The seen and unseen: the unintended impact of a conditional cash transfer program on prenatal sex selection

Javadekar, Sayli; Saxena, Kritika Article — Published Version The seen and unseen: the unintended impact of a conditional cash transfer program on prenatal sex selection Journal of Population Economics Provided in Cooperation with: Springer Nature Suggested Citation: Javadekar, Sayli; Saxena, Kritika (2025) : The seen and unseen: the unintended impact of a conditional cash transfer program on prenatal sex selection, Journal of Population Economics, ISSN 1432-1475, Springer, Berlin, Heidelberg, Vol. 38, Iss. 1, https://doi.org/10.1007/s00148-025-01091-6 This Version is available at: https://hdl.handle.net/10419/323177 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Vol.:(0123456789) Journal of Population Economics (2025) 38:32 https://doi.org/10.1007/s00148-025-01091-6 ORIGINAL PAPER The seen andunseen: theunintended impact ofaconditional cash transfer program onprenatal sex selection SayliJavadekar1· KritikaSaxena2 Received: 12 June 2023 / Accepted: 4 February 2025 / Published online: 22 February 2025 © The Author(s) 2025 Abstract This study examines the unintended consequences of the Janani Suraksha Yojana, a conditional cash transfer program in India, on prenatal sex-selective behaviour within a son-preference culture. This program unintentionally altered existing trends in prenatal sex selection through its simultaneous provision of cash incentives to households and community health workers as well as access to prenatal sex detection technology such as ultrasound scans. Using difference-in-differences and triple difference estimators we find that the program causes an increase in the likelihood of female births. Furthermore, we observe a rise in under-5 mortality for girls born at higher birth orders, suggesting a shift in discrimination against girls from prenatal to postnatal. Our calculations suggest that the net impact was approximately 300,000 girls surviving in treated states between 2006 and 2015. Finally, we find suggestive evidence that the involvement of community health workers in facilitating the program is a key driver of this trend. Overall, this study sheds light on the complex interplay between policy interventions, cultural norms, and gender disparities in shaping demographic outcomes. Keywords Prenatal sex selection· Missing girls· Sex-selective abortions· Community health workers· Janani Suraksha Yojana JEL Classification J13· J16· J18 Responsible editor: Kompal Sinha * Kritika Saxena [email protected] Sayli Javadekar javadekar.say[email protected] 1 Thoughtworks Gmbh, Munich, Germany 2 Department ofEconomics, Econometrics andFinance, University ofGroningen, Groningen, Netherlands S.Javadekar, K.Saxena 32 Page 2 of 34 1 Introduction The long history of son preference in India has resulted in nearly 63 million women missing from the country’s population, with almost 2 million missing across different age groups every year.1 This phenomenon of ‘missing women’ has the potential for socioeconomic disruption, such as a marriage market squeeze (Hesketh and Xing 2006), an increase in crime rates (Edlund et al. 2013), social stratification based on gender (Edlund 1999), and fewer health and educational investments in women (Jayachandran and Kuziemko 2011). The Indian government has introduced various schemes to reduce discrimination against women, including providing parents with financial incentives to have daughters. However, the effects of these policies are ambiguous (Anukriti 2018; Sekher 2012; Sinha and Yoong 2009). At the same time, the literature shows that access to ultrasound technology increases the likelihood of sex selection (Almond etal. 2013; Anukriti etal. 2022a, b). This paper demonstrates how accessible ultrasound technology, along with financial incentives provided under a nationwide safe motherhood program, interacts with the culture of son preference to influence the gender imbalance in India. We examine the causal relationship between a safe motherhood program and sex-selective behaviour among Indian parents and investigate the underlying mechanism that explains this relationship. The safe motherhood program, known as Janani Suraksha Yojna (JSY), was launched by the Indian Government in 2005 to reduce maternal and neonatal mortality. Mothers were given cash payments for every live birth in a health facility. The program also mandated that beneficiaries undergo at least three antenatal checkups, including ultrasound scans (a prenatal sex-determination technology). To facilitate the program, the government recruited health workers. They received financial incentives for every institutional delivery by registered mothers in their neighbourhood. The scheme, thus, provided simultaneous access to prenatal sex detection technology and cash incentives for institutional births. Therefore, the program had the potential to influence sex-selective behaviour among Indian parents unintentionally. To estimate the impact of the program on prenatal sex selection we use difference-in-differences (DID) and triple difference (DDD) estimators that exploit the variation in the timing of program implementation, eligibility of beneficiary households based on their socioeconomic status and geographic location, and the natural experiment created by sex of the firstborn child. Before the implementation of the program, states in India were categorized as lowor high-performing based on their state-specific institutional delivery rates. The eligibility criteria for program benefits varied by household socioeconomic characteristics across this classification. All women residing in low-performing states were eligible for the program; however, only those living below the poverty line (BPL) and belonging to the Schedule Castes or Schedule Tribes (SC/ST) in high-performing states could participate. Taking advantage of this variation in program access, we compare women living above the poverty line and not belonging to SC/ST from 1 Estimates in Economic Survey of India http:// mofapp. nic. in: 8080/ econo micsu rvey/ pdf/ 102118_ Chapt er_ 07_ ENGLI SH_ Vol_ 01_ 201718. pdf The seen andunseen: theunintended impact ofaconditional… Page 3 of 34 32 low-performing states with their counterparts from the high-performing states who were excluded from the program. In other words, non-BPL, non-SC/ST women from lowand high-performing states composed the treatment and control group, respectively.2 For our analysis, we created a mother-child panel using the reported fertility history of mothers from the Demographic and Health Survey of India (DHS) - 2015/16. The identification of our estimates is conditional on the inclusion of mother fixed effects that account for the systematic differences in the characteristics of mothers in our two comparison groups. We include child year of birth fixed effects that account for the unobserved time-varying factors that may influence the propensity of births to a mother over time. We also include child birth order fixed effects to account for the unobserved heterogeneity in the propensity to sex-select dissimilarly for different birth orders. To ensure the robustness of our estimator to any time-variant state-specific shocks and policies that could conflate with the program’s impact, we include state-year fixed effects in our DDD estimation. We measure sex-selective behaviour as the likelihood of female birth at every birth order for a mother. Our DID estimator shows that the program increased the likelihood of female births by 4.8 percentage points. The DDD estimates support this finding. We show that the families in the treatment group with a firstborn daughter see an increase in the likelihood of female births by 12.7 percentage points for birth orders 2 and above. This is a novel result considering the existing evidence on the greater prevalence of prenatal sex selection among the forward caste, non-poor families, and families with firstborn daughters (Borker etal. 2017; Anukriti 2018; Almond etal. 2019; Rosenblum 2013a, b). To establish the robustness of our