House Ownership as a Tool for Female Empowerment:
Evidence from India
Master Thesis
Supervised by Prof. Erik Plug
Matilde Marchesi 10840494
University of Amsterdam 15 August 2018
Abstract
This paper investigates the causal effect that house ownership has on women’s empowerment and subjection to domestic violence. I use an instrumental variable model to avoid issues of endogeneity related to the fact that specific household characteristics might affect the likelihood of women to acquire a property. The instrument used is the effect that a stamp duty concession, i.e. a tax cut on property taxes that some Indian states placed in favour of women, has on the share of female owners. This effect is estimated with a difference-in-differences specification. I find that the introduction of the stamp duty concession increases the share of women homeowners but the effect is not always relevant enough to be used as an instrument. Interestingly, the effect is strong enough to be used as an instrument for a subsample of wealthier and more educated women. I run the instrumental variable model on this subsample and I find that house ownership seems to increase women’s say in decisions concerning their children. However, there are concerns with respect to the validity of the underlying assumptions of the model.
This document is written by Student Matilde Marchesi who declares to take full responsibility for the contents of this document. I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Table of Content
1. Introduction 2
2. Literature Review 5
3. Stamp Duty Concessions 7
3.1 Stamp duties in Assam, Uttarakhand, and Uttar Pradesh. 7
3.2 Evidence from Uttar Pradesh. 8
4. Data 9
5. Methodology 13
6. Results 16
6.1 Results of first-stage regressions. 16
6.2 Results of reduced-form and structural equation. 20
6.3 Concerns about the underlying assumptions of the model. 22
7. Discussion 26
8. Conclusion 28
References 30
1. Introduction
In an overwhelming number of societies around the world, women occupy subordinate positions compared to men. They have long been associated with the private realm and domesticity, and have often been excluded from basic rights such as property and inheritance rights - let us only consider that in a highly developed country such as the United Kingdom equal inheritance rights between men and women were achieved in 1922. In India, gender discrimination in Hindu inheritance law was only abolished with the Amendment of the Hindu Succession Act, 2005, which now allows women to inherit property after marriage. It is not only the case for judicial law, traditions and customary laws can also stand in the way of gender equality (Domingo, 2013). In patriarchal societies parents still view daughters as less valuable means to accumulate assets in the long term: once women marry, they leave the parental home to move in with the husband’s family and it is the son who remains with the parents and inherits any property (Mishra &Sam, 2016). Practices like these have resulted in a very skewed distribution of assets between men and women. A report from the World Economic Forum Annual Meeting in 2017 states that women own less than 20% of the world’s land, and according to a survey in 34 developing countries by the United Nations Food and Agriculture Organisation this figure is as low as 10% (Villa, 2017). This unequal distribution has, in turn, created an environment in which it is harder for women to thrive economically (United Nations Development Programme (UNDP), 2014).
To counter this uneven distribution of assets, some Indian states have implemented policies to encourage female ownership of immovable assets. In particular, 10 out of the 28 federal states have introduced a reduction in the property tax to be paid at the time of property exchange, i.e. a concession on stamp duty fees. States have reported that the main objective of this scheme is to empower women socially and economically (as cited in Singh & Arora, 2013). Empowerment is intended as "a process which enhances the ability of disadvantaged ("powerless") individuals or groups to challenge and change (in their favour) existing power relationships that place them in subordinate economic, social, and political positions” (Agarwal, 1995). Indeed, ownership of productive resources should make women more secure and independent economically, increasing their ability to bargain within the household, and even reducing their likelihood to be subject to domestic violence (e.g. Singh & Arora, 2013; Panda, 2003; Pandey, 2003).
The existing literature typically disentangles the concepts of socio-political empowerment and domestic abuse, and it investigates two main research question: what is the
impact of women’s asset ownership on their social and economic empowerment? And, what is the impact of women’s asset ownership on their likelihood to be subject to domestic violence? However, the evidence that these studies provide is not always relevant for the creation of policies: limited sample size and lack of causal identification of the effect of asset ownership on the outcome give little indication on how to tackle these issues (Peterman, Pereira, Bleck, Palermo, & Yount, 2017; Domingo, 2013). Many existing studies also fail to recognise a potentially endogenous relationship between female asset ownership and empowerment and domestic violence. For instance, it is likely that household’s unobserved characteristics related to economic and cultural conditions are correlated with unobserved dimensions that could dictate women’s probability of owning assets, thus biasing the results (Petermal et al., 2017). Agarwal (1994) identifies other possible drivers of women’s likelihood of asset ownership in marriage practices, purdah, i.e. female seclusion, labour force participation, literacy and, more generally, gender relations within society.
The theory of change behind the mechanisms that lead female house owners to empowerment and decreased risk of experiencing abuse is not always explicitly presented. Domingo (2013) suggests a theory of change that is closely applicable to the enactment of stamp duty concessions in Indian states. The initial change is created with the introduction of policies to secure property rights. These policies bring about intermediate results: they create conditions for individuals to become legally empowered and affirm their individual legal identity, which allows them to use their assets as tradable to obtain access to socio-political capital and credit, expanding their economic possibilities. All these intermediate outcomes create an environment where individuals’ agency is enhanced, increasing individuals’ ability to make autonomous decisions to their benefit. This theory of change relies on the main assumption that asset ownership increases economic independence, and this, in turn, increases the ability of the owner to participate in decision-making processes. This assumption is in line with what is argued by Agarwal (1994). He rejects the view that property rights are intrinsically valuable (Cousins, 2009; Nyamu-Musembi 2007) but claims that their fundamental role is serving as an instrument to obtain economic independence, which enables the development of agency. In the case of India, the introduction of stamp duty concessions gives economic incentives to household to purchase assets in the woman’s name. The legal possession of the property changes the woman’s status, increasing her ability to bargain within the household and use the property as collateral for loans. Furthermore, asset ownership gives women an exit option from marriage that, even if not exercised, can be
threatening enough to discourage domestic violence (Panda & Agarwal, 2005), altering the gender power relations within the household.
The focal matter is unravelling the causal effect of asset ownership from the endogenous effect that other factors might have on the outcomes. In this thesis, I attempt to quantify the causal impact of women’s house ownership on their empowerment and on their subjection to domestic violence. I measure empowerment through women’s control over money for their own use and their decision-making relevance in the realm of household expenditures, child care, fertility, and mobility. I account for issues of endogeneity with the use of an instrumental variable (IV) model: I use the impact that the introduction of the stamp duty concession has on female asset ownership at the state level as an instrument for women’s house ownership in that state. I take advantage of the fact that some states have introduced the concession while others have not. I exploit this difference across states and estimate the impact of stamp duty concessions on asset ownership with a difference-in-differences (DiD) specification. This thesis provides further evidence to be added to the scarce existing literature aiming to estimate the causal effect of women’s asset ownership on empowerment and domestic violence incidence. An additional contribution of my study to the pool of existing research, is that, in running the IV model, I also quantify a causal impact of the stamp duty concession on the share of women owners, providing valuable information on the functioning of the scheme. States are foregoing tax revenues to implement the concession, so it is important to make sure that the scheme is actually resulting in an increased number of female asset owners.
