Abstract
Worldwide, 1.6 million girls are “missing” at birth every year. One policy tool to improve the sex ratio is a conditional cash transfer that pays parents to invest in daughters, but existing evidence on their effectiveness is sparse. Using a difference-in-differences framework, we evaluate the Dhanlakshmi scheme, an Indian CCT program that strongly encouraged girl births without restricting fertility. Dhanlakshmi improved the sex ratio at birth, with only a small fertility increase. The girl-birth effect was concentrated among the first two parities and partially persisted after the program was discontinued. Post-birth outcomes like immunization and education also improved.
I. Introduction
More than 1.6 million girls worldwide are “missing” at birth every year, a combination of cultural preferences for sons and the diffusion of cheap technology for prenatal sex determination (Bongaarts and Guilmoto 2015). Economic development and increasing educational attainment have been largely ineffective in reversing this trend because rising incomes and improved information access also reduce the barriers to sex-selective abortions (Jha et al. 2011; Jayachandran 2017). Prohibiting sex-selective abortions has done little to address the problem (Bhalotra and Cochrane 2010). In India, for example, illegal ultrasound providers routinely escape detection and prosecution. Governments and nonstate actors seeking to normalize the sex ratio among young children therefore need alternative policy tools.
One potentially effective intervention is a conditional cash transfer (CCT) that pays parents for having a girl child if she reaches certain milestones, such as immunization or school enrollment. Such programs not only incentivize parents to have a daughter, but also encourage them to invest in her post birth. India provides an ideal context to study the performance of such schemes. The country accounts for one-third of the world’s 126 million missing women and is the only one to have widely experimented with CCTs to reduce sex-ratio imbalances (Sekher 2010; Bongaarts and Guilmoto 2015; Kumar and Sinha 2020). Since the early 1990s, Indian governments have implemented more than 20 girl-child CCTs embodying a variety of designs.
Table 1 provides an overview of Indian girl-child CCT programs by start year, with their spatial coverage and scope. The first girl-child CCTs were introduced in 1992 in Rajasthan and Tamil Nadu. Since then, many others have been launched, some of which have been discontinued.1 All but two programs explicitly promote girl births, and about half provide payments for immunization, education, or marriage delay. Table 2 summarizes the restrictions associated with each scheme’s design. Most limit participation based on income, either through a below-the-poverty-line (BPL) or other specified income cutoff, and restrict the number of girls that can be enrolled from the same family. A number of programs also require proof of terminal family planning methods, such as sterilization, or impose a maximum number of children in the household. Households that fail to comply with fertility limits lose their eligibility. Payout timing, frequency, and generosity vary widely. The most common structure is a single payment to the girl when she reaches adulthood, usually conditional on having remained unmarried. The payment is typically financed through a government deposit with an insurance company when she is enrolled.2
Overview of Girls Child CCT Programs
Overview of Girl Child CCT Programs
Despite the frequent use of girl-child CCTs in India, little is known about their effectiveness, let alone the pros and cons of specific design features. One exception is Devi Rupak, a CCT program launched in Haryana in 2002. While the program is relatively generous, paying out meaningful monthly benefits for up to 20 years, it is highly restrictive. Devi Rupak limits benefits to households with either one child or two daughters and no sons, and it includes a sterilization requirement, with the objective of improving the sex ratio without increasing fertility.3 Anukriti (2018) studies the effects of this scheme using a difference-in-differences framework and finds that Devi Rupak actually worsened the sex ratio at first birth by 1–2 percent. Her work is a clear warning that imposing restrictions to attenuate the sex ratio–fertility trade-off may undermine the primary program goal.
Anukriti’s findings for Devi Rupak raise the question whether program designs that do not trade off girl births and family size could be more effective at improving the sex ratio. As Table 2 shows, more recent CCTs have apparently considered the possibility, dropping terminal family planning requirements or number-of-children restrictions. One program that exemplifies this more liberal approach is Dhanlakshmi. This scheme, which was introduced in 2008 and operated through 2013, provided cash payments for giving birth to a girl, immunizing her, enrolling her in school, and deferring her marriage until age 18. While its overall generosity was roughly on par with Devi Rupak, Dhanlakshmi did not impose any limits on family size or the number of beneficiaries, nor did it apply a means test. The only eligibility criterion was residency in an area where the program was implemented. In sharp contrast to Devi Rupak, Dhanlakshmi’s design put the goal of promoting the birth of girls at odds with the competing policy aim of lowering fertility.
In this paper, we examine whether Dhanlakshmi improved the sex ratio at birth, and if so, whether this change came at the cost of a fertility increase. Our analysis is focused on the state where the scheme was most effectively implemented, Punjab, which also has one of the most skewed sex ratios in India. We draw on data from repeated cross sections of large household surveys, as well as on census data. Our main analysis compares outcomes for mothers in the treated district of Fatehgarh Sahib with their counterparts in other Punjabi districts before and after the scheme was introduced. Our preferred estimates from the household survey data indicate that Dhanlakshmi increased the probability of a girl birth by 5.5 percentage points relative to the rest of Punjab, while raising the likelihood of any birth by only 0.85 percentage points. Under plausible assumptions about the 2011 Indian census data for the treated area, these estimates imply that Dhanlakshmi was responsible for more than 30 percent of girl births but only 5 percent of total births during the treatment period. All of the gains in girl births occur in the first two birth parities. The effect is concentrated in rural areas where the sex ratio imbalance is more severe and where more than two-thirds of the district population live. Significantly, we also show that Dhanlakshmi reduced the reported desired preference for a son, and that almost half of the girl-birth effect persists one to three years after the scheme ended. These findings are robust to a variety of specifications and control groups and are confirmed in the census data.
The sex-ratio imbalance among young children in India arises not only through sex-selective abortion but also from excess female mortality. Dhanlakshmi’s payout structure is clearly designed with this problem in mind. Thus, we follow the cohort that was born under the Dhanlakshmi program through ages five to seven, estimating its impact on immunization, survival to ages five to seven, and primary school enrollments. Although it is not directly compensated, we also examine whether Dhanlakshmi indirectly affected breastfeeding. We find that the scheme increased immunization rates by up to 7.1 percentage points, the girl share of the five-year-old age group by 4.5 percentage points, and primary school enrollment rates by four percentage points. The breastfeeding rate rose by 2.4 percentage points. Dhanlakshmi’s payouts promoted the same post-birth outcomes for girls born before the scheme was introduced. Extending our analysis of education outcomes to 8–14-year-olds, we show that Dhanlakshmi improved persistence, increasing completed years of schooling by about a third of a year. Finally, we present strong evidence that Dhanlakshmi led to improved health outcomes for mothers during pregnancy and delivery, reducing conditions ranging from excessive fatigue and high blood pressure during pregnancy to premature and prolonged labor during delivery.
