The British Academy Brian Barry Prize Essay: Can Religious Norms Undermine Effective Property Rights?: Evidence from Inheritance Rights of Widows in Colonial India

Abstract

Religious norms can undermine the effects of property rights institutions. Districts in colonial India that provided widows with rights to inherit the joint-family property of their deceased husband had significantly higher widow immolations than districts that did not. Religious elites (Brahmins) burnt disproportionately more widows, and widow immolations were higher in regions with a higher density of religious elites. The findings indicate that egalitarianism requires egalitarians. Elite norms embedded in religion can mediate the effects of property rights and lead to negative consequences.

Property rights institutions are widely believed to have a positive effect on development outcomes. ¹ Social scientists have long argued that cultural norms such as intrinsic values, ² social capital ³ and interpersonal trust ⁴ mediate the effect of institutions on development. ⁵ Nevertheless, there are limits to this mediation. Strong institutions can overcome cultural barriers. ⁶ Alternatively, institutions implement the cultural preferences of those in power. If property rights are in place, they already represent the preferences of the ruling coalition. As a consequence, strong property rights institutions created by those in power should increase the welfare of the recipients of property.

This article presents one channel through which this logic might not hold. It demonstrates that property rights institutions can have a negative effect on their recipients if a subset of affected elites, who have complete control over an alternative set of institutions that can counter property rights, use these institutions to protect their own interests. Using a newly constructed dataset on widow immolations (sati) from 1815–21, the practice of widows being immolated on the funeral pyre of their deceased husbands in colonial India, this article shows that widow immolations were higher in regions that provided inheritance rights to widows. Differences in the interpretation of Hindu religious texts led to two types of legal practice – Dayabhaga (which provided equal rights to widows in the joint-family property of their deceased husbands) and Mitakshara (which did not provide this right). ⁷ In the pre-colonial period, these practices were not effectively codified, which allowed them to differ from written texts. ⁸ During colonial rule, these practices were codified as Hindu law. ⁹ We show that districts governed by Dayabhaga law had a significantly higher number of widow immolations than those governed by Mitakshara law. Inheritance rights for widows led to an average increase of between 115 and 437 widow immolations in Dayabhaga districts from 1815 to 1821 (mean of 153 for a Dayabhaga district and twenty-five for a Mitakshara district). The results are robust across different interpretations of Dayabhaga regions, accommodating alternative explanations, including geographic, institutional and demographic controls, and using alternative estimation models.

Why did pro-widow property rights laws lead to such unintended consequences? The second part of the analysis claims that cultural norms and the ability to hardwire these norms in alternate institutions can mitigate the positive effect of property rights. The Dayabhaga law was introduced in a period of widespread elite patriarchy which allowed a subset of elites to use alternate institutions that more than compensated for the rights provided by the Dayabhaga law. The analysis presents several pieces of evidence to support this hypothesis. First, based on an individual-level dataset of 533 widow immolations in Bengal Presidency in 1824 for which caste and income data were recorded, 62 per cent of the widows were Brahmins (including Kayasthas), despite comprising only 5.3 per cent of the total population, which validates that religious elites primarily followed the norm. Secondly, descriptive evidence shows that Dayabhaga districts were infamous for a range of patriarchal social institutions – such as polygamy (Kulinism) and child marriage – among the religious elites. Thirdly, the analysis shows that widow immolations were disproportionately higher in Dayabhaga regions with a higher proportion of religious elites. We estimate that a one-percentage-point change in the proportion of Brahmins led to an average increase between fifty-six and 102 widow immolations in Dayabhaga districts. Finally, official government reports reveal that the Brahmins controlled the interpretation of Hindu law and developed a doctrine of legal sati that was accepted by the colonial state until its abolition in 1829.

The empirical approach adopted in this article has distinct advantages compared to the previous literature. Rosenblum ¹⁰ and Brule ¹¹ demonstrate that providing equal inheritance rights over agricultural land to women in five Indian states during the 1980s and 1990s led to an increase in female child mortality. They argue that inheritance rights increased the costs of daughters, causing parents to reduce investment in their health. These articles require two assumptions regarding parental behavior. First, that parents prefer to leave higher bequests to their sons. Secondly, that parents engage in a long-term optimization between current health expenditure on their infant daughters and expected gender-biased bequests to their children. In contrast to previous work, the analysis in this article provides a closer fit between the institution (widows’ inheritance of joint property), the competing beneficiaries (relatives, as other claimants of the joint property), the event (death of the husband) and the victims (widows). Secondly, an economic motive is sufficient to explain the outcome: relatives prefer to burn widows in order to increase their own share of the family inheritance after the death of the husband.

The identification strategy used in this article, and described in detail in the historical background section, is credible for two reasons. First, it relies on the fact that the inheritance laws were introduced independently of the local environment. Historians argue that the divergence in the laws was a consequence of the different interpretation of a term in the Manusmriti, a fundamental text of Hindu law, by scholars residing in different jurisdictions. The subsequent codification of Hindu law during the earlier period of colonial rule (before the nineteenth century) further reduced the diversity of inheritance norms and led the jurisdiction of the two types of schools to be firmly established in legal jurisprudence. Banerjee and Iyer ¹² use a similar identification strategy in their article on the long-term effect of land tenure institutions. In their analysis, the differences in the ideology of the administrators led to spatial variation in land tenure institutions, which in turn led to variation in development outcomes. Secondly, the identification strategy is a particularly effective way of isolating the independent effect of inheritance-linked economic shocks from other economic shocks because the rite required the husband to die first. Hence widow immolations cannot be attributed to the strategic need to get rid of unproductive individuals during income shocks such as crop failure or famine, as proposed by Miguel ¹³ and Oster. ¹⁴ In effect, our identification strategy relies on the exogenous assignment of the property rights institution, as well as the shock (death of the husband). In both types of districts, the counterfactual led to the same outcome: women were not immolated when husbands did not die.