The relevance of this topic is straightforward: it is necessary to understand what type of policies or schemes result in a true redistribution of assets across genders and help promote the empowerment of women and lessen their subjection to domestic violence. The results of this thesis show that the introduction of the stamp duty concession for women has the desired effect of increasing female house ownership, however, this effect is not as relevant as expected. The impact of the concession on the share of female homeowners is stronger on a subsample of wealthier and more educated women. Nonetheless, the effect of house ownership on empowerment is limited on this subsample as well, only increasing the decision-making relevance of women on child care. Notably, the analysis is likely to be violating its underlying assumption and results should be considered with caution.
This thesis is structured as follows: Section 2 is a literature review of studies assessing the impact of asset ownership on empowerment and violence. Section 3 describes stamp duties and presents the different rates across the Indian states considered. It also summarises
the finding of a case study on the effect of stamp duty on empowerment conducted in the state of Uttar Pradesh, India. Section 4 introduces the dataset used and its characteristics, while Section 5 presents the methodology. Section 6 presents the results obtained and discusses issues regarding the underlying assumptions of the analysis, Section 7 makes more general comments on the results obtained. Lastly, Section 8 concludes.
2. Literature Review
In recent years, several studies have addressed the role of property ownership on women’s empowerment and their subjection to domestic violence. Pandey (2010) focuses on the city environment of Kathmandu, Nepal, which in the years prior to the study saw a surprising increase in the number of women acquiring properties in their name. She investigates the relationship between asset ownership and women’s role in decision-making processes about household spending. After controlling for demographic information, education, value of the property, and having a bank account, she finds suggestive evidence that women sole owners have significantly higher say compared to joint owners. This significant effect vanishes when she adds controls for age, marital status, size of the household, having sons, and number of years the property was in the woman’s name (Pandey, 2010). Interestingly, Pandey (2010) also investigates the reason that prompted the increase in the number of female homeowners and finds that initial economic incentives were sufficient to bring about social change in favour of women. Nepalese law prevents any family member to claim back a property if it is in the name of a woman, so it became a strategic move for households to put houses and land in the wife’s name. Pandey (2010) reports how the control over assets increased women’s confidence enough to challenge discriminatory laws1 (Pandey, 2010). This paper is a great example of how economic incentives can sometimes lead to social change, albeit only among privileged women.
Mishra and Sam (2016) also focus on Nepal and investigate the causal impact of land ownership on women’s empowerment. They account for the possible endogeneity of ownership with inverse probability weights, and coarsened exact matching and instrumental variable methods. They used coarsened exact matching to create an instrument that is closely related to land ownership, but does not explain empowerment. They consider as dependent variable a measure of women's bargaining power quantified as their role in decisions about
1 Such as the inheritance law in place.
household expenditures2 and personal healthcare. The authors also look into the ability of women to visit family and friends. They find a significant and positive effect of land ownership on bargaining power but acknowledge that this effect might be limited to places where agriculture remains a prime source of employment for women.
For what concerns the effect of asset ownership on domestic violence, there exists a school of thought which suggests that female asset ownership can protect against violence: women are able to signal a level of economic independence that is credible enough to discourage domestic abuse under the threat of separation (Panda & Agarwal, 2005). This is especially true when there is legal evidence of her ownership and there exist laws which allow for divorce and support female inheritance of properties. However, others believe that, in settings where assets are strongly concentrated among men and violence is used as tool to discipline and control wives, female asset ownership might be viewed as threatening traditions and might trigger husbands to try and exercise even more control through violence (Peterman et al., 2017).
Panda and Agarwal (2005) report suggestive evidence that women who own any type of asset experience a lower incidence of domestic abuse (both physical and psychological). In particular, the reduction of violence because of asset ownership is especially pronounced in the case of house ownership. It is important to note that the study is set in the state of Kerala, India. Here, a considerable share of the population belongs to matrilineal societies which might be more responsive to issues of gender power imbalances. Bhattacharyya, Bedi, and Chhachhi (2011) replicate a similar study in a prevalently patriarchal Indian state, Uttar Pradesh. They limit their sample to female homeowners as they find house and land ownership to be highly correlated, but only house ownership to be significant on violence. They try to account for endogeneity of ownership by including a number of controls for observed factors. Although this solution does not truly solve issues of endogeneity, the authors find that women’s house ownership is linked to lower incidence of domestic violence. Kelkar, Gaikwad and Mandal (2015) investigate the incidence of domestic violence between landless and landowning women in India. Their study does not produce causal evidence but is nonetheless valuable in underlying the fact that social norms might drive results. In fact, they further split their sample between women living in matrilineal societies, i.e. Meghalaya, and patriarchal ones, i.e. Karnataka and Telangana. The authors find that, in general, landowning women experience less violence, but women owners in matrilineal
2Major household expenditures only, as the authors say that women are probably already in charge of smaller
societies report the lowest level of violence incidence. In patriarchal society women report that owning land has only marginally increased their decision-making role and their responsibility within the household, and report only a marginal reduction in physical abuse. Instead, in the matrilineal society most women reported a definite increase in their responsibility in buying and selling produces, and an increase in their ability to visit family and friends3 and decide where to work. The authors conclude that land ownership and a gender-egalitarian society were both relevant factors in reducing the incidence of violence.
More recently, Peterman et al. (2017) estimate the causal impact of asset ownership on domestic violence. They use data about women from 28 Demographic and Health Surveys (2010-2014) in a multivariate probit model that matches households based on wealth, as they believe it to be correlated with women’s asset ownership. The authors also control for individual characteristics (age, schooling, marital status), and household characteristic (household size and wealth, urban residence and other geographic positioning information). They exclude never married women as they could have more agency about their life decisions. They find inconclusive evidence of the impact of ownership on violence: violence is negatively associated with asset ownership in 3 countries, but the relationship is positive in 5 others, and inconclusive in the remaining 20. In this study, results remained inconclusive even when the data was disaggregated into type of property, whether it was jointly owned, and according to women’s age. These results suggest that the effect of property on domestic violence is highly context specific.
3. Stamp Duty Concessions
3.1 Stamp duties in Assam, Uttarakhand, and Uttar Pradesh.
In order to become the legal owner of an immovable property (house or land), the Indian government requires the payment of a property tax, i.e. stamp duty, to acknowledge the change of hands of the property. Stamp duties are a percentage of the market value of the property in question and are to be paid at the time of property exchange. Without the payment of the stamp duty, the sale agreement cannot serve as a legal document in case of a dispute. Stamp duty can be paid through non-judicial stamp papers, which are documents with details about the agreement purchasable from a licenced vendor. The cost of the papers is equal to the stamp duty to be paid in the agreement. Franking is another method in which the agreement is
3 In this type of society relatives might be closer to begin with, making a small distance much easier to cover and
printed and an authorised bank affixes a special adhesive on the paper which has value equal to the stamp duty. It is also possible to pay stamp duty online through e-stamping.