We are the first to evaluate the Dhanlakshmi scheme, and there are no other comparable studies of similarly designed girl-child CCTs. Our results therefore have important implications for employing CCTs to reduce son preference and raise the status of girls. First, they indicate that it is possible to place all of the incentives on having and caring for daughters without worrying too much about the fertility effects. Dhanlakshmi lowered the impact of son preference largely by inducing the substitution of daughters for sons, implying that many parents chose to carry a female fetus to term rather than aborting the pregnancy to have a son. That we find such sizable effects even in a comparatively rich state like Punjab is encouraging news for many of the other girl-child CCTs currently in operation in India. Most programs that were introduced since the mid-2000s have already done away with sterilization or fertility restrictions (though typically retaining some income restrictions and beneficiary limits), making them more in line with Dhanlakshmi than with Devi Rupak. Our results suggest that this has been a design move in the right direction.
Second, the results suggest that an unrestricted, competently implemented CCT, whose design clearly signals that encouraging girl births and daughter welfare are the goal, may contribute to a persistent cultural change. Dhanlakshmi’s lasting effect on girl births and the reduction in expressed son preference are consistent with a reduction in the stigma associated with having a daughter. Dhar, Jain, and Jayachandran (2019) document the significant role parents play in shaping the gender attitudes of their children in India, so such a shift in the revealed fertility preferences may have additional long-run benefits.4 Unlike nonmonetary interventions, such as a discussion and persuasion experiment (Dhar, Jain, and Jayachandran 2022), exposure to cable television (Jensen and Oster 2009), or the reservation of seats for female politicians (Beaman et al. 2009), Dhanlakshmi shows that monetary incentives for having a girl can also raise the status of girls and women. This complements and extends recent evidence that higher financial control for women improves female labor force participation and gender attitudes (Field et al. 2021).
II. The Dhanlakshmi Scheme
A. Dhanlakshmi Design: Unrestricted and Generous
The Dhanlakshmi Scheme was launched by the Indian Ministry of Women and Child Development in March 2008 as a pilot program in select areas. All girls born on or after November 19, 2008 were eligible to enroll in the program irrespective of household income or the number of siblings. The only requirement was residency in an area where the program is active. The Indian government chose not to scale up the program to the rest of the country, and the policy was discontinued in April 2013. However, the financial commitments to the enrolled households continue to be honored.5
The scheme provides cash transfers to beneficiary households for every girl from her birth until age 18, conditional on proof of having reached certain milestones. Girls born before November 19, 2008 could still enroll in Dhanlakshmi and benefit from later milestone payouts, starting with the milestones required for their age at enrollment. Table 3 provides an overview of the payments under the program (Brahme and Kumar 2015). A family received 5,000 rupees for showing proof of the birth’s registration and 1,250 rupees for immunizing the girl. Conditional on their daughter’s enrollment in school, the household would be eligible for a total of 3,500 rupees for completing primary school and 3,750 rupees for completing secondary school up to Grade 8. If parents can show that the girl has remained unmarried until her 18th birthday, the girl also receives an insurance maturity cover of 100,000 rupees. To put these payouts in perspective, average monthly household expenditures in Punjab’s treatment district Fatehgarh Sahib according to the 2007/08 National Sample Survey were 5,397 rupees, 5,094 rupees in rural areas, and 6,971 rupees in urban areas. Thus, registration alone was worth almost a month of household expenditures.
Dhanlakshmi Payouts with Conditions for Each Milestone
There are no direct monetary costs associated with meeting any milestone. Registration within 21 days after a child’s birth is free of charge, and children born in hospitals are typically automatically registered. There is no charge for immunizations under India’s Universal Immunization Program, and children can attend government school up to the eighth grade without having to pay tuition. Consequently, scheme payments can be applied to the indirect costs associated with such milestones, such as transportation expenses, or be used to subsidize a daughter’s general upbringing.6
B. Dhanlakshmi Implementation: Why We Focus on Punjab
Dhanlakshmi was introduced in 11 blocks (the next administrative level below the district) across seven Indian states: Andhra Pradesh, Bihar, Chhattisgarh, Jharkhand, Odisha, Punjab, and Uttar Pradesh. However, we focus exclusively on the scheme’s impact in Punjab for two reasons. First, our primary goal is to determine whether such an unrestricted and generous CCT can improve the sex ratio without significantly increasing fertility, and only Punjab had a severely skewed sex ratio prior to 2008. Second, Dhanlakshmi was much more effectively implemented in Punjab than in the other six states.
Prior to 2008, the district-level child sex ratio (CSR)—defined as the number of girls per 1,000 boys in the zero-to-six-year age group—was categorically different in Punjab. Based on the 2001 census, Fatehgarh Sahib’s CSR was 766, well below that of every other district where Dhanlakshmi was introduced and the lowest in all of India. In the other treated districts, the CSR was at or much closer to the expected biological parity. Further, Fatehgarh Sahib was explicitly chosen as the treated district, and Sirhind as the treated block, because of the high number of “missing girls” (Ministry of Women and Child Development 2015). Sirhind’s CSR in 2001 was 749, even lower than the district average.7
Sirhind accounted for the largest share of beneficiaries and budgeted allocations among the 11 Dhanlakshmi blocks and was the only treatment block that consistently recorded beneficiary enrollments. Awareness about registration sites in Punjab was among the highest of all treatment areas, partly because schoolteachers played a key role in information dissemination (Sekher and Ram 2015). Registration required relatively little time in Punjab. More than 60 percent of households reported it took less than a week, while in other states the majority of households waited more than four months (Sekher and Ram 2015). Existing state government schemes in Punjab were small and differed substantially in eligibility, so there was no confusion with other programs. The goals of Dhanlakshmi appear to have been well understood. In a survey, a significant share of potential beneficiaries in Punjab identified the scheme’s purpose as balancing the sex ratio (Sekher and Ram 2015). By contrast, there were substantial implementation challenges in the other six states, owing largely to Maoist violence.8
III. Research Design
A. Data and Basic Framework
Our research design is based largely on household data from the District Level Household and Facility Survey (DLHS) and the National Family Health Survey (NFHS) (IIPS 2010, 2017, undated).9 Both are large-scale surveys of health and fertility outcomes that are representative at the state level. They record all live births to women aged 15–49 for up to four years prior to the interview date. For each birth, the surveys report the sex, birth year, birth order, and whether it involved a single child, twins, or triplets. For each woman, the survey provides her age, education, caste, and religion, as well as her household’s wealth (measured by dwelling type) and location (whether urban or rural). We identify the years a woman was at risk of giving birth over the previous four years (essentially the years she was married), the birth(s) in a given year, and whether a birth was a girl.
