Background

Violence against women is a major social, and public health problem globally affecting an estimated 30 percent of women, aged 15 years and above. In 2018, about 852 million women experienced one or multiple forms of violence in their lifetime (World Health Organization, 2021). Rape is a common form of violence against women across nations, and in India, its incidence rate is about 4.8 per 100,000 women (National Crime Records Bureau, 2022). As per the 2012 statistics (United Nations Office on Drugs & Crime, 2023), per 100,000 people, 1.96 rapes were reported in India, while it was 1.44 in Canada, and 8.32 in Bhutan. Though the prevalence of rape is low compared to other countries, India is home to the second-largest female population in the world (World Bank, 2024). The aspect of female population size can also be seen in the reported rape numerals for 2012, which were 24,923 for India, 499 for Canada, and 60 for Bhutan (United Nations Office on Drugs & Crime, 2023).

Sexual violence affected an estimated 27.52 million Indian women aged 15 to 49 years in 2013 (Raj & McDougal, 2014). Psychological victimization of the rape survivor occurs before, during, and after rape (Chaudhury et al., 2017; Nieder et al., 2019). The emotional and cognitive burden associated with the fear of future violence extends beyond the survivor to other women in the community, resulting in endless self-censoring and self-regulation of behavior in daily lives (Mukherjee & Basu, 2017; Nieder et al., 2019; Pickup et al., 2001). The social and public health implications of rape are complex as it involves the survivors’ physical, mental, reproductive, and sexual health as well as the quality of life (Basile et al., 2021; Chaudhury et al., 2017; International Centre for Research on Women, 2004; Mukherjee & Basu, 2017). Rape can result in mental health consequences like anxiety, depression, suicidal ideations, substance abuse, and phobia about sex (International Centre for Research on Women, 2004; McCauley et al., 2013; Mukherjee & Basu, 2017; Pickup et al., 2001; Rai & Rai, 2020). Zinzow et al. (2012) using a nationally representative sample from the USA found that in rape survivors, the lifetime prevalence of posttraumatic stress disorder (PTSD) varied between 30–52 percent depending on the type of rape while that among non-victims was only 13 percent. The social consequences of rape include social isolation, stigma, social exclusion, difficulty in establishing social alliances, and the ability to work (International Centre for Research on Women, 2004; Mukherjee & Basu, 2017). The rise in reported rapes has had direct and reputational effects on the Indian economy, mainly through a drop in tourism and foreign direct investment (Jha & Sarangi, 2014). Like no other crime, rape includes transaction costs (the state’s expenditure on law enforcement and compensations), social, and psychological costs in one (Kamthan, 2013).

Newton and Felson (2015) state the need to study crimes from a spatio-temporal perspective; The activities or routines of offender and victim, in both time and space, may result in the occurrence of a crime like rape (Lauritsen & White, 2014; Newton & Felson, 2015). There is considerable spatial and temporal variation in the distribution of crimes against women (CAW) (especially rape) across India (Chauhan & Baraik, 2016), which is even more evident at the district level. The localization of CAW was indicated by studies such as Mukherjee et al. (2001), Kabiraj (2022), and Bala (2007) which found that most regions that fall under a certain crime level were spatially contiguous. According to Kabiraj (2022), apart from the local characteristics, the crime rate for a particular region was affected by the crime rates of its neighbors, mainly due to the “criminal behavior and harmful social norms that spread across contiguous regions over time”. Several studies exist to show the effect of environmental factors like weather and season on the temporal variation of CAW (Heo et al., 2022; Hewitt et al., 2022; Lauritsen & White, 2014; Mišák, 2022; Xu et al., 2021). Most sexual offenses like rape occur in summer than in any other season. In the Indian context, studies investigating the associations between time, weather, seasons, and violence against women are sparse. Sekhri and Storeygard (2014) found that a decline in annual rainfall increased domestic violence and dowry deaths. Chaudhary (2015) found that for Delhi, 51 percent of rapes reported in a prominent newspaper during 2011–13 occurred before evening, as opposed to 15.8% during the night. The reason could be that women have restricted mobility during the night compared to other times, which lessens the chance of being victimized.

