In India between 2001 and 2018, the majority of domestic violence cases were filed under ‘cruelty by husband or his relatives’, with the reported rate of this crime increasing by 53% over the 18 years. However, it is important to note that only some states recorded change in the reported rate with the almost stagnant reported rate of domestic violence in many states over time. Significant heterogeneity was seen in the pattern of the four types of crimes at the state-level. Overall, the mean persons arrested decreased irrespective of the crime during the period studied, and less than 7% of the filed cases had completed legal trial in 2018. We discuss the gaps identified in the reported data which unless addressed have major implications in the facilitating action to reduce domestic violence against women in India.
The rate of reported crime under all the considered categories excluding dowry deaths in 2018 in India in the NCRB was close to the 33% self-reported domestic violence reported by women in the national survey in 2015–16 , though there is an indication that the prevalence of domestic violence could be as high as 41% in India . The NCRB data provides passive surveillance with the source being the FIR filed by family/kin/community member with the police for a crime, and hence is dependent on the reporting from the community, which is known to be selective as women report less to the police for domestic violence due to various reasons including lack of social support, shame, and stigma [34,35,36,37]. These differences could account for differential rates of domestic violence between the police records and self-reporting of domestic violence in the surveys [3, 4]. Recently, it is also shown that how women are asked about domestic violence in surveys can also result in different estimates . Furthermore, the Principal Offence Rule followed by NCRB "hides" many cases of domestic violence as according to this Rule, each criminal incident is recorded as one crime. If many offences are registered in a single case, only the most heinous crime—one that attracts maximum punishment—is considered as counting unit [39, 40]. For example, an incident involving dowry death and cruelty by husband or relative will be reported in NCRB as dowry death as it warrants the maximum punishment, thereby, underreporting the number of cases with cruelty by husband or relative.
Cruelty by husband or his relatives
The cases under cruelty by husband or his relatives accounted for the majority of reported cases, and the rate of this reporting was comparatively higher in the middle-SDI states over the years studied. Previous research using field notes from cases reported to police indicate that victims are often in an environment that condones violence through active encouragement or tacit approval by the husband’s family members; and that many women lack social support as they experience violence from multiple perpetrators at home [34, 41]. It is plausible that this rate is higher in the middle-SDI states because material wealth is highly prized among the Indian middle class, and dowry is seen as an easy path to greater wealth and social status . A higher dowry demand, and a greater dissatisfaction from inability to meet these demands could possibly result in more domestic violence in these states [12, 42, 43]. Another possible factor in these states could be that the increasing female literacy in these states may be perceived as a threat to the prevalent power structures, prompting violence against women as a means to reinstate control [12, 44,45,46,47].
Abetment of suicide of women
The middle-SDI states also had a higher rate of reported cases under abetment to suicide. The link between abuse and suicidal behaviour is well established, with research indicating that three out of ten women who undergo domestic violence are likely to attempt suicide . Furthermore, a significantly higher suicide death rate is reported in Indian women than their global counterparts , and housewives account for the majority of these suicide deaths . Wide state-level variations in the suicide death rate for women are also reported , and the relationship between the prevalence of domestic violence and suicide death rate needs to be explored further.
In contrast to the increased reporting of cases of cruelty over time, the rate of dowry death cases decreased from 2001 to 2018, with the low-SDI states recording the highest rate of dowry deaths. The dip in these cases may have resulted from the 2010 judgment requiring prior harassment of the victim associated with a dowry shortfall which made it harder to register a dowry death but presumably also harder to prove beyond a reasonable doubt that it was a dowry death, and not in fact. Furthermore, qualitative research has shown that the families of dowry death cases deter from accusing the husband or his family due to fear of issues with up-bringing of the children of their daughter . Also dowry deaths or related suicide deaths are less likely to be reported by the natal family, who fear social stigma and negative impact on marriages of their other daughters [42, 49]. In this context, it is not easy to interpret the decreased number of cases of dowry deaths in India as actual fewer dowry deaths, for which more evidence is needed.
Very few cases were filed under PWDVA with the middle-SDI states reporting no cases during the period studied. While PWDVA defines domestic violence to include coercive behaviour as well as physical, sexual, emotional and economic abuse , in actuality only extreme forms of physical violence with evidence of injury are seen to evoke a legal response . Interviews with victims indicate that unless they were able to offer a dowry claim or show evidence of grave physical violence, the police were either reluctant to file an FIR or offer PWDVA as a legal recourse to them . It is also documented that the police, acting as social brokers, attempt to fit the reported domestic violence cases into ‘normal constructs’ frequently focusing on dowry harassment despite the broadened scope of the law as a recourse for domestic violence beyond dowry harassment . Thus, data under this crime heading is unlikely to reflect the true picture of domestic violence against women in India.
Status of the legal cases
The poor response of formal system to domestic violence is also reflected in the legal recourse as only 6.8% of the cases filed completed trials in 2018, with the majority of accused being acquitted. This bleak state of waiting, extended trials and low conviction is known to further discourage women from reporting . The legal process is also influenced by the patriarchy driven attitudes of the police and people in the legal systems , and their unwillingness to act on domestic violence cases which they view as “private matter,”[13, 44] such that many cases are not investigated, or dropped due to delay in filing . In other cases, the investigation is based on the statement of the husband or relatives rather than fingerprints , with the perpetrator of violence not even recorded in over 90% of the cases . Notably, little empirical research is available on the perceptions of abusive husbands and families on domestic violence that can facilitate intervention programs for abusive husbands .
Limitations and way forward
There are limitations to the data presented and the interpretation. The NCRB data depend on the availability and quality of data recorded by the police at the local level, which is known to have varied quality [9, 10, 18]. The findings have to be interpreted within in this limitation as it is not possible for us to comment on the extent of underreporting of data or the pattern of underreporting by type of crime, year or state. The heterogeneity in the NCRB data at the state level highlighted by the noisy trends or stagnant trend for certain states do not allow for a meaningful interpretation, and calls for a robust assessment of the reporting practices by the police and judiciary at state level to identify the gaps for inadequate documentation and underreporting that can facilitate appropriate corrective measures to improve data quality . We assumed the age group of affected women to be 15–49 years. Though majority of the cases are likely to be in this age group given the other available information, the unavailability of age of women affected by the type of crime, year and state restricts understanding of the target women for prevention and action. Currently, the data are available in heavily tabulated fixed formats that limit the extent of disaggregated analysis. Because of non-availability of data on number of victims for some years, we assumed the ratio of the number of cases to victims based on the available data for other years. More informative analyses may also be possible if the NCRB reports allow for anonymized individual level data to be available in the public domain, including repeat reports of domestic violence by individual women.
Despite NCRB being a passive surveillance source, efforts can be made to improve the quality of information collected by the police during their routine tasks to improve utilisation of these data for planning action. The World Health Organization injury surveillance guidelines could provide practical advice on collecting systematic data on domestic violence, which can be more comparable over time and location . Training and sensitisation of the police to address gender violence should also include standardisations in capturing of the data and the quality of data captured.
Disasters, natural or otherwise, disproportionately impact women and girls with some evidence suggesting that violence against women increases in disaster settings, however, there is a lack of rigorously designed and good quality studies that are needed to inform evidence-based policies and safeguard women and girls during and after disasters . There has also been suggestion of an increase in domestic violence against women during the Covid-19 pandemic, globally  and in India [54, 55]. In this context, the urgency to address the gaps highlighted in the NCRB data is even more for India to protect its women against domestic violence.