Our primary focus is on the effects of Dhanlakshmi on girl births and fertility. The 2007/08 and 2012/13 rounds of the DLHS are the sources for information on children born before and after Dhanlakshmi was introduced.10 The program was announced in early 2008, but it did not go into effect until November of the same year. We therefore take 2004–2007 as our pre-treatment period and 2009–2012 as our post-treatment period.
Our research design compares the change in each outcome before and after the scheme was introduced in the treated area of Punjab with the changes in areas not covered by the scheme (Zimmermann, Cornwell, and Biswas 2023). In our main analysis, the treated area is the district of Fatehgarh Sahib and the other Punjabi districts comprise the control group. Focusing on Punjab has the advantage that state-specific policies, as well as cultural and socioeconomic factors, are held constant. Importantly, there were no confounding district-level interventions in Punjab between 2004 and 2012.
Using the household survey data, we formalize the difference-in-differences (DID) setup in terms of a regression model of the form:
1where yidt is a (girl) birth indicator for woman i in district d and year t, DL equals one for women residing in the treatment area, and Post equals one for observations drawn from the treatment period, Xidt is a vector of covariates, and αd and πt are district and year fixed effects. The coefficient of interest is δ. We estimate Equation 1 first without and then with covariates. We assess the robustness of our main findings to alternative control groups comprised of districts in states that are plausible counterparts to Punjab. We base our primary inference on standard errors that are clustered at the district level. However, in settings with only one treated cluster, there is a concern that cluster-robust (CR) standard errors will tend to overreject the null hypothesis. One commonly suggested solution is to compute standard errors using the wild cluster bootstrap. Therefore, we also report p-values using the restricted (Wild R) and unrestricted (Wild U) wild cluster bootstrap, along with p-values based on the CR standard errors.
Because Dhanlakshmi was discontinued in 2013, we explore whether the effects of the scheme persisted when enrollment ended. For this analysis, we combine the DLHS 2012/13 data with the 2015/16 round of the NFHS. In this case, we compare the changes in attitudes about desired pregnancy outcomes, girl births, and fertility from the last three birth years completely covered in the DLHS (2010–2012) with the first three covered in the NFHS (2014–2016) for both Fatehgarh Sahib and the other Punjabi districts.
Using the same DID design, we can also examine Dhanlakshmi’s impact on the post-birth outcomes its payouts target. We follow the cohorts born under the scheme through age seven to estimate its effects on immunizations and enrollment in primary school. Similarly, we estimate Dhanlakshmi’s effect on the educational persistence of girls born before Dhanlakshmi. Finally, we consider the possibility that Dhanlakshmi led to improvements in mothers’ pregnancy-related health outcomes.
B. Main Analysis Samples
The Punjab sample constructed from the 2007/08 and 2012/13 rounds of the DLHS comprises 36,768 woman–years, 6,701 unique mothers, and 10,678 live births. The treated district, Fatehgarh Sahib, contributes 403 mothers, 180 (223) from the pre (post)-treatment period. The other 19 control districts supply 6298 mothers, 3041 (3257) from the pre (post)-periods. The live-birth total breaks down to 619 in Fatehgarh Sahib, 248 (371) pre (post), and 10,059 in the rest of the state, 4524 (5535) pre (post). The left panel of Table 4 provides the summary statistics for the variables used in our analysis.
Summary Statistics for Birth and Fertility Samples, Punjab
The two outcomes of primary interest are whether a mother gives birth to a girl, conditional on giving birth (isgirl), and whether she gives birth at all (isbirth). Comparing the pre/post means for these two variables in the treatment and control districts offers a preview of our main findings. The girl share of births is roughly the same magnitude in Fatehgarh Sahib and the rest of Punjab during the pre-treatment period. However, it rises 4.5 percentage points in Fatehgarh Sahib—from 0.456 to 0.501—while it remains essentially unchanged elsewhere. Fertility is declining across Punjab during this period, but this decrease is slightly greater outside of Fatehgarh Sahib.
The covariates include mother’s age and completed grades of schooling, religious affiliation, and whether a woman is low caste, a member of a Scheduled Tribe, resides in a rural or urban area, and lives in an engineered house.11 The treatment and control groups differ little in terms of many key characteristics. On average, mothers in Fatehgarh Sahib and in the rest of Punjab are similar in age and level of schooling, and the share of rural households is roughly the same. At the same time, Fatehgarh Sahib has a higher share of Sikhs and low-caste households. It also appears slightly wealthier, as proxied by house construction type.
To estimate the post-program effects of Dhanlakshmi, we join the 2015/16 round of the NFHS with the 2012/13 round of the DLHS. The Punjab sample for this analysis includes 15,170 woman–years, 5,269 unique mothers, and 5,721 live births. The right panel of Table 4 presents the summary statistics, and the bottom two rows give the breakdown in mothers and births pre/post in Fatehgarh Sahib and the rest of Punjab. DLHS and NFHS are very similar data sets in terms of survey design and sampling method, and they are routinely combined in district-level empirical analyses. See Anukriti (2018) for a recent example. The sample size is larger in the DLHS because it is explicitly meant to be representative at the district level.
The girl share of births in Fatehgarh Sahib declines by only 1.2 percentage points in the post- program period, but remains 1.4 percentage points higher than the post-program girl share in the other districts. Consistent with the sample represented in the left panel, we see the same similarities in mother’s age and level of schooling as well as the share of rural households in treated and control groups. Likewise, we find the same religious affiliation and wealth differences.
To explore the impact of Dhanlakshmi on the post-birth outcomes of immunization and education, we use the DLHS 2007/08 and 2012/13, as well as the NFHS 2015/16. For immunization outcomes, we combine the DLHS 2007/08 and 2012/13 rounds and restrict the sample to girls who are at least two years old, the age by which children should be fully immunized according to India’s Universal Immunization Program (Corsi et al. 2009). For primary school enrollment, we link the DLHS 2012/13 and NFHS 2015/16 to follow the girls born under Dhanlakshmi (between 2009–2011) through ages five to seven, the ages at which children typically start school.12 We also draw on DLHS 2012/13 and NFHS 2015/16 to investigate whether Dhanlakshmi improved educational outcomes for 8–14 year-olds who were born before the scheme was implemented (2002–2007).