The regional and temporal variation in contextual factors like socioeconomic conditions could also explain the heterogeneous spread of rapes across India (Bala, 2007; Chauhan & Baraik, 2016; Woetzel et al., 2015). The 2012 fatal Delhi ‘Nirbhaya’ gang rape case led to outrage and increased awareness among the public, change in legislative and judicial responses following which the rape reporting saw a massive increase, especially in districts of Northern India (Bandyopadhyay, 2018; Chhachhiya, 2017; McDougal et al., 2021). The district-level heterogeneity varied according to distance from Delhi, the presence of women police stations, the literacy sex ratio, and access to media.

Considering the above facts, for understanding the district-level heterogeneity in reported rapes across India from 2011 to 2020, the present study used spatial and space–time scan statistics to detect localized high rape clusters. The study aims to identify the spatial and spatio-temporal pattern of rape across India during the period 2011–20, with a focus on district-level analysis. The study also investigated the persistency of patterns, if any, across districts during the period. Identification of the high rape clusters could result in understanding the contextual factors of rape and, thereby helpful in designing, implementing, and evaluating appropriate regional interventions and prevention policies.

Data and methodology

Study area

At the time of Census 2011 (https://censusindia.gov.in/), India was divided into 28 states and seven Union territories including a total of 640 districts (Fig. 1). The country has a total area of 32,87,469 sq. km and a population density of 382 people per sq. km. According to the Census, the total population of India is 1,21,08,54,977 of which 58,75,84,719 (48.5%) are females. The sex ratio is 943 females per 1000 males. Around 31.14% of the Indian population is urban. The overall literacy rate among the population aged 7 years and above is 74% with the literacy rate for males being 82% and for the females being 65%. The work participation rate for females is 25.5% and for males is 53.3%. The rate of crime against women in India (under the Indian Penal Code, Special and Local Laws) increased from 38.9 in 2011 to 54.26 per 100,000 women in 2020 (National Crime Records Bureau, 2023).

Fig. 1
figure 1

States and Union Territories of India as per Census—2011

Full size image

The intersectionality of gender, caste, and class hierarchies paved the way for Brahminical Patriarchy in India where the subordination of women and suppression of their sexuality were necessitated for patrilineal succession of private property as well as to maintain caste purity (Chakravarti, 1993; Chapman, 2014; Dixit, 2020; Roy & Dastidar, 2018). According to Krishnan (2015) even in contemporary India, caste, politics, and capitalism play an active role in maintaining women’s subordination in society. Roy and Dastidar (2018) claimed that sexual violence against women is more likely in a society where women are considered inferior to men in the social and cultural contexts; They found social attitudes towards women as the most robust predictor for the incidence of rape in India.

According to Javidan (2021), structural sexism “acts as both a producer and product (or cause and effect) of the elements comprising it, which are mutually reinforcing to entrench sexism in societies.” Structural sexism is reproduced through gender socialization and conformity, ideology, as well as harm and violence, all of which are evident in India (Javidan, 2021; Nieder et al., 2019; Siddiqi, 2021; Singh & Aggarwal, 2020; Sivakumar & Manimekalai, 2021). With traditional gender roles favoring men in a patriarchal society like India, gender socialization taught across generations and institutions such as family, community, and mass media, tend to normalize as well as increase the risk of sexual violence against women. Also, the patriarchal backlash when women deviate from gendered expectations and norms could factor in the rise of sexual violence against women (Nieder et al., 2019; Singh & Aggarwal, 2020).

The acceptance of rape myths and gender stereotypes among even educated men and women in India, evident in law enforcement, Indian courts, and mass media (Fadnis, 2018; Kishore, 2023; Mukhopadhyay & Banerjee, 2021; Qureshi & Kim, 2018; Qureshi et al., 2021; Tannvi & Narayana, 2022), also perpetuates patriarchy and male supremacy in the Indian society in a vicious cycle. The traditional gender attitudes towards women and hostile sexism make rape myths more acceptable in Indian society (Hill & Marshall, 2018; Qureshi et al., 2021).

The majority of women in India experience benevolent sexism in their daily lives, mostly in the form of protective paternalism and complementary gender differentiation (Kishore, 2023; Shashwati et al., 2023; Siddiqi, 2021) with protective paternalism restricting social opportunities of women. Jain et al. (2020) found high levels of both benevolent and hostile sexism in Indian men. The normalization and continuation of a patriarchal society along with the influence of religion and ancestors as well as years of social conditioning form the root causes of sexism in India (Kishore, 2023). According to Mukherjee and Basu (2017), the common nature of public reaction to rape is protective in India which becomes ironic when most of the survivors were raped by people known to them.

