Please note: Two additional briefs related to the conference are published on the FREE policy briefs website – a brief on gender-based violence in conflict based on the panel discussion, and another sharing preliminary results from the recent FROGEE survey.

Introduction

While the concerns about domestic violence (DV) and intimate partner violence (IPV) have been gaining prominence since the start of the COVID-19 pandemic, they were further exacerbated by the devastating events happening in Ukraine. Times of crisis or conflict makes the issue more severe, however, gender-based violence is sadly prevalent at normal times too, and a major portion of it is DV and IPV. Limiting violence towards women requires understanding the determinants of DV and IPV and the channels through which they take effect. With this in mind, the Stockholm Institute of Transition Economics (SITE) together with the Centre for Economic Analysis (CenEA) and the FREE Network invited researchers to present their work relating to the economic and social context of domestic violence. This brief provides an account of what was shared at the conference.

Prevention of domestic violence: What works and what doesn’t?

Three presented studies geared toward evaluating policies aimed to limit violence against women.

Dick Durevall shared his findings on IPV and national policy programs in Colombia, focusing on the laws and policies implemented based on the UN campaign “UNiTE to End Violence Against Women” between 2010 and 2015. To evaluate the effect of these policies, he adopts a differences-in-differences design and compares provinces that had a gender policy before this renewed effort with those that did not. This builds on the idea that provinces that had an IPV policy strategy before UN recommendations were adopted are more efficient in implementing new such policies. It is found that self-reported physical violence falls from 20% to 16% between 2010 and 2015 in provinces that had IPV policies while this number remained at 18% in those that did not. While sexual violence decreased in both groups, provinces with IPV policies experienced a stronger reduction.

Accurate reporting is a key issue when it comes to IPV since it makes up the foundation for designing effective policy. Due to long-lasting and tiresome judicial procedures, threats, social barriers, or emotional costs, victims might choose not to report. Looking at the introduction of specialized IPV courts in Spain, Marta Martínez-Matute presented her paper on how institutions shape reporting. Bestowed with specialized staff, victim-oriented resources, and a swifter judicial process, these courts are specifically designed to deal with IPV cases. Martínez-Matute and co-author investigate if these resources make women more prone to report IPV by exploiting the sequential rollout of specialized courts. They use yearly court-level data on individual IPV cases between 2005 and 2018 in a staggered difference-in-differences framework with matched control districts. The results show that the introduction of an IPV court in a judicial district reduces the length of the judiciary process by 61% and increases the reported number of IPV cases by 22%. Ensuring that this increase is not fully driven by a rise in false reports, it is found that the share of dismissed IPV cases remains unchanged. Further, it is shown that the increase is driven by less severe IPV cases and not aggravated IPV offenses or homicides.

A distinctive feature of DV crimes is that there is a high degree of recidivism, with many women experiencing repeated violence from the same partner. However, little is known about how police should respond to such crimes to ensure safety to those victimized. From one perspective police arrests deter repeated DV crimes since they incapacitate perpetrators and allow police to investigate while offering safety to victims. However, some argue that this safety is merely temporary and that DV arrests might trigger offenders to retaliate against victims, leading to increased long-term DV. Against this reasoning, Victoria Endl-Geyer presented a study on the relationship between police arrests and DV dynamics in the UK. It uses highly granular administrative data on the population of DV incidents in the West Midlands which allows the researchers to observe the detailed information on the incidents’ timing and location as well as on police officers and their crime scene responses. It adopts an instrumental variables approach using the dispatch team’s previous propensity to arrest (measured as the weighted average arrest rate of officers in the team) as an instrument. The results provide evidence consistent with a deterrence effect. While regular OLS estimates show an insignificant impact, the IV results indicate that an on-scene arrest decreases repeat DV incidents by 25-26 percentage points. They find that the effect is the same when restricting the sample to incidents reported by a third party, supporting that this effect is not driven by a change in reporting behavior.

Factors of domestic violence and its mechanisms

Other studies presented at the conference focused less on policy assessment and more on identifying the determinants of IPV and DV.

Losing or obtaining a job causes a shock in the intra-relationship dynamics and changes the economic power balance between spouses. Deniz Sanin presented her paper on the DV effect of women’s employment in the context of Rwanda. Following the government-initiated National Coffee Strategy in 2002, the number of coffee mills in Rwanda increased from 5 to 213 over the course of ten years. This natural experiment allows studying the effect of having a paid job as it captures the shift from unpaid labor on a family farm to paid work on a mill, keeping job-related skills constant. Using survey data on both DV and labor market outcomes along with administrative data on DV hospitalizations, the study adopts a staggered difference-in-differences strategy and compares women before and after mill opening as well as within and outside of the catchment area (a buffer zone surrounding the mill). The results show that upon mill opening, the probability of working for cash increases and that of self-reporting domestic violence in the past 12 months decreases by 26% (relative to the baseline of 0.35). During the harvest months, the only period of the year in which the mills operate, hospitals are significantly less likely to admit DV patients compared to the month before the harvest season, suggesting that the initial results are not driven by reporting bias. Looking at the mechanisms, she finds evidence supporting an increased bargaining power explanation – women in catchment areas who are exposed to mill opening are more likely to have a bigger say in household decisions such as larger household purchases and contraception usage. Increases in husbands’ earnings and decreased exposure are also ruled out as possible channels since a decline in DV is also found among spouses where the husband works in a different occupation with no change in earnings.

