Too cold to cope, too hot to work: heterogeneous effects of extreme weather on intimate partner violence

This is our 8th blog post for our Job Market Paper Series blog for 2025-2026.

Julieth Saenz-Molina is a PhD candidate in Economics at Fordham University. Her research lies at the intersection of development and environmental economics, with a focus on how climate variability affects household welfare, labor markets, and gender-based violence in low- and middle-income countries.

Extreme temperature events are becoming more frequent and intense, affecting not only economies but also households. In places with limited social protection, weather shocks can translate into economic stress and conflict inside the home. My Job Market Paper studies how extreme weather can affect intimate partner violence (IPV) in Bolivia. A growing literature documents links between temperature shocks and violence, with evidence that extreme heat increases interpersonal and domestic violence across diverse settings. Much less is known about how both hot and cold shocks affect IPV, or how these effects vary across climatically distinct regions. 

The main takeaway of my analysis is that in low-altitude areas, both extremely cold and hot days raise IPV, though through different pathways.  By contrast, in the highlands, where temperatures are more stable and populations are adapted to cooler conditions, no clear response is observed. The results highlight the importance of climate adaptation policies that are gender-sensitive and grounded in local socio-economic context to address the social costs of climate variability.

Why Bolivia is a useful setting
Bolivia has one of the highest reported rates of intimate partner violence in Latin America, which makes it critical for understanding how environmental stress might exacerbate IPV. Bolivia spans tropical lowlands and high-altitude Andean regions. As Figure 1 shows, the lowlands experience frequent heat and episodic cold fronts called surazos, while the highlands are cooler and comparatively stable. This geographic contrast creates sharp differences in temperature ranges and exposure to climatic extremes, offering a natural setting to study how populations living in different environments experience weather variation.

Data and approach

I link the 2008 Bolivian Demographic and Health Survey to daily gridded temperatures from ERA5-Land, matching each DHS cluster to its nearest 10-kilometer climate grid cell. For every woman, I calculate the number of days in the previous 12 months that fell within 2°C ranges of daily maximum temperature, a flexible specification that captures nonlinear effects without imposing a functional form. Using narrow temperature bins allows the relationship between temperature and IPV to vary flexibly across the distribution and to be interpreted transparently. This approach is especially appropriate in Bolivia, where limited temperature seasonality—driven by the country’s proximity to the equator—means that even small deviations from usual conditions can represent substantial shocks for households.

I exploit natural variation in the timing of DHS interviews, which is driven by survey logistics rather than respondents’ characteristics. Because the DHS is nationally representative and households are randomly selected within clusters, interviews are scheduled independently of women’s behavior and outcomes. This allows me to compare women interviewed within the same region and month who faced slightly different temperature histories, generating quasi-random variation in recent exposure to hot and cold days.

To capture differences in climatic exposure, I estimate models separately for low- and high-altitude areas, using 1,000 meters above sea level as the cutoff. These regions differ not only in climate but also in socioeconomic composition, yet the key distinction lies in their exposure to temperature variability. In high-altitude areas, daily maximum temperatures rarely exceed 23 °C, while in low-altitude areas the range spans roughly 13 °C to 37 °C. Cold days in the tropics are rare and disruptive, while sustained heat is common. 

Main findings: heterogeneous effects by altitude
Average national effects on the incidence of physical and/or sexual intimate partner violence (IPV) mask substantial heterogeneity across regions. In high-altitude regions, estimates are small and statistically indistinguishable from zero across the temperature distribution. In low-altitude regions, the relationship is nonlinear and asymmetric. Ten additional days below 21 °C raise IPV incidence by about 3.6 percentage points relative to moderate temperatures, and ten additional days at or above 33 °C raise IPV by about 2.2 percentage points. By contrast, exposure to [21, 23) °C is associated with a reduction in IPV of roughly 6 percentage points. These magnitudes are policy-relevant and consistent with a de-adaptation mechanism discussed in recent work by Heutel et al. (2021) and Helo-Sarmiento et al. (2023): climate risk rises when conditions deviate from local norms. Populations accustomed to heat, for instance, may be less prepared to cope with sudden cold spells, which can intensify household stress and conflict.

How do temperature shocks translate into harm?
There are three main pathways in the lowlands.

1. Economic stress: In the coldest temperature ranges, both men and women shift away from non-agricultural work, and households experience declines in the DHS wealth index. Together, these patterns point to heightened economic stress within households. This pathway is consistent with a growing body of research linking environmental shocks—such as droughts and floods—to IPV through economic stress and income instability (Cools et al., 2020; Díaz and Saldarriaga, 2023; Epstein et al., 2020). 
2. Alcohol consumption: My results show that alcohol use is a key behavioral pathway through which temperatures translate into violence, especially during cold spells. Ten additional cold days below 21 °C increase the likelihood that the partner was intoxicated during violent incidents by about four percentage points. Research in Latin America consistently identifies male alcohol consumption as one of the strongest predictors of IPV (Angelucci, 2008). This alcohol-related channel is evident in both urban and rural areas, though the effect is larger in urban settings.
3. Women’s employment and bargaining power: My analysis shows that heat shocks reduce the probability that women work and lower their earnings relative to their partners. Prior research shows that declines in women’s relative earnings are associated with higher IPV risk, suggesting a plausible bargaining-power channel (Aizer, 2010).

These channels suggest that temperature affects IPV indirectly through income, employment, and behavior, rather than only through physiological discomfort. Whether migration or relocation away from extreme climates could mitigate IPV risk remains an open question, and one that warrants further study as climate-induced mobility becomes more common.

Who is most at risk?

Risk is not uniform within the lowlands.

• Rural and indigenous women face the sharpest increases after cold shocks. In these communities, ten additional cold days raise IPV by roughly 4 to 6 percentage points. Economic fragility and alcohol use reinforce one another.
• Urban women experience distinct patterns. Heat reduces women’s labor participation, which raises both the incidence and severity of IPV. Cold days, by contrast, do not increase overall IPV but are associated with more severe and alcohol-related violence.
• High-altitude households show minimal change, which is consistent with stronger long-run adaptation to a stable cool environment.

These heterogeneous effects make clear that climate impacts intersect with existing inequalities. Geography, livelihood, and identity shape who bears the brunt of climate stress.

Policy implications

As climate variability increases, households are likely to face more frequent exposure to atypical conditions, amplifying the social costs of weather extremes. In Bolivia, lowland households are accustomed to heat but face infrequent and abrupt cold surazos, which can catch them unprepared and heighten household stress. These results point to the need for climate adaptation policies that anticipate such localized risks and address unequal vulnerabilities.

Possible responses include:

• Expanding social protection and livelihood options. Cash transfers or public works programs that activate during cold spells can help cushion income losses, while income diversification reduces long-term exposure to climate shocks.
• Addressing behavioral risks. Cold spells coincide with higher alcohol-related IPV, suggesting opportunities for targeted prevention and enforcement during risky periods.
• Supporting women’s economic participation. Employment programs that stabilize women’s earnings when heat suppresses labor demand can help maintain household bargaining power.

The broader lesson is that climate policy must account for heterogeneity in vulnerability. Protecting women’s safety is a core part of building resilience, not an afterthought.

Featured image: The image is AI-generated using Sora and is intended to symbolically reflect climatic differences between highland and lowland regions, which are central to the analysis.