The Effects of India’s Gender Quota in Local Government on Rates of Reporting Rapes of Women from Scheduled Castes and Tribes

Nadia Kale

Advisor: Professor Anna Harvey

New York University

Politics Honors Thesis

Abstract

This research project asks whether increased female representation in India increases rates of reporting rapes of women from scheduled castes and tribes. The impact of female representation on the incidence of violence against women has yet to be extensively explored, due to the nonrandom assignment of female representation across electoral districts. In India, however, the Panchayati Raj Act of 1993 introduced a quota system in local levels of government, mandating that one third of seats be reserved for women. For several idiosyncratic reasons, the Act was implemented in different states at different times, creating increases in women’s representation that were as-if random. One recent study looked at the gender quota’s impact on crimes committed against all women and found an increase in rates of reporting (Iyer 2012). However, this study did not account for caste and socioeconomic distinctions that may influence which women are empowered to report. In exploring the impact of mandated increases in female representation on rates of reporting rapes of India’s most marginalized women, this project finds that while gender quotas have a positive impact on rates of reporting amongst all women, the same does not hold true for women from scheduled castes and tribes.

Acknowledgement

I am extremely grateful to Professor Harvey and Hannah Simpson for the time they dedicated to teaching me about quantitative methods and for their guidance in helping me write my thesis. What I have learned this year surpassed my academic goals and expectations and I hope to continue incorporating quantitative analysis into my future studies!

Statement of Research Question

My project asks whether increased female representation in India reduces violence against women from scheduled castes and tribes. The impact of female representation on the incidence of violence against women has yet to be extensively explored, due to the nonrandom assignment of female representation across electoral districts. In India, however, the Panchayati Raj Act of 1993 introduced a quota system in local levels of government, mandating that one third of seats be reserved for women. For several idiosyncratic reasons, the Act was implemented in different states at different times, creating increases in women’s representation that were as-if random. One recent study looked at the gender quota’s impact on crimes against women and found an increase in rates of reporting all types of crimes committed against women (Iyer 2012). However, this study did not account for class distinctions that may influence which women are empowered to report crimes committed against them. My project will explore the impact of the increased female representation mandated by the Panchayati Raj Act on rates of reported rapes of women from scheduled castes and tribes. This project will allow me to assess the impact of increased female representation on India's most marginalized women.

Literature Review

Rape is one of many violent crimes perpetrated against women by both known and unknown attackers, though the former is far more likely. In a telephone survey conducted from 1995 to 1996 on a total of 8,000 women in the District of Columbia, Kruttschnitt and Macmillan found that over three-quarters (78%) of attackers in violent crimes against women are known to their female victims (“Patterns of Violence Against Women: Risk Factors and Consequences” (2005)). Most studies of rape and other violent crimes against women thus focus on factors motivating victims’ intimate partners to commit such crimes.

Reported rates of violent crimes against women are the product of two kinds of factors: those responsible for the actual crimes, and those responsible for the rates of reporting these crimes. Most studies focus on the former set of factors. However, a few studies have looked at factors motivating reporting of violent crimes against women.

A theory often cited in explaining causes of violence against women is a community or region’s poverty level and/or relative level of development. Among the first quantitative analyses of the relationship between poverty and violence against women was Miles-Doan’s article, “Violence Between Spouses and Intimates: Does Neighborhood Context Matter?” (1998). In this study, the dependent variable analyzed is reported incidence of domestic violence, which includes a number of acts of aggression, such as rape. Using law enforcement data from 1992 and the 1990 census data from one county in Florida with exceptionally high death rates due to violence, Miles-Doan found that within Duval county, “neighborhoods with a high concentration of residents living in poverty [and] unemployed males…have drastically higher rates” of domestic violence than neighborhoods with comparatively lower concentrations (Miles-Doan, 1998; p. 637). The causal explanation that Miles-Doan presents to explain this observed neighborhood effect is that one’s geographical location influences the networks that one operates within and will consequently affect “prospects for employment, for public services, for educational advancement… and much more” (Miles-Doan, 1998; p. 626). In cases where prospects are low, the assumption is that there will be higher rates of domestic violence, because unemployment, a lack of public services and resources, and low levels of education are all risk factors associated with both domestic violence, and violence against women more generally (Campbell, 2005; Kruttschnitt, 2006).