Sources of data

The district-level rape data for the years spanning 2011 to 2020 were extracted from NCRB (National Crime Records Bureau, n.d.) and Open Government Data (OGD) platform (https://data.gov.in/), and district-wise female population data for the year 2011 was extracted from the Census of India (https://censusindia.gov.in/). The 2019 district information was not available for the state of West Bengal and hence the 2018 data was used in the crime data (National Crime Records Bureau, n.d.). The data sourced from railways, and special branches (like crime, narcotics, and CID) were excluded in the present study. The districts that split after 2011 were merged back into their original districts (n = 640) as per the 2011 census. To maintain comparability, the study considered for analysis only those cases reported explicitly under the heading ‘rape’ in the NCRB crime data. Till 2017, no separate information was provided for the rape of women above and below 18 years of age, hence, though the term ‘women’ is used in the analysis, the data also includes minors.

The latest district wise female population for India as well as the corresponding district-level shapefile (GitHub, 2014) was based on the Census 2011 and hence, we have considered 2011 as the base year. A 10-year study period (2011–20) was considered to uncover any recurring spatial or spatio-temporal patterns in rape clusters. The annual national and district-wise female populations between 2012 and 2020 were obtained by extrapolating the 2011 census data using the 2001–11 exponential growth rates. The district-level shapefile of India based on the 2011 census was used under the Creative Commons license and was obtained from GitHub (GitHub, 2014).

Data analysis

Zonal classification of states and union territories

For analysis, the study also considered the classification of states and union territories (UTs) as per the organization of Zonal councils in India (https://www.mha.gov.in/). States of Jammu & Kashmir, Himachal Pradesh, Haryana, Punjab, Rajasthan, UTs of Chandigarh, and the National Capital Territory of Delhi form the Northern zone. The North Eastern zone comprises Assam, Arunachal Pradesh, Manipur, Tripura, Mizoram, Meghalaya, Nagaland, and Sikkim. The Central zone includes Chhattisgarh, Madhya Pradesh, Uttarakhand, and Uttar Pradesh. Bihar, Jharkhand, Odisha, and West Bengal make up the Eastern zone, while the Western zone comprises Goa, Maharashtra, Gujarat, and the UTs of Dadra & Nagar Haveli and Daman & Diu. The Southern zone includes Andhra Pradesh, Tamil Nadu, Karnataka, Kerala, and the UTs of Andaman and Nicobar Islands, Lakshadweep, and Puducherry.

Crime rate and trends

The rape crime rate was calculated as the number of rapes per 100,000 females in the population and presented in tabular format at the national level and graphically at the zonal level.

Spatial and spatio-temporal clusters

Spatial scan statistics, an important cluster detection method, detects and evaluates the statistical significance of localised clusters in space, that cannot be explained, traditionally, by an underlying probability model defined by a null hypothesis of complete spatial randomness. The spatial scan statistic proposed by Kulldorff (1997) is a maximum likelihood ratio test statistic based on the numerous variable-size circular windows scanning the entire study region. Space–time scan statistic is a three-dimensional extension of the spatial scan statistic, which detects and evaluates clusters, in both space and time, that might have appeared anytime during the study period (Kulldorff et al., 1998). Spatial scan statistics has been widely utilized in fields such as spatial epidemiology, accidents, criminology, demography, disease surveillance, forestry and others (Costa & Kulldorff, 2009).

Spatial scan statistic based on the discrete Poisson model (Kulldorff, 1997) was used to detect statistically significant spatial clusters with a high risk of rape by year. To detect spatio-temporal clusters with high risks, a retrospective analysis using the space–time scan statistic based on the discrete Poisson model was performed (Kulldorff et al., 1998). Both analyses were performed in SaTScan v9.6 (https://www.satscan.org/) and calculations were based on 999 Monte Carlo replications. Datasets prepared for analyses in SaTScan contained the annual rape numerals (count data) as well as the projected annual female population per district (n = 640) for the years 2011 to 2020.

The maximum spatial cluster size of the circular scanning window for the purely spatial cluster was set to be a circle with a radius of 100 km (the default population size is 50% of the population at risk). For spatio-temporal analyses, a cylindrical scanning window was considered. The radius of its circular geographical base and height corresponding to a time interval were restricted by an upper limit of 100 km and three years, respectively. Among the statistically significant clusters identified based on the P value and log-likelihood ratio test (alpha = 0.05), the most likely (primary) cluster was the one with the maximum likelihood ratio. If a potential cluster detected was not statistically significant, then it can be dismissed as a probable chance occurrence and not a cause for alarm. Secondary clusters were reported based on the criterion of ‘no geographical overlap’ (Kulldorff, 2022).

