Inheritance Law Reform, Empowerment, and Human Capital Accumulation: Second-Generation Effects from India

Abstract Although many studies point towards significant positive impacts of Hindu Succession Act (HSA) reforms on females’ empowerment and access to human and physical capital, the fact that this reform also led to increased female mortality raises questions about long-term sustainability of reform effects. We use evidence from three states, one of which amended the HSA in 1994, to assess first- and second-generation effects of this reform using a triple-difference strategy. First-generation effects include greater likelihood of completing primary education, more assets brought into marriage, improved access to bank accounts, a lower share of female births, and higher female survival rates. Second-generation effects on education, time use, and health are robust and point estimates of education are larger than first-generation ones even after mothers’ endowments are controlled for, pointing to a sizeable and sustained empowerment effect.


Introduction
Parental bequests of material wealth or human capital are a key way to transfer assets across generations that impacts patterns of accumulation and overall development by affecting individuals' access to physical and human capital (Becker & Tomes, 1979;Stiglfitz and Weiss, 1981) and associated economic opportunities (Blinder, 1973). Inheritance laws affect not only the nature and size of inter-generational bequests (Ellul, Pagano, & Panunzi, 2010) but, in most contexts, also gendered patterns of wealth, bargaining power, and marital decision-making (Anderson & Bidner, 2015). If there are differences between males' and females' preferences, for example because females attach higher values to family needs or children's welfare and thus devote more of the resources at their disposal to these (Behrman, 1990;Strauss, Mwabu, & Beegle, 2000), and if, as often found, traditional norms provide men with preferential access to key household resources through inheritance, this is particularly relevant. In this case, the level of female wealth, including assets received by way of inheritance, will affect their bargaining power (Anderson & Eswaran, 2009) and future generations' human capital investment, social mobility (Davies, 1982;De Nardi, 2004;Kotlikoff & Summers, 1980), occupational choice, and asset accumulation (Cowell, 1998). The potentially farreaching effect of legal provisions regarding divorce (Stevenson & Wolfers, 2006) or access to political office (Chattopadhyay & Duflo, 2004) on female empowerment is well documented in the literature.
Given the key role of inheritance in inter-generational wealth transmission and the gender bias of the rules governing inheritance in many contexts, amending inheritance laws to overcome bias against females' ability to control resources is thus an attractive policy option. Changes in such laws and their impacts have been explored for Kenya (Harari, 2017), Indonesia (Carranza, 2012), and followed by a nation-wide effort to register land -Rwanda (Ali, Deininger, & Goldstein, 2014). Recognition of the potentially transformative role of improving women's inheritance rights was also a reason for some Indian states to, from 1987, amend the country's Hindu Succession Act (HSA) so that women's inheritance rights would be equal to those enjoyed by men.
Studies that analyse direct first generation impacts of the Hindu Succession Act amendment (HSAA)a reform adopted at a national scale in 2005from early adopting states generally find evidence of positive effects: the HSAA increased, without yet equalising, women's likelihood of inheriting land (Deininger, Goyal, & Nagarajan, 2013); enhanced women's educational attainment and the amounts of dowries received (Roy, 2015) and their participation in household decision making (Mookerjee, forthcoming); allowed women to supply more labour to high-paying jobs (Heath & Tan, 2016); and reduced intra-household violence (Amaral, 2015). Yet, as efforts to implement far-reaching legal changes will inevitably challenge long-established norms, impacts are not unequivocally positive. In India's patrilocal context (Khalil & Mookerjee, 2018), newly awarded rights were often 'gifted' to male siblings (Roy, 2015). The stress created by legislation to expand female inheritance rights seems to have manifested itself in higher levels of suicide and wife beating (Anderson & Genicot, 2015). By raising the relative cost of girls versus boys, the HSAA also reduced girls' survival ratios, most likely through sex-selective abortion (Bhalotra, Brule, & Roy, 2017;Rosenblum, 2015).
