Survey links eviction to elevated violence reports, leaving causal direction unresolved

Summary

The 2025 USVEX survey documents a seven-fold association between recent eviction and recent physical violence, but its cross-sectional design leaves the causal direction unresolved, admitting reverse causation, unobserved common causes, and selection effects as competing accounts of the same observational pattern.
Tulane’s Anita Raj and UC San Diego’s Jakana Thomas identify two distinct pathways — acute economic-shock vulnerability and early-life violence exposure — that the survey’s variables are structured to describe but not to adjudicate between.
The Trace calculated that the Trump administration had terminated, redirected or delayed more than $1bn in federal grants related to gun violence, public safety, and mental health, but the survey’s design cannot determine whether that documented funding posture has shifted population-level prevalence.
The 2025 USVEX survey documents population-level prevalence rates — 52% lifetime physical violence, 49% lifetime sexual violence, 19% of women reporting potentially lethal IPV — that make visible an architecture of exposure that crime-report data systematically understates.

The 2025 United States Violence Experiences Survey, conducted by NORC at the University of Chicago for researchers at Tulane’s Newcomb Institute and the University of California, San Diego, surveyed 2,720 adults across all 50 states between October and December 2025 in English and Spanish using a nationally representative sample, and was released in September 2026, the Guardian reported. The headline numbers — 52% lifetime physical violence, 49% lifetime sexual violence, 61% of women and 45% of men reporting lifetime intimate partner violence — situate the survey’s analytical contribution as making visible what crime-report data cannot capture. The survey’s central analytical claim is that recent eviction is associated with sharply elevated reports of recent violence, but the cross-sectional structure leaves the direction of causation unsettled.

What the survey documents

The survey reports that lifetime prevalence of physical violence is identical across genders at 52%, though it documents sharp differences in who perpetrated the violence and where it occurred. Women most often experienced physical violence at the hands of a spouse or romantic partner (61%); men most commonly experienced violence from a stranger (38%). Among men, 25% reported violence by a spouse or romantic partner and 24% by a family member. Both women (69%) and men (34%) most often reported experiencing physical violence at home; men also reported violence in public spaces at 30%.

Sexual violence affected 68% of women and 29% of men during their lifetimes, with strangers accounting for 58% of past-year cases involving women and 66% of cases involving men. Acquaintances or community members were the next most commonly reported perpetrators (47% for women, 41% for men). Locations differed by gender: 55% of women who experienced past-year sexual violence reported it occurred in public spaces; 64% of men said it occurred at home. The study noted that sexual harassment accounts for much of this violence.

Past-year prevalence was lower: 5% of adults experienced physical violence (about one in 20), with 3% reporting physical abuse not involving a weapon; 7% experienced sexual violence (about one in 14), with the rate at 9% among women and 4% among men.

Intimate partner violence produced the most severe gendered findings. The study reported that 19% of women — nearly one in five — reported potentially lethal IPV, including beating, burning, strangulation, or violence involving a knife or gun, compared with 3% of men. Raj framed the fatality-risk approach: “It shouldn’t take a fatality for you to understand risk for fatality. By assessing behaviors that women have experienced that could have killed them, you have a better understanding of how much risk for fatality from domestic violence there actually is.”

The eviction-violence association and its causal ambiguity

The report’s central analytical finding is that adults evicted in the prior year were seven times more likely to report recent physical violence, three times more likely to report recent sexual violence, and nearly five times more likely to report recent IPV than adults who had not been evicted. Raj described the mechanism: “What we really see are what we call economic shocks that increase their very acute vulnerability. So [those are] things like not having enough money to meet basic needs like food … an eviction, or having been homeless in the past year.”

The cross-sectional design does not license a single causal interpretation. At least three alternative causal structures are consistent with the same observational pattern. First, reverse causation: the relationship between eviction and recent violence could run in the opposite direction, with an abusive partner or household member producing both economic instability and physical danger, eviction a downstream consequence of victimization rather than a precipitating condition. Second, an unobserved common cause — substance use disorder, neighborhood disadvantage, a specific perpetrator’s pattern of conduct — may produce both economic precarity and recent violence; the survey measures substance use and neighborhood safety, but latent household-level variables cannot be ruled out from the published data. Third, selection effects may distort apparent prevalence if respondents who have experienced violence are more or less likely to complete a survey on the subject.

