Can Wage Legislation Lead To Women s Social Empowerment? A Difference-in-Difference Analysis of the Minimum Wage Law for

Transcription

1 Can Wage Legislation Lead To Women s Social Empowerment? A Difference-in-Difference Analysis of the Minimum Wage Law for Domestic Workers and Rates of Domestic Violence in India Ankita Banerjea International Relations Honors Thesis New York University Class of

2 Contents 1 Abstract 3 2 Acknowledgments 4 3 Introduction 5 4 Hypothesis 8 5 Literature Review 8 6 Background 14 7 Research Design Data Difference-in-Difference Estimation Methodology Results 27 9 Conclusion References 34 2

3 1 Abstract This paper investigates the impact of the establishment of a minimum wage for domestic workers in the state of Karnataka, India on the subsequent rates of domestic violence faced by women in said state. Given that domestic work is a largely female-dominated industry within India, this minimum wage establishment allows us to explore whether legislative changes of this form could have an effect on the social empowerment of women in terms of their bargaining power within the household. Using a difference-in-difference approach, I use the establishment of the minimum wage law for domestic workers as an exogenous source of variation and isolate its effect on the rates of domestic violence. The results of my analysis indicate that the treatment effect is negative and significant on the entirety of the population surveyed within the state, indicating that the minimum wage establishment led to a reduction in the rates of domestic violence experienced post-treatment. These results contribute to policy debates that are often divided between sides that believe economic and legislative reform can have a positive effect on women s empowerment and others that believe such measures make women s bargaining position worse. 3

4 2 Acknowledgments I would like to thank my thesis advisor Professor Michael Gilligan for his continued support and meaningful advice through this process. I am extremely appreciative of his dedication to his students and can safely say that this thesis would not have been possible without him. I would also like to thank my teaching assistant Hannah Simpson for her invaluable input with the statistical analysis component of this paper and her continued willingness to assist me through this process. 4

5 3 Introduction The issue of domestic violence, especially with the domestic partner and other family members being the perpetrators has been studied at length in a variety of disciplines, primarily sociology and criminology. While this is an issue that prevails world over, I am particularly interested in investigating the issue in the context of India. The vast body of literature that exists has over the past few decades has identified a number of community, household and individual-level risk factors that determine rates of domestic violence in India. Traditionally, higher levels of education and higher socioeconomic status are considered protective factors against women s risk of domestic violence. On the other hand, lower levels of education, behavioral and cultural factors such as higher alcohol intake of the husband, lower levels of dowry and lower number of male births have been associated with increased rates of violence. In recent years, there has been considerable interest in women s empowerment and how that links to domestic violence within the household. There is a debate among scholars on whether the effect of empowerment programs such as group-based saving and credit programs is positive or negative, with many arguing that they may actually increase rates of violence in the household, the so-called male backlash theory. A less studied question is whether economic conditions and indicators impact rates of domestic violence. In that vein, this paper attempts to analyze the impact of the institution of a minimum wage for domestic workers in the state of Karnataka, India on the rates of domestic violence experienced by women residing in the state. Since domestic work is a largely female dominated industry in India, the establishment of the minimum wage legislation in 2004 in this one state provides an opportunity to measure the impact of an exogenous change in female income on the status of women within the household, and see if implied financial empowerment in the form of higher, federally-regulated wages links to social empowerment as well. Figures 1 and 2 below show crime against women in India, and against domestic labor specifically (Fig. 2) based on values between , pointing to 5

6 the continued relevance of this issue and the need to examine whether changes in the past have affected the present rates of violence. Figure 1 Source: National Crime Records Bureau (NCRB), Ministry of Home affairs, Government of India 6

7 Figure 2 Source: National Crime Records Bureau (NCRB); The Wall Street Journal In the following sections, I will state my hypothesis, provide a brief literature review highlighting past works that have studied various facets of the issue at hand, as well as some works that mirror empirical strategies I will be following. I will also discuss the background pertinent to the establishment of minimum wage legislation across India and will be discussing the roll-out of the minimum wage scheme for domestic workers in Karnataka, which was the pioneering state to establish this program. Finally, I will present my data and research design, followed by the results and conclusion. 7