Such evidence of negative reform impacts and persistence of traditional norms raises concerns that positive first generation impacts of HSAA may not persist and be partly compensated or even outweighed by negative second-generation effects. Although the 2005 nation-wide extension of the HSAA makes this a first order policy issue, standard datasets such as the India Human Development Survey (IHDS) or the Demographic and Health Surveys (DHS) lack the information needed to analyse it. Thus, to the best of our knowledge, there are no studies exploring if HSAA effects persist beyond the first generation or the channels through which they may materialise. To address this issue, we use data from a unique survey conducted to specifically shed light on this issue by providing information on all individuals (not only those present in a given household) for three generations and including information (in this case on the time of the father's death) to identify a woman's eligibility to inherit according to the revised legal provisions. While resource limitations limited coverage to three states, Maharashtra (the reform state) where the HSA was amended in 1994 and Orissa as well as Uttar Pradesh (controls) where the HSAA became effective with the 2005 national expansion of this law, implying that we cannot claim national representativeness. With a population of 362 million as per the 2012 census, these make up more than a third of India's population and are home to more people that Germany, the UK, France, Italy, Spain, Netherlands, Belgium, and Greece combined and are thus of substantive relevance. 1 Using these data allows us to contribute to the literature by showing that the HSAA has positive short-and longer-term impacts on females: first-generation benefits include a greater likelihood of completing primary education, higher asset ownership due to more assets brought into marriage, and improved access to bank accounts. Our data support a reform-induced decline in the share of female births as found in other studies but show that this is accompanied by higher female survival rates, an outcome that is more consistent with higher levels of female bargaining power (Eswaran, 2002) than systematic neglect of female offspring post-reform. Second-generation effects on education, time use, and health are robust and point estimates of effects on education are larger than those for firstgeneration ones even if mothers' endowments are controlled for. While we find evidence of HSAAinduced resource transfers to femalesprimarily at the time of marriageeffects beyond those linked to direct resource transfers are sizeable as well, presumably reflecting an improved fall-back position by women empowered through the HSAA.
The paper is structured as follows. Section two reviews evidence of asset transfers across generations and legal provisions regarding inheritance in India including the HSAA. Section three discusses data and introduces issues of identification. Section four presents estimates of (first-generation) HSAA impacts on generation II's primary education completion, assets received at marriage, access to latrines and bank accounts as a proxy for empowerment, reproductive choices, as well as secondgeneration impacts on education and health outcomes, and discusses mechanisms that may underpin such effects. Section five concludes with implications for research and policy.

Background and motivation
Although inheritance of assets has long been identified as a key determinant of women's bargaining power, empirical studies to assess the direct or indirect impact of the inheritance reform are few and often weakly identified. Legal amendments to change female inheritance rights in India provide an exogenous source of variation that has been widely used to assess first-generation reform impacts. The mixed results from such studies provide a strong rationale for investigating persistence of reform effects in the second generation.

The importance of women's inheritance
A unitary household model may not adequately describe reality if preferences are heterogeneous and the distribution of resources within the household will affect parties' bargaining power (Anderson & Eswaran, 2009). Who in the household owns assets that generate key income streams or can claim access to such income will profoundly affect substantive outcomes including decisions on the use of household resources, fertility, and educational, health, and nutrition investments in future generations' welfare.
Women's bargaining power is of particular relevance for transfers of human and physical capital to the next generation if, as found in some studies, women devote higher proportions of income to family needs than men (Strauss et al., 2000). In this case, children benefit if their mothers control a larger share of family resources (Thomas, 1990). Support for this is provided by studies showing that (i) greater female bargaining power helped reduce fertility and child mortality in India (Dyson & Moore, 1983); (ii) receipt of pensions by females but not males affected girls' anthropometric status in South Africa (Duflo, 2003); (iii) higher female incomes after agricultural reforms increased girls' survival rates in China (Qian, 2008) and female empowerment affected contraceptive choice in Zambia (Ashraf, Field, & Lee, 2014); and (iv) exogenous increases in low-castes' female income significantly increased investment in schooling, particularly for girls (Luke & Munshi, 2011).
The literature points towards long-term effects from female education (Becker, Cinnirella, & Woessmann, 2013) and far-reaching economic impacts of regulations regarding inheritance on the level and nature of investment (Ellul et al., 2010). Yet, until recently, study of inheritance largely abstracted from recipients' gender and the conceptual discussion focused on other attributes but was often gender-blind. The wealth model (Becker & Tomes, 1979) predicts altruistic parents to provide children of different ability with amounts of human capital to equate marginal returns to schooling to those from financial assets. The strategic bequest model (Bernheim, Shleifer, & Summers, 1985) assumes parents bequeath assets to children in return for support in old-age; tests in a developed country setting lend general support to an equal allocation rule (Behrman & Rosenzweig, 2004).