results, we verify the identifying assumption. Our identifying assumption is that in the absence of the policy, the likelihood of female births evolves in a parallel manner in the treatment and control groups, conditional on mother fixed effects. To further validate our empirical strategy, we perform our analysis on the data collected before the launch of the program and find no program effects on mothers who never received the program benefits. These falsification tests validate our identification strategy and buttress our findings on the causal effect of the policy. Furthermore, we explore what these results mean for the survival and well-being of these additional girls. We find suggestive evidence that although more girls were born in treatment households, they were more likely to die before reaching 5 years of age. Surviving girls are likely to have poorer health and nutritional outcomes, increasing the gender gap in well-being among children. Although these results are not causal effects of the program, they provide additional insights into fertility dynamics in India, particularly among mothers in the treatment states. Our broad calculations indicate that the net effect of the program on female births is an overall increase of 300,000 girls born in treatment states between 2006 and 2015. 2 Caste groups in India are given a hierarchical classification: upper/forward castes, other backward castes, schedule castes, and schedule tribes. Non-SC/ST group includes upper/forward castes and other backward castes. Upper and forward caste are used interchangeably here. S.Javadekar, K.Saxena 32 Page 4 of 34 How did JSY influence the sex-selective preferences of Indian households? We hypothesize that the program worked through four possible channels. First, by mandating at least three antenatal checkups, JSY increased access to ultrasound technology among households who might have had limited or no access. Parents with strong son preferences may use these for sex selection by inducing abortions of unwanted female foetuses. Second, the financial incentives given to households for every live birth lowered the cost of bearing children. This is a motivator to not carry out sex selection and to give birth to their child, particularly during periods of economic shock. Third, health workers’ remuneration was linked to the number of beneficiaries registered for the policy and their deliveries at health centres. This is an incentive for health workers to dissuade parents from performing sex-selective abortions and to encourage them to give birth to their female children. Finally, the health workers maintained a JSY card to track every pregnancy in their neighbourhood. Fetal sex determination and sex-selective abortions are illegal in India. Hence, the registration and monitoring done by the health workers could deter the households from sex selecting.3 JSY thus could influence the willingness of parents to bear daughters by creating an unintentional trade-off along these different dimensions of the program.4 We use the Health Management Information System data obtained from the Ministry of Women and Child Development, Govt. of India. Using this approach, we created a dataset of all health workers at the district level from 2008 to 2015. We find suggestive evidence that neither access to ultrasound technology nor the financial incentives given to parents explain the increased propensity for having girls in the treatment states. We find descriptive evidence that the increase is explained by the presence of health workers. This result has important policy implications. This shows that intermediary health workers can play a vital role not just in delivering health services but also in fostering desirable outcomes. Another key result is the shift of the discriminatory behaviour directed at girls from prenatal to postnatal as a response to this policy. This is a reversal of the prevailing trend where access to ultrasound technology shifted discrimination against girls from postnatal to prenatal (Bhalotra and Cochrane 2010; Bhaskar 2007). Although this result is not encouraging, it shows that there is scope for policy to achieve desirable fertility outcomes even in the presence of conflicting cultural beliefs. This paper contributes to the extensive literature on missing women. Several existing papers examine the effect of ultrasound technology on sex ratios at birth and the relative well-being of female children (Chen etal. (2013); Anukriti etal. (2022a, b); Lin etal. (2014); Hu and Schlosser (2015); Almond etal. (2019); Bharadwaj and Lakdawala (2013); Valente (2014) Congdon Fors and Lindskog (2023)). All of these studies show that the increased availability of fetal gender identification technology induces parents to abort unwanted female foetuses. The surviving girls, therefore, 3 Pre-Conception and Pre-Natal Diagnostic Techniques (PCPNDT) Act, 1994 is an Act of the Parliament of India enacted to stop female foeticides and arrest the declining sex ratio in India. The act banned prenatal sex determination. 4 We test these various mechanisms, however given the availability of data and restrictions on empirical estimations, we can only provide supportive evidence supporting or rejecting these mechanisms. The seen andunseen: theunintended impact ofaconditional… Page 5 of 34 32 are wanted and acquire health investments. The structure of the program creates a tradeoff between access to technology for prenatal sex selection and health workers’ performance-based benefits and beneficiaries’ own cash transfers for live deliveries at health centres. We fill the gap in the literature by documenting the impact of the program on sex-selective behaviour for the population amongst whom the practice is most prevalent, the upper-caste and wealthy families in India (Bhalotra and Cochrane 2010). We find that the program causes a decline in sex-selective abortions in India. The second contribution of this paper is to the growing literature on the unintended consequences of public policies and programs (Ebenstein 2010; Buchmann etal. 2019). This literature evaluates how policies can create perverse incentives and have an unintentional impact on other socioeconomic outcomes. JSY was implemented to reduce maternal and neonatal deaths during deliveries. The scheme did not target improving gender equality at birth. Existing literature assessing the impact of JSY has studied its impact on the uptake of maternity services and maternal mortality (Powell-Jackson etal. 2015), fertility (Nandi and Laxminarayan 2016), maternal care (Sen etal. 2020), and the academic performance of children (Chatterjee and Poddar 2021). This paper is the first to study the impact of JSY on sex selection, an outcome it did not target, and understand the underlying mechanism. The final contribution of this paper is to the growing literature on the importance of community health workers, to achieve desirable maternal and child well-being objectives. Several studies have documented the impact of financial incentives given to community health workers on a reduction in child mortality and morbidity (Cohen etal. 2013; Björkman Nyqvist etal. 2019; Celhay etal. 2019; Brenner etal. 2011; Singh and Masters 2017). We add to this growing literature by showing suggestive evidence of the contribution of health workers in the reduction of prenatal sex selection in the treatment states. In terms of methodology, our paper is closest to Anukriti etal. (2022a, b), but our paper differs in three ways. First, we study how simultaneous access to prenatal sex detection technology and financial incentives to households and health workers affect prenatal sex-selective behaviour. The trade-off between these dimensions of the policy is an unintended consequence of the intervention designed to tackle low rates of institutional deliveries. This is the main analysis of our paper. Second, our analysis focuses on the prenatal sex-selective behaviour of the non-SC/ST and non-poor groups as opposed to their work which studies all the socioeconomic groups. Although our findings are for a specific socioeconomic group, existing evidence shows that prenatal sex-selective behaviour is more prominent for this group. Finally, we attempt to explain how policy mechanisms affect prenatal sex-selective behaviour. The mechanisms that explain these respective results are distinct. We find that community health workers played a prominent role in increasing the likelihood of female births. Their paper finds the decline in desired fertility and lower birth spacing as the driving factors of the decrease in the number of female births. This paper is organized as follows: The ‘Background and Data’ section provides background on son preference in India and discusses the data and descriptive statistics. The “Empirical Strategy” section introduces the empirical strategy used in the paper. The ‘4’ section is a discussion of the results. The ‘Robustness S.Javadekar, K.Saxena 32 Page 6 of 34 Tests’ section presents the robustness tests. The ‘Discussion and Additional Evidence’ section is a discussion on mortality and additional evidence. The ‘7’ section discusses and tests various mechanisms that explain the results and ‘8’ section concludes the paper with some policy recommendations. 