Rainfall shocks constitute one of the most important sources of crop production risk among smallholding farmers in developing countries. Ample research documents that rainfall shocks reduce agricultural productivity (Holden et al., 2001; Chavas et al., 2005; Hansen et al., 2011), generating sizable income losses and decreasing consumption (Wolpin, 1982; Paxson, 1992; Jacoby and Skoufias, 1998; Porter, 2012; Amare et al., 2018) and productive investments (Alem et al., 2010; Dercon and Christiaensen, 2011; Barrett and Carter, 2013; Shiferaw et al., 2015; Amare and Shiferaw, 2017). Even more, and despite being transitory, rainfall shocks can affect developmental outcomes in the short (Hoddinott and Kinsey, 2001; Maccini and Yang, 2009; Rocha and Soares, 2015; Rosales-Rueda, 2018) and long term (Holden et al., 2001; Stanke et al., 2013; Dinkelman, 2017; Shah and Steinberg, 2017; Joshi, 2019).

Less studied, however, are the immediate effects rainfall shocks can pose on intra-family dynamics. In particular, there is scant evidence on how rainfall shocks may affect family bonds. In this paper, we draw attention to the effect rainfall shocks have on one specific dimension of family relations: marital conflict. Specifically, we investigate whether and how exposure to rainfall shocks affects the prevalence of physical intimate partner violence (P-IPV) among women, uncovering the plausibility of multiple mechanisms driving the connection between rainfall shocks and the risk of experiencing P-IPV.

Assessing this impact is of primary interest as intimate partner violence (IPV) constitutes the most common form of violence against women (VAW) and is both a cause and a consequence of gender disparities that manifest in power relations within the couple. Worldwide, nearly 30 percent of women have suffered some form of IPV (Devries et al., 2013) and more than 35 percent of all murders of women are reported to be committed by an intimate partner (WHO, 2012).2 In this regard, IPV is a global concern encompassing a severe health problem (ICRW, 2009; Butchart et al., 2010; Garcia-Moreno and Watts, 2011).

We set our study in the context of rural households in the Andean region of Peru. Several facts motivate our focus on this particular setting. First, the prevalence rates of lifetime and recent IPV in rural Peru —- of 61 and 25 percent, respectively —- are in the upper tier among countries that record IPV (Garcia-Moreno et al., 2006; Devries et al., 2013). Second, with roughly 80 percent of employment in agriculture and 75 percent of rainfed crop production, living conditions in the Peruvian Andes are intrinsically tied to weather realizations (Ponce et al., 2015). Third, insufficient irrigation infrastructure and the use of ancient farming practices make crop production in this region particularly vulnerable to extreme rainfall events (Escobal, 2005). Fourth, limited access to credit markets in these areas makes it more difficult for households to cope with adverse income shocks (Trivelli, 2000).

Our analysis incorporates spatial data on historical rainfall levels from the University of Delaware’s climate data repository and women’s records on P-IPV from repeated annual cross-sections of the Peruvian Demographic and Health Surveys over 2005–2014. We use these data to track changes in the experience of instances of P-IPV occurring during the past 12 months among women exposed and not to dry or wet shocks during the cropping season in the municipality where they reside. We construct our measure of recent P-IPV using a slightly adapted version of the Conflict Tactic Scales (Straus, 1990) available in our data. We define dry and wet shocks as monthly rainfall falling below or above the 5th or 95th percentiles in the distribution of long-term (1950–2010) local monthly rainfall observed in this season, respectively.3

Our principal finding is that women in rural Peruvian Andes experience more P-IPV when exposed to dry (but not wet) shocks. In particular, the probability that a woman experiences instances of P-IPV during the past 12 months increases by 8.5 percentage points (65 percent) after being exposed to a dry shock during the cropping season. This estimate is statistically robust to a variety of falsification and sensitivity tests. We further delve into the details of the aggression and find that the probability that a woman experiences acts of “moderate” P-IPV, such as being pushed/shook, slapped, punched, or kicked/dragged by their partners, increases by 60 percent after dry shocks are observed during the cropping season. Even though these acts do not aim to cause permanent injury to the victim (Bott et al., 2012), we still find a 70 percent increase in women’s reports of sequelae from the physical abuse in the form of bruises or lesions on their bodies.

Borrowing from the literature on IPV (Ellsberg and Heise, 2005; Eswaran and Malhotra, 2011; Bott et al., 2012; Garcia-Moreno et al., 2015; Buller et al., 2018; Tankard and Iyengar, 2018; Peterman et al., 2020; Ranganathan et al., 2021), we identify three main pathways that link exposure to rainfall shocks with instances of P-IPV: (i) household economic insecurity, poverty-related stress, and emotional well-being; (ii) women’s empowerment; and (iii) exposure to aggressive partners.4 Naturally, P-IPV arises from the interplay between these different pathways, either reinforcing or offsetting (Roy et al., 2019). Among them, we assess as many mechanisms as our data allow us to do.

Consistent with our main result, we find that only the exposure to dry shocks induces changes in the mediating factors. We find that household income and consumption expenditures per capita decline by 20 and 15 percent, respectively, after the occurrence of a dry shock during the cropping season. These results indicate that income loss is one of the main drivers of the observed increase in P-IPV and point to the economic hardship generated by dry shocks as a principal mediating channel. We also document increases in men’s marital control, alcohol intake, and alcohol-related aggression — behaviors that link to lower financial command, stress, and anticipatory anxiety from higher economic uncertainty. Yet, we also find a decline in women’s employment and financial autonomy — a result that highlights the importance of female economic opportunities in determining P-IPV.

