Using district variation in COVID-19 lockdowns in India to assess their impact on violence against women


This page in:

As noted by many, lockdowns can be an effective way of controlling a pandemic, but they come with costs. These costs are usually expressed in terms of economic losses -- such as GDP declines or poverty increases. In their new NBER working paper, Saravana Ravindran and Manisha Shah study the impact of COVID-19 lockdowns on violence against women in India, or costs in terms of a ‘shadow pandemic’.

A lot has been written on the impact of COVID-19 on violence against women. Media and other reports point to a sharp increase, especially in domestic violence, from COVID-19, due to both confinement and lack of access to services. I can’t do justice to the range of work that has emerged on this topic since the onset of the pandemic; see the posts here as a launching point to a lot of other work. Some of this work draws on evidence from past epidemics, other pieces are speculative (though with compelling reasoning, in my view), and a much smaller set has some data underpinning it. Real-time data on impacts from direct reports is hampered by the challenge of collecting sensitive information via digital methods (phone or internet). Experts in this area that I know have cautioned about trying to do so (as articulated in this piece).

Ravindran and Shah quantify COVID-19 lockdown impacts in India on violence against women by using temporal and spatial variation in the government-mandated intensity of lockdowns. Districts are classified as red zone (strictest restrictions during lockdown), orange zone, or green zone (most lenient), based on factors such as the number of COVID-19 cases and the rate of increase in COVID-19 cases. The paper has more details on the restrictions associated with these three classifications. To verify that the classifications have some import, they compare district lockdown classification with state-level Google Community Mobility Reports (drawn from those who have opted-in to share their Location History for their Google Account), showing that after the national lockdown but before the classifications were declared (end of March), decreases in mobility were not different across districts by classification. By May 2020, after the May 1 government declaration of classifications, red zone districts experience a greater decrease in mobility.

The district lockdown classifications are combined with district-month data on complaints made to the National Commission for Women (NCW), for different types of violence against women: domestic violence, cybercrimes (to capture on-line stalking, bullying, sexual harassment, and sex trolling), harassment, and rape and sexual assault. These same data are also being used by other researchers and media to examine COVID-19 impacts in India. The data clearly show an increase in domestic violence-related complaints received by the NCW during the lockdown period. The paper notes that this is consistent with an increase in Google search activity for domestic violence-related terms using Google Trends data for India during this same period – though these data are not reported by district. The paper offers scant detail, unfortunately, on how the complaints are received, with reference to postal and phone-based modes of communication (including a Whatsapp number set up in April 2020).

A small number of studies (cited in the paper) for other countries also track changes in violence against women using data from police reports and helpline calls, though without the type of quasi-random subnational variation in government-mandated lock-down policies as in India and/or variation in GBV by type. The latter will prove to be an interesting aspect of this study.

Using a difference-in-differences strategy with district data spanning from January 2018–May 2020, they find a 0.47 SD (131%) increase in domestic violence complaints in May 2020 in districts with the strictest lockdown measures (red) relative to districts with the least lenient lockdown (green). The red zone districts also experienced a 0.70 SD (184%) increase in cybercrime complaints relative to green zone districts. On the other hand, sexual harassment and rape/sexual assault complaints both declined significantly in May 2020 in red zones compared to green zones. Impacts in orange zones compared to green are more mixed. The descriptive results of complaints by type, specifically the increase of domestic violence and decrease in sexual harassment and rape/sexual assault complaints, are similar to the diff-in-diff results. (The data also show that the parallel trends assumption holds.) Results are robust to inclusion of region/month/year fixed effects to address the concentration of green zones in East and Northeast India.

Sexual harassment and rape/sexual assault complaints are posited to decline during lockdowns because they occur in public spaces, on public transport, and in workplaces. This is consistent with the decline in complaints in red zones (with greater lockdown rules with greater limitations on transport and work options) relative to green ones. It is also consistent with research that documents the high rates of sexual harassment women in India face in public transport and in workplaces.

Not all red districts are the same. They differ along important dimensions with regards to attitudes held by men and by women toward violence against women. Ravindran and Shah show that these attitudes matter for the uptake in domestic violence during the lockdown, using basically a triple-difference empirical strategy. The increase in domestic violence complaints between red and green districts is larger in districts in which a greater proportion of husbands report that beating wives is justified, measured in the National Family Health Survey 2015–2016. On the other hand, the increase between red and green districts is smaller in districts in which a greater proportion of wives report that a husband beating his wife is justified. (Results are not always statistically significant.) The finding on women’s attitudes is consistent with a lower propensity of women to report domestic violence when domestic violence is viewed by them as justified. This points to the issue of reliability of the complaints data and concerns about under-reporting. The differential directions of impacts for domestic violence and cybercrime complaints compared to rape and sexual assault complaints suggest that measurement error would not explain the findings. But a larger concern would be that true impacts of lockdowns on domestic violence are even greater than measured here if lockdowns increase barriers to reporting incidents, as seems quite plausible. It may be bad form to repeat a link that was posted by David in a July weekly links post, but this note on the COVID-19 impacts on difficulty to contact women by phone in Indian is worth a repeat mention here).

This study can’t unpack the specific pathways by which lockdowns increase domestic violence, such as, economic insecurity and poverty-related stress, or stress from social isolation, or the inability of women to temporarily escape abusive partners, or increased exposure of women to potential perpetrators of violence. Nonetheless, putting some numbers on a shadow pandemic is important for informing policies to address it.

Be on the look-out for more studies on this topic, including several that were part of the NBER Summer Institute 2020.


Kathleen Beegle

Lead Economist with the World Bank Gender Group

Join the Conversation