2 Background anddata Discrimination against young girls in India is well documented, with formal records available as far back as the First Census of British India in 1871-72 (Waterfield 1875). Today this discrimination is reflected in skewed sex ratios at birth and child sex ratios. The natural sex ratio at birth for humans is estimated to be between 104 and 106 boys per 100 girls (Bhaskar 2007; Anderson and Ray 2010); however in India, the sex ratio at birth has increased from 108 boys per 100 girls in 1991 to 111 boys per 100 girls in 2011.5 This increasing shortfall in girls at birth is primarily due to the culture of son preference. This shortfall has also been documented in other Asian societies that are known to share India’s preference for boys over girls (Clark 2000; Almond etal. 2019). India has some religious and cultural norms that view sons as assets and daughters as liabilities. For instance, in Hinduism, the dominant religion in India, sons are expected to perform funeral rites when their parents die. In the absence of social security, older parents typically live with their sons, while their daughters live with their husbands’ families. Although daughters have a legal right to an equal inheritance of the family wealth, due to sticky social norms around marriage, households prefer to keep wealth in the family by bearing a son instead of bequeathing assets to a daughter who will eventually move to another household (Bhalotra etal. 2020; Roy 2015).6 Paying large dowries for daughters (Borker etal. 2017) and safety concerns also make it more costly for parents to have a daughter (Alfano 2017; Borker 2021; Anukriti etal. 2022a, b). Furthermore, there is some evidence that sons benefit from economic advantages in the labour market that daughters do not receive (Rosenblum 2013a, b). These norms shape households’ fertility preferences and are in turn reflected in the discriminatory behaviour of households towards daughters before and after their birth. Parents adjust the gender composition of their family via prenatal discrimination and postnatal discrimination. Before ultrasound technology was available in India, parents followed a fertility rule called the stopping rule, of having children until they reached their desired number of boys. As a result, girls were born into larger families with limited resources and therefore received lower investments (Jensen 2012; Arnold etal. 1998; Das Gupta and Mari Bhat 1997). This postnatal discrimination resulted in worse health outcomes and excess mortality among 5 The sex ratio at birth among many species including humans is biased towards males. 6 In 2005, Hindu Inheritance Act was amended to allow women to inherit wealth from their parents. Our results stay robust to this change. See Appendix C for details. The seen andunseen: theunintended impact ofaconditional… Page 7 of 34 32 young girls. With the advent of prenatal sex determination technology, parents can determine the sex of the foetus within 7 weeks of pregnancy.7 This allowed parents to abort unwanted female foetuses (Chen etal. 2013; Bhalotra and Cochrane 2010). Easy access to ultrasounds since the mid-1980s and an increasing preference for smaller families have led households to change their behaviour from postnatal discrimination to prenatal discrimination (Goodkind 1996; Kashyap 2019). A feature observed since the 1990s in India is that the sex ratio at birth is highly skewed towards males, particularly at higher birth orders (Gellatly and Petrie 2017; Visaria 2005; Das 1987; Nath 2023). Parents seldom sex-select at the first birth since they prefer to have a child of either gender over the possibility of not having a child. However, in the presence of son preference, parents whose firstborn is a daughter are more likely to have prenatal sex-selective abortions from the second birth onwards than are parents whose firstborn is a son. Figure1 plots the sex ratio at birth from 2000 to 2016 at various birth orders. The horizontal line at 106 is the reference line for the natural sex ratio at birth. The solid line plots the sex ratio at birth for children born at birth order one, i.e. the first-born children. This line closely follows the reference line indicating a balanced sex ratio for firstborn children. The dashed line and the dotted line plot the sex ratio at birth for children born at birth order two and birth order three or above, respectively. Both of these lines diverge increasingly Fig. 1 Sex ratio at birth, by birth order. Sex ratio is measured as the number of males per 100 females 7 PNSDT or fetal gender identification technology. S.Javadekar, K.Saxena 32 Page 8 of 34 from the reference line of the natural sex ratio, indicating that the sex ratio at birth for children born at higher birth orders is substantially distorted towards males. This distortion at higher parity suggests that sex selection is more prevalent for pregnancies at a higher order. Although the sex ratio imbalance for children born at higher birth orders is linked to prenatal sex determination technology like ultrasounds, the literature also discusses other channels that influence sex-selective behaviour among Indian households, such as the price of gold, dowry and marriage conventions and the religious identity of the political leader (Bhalotra etal. 2018, 2020). 2.1 Janani Suraksha Yojna In 2005, the Government of India launched Janani Suraksha Yojana, a conditional cash transfer program sponsored 100% by the national Government with a dual objective of reducing the number of maternal and neonatal deaths nationwide.8 This scheme promoted safe motherhood by providing cash incentives to women if they delivered their children either in government hospitals or in accredited private health institutions or at home under medical supervision.9 A further condition to receive the full cash incentive was that the mother should undergo at least three prenatal checkups that include ultrasound and amniocentesis, technologies used to determine fetal sex. By mandating ante-natal checkups, JSY enabled higher access and use of ultrasound technology even in areas that previously did not have access to it. Eligibility for the conditional cash transfer was dependent on the place of residence, income level, and the caste of the household. The scheme, implemented nationwide in April 2005, classified states as lowand high-performing based on the rates of institutional deliveries, i.e. the proportion of women who give birth at health centres as shown in Figure1. Low-performing states were states where the institutional delivery rate was less than 25%. These included Uttar Pradesh, Uttranchal, Bihar, Jharkhand, Madhya Pradesh, Chhattisgarh, Assam, Rajasthan, Orissa, and Jammu and Kashmir. The remaining states were classified as high-performing states. The objective of this program was to reduce maternal and child mortality rates by increasing the number of women who gave birth safely at health facilities (Joshi and Sivaram 2014). In low-performing states, all pregnant women were program beneficiaries and the benefits were paid regardless of whether the women delivered in a government hospital or a private accredited health centre and regardless of the birth order of their children. In high-performing states, only women who were classified as living below the poverty line (BPL) or belonging to a scheduled caste or scheduled tribe (SC/ ST) were eligible for program benefits. Eligibility in these states was restricted to 8 JSY is a modified graded version of the National Maternity Benefit Scheme which uniformly provided all below poverty line women throughout the country with Rs 500 per live birth up to two live births. This Scheme was suspended after JSY was launched. Since our comparison groups do not comprise women below the poverty line, our estimates are not affected by the earlier scheme. 