The cultural explanation offered in this article differs from cultural traditions that allude to social capital, intrinsic values or interpersonal trust as causal variables of development outcomes. ¹⁵ In this article, culture mediates the effect of institutions because patriarchal religious elites who have political power use informal institutions to counter formal institutions. This approach is also different from the constructivist tradition in the literature on ethnicity and identity. ¹⁶ While this tradition typically relies on the political salience of cultural identity due to some underlying demographic characteristic such as the size of the electorate, in this article the political salience of culture emerges from religious elites’ ability to use cultural norms to constrain fundamental economic and political institutions. ¹⁷

The following section discusses the relationship between property rights and gender outcomes, and the cultural mechanisms that mediate this relationship. Next, we provide a historical background of the inheritance rights institutions and the religious practice of widow immolations in colonial India. We then describe the data, empirical strategy and results pertaining to the two main hypotheses in the article: that widow immolations were (1) higher in districts under Dayabhaga law and immolation (2) higher in Dayabhaga districts that had a higher density of religious elites. Finally, the conclusion draws out implications for the contemporary discourse on institutional reform.

PROPERTY RIGHTS, CULTURE AND GENDER OUTCOMES

Inheritance rights are an important category of property rights institutions. ¹⁸ Gender bias in inheritance rights legislation is widespread across countries. ¹⁹ Improving women’s inheritance rights has positive effects on development outcomes. ²⁰ Equalizing the inheritance rights of sons and daughters in joint household property significantly increases the daughters’ likelihood to inherit land, improves their educational outcomes, ²¹ increases the autonomy of women within the household ²² and decreases violence against women. ²³ Apart from a direct positive effect on women’s wealth through economic channels, these studies indicate positive second-order effects that operate through a socialization channel – that is, by instituting greater gender equality in access to property rights, families are socialized to improve gender equality in other spheres.

Political scientists argue that the effects of pro-gender institutional reform on gender outcomes are also mediated by culture, more specifically by social norms of patriarchy. Researchers have presented several mechanisms. First, gender norms and pro-gender institutions can be endogenous; that is, institutions change as norms of patriarchy change and vice versa. For instance, franchise extensions for women emerged in periods characterized by women’s greater participation in public life. ²⁴

Secondly, pro-gender institutions can bring about an independent change in gender norms, and as a result change gender outcomes. For example, in Western India, electoral quotas for women still provided benefits even after their withdrawal; quotas allow parties and voters to learn that women can win elections and run offices. ²⁵ Similarly, quotas for women in council elections in India led to an improvement in development outcomes for women; that is, infrastructure investments are increasingly directed toward women’s needs. ²⁶ In both of these cases, positive welfare effects are a result of a change in gender norms. The devolution of decision-making power in favor of women leads to a positive change in gender outcomes.

Thirdly, pro-gender institutional reform need not bring about a change in gender norms, yet change gender outcomes. For example, a change in property rights in favor of women in rural India did not increase the amount of land inherited by women in general, though women’s land ownership increased in regions where women were represented in the local government. ²⁷ The local official’s potential for complicity in resisting reform is high where officials are accountable to male household heads who benefit from the status quo. Similarly, a survey of 744 married women in urban South India shows that enhancing women’s social or economic resources through self-help groups or vocational training increases their risk of domestic violence. ²⁸ In these studies pro-gender institutional reforms did not change gender norms, and the reforms led to positive effects in regions where prevailing norms were favorable toward women and negative effects in regions where the prevailing norms were more patriarchal.

In addition to the effect of societal norms, it is also possible for cultural elites to mediate pro-gender institutional reforms either through the socialization channel or the institutional channel. Through the socialization channel, elites can encourage or discourage the adoption of institutions using their ability to influence the behavior of non-elites. ²⁹ The role of the Catholic Church in encouraging women’s inheritance rights in the nineteenth century is a case in point. The church defended the right of women to inherit a portion of their husband’s property primarily out of its own economic interests. ³⁰ The church not only became a natural candidate for their bequests, but by exalting virginity, a religious norm, the church was able to create a sizable class of female faithfuls with properties inherited from their husbands and fathers. Secondly, cultural elites can also mediate the effect of institutional reform using an institutional channel, either formal or informal. For instance, elites in the Middle East used waqfs (an unincorporated trust with high setup costs) to shelter personal assets from egalitarian inheritance institutions under Islamic law. ³¹ In line with the elite-led institutional channel, religious elites in colonial India could have used widow immolations, a religious institution, to counter the effects of formal property rights institutions.