8 4 Hypothesis There are three plausible hypotheses that can be tested through this study. The first hypothesis is that the institution of the minimum wage law led to a decrease in the frequency of beatings experienced by women in the treatment state. This would imply that a formal, higher wage led to an increase in the female bargaining power in the household whereby these women were able to attain a sense of financial independence from their partners, or potentially choose to opt out of abusive relationships, leading to a reduction in the violence. The second hypothesis is that the women living in the treatment state experienced an increase in the frequency of beatings post-treatment, relative to those in the control state. This would be the case if the institution of the minimum wage law increased the bargaining power of the male, relative to the female which could manifest into higher frequencies of violence. This is commensurate with the male backlash theory popularized by several scholars. The third hypothesis is that the minimum wage law had no effect on the frequency of beatings experienced in the treatment state. This could either be because the minimum wage law was insignificant in relation to domestic violence or because the effects on bargaining power could potentially cancel each other out. 5 Literature Review There is a large empirical body of evidence that explores factors influencing violence within the household, often with a variety of different conclusions. In an extensive study for the World Bank in Dhaka, Bangladesh, Rachel Heath finds a positive correlation between being employed and domestic violence for women with low levels of education. These results are consistent with the argument that women with low bargaining power in the household may face violence when trying to enter the labor force for the first time, as the partner tries to oppose or block the potential increase in bargaining power that the woman might 8

9 receive by working. In their study, authors Koenig, Campbell, Jejeebhoy, Stephenson and Ahmed (2006) examine individual and community-level influences on domestic violence in the state of Uttar Pradesh in India. They used several individual-level determinants such as income, education, household assets and area of residence as their independent variables and using multilevel modelling, were able to estimate the effect of those factors on both physical and sexual abuse. An interesting implication of their results was that while higher levels of education among both men and women proved to be protective sources against physical violence, there was no such implication for sexual violence. Inter-generational transmission of domestic violence also seemed to be of importance as the authors found that husbands who had witnessed their mothers being beaten by their fathers were 4.7% more likely to beat their own wives, relative to men who had not witnessed such events 1. While this study points to a number of contextual norms that expose women to higher risks of domestic violence, it does not specifically point to the effect of women s financial literacy or access to employment on domestic violence. Authors Babu and Kar (2009) conducted a population-based study of married men and women across 4 states in Eastern India and interviewed them to collect data on physical, psychological and sexual forms of domestic violence. They employed a multivariate regression model to analyze the effect of several socioeconomic characteristics on prevalence of domestic violence. Similar to other studies, their results also point to higher household income and education levels being associated with lower rates of domestic violence, while several noneconomic factors such as lower caste and religious status are associated with higher prevalence of violence. A key finding from their analysis is that rates of violence are significantly higher for women who are housewives as compared with those who are employed in a salaried job, working as a farmer or even as domestic labor 2. 1 Koenig, Michael A., Jacquelyn Campbell, Rob Stephenson, and Saifuddin Ahmed. Individual and Contextual Determinants of Domestic Violence in North India. N.p., 15 Jan Web. 9 Nov Babu, Bontha V., and Shantanu K. Kar. Domestic Violence against Women in Eastern India: A Population-based Study on Prevalence and Related Issues. BMC Public Health. BioMed Central Ltd, 9 May Web. 10 Nov

10 Several scholars have studied the relationship between male and female incomes and spousal violence in other countries. Economists have used a number of methods to overcome the problem of endogenous wages. For instance, Helen V. Tauchen, Ann D.Witte and Sharon K. Long (1991) use a Stackleberg type model where the assailant maximizes expected utility subject to the stochastic reaction function of the victim 3. Their results suggest that an increase in the male partner s income serves to increase violence in the household over time. With regards to the female employment patterns, an interesting result was found, depending on whether the family had a high-income or whether it had a low-income. In high income families, an increase in the time that a woman spends working is associated with a significant increase in the frequency of violence. In contrast, if a woman from a low-income family spends more time employed or working, the frequency of violence she experiences actually decreases. Perhaps the central conclusion of the study points to the hypothesis that overall, improvements in the woman s opportunities outside of the relationship significantly reduce the level of violence. The authors also study the woman s behavior and reaction to different forms of violence and acknowledge that non-economic rather than employment variables are more closely related to female behavioral patterns if violence ensues. For instance, a woman is more likely to disobey her partner if she has more children, which tends to lead to higher rates of violence in the household. Bowlus and Seitz (2006) use a structural methods approach to discuss what drives men to abuse their wives and what keeps women in some abusive marriages 4. Their data is based on the Canadian Violence Against Women Survey (VAWS), which provides information on marriage, domestic violence in current and past relationships, violence in the family backgrounds of women and their spouses as well as the female s current employment behavior. The authors are able to control for observed and unobserved characteristics by taking advantage of the information on the timing of marriage, abuse and employment, and are 3 Tauchen, Helen, Ann Dryden Witte, and Sharon Long. Domestic Violence: A Non-random Affair. National Bureau of Economic Research (1985): n. pag. Web. 4 Bowlus, Audra J., and Shannon Seitz. Domestic Violence, Employment, And Divorce. International Economic Review 47.4 (2006): Web. 10

