When Layoffs Hit, Domestic Violence Follows
Losing one's job leads to a sharp and persistent increase in the risk of violence against women, regardless of who in the family is left unemployed. After a mass layoff, the likelihood of a man committing violence against his partner increases by 32%, while the likelihood of a woman being a victim increases by as much as 56%. This is a vertical jump recorded in millions of administrative data in Brazil, cross-referenced between courts, hospitals, and employment records from 2009 to 2018. And these are not one-off incidents: the escalation lasts for years. These figures clearly show how employment crises creep into homes and change family life.
A unique dataset and a causal relationship
In their study, Sonia Bhalotra (University of Warwick), Diogo G. C. Britto (Bocconi), Paolo Pinotti (Bocconi), and Breno Sampaio (Universidade Federal de Pernambuco) – draw on a rare wealth of information: all reports of domestic violence, all requests for protection, all hospitalizations and health notifications for violence, and all formal employment relationships in the country. It is an ideal condition for isolating the pure effect of job loss.
“The effects are strong, immediate, and persistent,” explains Diogo Britto, director of CLEAN (Crime: Law and Economic Analysis), a research unit at Bocconi's Baffi Center. “The likelihood of domestic violence increases when either the man or the woman loses their job. And this goes against the main economic theories on the subject.”
Plummeting income and forced exposure at home
The combination of the two effects is crucial. Job loss causes sharp and lasting drops in income—as much as 70% in the first year—and increases the time partners spend together at home, in a most stressful circumstance. “The reality that emerges from the data is simpler and harsher: what matters is the income shock and the forced time spent together,” adds Britto. The main theories—from bargaining to backlash—fail to explain the symmetrical effects of male and female job loss found in the study.
Unemployment benefits are not enough
The research also highlights the role of protections. But with a surprising result: unemployment benefits do not reduce domestic violence and may increase it after handouts end. By employing an administrative threshold that determines access to the new benefit, the authors produce one of the few truly causal analyses available on the subject. “During the months when the benefit is received, the effects cancel each other out,” explains Paolo Pinotti, professor of economics and Dean of the Faculty at Bocconi. “But when the benefit ends, the income effect disappears while the exposure effect remains, because those receiving the benefit return to work later. And that’s when violence increases.”
Severance pay works if it is large enough
The paper shows that when workers receive high severance pay—for example, thanks to seniority, which in Brazil can amount to seven months’ salary—the effect of job loss on violence virtually disappears. The available income at the time of the employment shock is therefore a real, not theoretical, shock absorber.
What politicians should do
The message is clear: protections must be designed not only to compensate for income, but also to reduce periods of inactivity. It is not enough to support consumption in the short term if this extends the period spent staying at home without work. We need benefits linked to active policies, mandatory training, incentives for re-employment, monitoring of the actual duration of unemployment, and constant coordination with anti-violence services.
The structural issue: violence is sensitive to economic shocks
The work of Bhalotra, Britto, Pinotti, and Sampaio shows that domestic violence is deeply sensitive to economic shocks, and that ignoring this link means failing to see an essential part of the problem. In times of increasing job instability, labor policies and anti-violence policies can no longer exist separately: the data show that, for many women, they are two sides of the same reality.