IV. Results
First, we present Dhanlakshmi’s effects on birth outcomes—girl births and fertility. Then we turn to the results for the post-birth outcomes of immunization, survival to primary school age and education. Finally, we consider whether Dhanlakshmi improved mothers’ health during pregnancy and birth.
A. Birth Outcomes
In this section, we discuss our main results on girl births and fertility and show how they vary with birth order and residential location. We also provide evidence that Dhanlakshmi’s impact on girl births persisted after the scheme was terminated.
1. Girl births and fertility
Table 5 reports the estimates of Dhanlakshmi’s effects on girl births and fertility from using our sample of mothers from the 2007/08 and 2012/13 DLHS waves. The simple DID coefficient estimate reported in Column 1 of Panel A suggests that Dhanlakshmi increased the likelihood of a girl birth by 4.5 percentage points in Fatehgarh Sahib during the treatment period. Adding district and year fixed effects, along with controls for birth order, mother’s age and education, and household’s caste, religion, residential location, and wealth (Column 2), causes the estimated effect to rise by about a percentage point. In each case, the estimated DID coefficient is statistically significant at the 1 percent level using the conventional cluster-robust (CR) standard errors.
Effects of Dhanlakshmi on Girl Births and Fertility
As emphasized by Anukriti (2018), it is reasonable to suppose that promoting girl births may generate higher overall fertility if sons and daughters are imperfect substitutes or if parents desire a minimum number of sons. Devi Rupak was designed to mitigate this trade-off, limiting benefits to households with either one child or two daughters and no sons and requiring sterilization. Anukriti (2018) shows these restrictions essentially counteract the impact of otherwise generous incentives to have daughters, leading to an unintended rise in the sex ratio.
While Dhanlakshmi improved the prospects for girl births, the lack of any countervailing incentives against higher fertility could potentially undermine its impact on the sex ratio. Panel B of Table 5 reports the estimated effects of Dhanlakshmi on fertility. Columns 1 and 2 repeat the same specifications as in Panel A and show a positive and statistically significant effect of Dhanlakshmi on births, although it is relatively small. The simple DID estimate is about half of a percentage point, and the estimate from the fully controlled model is about 0.9 percentage points. The coefficient estimate from the fully controlled model is significant at the 1 percent level (based on the CR standard errors), while the p-value of the simple DID estimate is 0.065.
Overall, the message obtained from the Punjab sample is that Dhanlakshmi led to a five percentage point increase in girl births with less than a one percentage point increase in fertility. However, as discussed in Section III, because there is a concern that CR standard errors tend to overreject the null hypothesis, we also provide inference based on the wild cluster bootstrap.13 For girl births (Panel A), Wild U rejects the null at close to the 5 percent level, and Wild R rejects at the 10 percent level in the fully controlled model. For fertility (Panel B), both bootstrap procedures fail to reject the null for fertility. We might view these p-values as roughly providing bounds on the true inference because Wild R will underreject, and Wild U will overreject when there is just one treated cluster (MacKinnon and Webb 2017).14
A separate concern with our analysis might be the existence of other girl-child CCTs operating in Punjab contemporaneously with Dhanlakshmi. As indicated in Table 2, Punjab had two such girl-child programs, Kanya Jagriti Jyoti Scheme and Balri Rakshak Yojana. Kanya Jagriti Jyoti had about 8,000 beneficiaries per year across the state, according to administrative sources, and there is no evidence of any material changes in the scheme during Dhanlakshmi (Sekher 2010). Balri Rakshak Yojana had few beneficiaries—only 306 between 2005 and 2010—and was discontinued in 2014. Overall, Sekher (2010) describes officials in the state of Punjab as generally uninterested in implementing other girl-child CCTs, explaining their low relative enrollment rates. In any event, these ongoing schemes can be held constant in our DID setup.
We also explore the robustness of the results produced by the Punjab sample by carrying out our analysis with control groups constructed from districts in other states. While the existence of ongoing, contemporaneous girl-child CCTs across India limits the pool of alternatives, we assemble three different groups from the states that were included in DLHS 2012/13. The first includes those states that had no girl-child CCT operating during the period of our analysis: Arunachal Pradesh, Kerala, Maharashtra, Manipur, Meghalaya, Mizoram, Nagaland, Sikkim, Tripura, and West Bengal. We label this group “None.” The second uses those states with a girl-child CCT but where the status of the CCT did not change over the analysis period: Haryana, Karnataka, and Tamil Nadu. We designate this the “No Changes” group. Clearly, the analysis carried out with the second group is conceptually different, as it compares Fatehgarh Sahib to districts that always had a girl-child CCT. While it is possible that any of the other schemes induced time-varying differences with the treatment area, we have no evidence of any. A potential problem with the “None” and “No Changes” groups is that they contain states that are culturally different from Punjab. Therefore, we construct a third group from states that are culturally more similar, especially in terms of son preference. These are the northwest states of Haryana and Himachal Pradesh, hence the “Northwest” label. This is consistent with the approach of Anukriti (2018) and Sinha and Yoong (2009), who use control groups comprising neighboring states on the basis of their similarity to the treated area.15
Columns 3–5 in Panels A and B of Table 5 present the results from the fully controlled model, using the three alternative control groups, “None,” “No Changes,” and “Northwest,” respectively. The estimated effects of Dhanlakshmi are remarkably robust across the three alternatives. The DID estimate for female births ranges from 4.7 to 5.4 percentage points; for fertility, the range is 0.3–0.8 percentage points. All the coefficient estimates are significant at the 1 percent level based on the CR standard errors.16 Both Wild U and Wild R p-values generally support the statistical significance of the girl-birth findings, while indicating less precision for the fertility estimates. Overall, the findings from the DLHS data strongly suggest that Dhanlakshmi led to more girl births, with only a slight rise in fertility.17
As an additional robustness check, we also explore whether the effects of Dhanlakshmi are evident in census data. As discussed in Section II, the scheme was actually implemented in Sirhind, one of five blocks in Fatehgarh Sahib. An advantage of the census data is that they are available at the block level, so we can examine the CCT’s effect on Sirhind directly. Specifically, we carry out a DID analysis on the share of girls among children aged zero to six and the size of the zero-to-six–year-old population, comparing Sirhind to the other Punjabi blocks in the 2001 and 2011 censuses. Our census sample is constructed from the primary census abstracts. Using the census administrative atlas, we reconcile all block boundary changes across the two years, creating a panel of 72 blocks.18 The controls in this specification include male and female literacy rates, labor force participation rates by gender, the rural population share, and the population shares belonging to Scheduled Castes or Scheduled Tribes. Standard errors are clustered at the block level.
