main challenges. First, one would need data on domestic violence incidents, data on mental health, and other wellbeing outcomes for victims and their children over time, as well as the ability to link the two types of datasets together. Second, there is the problem of selection bias, since domestic violence is not random: families reporting domestic violence may have had worse mental health and wellbeing even in the absence of domestic violence.

We make progress on these challenges by using rich Norwegian data and an event study approach. Our work draws on several strengths of Norwegian registry data. We construct a panel dataset containing complete administrative records of all police reports related to domestic violence over a 22-year period. We are able to link this data to a rich set of outcomes in both the short and medium run, including data on mental health for victims and children, labour market outcomes for victims, and schooling outcomes for children. Importantly, we can match victims to their children.

To solve the selection bias problem, our primary analysis compares outcomes before and after a police report, using a well-defined control group to construct a counterfactual. Specifically, we use families who will report a domestic violence event in the future as a control group for families who report a domestic violence event today. This characterisation of the control group recognizes the fact that families who never report a domestic violence incident are unlikely to have similar unobservable characteristics as families who do. Instead, the design leverages differential timing of a domestic violence report to estimate effects.

Interpreting our treatment (i.e. the reporting event) and the resulting treatment effect is not straightforward for several reasons. First, treatment could reflect violence which was either more severe or different (e.g. someone saw it) and hence resulted in a police report. Second, the report itself could trigger access to shelters, social support programmes, or child services. Our estimates will capture both of these effects. Third, not all domestic violence is reported to the police, and we are unable to study the effects of unreported domestic violence. Relatedly, reported domestic violence may reflect a longer build-up of violence and conflict within the household. If this is the case, the treatment effects we estimate will capture the incremental increase in violence at the time of a report plus the reporting effect. However, if this were the case, we would also expect to see differential pre-trends in outcomes, which we generally do not.

Domestic violence has negative consequences

The overarching theme of our findings is that there are large and persistent negative effects of a domestic violence episode in Norway, both for the victim and their children. There are worse mental health outcomes for victims and their children, as well as worse labour market outcomes for victims and educational outcomes for children. We explore each of these margins in more detail below. These negative effects exist despite well-funded social support programmes for victims and their families in Norway, suggesting that domestic violence events have negative consequences that are difficult to fully ameliorate afterwards. Hence, policies or changes in societal norms which could prevent domestic violence in the first place could have a high social return.

Changes in the home environment

Our first finding is that the domestic violence event has a large effect on the home environment. The event triggers a large increase in divorce (50%), an even larger decrease in new marriages (-79%), and an increase in moving (3%). These percentages are the average effects for one to five years after the event. There is also a drop in financial resources available to victims and their children; per capita household consumption expenditure drops by 9%. Most of this is due to a drop in spousal income, which is not made up for by higher transfer payments (which include government programmes and private alimony/child support payments).

To illustrate the event study approach we use, consider plots for the effects on divorce and household consumption over time in Figure 1. The first panel shows the likelihood of being married, given an individual married in the year 2000 (long before either treatment or control households have been victimised). The dots are the differences in marriage rates for those who report a domestic violence incident in a year (normalised at zero in the graph), compared to those who will experience a domestic violence incident at least six years later.

Panel (a) shows that prior to the domestic violence event, there is not a noticeable difference in the probability of remaining married between treatment and control women. However, in the year of the event, marriage rates fall by roughly 7 percentage points for victimised women versus the controls. There is an even larger 39 percentage point reduction in the ensuing years, which translates to a 50% reduction in remaining married relative to the mean. Panel (b) shows a similar pattern over time for per capita consumption, which falls by almost 30,000 Norwegian Kroner, or 9% relative to the mean.

Figure 1 Event study graphs for divorce and consumption

Figure 1 Event study graphs for divorce and consumption

Note: Per capita consumption expenditure is measured in thousands of Norwegian Kroner and uses EU weights for adults versus children. Vertical lines denote 90% confidence intervals.


Mental health of victims and their children

We next turn to mental health outcomes. As seen in Figure 2, for victims there is a 15 percentage point increase in receiving at least one mental health diagnosis in the year of the event, with a tapering off after three years. This translates to a 35% increase relative to the mean in the year of the event. Similar analyses reveal that this mental health effect is driven by mood, depression, sleep, and anxiety disorders. For children, there is an immediate increase in mental health diagnoses, with a sustained average increase of 2.6 percentage points, or 15% relative to the mean, in the four years after the event. This statistically significant average increase for children is driven by a similar set of diagnoses as for victims.


Figure 2 Event study graphs for victim and child mental health, defined as having at least one mental health-related doctor visit in a year

Figure 2 Event study graphs for victim and child mental health, defined as having at least one mental health-related doctor visit in a year

Note: Vertical lines denote 90% confidence intervals.

Other outcomes for victims and their children

Other outcomes for victims and their children also point to a decline in wellbeing. Victims experience more doctor visits, lower employment, reduced earnings, and greater use of disability insurance. For children, reporting increases the use of child protective services (such as counselling and home assistance programmes) and the likelihood that they will be placed in foster care.

One important academic outcome for children is performance on a consequential nationwide exam used for admission to high school. Students take this exam in the last year of junior high school, when they are age 15 or 16. To study this outcome, we use a different research approach, as the outcome is observed only once. In this case, we take advantage of the timing of the domestic violence report, leveraging whether it occurs just before or just after students take the exam. We employ what economists call a regression discontinuity design, which compares outcomes immediately before and after the test date. Figure 3 illustrates this approach. The x-axis is the test date minus the date of the domestic violence reporting event, and the y-axis is a student’s exam score (normalised to be mean zero and standard deviation one). Children to the left of the vertical line in the graph take the exam before the domestic violence incident, while those to the right take the exam afterwards. The figure reveals that a domestic violence reporting event reduces exam scores by 8% of a standard deviation. A similar analysis reveals that a domestic violence event also lowers the chances of completing the first year of high school on time by 8%.

Figure 3 Regression discontinuity graph for child exam scores

Figure 3 Regression discontinuity graph for child exam scores

Note: Dots are average scores in 6-month bins, dotted lines are the RD regression lines, and solid lines are 90% confidence intervals.

To conclude, our analyses find large costs of a domestic violence reporting event for victims and their children. Standard regression analyses (i.e. ordinary least squares) yield substantially different estimates, highlighting the difficulty of accounting for selection bias into victimisation and motivating the use of our alternative approaches to estimate causal effects.

Authors’ note: This column is based on prior research by the authors (Bhuller et al. 2022), which received generous financial support from the European Research Council.


Adams-Prassl, A, T Boneva, M Golin and C Rauh (2021), “Lockdowns widen the gender gap in mental health”,, 27 April.

Biasi, B, M S Dahl and P Moser (2021), “Career effects of mental health”, NBER Working Paper No. w29031.

Hvide, H K and J Johnsen (2021), “COVID-19 and mental health: a longitudinal population Study through 2020”,, 7 July.

Tsutsui, Y and E Yamamura (2020), “COVID-19, mental health, and domestic violence: Evidence from Japan”,, 22 June.

World Health Organization (2021), Global and regional estimates of violence against women: prevalence and health effects of intimate partner violence and non-partner sexual violence.

Bhuller, M, G B Dahl, K V Løken and M Mogstad (2022), “Domestic Violence and the Mental Health and Wellbeing of Victims and Their Children”, NBER Working Paper No. 20792.

Currie, J, M Stabile, P Manivong and L L Roos (2010), “Child Health and Young Adult Outcomes”, The Journal of Human Resources 45(3): 517–548