Intimate partner violence (IPV) is a significant public health problem, being the most common form of violence experienced by women and imposing adverse consequences for the health of the victims and their children (World Health Organization, 2013). According to the National Intimate Partner and Sexual Violence Survey (NISVS), 6.6 percent of women in the United States report experiencing IPV in the past 12 months, reaching a lifetime prevalence of 37 percent. An important risk factor associated with IPV perpetration is substance abuse, which can trigger aggressive behavior and worsen impulse control problems (Castilla and Murphy, 2022, Chalfin et al., 2021, Angelucci and Heath, 2020). With the U.S. facing an epidemic of opioid overdose, public health experts raised concerns about the role that opioid misuse plays in facilitating IPV (Warshaw et al., 2014, Packard and Warshaw, 2018).1 While increasing trends in opioid misuse are causing a serious public health crisis across the U.S., their consequences for IPV have not been explored systematically.

This paper examines the effects of opioid misuse on IPV by studying the reformulation of the main legal opiate—OxyContin—into an abuse-deterrent form in 2010, a major supply-side intervention implemented in the U.S. to curb excessive prescription of opioids and reduce their addictive potential. We provide the first study on the spillover effects of the OxyContin reformulation on domestic violence by intimate partners, and inform how a supply-side shock that disrupted access for one particular, albeit important, segment of the opioid market, generated downstream impacts on interpersonal violence and women’s well-being. We combine IPV data from the National Incident Based Reporting System (NIBRS) from 2006 to 2019, which includes incident-based reports to law enforcement agencies, with county-level opioid prescriptions prior to 2010, the year in which OxyContin was reformulated. We capitalize on the baseline spatial variation in treatment exposure within a difference-in-differences (DID) framework to examine whether areas that were more exposed to prescription opioids prior to reformulation experienced differential changes in IPV outcomes after the reformulation. The NIBRS data also enable us to identify incidents in which the perpetrator was suspected of using heroin. We use this information to construct a county-level measure of heroin-involved IPV: the number of IPV incidents where law enforcement reported suspected heroin use by the perpetrator, scaled per 1000 residents in the county.

We find that the reformulation of OxyContin into an abuse-deterrent form led to a significant relative decline in the rate of IPV experienced by women in counties with greater exposure to prescription opioids prior to the reformulation. We show that these declines occur after the policy change, and they are driven primarily by non-Hispanic Whites. The coefficient estimates imply that a one standard deviation increase in pre-reformulation exposure yields a relative decrease of 7.5 percent annually in the IPV rate following OxyContin’s reformulation. We also document corollary declines in injuries and arrests related to IPV (7.3 percent and 6.1 percent, respectively), indicating that the effects are reflective of an actual decline in the incidence of IPV rather than a shift in reporting behaviors. The overall decline in IPV, however, masks a significant uptick in IPV incidents where the perpetrator was suspected of using heroin, particularly in more urban areas. These findings highlight the importance of identifying populations at a higher risk of substitution to illicit opioids post-reformulation and mitigating this risk with evidence-based policies.

Exploring potential channels, we document that the reformulation reduced the IPV rate primarily in states with less-developed illicit opioid markets, where options for substitution towards illicit drugs were much more limited ex ante. In states with larger and more developed illicit opioid markets, we find no evidence of a decline in the IPV rate; conversely, these locations which would offer greater substitution possibilities towards illicit drugs actually experienced an increased rate of heroin-involved IPV following the reformulation. Moreover, if the primary mechanism for reducing IPV prevalence is the decline in prescription opioid misuse, we would expect to see larger reductions in IPV among demographic groups and locations that initially had higher rates of prescription opioid misuse and thus benefited more from the reformulation. Our pattern of results is consistent with this mechanism: sub-populations (non-Hispanic Whites; younger adults) and localities (lower-educated; high-poverty), which experienced higher rates of opioid prescribing and misuse at baseline, accrued the largest benefits in terms of lower IPV rates. Finally, to the extent that the OxyContin reformulation resulted in a decline in employment and labor force participation rates of both men and women at similar rates (Cho et al., 2021, Harris et al., 2020, Aliprantis et al., 2023), we would expect this channel to increase IPV risk through worsened financial distress at the household level. Our baseline specification adds controls for local area unemployment and labor force participation rates, which only marginally impacts our estimates. Hence, we largely rule out shifts in labor market outcomes as a key channel underlying our main results.