In societies all over the world, land is a key asset (Deininger, 2003;Doss, Truong, Nabanoga, & Namaalwa, 2012) and inheritance is the main mechanism for acquiring it. Inheritance of land thus determines access to economic opportunities, the ability to bear risk, and the ability to affect decision-making within the household and the community. Yet, women's ability to inherit land or have their land rights documented is often restricted (Platteau and Baland, 2001), affecting social and Second-generation effects of the HSAA 2551 economic outcomes (La Ferrara, 2003) through a range of channels (Deere, Oduro, Swaminathan, & Doss, 2013), including the fact that women are disproportionately affected byinheritance relatedland conflicts (Deininger & Castagnini, 2006). While this has led scholars to recommend legal change to make women's inheritance rights more secure and in doing so reduce long-standing gender discrimination (Cooper & Bird, 2012), impacts of legal interventions on deeply engrained behavioural patterns were often limited (Anderson, 2003) and property rights reforms proved difficult to sustain over time (Galiani, 2011). Evidence of tangible and sustained positive impact and avoidance of unintended negative consequences is essential to make a case for such reforms. Yet, making this case empirically is difficult as standard surveys do not contain the information needed to properly identify such effects (Quisumbing, 2001).
Exogenous changes in legal provisions that affect women's ability to control key family assets can help fill this void and advance understanding of intra-household bargaining. For example, changes of US divorce laws that made exit easier for women are analytically equivalent to an asset transfer (Chiappori, Fortin, & Lacroix, 2002) and raise similar analytical issues. Studies often use states' adoption of such provision at different points in time to achieve identification (Allen, 1992;Friedberg, 1998;Peters, 1986;Wolfers, 2006). Results suggest far-reaching impacts on spouses' bargaining power in existing marriages and outcome variables such as domestic violence (Stevenson & Wolfers, 2006) and female labour force participation (Stevenson & Wolfers, 2007). We use a similar strategy for inheritance below and combine this with innovative data to allow us to trace second-generation effects.

Women's inheritance rights in India
Although India's Constitution mandates gender equality, inheritance was traditionally strongly biased against women. The 1956 HSA distinguished individual property from joint ancestral assets including land (Agarwal, 1994). Rights to the latter were limited to the coparcenary, a set of individuals that comprised only males, severely limiting females' ability to inherit joint property. 2 From 1986 states started amending this Act through a set of measuresreferred to as the HSAAthat stipulated that any coparceners' daughters will acquire coparcenary rights by birth, 3 making their status de facto equal to that of sons. We use the change generated by this law, which was adopted nationally in 2005, to explore first-generation impacts on women affected by this legislation as well as second-generation effects on their children.
Several studies explored first-generation effects of the HSAA for states that introduced the reform before it became national law. Comparing males and females in the same household for those that did and did not benefit from the reform suggests that the HSAA increased the likelihood of inheriting land for direct beneficiaries (Deininger et al., 2013) as well as their access to physical (dowries) or human capital (education) post-reform (Roy, 2015). 4 The reform reduced domestic violence (Amaral, 2015) and improved women's decision making, predominantly through a shift in family structure and at the cost of the older generation (Mookerjee, forthcoming). Exposure to the HSSA is also found to have raised women's autonomy and labour supply, especially into high paying jobs (Heath & Tan, 2016). At the same time, the inheritance reform reduced girls' survival ratios, possibly by leading to sex-selective abortion that resulted from an increased cost of raising girls as compared to boys for landowning households (Rosenblum, 2015). The reform increased excess female infant mortality and son-based fertility stopping (Bhalotra et al., 2017). Stronger female inheritance rights are also associated with a decrease in the difference between female and male suicide rates, consistent with the notion that this reform increased conflict within households, a notion that is consistent with an increase in the incidence of wife beating (Anderson & Genicot, 2015).
While higher maternal education has been shown to increase the use of prenatal care and infant health, (Currie & Moretti, 2003), use of standard datasets to assess second-generation impacts of the inheritance law reform has to contend with a number of issues relating to the lack of information on eligibility or receipt of inheritance or other assets and the difficulty of establishing causality in the presence of unobserved variables and mismeasurement. Bose and Das (2017) find unambiguously positive first-generation impacts but a significant negative HSAA effect on boys' and no effect on girls' education. Whether, as the authors suggest, this is due to more empowered mothers who realise the high opportunity cost (or, widespread teacher absenteeism lowering effectiveness) of school attendance using their increased bargaining power to pull sons out of school to allow them to use their time more productively cannot be ascertained without more specific information. Insofar as they lack information on a woman's eligibility to benefit from the HSAA (that is, if she was unmarried and her father was still alive when relevant legislation was passed) or intermediate variables, standard surveys may be ill-suited to provide answers and indirect evidence needs to be relied on. Tommasi, Lewbel, and Calvi (2017) develop an estimator for local average treatment effects that uses the inheritance law change as an instrument for broader female empowerment, finding first-generation benefits for affected women in terms of higher BMI and lower anaemia and second-generation benefits in terms of their children suffering less from cough, fever, or diarrhoea.