The last column of Table 5 gives the census results. The DID estimate in Panel A is 0.0062 and significant at the 1 percent level, suggesting that Dhanlakshmi increased the share of girls in the zero-to-six-year-old population by about 0.62 percentage points. While the census estimate implies a smaller program effect, it is attenuated relative to the household survey result because the zero-to-six age group includes children born between 2005 and early 2011. However, Dhanlakshmi could have caused births only after 2008. At the same time, we find an imprecisely estimated negative impact of Dhanlakshmi on the zero-to-six-year-old population, offering further support for the conclusion that the introduction of Dhanlakshmi did not lead to a large fertility increase.19 As we demonstrate below, the number of additional births implied by the household results lies within the confidence interval of the census fertility results.20
We also use the block-level information in the census data to test for spillovers from the treatment area to surrounding control districts or blocks. In principle, we can effectively rule out direct spillovers in the form of payments to nonresident households because beneficiaries must produce a certificate of residence in Sirhind to qualify, and they lose coverage if they move away. Indirect influence of gender attitudes in surrounding areas is another matter. We investigate this possibility, varying the analysis in Column 6 of Table 5 in three ways: (i) dropping Sirhind and designating the other four Fatehgarh Sahib blocks as the treated area, (ii) using the three blocks outside of Fatehgarh Sahib contiguous to Sirhind as the treated area, and (iii) dropping all blocks contiguous to Sirhind. For the girl-share regression, cases (i) and (ii) deliver fairly precise zeroes. The DID coefficient estimates are 0.0077 and 0.0030, with p-values of 0.209 and 0.432. Case (iii) replicates the result in Column 6 of Table 5; the DID coefficient estimate remains 0.0062 and is statistically significant at the 1 percent level. For the zero-to-six population regression, cases (i) and (ii) produce highly imprecise estimated DID coefficients, while case (iii) (again) closely approximates the Table 5 finding.21
2. Birth order and residential location
Now we consider two important dimensions of heterogeneity in the impact of Dhanlakshmi, birth order and residential location. Traditionally, the sex ratio for second births is even more skewed. We test for birth-order effects by interacting the treatment variable with indicators for the second and higher-order children.22 Column 1 of Table 6 reports the results for the Punjab sample. Dhanlakshmi improved the prospect of a girl birth with the first child by 4.3 percentage points and an additional 5.8 percentage points with the second child, with p-values well below 0.05. The estimated coefficient of the interaction between the program and higher (more than two) birth-order indicators is roughly the same magnitude as the main effect, but negative and not statistically significant.
Dhanlakshmi Effects on Girl Births and Fertility by Birth Order and Urban/Rural Residency, DLHS 2007/08 and 2012/13
In India, sex-ratio imbalances are often more concentrated in urban areas, where there is better access to sex-selection technology and higher incomes to pay for it. However, in Fatehgarh Sahib, the pre-treatment sex-ratio imbalance was larger in rural areas, where more than two-thirds of the district population lives. Columns 2 and 3 show how the impacts of Dhanlakshmi on girl births and fertility vary by urban residency using the Punjab sample. We find the CCT’s effect on girl births is highly concentrated in rural areas, where it raised the likelihood of having a daughter by 7.4 percentage points, compared with less than a percentage point in urban areas. Both main and interaction coefficient estimates are significant at the 5 percent level. Consistent with these findings, the overall small, positive fertility effect is also concentrated in rural areas.
3. Son preference and post-program effects
Although Dhanlakshmi stopped enrolling new households in 2013, it is plausible that the girl-birth effects documented in Section IV.A.1 could have persisted to some degree through a change in attitudes. To probe this possibility, we first use the DLHS questions on desired pregnancy outcomes to estimate the effect of Dhanlakshmi on a mother’s stated preference for a boy child. Panel A of Table 7 presents the results from the Punjab sample (Columns 1–2) and alternative control group samples (Columns 3–5), in parallel with Table 5. Here, we also condition on the number of daughters from previous births. The simple DID estimate produced by the Punjab sample is −0.329 and significant at the 10 percent level; with controls it is −0.539 and significant at the 1 percent level. The implication is that Dhanlakshmi reduced the expressed desire for a boy child by 3.3–5.4 percentage points. The evidence from the alternative control groups is weaker, but it generally supports the claim that the scheme reduced the intensity of son preference. In addition, we find that the son-preference effect does not vary significantly with whether there is a daughter in the household. Panel B of Table 7 repeats the analysis of Panel A, adding an interaction with a “has daughter” indicator. While mothers with daughters are more likely to prefer a son (on the margin), they are not significantly more likely to say so under Dhanlakshmi.