Cultural elites could have found it particularly feasible to vitiate the effect of inheritance rights during the period when women appeared less likely to exercise these rights. Other forms of violence against women, such as domestic violence, are positively related to lower financial autonomy, ³² lower incomes, ³³ lack of property ³⁴ and lower labor force participation of female partners. ³⁵ The indicated mechanism is that spousal abuse increases with a decrease in the outside options or bargaining power of female partners. A similar argument is proposed regarding dowry. Husbands use violence against wives to extract transfers from their wives’ parents. ³⁶ In line with these results, widow immolations, both real and threatened, could have served as a mechanism to prevent widows from exercising their inheritance rights in a period when they had limited outside options. In the 1820s, widows were prohibited from remarriage until the East India Company enacted the Hindu Widows’ Remarriage Act in 1856. Women’s participation in the labor market was also quite limited.

Negative income shocks can also increase violence against women (dowry deaths) if households eliminate unproductive partners to smooth household consumption. ³⁷ In line with the income-smoothing theory, the death of a productive member of the joint family could have increased the burden on other productive members of the family. This burden could have been more pronounced if the relatives did not inherit any property from the deceased or if the deceased was not substituted with a new productive member through remarriage (since widow remarriage was not legalized until 1856). In such an environment, the benefits of widow immolations for the remaining relatives of the deceased could have been high.

HISTORICAL BACKGROUND

The Divergence in Inheritance Laws

The officials of the East India Company considered the Manusmriti or Ordinances of Manu (approx. 500 CE) as a fundamental text of Hindu law. William Jones, the founder of the Asiatic Society, selected it as the first Hindu legal text for translation into English in 1794. ³⁸ The Manusmriti is divided into twelve chapters that provide norms of behavior for individuals and the state. These norms govern education, marriage, economic transactions (private morality), gender relations, religion, governance, military, judiciary, class relations, immigration and crime. Dayabhaga and Mitakshara, written by Jimutavahana (approx. 1200–1500 CE) and Vijnaneshwara (approx. 1200 CE), respectively, are commentaries on the Manusmriti and other legal texts. These commentaries mainly discuss inheritance rules. In the context of widows’ inheritance rights, the differences in the two schools arose because the two scholars interpreted the text in the Manusmriti differently. ³⁹ Chapter 9 of the Manusmriti includes rules on inheritance; verse 187 states ‘He who inherits the property, also offers the pinda.’ ⁴⁰ Vijnaneshwara, in the Mitakshara, interpreted pinda as flesh and blood due to which widows were not permitted to have coparcenary rights (the rights to inherit joint property) since they were not related by blood to the deceased. Jimutavahana, in Dayabhaga, interpreted pinda as the balls of rice offered as oblations, which permitted widows as heirs to joint property. ⁴¹ Historians in the late 19^th century argued Jimutavahana was patronized by Adisura Sen, the ruler of Bengal, hence Dayabhaga prevailed in Bengal, while Mitakshara law and its related schools were adopted under the Chalukya dynasty in south, west and central India. ⁴² The codification of practice into formal legislation during colonial rule led to the stricter adoption of these laws, compared to possible differences between written texts and prevailing practices in earlier periods. ⁴³ Warren Hastings, the first governor general of Bengal, commissioned a digest of Hindu Law in 1775, A Code of Gentoo Laws or Ordinations of the Pundits, which was developed by a committee of Hindu Brahmins. ⁴⁴ In 1798, Charles Cornwallis, the third governor general, commissioned a Digest of Hindu Law on Contracts and Successions, which was also guided by Hindu Brahmins. ⁴⁵ Since the administrators relied on a small committee of Brahmins, the codified laws were biased in favor of texts favored by a subset of that caste and related less to the practices of non-Brahmin Hindus. ⁴⁶

The Religious Practice of Widow Immolation (Sati)

Prior to the eighteenth century, sati was frequently documented in the accounts of European travelers. ⁴⁷ One such description is also found in the travel accounts of Marco Polo, the Italian explorer, toward the beginning of the thirteenth century. ⁴⁸ The Mughals, from Muhammed Tughlak (1325–1351) until Aurangzeb (1618–1707), were critical of the practice and followed a policy of dissuasion. The Mughal Emperor Akbar (1542–1605) is noted to have said ‘It is a strange commentary on the magnanimity of men that they should see their deliverance through the self-sacrifice of their wives.’ ⁴⁹ Under Mughal rule, widows were asked to seek permission, with the intent of using bureaucratic hurdles to discourage them. ⁵⁰ An absolute ban was not initiated due to fear of a Hindu revolt. ⁵¹

In 1787, a widow immolation was described in an official report of the East India Company. An official policy toward widow immolations emerged in 1812, eight years after Lord Wellesley, the governor general of India at the time, referred the issue to the Nizamat Adalat, the High Court established by the East India Company. ⁵² One of the earliest debates regarding widow immolations pertained to its legality. Colebrooke, responsible for translating Hindu legal texts including the Dayabhaga and the Mitakshara, observed that Hindu law permitted widow immolations under specific conditions but did not prescribe it. According to Hindu law, widow immolations were permitted if the widow was (1) not forced, (2) not intoxicated, (3) not pregnant and (4) did not have a child under three years whose nurture could not be provided for. ⁵³ The policy in 1812 categorized widow immolations as legal or illegal. The intent was to target the ‘gross abuses that had crept in due to the ignoble intentions of the Brahmins and the instigations of the greedy relations of deceased husbands’. ⁵⁴