11 therefore able to determine the extent to which the correlations observed in the raw data are due to causal relationships 5. Their findings are consistent with the implications of intergenerational transmission of domestic violence as discussed in Koenig et.al and they find that the likelihood of abusing one s wife is times greater, depending on the age of the wife. This paper is particularly interesting as it looks at employment as a dependent variable, and assesses the effect that domestic violence has on the decision to seek employment on the part of the female. In this regard, Bowlus and Seitz find that women are able to reduce the likelihood of violence by working, but only before the abuse actually arises in the relationship. This implies that abused women would be less likely to use employment as a means to prevent further violence, and look to more permanent solutions, such as divorce, instead. Another interesting study using the Canadian VAWS was conducted by Macmillan and Gartner (1999) where they sought to determine the impact of labor force participation on violence against women. The primary causal model tested by them was based on the symbolic relationship of employment whereby employment is viewed as a status symbol within a marriage. Their results suggest that female employment may increase the risk of abuse against the wife if the husband happens to be unemployed, while the risk is significantly lower when the husband, too, is employed. They also included controls such as husband s education with higher levels of education implying lower rates of violence 6. In 2010, Anna Aizer built on the work of her peers and analyzed the impact of the gender wage gap on domestic violence in the United States. Economic theory states that increases in the woman s relative wage increases her bargaining power within the household and is associated with lower rates of domestic violence. To test these predictions and overcome the problem of endogenous wages, she exploits exogenous changes in the demand for labor in female-dominated industries relative to male-dominated ones to estimate the im- 5 Bowlus, Audra J., and Shannon Seitz. Domestic Violence, Employment, And Divorce. International Economic Review 47.4 (2006): Web. 6 Macmillan, Ross, and Rosemary Gartner. When She Brings Home the Bacon: Labor-Force Participation and the Risk of Spousal Violence against Women. Journal of Marriage and the Family 61.4 (1999): 947. Web. 11

12 pact of the gender wage gap on levels of violence 7. Instead of using self-reported values of domestic violence, she uses data on female hospitalizations due to assault, which she believes to be more reliable. Her results prove to be consistent with the traditional bargaining power theory, and therefore, inconsistent with the male backlash theory wherein men retaliate when women are earning higher wages as they find their traditional gender roles in question 8. Wages can be an interesting determinant of a number of societal behaviors, and an extensive amount of literature exists on the effect of changes in wage legislation on other socioeconomic factors. Perhaps one of the most well-known studies of this nature is Card and Krueger (1994) where the authors analyze the impact of the change in the minimum wage in New Jersey in 1992 on employment rates, specifically in the fast-food industry. They use a control group of fast-food restaurants in eastern Pennsylvania, which is extremely similar to New Jersey in terms of its employment climate, and use this to difference out any seasonal employment effects 9. To estimate the change in employment, the authors use a regression model in which employment rate between wave 1 and wave 2 is estimated using a dummy variable that is indicative of whether the store is in New Jersey or Pennsylvania (taking a value of 1 for New Jersey) and a set of characteristics of each store, such as the fraction of full-time workers, number of hours open per weekday, number of cash registers etc. This difference in difference approach leads them to the conclusion that there was no evidence that the rise in the minimum wage in New Jersey led to a decrease in the employment rate. In fact, most of their analysis points to the fact that the increase in minimum wage led to an increase in the employment of low-wage workers in New Jersey. It appears that prices of fast-food did go up during this period which suggests that much of the burden of the higher wages was passed onto consumers, but this cannot conclusively be proven through the model. 7 Aizer, Anna. The Gender Wage Gap and Domestic Violence. American Economic Review (2010): Web. 8 Ibid. 9 Card, David, and Alan Krueger. Minimum Wages and Employment: A Case Study of the Fast Food Industry in New Jersey and Pennsylvania. (1993): n. pag. Web. 12