Dhanlakshmi Effects on Son Preference, DLHS 2007/08 and 2012/13
Next, we directly estimate the post-program effects of the scheme, comparing changes in girl births and fertility from 2010–2012 with those from 2014–2016 using the Punjab sample. Table 8 reports the results. For girl births, the simple DID estimate is −0.030 and statistically significant at the 5 percent level. The DID estimate from the fully controlled model is similar in magnitude (−0.023), though less precise (with a p-value of 0.126). Taken together, these results imply that almost half of the overall effect on girl births reported in Table 5 persisted one to three years after Dhanlakshmi ended, which is consistent with the attitude change documented in Table 7. In Column 3, we see that the lasting impact lands entirely on the first child, where the prospect of having a daughter is higher by 7.3 percentage points. In contrast, the estimated post-program effect on fertility essentially reverses the small positive result shown in Table 5.23
Post-Program Effects of Dhanlakshmi on Girl Births and Fertility, DLHS 12/13 and NFHS 15/16
4. Discussion: Effect sizes and treatment intensity
The findings reported in Tables 5–8 clearly point to a Dhanlakshmi-led improvement in Sirhind’s child sex ratio. The scheme’s impact is sufficiently strong to be picked up at the district level in the household survey data, even though Sirhind is only one of five blocks in Fatehgarh Sahib. Here, we provide the results from a simple back-of-the-envelope calculation of the number of girls born in Sirhind because of Dhanlakshmi.24
Based on the simple comparison of means between treatment and control groups before and after the implementation of Dhanlakshmi from Table 4, the program led to one extra birth in Fatehgarh Sahib per year (an increase of 1.64 percent), but to three additional girl births. All of these girls were born due to the improvement in the sex ratio rather than the small fertility increase. Using 2011 district-level census information to provide a rough translation from the household-survey sample numbers to Fatehgarh Sahib’s relevant population age group, this implies an estimated 584 additional children and 1,752 additional girls were born in the four years of Dhanlakshmi.25 If all of these births came from Sirhind, then Dhanlakshmi accounted for 4.85 percent of all births and 32.13 percent of all girl births.26
The high treatment intensity implied by these calculations is supported by the descriptive assessments of Dhanlakshmi compiled in Sekher (2010) and Sekher and Ram (2015). When surveying Sirhind residents about Dhanlakshmi, Sekher and Ram (2015) reported that finding nonbeneficiary households was a challenge because the scheme had been implemented so well. In a number of villages, they found that all eligible girls had been enrolled in the program. They document continued consistent communication about Dhanlakshmi in Sirhind through print media, electronic media, and posters, with an integral role for health center workers and local government officials. As a summary illustration of Dhanlakshmi’s impact on Sirhind, Sekher and Ram (2015) offer a quote from one survey respondent: “Dhanlakshmi scheme benefits the girls in a major way. It would help reduce the level of female foeticides. The scheme would help to improve equality in treatment between boys and girls.”27
B. Post-Birth Outcomes
Dhanlakshmi not only promoted girl births, but their health and education as well. The scheme paid 1,250 rupees if the child received the full complement of immunizations scheduled for the first 24 months and 1,000 rupees if she enrolled in primary school. In addition, there were annual payments of 500 rupees for each primary school grade completed, another 1,500 rupees for enrolling in middle school, and 750 rupees for each middle school grade completed. Using the DLHS 2012/13 and NFHS 2015/16, we follow the cohorts born under the Dhanlakshmi scheme through age seven, which allows us to examine the program’s effects on immunization, survival to primary school age, and primary school enrollment. It is possible that the program’s overall encouragement of postnatal care through its payouts to immunization led a higher incidence of valuable uncompensated activities as well. We explore this possibility by also estimating Dhanlakshmi’s effect on breastfeeding.28 Table 9 reports our findings for immunization, breastfeeding, survival, and primary school enrollment, comparing children born under Dhanlakshmi with birth cohorts whose outcomes could not be influenced by the scheme. Each column coincides with the fully controlled specification using data from Punjab (as in Column 2 of Table 5).
Dhanlakshmi Effects on Immunization, Survival and School Attendance, DLHS 2007/08 2012/13 and NHFS 2015/16
1. Immunization and breastfeeding
According to Corsi et al. (2009), India’s Universal Immunization Program during our study period included one dose of BCG (tuberculosis), three doses of OPV (polio) and DPT (diptheria, pertussis and tetanus), and one dose of measles. We bundle these vaccinations into three “rounds,” corresponding roughly to three distinct phases in their age-appropriate classification: BCG + DPT1 (Round 1), DPT 2–3 + OPV 1–3 (Round 2) and measles (Round 3). Round-1 vaccinations are supposed to be given at or shortly after birth and show the largest gender gap in India (Corsi et al. 2009). Round-2 vaccinations would occur within a couple of months after birth, with Round-3’s being given at nine months or later.
While the Dhanlakshmi program specifies that children should be fully vaccinated by the time they are 24 months old, the age ranges for the individual immunization milestones do not directly map to these recommended age-appropriate immunization rounds. We therefore restrict the sample to girls who are at least 24 months old and test the impact of the CCT on vaccination rates in that sample. We find that Dhanlakshmi increased the probability a girl received her Round-1 vaccinations by 7.1 percentage points (Column 1), and the result is statistically significant at slightly above the 1 percent level. Columns 2 and 3 indicate that the scheme also improved second- and third-round vaccination prospects by about five percentage points, with p-values below 0.05. These effects are large given that baseline vaccination rates are around 40 percent in the sample.29
Using the immunization sample, we find evidence that Dhanlakshmi also increased the likelihood a girl child is breastfed. Column 4 reports this result, which suggests that the probability of breastfeeding rose 2.4 percentage points with a p-value of 0.051.
2. Survival
Where son preference is strong, there is commonly a gender disparity in postnatal mortality because some households substantially underinvest in their daughters’ care. Guilmoto et al. (2018) calculate the excess female under-five mortality rates (U5MR) for each Indian state for the 2000–2005 period, just before the introduction of Dhanlakshmi. These U5MR estimates capture the difference between observed and expected (biological) female mortality rates. Guilmoto et al. (2018) report a U5MR in Punjab of 14.3 per 1,000 live births. Many of these deaths are believed to be the direct result of low vaccination rates and parental neglect. The staggered Dhanlakshmi immunization and education payouts are clearly designed to combat this discrimination by incentivizing parents to continuously make investments in their daughters, which should improve girls’ survival chances.
We explore the effectiveness of these payouts in increasing postnatal survival by repeating the girl-birth analysis for primary-school-age children before and after the scheme was introduced. Specifically, we use the NFHS 2015/16 to follow children born during Dhanlakshmi (2009– 2011) until they reach primary school age (five to seven years old). Likewise, we use the DLHS 2012/13 to follow children born early enough (2000–2002) that Dhanlakshmi could not affect their post-birth outcomes. Column 5 reports the estimated effect of Dhanlakshmi on the probability that a primary-school-age child is a girl. We find that the girl share of this age group increased by statistically significant 4.5 percentage points under the scheme, which is more than 80 percent of the girl-birth effect given in Table 5. Consistent with the immunization and breastfeeding findings, it appears that Dhanlakshmi not only led to a rise in girl births, but also aided their postnatal survival.
3. Education
Next, we examine the role Dhanlakshmi played in promoting school enrollment. Both the DLHS 2012/13 and NFHS 2015/16 ask each household whether their school-age children are in school at the time of the survey. Combining these data similarly to the survival sample, we estimate the effect of the scheme on the probability a five-to-seven-year-old girl was enrolled in school (was “in school now” (isn) according to the survey). Column 6 reports the result, which suggests Dhanlakshmi increased the likelihood a girl started primary school by four percentage points. The estimated education effect is significant at the 1 percent level.