The study also considered the Mann–Whitney (MW) spatial scan statistic based on the Mann–Whitney statistic (Cucala, 2016) to detect the most likely high clusters of rape for each year. The distribution of rape rates in India (continuous in nature) was skewed during the years of the study period. Therefore, the nonparametric MW spatial scan statistic which does not make any assumptions regarding the distribution of the continuous data was also considered. The analysis was performed in R software and the maximum radius of the circular scanning window was kept at 100 km. Statistical significance was based on the P value calculated through 999 Monte Carlo replications. P value < 0.05 was considered statistically significant.

Cluster frequency

Cluster frequency was analyzed by examining the number of times a particular district was part of either a primary or secondary statistically significant spatial cluster (based on the Poisson model) detected during the study period 2011–2020. The analysis was done to understand the patterns of persistence, if any, in the spatial clustering of high-risk districts.

The main results of spatial, spatio-temporal as well as cluster frequency analyses were summarized in tabular format. The results of cluster frequency analysis and the statistically significant spatial clusters detected based on the Poisson model by year were also mapped using QGIS 3.10.11 software (QGIS.org, 2020).

Results

Trends in the rate of rape crime

The overall rape crime rate in India (per 100,000 women) shows a mixed trend, with an increase from 2011 to 2014, and a decline from 2016 to 2020 (see Table 1). There was a decline of 7.4% in the year 2015 over 2014 and an increase of 10.6% in 2016 over 2015. The rape rates are at their lowest value during the study period in 2020. State-wise, Mizoram reported the highest rape rates in 2011 and 2012, Delhi from 2013 to 2015, and Sikkim in 2016. Madhya Pradesh had the highest rape rate in 2017, A & N (Andaman & Nicobar) Islands in 2018, Chandigarh in 2019, and Rajasthan in 2020. During the study period, the highest value of rape rates (per 100,000 women), state-wise, ranged between 13.47 in 2020 to 29.99 in 2016. Figure 2 depicts the zone-wise graphical representation of the trends in rape crime rates from 2011 to 2020 for the states and UTs in India.

Table 1 Trends in annually reported rapes of women in India, 2011–2020
Full size table
Fig. 2
figure 2

Trends in rape rates (per 100,000 women) for the states and UTs in India during 2011–2020 as per the zonal classification-A. Central zone, B. Eastern zone, C. North Eastern zone, D. Northern zone, E. Southern zone, and F. Western zone

Full size image

Spatial clusters of rape from 2011 to 2020

Kulldorff’s spatial scan statistic

The number of significant clusters detected by the spatial scan statistic increased from 39 in 2011 to 51 in 2016 and fluctuated between 47 and 41 in the years 2017 to 2020, as seen in Figs. 3, 4 and Table 2. The number of districts contributing to the annual most likely cluster ranged from 6 (in 2020) to 20 (in 2019). Most likely clusters in 2011 and 2012 were concentrated in the Eastern and North Eastern zones (West Bengal and Assam in 2011; Assam and Meghalaya in 2012).

Fig. 3
figure 3

Statistically significant high rape spatial clusters by year during 2011–16

Full size image
Fig. 4
figure 4

Statistically significant high rape spatial clusters by year during 2017–2020

Full size image
Table 2 High rape spatial clusters detected using spatial scan statistic based on discrete Poisson model
Full size table

Most likely clusters were mainly concentrated in the Northern (Haryana, Rajasthan, and Delhi) and Central zone (Uttar Pradesh) from 2013 to 2019 except in 2016 (when Rajasthan was not part of the cluster) and 2020 when the most likely cluster was distributed across six districts of Rajasthan. From 2014 to 2018, with exception of 2016, the location of the most likely cluster remained constant and was spread across Delhi, six districts from Haryana, and two districts each from Rajasthan and Uttar Pradesh.

The total number of districts contributing to the spatial clusters during the study period ranged between 150 to 205. Though considering the contribution of districts, the extent of clustering was mainly distributed across the Central, North, Eastern, and North-Eastern zones, there was a difference in the size and spatial location of the clusters across the years (Figs. 3, 4). In the Southern zone (except in 2020, when only districts of Andhra Pradesh were included), the likely clusters were mainly distributed across districts of Kerala.