13 It was this study that inspired me to look more closely at minimum wage legislation in India and tie that in with my area of research interest. While the establishment and enforcement of minimum wage laws in India remains scarce, the most important piece of legislation is the Minimum Wages Act of 1948 through which the minimum wage is periodically determined by both the Central and Provincial governments. Until recently, this act did not have any provision for domestic workers as domestic work was considered to be an informal and personalized service provided within a household, and therefore not included as an employment category 10. However, between the years of four states in India began to enforce minimum wages for domestic workers. Gudibande and Jacob (2015) analyze the impact of this legislation over the four states of Andhra Pradesh, Bihar, Karnataka and Rajasthan. They use data from the National Sample Survey (NSS) of India from years between in order to analyze two surveys from before the treatment and two from after the treatment, the treatment being the minimum wage legislation that started in Using a difference-in-difference empirical model, they estimate their outcome variables between the control and treatment groups. In addition, they also use a matching procedure to find counterparts in the treatment and control group before and after the treatment. Using the matching procedure and the difference-in-difference method collectively, specifically through the difference in difference of the outcome variables between these treatment and control group members before and after the treatment, the authors are able to measure the impact of the minimum wage legislation 11. The outcome variables studied were log wages and unemployment, and they find a statistically significant positive effect of the minimum wage legislation in the treatment states. They also use interaction terms such as a gender and caste interaction to capture any additional discriminatory effects on women of lower caste. The aforementioned positive effects continue to exist even while including the interaction between covariates. 10 Gudibande, Rohan, and Arun Jacob. Minimum Wage Law for Domestic Workers: Impact Evaluation of the Indian Experience. SSRN Electronic Journal (2015): n. pag. Web. 11 Ibid. 13

14 Mathur and Slavov (2016) investigates the impact of two specific legislative actions aimed at empowering women on rates of domestic violence in India 12. Similar to my own paper, they use data from the National Family Health Survey for information on rates of domestic violence. They look at changes in inheritance laws and the establishment of a nation-wide program to increase female political representation and find no significant effect of the first legislation but do find that the possibility of increased political representation increases the likelihood of violence, which they hypothesize to be a result of the male backlash theory or of the women s increased willingness to come forward and report these crimes. While the above mentioned studies are comprehensive in their own way, there doesn t seem to be a study of the effect of the minimum wage legislation for domestic workers on rates of domestic violence in those treatment states in India. In a sense, these studies have all influenced my own, with some focusing specifically on my dependent variable and others on my independent variables. Yet my study provides a unique perspective on whether minimum wage legislation in a female-dominated industry can be a source of women s empowerment within the household. 6 Background The Minimum Wages Act of India, 1948 legally grants a minimum wage for workers working in different industries listed in the employment schedule of the government, including both state and federal governments. The law applies to both the formal and the informal sectors as long as the activity belongs to the employment schedule list 13. Besler and Rani (2008) find that the central and state governments in India set about 1171 different minimum wages (with 48 belonging to the center and 1123 to the state governments). 12 Mathur, Aparna, and Sita N. Slavov. The Role of Legislative Change in Reducing Domestic Violence against Women in India. American Enterprise Institute for Public Policy Research (2016): n. pag. American Enterprise Institute (AEI), May Web. 13 Soundararajan, Vidhya. Minimum Wages and Enforcement in India - Inverted U-Shaped Employment Effects. In 8th IZA/World Bank Conference on Employment and Development, Deutsche Post DHL Convention Center, Bonn,

15 This implies that a domestic worker cannot take the help of labor laws or courts in case of a dispute with their employer, which naturally excludes them from the National Minimum Wages Act (NMW) 14. For the purpose of policy in India, a domestic worker is defined as a person who is employed for remuneration either in cash or kind, in any household, through any agency or directly, either or a temporary, permanent or part-time basis to do household work, including tasks like cooking, cleaning, child and elderly care but is not a member of the family of the employer. Other categories of domestic workers such as watchmen, personal drivers and gardeners were not included in this definition, since it was felt that it would be more appropriate to limit the definition to the most widely understood tasks of domestic workers, especially in the context of India 15. This also works in favor of my argument because the categories of domestic work that are included are largely female-centric, as compared to occupations such as watchmen and gardeners, which are more male-centric. In other words, domestic work is largely viewed by all shareholders as a non-technical and unskilled occupation. While the number of workers in this sector has not been accurately been measured, the data analysis of the NSSO (61st Round, ) reveals an approximate figure of approximately 4.2 million domestic workers in the country, with a large number being female 16. The reasons for the exclusion of domestic workers from National Minimum Wages Act can be traced back to several arguments over decades in the Indian parliament. Some of the arguments against the enforcement of minimum wage for domestic workers include a large scale loss of employment opportunities of domestic workers due to lowering of demand in case of higher wages, the difficulty in enforcing the laws protecting the minimum wage of the workers due to the location of work being a private space and the informal 14 Neetha, N. Contours of Domestic Service: Characteristics, Work Relations and Regulation. The Indian Journal of Labour Economic, 52, N., Neetha. Minimum Wage Setting Practices in Domestic Work: An Inter-State Analysis. Geneva: ILO, International Labor Organization Web. 16 Domestic Workers in India. Domestic Workers in India WIEGO. Women in Informal Employment: Globalizing and Organizing (WIEGO). Web. 03 Dec