An important feature of Dhanlakshmi’s design was the ability to enroll older girls who were born before the program was introduced. While those girls could not benefit retroactively from payouts for earlier milestones, they were eligible for any remaining age-appropriate payments. Most of these financial incentives center on school attendance and grade completion through the end of middle school. We investigate Dhanlakshmi’s impact on these older girls, focusing on 8–14-year-olds in the NFHS 2015/16 and their counterparts in the DLHS 2007/08. Children who started school on time should have received the last education payout for eighth-grade attendance (the final year of middle school) at around age 14. Compulsory schooling in India covers 6–14-year-olds, although many girls drop out earlier.
Table 10 reports the estimated effects of Dhanlakshmi on isn and completed years of schooling (yos). The scheme apparently had no impact on the enrollment of older girls (Column 1), but it did raise their average years of schooling (Column 2). The DID estimate of the yos effect, which is statistically significant at the 1 percent level, indicates an increase of roughly a third of a year. These findings suggest that enrollment patterns for this age group were set before Dhanlakshmi started, but the scheme successfully encouraged the persistence of those who were already enrolled.
Dhanlakshmi Effects on the Education Outcomes of Older Girls, DLHS 2012/13 and NFHS 2015/16
C. Mother Effects
Milazzo (2018) shows that the morbidity and mortality of adult women in India are partly linked to son preference. Taking Milazzo’s findings together with our evidence that Dhanlakshmi reduced son preference, it is natural to ask whether the scheme had beneficial effects on maternal health. The DLHS allows us to address this question through its broad coverage of pregnancy, delivery, and post-delivery treatments and conditions.
As a final exercise, we use our Punjab sample to investigate the extent to which Dhanlakshmi affected them. For most of the surveyed conditions, all but 50 of the 6701 mothers in the Punjab sample are represented. Only in the case of iron supplements would there be a concern about survey nonresponse. Tables 11 and 12 present our findings. Table 11 is concerned with pregnancy treatments (iron supplements and tetanus shots) and conditions (ranging from swelling of hands, feet, and face to vaginal discharge) surveyed by the DLHS, while Table 12 covers a range of delivery and post-delivery conditions (including labor problems, high blood pressure, abdominal pain, and severe headache), along with indicators of whether delivery was normal and took place at a medical facility. The treatments are recommended, and each condition is defined in a manner such that a reduction amounts to an improvement in maternal health.
Dhanlakshmi Effects on Prenatal and Pregnancy Outcomes, DLHS 2007/08 and 2012/2013
Program Effect on Delivery and Post-Delivery Care, DLHS 2007/08 and 2012/13
The obvious takeaway is that Dhanlakshmi improved maternal health in almost every dimension we examine. Dhanlakshmi increased mothers’ use of iron supplements and acceptance of tetanus shots. We find statistically significant reductions in the likelihood of ten of the 13 negative pregnancy conditions, with half of the estimated effect sizes exceeding five percentage points. Similarly, we find statistically significant declines in the probability of eight of the 12 adverse delivery and post-delivery outcomes. The impact of Dhanlakshmi on high blood pressure, excessive bleeding, prolonged labor, and obstructed labor is especially noteworthy, as these conditions account for almost two-thirds of all maternal deaths in India.30 At the same time, we find no effect of Dhanlakshmi on normal delivery and evidence that the scheme decreased the likelihood of delivery at a medical center. Given the overall positive impact of Dhanlakshmi on maternal health, the medical center result might reflect a reduced need for the services provided by such a facility.
V. Conclusion
In the early 1990s, Amartya Sen coined the term “missing women” to describe the large sex-ratio imbalance between men and women in some countries (Sen 2010). Since then, the problem has received high levels of attention by researchers, practitioners, and policymakers. Despite the scale, salience, and urgency of skewed sex ratios in countries such as India, one of the most common policy tools meant to combat son preference—conditional cash transfers—has hardly been studied. India’s earliest girl-child CCTs were introduced in 1992, and to this day Indian governments have implemented more than 20 such programs. Nevertheless, little is known about their effectiveness or best design practices.
In this paper, we study the impact of Dhanlakshmi, one of the most liberal and generous of the Indian CCTs. Using a difference-in-differences empirical strategy, we show that the scheme had a substantial positive effect on girl births, with only a slight increase in fertility. Thus, most of the sex-ratio improvement came from parents substituting daughters for sons. Some parents who would have otherwise pursued abortion or post-birth feticide apparently chose to have daughters because of the scheme’s monetary incentives. We also show that half of the girl-birth effect persisted even after the program was discontinued, and this persistence coincided with a reduction in self-reported son preference.
Dhanlakshmi not only promoted girl births, but encouraged post-birth investments by providing payments for reaching immunization and education milestones. Girls born under the scheme were more likely to receive their scheduled immunizations, survive to primary school age, and enroll in primary school. These benefits were not limited to girls born under Dhanlakshmi, but accrued to girls who were alive when the program was implemented as well. Dhanlakshmi’s incentives for girl births and post-birth investments also spilled over to mothers in the form of improved health during pregnancy and childbirth.
A key implication of these findings is that in India CCT design can focus squarely on promoting girl births and post-birth investments in daughters and forgo the fertility restrictions and sterilization requirements widely adopted by previous schemes. The limits on fertility are perhaps understandable in the early CCTs introduced in the 1990s, when the national fertility rate was still over 3.5. However, the fertility rate was falling at the time, and it has continued to drop, down to 2.0 as of 2022 based on the latest NFHS data. The point is that perhaps the fertility limits were largely inframarginal to other forces driving family size. As Anukriti (2018) suggests, including them may have undermined the main goal of improving the sex ratio. In any event, most girl-child CCTs implemented since the late-2000s have dropped the most severe limits, a move our results support.
Acknowledgments
The authors thank Vanisha Kudumuri, Monthida Napier, Sarah Pool, and Marleena Tamminen for excellent research assistance. They also thank S. Anukriti and participants at the Annual Conference on Economic Growth and Development (Delhi), the Bangladesh Institute of Development Studies Conference on Development (Dhaka), and the Florida International University seminar series for valuable feedback and suggestions. Lakshmi is the Hindu goddess of wealth and prosperity. All authors have no relevant or material interests that relate to the research described in this paper. This paper uses proprietary DLHS survey data, managed by the International Institute for Population Sciences (IIPS). Data can be obtained by filing a request directly with IIPS (http://rchiips.org/pdf/data_request_form.pdf). The NFHS survey data is open access and can be downloaded from https://dhsprogram.com/methodology/survey/survey-display-355.cfm. Census data are available at https://censusindia.gov.in/census.website/data/census-tables. The code for creating the analysis samples and replicating the results of the paper is available at https://www.openicpsr.org/openicpsr/project/186862/version/V1/view.