A second debate prior to the abolition of widow immolations pertained to the voluntary nature of the act. Official communication in the 1820s reveals that colonial administrators believed widows were largely forced into the act despite the law categorizing involuntary widow immolations as illegal. Colonial accounts of widow immolations point to the common practices of tying widows to the funeral pyre, pushing them onto the pyre during their attempts to escape and using bamboo sticks to hold them on the pyre as evidence of the involuntary nature of the act. ⁵⁵

[…] the victim was laid down by the side of her deceased husband; and if she was not bound by her clothes with the deceased, yet they were so wrapped about her and the deceased that it would have been difficult for her to disentangle herself from them; upon her were then laid logs of wood, some of them of a size and weight such as but one man could lift and carry: a minute or two elapsed in this portion of the ceremony, when the pile was set fire to, and immediately after a large green bamboo was placed across and held down by two men on each side, so that it rendered any effort on the part of the victim to escape wholly unavailing. Scarcely had the funeral pyre been lighted, when she made a most violent attempt to escape from the flames, but the nature and construction of the pile rendered any immediate effort to release her from the flames, with the chance of escaping with her life, quite impracticable. ⁵⁶

The Relationship Between Inheritance Laws and Widow Immolations

Colonial administrative reports reveal that the government was aware of the relationship between Dayabhaga law and widow immolations. ⁵⁷ Judges, in court cases on civil disputes relating to property in the Sudr Dewanee Adalat, the High Court of Civil and Revenue Jurisdiction established in 1772, often blamed Dayabhaga law for the widow immolations.

There can be no doubt that the prevalence [of sati] may be ascribed to the cupidity of the male relations of the deceased, between whom and the property the right of the widow intervened under the law of the Dayabhaga. Wherever this law prevailed, that is throughout Bengal Proper, sati was of continual occurrence; while in Berar and Benaras and Cuttack, where the rule of the Mitakshara was observed, which excludes widows altogether from inheritance it was rare, evidently because the male relations had not the same direct personal interest in the destruction of the widow. ⁵⁸

I do not hesitate in offering my opinion that a law for the abolition [of sati] would only be objected to by the heirs, who derive worldly profit from the custom. ⁵⁹

DATA AND EMPIRICAL STRATEGY

The first part of the empirical analysis demonstrates that positive property rights institutions led to negative effects on their intended beneficiaries. Widow immolations were more prevalent in districts where widows received inheritance rights in the joint-family property of the deceased husband compared to districts where they did not receive these rights. To test this or towards testing this hypothesis, the dependent variable, WidowImmolations _i , is the count of the total number of widow immolations from 1815 to 1821 at the district level, as listed in the returns annexed in the Parliamentary Papers on Widow Immolation. ⁶⁰ In 1815, the Nizamat Adalat ordered the district officials to report all widow immolations, a practice that continued until its abolition in 1829. The statistics are available for seventy-nine districts of British India in the three presidencies, Bengal, Madras and Bombay. In the first ten years, the courts recorded 6,632 widow immolations, 5,997 of which were in the Bengal Presidency. ⁶¹ In Bengal, the statistical returns provide data from 1815 to 1828, in Bombay from 1815 to 1827, and in Madras from 1815 to 1821. The analysis restricts the number of satis to those up to 1821, which is the last year for which data were collected uniformly across British India. ⁶²

The main independent variable, DayabhagaCont1 _i , is a proxy for the proportion of the district governed by the positive inheritance rights law, Dayabhaga. Since ‘the Bengal School [Dayabhaga] prevails over Bengal proper and is co-extensive with the Bengali language’, ⁶³ DayabhagaCont1 _i is measured as the proportion of the population in a district that speaks Bengali as their first language according to the 1881 Census of British India, which included questions on language for the first time. The Dayabhaga law was applicable only to Hindus in Bengal. Hence the analysis develops an alternative variable, DayabhagaCont2 _i , which is an interaction of the proportion of Bengali speakers and the proportion of Hindus in a district in 1881. While the continuous measures of Dayabhaga rely on 1881 data, there is no reason to believe that the distribution of language or religion changed materially within this period or that widow immolations affected the distribution of language or religion within the regions. As an alternative measure, the analysis develops a dummy variable, DayabhagaDummy _i , which takes a value of 1 if the district is a part of ‘Bengal Proper’ or Bengal Province in the Bengal Presidency. In the analysis, among the fifty-nine districts in Bengal Presidency, nineteen form part of Bengal Province.

While the results rely on an arguably credible two-stage identification strategy as described above, it is likely that the districts governed by the two types of inheritance laws might be different in other characteristics, and that these characteristics might effect the number of widow immolations. We identify these characteristics based on the hypotheses for the variation in widow immolations in the existing literature and use them as controls, after presenting the results without controls. The control variables are divided into two categories, pre-treatment and post-treatment. Systematic district-level data on demographic and economic variables are not available prior to the first Census of India conducted in 1871, forty years after the abolition of sati. Hence the analysis uses post-treatment variables under the assumption that widow immolations would not have affected these variables. The results and analysis do not rely on post-treatment controls, but use them as additional proxies of pre-treatment controls. By using the control variables we are able to rule out or account for alternative explanations of widow immolations, as well as control for geographic and demographic variation.