Despite Miles-Doan’s findings in favor of neighborhood effect theories, there is no way to thoroughly distinguish whether the observed results are truly due to a neighborhood effect, which asserts that it is the poverty and underdeveloped nature of a specific community that motivates a higher number of domestic violence cases. The reason that no concrete conclusions can be determined is because the conditions of each neighborhood are non-random, which presents a selection problem. Without randomization of economic conditions, it is not possible to accurately deduce what effect a neighborhood’s level of development or affluence has on rates of domestic violence, because extenuating factors that may influence neighborhood conditions may also be exogenous variables that affect levels of domestic violence. One possibility is that a historically high concentration of families with issues of domestic violence could affect a neighborhood’s economic conditions, which would imply reverse causality.

Another investigation of the relationship between economic conditions and domestic violence is Aizer’s article, “The Gender Wage Gap and Domestic Violence” (2010). Aizer’s study uses an instrumental variable design, exploiting “exogenous changes in the demand for labor in female-dominated industries” to estimate the effects of a decreasing male-female wage gap on domestic violence (Aizer, 2010; p.1). Aizer’s measure of the female-male wage gap is constructed to reflect a particular county’s proportions of male and female workers in a given industry in that county. This is then indexed by the statewide wage for that industry, which Aizer argues makes the measure of the wage gap exogenous, because of the fact that she is using state-wide wages averaged across all industries, as opposed to using county-specific wages, which would not be random when observing those counties.

The instrumental variable used in Aizer’s investigation is derived from the same strategy of indexing county-specific proportions of workers in a given industry by statewide growth in that industry. Again, the argument is that this variable is exogenous to county-specific conditions, because Aizer uses statewide growth in an industry, as opposed to county-specific growth.

Aizer ultimately finds that over a span of fifteen years, from 1995-2010, violence against women declined as employment and earnings amongst women increased. More specifically, Aizer concludes that a decline in the wage gap witnessed over the same time period can explain 9 percent of the reduction in violence against women (Aizer, 2010; pp. 18). As with Miles-Doan’s article, Aizer’s dependent variable is rates of domestic violence, which includes intimate partner rape. The problem with Aizer’s analysis, however, is that the proportion of workers in a given industry could be dependent on a number of variables that are not accounted for in this research design. One possibility is that in areas where there is more domestic violence, one might observe a higher proportion of women working in service jobs, which are historically considered part of a lower wage female-dominated industry.

Further exploring the association between economic development and violence against women, Hackett’s article, “Domestic Violence Against Women: Statistical Analysis of Crimes Across India” (2011), uses the National Crime Records Bureau of India’s “Crimes against Women” data to analyze how a state’s level of development impacts certain types of crime rates against women. Using multivariate linear regressions involving a number of development indicator variables, Hackett looks specifically at dowry deaths and cruelty (wife abuse) to analyze potential causal effects. Both dowry deaths and cruelty are forms of intimate partner violence perpetrated against female victims by members of their immediate family. Although not limited to sexual violence, cruelty as a form of intimate partner violence accounts for rapes committed against married women by their partners and other family members.

The independent variables that Hackett employs, each of which groups together a number of variables within them, are human development, gender-equality development, and urban development. Here, Hackett uses a factor analysis approach, taking a number of state indicators, such as female literacy rate, child sex ratio, percentage of population with electricity, and female employment, and weighting them into three groups, with each group representing a specific type of development. Results indicate that states with higher rates of urbanization, health, and education have lower rates of dowry death and cruelty. Further, in regards to the gender-equality development factor, Hackett found that the less developed a state is in terms of gender-equality, the higher the incidence of dowry deaths and cruelty.