A longitudinal design tracking individuals through documented economic shocks and measuring violence exposure at fixed intervals before and after — rather than the cross-sectional co-occurrence reported — would discriminate among these accounts, as would an instrumental-variable analysis exploiting plausibly exogenous variation in eviction risk, such as housing-court randomization or natural experiments in tenant-protection laws. The published seven-fold figure appears to be an unadjusted association; the report itself uses the language of “closely linked” rather than claiming a quantified causal effect, and does not specify which confounders the published estimate conditions on.

Two pathways

Raj identified two distinct temporal pathways through which vulnerability operates. The acute-vulnerability pathway runs through economic shocks — eviction, inability to meet basic needs, recent homelessness. The early-life pathway runs in the opposite temporal direction: “When you’ve had those experiences early in life, you are more vulnerable to them throughout your life, and so that’s really a piece of this.” The survey’s findings support both: experiences of violence rose sharply during adolescence and continued climbing into young adulthood, and by young adulthood, nearly one-third of adults reported having experienced sexual violence. In a structural reading, early-life exposure functions as a cause of lifetime prevalence; adult economic hardship is then a partial mediator, since early-life violence raises the probability of later economic precarity, which the survey’s variables treat as an exposure in its own right. The two pathways form a closed loop in which the cause of one period is partly the effect of an earlier one.

Mental-health consequences and reporting rates

Among those who experienced past-year physical violence, 48% of women and 26% of men said the experience resulted in anxiety or depression. Among women who experienced past-year sexual violence, 42% reported depression or anxiety; 39% of men reported the same. The survey found that 4% of women and 10% of men who experienced past-year sexual violence considered suicide — a gendered reversal the report presents without elaboration.

Filing rates after past-year physical violence were 10% among women and 17% among men — a minority of those affected filed formal complaints. Raj made the measurement-constraint point: “That’s a fraction of people who are affected. We also don’t want to find out about it at that point because if we’re finding out at that point, the risk for fatality has now gone up.” The methodological consequence is that crime-report data understates prevalence along a non-random margin, biased toward incidents severe enough to warrant reporting and toward populations with sufficient service access to file. Underreporting can be modeled as a measurement constraint conditioned on both violence itself and the availability of services, mental health support, and economic stability; conditioning on filing, as crime-report data implicitly does, opens a path from violence through service availability and economic hardship.

The documented federal grant posture

The survey’s release came months after The Trace — a non-profit newsroom dedicated to reporting on gun violence — had calculated that the Trump administration had terminated, redirected or delayed more than $1bn in federal grants related to gun violence, public safety, mental health and other efforts to save lives. The survey’s variables and The Trace’s documented grant posture address overlapping channels — mental-health service availability, economic-hardship amelioration — though the alignment is the analyst’s reading rather than a statement by either source.

The Trump administration appears in the published reporting only through The Trace’s calculation. The published reporting does not contain direct comment from administration officials on either the funding actions or the survey’s findings, and the administration’s own framing of the funding changes is not developed in the source article. Whether the documented grant posture has shifted population-level prevalence is a treatment-effect question the survey’s variables are structured to inform but not to adjudicate directly.

Interest mapping — documented positions

Thomas, a political science professor at UC San Diego and a co-author of the study, framed the prevention logic: “These findings show that violence prevention cannot occur in isolation. It must be part of a broader strategy that also addresses economic hardship, mental health and other forms of social inequality associated with violence.” Tulane’s release statement says the national study grew out of work the team had previously done focused on Louisiana, “before seeing many of the same patterns emerge nationally.”

The Louisiana Coalition Against Domestic Violence has documented that “the state’s rate of women murdered by men has historically been significantly higher than the national average.” Louisiana’s department of health had warned, two years before the Houma killings, that intimate partner violence was “a significant public health concern in the state.”