Economic-based arguments hypothesize that widow immolations were a consequence of income shocks due to low agricultural productivity and famine. In Bengal, immolations have been attributed to frequent drought – ‘the rite was a primitive Malthusian means of population control in famine-ridden Bengal’. ⁶⁴ The analysis uses soil type (Alluvial _i ) and famine exposure (Famine _i ) to control for these alternative economic arguments. Regions with a higher density of alluvial soil are generally considered to be more productive. ⁶⁵ Famine exposure is the count of the number of famine years prior to 1821 as recorded in the Administration Manual of the Madras Presidency (1885). In addition, the analysis uses land revenue per capita (LandRevenuePerCapita _i ) and land revenue per square mile in 1871 (LandRevenuePerSq.Mile _i ) as an alternative measure of poverty. Poorer regions are likely to provide lesser land revenue per capita than richer regions. If these alternative economic explanations determined widow immolations, the coefficient is expected to be negative on Alluvial _i , LandRevenuePerCapita _i and LandRevenuePerSq.Mile _i , and positive on Famine _i .

Widow immolations have also been attributed to European cultural influence through two possible channels. First, religious elites could have revived a previously unpopular practice as a response against European cultural interventions. ⁶⁶ After the government instituted a monitoring system, satis rose sharply. Religious elites could have encouraged the practice in response to colonial intervention in the cultural sphere. ⁶⁷ Colonel James Sutherland, a resident of the Gwalior State, makes a similar claim regarding attempts to discourage the practice in the Princely States. Sutherland believed that inconclusive attempts by the colonial state to limit widow immolations between 1813 and 1829 reinforced rather than reduced the practice. ⁶⁸ Secondly, the classification of the practice into legal and illegal forms and the resultant publicity due to the issuance of periodic circulars could have lent support to the practice. ⁶⁹

To account for the possibility that widow immolations could have been a response to the cultural influence of the colonizers, the analysis uses two variables, Coastal _i and DirectLandTenureProp _i . ⁷⁰ Coastal regions are more likely to be exposed to colonial officials since the East India Company was heavily involved in sea trade. Land tenure systems mattered because they determined the proximity of the British to the cultivator. Cultivable land in British India was classified into two broad types of land tenure regimes, cultivator based (direct) and landlord based (indirect). The cultivators had direct relationships with the colonial state in regions where land tenure was granted directly to the cultivators as opposed to the landlords. Since cultivator-based land tenure systems also required the colonial state to develop a more entrenched administrative structure than landlord-based systems, British cultural intervention is likely to be higher under the former system. ⁷¹ As an alternative, the analysis uses the number of Europeans (No.Europeans _i ) and the percent of Europeans (EuropeanPercent _i ) in a district enumerated in the first Census of India conducted in 1871, which is again a post-treatment variable but better captures European cultural influence. If cultural influence increased widow immolations, the coefficient is expected to be positive on Coastal _i , DirectLandTenureProp _i , No.Europeans _i and EuropeanPercent _i .

While Brahmin density is able to control for differences in the proportion of elites across districts, and the patriarchy that is specific to elites, it is also possible that society-level differences in patriarchy across districts might explain widow immolations. To test whether the overall level of patriarchy in a society determined the frequency of the practice, the analysis uses the ratio of male to female scholars at school in a district (MaletoFemaleScholars _i ). The number of children at school included children across faiths and castes. If differences in the level of societal patriarchy determined widow immolations, we should expect the coefficient on MaletoFemaleScholars _i to be positive.

In addition, the analysis controls for geographic variation. During the period of analysis, the East India Company governed British India by dividing it into three presidencies, Bombay, Madras and Bengal. ⁷² Presidency fixed effects control for any administrative variation across these governing units. Population _i is used to control for variation in population across the districts. The population data is again based on the 1871 Census. The Descriptive Statistics are presented in Appendix 1.

The balance statistics in Table 1 show that Dayabhaga regions had a lower proportion of Brahmins, a lower proportion of Europeans, lower per capita land revenue, a lower ratio of male to female scholars and a lower proportion of direct land tenure. The two sets of districts were not entirely similar. Nevertheless, by including these controls in the regressions we are able to account for the variation in widow immolations due to differences in these characteristics.

Table 1 Balance Statistics on Main Covariates

Note: the balance statistics for DayabhagaCont2 are similar to those for DayabhagaCont1.

To examine whether Dayabhaga law, which provided equal rights to widows in the joint-family property of their deceased husband, led to an increase in widow immolations, the following specification is used:

(1) $$WidowImmolations_{i} \,{\equals}\, \alpha _{0} {\plus}\alpha _{1} \left( {Dayabhaga_{i} } \right){\plus}X_{{ij}} \gamma {\plus}{\varepsilon}_{t} ,$$

where WidowImmolations _i represents the number of widows burnt in district i from 1815 to 1821, Dayabhaga _i is the proportion of the district governed by Dayabhaga law, and X _ij is a set of j economic, demographic, geographic and institutional controls. In Equation 1, a change in a district from non-Dayabhaga to Dayabhaga leads to an α ₁ increase in the number of widow immolations from 1815 to 1821. Based on the hypothesis that the Dayabhaga law increased the number of widow immolations because it provided widows with inheritance rights in the joint property of their deceased husbands, which gave the relatives of the deceased husband an incentive to kill them, we should expect the effect of Dayabhaga _i on WidowImmolations _i to be positive (α ₁>0).