9 This included government health centres such as Sub centres/Primary Health centres/Community Health centres/First Referral Units/general wards of the district or state hospitals. The seen andunseen: theunintended impact ofaconditional… Page 15 of 34 32 of these arguments support the case for natural experiments created by the sex of the firstborn and satisfy the first condition mentioned above. Further, while having a firstborn girl or boy does not inherently confer additional benefits for accessing the JSY program, families with firstborn girls may have an incentive to utilize prenatal sex-selection technologies offered through the program for sex-selection (Akbulut-Yuksel and Rosenblum 2023; Anukriti et al. 2022a, b; Bhalotra and Cochrane 2010). Therefore, the group with firstborn girls the group potentially benefits from the program in both LPS and HPS states compared to the firstborn boys group and hence satisfies the second condition. The triple difference estimator is equivalent to taking the difference between two difference-in-difference estimators (Olden and Møen 2022). It first takes the difference between first girl and first boy families over time in treatment group and control group separately and then takes the difference if these two differences. Since the first girl families and the first boy families in the treatment and control states experience the same state-specific time-varying effects. By doing this the triple difference Table 2 Balance test The baseline descriptive statistics are for families with firstborn girl and firstborn boy in rural areas during pre-program years 2000–2005. In the last column, we have the coefficients for the regression of the respective variable on the indicator FirstGirl. Standard errors clustered at the state level. The variable self-reported ultrasound use has missing values hence the number of observations is different than the rest of the variables * p < 0.1; **p < 0.05; ***p < 0.01 First girl families First boy families Diff NMean SD NMean SD Hindu 35,539 0.76 0.43 38,284 0.76 0.43 − 0.002 Muslim 35,539 0.11 0.31 38,284 0.11 0.31 0.002 Forward caste 35,539 0.18 0.38 38,284 0.17 0.38 0.003 OBC 35,539 0.38 0.49 38,284 0.39 0.49 − 0.006** Mother’s education 35,539 4.47 4.59 38,284 4.49 4.58 − 0.017 Sex of household head 35,539 1.12 0.33 38,284 1.12 0.33 0.000 Age of household head 35,539 44.16 13.07 38,284 44.13 13.08 0.032 Self-reported ultrasound use 1771 0.29 0.45 1519 0.23 0.42 0.058*** Poorest 35,539 0.22 0.42 38,284 0.22 0.41 0.004 Poorer 35,539 0.21 0.40 38,284 0.21 0.41 − 0.001 Middle 35,539 0.20 0.40 38,284 0.20 0.40 0.004 Richer 35,539 0.19 0.39 38,284 0.19 0.39 − 0.003 Richest 35,539 0.18 0.39 38,284 0.19 0.39 − 0.004 Electricity 35,539 0.96 0.94 38,284 0.96 0.93 0.002 Truck 35,539 0.18 0.99 38,284 0.18 0.99 0.000 Fridge 35,539 0.32 1.03 38,284 0.32 1.02 − 0.005 Cycle 35,539 0.66 1.03 38,284 0.70 1.02 − 0.040*** TV 35,539 0.66 1.03 38,284 0.67 1.02 − 0.004 Radio 35,539 0.22 1.01 38,284 0.22 1.00 0.002 S.Javadekar, K.Saxena 32 Page 16 of 34 estimator allows us to account for state-specific confounding effects which we could not in our difference in difference estimation. Like the DID estimation, the DDD estimator also requires a parallel trend assumption for the estimated effect to have a causal interpretation. Although DDD is the difference between two difference-in-differences (difference between first girl and first boy families over time and difference between treatment and control groups over time), it does not require two parallel trends assumption (Olden and Møen 2022). We discuss this identifying assumption in the ‘Identification Assumption’ section. We run the following triple difference specification where Treat × Post interacts with an indicator for first girl families given by FirstGirl. The triple difference specification estimated is as follows: The DDD coefficient β1 captures the difference in the likelihood of female births between families with firstborn daughters in the treatment and control group. We include mother fixed effects, birth order fixed effects, year-of-birth fixed effects, and state-year-of-birth fixed effects. Furthermore, we estimate the above DID and DDD (2) Girl bist = 𝛽 0+ 𝛽 1 Treat is × Post t× FirstGirl i +𝛽2Postt×FirstGirli+𝛽3Treatis ×Postt +𝛽4FirstGirli×Treatis +𝛽5Treatis +𝛽6Postt+𝛽7FirstGirl i + Stateyearst +𝛿 i +𝜆 t +𝜃 b + ebits Table 3 Main results: estimation results for difference-in-differences estimation The table reports the difference-in-differences estimation coefficient of the impact of the JSY on the likelihood of observing that the child born is a girl. Treat is the dummy variable that takes the value 1 if the mother is from our treatment group. Post compares post-program years (2006–2015) to the pre-program years (2000–2005). Post 2006−10 and Post2011−15 are the early (2006–2010) and late diffusion (2011–2015) periods of the program. The main FEs include mother, birth order, and year of birth fixed effects as indicated. Season FEs are birth month fixed effects and Season-Year FEs are month and year-specific birth month fixed effects. All standard errors are clustered bootstrapped (with 1000 reps) at the state level and reported in parentheses * p < 0.1; **p < 0.05; ***p < 0.01 (1) (2) (3) (4) (5) Treat × Post 0.048** 0.048** 0.048** (0.023) (0.023) (0.023) Treat × Post2006-10 0.041** 0.041* (0.020) (0.021) Treat × Post2011-15 0.086** 0.086** (0.036) (0.038) No. of Obs 150,757 150,757 150,757 150,757 150,757 Main FE Yes Yes Yes Yes Yes Season FE No No Yes Yes No Season and Year FE No No No No Yes The seen andunseen: theunintended impact ofaconditional… Page 17 of 34 32 Table 4 Main results: Estimation results for triple difference estimation The table reports triple difference estimation coefficient of the impact of the JSY on the likelihood of observing the child born to first girl families is a girl. Dependent variable is girl. Treat is the dummy variable that takes the value 1 if the mother is from our treatment group. Similarly, FirstGirl is an indicator for if the woman’s firstborn child was a girl. Post compares post-program years (2006–2015) to the pre-program years (2000–2005). Post2006−10 and Post2011−15 are the early (2006–2010) and late diffusion (2011–2015) periods of the program. The main FEs include mother, birth order, and year of birth fixed effects as indicated. The state-year trend is the state specific time trend, and the state year FE is the State Year specific fixed effect. Season FEs are birth month fixed effects, and Season Year FEs are birth month and year-specific fixed effects. All standard errors are clustered bootstrapped (with 1000 reps) at the state level and reported in parentheses * p < 0.1; **p < 0.05; ***p < 0.01 (1) (2) (3) (4) (5) Dep (6) (7) (8) (9) Treat × Post × FirstGirl 0.126** 0.105* 0.114** 0.127** 0.126** (0.056) (0.058) (0.057) (0.055) (0.056) Treat × Post2006-10 × FirstGirl 0.116** 0.097* 0.107* 0.116** (0.056) (0.058) (0.056) (0.056) Treat × Post2011-15 × FirstGirl 0.183*** 0.152** 0.163** 0.184*** (0.069) (0.069) (0.067) (0.070) Treat × Post − 0.011 0.000 − 0.080 − 0.012 − 0.013 (0.050) (0.000) (0.053) (0.051) (0.053) Post × FirstGirl − 0.181*** − 0.147*** − 0.157*** − 0.181*** − 0.180*** (0.053) (0.056) (0.055) (0.053) (0.054) Treat × Post2006-10 − 0.017 0.000 − 0.063 − 0.018 (0.049) (0.000) (0.048) (0.048) Treat × Post2011-15 0.014 0.000 − 0.079 0.012 (0.064) (0.000) (0.060) (0.064) No. of Obs 63,250 63,250 63,204 63,204 63,250 63,250 63,232 63,232 63,250 Main FE Yes Yes Yes Yes Yes Yes Yes Yes Yes State Year FE No No Yes Yes No No No No No State Year Trend No No No No Yes Yes No No No Season FE No No No No No No Yes Yes No Season Year FE No No No No No No No No Yes S.Javadekar, K.Saxena 32 Page 18 of 34 specifications by classifying the post-JSY years into the early and late diffusion periods. This is done for two reasons. First, as additional features were added to JSY in 2011, we can see how the impact changed over the two diffusion periods. Second, we have information on the anthropometric outcomes for children born in the late diffusion period. By classifying the effects into diffusion periods we can tie the effect of the program on the sex ratio at birth for this cohort to their average anthropometric welfare outcomes. 