State agencies whose funding was terminated or redirected have institutional interests in maintaining program capacity and in survey findings that document the conditions those programs addressed. Survivors of violence carry substantive economic and safety interests, procedural interests in reporting, remedy, and service access, relational interests in family and intimate-partner stability, and recognition interests that the framing applied to their experiences either advances or diminishes. The procedural interests of survivors are not directly addressed by the survey, which measures prevalence rather than service access; the gap between formal procedural engagement and the underlying population affected is itself the underreporting finding.

Beneficiaries of the terminated or redirected grants are not named in the source article; the substrate does not establish identifiable beneficiaries of the funding actions on the other side of the ledger.

Asymmetries in visibility

Several interest profiles are less visible in the published framing. Men who experienced sexual violence — 29% lifetime, 4% past year — appear in the data but receive less narrative prominence than women’s experiences; the location split (64% of men reporting sexual violence occurring at home versus 55% of women reporting public spaces) is reported but not elaborated into a distinct analysis.

Rural respondents, specific racial and ethnic subgroups, and LGBTQ+ respondents are not disaggregated in the published summaries; their distinct interest profiles and exposure patterns cannot be inferred from the published reporting. The Louisiana-specific frame — state-level rates of intimate-partner homicide, the state’s prior warning, and the geographic proximity of the Newcomb Institute to Houma — remains present in the reporting without being developed into a state-level causal analysis.

The Houma killings — intersection without adjudication

The Houma killings occurred three days before the study’s release and about 60 miles from the Newcomb Institute. A stabbing at a home in Houma, Louisiana, killed a woman, her daughter, her son, and his girlfriend and wounded a six-year-old; the man authorities arrested had reportedly fathered a child with one of the victims. The incident is not in the survey’s data, but it sits at the intersection of two of the survey’s variables — intimate partner violence at 61% lifetime prevalence among women, and economic-shock vulnerability — without the survey being able to adjudicate which pathway applies. What the survey can adjudicate, and what its variables are structured to measure, is the population-level architecture in which such incidents occur.

Boundary questions the data leave unresolved

The cross-sectional structure cannot adjudicate among the causal accounts the eviction data suggest. The categorization of perpetrators — stranger, acquaintance, family member, spouse or romantic partner — depends on respondent report and is not externally validated. The mental-health consequences reported cannot establish whether violence caused the mental-health outcomes or whether prior mental-health conditions elevated the likelihood of exposure. Lifetime prevalence of physical violence is identical across genders at 52%, but the survey does not adjudicate the causes of the gendered asymmetries in perpetrator category and location.

Frame audit — how this is being framed

The published framing presents USVEX as a prevalence study that makes visible what crime-report data cannot capture, foregrounding economic-shock vulnerability as the survey’s central finding and aligning that finding with The Trace’s documentation of reduced federal grant funding in the same categories. Researchers’ quoted statements tie the survey’s intervention logic — that “violence prevention cannot occur in isolation” — to the categories of grant The Trace’s reporting indicates have been reduced. The article does not develop the administration’s framing of the funding changes or any counter-position from the administration; the published reporting does not contain direct comment from administration officials on either the funding actions or the survey’s findings.

The Houma killings provide temporal and geographic proximity to the release, with the Newcomb Institute located about 60 miles from Houma. The article does not classify the Houma incident by causal pathway — IPV versus economic-shock — and the survey’s data cannot adjudicate between them.

Survey’s contribution and what it leaves open

The survey’s contribution is the documentation of co-occurrence at population scale — prevalence rates, gendered patterns of perpetration, and the intersection with economic hardship — and the identification of policy categories — economic support, mental health, neighborhood safety — that co-occur with the violence the survey measures. The substantive analytical contribution of USVEX is to make that architecture visible in a way that crime-report data cannot. Whether the relationship among these categories is causal, and in which direction, remains a question the data as published do not resolve.

Analytical techniques used in this piece

This analysis applies the methods below. Each links to a short, plain-English explainer you can read and reuse.

Causal DAG: Maps cause and effect as an explicit directed graph, exposing confounders and mediators (Pearl).
Cui Bono — Who Benefits: Asks who gains and who pays from a state of affairs, decision, or claim.
Interest Mapping: Separates parties’ stated positions from their underlying interests (Fisher & Ury).