MW spatial scan statistic

The number of districts in the annual most likely cluster detected using the MW scan statistic ranged from nine to 23 (see Table 3). The most likely cluster in 2011 detected by MW scan statistic, a slightly bigger cluster compared to Kulldorff’s method, strictly covered only districts of the North Eastern zone (Assam and Meghalaya). The detected cluster had several districts in common with the first two likely clusters from the North Eastern zone in 2011 and the primary cluster in 2012 identified by Kulldorff’s method. Barpeta, Bongaigaon, and Goalpara districts of Assam contributed to clusters detected using both methods for the two years.

Table 3 High rape primary spatial clusters detected using MW scan statistic
Full size table

Though the most likely cluster detected by the MW scan statistic in 2012 was distributed across the North Eastern zone like Kulldorff’s spatial scan statistic, it shifted in location to cover districts from Assam, Mizoram, and Tripura. The second likely cluster from the North Eastern zone identified by Kulldorff’s spatial scan statistic in 2012 covered that exact location. The third likely cluster from the North Eastern zone identified by Kulldorff’s method in 2011 covered several districts in common with the primary cluster identified through the MW method in 2012.

The MW scan statistic detected a slightly bigger primary cluster in 2013 than the one detected by Kulldorff’s method. The cluster was concentrated in the Northern (Haryana, Rajasthan, and Delhi) zone covering twenty districts. In 2014, the location of the most likely cluster detected by the MW method was spread across the Northern and Central zones covering Delhi, six districts from Haryana, two districts from Rajasthan, and two districts from Uttar Pradesh. In 2015 and 2020, the location of the cluster remained the same as in 2013. In 2016, the cluster was spread across the northern and central zones covering Delhi, nine districts from Haryana, and three from Uttar Pradesh.

Similarly, in 2017, the location of the most likely cluster (though a bigger cluster compared to Kulldorff’s method) was spread across Delhi, seven districts from Haryana, two districts from Rajasthan, and five from Uttar Pradesh. In 2018 and 2019 the location of the most likely cluster remained constant and was spread across Delhi, ten districts from Haryana, one district each from Rajasthan and Uttar Pradesh.

From 2013 to 2019, the primary clusters were located across the Northern and Central zones. In 2020, the primary cluster detected by the MW method strictly covered districts of the Northern zone, like that by Kulldorff’s method. Though the latter cluster included only districts of Rajasthan, the former was a much bigger one covering ten districts of Haryana, Delhi, and one district from Rajasthan.

Spatial pattern of rape clusters

Figure 5 and Table 4 illustrate the frequency of spatial rape clusters (Kulldorff’s spatial scan statistic based on the Poisson model) across districts during the study period. Approximately, 19.4% of districts (n = 640) contributed to high-risk clusters for more than six years. It is observed that districts in the Central, Northern, and North Eastern zones of the country contributed most frequently to the clusters detected from 2011 to 2020.

Fig. 5
figure 5

Persistent spatial pattern of rape clusters across districts of India, 2011–20

Full size image
Table 4 Persistence of spatial rape clusters across districts of India, 2011–20
Full size table

In the Central zone, districts from Madhya Pradesh and Chhattisgarh contributed the most to high-risk clusters. The persistent spatial clusters from the Eastern zone were spread across districts of Jharkhand and Odisha. From the Northern zone, the entire Delhi and Chandigarh had cluster frequencies of ten and nine, respectively. Apart from Delhi, Haryana, and Rajasthan contributed the most to the persistent clusters during the study period. In the North Eastern zone, districts from Assam contributed mostly frequently to the high-risk clusters.

In the Southern zone, Hyderabad from Andhra Pradesh persisted as a rape cluster during the ten years, while districts from Kerala contributed the most to the high-risk rape clusters. Goa and Maharashtra contributed mostly to the spatial clusters detected in the Western zone.

Spatio-temporal clusters of rape

Seventy-six statistically significant spatio-temporal clusters (one most likely cluster and 75 secondary likely clusters) were detected in India from 2011 to 2020 with the highest log-likelihood ratio (LLR) being 5560.09 (P < 0.001) and the smallest being 13.34 (P < 0.05). The most likely cluster was observed within the time frame 2014 to 2016 covering 19 districts mainly from the northern zone (see Table 5).