Only 10 out of the 28 Indian states have introduced a stamp duty concessions scheme for women. These 10 states enacted the scheme in the years between 2003 and 2008, with Delhi being the pioneer and introducing the stamp duty cut in 2003. Yet, data on other dimensions relevant for my analysis is only available for the years of 2005 and 2012, therefore, I focus only on those states that introduced the scheme between those years. These states are Assam, Uttarakhand, and Uttar Pradesh. In Assam, the scheme was introduced in June 2009 reducing women’s rate from 6% to 5% of the market value of the property. Men’s rate is 6%. In Uttarakhand the scheme was introduced earlier, in August 2006, and women’s rate was reduced to 4.5% for a property value of up to 20 lakhs4. The concession was amended in October 2009, further reducing women’s rate down to 3.75%5. Men’s rate is 5%. In Uttar Pradesh, the scheme was introduced in February 2006 and the concession reduced women’s rate from 5% to 4%, while men’s rate is 5%.
3.2 Evidence from Uttar Pradesh.
The literature on the impact of stamp duty on women’s house or land ownership is extremely limited. Singh and Arora (2013) are the only ones to explore the consequences of the introduction of the scheme in Uttar Pradesh. They report positive results and write that the tax cut motivated many families to register properties in the woman’s name. They further show that the district with the highest number of women beneficiaries was the capital’s, Dehradun, while the one with the least number of women beneficiaries was Champawat, a rural district located at a higher altitude. This finding suggests that mountainous and more remote areas could be less reached by the policy.
Singh and Arora (2013) report that the majority of women beneficiaries of the scheme were literate, with a considerable share of them having reached higher education levels. Most of them did not have formal employment and reported to be housemakers. Among the women who acquired a property, a large majority obtained a house or a house plot rather than land. Most assets were purchased for residential purpose and respondents indicated that they were looking to obtain economic security (61%) or take advantage of the scheme (31%). The authors report that awareness of the scheme was fairly low with only a third of women who
4 Lakhs are the equivalent of 100’00 rupees. After this amount the rate goes to 6%.
acquired a property knowing about it. Singh and Arora (2013) also describe the value of the properties purchased and find that most of people, 75%, spent up to 5 lakhs and only 2% of buyers spent more than 10 lakhs.
From this case study, it is also evident that most women were not acting on their own accord but found themselves obtaining a property because of economic incentives which the husbands wanted to reap. Indeed, more than two thirds of women reported that the decision to buy the house in their name was their husband’s and only 5% of them were sure that they would have purchased the property in their name in the absence of the stamp duty concession. Singh and Arora (2013) ask respondents who acquired an immovable property to comment on their wellbeing and perception of their empowerment. They find that the majority of women owners felt economically more secure and more able to make decisions within the household. These women also felt more accepted and respected within the family and felt that their say with regards to children’s education, health issues, and marriage decisions was held in higher consideration than before. Only one third reported feeling able to participate in the decision about the number of children to have. Half the women owners felt more secure about the risk of being evicted or disowned, and reported higher self-esteem and being less subject to violence from their partner. Overall, 55% percent of women owners reported a positive change in their economic condition, 30% reported a positive change in their personal life, and 20% reported a positive change in their social life. The study concludes that, although the loss in government revenue is significant, the gender empowerment resulting from the scheme outweighs the costs. The authors claim that the concession should continue for the time being.
4. Data
In this paper, I use data from the Indian Human Development Survey (IHDS) which is a joint project of the University of Maryland and the National Council of Applied Economic Research (NCAER) in New Delhi. This dataset is publicly available online through the Inter-university Consortium for Political and Social Research (ICPSR). There are two waves of the survey: the first round of data was collected in 2004-2005, while the second round was collected in 2011-2012. Funding for the surveys was provided by the National Institutes of Health and the Ford Foundation. The surveys are nationally representative as they interview a stratified sample of approximately 40’000 households across India. The surveys are also multitopic: for instance, they ask question regarding, education, health, marriage, fertility,
gender relations, and children’s education. From the IHDSs, I was able to construct a panel dataset: indeed, 83% of household interviewed in the first wave were re-interviewed in the second wave as well.
I restrict the dataset to only the information needed for my analysis. I discard data regarding divorced or unmarried women as they are likely to have significantly different agency in their choices (National Family Health Survey, 2006). I also exclude observations of women owning a house at baseline as the effect I am interested in is on women who initially are non-owners. I restrict my sample to those states on which I was able to retrieve clear information about stamp duty policies. From this subsample of states, I further exclude Mizoram. There, the Transfer of Property Act, 1992 has not been enacted and other laws regulate property exchanges. I also discard the states of Delhi, Gujarat, Madhya Pradesh, Rajasthan, Punjab, and Puducherry. These states have introduced stamp duty concession prior to the first wave of the IHDS and their trends in women’s house ownership are likely to violate the common trend assumption needed to run a DiD regression. Lastly, I exclude all observations from Meghalaya and Kerala as they are known to have a majority of the population living in matrilineal societies (Kelkar et al., 2015). It is important that in all the states considered there is a predominantly patriarchal system. A comparable context reduces the likelihood of the results to pick up on unobserved differences in social norms (Agarwal, 1994; Kelkar et al., 2015; Peterman et al., 2017). Patriarchy is characterised by patrilocal marriages in which women move away from their home and go live with their spouse’s family, by arranged marriages which require the payment of a dowry to the husband’s family, by legitimacy of the father’s lineage, and men’s control over land and property. In this type of setting men are also in control of women’s freedom: women often need their husband’s permission to visit friends and family, to work, or conduct any other type of financial transaction (Kelkar et al., 2015).
I measure empowerment outcomes as the woman’s ability to control money, and make decisions. More specifically, I use binary variables indicating whether she has money in cash at her disposal, weather she has a say in household expenditures, and whether she has a say in the number of children to have. I construct 2 binary variables about whether she feels she has a say with respect to childrens’ care and her personal mobility. For what concerns children’s care, I code as 1 if she has a say on what to do if the child is ill or who to marry the child to. For the variable about personal mobility, I group information on whether she has to ask for permission to go to the health centre, the shop, or a friend’s house. I code the variable as 1 if she has to ask for permission to go to any of these places, and 0 otherwise.
I also consider as outcome information about whether she is physically punished by her husband for failing to look after the house and children, cooking badly, having extramarital relations, having no dowry, or being disobedient in some other way. I similarly construct a binary variable coded 1 if she is beaten for one of the reasons above, and 0 otherwise.
Furthermore, I use binary variables for time, states, and ownership. The year corresponding to the first survey wave is coded 0, while the second year is coded 1. Those states in which the stamp duty concession was introduces are identified by a 1 and correspond to the treatment states, while 0 is for those states in which the concession was not introduced, i.e. control states. Female house ownership is coded as 1 if she owns a house (jointly or not) and 0 otherwise. Table 1 shows descriptive statistics of all variables used in treatment and control states.