Rather than studying the impact of women’s employment status, Cristina Clerici shared a related paper that focuses on male unemployment. To investigate its effect on IPV, the study exploits the exogenous shock to employment caused by COVID-19 containment measures in Uganda. The authors collect individual-level data via phone surveys on the incidence of IPV among food vendors, including information on husbands’ sector of employment. To identify a causal DV effect of male employment exit, the authors distinguish between two groups of women with similar pre-lockdown experiences of abuse: those with spouses employed in sectors where operations were halted by COVID-19 lockdowns (construction workers, taxi drivers, etc.) and those with spouses who were unaffected (food vendors, farmers, etc.). The results show that male unemployment increases the probability of experiencing physical violence by 4.9 percentage points, corresponding to a 45% increase relative to the average likelihood. The effect cannot be explained by increased exposure (the man being more at home) – affected and unaffected women spend on average an equal number of nights in the market, which could be used as a coping mechanism. This suggests it is the change in unemployment status itself that drives the increase in DV.

While most of the literature on domestic abuse has documented that its drivers often come from changing life conditions of the victim or perpetrator, there is broad anecdotal evidence that exogenous events can lead to exacerbations in domestic violence as well. Ria Ivandic presented her paper that documents a causal link between major football games and domestic violence in England. The authors use a dataset on the universe of calls and crimes in the Greater Manchester area. The data provides a time series on the incidence of different types of domestic abuse with information on the timing, relationship to the accused, and individual characteristics of the victim and perpetrator, including whether the perpetrator was under the influence of alcohol at the time of the incident. They adopt an event study approach focusing on the hours surrounding a game and document a substitution effect in that the two-hour duration of a football game is associated with a 5% decline in DV incidents. However, following the game, the initial decrease is offset as DV incidents start increasing and culminate after 10-12 hours, eventually leading to an aggregate positive effect which constitutes a 2.8% hourly increase on days when games are played.

The authors argue that alcohol consumption, rather than emotions, is the main mechanism through which domestic violence is affected by sporting events. Supporting this hypothesis, they first find that the outcome of the game or the associated element of surprise (measured using the ex-ante probability of winning a game through betting markets) does not affect the probability of DV occurring. Second, they show that the increase in DV following a game is solely driven by an increase in alcohol-related DV incidents, while those committed by non-alcoholized men remain constant. Further strengthening this finding, it is shown that for games scheduled early in the day, when perpetrators can start drinking sooner and continue throughout the day, they find a significant increase in DV incidents committed by alcoholized perpetrators while this is not the case for late-scheduled games.

The role of women’s empowerment

In the literature on gender-based violence, there is a common disposition to think about women’s empowerment as a central element of DV mitigation. However, theories point in opposite directions making the effect of women’s economic empowerment rather unclear. On one end of the spectrum, there are bargaining theories indicating that an increase in women’s employment opportunities or income should have a negative effect on DV by creating outside options or increasing the bargaining power in a relationship. At the other end, there are backslash theories arguing that enhancing women’s financial empowerment may further exacerbate violence by undermining the role of the breadwinner, triggering male partners to retaliate with the use of violence in order to restore the power balance. Going in the same direction, theories of instrumental violence point towards that the male partner might also use violence to extract resources.

In her keynote lecture, Bilge Erten outlined the evidence relating to DV and women’s empowerment and discussed to what extent and in which contexts these theories are supported.

The evidence of a positive or negative effect of empowerment may depend on which aspect of it is studied. Education is seen as an important one because it has the potential to raise women’s self-awareness of IPV, increase the likelihood of matching with a well-educated partner (which is negatively correlated to abusive behavior), and improve labor market outcomes. Although evidence is scarce in this area, Erten shared her own findings on the causal effect of education reform on IPV in Turkey. In line with instrumental violence theories, it is found that, while women in cohorts affected by the reform performed better in the labor market, they experienced more psychological violence and financial control behavior, and there was no sign of an effect on DV attitudes, partner-match quality or marriage decisions.

What we know about women’s empowerment and DV is also different across countries. When it comes to the effect of employment, findings from developed countries are generally consistent with bargaining theory explanations while what is found in the developing world is more mixed. This is also the case for studies on unilateral divorce laws – while a negative effect on IPV has been documented in the United States, a positive effect of these laws is found in Mexico.

Assessing the literature on the income effect leads to a somewhat ambiguous verdict too. Although generally, most studies confirm that overall violence declines with women’s income, there is often heterogeneity in the effect. It has for instance been found that the sign of the income effect from cash transfers on DV changes from negative to positive as the size of the transfer increases.

Finally, Erten provided some important policy considerations. There is evidently a widespread backlash problem that can arise after a policy intervention of the types discussed above. Policymakers need to think more about monitoring and protecting victims from more violence when implementing such a policy. Further research assessing post-intervention is also needed to identify interventions that are the most effective in minimizing domestic violence. In particular, a change in broad social norms around gender roles should be a desirable outcome, to the effect that a new, improved status of women in society and in the household becomes more culturally acceptable and needs not lead to backlash. In the case of expressive violence (that is not a rational, calculated response but rather a compulsion in the heat of the moment), mental health interventions should also be considered.

Concluding remarks

As highlighted by the 2022 FROGEE conference, domestic violence not only has been put in the spotlight following the pandemic or the ongoing conflict in Ukraine, but is widespread across the globe in regular conditions too. The mixed findings shared at the conference suggested that policies limiting gender-based violence should be designed with respect to the cultural and social setting where they are to be implemented as the heterogeneity is very high across contexts. Although research has come a long way, the conference stressed that there is much more to be done, in terms of not only knowledge but also the political will and commitment to seriously address the issue of gender-based violence.

The presentations held at the conference can be viewed at this link and a separate policy brief based on the panel discussion on gender-based violence in times of conflict can be found here.

Disclaimer: Opinions expressed in events, policy briefs, working papers and other publications are those of the authors and/or speakers; they do not necessarily reflect those of SITE, the FREE Network and its research institutes.

Photo by Anete Lusina, Pexels.com