Data and descriptive statistics
To analyse effects of the inheritance reform, we use data from a 2011 follow up to the 2007 Rural Economic and Demographic Survey (REDS) conducted by India's National Council for Applied Economic Research (NCAER) in Maharashtra, Uttar Pradesh, and Orissa. The household survey collected information on three generations, the head and spouse at the time of the survey (generation II), their parents (generation I), and children (generation III). The sample comprises 1,204 households and 2,931 generation III descendants. Figure 1 plots the density distribution of birth years for the sample, pointing towards an average age of 79, 49, and 18 years for individuals in generations I, II, and III, when data were collected.
Beyond demographic and educational data routinely collected in multi-purpose household surveys, our survey includes information on time use during a typical day for all individuals in generations II and III. For those in generation II, we have information on the value of assets (including land) brought into marriage, access to individual bank accounts and latrines, spending on education for each co-resident child, and the survival status or year of death of each parent. This is complemented with household-level data on the amount of resources spent for each child Second-generation effects of the HSAA 2553 on education (books/stationery, transport/hostel, and private coaching/tuition) and to cure preventable diseases, a variable we use as a proxy for attention to preventive rather than curative health. Moreover, the survey also includes the number of children born and still alive by gender for all generation II women below the age of 60, providing information on fertility (the number of children and the share of daughters born) as well as sons' and daughters' survival rates. For generation III individuals, including not co-resident ones, information on the extent of primary school completion is also available. Descriptive statistics on characteristics of interest for individuals in generation II and III when the HSAA reform became effective (Maharashtra in 1994 versus 2005 in the control states) are in Table 1. Data for the same variables by state are in Appendix Table A1. We note that 62 per cent of generation II males' fathers and 76 per cent of generation II females' fathers passed away after 1994 or are still alive so that the HSAA would apply, providing sufficient variation to identify inheritance-induced effects.
Although descriptive analysis cannot substitute for proper econometric analysis with controls and fixed effects as needed, variation comparing between reform and non-reform states in these can illustrate the nature of the data and point towards interesting variables for analysis. On average, generation II males received more than six years of education and females in this generation lag their husbands by some three years. Amounts of gifts received at the time of marriage and controlled by individual marriage partners (that is, females' stridhan) 5 were more for males than females on average (0.61 versus 0.39). Females have higher levels of endowments with human capital (3.83 versus 3.37) and physical assets (0.43 versus 0.38) in Maharashtra than non-reform states, a pattern reversed for males, suggesting that prima facie the HSAA may have been effective. If increases in endowment or HSAA effects through channels including actual or potential inheritance translated into bargaining power within the household for females, our variables measuring intra-household bargaining power should capture some systematical differences between Maharashtra and non-reform states. Again, there is some descriptive evidence to support this: households in the reform state are 36 per cent more likely to have a latrine, a good that is strongly preferred by females who disproportionately benefit from it (Stopnitzky, 2017).
As inheritance legislation has been shown to affect reproductive choices, we use information on the number and the sex ratio of children by generation II females based on survey information regarding the pregnancy history for ever married women aged 15-60. At a descriptive level, doing so points towards significantly lower levels of fertility (2.78 versus 3.46) but insignificant differences in survival rates or sex ratios in the reform state, providing another variable that is a candidate for more detailed checking in a regression framework.

Analytical approach
We identify second-generation impacts by comparing male and female siblings in a household where the mother was affected by the HSAA with siblings in a household where this was not the case, adding relevant controls and fixed effects (including household fixed effects and gender and state specific year of birth effects). While lack of information on generation II siblings precludes use of household fixed effect, resulting in estimates that are less robust than those for second-generation effects, first-generation effects are also reported to allow comparison to the broader literature.

Identifying first-generation effects
To explore if first-generation impacts could underpin second-generation effects, we estimate how the HSAA affected outcomes in terms of educational completion, share of assets brought into marriage, access to an independent bank account and a sanitary latrine, as well as reproductive choices, by females in generation II.