Table 5 Spatiotemporal clusters of rape in India (only the first ten out of the 76 statistically significant clusters described)
Full size table

A total of 59 clusters were observed across time frames (n = 8) spanning 3 years. The highest number of clusters was detected within the time frame 2018 to 2020 (n = 14) and the least within 2011 to 2013, 2015 to 2017, and 2017 to 2019 (n = 3). The remaining 17 clusters were observed across time frames of length two years (n = 5) and one (n = 6).

The highest number of districts covered by the spatio-temporal clusters were from the Northern (n = 78), Central (n = 64), and North Eastern zone (n = 52). The detected most likely cluster was situated across the same location as the most likely spatial clusters from 2014 to 2018, with exception of 2016. The clusters observed within the time frame 2015 to 2017 covered only districts of Odisha, while the clusters within the time frame 2016 to 2018 were distributed mainly across districts of the Central zone.

Discussion

The present study applied spatial scan statistics to investigate the spatial and spatio-temporal patterns of rape from 2011 to 2020 at the district level in India. The results of the study indicate significant local clustering of high rape rates mainly across geographically contiguous regions across time and space in the country. The findings agree with results from previous studies that observed the localization of CAW (including rape) in India (Bala, 2007; Kabiraj, 2022; Krishnaswamy et al., 2013; Mukherjee et al., 2001), albeit with a different methodology. Previous studies have investigated the temporal and spatial pattern of rape and CAW in India mainly through trend analysis and mapping through tabular, graphical, and cartographic representation methods (Bala, 2007; Bhattacharyya et al., 2021; Chauhan & Baraik, 2016; Chhachhiya, 2017; Dandona et al., 2022; Kabiraj, 2022; Kahlon, 2014; Krishnaswamy et al., 2013; Lolayekar et al., 2022; McDougal et al., 2021; Mitra & Saha, 2018; Mukherjee & Basu, 2017; Mukherjee et al., 2001). The districts/states with high rates of crime were mapped based on cut-off values which differed across studies (Jha, 2015; Krishnaswamy et al., 2013; McDougal et al., 2021; Mukherjee et al., 2001). Simple statistical techniques like summary measures and the composite index were also used to describe the patterns of the crime (Bala, 2007; Chaudhary, 2015; Dey & Modak, 2015; Jha, 2015; Maity & Roy, 2021; Roy & Dastidar, 2018; Sharma et al., 2018; Singh, 2016; Wani et al., 2022). Kabiraj (2022) used Moran’s I index to test for spatial autocorrelation in the pattern of CAW. The K-means clustering method was used by Jasper and Govardanan (2022) to identify the state-wise clusters of CAW for 2013, while Singh et al. (2020) used K-means to identify clusters of different CAW including rape for the 2001–2010 period at the state level. Different spatio-temporal areal models were also used to study districts with a high risk of certain CAW including rape in selected Indian states (Vicente et al., 2018, 2020, 2021). The present study is distinct from the earlier studies on rape crime in India for the following reasons: First, the study made use of the cluster detection technique known as spatial scan statistic, which not only detects but also evaluates the statistical significance of the identified local clusters. Second, the NCRB rape data for all the years from 2011 to 2020 were studied, and significant high-risk spatial and spatio-temporal clusters of rape in India were identified at the district level. Third, the high-risk districts of rape in India were identified based on their contribution to spatial clusters during the years of the study period.

From the results of spatial analysis using Kulldorff’s spatial scan statistic, a variation in the size and location of significant spatial clusters can be seen over the years. The clustering was evident across the Central, North, North Eastern, and Eastern zones. In the Southern zone, except in 2020, the likely clusters were concentrated in Kerala. The primary high rape spatial clusters (both Kulldorff’s and MW methods) indicate a shift in the pattern during the study period, with clusters located across the Central and Northern zone from 2013 onwards, while in 2011–12 wherein the primary clusters were spread across the North Eastern zone. Thirty-seven districts of India from the North (17), Central (12), North East (7), and South (1) showed a tendency to cluster in all years of the study period.

The spatio-temporal clusters were majorly located across the Northern, Central, and Eastern districts of India. The districts from the North Eastern zone were part of spatio-temporal clusters spanning only the first half of the study period, except for Sikkim during 2016. In sharp contrast to other years, Sikkim had the highest rape rate in the country in 2016. An interesting trend seen in the data for Sikkim is that while the North district of Sikkim had the lowest rape rates in the first half of the study period, during the second half, the district had higher rape rates than the West district of Sikkim.