16 employment relation between the domestic worker and the private household 17. However, during the last couple of decades, the pressure on the Indian government to respond to the issues and discrimination faced by domestic workers had become more severe, which led to a differing response across states. The most important of the interventions that followed was the inclusion of domestic workers in a list of scheduled employment under the National Minimum Wages Act in a total of seven states across the country, namely Karnataka, Andhra Pradesh, Kerala, Bihar, Jharkhand and Odisha. The state of Karnataka was the forerunner in extending the National Minimum Wages Act to include domestic workers. According to an ILO report (Eluri and Singh, 2013), there was a historically strong grassroots movement in Karnataka pushing for this legislation. The formation of the Karnataka Domestic Workers Congress and the Domestic Workers Brigade played an important role in organizing domestic workers, especially women, and encouraged them to fight for higher wages 18. Based on another ILO report, (Neetha N., 2015) the preliminary wage rates in Karnataka were arrived at through taking into account the cost of living and basic human caloric intake. The main suggestion coming from the abovementioned unions was with regard to accounting for the size of the employer s household in calculating the minimum wages, which combined with the recommendations of the Minimum Wage Advisory Board, led to the establishment of the preliminary wages in Karnataka 19. It is also observed that these wages were fixed per hour or per day as opposed to a monthly calculation and the statutory minimum wage finally came into effect from April 1, The other six states gradually followed suit in terms of instituting this policy, with Andhra Pradesh and Bihar in 2007, Rajasthan in 2008, Jharkhand and Kerala in 2011 and Odisha in Gudibande, Rohan, and Arun Jacob. Minimum Wage Law for Domestic Workers: Impact Evaluation of the Indian Experience. SSRN Electronic Journal (2015): n. pag. Web. 18 Ibid. 19 N., Neetha. Minimum Wage Setting Practices in Domestic Work: An Inter-state Analysis. Geneva: ILO, International Labor Organization Web. 16

17 7 Research Design 7.1 Data My dataset consists of data from various phases of the National Family Health Survey conducted in years (NFHS-2) and (NFHS-3). The NFHS is a large-scale, multi-round survey conducted in a representative sample of households across India. It is a collaborative project of the International Institute for Population Sciences (IIPS), Mumbai, India; ORC Macro, Calverton, Maryland, USA and the East-West Center, Honolulu, Hawaii, USA. The Ministry of Health and Family Welfare (MOHFW), Government of India, designated IIPS as the nodal agency, responsible for providing coordination and technical guidance for the NFHS. NFHS was funded by the United States Agency for International Development (USAID) with supplementary support from United Nations Children s Fund (UNICEF). The Second National Family Health Survey (NFHS-2) was conducted between in the then 26 states of India with added features on the quality of health and family planning services, domestic violence, reproductive health, anemia, the nutrition of women, and the status of women. The Third National Family Health Survey (NFHS-3) was carried out in Eighteen Research Organizations including five Population Research Centers carried out the survey in 29 states of India. The funding for NFHS-3 is provided by USAID, DFID, the Bill and Melinda Gates Foundation, UNICEF, UNFPA, and MOHFW, GOI. ORC Macro, USA, is providing technical assistance for NFHS-3, and the National AIDS Control Organization (NACO) and the National AIDS Research Institute (NARI) have provided technical assistance for the HIV component. Both surveys used uniform sample designs, questionnaires, field procedures and procedures for biomarker measurements throughout the country to facilitate comparability across states and ensure the highest possible data quality. An overview of the two surveys are shown in tables 1 and 2 respectively: 17

18 Table 1 Source: NFHS 2 Report: International Institute for Population Sciences, Mumbai. 18

19 Table 2 Source: NFHS 3 Report: International Institute for Population Sciences, Mumbai. For the purpose of this study, I will be using data from the Women s Questionnaire in both surveys which collected information from women between ages about reproductive behavior, quality of care, sources of family planning and women s status within the household. I pay attention specifically to the section about women s status which includes questions about the frequency of abuse, whether the respondent has been beaten by other 19

20 members of the family and whether the respondent feels the beating is justified for a variety of reasons. The data also provides demographic information about the state in which the respondent resides, education, literacy rates, household amenities, number of children and other relevant information which allows for a variety of controls in the analysis. Domestic violence is seen as an issue faced in nearly every state in India. Based on the NFHS-3 survey conducted in , overall one-third of women between the ages of have experienced physical violence and about 1 in 10 have experienced sexual violence. From a study of the data collected, it appears that married women are more likely to experience physical or sexual violence from their husbands more than any other family member. In addition, nearly half (46 percent) of married women with no education have experienced spousal violence. Similarly, nearly half (47 percent) of women whose husbands have no education have experienced some form of violence. Spousal and familial domestic violence varies greatly by state in the NFHS-3 survey, with the states of Karnataka and Andhra Pradesh both being towards the middle of the spectrum while the prevalence ranges from 6% in Himachal Pradesh to 59% in Bihar Description of the surveys and the tables draw heavily from the IIPS website and reports of both phases of the National Family Heath Survey (NFHS). National Family Health Survey. National Family Health Survey. International Institute for Population Sciences (IIPS), Web. 20