4 Results Table3 presents the results for the DID estimation, and Table4 shows the results for our triple difference estimator. In the first column of Table3, the post-program years 2006 to 2015 are compared with the pre-program years 2000 to 2005. In the second column, the post-program years are divided into a late diffusion period (2011–2015) and an early diffusion period (2006–2010) and compared to the reference pre-program years. The key variables of interest are Treatis × Post, Treatis × Post2006-10, and Treatis × Post2011-15. Columns 1, 3, and 5 of Table 3 show that the likelihood of a female birth increased by 4.8 percentage points in the treatment group. This translates to a nearly 10% increase in the number of girls born to mothers in the treatment group. When we look at the early and late diffusion periods of the policy, we see that in the early diffusion period, this likelihood increases by 4 percentage points while in the later period increases by 8.6 percentage points. This result is interesting because it shows a reduction in sex-selective behaviour among the groups that have been known in the literature to sex select, i.e. non-SC/ST and non-BPL groups. The key coefficients of interest are the triple difference estimators. Similar to the DID specification, we first look at the post-policy period from 2006 to 2015 in Table4 in columns 1,3,5,7, and 9 and then we differentiate between the early and late diffusion periods in columns 2,4,6, and 8 of Table4. We see that the program led to an increase in the likelihood of female births from birth order 2 onwards for families with a first born female child in the treatment group by 12.6 percentage points. There was an increase of almost 18.3 percentage points in the later diffusion period and 11.6 percentage points in the early diffusion period (column 2). We add state-year fixed effects and state-year trends to our specifications. After including state-year fixed effects this estimate reduces to 10.5 percentage points (column 3) with increases of 9.7 and 15.2 percentage points in the likelihood of female births in the earlier and later diffusion periods (column 4). This is a more conservative specification as it controls for state-specific time-varying confounders. This suggests that for families with first-born daughters in the treatment group, the increase in the number of girls after 2005 was nearly 23%, compared to families with a first-born boy. In columns 7 and 8 we include month of birth fixed effect and in column 9 we additionally include birth month-year specific fixed effects to account for seasonality of births (Boland etal. 2020; Krombholz 2023). The seen andunseen: theunintended impact ofaconditional… Page 19 of 34 32 Though our triple difference estimate suggests increased likelihood of birth of girls, the coefficient on Post × FirstGirl in all our specifications shows that in the control group first girl families after the year 2005 were significantly less likely to have second birth of a girl compared to first boy families before 2005. Overall, our results suggest that an unintentional impact of the program is the reduction in sex selective abortions and an increase in the probability of girls being born, in families eligible for treatment. We also see that most of the positive results are driven by the larger impacts in the later diffusion periods. 5 Robustness tests 5.1 Identification assumption A key assumption of a DID estimation is that in the absence of the program, the outcome variable in the treatment and control groups has parallel trends, i.e. the outcome variable would have evolved in the same way for both groups. For validity of our analysis, the probability of having a girl at the next birth should not be significantly different across mothers in the treatment and control groups during the preprogram years. To test this we run a specification where the effect of the program is allowed to vary by year, as in an event study analysis. This approach is recommended and widely used for detecting pretrends (Roth etal. 2023). For us to be confident that the program had a causal impact on the sex-selective behaviour of mothers, we should not observe any significant differences in the probability of having a girl in the comparison groups prior to the program. Significant differences, if any, should only occur after the program if the program has any effect on sex-selective abortions. To check this, we estimate the following specification for a DID and a DDD: Fig. 2 Test for parallel trends. a Plots estimated difference in the likelihood of girl births between treatment and control groups, conditional on mother fixed effects. b Plots the estimated differences in the likelihood of girl births to mothers with first girls in treatment groups with their counterparts in the control group. The dashed red line represents the year of the JSY program. The joint test of the significance of the lead years of the program yielded a F-Statistic of 1.26 and 1.74 respectively, implying that before the program, the number of girl births evolved similarly in the two groups S.Javadekar, K.Saxena 32 Page 20 of 34 Figure2a shows the likelihood of a girl being born to a mother in the treatment or control groups is not significantly different for years prior to 2005. Similarly, Figure2b shows the likelihood of giving birth to a girl is not significantly different for first-girl families between the treatment and control groups. Conditional on mother fixed effects, we find no significant differences in the likelihood of birth of girls between the treatment and control groups before program implementation in 2005. The differences in outcomes become significant only after 2009. The joint test of the significance of the lead years of the program yielded p-value of 0.3 and 0.163, implying that prior to the program, the number of girl births evolved similarly in the two groups. While testing for pretrends alone is not sufficient we also support our results with the placebo tests discussed in the next subsection. Another factor that could bias our results and threaten our identification strategy is whether the implementation of the JSY was anticipated before 2005. We would then be conflating our estimate with the households’ expectations. If this were the case, then households in the treatment group should have changed their fertility behaviour prior to 2005 and we should see a decrease in female births. However, if households in the treatment group did not change their behaviour prior to 2005 differently than did households in the control group, i.e. the probability of female births was similar in both groups prior to 2005, we can say that households did not anticipate the programme and that the year of implementation was exogenous. In Figure2, we show that the difference in the number of female births prior to 2005 (3) Girl bist =𝛽0+ ∑2015 j = 2000 𝛽jTreatis ×Yearj+𝛿i+𝜆t+𝜃b+e bist (4) Girl bist =𝛽0+ 2015 ∑ j = 2000 𝛽jTreatis ×Yearj×FirstGirli+𝜙st +𝛿i+𝜆t+𝜃b+e bist Fig. 3 Falsification tests using DHS—2015/16. This figure reports coefficients of the difference-in-difference analysis assuming years from 1990 to 2004 as program years and checks if the likelihood of female birth is different across the treatment and control group. All regressions contain mother, birth, and year fixed effects. Standard errors are clustered at the state level. The F-statistics of the pre-treatment years jointly is 1.04 The seen andunseen: theunintended impact ofaconditional… Page 21 of 34 32 was not significant, indicating that households did not anticipate the program and change their fertility behaviour. 