(1) (2) (1) (2) Control States Treatment States Control States Treatment States Cash-in-Hand 0.80 0.864 Decisions Household 0.873 0.657
[0.403] [0.343] [0.369] [0.475] Decision Fertility 0.199 0.195 Decisions Children 0.374 0.241
[0.399] [0.397] [0.484] [0.428] Asks for Mobility 0.763 0.792 Abuse 0.807 0.835
[0.425] [0.406] [0.425] [0.371] Education 4.835 3.246 Purdah Practice 0.385 0.810
[4.611] [4.411] [0.478] [0.393] Age 33.104 32.933 Girls’ educ. same as boys 0.820 0.859
[7.758] [7.568] [0.384] [0.348] Household head’s wife 0.852 0.812 Age at marriage 17.803 16.752 [0.355] [0.391] [3.520] [3.564] Number of children 2.444 3.207 Any say in spouse choice 0.466 0.327
[1.415] [1.928] [0.499] [0.469] TV 2.263 1.743 Couple lives alone 0.090 0.127
[0.840] [0.950] [0.286] [0.333] Radio 1.577 1.558 Husband worse off 0.088 0.079
[0.739] [0.832] [0.284] [0.270] Highest male education 7.047 6.459 Husband not a relative 0.732 0.815
[4.882] [5.032] [0.443] [0.389] Hindu 0.830 0.751 Financial help from children 0.122 0.072
[0.376] [0.432] [0.327] [0.259] Brahmin 0.032 0.077 Has bank account 0.169 0.154
[0.175] [0.266] [0.327] [0.361] OBCa 0.407 0.427 Knows government employees 0.306 0.294
[0.491] [0.495] [0.461] [0.456] Scheduled caste 0.224 0.204 Families care for widows 0.921 0.969
[0.417] [0.403] [0.270] [0.172] Scheduled tribe 0.074 0.027 Girls harassment common 0.060 0.065
[0.262] [0.163] [0.239] [0.246] Incomeb 48,901 41,276 Self-help group 0.125 0.027
[64,173] [53,646] [0.331] [0.161] Water within household 1.405 1.648 Credit or saving group 0.079 0.030
[0.492] [0.478] [0.270] [0.171] House type 2.821 2.810 Development group or NGO 0.018 0.008
[1.681] [1.699] [0.132] [0.089] Sanitation facilities type 1.831 1.820 Caste association 0.102 0.099
[1.225] [1.061] [0.303] [0.298] Urban 0.385 0.283 Cooperative 0.057 0.025 [0.487] [0.451] [0.231] [0.155] Trade union 0.076 0.013 [0.265] [0.113] Observations 5,556 2,022 5,556 2,022 ∗p < 0.001
Notes: Column (1) and (2) show the means of a number of control variables at baseline in control and treatment states respectively.
Standard deviations are in parenthesis.
aOther Backward Castes
5. Methodology
In this section, I present the methodological approach I use to identify the causal effect of women’s house ownership on their empowerment and violence incidence experienced. As shown by the results obtained in previous studies, the effect of female house ownership on empowerment and domestic abuse suffers from endogeneity issues. In this case, endogeneity is caused by reverse causality: women’s ownership of a house is a tool for their empowerment but at the same time more empowered women are more likely to acquire a property (e.g. Peterman et al., 2017; Mishra & Sam, 2016; Bhattacharyya, Bedi, & Chhachhi, 2011). To obtain an estimate which only shows the true effect of house ownership on the outcomes and excludes the endogenous part, I adopt an IV model. I use the impact that the introduction of the stamp duty concession has on female asset ownership as an instrument for women’s house ownership. As previously mentioned, I can exploit the fact that some states have introduced the concession scheme while others have not. In order to estimate more precisely the causal effect of the stamp duty concession on house ownership, I adopt a DiD specification. Having data for only two points in time, I run a straightforward DiD in which I compare treatment and control states at baseline, and after the introduction of the stamp duty scheme. In this way, when regressing the instrument on the outcome variables, I am able to capture the effect of stamp duty related to ownership and avoid including the effect that individual characteristics could have on the outcomes. Estimating the direct impact of house ownership on empowerment and violence incidence would produce results biased by the individual characteristics of homeowners. Instead, the IV model estimates indirectly the impact that house ownership has on outcomes through the effect that the stamp duty concession has on the share of female homeowners in treatment states.
The IV model is composed of three main regressions. The first-stage regression estimates the effect of the instrument, i.e. the stamp duty introduction, on the endogenous regressor, i.e. house ownership. The reduced-form regression estimates the effect of the instrument on the outcomes, i.e. empowerment and domestic violence. The structural equation is equal to the ratio between the first-stage and the reduced-form, and gives the effect of the exogenous component of the regressor on the outcome. Equation (1) shows the first-stage regression:
!"# = %&" + (&)#+ *&+"+ ,&(+"∗ )#) + 0&1"+ 2&"# (1)
!"# is the dependent binary variable for female house ownership. )# is a binary variable indicating the year of the survey, +" is a binary variable for treatment state, and +"∗ )# is the
interaction between time and treatment, i.e. the instrument. 1" is a set of controls. The
coefficients (& and *& respectively capture the effect of time and the differences between states. ,& is the coefficient of interest as it captures the actual effect of the tax scheme on women’s house ownership. Lastly, 0& refers to the part of the estimate described by control variables. Taking advantage of the panel form of the dataset, I use individual fixed effects in every step of the model to reduce heterogeneity related to unobserved time-invariant characteristics of the individuals. The use of fixed effects does not allow the investigation of the effect of time-invariant variables on the outcome: assuming that families do not move across control and treatment states, the state of residence +" is a time-invariant variable and is
therefore omitted in the results. To assess the robustness of the results, I ran Equation (1) three times: with no controls, and with the introduction of two sets of controls which are represented by 1&" and 14".
The first set of controls follows closely the work of Mishra and Sam (2016) and includes women’s individual characteristics and household characteristics. Individual characteristics are education, age, whether she is married to the household head, the number of children she has, and how much media exposure she receives6. Household characteristics
instead are information such husband’s education, the caste of the family and their religion, household wealth, and whether they reside in a rural or urban area. Education is a relevant control because it has been found to increases the recipient’s sense of independence and self-worth (Kabeer, 1999; Kishor & Gupta, 2004; Trommlerova, Klasen, & Leßmann, 2015). Kelkar et al. (2015) also find education to be a relevant factor in determining baseline levels of domestic abuse. Pandey (2010) finds that property ownership is more prevalent among women who have received a certain level of education7. In southern Asia, age can often be a
proxy for authority (Basu & Koolwal, 2005; Kabeer, 1999; Kishor & Gupta, 2004) and Singh and Arora (2013) also find that women’s agency increases with age. Being the wife of the household head could also increase her authority as she can try to influence the main decision-maker in the family (Mishra & Sam, 2016). Media exposure through television or radio can have a significant impact as it can be the only vessel of information from the outside world that women receive (Aslam & Kingdon, 2012; Basu & Koolwal, 2005; Kishor & Gupta, 2004). Because of lack of information about the education of the husband, I use instead the highest level of education attained by a man in the household as a proxy. The effect of husband’s education does not have a straightforward direction but it has often been