21 Table 3 Source: NFHS 3 Report: International Institute for Population Sciences, Mumbai. Also drawing from the NFHS-3 report on domestic violence faced by women at large in India, the following visual provides a screenshot of the overall data and various factors that were examined as a part of the DHS survey: 21

22 Table 4 Source: NFHS 3 Report: International Institute for Population Sciences, Mumbai. 22

23 Finally, an overview of the variables from the dataset used in the regression analysis are shown below in Table 5, along with summary statistics for each: Mean Standard Deviation Minimum Maximum State Treated Time Treated State Time Education Level Marital Status Partner s Education N Table Difference-in-Difference Estimation Since Ashenfelter and Card (1985) and Card and Krueger (1994), the use of a differencein-difference (DID) approach to study causal estimates has become very widespread. The simplest set up of the methodology is where outcomes are observed for two groups in two different time periods. One of the groups is exposed to the treatment in the second period but not the first, and the second group is not exposed to the treatment during either period. In contrast to a time-series or a cross-section analysis of a treatment effect, a difference-indifference approach typically uses panel data to measure the differences between the treatment and control groups of the changes in the outcome variable over a period of time. As shown in the figure below, the outcome in the treatment group is measured by line P whereas the outcome of the control group is measured by line S. The dependent variable is first measured in both groups at period 1 before the treatment and then at period 2 after the treatment group has received the treatment but the control group has not. DID then calculates the normal difference in the outcome variable between the two groups i.e., 23

24 the difference that would still exist if the treatment had never occurred, represented by the dotted line Q 21. The treatment effect is then the observed outcome and the normal outcome; the difference between P2 and Q. Figure 3 All the assumptions of the OLS model apply to the DID model, but perhaps the most important addition is that of the parallel trends assumption. The explanation above implicitly assumes that the parallel trends assumption holds, since the change over time between the two groups would remain the same without the treatment. In reality, this proves to be a challenge because it can be difficult to ensure that nothing other than the treatment has changed in either of the groups. Moreover, depending on how the two groups are selected, this method of estimation may still be subject to certain biases such as auto correlation and reverse causality. 21 Imbens, and Wooldridge. Lecture 10, Tuesday, July 31st, Pm Difference-in-Differences Estimation (n.d.): n. pag. National Bureau of Economic Research, Web. 24

25 7.3 Methodology In order to measure my dependent variable, I have constructed a measure of domestic violence experienced by women as reported by them in the survey as the frequency of beating experienced over the past 12 months in the survey and compared that to the frequency of beating over the past 12 months, as measured by the survey. This includes beating from spouses as well as other family members. To measure this change, I have used a difference-in-difference approach, as discussed above. A key assumption of this strategy is that the parallel trends assumption holds i.e., the average change in the comparison group is the same as the average change in the counter-factual group when there is no treatment. Having survey data from and then allows me to assess the change in responses between the two time periods, before and after the establishment of the minimum wage i.e., the treatment in For the purpose of this paper, I have used Andhra Pradesh as the control state because it is largely similar to its bordering state Karnataka in terms of geographical landmass, population, GDP and other societal and economic influences, with the primary difference being that Andhra Pradesh did not have a minimum wage for domestic workers established in the year of The difference in GDP has remained at an average of approximately 1.65 Crore Rupees or $250,000 and the 10-year trend from appears to remain relatively parallel, suggesting that there were no economic shocks affecting the states during this time (see figure below). Ideally, we would like to see the same trend exhibited by the domestic violence rates in the two states with the absence of the treatment, but this would not be possible given that the treatment did indeed occur. Moreover, I have also included certain co-variates in my secondary regressions to ensure greater robustness, and to control for any differences between the treatment and control state that could affect the parallel trends assumption. Based on these facts, it is assumed that the parallel trends assumption holds true. 25

26 Figure 4 My primary regression equation is as follows: y = β o + β 1 x 1 + β 2 x 2 + β 3 x 1 x 2 + ε (1) With the addition of controls for a more robust estimate, I used the following secondary regression equation: y = β o + β 1 x 1 + β 2 x 2 + β 3 x 1 x 2 + β 4 x 3 + β 5 x 4 + β 6 x 5 + ε (2) Where: y = Change in Domestic Violence Rates x 1 = State Treated Dummy x 2 = Time Treated Dummy x 1 x 2 = State Time (Treatment Effect) x 3 = Respondent s Education x 4 = Marital Status 26