One problem with Hackett’s study is that the independent variables identified using the factor analysis approach are nonrandom across the Indian states being analyzed, which creates a causal inference problem. Any number of extenuating factors could impact one of Hackett’s three independent variables, as well as cruelty and dowry death rates.

Contrary to Hackett’s findings, Johnson proposes that improved socioeconomic conditions might instead lead to increases in violence against women. In his article, “Rape and Gender Conflict in a Patriarchal State” (2014), Johnson examines the empirical relationship between female socioeconomic and political power and rape rates in Kansas. He hypothesizes that as women begin to progress towards equality, both economically and politically, men react to increased competition within their community cohort by trying to thwart such advancement and asserting their dominance and superiority. Johnson’s results suggest that there is a positive and strong correlation between county rape rates and female sociopolitical power. According to Johnson’s findings, controlling for overall violent crime rates and specific county characteristics, an increase in the number of female-headed households, female-owned businesses, and female politicians and police officers motivated an increase in rapes across all Kansas counties.

However, Johnson’s causal story is inconclusive. Because the increase in female sociopolitical power, which Johnson observes through a number of variables, is non-random across the counties within Kansas, no accurate conclusions can be drawn as to how increased socioeconomic status for women affects rape rates across the state. Moreover, Johnson does not further investigate the possibility that the increase in rates is actually an increase in reporting of crimes committed against women, which in itself could be a result of increased socioeconomic and political status of women in the historically patriarchal state of Kansas. Unlike Aizer and Hackett’s articles, where the argument is that an increase in status for women, whether in levels of bargaining power or equality, leads to a consequent decrease in rates of violence against women, Johnson’s article does not consider the psychological effects of elevated status on women. Thus, his assumption of backlash needs to be further explored.

Similar to Johnson’s findings, Iyer et al’s article, “The Power of Political Voice: Women’s Political Representation and Crime in India” (2012) finds that increased political representation for women in local government increases rates of violence against women across states in India. Despite these results, however, Iyer et al conclude, after further analysis, that the observed increases in rates of violence against women are actually increases in rates of reporting, which suggests a much more positive effect of rising status for women in patriarchal societies.

In order to investigate the effects of increased political representation for women on rates of violence against women, Iyer et al take advantage of India’s as-if random gender quota, which eliminates the problems of non-randomization that Johnson faced in his study. India’s gender quota was mandated for all 17 of India’s major states in the 1993 amendment to the Panchayati Raj Act. This presents an opportunity for an as-if random analysis, because the reservation of seats was mandated at the federal level, so the sudden increase in female political representatives is consistent across states and cannot be attributed to variation in state conditions, which otherwise might affect each state’s rates of violence against women. Further, the variation in dates of implementation of the reservations for women across states addresses potential endogeneity of the passage of the Panchayati Raj Act itself. Unlike other policy implementations used as treatments that may have been implemented due to a specific incident occurring at a particular time or in a particular state, the Panchayati Raj’s 73rd amendment was passed and implemented over a span of years with no potential confounding variable that both led to its initiation and will also impact crime rates against women.

Comparing state-level crime rates pre and post reservations for women, Iyer et al gauge the impact of increased female political representation while controlling for a number of factors, such as literacy rates, per capita incomes, male-female ratio, level of urbanization, and size of police force. Iyer et al’s independent variable was reservations for women, and their dependent variable was overall rates of violence against all women across Indian states.

Results from these regressions indicate that “political representation for women is associated with a large and significant increase in the documented crimes against women” (Iyer, 2012; p.176). There are two possible causal mechanisms to explain these relationships. The first is backlash theory, which assumes that the increase in crimes against women is a result of a rise in hostility towards women due to their rising status, or that in the presence of increased representation for women, there is a decrease in overall law and order. The second potential causal mechanism is that the observed increase in documented crimes is actually an increase in rates of reporting. This hypothesis assumes that with an increase in female representation, more women feel empowered to come forward and report violent crimes committed against them. In this case, increased levels of confidence that a victim’s claims will be handled responsibly and that potential backlash from reporting is no longer a threat are possible explanations.