For school completion, we use a difference-in-difference (DID) estimation where the first difference is between 'young' and 'old' age groups, defined as those who should have completed relevant education decisions before or after the HSAA came into effect in 1994, and the second difference between states that did or did not change the HSA. With subscripts, i, j, and k denoting individuals, households, and villages, the DID estimation equation for school completion is: Second-generation effects of the HSAA 2555 where Y ijk is an indicator variable of whether or not the individual completed primary or secondary education, α k is a village-level fixed effect controlling for time-invariant characteristics at this level, 7 G ijk is a dummy that equals one if the individual is in the 'young' cohort that should have been affected by the HSAA (those below 30 years old in 2011 for primary and those below 34 years old in 2011 for secondary implying that they were below 14 or 18 years of age in 1994, respectively), 8 and M k is a reform state dummy. We estimate Equation (1) for male and female samples separately and note that in the female sample β 2 denotes the estimated first-generation impact of the HSAA on educational completion. We run a placebo test by comparing generation II individuals who already completed primary or secondary school when the HSAA came into force. To do so, we include in G ijk an indicator variable for those 31 to 40 years old in 2011 for primary and those aged 35 to 40 in 2011 for secondary school completion. 9 For assets transferred at time of marriage, a triple difference strategy is used with differences between (i) reform and non-reform states; (ii) 'young' and 'old' cohorts whose date of marriage (after or before 1994) implies decisions on asset transfers were or were not informed by the HSAA; and (iii) whether the father died before 1994 so that an inheritance had already been triggered. 10 The estimation equation is where, Y ijk denotes the logarithm of gifts received by individual i at the time of marriage, D ijk is a dummy of whether the individuals' father died before 1994 so s/he would not have benefited from the reform even in the reform state, G ijk is an indicator variable for the 'young' cohort (generation II individuals below age 48 in 2011 who were less than 30 years in 1994 so that their marriage decisions were affected by the reform), 11 M k is a reform state dummy, and λ ijk is a vector of state specific year of birth fixed effects to control for time-varying shocks or aggregate effects by state. Equation (2) is estimated separately for female and male samples; for the female case, β 6 is the estimate of the firstgeneration impact of the HSAA on assets brought into marriage. To show validity of our estimation, we report the comparison between fathers who died one to four years prior to the inheritance reform (between 1990 and 1993) and those who died earlier (before 1990). To address concerns about the reform state having historically been more gender progressive, we apply a DID estimation strategy exclusively to the Maharashtra sample only. A similar strategy can be used for empowerment and reproductive choices. The only difference to Equation (2) is that we are unable to compare between cohorts. We thus apply a DID estimation with differences between (i) reform and non-reform states; and (ii) whether or not the father died before 1994 so that an inheritance case had already been triggered. As the basic specification is a DID, we do not report results for the Maharashtra sample. In addition to the placebo test (fathers died between 1990 and 1993 versus fathers died before 1990), we also explore mechanisms of empowerment as well as the wealth effect in terms of increasing household assets. Specifically, we control for assets brought into marriage and the difference in education between the spouse and her husband, two measures of empowerment that have traditionally been used in the literature, to explore whether the HSAA affected mothers through these or other channels. As wealthier households may be more gender progressive in general, we also control for household assets to capture potential wealth effects.

Identifying second-generation effects
To identify second-generation effects of the inheritance reform, we explore time allocation, educational spending, completion of primary education, and resources spent on curing preventable diseases, possibly due to insufficient preventive care, for generation III individuals. We use a triple difference estimation strategy where the three relevant differences are (i) between generation III males and females within the same household; (ii) between generation III individuals whose mothers' fathers died before 1994 and those who died after 1994 or are still alive; and (iii) between generation III individuals from the reform state (Maharashtra) and non-reform states of Orissa and Uttar Pradesh. The estimating equation is: where Y ij represents the outcome variable of interest (of individual i from household j) as discussed above, 12 α j is a household-level fixed effect that controls for commonalities within a generation II household. 13 F ij and MD j are indicator variables, respectively, for generation III females (compared to their male siblings), and whether their mothers' fathers died before or after 1994 which, in the reform state (but not elsewhere) would imply that mothers either did not or did benefit from reforms, and φ ij includes vectors of gender and state specific year of birth fixed effects and a vector of gender and state specific order of birth fixed effects to control for time-variant and order-variant aggregate effects or shocks by gender and state, respectively. δ 3 and δ 4 , the coefficients on their interactions with a reform state dummy M j are parameters of interest that capture the direct impacts of the reform on girls (first-generation impacts) and the indirect impacts of HSAA-induced empowerment of generation II females from inheritance (second-generation impacts), respectively. 14 To explore robustness of our estimation and identify mechanisms that might underpin the impacts of the HSAA, we also report results from alternative specifications in line with those for generation II discussed above as follows: first, we run a placebo test by comparing, for subsamples with sufficiently large sample sizes, generation III individuals whose maternal grandfathers died prior to the reform. In this case, MD j also includes an indicator variable for whether mothers' fathers died one to four years prior to the inheritance reform (between 1990 and 1993). Second, a DID estimation for the Maharashtra sample to address concerns about the reform state having historically been more gender progressive. Third, we control for assets brought into marriage and differences in education between the spouse and her husband to explore whether HSAA affected mothers through these or other channels. Finally, we control for household assets and the number of generation III individuals to capture potential direct and indirect wealth effects.