27 x 5 = Partner s Education ε and ε = Error Term This empirical strategy allows me to make a causal claim by studying the differential effect of the treatment (the minimum wage legislation) on the treatment group i.e., Karnataka respondents versus the control group respondents in Andhra Pradesh. As a test of robustness, I have included control variables such as education level of the respondent which is measured by enrollment in primary, secondary and tertiary educational institutions in the country. I have also used marital status as a control, which provides an insight into how many women were actually married when exposed to violence, which could point to other risk factors such as dowry in the South Asian context. Finally, I have also controlled for partner s education which is measured the same way as respondent s education by enrollment. 8 Results Regression results are shown in Tables 6-10 (below). Table 6 shows the results of my primary regression equation, based on the original specification of the difference-indifference design. From Table 6, column (1) it appears that the effect of the treatment i.e., the institution of a minimum wage for domestic workers in the treatment state of Karnataka caused a decrease of in the frequency of violence between 1999 and 2005 as measured by the coefficient on the interaction term i.e., the difference in frequency of beating between the treatment and control state in time 1 versus time 0. The mean frequency of violence in Karnataka at time 1 is the difference between the coefficients on the constant and the interaction term, i.e., These results are statistically significant at the 99% level. Table 6, column (2) shows results of the primary regression equation, with the addition of the controls: respondent s education, marital status and partner s education, all of which could be associated with risk factors contributing to domestic violence within the household. The effect of the treatment i.e., the coefficient on the interaction term continues to be negative 27

28 and significant at the 99% level, in this case with a value of The following tables are all iterations of the previous regressions, that attempt to isolate the subset of respondents for whom the minimum wage law was intended i.e., domestic and unskilled workers. Table 7 contains results from a subset of the data that looks only at respondents with an educational level of primary school or less. Respondents falling into this category are assumed to be working in some kind of unskilled occupation, which is encompassed in the minimum wage law established in Sub-setting out the population this way allows me to explore whether the treatment had a differential effect, based on levels of education. The results suggest that the treatment effect continues to be negative and significant at the 99% confidence level, with the effect of the treatment remaining largely the same. Table 7, column (2) adds controls to the regression equation looking at respondents with a primary school education or less and the coefficient remains roughly the same and continues to be significant. It is important to note that the post-treatment controls included may have absorbed the effect of the treatment itself and the composition of the groups may have changed in the aftermath, potentially introducing bias into the results. It is unlikely that the education levels of the respondent and partner were affected by this minimum wage law, given the relatively short time period after the treatment. However, marital status is a variable that could have been affected by the legislation, potentially leading to a larger number of women opting out of their marriages or deciding to stay single. Table 8 reports the results of the same regression, looking specifically at the subset of the female population that reported their occupation as maids and other housekeeping services. The results show no change in the value of the coefficient on the interaction term. However, this result is not statistically significant. This could be attributed to the fact that the sample size reduced significantly from over 12,000 to 319 which would cause the standard error to inflate. It is also possible that the composition of this group may have changed in the post-treatment period, leading to women substituting in from other comparable jobs, given the wage increase. 28

29 Variables (1) (2) Frequency of Beating Frequency of Beating State Treated *** ** ( ) ( ) Time Treated *** *** ( ) ( ) State Time *** *** ( ) ( ) Respondent s Education Level ** ( ) Marital Status *** ( ) Partner s Education Level ** ( ) Constant 0.128*** *** ( ) ( ) Observations 21,542 18,654 R-squared Table 6 Robust standard errors in parentheses ***p<0.01, **p<0.05, *p<0.1 29

30 Variables (1) (2) Frequency of Beating Frequency of Beating State Treated ** ** ( ) ( ) Time Treated *** *** ( ) ( ) State Time *** *** (0.0130) (0.0133) Respondent s Education Level *** ( ) Marital Status 0.107*** ( ) Partner s Education Level *** ( ) Constant 0.147*** *** ( ) ( ) Observations 12,046 11,583 R-squared Table 7 Robust standard errors in parentheses ***p<0.01, **p<0.05, *p<0.1 30

31 Variables (1) (2) Frequency of Beating Frequency of Beating State Treated (0.0734) (0.0725) Time Treated 0.131** 0.159** (0.0621) (0.0635) State Time (0.0946) (0.0999) Respondent s Education Level (0.0350) Marital Status 0.141*** (0.0471) Partner s Education Level (0.0185) Constant (0.0546) (0.0663) Observations R-squared Table 8 Robust standard errors in parentheses ***p<0.01, **p<0.05, *p<0.1 31