Footnotes
↵1. CCTs listed without end dates were still active during our sample period, according to Sekher (2010).
↵2. Trust in the payout promise is crucial. One of the earliest programs, Rajasthan’s Rajalakshmi scheme, was popular among households in Rajasthan, and some other programs, such as Haryana’s Apni Beti Apna Dhan, were inspired by it. However, Rajalakshmi had to be terminated in 2000 because the insurance company could not sustain the high promised interest rate (14–17 percent per year). This experience is said to have undermined the appeal of similar schemes. Rajasthan has not implemented another girl-child CCT since, even though the state has one of the most skewed child sex ratios in India. Because of this structure, most programs only enroll girls shortly after birth or within a window of three to five years after birth.
↵3. Sinha and Yoong (2009) study a similarly restrictive program in Haryana from the mid-1990s, Apni Beti Apna Dhan, which requires BPL status and limits the number of girl beneficiaries per household.
↵4. In other contexts, just having a daughter or sister has been shown to shape gender norms. See, for example, Washington (2008) and Healy and Malhotra (2013).
↵5. See Online Appendix A for details on Dhanlakshmi’s discontinuation.
↵6. See Online Appendix B.1 for details on the scheme’s approach to beneficiary recruitment and enrollment.
↵7. See Online Appendix B.2 for a fuller account of Punjab’s skewed CSR and the ways it is distinct among the Dhanlakshmi states.
↵8. See Online Appendix B.3.
↵9. The India Human Development Survey (IHDS) 2011/12 cannot be used for the analysis since it contains only 39 observations from the district of Fatehgarh Sahib. This is before any additional data restrictions, such as limiting the sample to potential mothers.
↵10. DLHS 2012/13 is publicly available from the Indian Ministry of Family Health and Welfare website. The 2007/08 round of the survey is available on request from IIPS (Indian Institute of Population Studies).
↵11. We use the house construction type as a proxy for wealth because the DLHS does not collect wealth or income information.
↵12. Online Appendix Tables D.5 and D.6 provide the summary statistics for these samples.
↵13. A more conservative way of dealing with intracluster correlation is to collapse the analysis to the district level (Bertrand, Duflo, and Mullainathan 2004). The results in Table 5 are fully robust to this exercise. See Online Appendix C.2 for the details pertaining to this analysis.
↵14. Cluster size also matters to the performance of these bootstrap procedures. Both underreject when cluster sizes vary relative to the equal cluster-sizes case, although the impact of cluster size on rejection frequencies is smaller for Wild U than Wild R (MacKinnon and Webb 2020). Further, the range of underrejection for Wild R is narrower when the largest cluster(s) are treated (MacKinnon and Webb 2017), which is the case here. Fatehgarh Sahib is the third largest cluster overall and the second largest in the post-treatment period. Taken together, these results suggests that the “bounds” provided by the Wild U and Wild R p-values can be viewed as conservative.
↵15. Online Appendix Table D.3 provides the summary statistics for the alternative control groups.
↵16. Robustness across alternative control groups is a general theme for all of our analysis. Online Appendix D.4 reports these results for the rest of the analysis in Sections IV.A and IV.B.
↵17. The pattern for the bootstrap inference we observe in Table 5 is evident in almost all of the DLHS-based program-effect estimates. Therefore, in the interest of readability, we relegate the remaining bootstrap results to Online Appendix D.5.
↵18. See Online Appendix Table D.4 for the summary statistics.
↵19. Replicating the analysis in Column 6 for the alternative control groups produces statistically significant girl-share effect estimates of very similar magnitudes. The zero-to-six-year-old population results are more mixed in magnitude and precision. See Online Appendix Table D.10.
↵20. The difference between the estimated effects across the two samples is also affected by the implicit weighting scheme. The DLHS is representative at the district level, so more women from larger districts are included in the analysis sample. The block-level regression using the census data weights all blocks equally because there is one observation for each block. Unfortunately, information on the number of women of reproductive age is not available in the 2001 census at this disaggregated level, so we cannot reweight our observations. Collapsing the household survey data to the district and weighting all districts equally produces qualitatively the same results. See Online Appendix Table D.7.
↵21. See Online Appendix Table D.8.
↵22. About 50 percent of births in our sample are first births, followed by 35 percent and 15 percent at birth orders two and three or higher, respectively.
↵23. We also estimated the post-program effect by comparing pre-program births from DLHS 2007/08 with post-program births from NFHS 2015/16. The results are qualitatively and quantitatively very similar. See Online Appendix Table D.9.
↵24. For more details, see Online Appendix C.1.
↵25. Keeping in mind that the census was held in 2011, while the household survey data covers births up to 2012, the corresponding number of additional births is 438. This figure falls inside the 95 percent confidence interval of the DID estimate in Column (6) of Table 5: [−3,680.39, 468.95].
↵26. The evidence presented in Section IV.A.1 for no spillovers outside of Sirhind supports this assumption.
↵27. Online Appendix C.3 describes their survey and summarizes responses on gender attitudes.
↵28. Given the Dhanlakshmi’s payouts, it would be desirable to show their first-stage effects on household or medical expenditures. Unfortunately, neither the DLHS nor NFHS collects expenditure data. The NSS does but contains many fewer births and less granular birth information. Nevertheless, we can use three so-called “thick” rounds—NSS 2007/08, 2009/10, and 2011/12—that provide a large enough sample to potentially identify first-stage expenditure effects. We find that Dhanlakshmi led to higher household expenditures in general, and medical expenditures in particular, for households with newborn girls, consistent with the estimated post-birth effects documented in this section. For households with older girls, medical expenditures fell, as one might expect if early spending and higher immunization and breastfeeding rates improved infant health.
↵29. In this case, the bootstrap inference might lead to more circumspection. See Columns 1–3 in Online Appendix Table D.16. Nevertheless, the effect magnitudes remain economically significant.
↵30. See https://www.unicef.org/india/what-we-do/maternal-health (accessed December 19, 2024).
- Received October 2021.
- Accepted October 2022.
This open access article is distributed under the terms of the CC-BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0) and is freely available online at: https://jhr.uwpress.org.
Jump to section
Related Articles
Cited By...
- No citing articles found.