The second part of the analysis presents a mechanism to explain why more widow immolations took place in districts where widows received inheritance rights – elite patriarchy. Widow immolations were higher in Dayabhaga regions because Hindu religious elites (Brahmins), who were more patriarchal, were able to use religious norms to counter the inheritance legislation. To test the mechanism, the analysis interacts the percent of Brahmins (BrahminPercent _i ) in a district with the proportion of the district under positive inheritance rights, Dayabhaga _i . The correlation between BrahminPercent _i recorded in the first census in 1871 and the census in 1931, which recorded caste data for Brahmins for the last time, is 93 per cent, which supports the use of post-treatment demographic variables from the 1871 Census to capture the approximate demography in 1815. To demonstrate that widow immolations were higher in Dayabhaga districts with a higher density of religious elites, the analysis uses the following specification.

(2) $$WidowImmolations_{i} \,{\equals}\,\beta _{0} {\plus}\beta _{1} \left( {Brahmin_{i} X\,Dayabhaga_{i} } \right){\plus}\beta _{2} \,Dayabhaga_{i} {\plus}\beta _{3} \,Brahmin_{i} {\plus}X_{{ij}} \gamma {\plus}{\varepsilon}_{t} $$

where WidowImmolations _i represents the number of widows burnt in district i from 1815 to 1821, Brahmin _i is a variable that captures the percentage of religious elites in the population, and X _ij is a set of j economic, demographic, geographic and institutional controls. In Equation 2, a one-percentage-point change in the density of religious elites in Dayabhaga districts is correlated with a β ₁ increase in the number of widow immolations in Dayabhaga districts from 1815 to 1821. Based on the hypothesis that patriarchal religious elites were responsible for the increase in the number of widow immolations in Dayabhaga districts, we should expect the effect of Brahmin _i × Dayabhaga _i on WidowImmolations _i to be positive (β ₁>0).

The main effect and the mechanism are estimated using a linear regression model. Since the variance-to-mean ratio is greater than one (2.46), the results are also estimated using a negative binomial model in the Appendix. The results use three measures of Dayabhaga jurisdiction as explained above to account for differences in the legal interpretation of Dayabhaga jurisdiction. For each measure of the jurisdiction, four types of models are used. The first model shows the uncontaminated results with no controls. The second model shows the results with pre-treatment controls. The third model shows the results with post-treatment controls since the first census was conducted in 1871. Finally, the fourth model includes both pre-treatment and post-treatment controls with the presumption that these variables capture different data-generating processes.

MAIN RESULT

Based on the above discussion, the primary claim made here is that positive inheritance rights can have a negative effect on their intended beneficiaries. Widow immolations were higher in districts governed by an inheritance law that provided equal inheritance rights to widows in the joint-family property of their deceased husband. Table 2, which presents the estimates associated with Equation 1, shows that the results are in line with this hypothesis. The estimated coefficient α ₁ is consistently positive and significant across different measures of the jurisdiction of the positive inheritance rights law (Dayabhaga) and different sets of control variables. In models that use a dummy variable to indicate Dayabhaga jurisdiction (Columns 1, 2, 3 and 4 of Table 2), the results demonstrate that 115.23 to 185.87 additional widows were burnt in Dayabhaga districts. In models that define the jurisdiction of Dayabhaga as the proportion of Bengali speakers (Columns 5, 6, 7 and 8), the results demonstrate that 131.92 to 215.64 additional widows were burnt in Dayabhaga districts. In models that define the jurisdiction of Dayabhaga as the product of the proportion of Hindus and the proportion of Bengali speakers (Columns 9, 10, 11, and 12), the size of the estimated effect increases substantially. This measure of Dayabhaga is the closest to the expected jurisdiction for two reasons. First, Dayabhaga law was applicable only to Hindus, and secondly, it applied only to Bengalis. By multiplying the Hindu proportion by the Bengali proportion, the measure is able to account for those regions in Bengal that had a larger Muslim population and as a result were not governed by the Hindu Dayabhaga law. The results demonstrate that 363.89 to 436.71 additional widows were burnt in Dayabhaga regions. ⁷³ Among the control variables, the density of religious elites is significant across all models. Based on the estimates in Table 2, a one-percentage-point change in the proportion of Brahmins in a district correlates to an increase of 10.91 to 15.49 widow immolations.

Table 2 Effect of Inheritance Laws on Widow Immolations (OLS)

Note: DayabhagaCont1 is a continuous variable coded as the proportion of Bengali-speaking population in a district in 1881. DayabhagaCont2 is a continuous variable coded as DayabhagaCont1 multiplied by the proportion of Hindus in a district in 1881. Standard errors in parentheses. *p<0.1, **p<0.05, ***p<0.01.