Our results speak to the vulnerability of women to the occurrence of weather shocks and identify pathways through which social policies can amend this condition. This situation becomes more relevant in the face of climate change as its negative impacts are likely to be harsher among rural populations in developing countries (Morton et al., 2014).5 A back-of-the-envelope calculation using weather projections from the National Center for Atmospheric Research suggests that the prevalence of P-IPV would increase by 3.5 to 4 percentage points (or 25–30 percent) due to more frequent dry shocks that will be observed in the Peruvian Andes over 2025–2034, provided that no adaptation or mitigation strategies will be put in place in the face of climate change.

We contribute to four different strands of the literature. First, our study adds to the literature on rainfall shocks and IPV providing evidence on causal pathways. Previous studies also established a connection between dry shocks and increased IPV, but none have addressed the causal pathways. In particular, three other studies have also evaluated this effect finding mixed results in the context of sub-Saharan African countries. While Cools et al. (2020) find no impact, Abiona and Koppensteiner (2018) and Epstein et al. (2020) document that exposure to dry shocks increases the probability that women experience P-IPV by between 25 to 50 percent.6 We introduce three important extensions that make our study more informative. In terms of methodology, our data integrates information on precipitation across geography and over time, which permits for a clear identification of our effect of interest by harnessing temporary, local variation in rainfall. Concerning the context of study, we provide novel evidence from the rural Andes in Peru, a developing country outside sub-Saharan Africa with a different culture and custom setting. As regards the interpretation of the results, the wealth of data we utilize allows us to thoroughly analyze causal pathways and mechanisms and shed light on the likely channels linking rainfall shocks with P-IPV.

Our findings reinforce the notion that household economic insecurity and poverty-related stress are critical determinants of IPV (Fox et al., 2002; Schneider et al., 2016).7 Also on this matter, we highlight the relevance of female financial autonomy and economic opportunities in balancing the woman’s power position within the relationship, which is particularly relevant in times of economic hardship (Aizer, 2010; Anderberg et al., 2016; Krupoff et al., 2017; Tur-Prats, 2021). Our results, however, contrast those from previous work, arguing that IPV could result from a male retaliation to a lost authority or gender identity as the “family provider” (Chin, 2012; Guarnieri and Rainer, 2018; Garcia-Ramos, 2021).

Second, our study relates to new research showing that economic uncertainty from rainfall shocks can bring forth the existence of gendered power dynamics. This situation manifests, for instance, in the higher probability that parents “marry off” their daughters in societies where marriages involve financial transfers across families (Corno et al., 2020; Corno and Voena, 2021) or the lower probability that parents enroll their daughters in primary schools (Björkman-Nyqvist, 2013) after the occurrence of rainfall shocks. In this line of work, we uncover gender-based relations of power among couples that become apparent in behaviors such as increased marital control from men.

Third, our study provides novel evidence to the climate-conflict literature. For the most part, studies focus on the conflict between unrelated individuals and do not provide hints on the causes of conflict outbreaks due to the lack of history of interactions among disputing parties (Hsiang et al., 2013; Burke et al., 2015). By focusing on marital conflict, we can assess individual and marital relationship dynamics to reveal details on the likely causes of disputes. We emphasize the relevance of loss-of-control on emotions when confronting weather shocks, consistent with recent advances in this area (Burke et al., 2018; Mullins and White, 2019; Baylis, 2020).

From a broader perspective, our paper contributes to a more extensive literature that analyzes the effects of economic shocks and programming on VAW, in general, and on IPV, in particular. As long as these shocks affect intra-household dynamics and the economic and social status of women and men differently, they can affect IPV. Besides the strand of the literature focusing on the effects of rainfall shocks on VAW and IPV, several other papers study the relationship between VAW and IPV and labor market shocks (Aizer, 2010; Heath, 2014; Anderberg et al., 2016; Kotsadam and Villanger, 2020; Bhalotra et al., 2021b; Bhalotra et al., 2021a; Tur-Prats, 2021; Erten and Keskin, 2021a, Erten and Keskin, 2021b), microcredit and the financial inclusion of the poor (Vaessen et al., 2014; Orton et al., 2016; Duvendack and Mader, 2020; Limbikani et al., 2021), and (conditional or unconditional) cash and in-kind transfers (Buller et al., 2018; Roy et al., 2019; Haushofer and Shapiro, 2019; Heath et al., 2020; Díaz and Saldarriaga, 2022). More recently, a flurry of papers explores the effects of health-related policies to contain the COVID-19 pandemic worldwide on VAW and IPV (Peterman and O’Donnell, 2020; Peterman et al., 2020; Peterman et al., 2020a, Peterman et al., 2020b; Bourgault et al., 2021).8

Finally, our study has implications for the literature assessing the lifetime effects of weather shocks. Despite the extensive evidence on the negative repercussions on the human capital of individuals exposed to weather shocks during critical developmental stages of life, its causes are not yet fully understood. In this regard, we stress the relevance of the family environment, in general, and intra-family conflict, in particular, as a transmission channel from early-in-life exposure to weather shocks to later-in-life developmental outcomes (Currie and Vogl, 2013; Aizer and Currie, 2014; Doyle and Aizer, 2018).

We organize the paper as follows. Section 2 describes the data and the definition of the variables utilized in the analysis. Section 3 presents our empirical strategy for estimating the effect of exposure to rainfall shocks during the cropping season on women’s probability of experiencing P-IPV. Section 4 reports our main results, several robustness checks, and additional analyses. Section 5 presents our study of causal pathways and mechanisms potentially explaining our main results. Section 6 concludes.