As the reform was implemented by state, we allow error terms ( ijk and ij ) in all the specifications discussed above to be correlated within state and cluster by state. Moreover, to address the small number of clusters (three states), we report results of wild-clustering bootstrap following the method proposed by Cameron, Gelbach, and Miller (2008) and adopted in empirical studies (Nandi & Deolalikar, 2013).

Econometric results
Implementing the above regressions provides evidence in support of positive effects of inheritance reform on both direct beneficiaries and the next generation. Female empowerment via an improved fall-back position is the most likely reason for such effects. The finding of reform-induced reductions in female birth rates as well as increased survival rates for girls supports that inheritance is closely linked to reproductive choice while at the same time suggesting that more in-depth study will be of interest.

First-generation effects of the inheritance reform
Results for primary and secondary school completion by generation II females (cols. 1 and 2) and males (cols. 3 and 4) in Table 2 point towards significant and quantitatively large effects on primary completion. The estimated HSAA-induced increase in the likelihood of primary completion by females is 14 per cent (col. 1), compared to an insignificant impact for males (cols. 3 and 4). With Second-generation effects of the HSAA 2557 a negative highly significant point estimate (col. 2), the placebo test, based on interacting a reform dummy with the age group just above the cut-off, suggests little concern about contamination although the magnitude of the estimate reform effect declines somewhat. By comparison, estimated reform impacts on females' secondary education are positive but insignificant.
For assets brought into marriage, Table 3 presents results for females and males in cols. 1-2 and 3-4, respectively. Estimates point towards a HSAA-induced increase in the level of assets brought into marriage by females of about 65 per cent (col. 1), compared to an insignificant and negative change in the amount of assets contributed by husbands (col. 3). The placebo test suggests that females whose fathers died one to five years before the the HSAA coming into force did not gain access to such a high level of assets (col. 2), supporting the notion that the effect is reform-induced. Estimates based on the Maharashtra sample are insignificant and quantitatively smaller, but still with a magnitude of 30 to 40 per cent. To the extent that the share of assets brought into marriage will affect individuals' future bargaining power, this would be consistent with the notion that the HSAA empowered direct female beneficiaries, thus increasing the plausibility of downstream second-generation effects.
Results for the likelihood of females having a bank account or the household having a sanitary latrine in Table 4 suggest that the HSAA increased the likelihood of females having an individual bank account by 5 per cent (col. 1), an estimate that is robust to inclusion of other covariates (col. 2). Although households where the female spouses had benefited from the HSAA were not significantly more likely to have sanitary latrines (cols. 4 and 5), a good that is strongly preferred by women, positive education gaps between mother and father increased the tendency by 2 per cent. This is consistent with an intra-household externality whereby lack of authority may preclude women's  (Cameron et al., 2008). *Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent.
preferences from translating into action similar to what was found elsewhere for cook-stoves (Miller & Mobarak, 2013), suggesting that external forces that empower women may help translate preferences into action. It has been suggested that, even if it improves outcomes for girls alive at the time of survey, the HSAA may affect reproductive choices and female mortality. To test for such effects, regressions on the number of children and the share born to the 717 female spouses aged 15-60 in our sample as well as their sons' and daughters' survival rates are reported in Table 5. Results point towards a reduction of the share of daughters born (cols. 4-6 of panel A), consistent with sex-selective abortion. At the same time, we find a significant 10 per cent increase in girls' survival rate (col. 1 of panel B) compared to an unchanged  (Cameron et al., 2008). In panel B, figures in brackets are robust standard errors. *Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent.
Second-generation effects of the HSAA 2559 survival rates by boys (col. 4 of panel B), a result robust to including other covariates. This is consistent with a model where parents invest in children mainly for old age security. As in the South Asian context, such security is provided by boys rather than girls, greater female autonomy may be associated with higher survival rates for girls but also sex-selective abortion (Eswaran, 2002).