While the main analysis included controls for alternative explanations, the results in the Appendix use additional variables to test the dominant alternative explanations regarding widow immolations – namely cultural interference by the colonizers and poverty. The results in Appendix 4 do not support the hypothesis that colonial presence led to an increase in the number of widow immolations. One reason could be that districts with a higher European presence such as coastal regions were also trading regions and socialized to a different set of norms that was less favorable towards widow immolations. Medieval trading ports in India were also less likely to experience Hindu–Muslim religious violence between 1850–1950 because inter-religious co-operation to share gains from trade led to a sustained legacy of religious tolerance. ⁷⁴ The results in Appendix 5 show no support for income- or poverty-based arguments for widow immolations. Although similar forms of ritual violence such as witch hunts have been attributed to the need to get rid of unproductive persons in periods of income shocks, ⁷⁵ there is an obvious reason why the income-smoothing or ‘eliminating unproductive persons’ hypothesis does not apply here. The religious rite required the husband to die first, and in colonial India in the early nineteenth century the husband was more likely to be the productive member of the family for economic activity outside the household.

MECHANISM

The mechanism hypothesized in this article to explain why positive institutions can have a negative effect on their intended beneficiaries is that elites’ cultural norms mediate this effect. This mediation is material because elites have the political power to institute their norms to counter institutions. Widow immolations were higher in Dayabhaga districts because Brahmin norms were more patriarchal, and Brahmins were able to use the rite to counter the inheritance law. To test this mechanism, we demonstrate that the effect of the pro-widow inheritance law on widow immolations intensifies with an increase in the density of religious elites. Table 3, which presents the estimates associated with Equation 2, supports the hypothesized mechanism that widow immolations increase in Dayabhaga regions with a higher density of Brahmins. The estimated coefficient β ₁ is consistently positive and significant across different measures of the jurisdiction of the positive inheritance law (Dayabhaga) and different sets of control variables. The estimated coefficients are significant when a dummy variable is used as a measure of Dayabhaga jurisdiction (Columns 1, 2, 3, and 4). The models demonstrate that a one-percentage-point increase in Brahmin density is correlated with 55.98 to 57.49 additional widow immolations in Dayabhaga districts. In the columns where the jurisdiction of Dayabhaga is defined as the proportion of Bengali speakers (Columns 5, 6, 7, and 8), the models demonstrate that a one-percentage-point increase in the density of Brahmins is correlated with 76.27 to 78.24 additional widow immolations in Dayabhaga districts. Finally, in models that define the jurisdiction of Dayabhaga as the product of the proportion of Hindus and the proportion of Bengali speakers (Columns 9, 10, 11 and 12), the estimated effect size increases, arguably because it takes into account districts that have a high proportion of Muslims who also speak Bengali but are not governed by Dayabhaga law. The models demonstrate that a one-percentage point increase in Brahmin density in Dayabhaga districts is correlated with 93.90 to 101.90 additional widow immolations. The results in Table 3 generally support the hypothesis that the effect of the pro-widow inheritance law on widow immolations intensifies as the density of religious elites in a Dayabhaga district increases. ⁷⁶

Table 3 Mechanism: Heterogeneous Effect of Religious Elites on Widow Immolations

Note: DayabhagaConti is a continuous variable coded as the proportion of Bengali-speaking population in a district. DayabhagaCont2i is a continuous variable coded as DayabhagaConti multiplied by the proportion of Hindus in a district. Standard errors in parentheses. *p<0.1, **p<0.05, ***p<0.01.

While the results empirically demonstrate that elite density mattered, additional evidence shows that elites were indeed more patriarchal. Based on an individual-level dataset of 533 widow immolations in Bengal Presidency in 1824 for which caste and income data were recorded, 62 per cent of the widows were Brahmins (including kayasthas) despite comprising only 5.3 per cent of the total population, validating the hypothesis that the practice was primarily used by Brahmin elites against Brahmin women. ⁷⁷ Descriptive evidence from official colonial documents also suggests that religious elites in Dayabhaga regions were more patriarchal. Historians argue that the social environment in Bengal, the hotbed of widow immolations, was characterized by widespread patriarchy among the Brahmins. ‘Extreme hypergamy of the Brahmins in Bengal led to a large number of young widows unable to remarry due to restrictive laws.’ ⁷⁸ A survey of 426 villages in Bengal counted 447 Kulin Brahmins having 3,360 wives, averaging 7.5 wives each. ⁷⁹ Kulinism (polygamy) was partially abolished in Bengal in 1872, much later than the abolition of sati in 1829. ⁸⁰ The Hindu Widows’ Remarriage Act 1856 (Act XV) that legitimized widow remarriage was introduced twenty-six years after sati was criminalized.

Finally, religious elites were not only more patriarchal; they were also in a position to institute patriarchal norms within religious rites. The colonial state generally followed a policy of non-interference in the cultural realm in the earlier years of colonial rule. This allowed religious elites to control religious norms. Religious elites exclusively informed the codes of Hindu law, as practiced in the Courts of Nizamut Adalat and Dewanee Adalat, High Courts of Criminal and Civil disputes, respectively. In 1813, the Nizamut Adalat relied on Sanskrit legal texts as interpreted by Hindu Brahmins to issue directions regarding the legality of widow immolations. ⁸¹ The formal codification of widow immolation into legal and illegal types provided a sanctified religious mechanism to eliminate widows and arguably also sent a signal to existing widows to discourage their pursuit of inheritance. Taken together, the analysis provides evidence that Brahmins were more patriarchal (more active in widow immolations and in practicing other forms of patriarchy) and able to institute religious rites to counter inheritance rights laws, and that widow immolations were higher in Dayabhaga districts with a higher density of Brahmins.