6 Media exposure is considered as whether she has regular access to television and/or radio.
considered a relevant determinant of land ownership (Basu & Koolwal, 2005; Kishor & Gupta, 2004). Again, following previous studies, I include covariates for caste and religion (Basu & Koolwal, 2005; Kabeer, 1999; Trommlerova et al., 2015; Pandey, 2010). An important determinant in ownership and female empowerment is household wealth as this could translate in more resources available to women and more access to information (Basu & Koolwal, 2005; Kabeer, 1999). Kelkar et al. (2015) find income to be relevant in determining baseline abuse. As suggested by Rutstein & Johnson (2004), I consider wealth in terms of household income, and categorical variables indicating the type of housing, water access and sanitation facilities. Lastly, women residing in rural areas might occupy even more subordinated positions to men than in urban places (Kabeer, 1999; Kishor & Gupta, 2004; Trommlerova et al., 2015) or might have less access to information (Singh and Arora, 2013).
The second set of covariates that I use includes all of the above with the addition of whether the household practices purdah (Agarwal, 1994), is joint or nuclear, whether it believes that boys and girls should receive the same education, has government employees in its close-knit circle, and if parents expect financial support from children. I also add information about whether the spouses are relatives, what the husband’s economic status is compared to the wife’s, her age at the time of marriage, and if she had any choice about her spouse. I include individual characteristics such as whether she has a bank account (Pandey, 2010) or is member of groups that could be a source of support, e.g. self-help groups, development NGOs, or credit/savings groups. Self-help groups could counter violence by giving women "organisational strength to resist male violence" (Kelkar et al., 2015, p. 50). Lastly, I include information about the type of community the household lives in, adding covariates for the level of harassment of girls and if families are expected to care for their widows. These controls are intended to further capture the type of gender relations at play in women’s environment, however, there are concerns that some of them might not be appropriate as concessions might impacts norms on gender equality.
In the reduced form of the IV model, I repeat the same type of DiD specifications: a stripped-down version, and two regressions with different sets of covariates. The basic form of the equation is as follows:
5"# = %4"+ (4)#+ *4+" + ,4(+" ∗ )#) + 041"+ 24"# (2)
where 5"# is the outcome variable and the other terms are as above. Lastly, the structural equation of the IV model is:
where 7 is the coefficient of interest of the regressor !8"#, female house ownership, which now captures only the effect of ownership derived from the instrument and exclude the endogenous component. 1"# contains information about the covariates described above and the binary time variable.
This model relies on five main assumptions: relevance, independence, exclusion, monotonicity, and the absence of spillovers. Relevance refers to the fact that the introduction of the stamp duty has to be correlated with the share of female homeowners. This assumption can be tested with the first-stage regression. Independence indicates that the instrument should be randomly assigned and should not be the results of selection bias. This assumption can be partially tested with a balance test. The third assumption, exclusion, refers to the fact that the stamp duty concession must not have a direct effect on empowerment and violence incidence but must solely act through house ownership. In other terms, the instrument cannot be correlated with the error term. This assumption cannot be empirically tested but some considerations can be made. Monotonicity indicates the fact that the effect of the stamp duty concession on house ownership should always have the same direction. Indeed, the introduction of the policy should always result in an increase in the share of female homeowners. Concerns about this assumption can be addressed by running the first-stage regression on different subsamples of the observations. Lastly, the absence of spillovers imposes that receiving treatment must not have an effect on those near the treated. This is not a concern in this analysis as the introduction of stamp duty concession in one state is unlikely to affect empowerment and violence incidence in a neighbouring state. In fact, there is one more assumption to consider: the DiD specification requires a common trend in women’s house ownership between treatment and control states prior to the introduction of the stamp duty concession scheme. This assumption cannot be empirically tested because of lack of data. Concerns about the violation of these assumptions are discussed in Section 6.3.
6. Results
6.1 Results of first-stage regressions.
The analysis hinges on the effect that the stamp duty concession has on the share of female owners. This is shown in Table 2 with the results of the first-stage regression on the whole sample. The estimates in Column (1) are found with the stripped-down version of the regression. In Column (2) and (3) I control for the sets of covariates 1&" and 14" respectively. The estimates of interest are the ones corresponding to the interaction variable State*Year.
Table 2: First-Stage Results: the Effect of Treatment on Female House Ownership
(1) (2) (3)
No Controls Control Set X1i Control Set X2i
State - - -Year 0.186∗∗ 0.157∗∗ 0.115∗∗ (0.005) (0.020) (0.024) State*Year -0.029∗∗ 0.018 0.034+ (0.010) (0.034) (0.019) F-Statistics -8.24 0.29 3.39 Observations 14701 6448 9243
Standard errors in parentheses
+p < 0.10,∗p < 0.05,∗∗p < 0.01
Notes: The table shows first-stage regressions of stamp duty concession on
fe-male house ownership using a difference-in-difference specification. Column (1) show the results of the regression using no controls. Column (2) shows the results obtained with the addition of the set of controls X1i. Column (3) shows the results
obtained with the addition of the set of controls X2i. The F-statistics refers to the
significance of the instrument State*Year. The variable State is omitted because of the fixed-effects design.
The stripped DiD regression produces an unexpected result: the introduction of the policy has negative and significant effects on female house ownership, reducing it by almost 3%. However, once I introduce controls for relevant covariates that could influence the likelihood of female ownership, the estimate becomes positive, and even significant at the 10 percent level in the last specification. On the full sample, none of the specifications produces a strong enough instrument: a rule of thumb indicates that a strong instrument should have an F-statistic of at least 10 (Taiger & Stock, 1997). These results show that the introduction of the stamp duty concessions does not have the expected results: stamp duties are not as relevant as expected in influencing the share of women homeowners. This suggests that the scheme is ineffective at achieving its aim.
Yet, as pointed out by Pandey (2010), the scheme is more likely to affect individuals who have more resources and are more educated. For this reason, I run the three specifications of the first-stage regression on three subsamples of observations. Panel A includes women whose household have an income greater or equal to the third income quintile8, i.e. the top 60% of household incomes. Panel B includes women who have received at least 3 years of education9, and Panel C includes women who fall under both household income and education restrictions. Table 3 shows the results obtained. Again, results are only significant at the 10 and 5 percent level when including 14". In line with expectations, the introduction of
the stamp duty scheme has a greater effect on these subsamples. In particular, women who fulfil both income and education restrictions seem to be the most affected. Although none of the regressions show an F-statistic large enough to consider the instrument strong, Panel C in Column (3) produces an F-statistic that is close enough to 10 to continue with the remaining part of the analysis.