Second-generation effects of inheritance reform
The samples to explore how the HSAA might have affected generation III girls' outcomes either directly or by empowering their (generation II) mothers differ slightly from each other due to different age cut-offs. To test for HSAA-induced impacts on time allocation and education expenses, we use data from 791 generation III individuals from six to 14 years old in 449 generation II households (sample I). To examine school completion, information from 1,919 generation III individuals 15 to 30 years old in 768 generation II households is used (sample II). Finally, to estimate the incidence of HSAA-induced health effects, measured as the need of treating diseases that could have been prevented with proper care, we use 927 generation III individuals zero to 14 years old in 471 generation II households (sample III). Results for time allocation in generation III are reported in Table 6. We note that, compared to their male siblings, 6-14 years old girls in the reform state whose maternal grandfather is still alive at the time of survey or passed away after 1994 so that their mothers would be affected by the HSAA spent some 6 per cent more time on study (col. 3) than those in households where mothers had been unaffected by the HSAA. With about 1.5 hours per day of additional time spent studying (including school attendance), this suggests a second-generation effect of non-negligible  (Cameron et al., 2008). *Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent.  (Cameron et al., 2008). *Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent.
magnitude. 15 While the reform indeed affected time allocation in generation III through other channels (col. 4), we note that in this case the point estimate of HSAA-effects increases to 14 per cent. Results also suggest that increased time spent studying is entirely compensated for by a reduction of leisure (cols. 5 and 6). This suggests that the HSAA empowered mothers beyond Notes: Dependent variable is the share of time spent on productive work, education and study, and leisure, respectively. All regressions include gender and state specific year of birth fixed effects, gender and sate specific order of birth fixed effects, and household fixed effects. In panel A, figures in brackets are standard errors clustered by state using wild-clustering bootstrap (Cameron et al., 2008). In panel B, figures in brackets are robust standard errors. *Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent. resources transferred to them, thus helping to increase the amount of education consumed by girls, in line with recent emphasis on the far-reaching impacts of female empowerment (Diebolt & Perrin, 2013). With the possible exception of girls benefitting from the reform spending more on leisure, estimated direct effects on girls whose mothers were not affected by the HSAA are weak throughout. Robustness checks using the Maharashtra sample suggest a second-generation effect with point estimates for time on study of 11 per cent (col. 4) and 3 per cent (col. 3) with and without controlling for other channels, respectively. Evidence for total educational spending as well as spending on books and stationary, the main expense categories, in Table 7 point towards strongly positive second-generation effects from reform indirectly through mother's inheritance. For overall spending, having a mother who benefited from the HSAA is estimated to triple spending on daughters' education through a female empowerment effect (col. 1). This is mainly contributed by the increase in the expense on books and stationary (col. 3). The significant and negative direct effects on girls whose mothers were not affected by the HSAA (with magnitude similar or even bigger than the indirect effects through mother's empowerment) are consistent with the previous result concerning time spent on study. Although the smaller sample size increases the variance leading to insignificant estimates, the robustness check based on the Maharashtra sample also suggests a doubling of spending on daughters' education, mainly contributed by the expense of books and stationary.
Results for completion of primary education in Table 8 point towards considerable bias (of 38%) against females compared to males in the same household for primary completion. Estimated secondgeneration effects from mother's inheritance (11%) suggest the HSAA helped to reduce such against girls' primary education (col. 1) and mother's education is an important channel (col.2). The variation in terms of death of maternal grandfathers in the five years prior to the HSAA passage allows us to run placebo tests that by and large confirm our results (col. 3). Estimates from the Maharashtra sample also point towards considerable bias (of 34-35%) against females compared to males in the same household for primary completion, consistent with the results from the three state regression. The estimated magnitude of the second-generation indirect effect is also comparable to what was found in the three-state sample.
While our data lack detailed information on health status, insufficient preventive care is likely to lead to greater incidence of otherwise preventable diseases that would in turn require higher levels of curative spending. Contrary to the finding of the inheritance reform having increased secondgeneration female mortality (Rosenblum, 2015), and although sample sizes are small, results for the sample of individuals below age six in Table 9 point towards second-generation effects whereby mothers' exposure to the HSAA reduced the likelihood of their daughters requiring treatment for preventable diseases (col. 3), a result that is robust to inclusion of other covariates (col. 4).

Conclusion and policy implications
Although studies point towards positive impacts of India's inheritance reforms on female empowerment, education, and asset access, evidence of the reform having led to higher infant mortality suggests that exploring longer-term effects of the HSAA is of great policy relevance.