32 9 Conclusion This thesis sought to investigate whether the roll out of a minimum wage law for domestic workers in the state of Karnataka, India had an effect on the lives of the female population in the form of their perceived bargaining power within the household. I employed a difference-in-difference strategy to identify the effect of the treatment on the female population sample provided by 2 stages of the National Family Health Survey of India. Through this, my aim was to be able to contribute to the already existing body of literature on domestic violence and the relatively few studies focusing on the socioeconomic effects of this minimum wage scheme for domestic workers. The results of this study point to the validity of the first stated hypothesis that the institution of the minimum wage law for domestic workers led to a decrease in the overall rates of domestic violence experienced by women residing in the treatment state of Karnataka. Though results of such studies have been mixed in the past, these results suggest that such wage legislation can indeed be beneficial for a woman s bargaining power within the household. It can be said that this sense of financial independence from the spouse or partner allows her to be more cognizant of abuse and more likely to seek help or separate herself from the perpetrator or even the marriage. However, unlike Card and Krueger s study, my data did not allow me to narrow the focus down to districts that lay exactly on either side of the border of my treatment and control states. This would have helped ensure greater consistency and homogeneity between the two states. I believe this research could be refined and replicated in the future with data that is more thorough and provides a district-level focus. A new survey asking targeted questions about domestic violence may also provide a more accurate picture of women s experiences across India. Moreover, it would also be beneficial to find a way around the minimum wage legislation s effect of increasing the pool of domestic workers post-treatment, in order to see how it affected domestic violence in the lives of the subgroup that were domestic workers 32

33 both before and after the treatment. Yet, the effect of the minimum wage roll-out on the larger population cannot be disputed. My results suggest that financial regulation of this kind can reduce levels of domestic violence. This is contrary to fears that increasing women s wages can promote a male backlash against them, manifesting in higher levels of violence. These results contribute to policy debate on these issues and suggest that financial legislation or social empowerment schemes for women can indeed produce the desired outcomes in terms of upward social mobility. 33

34 10 References Aizer, Anna. The Gender Wage Gap and Domestic Violence. American Economic Review (2010): Web. Babu, Bontha V., and Shantanu K. Kar. Domestic Violence against Women in Eastern India: A Population-based Study on Prevalence and Related Issues. BMC Public Health. BioMed Central Ltd, 9 May Web. 10 Nov Bowlus, Audra J., and Shannon Seitz. Domestic Violence, Employment, And Divorce. International Economic Review 47.4 (2006): Web. Card, David, and Alan Krueger. Minimum Wages and Employment: A Case Study of the Fast Food Industry in New Jersey and Pennsylvania. (1993): n. pag. Web. Domestic Workers in India WIEGO. Women in Informal Employment: Globalizing and Organizing (WIEGO), n.d. Web. 03 Dec Gudibande, Rohan, and Arun Jacob. Minimum Wage Law for Domestic Workers: Impact Evaluation of the IndianExperience. SSRN Electronic Journal (2015): n. pag. Web. Heath, Rachel. Women s Access to Labor Market Opportunities, Control of Household Resources, and Domestic Violence. The World Bank, Policy Research Working Paper 6149, Imbens, and Wooldridge. Lecture 10, Tuesday, July 31st, Pm Difference-in- Differences Estimation (n.d.): n. pag. National Bureau of Economic Research, Web. 34

35 Koenig, Michael A., Jacquelyn Campbell, Rob Stephenson, and Saifuddin Ahmed. Individual and Contextual Determinants of Domestic Violence in North India. N.p., 15 Jan Web. 9 Nov Macmillan, Ross, and Rosemary Gartner. When She Brings Home the Bacon: Labor- Force Participation and the Risk of Spousal Violence against Women. Journal of Marriage and the Family 61.4 (1999): 947. Web. Mathur, Aparna, and Sita N. Slavov. The Role of Legislative Change in Reducing Domestic Violence against Women in India. American Enterprise Institute for Public Policy Research (2016): n. pag. American Enterprise Institute (AEI), May Web. N., Neetha. Minimum Wage Setting Practices in Domestic Work: An Inter-State Analysis. Geneva: ILO, International Labor Organization Web. National Family Health Survey. National Family Health Survey. International Institute for Population Sciences (IIPS), Web. Sherman, Brittany. The Effect of Increases in the Federal Minimum Wage on Domestic Violence Rates. NYU Politics and IR (2015). May Web. Tauchen, Helen, Ann Dryden Witte, and Sharon Long. Domestic Violence: A Nonrandom Affair. National Bureau of Economic Research (1985): n. pag. Web. 35