Jha (2015) mapped the average rate of rape crime from 2008–2012 and found that the very high (> 5.2) crime belt included Mizoram and Tripura, while the high crime belt (3.7 to 5.2) was visible in two regions including Madhya Pradesh. The other region included Meghalaya and Assam. In the present study, certain districts of Mizoram, Tripura, Assam, and Meghalaya were part of several top ten secondary clusters in both 2011 and 2012, respectively. The second most likely cluster in 2011 and 2012 included districts of Madhya Pradesh, with more than sixty percent of districts contributing to spatial clusters in both years.

A study by Vicente et al. (2018) found that districts in the West and Central part of Uttar Pradesh presented with high incidence risks for rape in 2013–14. During this period, a total of three districts from West and South-West Uttar Pradesh were part of spatial clusters in the present study. Another study by Vicente et al. (2020) focused on rape and dowry deaths in Uttar Pradesh during 2001–2014 using a multivariate spatio-temporal model and observed a Northwest- Southeast gradient for high risk of rape with a majority in the Northwest. In the present study, three districts from the South, and two districts from the Northwest were part of statistically significant spatial clusters during 2011–14. In a similar study, Vicente et al. (2021) observed a clear Northeast- Southwest gradient for high risk of rape in Maharashtra upon analyzing data from 2001–2013 with a majority in the Northeast. In agreement, the current study also found that the districts of Maharashtra which contributed to spatial clusters during the years 2011 to 2013 were strictly from the Northeast and Southwest parts of the state.

A study by Maity (2019), while examining the relative performance of 28 major states, in controlling rape for the two decadal periods marked by 2001 and 2011, found that in both time points Gujarat (from the Western zone) ranked first. Only two districts of Gujarat were ever part of high rape clusters identified in the present study. For the decade ending in 2011 (basis of discussion hereafter), Goa and Kerala were the worst performers from the Western and Southern zones, respectively. Accordingly, the present study found these two states to be the only ones from the corresponding zones to have contributed more than 70% of the districts to spatial clusters in more than five years of the study period. Madhya Pradesh and Chhattisgarh were the poor performers in rape control in the Central zone while Uttarakhand performed best. According to the present study, 70% (n = 50), 61.1% (n = 18), and 0% (n = 13) of districts from these three states, respectively, had a cluster frequency greater than five. According to Maity (2019), Bihar was the best while Jharkhand, and Odisha were the poor performers from the Eastern zone. Only Katihar district from Bihar was ever part of a spatial cluster (2012), while 79.2% (n = 24) and 86.7% (n = 30) of districts from Jharkhand and Odisha, respectively, were part of spatial clusters during the study period.

Two states each from the Northern zone occupied nearby positions in terms of best and worst performers in rape control. While 6.7% (n = 30) of districts of Jammu Kashmir had cluster frequency greater than four, for Punjab it was 15% (n = 20). According to Maity (2019), Haryana performed slightly better than Himachal Pradesh in terms of efficiency ranks (19 versus 20). The present study found that the percentage of districts with cluster frequency greater than four in the two states, respectively, were 66.7% (n = 21) and 33.3% (n = 12). In North Eastern zone, Assam (n = 28) was the worst performer followed by Mizoram (n = 8) and Tripura (n = 4), while the best performer as per Maity (2019) was Nagaland (n = 11) followed by Manipur (n = 9). The percentage of districts from these states which ever appeared in any spatial cluster during the study period was 100 (Assam, Mizoram, and Tripura), 44.4 (Manipur), and 18.2 (Nagaland). Considering space–time clusters in a 3-year moving window from 2011 to 2015, the maximum contribution was from districts in the North Eastern zone. All eight states from the North Eastern zone contributed to spatial clusters from 2011 to 2013.

Combining the results from spatial as well as spatio-temporal analyses, pertinent questions arise, such as, what changed for Sikkim in 2016 to have the highest reported rape rates in the country? Why Hyderabad in Andhra Pradesh persisted as a cluster for all years of the study period? What are the underlying reasons behind the spatial contiguity in the distribution of rape in the Northern and Central zones?

Female precipitation, male pathology, male sexuality, and male hostility form the individual causes of rape, while the societal causes include society/socialization (social processes) and male dominance (Cowan, 2000). In the Indian context, the first three causes, which also function as rape myths, provide indirect social approval and contribute to the rising incidence of rape (Jha & Sarangi, 2014).

