In the tenth post of I4I’s month-long campaign to mark International Women’s Day 2023, Seetha Menon investigates the causal relationship between domestic violence and increased risk of cardiovascular disease among women. Using data from NFHS-4, and instrumenting the price of gold at the time of marriage as a source of variation in domestic violence, she finds a positive effect of domestic violence on hypertension in women but finds no effect on the partnered men.

Join the conversation using #Ideas4Women

Cardiovascular disease (CVD) is the leading cause of deaths around the world (Roth et al. 2018). People at risk for CVD often have early markers such as hypertension or diabetes, that can be relatively easily measured, monitored and controlled. Identifying those at risk of CVD has the potential to prevent premature deaths.

Previous studies have shown that exposure to violence, such as the 9/11 attack or the Boston marathon bombings, is inherently stressful (Liu et al. 2014) and has the potential to increase acute stress even in people who are only exposed to media coverage of such an event (Holman et al., 2014; Thompson et al. 2019). 

Similarly, domestic violence victims often endure long-term exposure to violence, which has the potential to induce chronic stress. However, the effect of domestic violence on cardiovascular risk is largely unknown. In a recent study (Menon 2022), I estimate the causal effect of domestic violence on cardiovascular risk, using an instrumental variables strategy¹. I find that, all else being equal, women who suffer physical violence, have a higher risk for hypertension. Conversely, I also find that the matched men in these relationships – both the husbands who are perpetrators of domestic violence, and the husbands who are non-perpetrators – do not suffer a similar detrimental increase in CVD risk.²

Mechanisms through which domestic violence impacts CVD 

There are two mechanisms through which domestic violence can increase CVD risk in victims: a biological mechanism and a behavioural mechanism. In the first, violence exposure is linked to not feeling safe, which in turn is associated with increases in basal cortisol levels (Johnson et al. 2008, Alhalal and Falatah 2020), post-traumatic stress disorder (PTSD), anxiety and depression (Golding 1999, Ang 2021). Over time, this increases a person’s risk of cardiometabolic morbidities, including hypertension and diabetes, amongst others (Harris et al. 2017). The second mechanism through which domestic violence can impact CVD risk is through behavioural changes. These include adopting maladaptive coping mechanisms, such as smoking, alcohol abuse and eating disorders. These can increase a person’s vulnerability to a range of cardiometabolic disorders including hypertension and diabetes (Breiding et al. 2008). Nonetheless, emerging epidemiological evidence on the effect of domestic violence on CVD is still relatively limited and inconclusive. 

Study

I use data from National Family and Health Survey 4 (NFHS-4) which was fielded between January 2015 and December 2016. NFHS-4 is a nationally representative household sample that contains a rich variety of information including individual, partner, and household characteristics, a domestic violence module, and for the first time, a biomarker module. Over 97% of eligible women and 95% of partnered eligible men had their biomarkers recorded. Two binary indicators for hypertension and Type-2 diabetes are used as outcomes for risk of CVD. 

Identifying the effect of domestic violence on cardiovascular risk is challenging for at least two reasons. First, there is a likelihood of omitted variable bias, because variables such as lifestyle choices (including diet and exercise) and stress are associated with both domestic violence and CVD risk. Second, it is hard to elicit a precise measure of domestic violence in surveys, and there are likely to be measurement errors, which may be systematic. For instance, people with a high socioeconomic status may systematically under report domestic violence. To overcome this, I use the price of gold at the time of marriage as a plausibly exogenous source of variation in domestic violence. Specifically, I use the deviation of the price of gold at the time of marriage from its long-term trend as a source of exogenous variation, to proxy for initial marital endowment (Menon 2020), and show that an unusually high price of gold at the time of marriage is associated with a higher risk of domestic violence. 

I argue that this instrument is plausibly exogenous for four reasons. First, the price of gold is set at the time of marriage, well before the measurement of domestic violence or cardiovascular risk. Second, the price of gold is external to the country and determined by the London Price Fix twice a day. Third, volatility in the price of gold is high and there is no compelling reason to believe that the average individual is able to accurately predict or hedge against gold prices. Fourth, by using the distance of the short-term stochastic component from the trend³ concerns about potential common evolution of gold prices and economic development are attenuated. Furthermore, I relax this strict exclusion restriction and show that the findings are robust to a relatively large degree of instrument endogeneity⁴. 

Findings

I find a positive effect of domestic violence on hypertension in women. In the most stringent specification, being a victim of domestic violence translates to a marginal effect of 0.8 percentage points when all explanatory variables are set at their mean. There is no evidence that domestic violence affects the likelihood of developing Type-2 diabetes. This result is robust to using alternative cutoffs for a Type-2 diabetes diagnosis. These two findings are robust to using alternative estimators. I also find that the effect size is higher when examining contemporary physical violence which occurs within a year before the survey, than when considering the presence of physical violence at any time during the marriage, providing suggestive evidence that contemporary measures of domestic violence is more linked to the contemporary measures of CVD. 