Drawing on a specially implemented survey to be able to identify empowerment effects, we find that females benefiting from the inheritance reform were 8 to 13 per cent more likely to have completed primary education received 65 to 77 per cent more assets at marriage, and were 4 to 7 per cent more likely to have a bank account. Second-generation effects on educational inputs and results as well as health status by beneficiaries' children, identified by comparing outcomes for females relative to their male siblings between households in which the mother was or was not affected by the reform, point to significant reform effects. Thanks to the reform, spending on education for beneficiary mothers' daughters was 2.2. to 2.6 times higher than for non-beneficiaries; they spent an additional 1.5 hours/day studying, and were 12 per cent more likely to complete primary school, slightly above what was found for the first-generation. Most of these effects seem to result from the HSAA having improved women's fall-back position beyond the direct transfer of resources.
A reduction in the share of daughters born to females who benefited from the reform together with an increase in these daughters' survival rate illustrates far-reaching impacts of the inheritance reform  (Cameron et al., 2008). In panel B, figures in brackets are robust standard errors. *Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent.  (Cameron et al., 2008). In panel B, figures in brackets are robust standard errors. *Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent.
Second-generation effects of the HSAA 2565 on reproductive outcomes. In a setting where old-age security is provided by boys, this could be consistent with a model of reform having increased female bargaining power rather than a universal neglect of female offspring post-reform. Exploring whether the HSAA's fertility impacts vary with the extent to which women have access to independent old age insurance would be a straightforward way to further clarify this issue and, in doing so, appreciate potential and limitations of the  (Cameron et al., 2008). In panel B, figures in brackets are robust standard errors. *Significant at 10 per cent; **significant at 5 per cent; ***significant at 1 per cent.
inheritance reform as a tool to promote gender equity in different socio-economic contexts well beyond India.
1. As budget limitations did not allow us to cover more than three states, they were chosen to maximise regional coverage and variability in terms of gender equality using the 'gendering human development' index by the Ministry of Women and Child Development, Government of India. As per this index, Maharashtra is the best performing reform state (excluding Kerala, which has a different system of inheritance) and Orissa and Uttar Pradesh are the worst performers in the Eastern and Central Zones, respectively. This allows us to compare the best with the worst states. 2. Our focus on intestate inheritance is justified by the fact that many rural Indians lack the knowledge or resources to have a will formally registered. 3. The Act was amended in 1986 in Andhra Pradesh, in 1989 in Tamil Nadu, and in 1994 in Maharashtra and Karnataka.
Kerala had abolished the joint family property system in 1976 (Agarwal, 1994). 4. Education results may partly be due to potential state-and gender-specific trends that are independent of inheritance reforms and results on inheritance and dowry may be driven by the fact that at the time when data were collected girls in the sample exposed to the reform were still relatively young (15 years in KA and MA; 20 years in TN, 23 years in AP, and 33 years in KE) when data were collected in 1999 and that women who had inherited land may no longer live with their parents. 5. As Agarwal (1994) explains, gifts received at the marriage ceremony are one important category of strdihan and Hindu females are absolute owners of such property. 6. M k is omitted in this village fixed effect model because the state dummy variable is constant within a village. 7. Note that we cannot include household fixed effects in this regression as males and females are from different households and match endogenously in the marriage market. See the discussion on this below. 8. Note that all the individuals in generation II in our sample are older than 20 years old and already completed the secondary school. We restrict the sample to include generation II individuals 45 years old and younger to exclude those who are too old to be comparable with the 'young' group. 9. Results for the Maharashtra sample only are not reported as this would leave us only with a simple cohort comparison. 10. The young cohort for school completion is younger than those for assets brought into marriage. People completed primary and secondary schools at about 14 and 18, respectively, and education decisions are relevant before those ages. However, people normally got married at 20 when marriage decisions including the gift transfer are made. Not surprisingly, there is no difference in the timing of father's death for the young cohort (all of them had a father alive in 1994) so to maintain comparability and consider only individuals who would have unambiguously benefited from the HSAA, we also exclude individuals whose fathers died before 1994 from the old cohort. 11. We do not know the year of marriage for generation II individuals, so we use age for proxy supposing people got married before 30. Note that we lose variations in terms of the timing of father's death when we focus on younger samples. 12. For school completion, we focus on generation III individuals 15-30 years old who were less than 14 years old in 1994 for primary education. 13. M j is not included by itself in the household fixed effect model because it does not vary within a household. MD j and MD j *M j are not included in the model either for the same reason. 14. Note that fathers of generation III individuals, that is, generation II males were all alive at the time of the survey, so we expect generation III females to directly benefit from the reform as generation II females whose fathers died after 1994 or were still alive at the time of the survey. 15. With a sample mean of 28 per cent for the variable (time spent studying), this represents an economically very meaningful increase of more than 20 per cent.
Second-generation effects of the HSAA 2567 Second-generation effects of the HSAA 2571