CONCLUSION

The beastliness of the Brahmins reaches its height in the ‘Suttee’ or widow burning. Strange considers this practice to be a ‘malus usus,’ not ‘law,’ since in the Manu and other high authorities there is no mention of it […]. The matter is clear: the suttee is simply religious murder, in part to bring the inheritance into the hands of the (spiritual) Brahmins for the religious ceremonies for the deceased husband and in part through Brahmin legislation to transfer the inheritance of the widow to the closest in the genes, the nearer family of the husband […]. Although suttee an innovation introduced by the Brahmins, in the Brahmin mind this innovation was conceived as a survival from the older barbarians (who had buried a man with his possessions)! Let it rest. ⁸²

Henry Maine, ⁸³ in his seminal work on comparative law, noted ‘there are in fact clear indications of a sustained general effort on the part of Brahminical writers on mixed law and religion, to limit the privileges of women which they seem to have found recognized by older authorities’. Marx, generally critical of Maine’s theories, aligned with his view regarding the role of Brahmins in the systematic suppression of women’s rights in India. Material interests such as property and income from sati were only one part of the story – a point that has not been emphasized in this article. Elite patriarchy was over and above elite materialism. If a widow was rendered powerless purely on the grounds of the death of her husband and the collusion of the relatives interested in the property, a norm of widower immolations should also have emerged. However, widowers were not burnt. The limited participation of women in the public and economic spheres could have prevented the collective action necessary for the emergence of the norm of widower immolation. ⁸⁴ Similar to other practices such as polygamy, dowry and child marriage, widow immolations were symptomatic of the Brahmin patriarchal norms prevalent during the period.

We show that property rights institutions in colonial India that provided inheritance rights to widows in the joint-family property of the deceased husband had negative consequences. Widow immolations were higher in districts under Dayabhaga law. In line with the intuition of many nineteenth century scholars and legal practitioners who studied the Hindu practice of widow immolations, our analysis suggests that elite religious norms and elite control over alternate institutions are one possible channel. The patriarchy of religious elites and their ability to use sati against politically established property rights institutions caused positive property rights to have negative effects on their recipients – a possible case of an illiberal religion acting against a liberal state. ⁸⁵

More importantly, the results point to dilemmas in institutional reform. If norms and institutions are in opposition, as indicated by the misalignment of patriarchal norms and positive inheritance institutions discussed in this article, how should policies address this misalignment? This article indicates that it matters whose norms are misaligned – elite or the masses – and who is in control. If the norms of the citizens are misaligned (for example, if citizens’ norms are more conservative) and the citizens are in control, such as in democracies, civic mobilization is largely believed to be useful for strengthening institutions. In the context of gender rights, civil society mobilization is considered to be more influential in combating violence against women than legislative institutions. ⁸⁶ The underlying presumption is that civil society actors in democracies, aided by democratic liberties such as freedom of speech, can use civic engagement to change norms. A change in norms, in turn, can cause institutions to be more effective. However if elites are in control, as this article indicates, mobilization might not always be effective. Brahmins countered positive property rights institutions with sati. If elite norms are regressive and elites can hardwire regressive norms in alternative institutions that are not accountable to the public, civic mobilization will fall short. In these situations, it might be necessary to reduce elites’ influence on institutions.

Political scientists point to institutional secularization as a way to reduce the influence of religious elites. ⁸⁷ Again, for institutional secularization to work it matters whose norms are misaligned and who is in control. Secularization from above might lead to negative consequences in an environment where the norms of the masses are conservative and elites are in control. For example, Turkish municipalities governed by non-secular Islamic parties have large positive effects on women’s education and employment. ⁸⁸ The evidence indicates that poor and orthodox Muslims face significant barriers to participation in municipalities where secular parties enact policies such as headscarf bans. Hence education outcomes improve in regions where Islamic parties win because these secular policies are not implemented and parents are more likely to send their girls to school. In the case of Turkey, secular elites, who had political power, attempted to secularize the masses from above by instituting a ban on headscarves. However, if conservative elites have political power and there is no democratic mechanism for changing norms, like in the case of non-democratic colonial India, how will secularization be brought about? Elites are likely to block reform if they expect reforms to reduce their political power. ⁸⁹ The government was able to ban widow immolations in colonial India forty years after their first official report on the practice. One reason for this delay was the political power of religious elites. In the 1820s, the religious elites split into reformist and conservative factions. The reformist faction, led by Ram Mohan Roy, advocated the abolition of sati, which allowed the colonial state to ban the practice in 1829.

The juxtaposition of religious elites with political power and a non-democratic regime renders a skeptical view of mobilization-based efforts to change cultural norms. In the context of gender rights, some Islamic states are believed to privilege practices that deprive women of property rights, which leads to a lack of progress in women’s rights. ⁹⁰ A common recommendation is that advocates should concentrate their efforts on increasing women’s education and labor force participation. ⁹¹ However, similar to the environment in the earlier days of colonial rule in India, elites in some non-democratic Islamic states control both institutions and norms (by controlling the ability of groups to mobilize). This ability leads to a double bind: institutional secularization might not precede mobilization because conservative elites control institutions, and mobilization might not precede institutional secularization because the democratic mechanism of mobilization is strictly controlled. ⁹² In these situations, a progressive citizenry might be no match for a conservative elite.