8 This corresponds to the mean income of the treatment states.
Table 3: First-Stage Results: the Effect of Treatment on Female House Ownership
(1) (2) (3)
No Controls Control Set X1i Control Set X2i
Panel A: Top 60% of household incomes
State - - -Year 0.170∗∗ 0.147∗∗ 0.099∗∗ (0.007) (0.023) (0.030) State*Year -0.004 0.043 0.055∗ (0.013) (0.038) (0.024) F-Statistics -0.08 1.28 5.20 Observations 8646 4446 5752
Panel B: Education ≥ 3 years
State - - -Year 0.178∗∗ 0.162∗∗ 0.119∗∗ (0.008) (0.026) (0.035) State*Year -0.001 -0.003 0.075∗ (0.018) (0.042) (0.029) F-Statistics -0.00 -0.00 6.50 Observations 7202 4359 5520
Panel C: Top 60% of household incomes and Education ≥ 3 years
State - - -Year 0.168∗∗ 0.153∗∗ 0.126∗∗ (0.008) (0.029) (0.040) State*Year 0.024 0.022 0.101∗∗ (0.020) (0.047) (0.034) F-Statistics 1.42 0.20 8.76 Observations 5801 3304 4016
Standard errors in parentheses
+p < 0.10,∗p < 0.05,∗∗p < 0.01
Notes: The table shows first-stage regressions of stamp duty concession on female house
own-ership using a difference-in-difference specification across three subsamples. Column (1) show the results of the regression using no controls. Column (2) shows the results obtained with the addition of the set of controls X1i. Column (3) shows the results obtained with the addition of
the set of controls X2i. The F-statistics refers to the significance of the instrument State*Year.
6.2 Results of reduced-form and structural equation.
The results from the first-stage show that the impact of the introduction of stamp duty scheme is less significant than anticipated. Yet, on a subsample of women who are more sensitive to the scheme, i.e. the richer and more educated women of Panel C, there seems to be a stronger effect. Hence, I run the reduced-form and structural equation with the set of controls 14" on the reduced sample of Panel C. The results produced here show the impact
that homeownership has on empowerment and domestic violence for the subsample which is most responsive to the introduction of the concession.
In the reduced-form, I find that the instrument is significant on decisions about child care and physical abuse, as shown in Table 4. The direction of the estimate about children’s decision is positive, consistently with expectation. Interestingly, the estimate for abuse is also positive. This could mean two things: either house owning women experience more violence as a form of retaliation from their husbands, or house owning women are more empowered and report violence more frequently (Kelkar et al., 2015). It should be noted that the issue of underreported violence is severe. Kelkar et al. (2015) also explain that marital violence is often not considered as such but it is simply seen as a way for husbands to discipline their wives. Table 5 shows the result of the structural equation: I find that only the estimate for decisions about children is significant at the 10% level. Nonetheless, I also find that most estimates are larger than expected. Given the fact that the F-statistic is lower than 10, results are likely to be capturing trends from unobserved dimensions rather than showing a very large increase in the outcome.
Table 4: Reduced-Form Results: the Effect of Stamp Duty Cut on Outcome Variables (1) (2) (3) (4) (5) (6) Cash-in-Hand Decisions Household Decisions Fertility Decisions Children Needs Permission to Leave House Abuse State - - - -Year 0.085∗ -0.017 0.044 -0.250∗∗ -0.415∗∗ 0.030 (0.039) (0.054) (0.059) (0.067) (0.110) (0.049) State*Year 0.008 -0.056 0.027 0.125∗∗ -0.103 0.089∗ (0.029) (0.045) (0.048) (0.048) (0.094) (0.035) Baseline Mean(%) 87.3 68.6 23.1 25.1 81.0 72.8 Observations 4104 4109 4109 4109 2878 4108
Standard errors in parentheses
+p < 0.10,∗p < 0.05,∗∗p < 0.01
Notes: The table shows the estimates for the reduced-form regression on the outcome variables of empowerment and
violence incidence. The variable State is omitted because of the fixed-effects design. Baseline Mean refers to the average of the outcome variable at baseline in the sub-sample considered. All regressions are controlled for X2i.
Table 5: Instrumental Variable Results: the Effect of Female House Ownership on Outcome Variables
(1) (2) (3) (4) (5) (6) Cash in Hand Decisions Household Decisions Fertility Decisions Children Needs Permission to Leave House Abuse Female Ownership 0.161 -0.614 0.189 1.185+ -2.120 0.720 (0.304) (0.519) (0.520) (0.668) (1.911) (0.455) Percentage Impact 0.2 -1.0 0.8 4.7 2.6 1.0 Baseline Mean(%) 87.3 68.6 23.1 25.1 81.0 72.8 Observations 4011 4016 4016 4016 2814 4015
Standard errors in parentheses
+p < 0.10,∗p < 0.05,∗∗p < 0.01
Notes: The table shows the results of the instrumental variables regression where Female Ownership is
instru-mented by the effect of the stamp duty concession on the number of female house owners. Baseline Mean refers to the average of the outcome variable at baseline in the sub-sample considered. All regressions are controlled for X2i.
6.3 Concerns about the underlying assumptions of the model.
As mentioned in Section 5, the model relies on a number of assumptions, namely: relevance, exclusion, monotonicity, the absence of spillovers, independence, and the presence of a common trend in the dependent variable of the first-stage. The relevance assumption is not necessarily violated as it is shown that the introduction of the scheme is correlated with female house ownership. Nonetheless, the impact of the scheme is too weak across all specifications to produce a truly relevant instrument. The F-statistics found increases once subsamples of women who are more responsive to the scheme are considered, yet they never reach a value of 10 or larger. The effect of stamp duty concessions on women’s house ownership is a weak instrument, which indicates that the predictive power of the first-stage is not sufficient to draw conclusions from the IV estimates.
The assumption regarding exclusion of the instrument does not seem to be violated but this statement cannot be proven empirically. The introduction of the stamp duty scheme seems unlikely to have affected empowerment directly. Instead, it has probably acted through Domingo’s (2013) theory of change: the concession increases women’s house ownership through the creation of economic incentives for the household, which decides to register the property in the woman’s name. In turn, having a property increases woman’s economic security and creates the conditions for her to increase her empowerment and reduce her subjection to violence.
Monotonicity seems to be respected. Indeed, once controls are included, the effect of the stamp duty concession on women’s house ownership remains positive across all subsamples. Spillover effects are unlikely to be an issue in this scenario: as previously mentioned, the introduction of stamp duty concessions does not affect outcome results in a neighbouring country.
To be able to comment on the violation of the independence assumption, I run a balancing test: I estimate differences in means for all covariates across treatment and control groups with a t-test, and I correct for the significant level of the p-values obtained with the Bonferroni method (Abdi, 2007). When comparing treatment and control groups along many dimensions, it is likely that some differences will be found by chance. To avoid overestimating the number of significant mean differences, I follow the Bonferroni correction and adjust the significance level required to determine if the comparison groups are truly different along that variable: I divide the canonical level of 0.05 by the number of variables considered and find that the threshold for the p-value is 0.001 (Abdi, 2007).