As an extension to this study, I also examine the effect of domestic violence on CVD amongst the partnered husbands of the women in the analytical sample. This is interesting for two reasons. First, there is considerable evidence of concordance in health behaviours and health outcomes amongst spouses (Davillas and Pudney 2017, Fadlon and Nielsen 2019, Saarela et al. 2019), which suggests that men who are partnered with women with high probabilities of cardiovascular risk would themselves have a high probability of cardiovascular risk. Second, previous research suggests that domestic violence may increase the risk of hypertension in the perpetrators (Lindman et al. 1992, O’Neil and Scovelle 2018). I do not find any evidence of a negative effect of perpetrating domestic violence on CVD amongst the partnered men. This is despite the fact that the matched men in the sample have double the prevalence of both hypertension and Type-2 diabetes as compared to the matched women. 

Conclusion

These findings suggest that, in addition to the apparent physical effects, the hidden health burden of domestic violence is likely to be higher than previously thought. The magnitude of the effects found in this paper is likely to be a lower bound, as women with a higher predisposition to heart disease, and women who were subject to the most extreme cases of domestic violence, will not be part of the sample. Future work to identify the mechanism through which this effect operates requires detailed data on biomarkers of stress, in addition to data on domestic violence.

Notes:

1. An instrumental variable is used in empirical analysis to address endogeneity concerns. An instrument is correlated with the explanatory factor but does not directly affect the outcome of interest, and thus can be used to measure the true causal relationship between the explanatory factor (in this case, domestic violence) and the outcome of interest (CVD risk).
2. While there may be alternative/multiple perpetrators of domestic violence, this study exclusively examines domestic violence from the husband only
3. I use the Hodrick-Prescott high pass filter to separate the time series of the price of gold into its trend and stochastic (random) components.
4. I do this by estimating bounds of the second stage effect to allow a degree of endogeneity in the instrument.

Further Reading

• Alhalal, Eman and Rwawih Falatah (2020), “Intimate partner violence and hair cortisol concentration: A biomarker for HPA axis function”, Psychoneuroendocrinology, 122: 104897.
• Ang, Desmond (2021), “The effects of police violence on inner-city students”, The Quarterly Journal of Economics, 136(1): 115-168.
• Breiding, Matthew J, Michele C Black and George W Ryan (2008), “Chronic disease and health risk behaviors associated with intimate partner violence - 18 US states/territories, 2005”, Annals of Epidemiology, 18(7): 538-544.
• Davillas, Apostolos and Stephen Pudney (2017), “Concordance of health states in couples: analysis of self-reported, nurse administered and blood-based biomarker data in the UK Understanding Society panel”, Journal of Health Economics, 56: 87-102.
• Fadlon, Itzik and Torben Heien Nielsen (2019), “Family Health Behaviors”, American Economic Review, 109(9): 3162-91.
• Golding, Jacqueline M (1999), “Intimate Partner Violence as a Risk Factor for Mental Disorders: A Meta-Analysis”, Journal of Family Violence, 14(2): 99-132.
• Harris, Melissa L, Christopher Oldmeadow, Alexis Hure, Judy Luu, Deborah Loxton and John Attia (2017), “Stress increases the risk of type 2 diabetes onset in women: A 12-year longitudinal study using causal modelling”, PloS One, 12(2).
• Holman, E Alison, Dana Rose Garfin and Roxane Cohen Silver (2014), “Media’s role in broadcasting acute stress following the Boston Marathon bombings”, Proceedings of the National Academy of Sciences, 111(1): 93-98.
• Johnson, Dawn M, Douglas L Delahanty and Keri Pinna (2008), “The cortisol awakening response as a function of PTSD severity and abuse chronicity in sheltered battered women”, Journal of Anxiety Disorders, 22(5): 793-800.
• Lindman, Ralf, Bettina von der Pahlen, Bjorn O¨st, and CJ Peter Eriksson (1992), “Serum testosterone, cortisol, glucose, and ethanol in males arrested for spouse abuse”, Aggressive Behavior, 18(6), 393-400.
• Liu, Bian, Lukman H Tarigan, Evelyn J Bromet, and Hyun Kim (2014), “World Trade Center Disaster Exposure-Related Probable Posttraumatic Stress Disorder among Responders and Civilians: A Meta-Analysis”target="_blank" rel="noopener noreferrer", PloS One, 9(7).
• Menon, Seetha (2020), “The effect of marital endowments on domestic violence in India”, Journal of Development Economics, 143: 102389.
• Menon, Seetha (2022), “The effect of domestic violence on cardiovascular risk”, Review of Economics of the Household. Pre-print version available here.
• O’Neil, Adrienne and Anna J Scovelle (2018), “Intimate Partner Violence perpetration and cardiovascular risk: A systematic review”, Preventive Medicine Reports, 10: 15-19.
• ·Roth, Gregory A, et al. (2018), “Global, regional, and national age-sex-specific mortality for 282 causes of death in 195 countries and territories, 1980–2017: a systematic analysis for the Global Burden of Disease Study 2017”, The Lancet, 392(10159): 1736-1788.
• Saarela, Jan, Maria Stanfors, and Mikael Rostila (2019), “In sickness or in health? Register-based evidence on partners’ mutual receipt of sickness allowance and disability pension”, Social Science & Medicine, 240: 112576.
• Thompson, Rebecca R, Nickolas M Jones, E Alison Holman, and Roxane Cohen Silver (2019), “Media exposure to mass violence events can fuel a cycle of distress”, Science Advances, 5(4).