5.2 Falsification Tests If our empirical strategy identifies the causal impact of the program on the fertility decisions of mothers, then we should not be able to see any effect on mothers who never received the program. Our first falsification test involves individual assumptions for each year from 1990 to 2004, i.e. years prior to 2005, to be the program year. Assuming that each of the years was the year when the JSY was implemented, we checked the impact of the program on the likelihood of girl births across treatment and control mothers. Figure3 plots the coefficient for each year and we can see that the probability of girl birth across treatment and the control groups is not significantly different for any of the years except 1996 and 1997. The significant difference in these 2 years could be due to the structural break in 1995 when ultrasound technology became widely available in India (Bhalotra and Cochrane 2010). However, the effect of this structural break did not last long and dissipated after 1997 as can be seen in Figure3. The coefficients for the remaining years are not significant and the differences in the outcome only appear after 2005, i.e. after the JSY was implemented, suggesting that what we are capturing is the causal effect of the JSY. The second falsification test is to run our triple DID specification on DHS2005/06. Since this survey was completed by 2005–2006, the women interviewed in this sample never received the program. This idea is similar to the test above. We should not find any effect of the program on women who never received the program. Here, we assume 1995 as the year the JSY was implemented and compare children born up to 10 years after 1995 with children born up to 5 years prior to 1995. Our sample consists of mothers who made their fertility decisions from 1990 onwards, since we assume 1995 to be the year that the program was rolled out. We compare children born between 1996 and 2000 (our assumed early diffusion period) and between 2001 and 2005 (our late diffusion period) with those born between 1990 and 1995. One reason for this is that if there are any reporting biases in fertility for children born more than 10 years prior to the survey year, then these biases should be the same in any DHS sample. Hence, if our main results are driven by reporting bias, then we will also see significant differences in the outcome of our DHS-III estimation results. We estimate the following specification for DHS-III: Table5 shows the results of our falsification test on mothers whose fertility decisions were made in 1990.15 In both columns, we see that the likelihood of giving (5) Girl bits = 𝛽 0+ 𝛽 1 Treat is × Post 1996−00,t× FirstGirl i+ 𝛽 2 Treat is × Post 2001−05,t× FirstGirli +𝜙 st +𝛿 i +𝜆 t +𝜃 b +e bits 15 An additional falsification test assuming the year 2000 to be the treatment year for the DHS III sample is shown in the Appendix. S.Javadekar, K.Saxena 32 Page 22 of 34 birth to a girl is not significantly different for families whose first child was a girl across the treatment and the control groups. A lack of significance will indicate that our empirical strategy is to capture only the program effect. Both falsification tests support our claim of causal identification of the program effect on the likelihood of girl births with the empirical strategy we employ. 6 Discussion andadditional evidence The previous section described the causal impact of the JSY on sex-selective abortions in India. The program caused an increase in the number of girls born to families eligible to receive the JSY benefits, indicating that the mechanism of access to prenatal sex determination technologies was not dominant. Previous work has shown that in societies with a preference for male children, girls suffer from lower welfare in families that follow the stopping rule and have more girls than they desire. This discrimination is starker for girls at higher birth orders. In this section, we therefore test the hypothesis that girls born in families with son preference will be worse off. Table 5 Falsification test: triple difference estimation using DHS 2005–06 The table reports the triple difference results for the falsification tests using DHS 2005–06 data collected prior to the implementation of the program. In columns (1) and (2), we assume 1995 to be the year of program implementation. In column (1), we consider years 1996–2000 and years 2001–2005 as early and late diffusion periods. These are compared to the pre-program period 1990–1995. In column (2), we assume years 1996–2005 as post-program years. In column (3), we assume 2000 as the year of program implementation. Post-program years 2001–2005 are compared to pre-program years 1996–2000. Treat is the dummy variable that takes the value 1 if the mother is from the treatment group. Similarly, FirstGirl indicates if the woman’s firstborn child was a girl. FE contains mother, birth, and year fixed effects. All triple difference estimates are for children at parity 2 onward. Main FE contains mother, birth, and year fixed effects. Standard errors in parenthesis are clustered at the state level * p < 0.1; **p < 0.05; ***p < 0.01 (1) (2) (3) Treat × Post1996-00 × FirstGirl − 0.036 (0.64) Treat × Post2001-05 × FirstGirl − 0.092 (0.82) Treat × Post1995-05 × FirstGirl − 0.048 (0.67) Treat × Post2001-05 × FirstGirl − 0.052 (0.04) Main FE Yes Yes Yes No. of Obs 15,524 15,524 11,987 The seen andunseen: theunintended impact ofaconditional… Page 23 of 34 32 Table 6 Estimation results for mortality for children under 1year The table reports the mortality outcomes for children below age 1. Columns 1 and 2 record the likelihood of girls dying before reaching age 1. Columns 3 and 4 record the likelihood of girls born at parity 2 and above dying before reaching age 1. Columns 5 and 6 record the likelihood of girls born at parity 3 and above dying before reaching age 1. Treat that takes the value 1 if the mother is from our treatment group. Post compares post-program years (2006–2015) to the pre-program years (2000–2005). Post2006-10 and Post2011-15 are the early (2006–2010) and late diffusion (2011–2015) periods of the program. The main FEs include mother, birth, and year fixed effects. Standard errors in parentheses are clustered bootstrapped (1000 reps) at the state level * p < 0.1; **p < 0.05; ***p < 0.01 All Births Parity > 1 Parity > 2 (1) (2) (3) (4) (5) (6) Treat × Post × Girl 0.007 0.020 0.042 (0.007) (0.013) (0.030) Treat × Post2006-10 × Girl 0.009 0.025** 0.071** (0.007) (0.012) (0.033) Treat × Post2011-15 × Girl 0.005 0.017 0.009 (0.008) (0.018) (0.038) No. of Obs 150,757 150,757 63,250 63,250 23,275 23,275 Main FE Yes Yes Yes Yes Yes Yes Table 7 Estimation results for mortality for children under 5year The table reports the likelihood of a girl in treatment group dying before she reaches age 5. Columns 3 and 4 record the likelihood of girls born at parity 2 and above dying before reaching age 5. Columns 5 and 6 record the likelihood of girls born at parity 3 and above dying before reaching age 5. Treat that takes the value 1 if the mother is from our treatment group. Post compares post-program years (2006– 2015) to the pre-program years (2000–2005). Post2006-10 and Post2011-15 are the early (2006–2010) and late diffusion (2011–2015) periods of the program. The main FEs include mother, birth, and year fixed effects. Standard errors in parentheses are clustered bootstrapped (1000 reps) at the state level * p < 0.1; **p < 0.05; ***p < 0.01 All Births Parity > 1 Parity > 2 (1) (2) (3) (4) (5) (6) Treat × Post × Girl 0.005 0.013 0.058* (0.007) (0.014) (0.034) Treat × Post2006-10 × Girl 0.008 0.020 0.091** (0.007) (0.015) (0.036) Treat × Post2011-15 × Girl 0.002 0.009 0.018 (0.009) (0.018) (0.039) No. of Obs 150,757 150,757 63,250 63,250 23,275 23,275 Main FE Yes Yes Yes Yes Yes Yes S.Javadekar, K.Saxena 32 Page 24 of 34 6.1 Impact onInfant Mortality We look at the under 5 mortality of children born to women in our sample. Biologically, mortality is greater among boys than girls between the age of 0 and 1 (Kraemer 2000); therefore, if we observe higher mortality for girls than boys, it would indicate that girls are being neglected. Using our difference-in-differences estimator, we tested whether the program increased child mortality for girls. We estimate the model: The results in Tables 6 and 7 show whether there are disproportionately more girls among infants who died in their first year or before reaching five years of age. For each of these samples, the first two columns show the results for all infants in rural India irrespective of their birth order. Columns 3 and 4 show the results for all infants who were born at birth order greater than 1. The last two columns show results for all infants born at birth order greater than 2. We make this distinction by birth order because girls at a higher birth order tend to die more than boys, due to neglect and discrimination. We find that for both age groups, the probability that the deceased child is a girl is positive for all birth orders. For girls born at a parity greater than 1, the likelihood of a girl dying is 2 percentage points greater in the treatment group. This is more prominent in the earlier diffusion period and for girls born at a parity greater than 2. The likelihood of a girl dying before reaching the age of 5 is nearly 6 percentage points greater in the treatment group after the program. An interesting observation is that the significant difference in mortality between girls and boys disappears when we look at the late diffusion period. This could be due to the additional feature of providing nutritional supplements to infants that were added to the program in 2011. 7 Mechanisms 7.1 Ultrasound access channel According to the literature, one of the main channels that impacts households’ sexselective fertility decisions is access to pre-natal sex determination technologies such as ultrasounds. All program beneficiaries were expected to undergo three antenatal checkups that included ultrasound scans. Although discovering the gender of the foetus was not the purpose of the scans, parents might use this information and abort unwanted female foetuses. Since we cannot observe who uses the technology to determine the sex of the foetus and who uses it to satisfy the programme condition, we can hypothesize that if more people were using this aspect of the programme to sex select, we should see this channel to lead to on average a significantly lower probability of girls being born on average in the treatment group. Using the DHS2015/16 data, we obtained information on which mothers reported having used ultrasound technology. Column 2 in Table8 shows the results (6) Deadbit = 𝛽0 + 𝛽1Treati × Postt × Girli + 𝛽2Treati + 𝛽3Postt + Stateyearst + 𝛿i + 𝜆t + 𝜃b + ebit The seen andunseen: theunintended impact ofaconditional… Page 31 of 34 32 8 Conclusion andpolicy recommendations This paper examined the impact of the JSY conditional cash transfer program on the fertility decisions of mothers in rural India. More specifically, this study provides causal evidence of the impact of the JSY on sex-selective behaviour among Indian households. The results show that, contrary to previous work on sex selection, this program led to an increase in the probability of having a girl at each birth order for mothers eligible for the program. The magnitude is especially larger in families who according to the literature have a greater incentive to sex select, i.e. those whose first child is a daughter. Although overall in the country, there has been an increase in the prevalence of sex-selective abortions, the JSY managed to reduce this practice among families who qualified for the program. The results also showed that while there were more girls being born to families in LPS; these girls are also more likely to die before reaching the age of 5 years. These findings indicate that though there are improvements in birth outcomes for girls as a result of the program; it may be the case that discrimination against them continues and shifts from prenatal to postnatal discrimination. Our results show that in the 0–4 age group, 424,825 women were missing from the population. However, this is an improvement of nearly 300,000 women compared to 724,997 missing women in the same age group a decade prior to the program. While there still is a very large number of missing girls in the country, the policy contributed to reducing this number. The channel that leads to this result is the one driven by community health workers (ASHA) that were appointed as part of the program to assist pregnancies in their neighbourhood. Since these workers record each pregnancy for beneficiaries of the program and get financial incentives for every live birth of beneficiaries at health institutions, they act as deterrents for couples to selectively abort their foetuses. This result supports the emerging evidence on the role that health workers play in efficient public good distribution and in supporting health programs. The effectiveness of community health workers in reducing the practice of prenatal sex-selective abortions either due to parental fear of being reported if they undergo a sex-selective abortion or ASHA’s pressure on parents to not abort the child as her payment is conditional on a beneficiary’s delivery in a hospital. This is an important piece of evidence in a country that has been unsuccessfully trying to reduce female foeticide through laws against sex-selective abortions or financial incentives to bear girls. However, our results should be taken with caution as we do not claim that the intervention of health workers shifted parental son preference in India. Our results point to the fact that parental son preference was merely substituted by postnatal excess girl mortality between the ages of 24 and 59 months. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1007/ s0014802501091-6. Acknowledgements The authors thank Giacomo de Giorgi and Lore Vandewalle for their supervision and support. This paper benefited from comments by Christelle Dumas, Eliana La Ferrara, Jaya Krishnakumar, Travis Lybbert, Tobias Mueller, Dilip Mukherjee, Michele Pellizzari, Debraj Ray, Alessandro Tarozzi, and Aleksey Tetenov. We are thankful to all the participants of the brown bag seminar at the S.Javadekar, K.Saxena 32 Page 32 of 34 University of Geneva and the Graduate Institute of International and Development Studies, NCDE 2019, SSDev 2019, DENS Switzerland 2019, the SSES 2019 6th DENeB Berlin 2019, and 2019 ISI conference New Delhi. Biplob Biswas’s Python expertise was indispensable. We are thankful to Vikram Bahure for sharing the CHIRPS data. Finally, we thank the three anonymous referees and editor Kompal Sinha for their valuable comments. Data availability The data and do file will be made available upon request. Declarations Conflict of interest The authors declare no competing interests. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/ licenses/by/4.0/. References Akbulut-Yuksel M, Rosenblum D (2023) Estimating the effects of expanding ultrasound use on sex selection in India. J Dev Stud 59(4):516–532 Alfano M (2017) Daughters, dowries, deliveries: the effect of marital payments on fertility choices in India. J Dev Econ 125:89–104 Almond D, Li H, Zhang S (2019) Land reform and sex selection in China. J Polit Econ 127(2):560–585 Almond, D., H. Li, and S. Zhang (2013). Land reform and sex selection in China. Technical report, National Bureau of Economic Research Anderson S, Ray D (2010) Missing women: age and disease. Rev Econ Stud 77(4):1262–1300 Anukriti S (2018) Financial incentives and the fertility-sex ratio trade-off. Am Econ J Appl Econ 10(2):27–57 Anukriti S, Bhalotra S, Tam EH (2022a) On the quantity and quality of girls: fertility, parental investments and mortality. Econ J 132(641):1–36 Anukriti S, Kwon S, Prakash N (2022b) Saving for dowry: evidence from rural India. J Dev Econ 154:102750 Arnold F, Choe MK, Roy TK (1998) Son preference, the family-building process and child mortality in India. Popul Stud 52(3):301–315 Banerjee S, Sen G (2024) Persistent effects of a conditional cash transfer: a case of empowering women through Kanyashree in India. J Popul Econ 37(4):66 Bertrand M, Duflo E, Mullainathan S (2004) How much should we trust differences-in-differences estimates? Q J Econ 119(1):249–275 Bhalotra S, Chakravarty A, Gulesci S (2020) The price of gold: dowry and death in India. J Dev Econ 143:102413 Bhalotra SR, Cochrane T (2010) Where have all the young girls gone? Identification of sex selection in India. https:// doi. org/ 10. 2139/ ssrn. 17311 85 Bhalotra SR, Clots-Figueras I, Iyer L (2018) Religion and abortion: the role of politician identity. https:// doi. org/ 10. 2139/ ssrn. 31172 96 Bharadwaj P, Lakdawala LK (2013) Discrimination begins in the womb: evidence of sex-selective prenatal investments. J Human Resour 48(1):71–113 Bhaskar V (2007) Parental choice and gender balance. University College London The seen andunseen: theunintended impact ofaconditional… Page 33 of 34 32 BjörkmanNyqvist M, Guariso A, Svensson J, Yanagizawa-Drott D (2019) Reducing child mortality in the last mile: experimental evidence on community health promoters in Uganda. Am Econ J Appl Econ 11(3):155–192 Boland MR, Fieder M, John LH, Rijnbeek PR, Huber S (2020) Female reproductive performance and maternal birth month: a comprehensive meta-analysis exploring multiple seasonal mechanisms. Sci Rep 10(1):555 Borker G, Eeckhout J, Luke N, Minz S, Munshi K, Swaminathan S (2017) Wealth, marriage, and sex selection.Unpublished manuscript Borker G (2021) Safety first: perceived risk of street harassment and educational choices of women Brenner JL, Kabakyenga J, Kyomuhangi T, Wotton KA, Pim C, Ntaro M, Bagenda FN, Gad NR, Godel J, Kayizzi J etal (2011) Can volunteer community health workers decrease child morbidity and mortality in southwestern Uganda? An Impact Evaluation Plos One 6(12):e27997 Buchmann N, Field EM, Glennerster R, Hussam RN (2019) Throwing the baby out with the drinking water: unintended consequences of arsenic mitigation efforts in Bangladesh(No. w25729). National Bureau of Economic Research Celhay PA, Gertler PJ, Giovagnoli P, Vermeersch C (2019) Long-run effects of temporary incentives on medical care productivity. Am Econ J Appl Econ 11(3):92–127 Chatterjee S, Poddar P (2021) From safe motherhood to cognitive ability: exploring intrahousehold and intergenerational spillovers. Economica 88(352):1075–1106 Chen Y, Li H, Meng L (2013) Prenatal sex selection and missing girls in China: evidence from the diffusion of diagnostic ultrasound. J Human Resour 48(1):36–70 Clark S (2000) Son preference and sex composition of children: evidence from India. Demography 37(1):95–108 Cohen A, Dehejia R, Romanov D (2013) Financial incentives and fertility. Rev Econ Stat 95(1):1–20 Congdon Fors H, Lindskog A (2023) Son preference and education inequalities in India: the role of gender-biased fertility strategies and preferential treatment of boys. J Popul Econ 36(3):1431–1460 Das N (1987) Sex preference and fertility behavior: a study of recent Indian data. Demography 24(4):517–530 Das Gupta M, Mari Bhat P (1997) Fertility decline and increased manifestation of sex bias in India. Popul Stud 51(3):307–315 Ebenstein A (2010) The “missing girls” of China and the unintended consequences of the one child policy. J Human Resour 45(1):87–115 Edlund L (1999) Son preference, sex ratios, and marriage patterns. J Polit Econ 107(6):1275–1304 Edlund L, Li H, Yi J, Zhang J (2013) Sex ratios and crime: evidence from China. Rev Econ Stat 95(5):1520–1534 Gellatly C, Petrie M (2017) Prenatal sex selection and female infant mortality are more common in India after firstborn and second-born daughters. J Epidemiol Commun Health 71(3):269–274 Goodkind D (1996) On substituting sex preference strategies in East Asia: does prenatal sex selection reduce postnatal discrimination? Popul Dev Rev 22(1):111–126 Hesketh T, Xing ZW (2006) Abnormal sex ratios in human populations: causes and consequences. Proc Natl Acad Sci 103(36):13271–13275 Hu L, Schlosser A (2015) Prenatal sex selection and girls’ well-being: evidence from India. Econ J 125(587):1227–1261 Jayachandran S, Kuziemko I (2011) Why do mothers breastfeed girls less than boys? Evidence and implications for child health in India. Q J Econ 126(3):1485–1538 Jensen R (2012) Another mouth to feed? The effects of (in) fertility on malnutrition. Cesifo Econ Stud 58(2):322–347 Joshi S, Sivaram A (2014) Does it pay to deliver? An evaluation of India’s safe motherhood program. World Dev 64:434–447 Kashyap R (2019) Is prenatal sex selection associated with lower female child mortality? Popul Stud 73(1):57–78 Kraemer S (2000) The fragile male. BMJ 321(7276):1609–1612 Krombholz H (2023) Global patterns of seasonal variation in human birth rates-influence of climate and economic and social factors.https:// doi. org/ 10. 23668/ psych archi ves. 13203 Lin M-J, Liu J-T, Qian N (2014) More missing women, fewer dying girls: the impact of sex-selective abortion on sex at birth and relative female mortality in Taiwan. J Eur Econ Assoc 12(4):899–926 Nandi A, Laxminarayan R (2016) The unintended effects of cash transfers on fertility: evidence from the safe motherhood scheme in India. J Popul Econ 29(2):457–491 S.Javadekar, K.Saxena 32 Page 34 of 34 Nath S (2023) Explaining third birth patterns in India: causal effects of sibling sex composition. J Popul Econ 36(4):2169–2203 Olden A, Møen J (2022) The triple difference estimator. Economet J 25(3):531–553 Powell-Jackson T, Mazumdar S, Mills A (2015) Financial incentives in health: new evidence from India’s Janani Suraksha Yojana. J Health Econ 43:154–169 Ritchie H, Roser M (2019) Gender ratio. Our World in Data. https:// ourwo rldin data. org/ genderratio . Accessed May 2019 Rose E (1999) Consumption smoothing and excess female mortality in rural India. Rev Econ Stat 81(1):41–49 Rosenblum D (2013a) The effect of fertility decisions on excess female mortality in India. J Popul Econ 26(1):147–180 Rosenblum D (2013b) Economic incentives for sex-selective abortion in India. Canadian Centre for Health Econ 2014–13 Roth J, Sant’Anna PH, Bilinski A, Poe J (2023) What’s trending in difference-in-differences? A synthesis of the recent econometrics literature. Journal of Econometrics 235(2):2218–2244 Roy S (2015) Empowering women? Inheritance rights, female education and dowry payments in India. J Dev Econ 114:233–251 Sekher TV (2012) Ladlis and Lakshmis: financial incentive schemes for the girl child. Economic and Political Weekly 58–65 Sekhri S, Storeygard A (2014) Dowry deaths: response to weather variability in India. J Dev Econ 111:212–223 Sen S, Chatterjee S, Mohanty SK etal (2020) Unintended effects of Janani Suraksha Yojana on maternal care in India. SSM-Population Health 11:100619 Singh P, Masters WA (2017) Impact of caregiver incentives on child health: evidence from an experiment with anganwadi workers in India. J Health Econ 55:219–231 Sinha N, Yoong J (2009) Long-term financial incentives and investment in daughters: evidence from conditional cash transfers in North India. The World Bank Valente C (2014) Access to abortion, investments in neonatal health, and sex-selection: evidence from Nepal. J Dev Econ 107:225–243 Visaria L (2005) Female deficit in India: role of prevention of sex selective abortion act. In: CEPEDCICREDINED Seminar on Female Deficit in Asia: T rends and Perspectives, Singapore, pp. 5–7. Citeseer Waterfield H (1875) Memorandum on the census of British India 1871–72.Eyre and Spottiswoode Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.