Table 6 and 7 show the means and mean differences between control and treatment states at baseline for the covariates described in the methodology section. There are significant differences between the states along the vast majority of variables even at the adjusted threshold of ; < 0.001. These results suggest that the independence assumption might be violated. I find that women in treatment and control states are very different across covariates that are likely to influence the likelihood of female ownership (e.g. Peterman et al., 2017; Kelkar et al., 2015): women in treatment states live in poorer households, are less educated, are much more likely to practice purdah, and are less able to make decisions for themselves, e.g. about their spouse. I hypothesize that states might have selected into treatment and chosen to introduce the scheme precisely because of population’s characteristics: governments might have wanted to improve the social role of females, who might be more subdued to men compared to women living in other states. The fact that the differences in means are at an individual level while treatment is at a state level increases the difficulty in making conclusive claims about the violation of this assumption.
(1) (2) (3) Control States Treatment States Difference (1)-(2)
Woman’s Characteristics
Education 4.835 3.246 1.587∗
[4.611] [4.411] (0.121)
Age 33.104 32.933 0.172
[7.758] [7.568] (0.200) Household head’s wife 0.852 0.812 0.0401∗
[0.355] [0.391] (0.009) Number of children 2.444 3.207 -0.763∗ [1.415] [1.928] (0.040) Media Exposure TV 2.263 1.743 0.520∗ [0.840] [0.950] (0.023) Radio 1.577 1.558 0.019 [0.739] [0.832] (0.020) Household’s Characteristics
Highest male education 7.047 6.459 0.588∗
[4.882] [5.032] (0.130) Hindu 0.830 0.751 0.079∗ [0.376] [0.432] (0.010) Caste Brahminb 0.032 0.077 -0.045∗ [0.175] [0.266] (0.005) OBC 0.407 0.427 -0.019 [0.491] [0.495] (0.013) Scheduled caste 0.224 0.204 0.0202 [0.417] [0.403] (0.011) Scheduled tribe 0.074 0.027 0.047∗ [0.262] [0.163] (0.006) Incomea 48,901 41,276 7,625∗ [64,173] [53,646] (1,611) Water within household 1.405 1.648 -0.243∗
[0.492] [0.478] (0.013) House type 2.821 2.810 0.011
[1.681] [1.699] (0.044) Sanitation facilities type 1.831 1.820 0.011
[1.225] [1.061] (0.031)
Urban 0.385 0.283 0.101∗
[0.487] [0.451] (0.012) Observations 5,556 2,022
-∗p < 0.001
Notes: Column (1) and (2) show the means of a number of control variables at baseline in
control and treatment states respectively. Standard deviations are in parenthesis. Column (3) shows the difference in means between control and treatment states. Standard errors are in parenthesis.
aThe variable Income is measured in Rupees. bOther Backward Castes
Table 7: Balancing Test of Control Variables
(1) (2) (3)
Control States Treatment States Difference (1)-(2) Purdah Practice 0.385 0.810 -0.458∗
[0.478] [0.393] (0.012) Girls’ education same as boys 0.820 0.859 -0.039∗
[0.384] [0.348] (0.010) Age at marriage 17.803 16.752 1.051∗
[3.520] [3.564] (0.092) Any say in spouse choice 0.466 0.327 0.139∗
[0.499] [0.469] (0.013) Couple lives alone after marriage 0.090 0.127 -0.038∗
[0.286] [0.333] (0.008) Husband economically worse off 0.088 0.079 0.009
[0.284] [0.270] (0.007) Husband is not a relative 0.732 0.815 -0.082∗
[0.443] [0.389] (0.013) Expects financial support from children 0.122 0.072 0.050∗
[0.327] [0.259] (0.013) Has a bank account 0.169 0.154 0.015
[0.327] [0.361] (0.010) Knows government employees 0.306 0.294 0.013
[0.461] [0.456] (0.012) Is member of:
Self-help group 0.125 0.027 0.099∗
[0.331] [0.161] (0.008) Credit or saving group 0.079 0.030 0.049∗
[0.270] [0.171] (0.006) Development group or NGO 0.018 0.008 0.010
[0.132] [0.089] (0.003) Caste association 0.102 0.099 0.004
[0.303] [0.298] (0.008) Cooperative 0.057 0.025 0.032∗
[0.231] [0.155] (0.005) Trade union or business group 0.076 0.013 0.063∗
[0.265] [0.113] (0.006) Within the community
Families expected to care for widows 0.921 0.969 -0.0487∗
[0.270] [0.172] (0.006) Girls harassment is common 0.060 .065 -0.004 [0.239] [0.246] (0.006)
Observations 5,556 2,022
-∗p < 0.001
Notes: Column (1) and (2) show the means of a number of control variables at baseline in control and
treatment states respectively. Standard deviations are in parenthesis. Column (3) shows the difference in means between control and treatment states. Standard errors are in parenthesis.
The many differences found between control and treatment groups could be due to the fact that treatment states are northern states, while control states are both northern and southern. Patriarchal culture in the northern states is stronger than in southern ones: Agarwal (1998, p.92) says that northern states are “characterized by lower female labour participation (and higher gender disparities in participation), a higher incidence of dowry, greater intra-household discrimination against female children, and lower (to male) survival chances than southern states”. These results are also supported by Bhattacharyya, Bedi, and Chhachhi (2011) who find that southern and northern states are not culturally comparable: although still patriarchal societies, southern states provide an environment in which it is easier for women to thrive. However, this does not seem to be the case in this analysis. I run the mean comparison tests considering only the northern states in the control group. Table 8 and 9 in the Appendix show that the number of significantly different means remains large: women in the treatment group have 1.2 years less of education, are less exposed to media, and are less likely to have a bank account. Households in treatment states earn 12,000 Rs less compared to control households, and are more frequently located in rural areas. These result support the hypothesis that there might have been a selection into treatment.
The possible violation of the independence assumption is closely linked to comments on the common trend assumption needed for the DiD specification. In theory, a DiD model is not concerned with baseline differences across control and treatment states but only needs a common trend in the dependent variable, i.e. female house ownership. Nonetheless, the possible selection into treatment of some states is suggestive of the fact that these states might have wished to “catch up” with the rest of the country. Thus, it is possible that female house ownership in control states was growing at a faster rate before the introduction of the policy compared to treatment states.
7. Discussion
In addition to the comments made about the reliability of the assumption of the model used, there are some more general comments that can be made with respect to the results. First and foremost, that fact that, as in the case of previous research and as suggested by Peterman at al. (2017), the results of this thesis might suffer from lack of generalisability and transferability to different settings.