We make several contributions to the literature. First, despite the well-known associations, most of the previous studies that document the relationship between opioid misuse and IPV are based on small sample sizes and fail to account for selection bias and reverse causality (Hughes et al., 2019, Stone and Rothman, 2019, Pryor et al., 2021). Our empirical setup allows us to estimate the effects of an exogenous supply-side intervention targeting opioid misuse on the risk of IPV victimization.

Second, our study contributes to the literature on the broader repercussions of the opioid crisis on families. Gihleb et al. (2022) find that must-access Prescription Drug Monitoring Programs (PDMPs), a similar supply-side shock that constrained access to prescription opioids for misuse purposes, reduced entry into foster care. Emerging evidence (Dave et al., 2025, Barbos and Sun, 2021) linking must-access PDMPs to IPV points to results consistent with ours, that controlling the supply of Rx opioids (albeit via a different lever and margin, by targeting prescribers) has led to a net decline in women’s exposure to domestic violence. A closely related paper to ours is Evans et al. (2022), which finds that counties with greater initial rates of prescription opioid usage experienced an increase in child maltreatment after OxyContin reformulation. While IPV and child maltreatment are both different forms of family abuse, empirically the rates of IPV and child maltreatment within a county are almost orthogonal to each other.2 Moreover, a comparison of our NIBRS sample to the sample in Evans et al. (2022) reveals that the counties in our sample are more rural and have smaller illicit markets (Appendix Table A1). These differences in sample composition may help explain the contrasting findings for different forms of family abuse between the two studies. Specifically, the relatively more rural composition of our sample, characterized by a smaller illicit drug market, could result in lower rates of substitution into heroin or other illicit opioids, thereby attenuating the reformulation’s adverse spillovers on family abuse. In contrast, the more urban sample in Evans et al. (2022), with greater access to illicit markets, may have experienced higher substitution, thereby amplifying the negative impacts on child maltreatment. Moreover, consistent with their more urban profile, the counties in Evans et al. (2022) contain a younger population, who are more likely to have children in the household. These differences in sample composition may explain the higher estimated risk of child abuse and neglect in the Evans et al. study, and also underscore the role played by the potential for substitution from prescription opioid misuse to illicit opioids in driving any beneficial vs. adverse impacts, on the net, of supply-side interventions.

Our paper also relates to the broader health and crime literature that has documented adverse consequences of OxyContin reformulation on infectious diseases (Beheshti, 2019), food insecurity (Heflin and Sun, 2022), child welfare utilization and caregiving patterns (Mackenzie-Liu, 2021, Laurito, 2024), and suicide and homicide rates (Powell, 2023, Park, 2025). In contrast to these studies, which focus on downstream health and socioeconomic harms, we provide the first evidence on the spillover effects of OxyContin reformulation on violent behaviors from intimate partners, highlighting interpersonal violence as an important but previously overlooked dimension of the reformulation’s consequences.

Finally, we contribute to the growing literature on factors affecting IPV prevalence, ranging from the effects of cash transfers (Bobonis et al., 2013), labor market shocks (Aizer, 2010, Erten and Keskin, 2021a, Sanin, 2021), education (Erten and Keskin, 2018), divorce laws (Stevenson and Wolfers, 2006, García-Ramos, 2021), women’s political participation (Stern and Erten, 2024, Anukriti et al., 2025), COVID-19 public health crisis (Erten et al., 2022, Bullinger et al., 2021), and trade shocks (Erten and Keskin, 2021b). Evidence on the effects of substance use on IPV is rare, and focuses on alcohol use (Castilla et al., 2022, Markowitz, 2000). Using a randomized control trial in rural Kenya, Castilla et al. (2022) find that the reduction in alcohol use lowers sexual violence. A related study recently found that while higher temperatures increase IPV risk, the OxyContin reformulation mitigates this effect by reducing the complementary use of other substances, such as alcohol, on hot days (Pavanello and Zappalà, 2024).

This paper is organized as follows. Section 2 provides a brief description of the OxyContin reformulation and the data used for the analysis. Section 3 presents the identification strategy and the empirical results. Section 4 concludes with a discussion of our findings.