According to several studies (Maity, 2019; Mitra & Saha, 2018; Qureshi & Kim, 2018; Roy & Dastidar, 2018), the social perspectives towards women (such as patriarchy, unhealthy gender relations, and roles) are the leading reasons behind rape crime in India. There is state-wise heterogeneity in system inputs and outcomes related to rape control; lack of investment and less efficient performance of the states could be the reasons for poor rape control outcomes (Maity, 2019). The study also explains that states with similar system inputs and control infrastructure but differ in different socio-economic indicators also differ in rape control. The regional variation in socioeconomic, demographic, cultural, and physical characteristics as well as gender equality (Chauhan & Baraik, 2016; Jha, 2015; Woetzel et al., 2015) could also be one of the reasons behind the spatial variation in the distribution of rape cases in India. Several studies have investigated various factors like female literacy, sex ratio, female workforce participation, the presence of women police stations, urbanization, etc. However, concerning India, for most of these possible causative factors, no clear relationship can be established due to the heterogeneity in the distribution as well as in the combination of the (levels of) factors across regions (Bhattacharyya et al., 2021; Chauhan & Baraik, 2016; Chhachhiya, 2017; Jha, 2015; Kaur et al., 2016; Krishnaswamy et al., 2013).

The regional changes in the temporal distribution of rapes across India could be due to the effects of various regional and national policies (Vicente et al., 2018) and events, as observed in the case of Delhi in 2013 and Sikkim in 2016. An increase in reported rape rates, post Nirbhaya case in 2012, (McDougal et al., 2021) can be seen especially in Delhi and states closer to it. The present study observed an increase of 127% in 2013 over 2012 for Delhi and an increase of 1717.6% in 2016 over 2015 for Sikkim. Possible reasons for the steep rise in rape rates could be the strict instructions given by the state government to the police and awareness among the public following a high-profile rape case in North Sikkim in 2016 (Press Trust of India, 2016); An amendment to the state victim compensation scheme for rape survivors in Sikkim also came into effect in the same year (sikkimslsa.nic.in). Hence, the need for targeted studies focusing on identifying the localized clusters of high rape to understand the dynamics of rape in India.

Limitations

The limitations of the present study are mainly related to shortcomings in the available and vastly underreported rape crime data. The reported cases of rape need not necessarily be an actual reflection of the true incidence of rape crime in India (Lolayekar et al., 2022). As in many other countries, this discrepancy with actual data could be due to “under-recording by state agencies and under-reporting by women survivors” (Pickup et al., 2001). Due to the nonavailability of data in 2019 for West Bengal at the district level, data for 2018 was used. The second limitation is that NCRB only provides year-wise data regarding rapes in India. No other temporal information is available for detailed analysis. The third limitation is the denominator used for calculating rape rates. The linear interpolation technique was used to obtain the ‘complete’ female population for the years during the study period. Another potential limitation could be the use of circular scanning windows in detecting irregular clusters. Despite these limitations, the results of the present study could provide valuable information regarding the regional spatial distribution of rape in India and be beneficial to rape control strategies.

Conclusions

The present study detected localized spatial and spatio-temporal clusters of high rape rates in India, at district levels. The study also investigated the pattern of persistence across districts and identified the high-risk areas. The detected spatial and spatio-temporal local clusters were majorly spread across the districts in the Central, North, North Eastern, and Eastern zones of India during the study period. While around 23.6% of all districts in India contributed to spatial clusters in more than half of the study period, around 43.8% of the districts were not part of any cluster. The heterogeneous distribution of rapes in India, both in space and time, could be due to the spatio-temporal regional variations in different determinants of rape. Identification of the localized high-risk districts will be beneficial in understanding the numerous factors of rape and their combinations operating at regional levels. This contextual information is relevant to the formulation of various national and regional policies to control rape in the country. The results of the current study could also aid State governments as well as law enforcement agencies in designing and investing in rape control strategies focusing on these high-risk districts so that there is an optimal allocation of resources. The Zonal councils of India could make efforts to understand the dynamics of the clustering of rape crime in India. Future research should focus on the reasons behind the increased presence of rape clusters in certain zones, as well as the absence or diminished presence in the Western and Southern zones. As rape is often a severely under-reported crime, the absence of localised rape clusters does not necessarily mean the absence of the crime in an area. Investigating the spatial variations in rape reporting behaviour across the country could also be of future interest. Rape is a blatant violation of a woman’s basic human rights and for a developing country like India, the social, public health, and economic dimensions of rape could impede development.