Moreover, the data used may not have been detailed enough to control for fundamental elements such as women’s employment and wage. The data used lacks
information about how long the house has been in the woman’s possession and if it was acquired before or after marriage. These differences could deeply affect the type of power relation existing between husband and wife (Pandey, 2010). Furthermore, it is possible that there is a non-linear relationship between female asset ownership and domestic abuse as husbands’ responses to their wives’ being property owners might change over time. Yet another piece of information that is not present in the data but that could have an impact in men’s response to women’s asset ownership is men’s asset ownership (Peterman et al., 2017). The economic incentives offered in the tax reduction may not have been sufficient to push a large enough number of families to go against traditions and register the property in the woman’s name. In the case of Uttar Pradesh, families are only able to save Rs. 2’500 on the first Rs. 250’000 of the value of the property. This is equivalent to almost 7% of the average income at baseline in the state10. In Uttarakhand concessions were initially only Rs. 1’000, but, thanks to the amendment in 2009, families could save Rs 2’500 on the first Rs. 200’000. This is equivalent to a 2% of the state’s average baseline income before 2009, and 5% after 200911. Lastly, in Assam there is no threshold on the value of property: stamp duty related savings can go from Rs. 2’500 for a property valued Rs. 250’000, up to, for instance, Rs. 10’000 for a property valued Rs. 1’000’000. The concession on this interval go from 5% to 20% of the state’s average income at baseline12. These incentives might not have been large enough to cause the social change described in Pandey (2010).
It is also possible that the incentives would have been sufficient to provoke social change, but limited success of the scheme may be due to the fact that information about the concessions was not sufficiently spread among women. Singh and Arora (2013) do report fairly low rates of awareness about the scheme which suggest that the impact on the number of women’s owners could have been larger had more people known about it.
The impact of the concession could have also been reduced by Power of Attorney being used to transfer properties. Power of Attorney gives power to a person to act on behalf of the person executing the right on the property. Legally, it cannot be used to transfer right,
10 Concession for women are 4% of property value until Rs. 250’000 compared to men’s 5%, but after the
threshold the rate is the same for men and women. This mean that on the first Rs. 250’000 women have to pay a stamp duty trax of 10’000 while men’s is 12’500. Average state’s income at baseline is Rs. 38’000.
11 Concessions for women were 4.5% but later reduced to 3.75% on the first Rs. 200’000 compared to men’s
5%. Women had to pay a stamp duty of Rs. 9’000 to later be reduced to Rs. 7’500 while men’s if Rs. 10’000. The state’s average income at baseline is Rs. 50’243.
12 In Assam stamp duty rates are 5% for women and 6% for men. This means that for a property of value Rs.
250’000 women have to pay Rs. 12’500 while men have to pay Rs. 15’000. For a much more expensive property of value Rs. 1’000’000 women have to pay Rs. 50’000 while men Rs. 60’000. The average income at baseline is Rs. 49’636.
title, or interest of an immovable property but Singh and Arora (2013) report that a fifth of respondents still used it to transfer the property to another family member13. The authors also report that a number of women do not apply for the required mutation, i.e. change the name of tax payer in the government records, as this is a lengthy and complicated process. This could complicate the exercise of property rights and reduce the impact of ownership.
Lastly, the impact of house ownership on domestic violence might have been insignificant because the exit option of divorce is not a credible threat. Although divorce is possible under Hindu law, women can ask for it under very extreme grounds (Virdi, 2009) and divorcees must suffer a real social stigma (Grover, 2011).
8. Conclusion
In this thesis, I attempt to estimate the causal impact that women’s house ownership has on their empowerment and subjection to domestic violence. I consider the empowerment dimensions of money control, decision-making agency over household, children, fertility, and independent mobility. The relationship between female house ownership, empowerment, and violence incidence is prone to reverse causality or selection bias. To avoid this endogeneity, I use an IV method. The instrument I use is the introduction of a concession for stamp duty fees in certain Indian states. I find that the impact of the concession scheme is positive and significant once I introduce several controls to address observed factors that are likely to be influencing the results. I find that the causal effect of the scheme is not relevant enough to produce a strong instrument. I do obtain a stronger instrument on one specific subsample of wealthier and more educated women who are more responsive to the scheme. I complete the IV analysis on this subsample and I find that the IV estimates produced are large. The instrument is probably not strong enough, and the results are likely to be picking up on trends from unobserved dimensions.
The results obtained suggest that the introduction of the concession scheme is not as relevant as expected in increasing the share of women homeowners. This, in turn, means that it could not be used as a tool for empowerment and violence reduction. The concession scheme remains an ineffective way to achieve the desired outcomes even on the subsample of women who were most responsive to its introduction. Although stamp duties have a more significant impact on wealthier and more educated women, they produce a very small – if present at all – improvement in empowerment and no improvement in domestic violence. In
sum, stamp duty concessions are an ineffective way to promote female ownership, and, in turn, they are an inefficient instrument to achieve female empowerment and the reduction of domestic violence.
Further research on the causal implication that women’s asset ownership can have on their empowerment and subjection to domestic violence is crucial to the creation and enactment of policies that will increase women’s agency and reduce domestic violence. Moreover, since the policy mostly affects women who already are in a somewhat privileged position, a more in-depth investigation on the effect of the scheme should focus on how to aid those women who are in greater need of support.
References
Abdi, H. (2007). Bonferroni and Šidák corrections for multiple comparisons. Encyclopedia of
measurement and statistics, 3, 103-107.
Agarwal, B. (1994). A Field of One’s Own: Gender and Land Rights in South Asia. Cambridge University Press, Cambridge, UK.
Agarwal, B. (1995). Some comments on “Research and policy questions on women and land rights.” Submission to Gender-Prop Email Conference, September 20, International Food Policy Research Institute, Washington, DC.
Basu, A. M., & Koolwal, G. B. (2005). Two concepts of female empowerment: Some leads From DHS data on women’s status and reproductive health. A focus on gender
Collected papers on gender using DHS data, ORC Macro, Calverton MD, 15-33. Bhattacharyya, M., Bedi, A. S., & Chhachhi, A. (2011). Marital Violence and Women’s
Employment and Property Status: Evidence from North Indian Villages. World Development, 39(9), 1676–1689. http://doi.org/10.1016/j.worlddev.2011.02.001 Cousins, B. (2009). Capitalism obscured: The limits of law and rights-based approaches to
poverty reduction and development. Journal of Peasant Studies, 36(4), 893-908. Desai, S., Vanneman, R., & National Council of Applied Economic Research, New Delhi.
India. (2005). Human Development Survey (IHDS). Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2017-05-10. https://doi.org/10.3886/ICPSR22626.v11
Desai, S., & Vanneman, R. (2011-12). India Human Development Survey-II (IHDS-II). Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2017-05-26. https://doi.org/10.3886/ICPSR36151.v5
Domingo, P. (2013). Property Rights and Development: Property Rights and Social, Political and Economic Empowerment. Overseas Development Institute.
Grover, S. (2011). ‘Purani aur nai shaadi:’Separation, Divorce, and Remarriage in the Lives of the Urban Poor in New Delhi. Asian Journal of Women's Studies, 17(1), 67-99.
Kabeer, N. (1999). Resources, agency, achievements: Reflections on the measurement of women’s empowerment. Development & Change, 30(3), 435.
Kelkar, G., Gaikwad, S., & Mandal, S. (2015). Women’s Asset Ownership and Reduction in Gender-based Violence. Landesa Rural Development Institute.
Kishor, S., & Gupta, K. (2004). Women’s empowerment in India and its states: Evidence from the NFHS. Economic and Political Weekly, 39(7), 694–712.
