Two studies link heat and drought to domestic violence, and disagree on the details

Climate-health research has focused overwhelmingly on physiological harm — heat stroke, cardiovascular strain, kidney injury. Two studies published within nine days of each other in August examine a different pathway, and one that is harder to measure: whether extreme weather is associated with violence inside households.

Both find associations. They also disagree in a way that is worth reading carefully rather than averaging away.

New Orleans, 2011 to 2021

The first study analysed 150,523 domestic violence-related calls to the New Orleans Police Department between 1 January 2011 and 31 December 2021 [s1]. Of those, 69.6% were classified as domestic disturbances and 22.4% as simple battery [s1].

The design was a spatially weighted time-stratified case-crossover — a method that compares a person or place against itself on other days, which removes the influence of anything that does not change day to day, such as neighbourhood characteristics or policing patterns.

Heat was measured using the Universal Thermal Climate Index, which combines temperature, humidity, wind and radiation into a single figure closer to what a body experiences than air temperature alone. Six definitions of extreme heat were tested, crossing two thresholds — 30 °C and the 90th percentile — with durations of one, three or five consecutive days [s1]. Calls were aggregated to zip code days and analysed with conditional logistic regression weighted by cases per zip code day [s1].

Associations were consistently positive [s1]. When mean UTCI exceeded 30 °C, the likelihood of a domestic violence call rose 4% (odds ratio, 1.04; 95% CI, 1.02 to 1.07) [s1]. When mean UTCI exceeded the 90th percentile for five consecutive days, it rose 7% (odds ratio, 1.07; 95% CI, 1.03 to 1.12) [s1] — duration mattered more than a single hot day.

The attributable estimate is modest: eliminating such heat wave events would most likely have prevented approximately 245.0 calls (95% CI, 105.1 to 370.9) across the eleven-year study period [s1]. Against 150,523 total calls, that is a small fraction, and stating it plainly matters more than the percentages do.

Forty-two countries, 2003 to 2020

The second study used multicluster survey data from women in 42 low- and middle-income countries across sub-Saharan Africa, South and Southeast Asia, Latin America and the Caribbean, surveyed between 2003 and 2020 [s2]. The analysis covered 494,471 women with a mean age of 32.1 years (SD 8.5) [s2].

Reported prevalence of intimate partner violence in the 12 months before the survey was 28.3% for any form (139,901 women), 16.8% emotional (83,100), 19.5% physical (96,531) and 7.4% sexual (36,804) [s2].

Drought was measured using the standardized precipitation evapotranspiration index at timescales of 1, 3, 6 and 12 months; extreme heat days were defined against cluster-specific distributions at the 90th, 92.5th, 95th and 97.5th percentiles [s2].

Drought exposure at every timescale was associated with increased risk of intimate partner violence, with the largest effect at 12 months (relative risk, 1.07; 95% CI, 1.06 to 1.09) [s2]. Both mild drought (RR, 1.08; 95% CI, 1.06 to 1.09) and severe drought (RR, 1.06; 95% CI, 1.04 to 1.08) were associated with any intimate partner violence [s2]. By violence type, the largest effects were for drought with emotional violence at the one-month scale and with physical violence at the 12-month scale [s2].

The finding that complicates the story

The intuitive expectation is that drought and extreme heat compound each other. The study tested this by estimating relative excess risk due to interaction in two ways, and found a negative and significant interaction between drought and extreme heat, defined at the 97.5th percentile threshold [s2].

A negative interaction means the combined association was weaker than the sum of the separate ones. The study does not explain why, and neither should this account. Possible readings — behavioural changes under extreme heat, measurement artefacts in survey recall, differing timescales between a multi-month drought and a handful of extreme days — are speculation, not findings.

What it does establish is that "climate stress increases domestic violence" is too crude a summary of the evidence. The relationship is not a simple dose-response.

What neither study can show

Both are cross-sectional or case-crossover analyses of administrative and survey data, and neither can establish causation for any individual episode of violence.

The New Orleans study measures calls to law enforcement, which is a measure of reported and police-involved incidents rather than of violence itself [s1]. Reporting behaviour is not constant — it varies with trust in police, with whether people are home, and with who else is around to call. Heat changes all of those.

The 42-country study relies on self-reported intimate partner violence in household surveys [s2], a measure with well-documented under-reporting that varies by culture, by interview conditions and by whether the survey is conducted privately.

Both studies' geographic scope is also a limit in opposite directions: one city over eleven years [s1], versus 42 countries pooled across 17 years [s2].

Why it is being studied now

The New Orleans authors describe evidence specific to domestic violence as limited, particularly in climate-vulnerable urban areas [s1], and conclude that climate adaptation should be integrated with violence prevention strategies during extreme heat events [s1]. The 42-country authors reach a similar policy conclusion: given the increase in extreme events under climate change, there is a pressing need for enhanced initiatives to prevent domestic violence [s2].

Neither claims that heat causes violence. What both suggest is that heat-health action plans, which are typically built around cooling centres, hospital surge capacity and warnings to vulnerable individuals, may be aimed at an incomplete list of harms. Domestic violence services are not usually part of heat preparedness planning; on this evidence, the case for including them is worth examining even if the effect sizes are small.

The effect sizes are small. A 4% to 7% increase in call likelihood, and a 6% to 8% increase in intimate partner violence risk with drought, are the kinds of numbers that only translate into meaningful absolute burden because the underlying prevalence is enormous — 28.3% of nearly half a million surveyed women reporting intimate partner violence in a single year [s2].

Sources

• Extreme Heat and Calls to Law Enforcement Related to Domestic Violence — JAMA Network Open, 2025-08-29
• Drought, Extreme Heat, and Intimate Partner Violence in Low- and Middle-Income Countries — JAMA Network Open, 2025-08-20