Financial Incentives and Fertility Choices: Evidence from India

Abstract

This article estimates the impact of a maternity cash transfer programme, Indira Gandhi Matritva Sahyog Yojana, implemented in India since 2011, on fertility choices. Since the scheme restricts the benefits to the first two births, we ask whether there is an impact on the likelihood of birth post implementation and if this behaviour is driven by son preference. We also test whether the scheme has affected the likelihood of female births. Our results give evidence in support of ‘stopping rule’ and we find that treated households reduce births only when their first two births are sons. However, we do not find any statistically significant impact of the likelihood of a female birth. We further find that post the implementation of the scheme, mothers and children are less likely to get better care for higher order births.

JEL Codes: J13, J18

Keywords

Indira Gandhi Matritva Sahyog Yojana conditional cash transfers son preference fertility sex ratio

I. Introduction

Can financial incentives alter fertility choices and the likelihood of a female being born in a household? We ask this question in the context of Indira Gandhi Matritva Sahyog Yojana (IGMSY), globally the largest maternity and child benefit scheme, introduced in India since 2010–2011. The scheme provides income support to mothers in the period extending from the last trimester of pregnancy through the first 6 months of the child’s life, for the first two live births. Beneficiaries would get direct cash transfer if they followed the conditions of the scheme concerning registration of pregnancy, birth, ante natal visits, vaccination of the child, breast feeding among others. However, all benefits of the scheme are limited to the first two births for a woman. Focus of the scheme on first two births might act as a signal to parents and prevent them from planning more than two children. One can also argue that choices related to fertility behaviour, particularly in a country like India are deep rooted in its social and cultural norms and financial incentives might not have any impact in influencing the same.

It is, therefore, worth investigating whether transmission of any such fertility influencing signal is dependent on underlying social norms driven by a preference for sons. If that is so, then, whether the household stops childbirth after two children could be dependent on the gender of the first two births. This follows from studies like Clark (2000) which find that in India smaller families have higher proportion of sons compared to larger families. There could be another unintended impact of schemes like IGMSY which require expecting mothers to undertake at least three ante natal check-ups. Such a criteria also gives access to prenatal sex determination technology (Bhalotra & Cochrane (2010), Javadekar & Saxena (2019)).

Against this background, the objective of the article is to test the impact of IGMSY on fertility choices, likelihood of a female birth and impact (if any) on higher order births in India, post 2011. Prior to IGMSY, several schemes implemented by State and central governments have covered maternal and neonatal child health care, sex ratio at birth such as Integrated Child Development Services (ICDS) since 1975, Devi Rupak in Haryana since 2002, Janani Suraksha Yojana (JSY) since 2005, Ladli Laxmi Yojana (LLY) in Madhya Pradesh since 2006, etc. among many others. The impact of such schemes have been explored in studies, such as Anukriti (2018), Javadekar and Saxena (2019) and Jain (2019).

With regards to impact of IGMSY, Ghosh and Kochhar (2018) find significant improvement in health outcomes of children, and increased birth gaps between two consecutive births in Bihar. von Haaren and Klonner (2020) find positive effects on long-term health care utilisation and infant immunisation. However, to the best of our knowledge, this is the first study which explores the unintended impact of IGMSY—on likelihood of births, sex ratio and maternal and neo-natal care for higher order births. Specifically, we ask three questions—(a) did IGMSY reduce the probability of women having a child post 2011 and did son preference have any role in that, (b) did the likelihood of female birth change after the scheme and once again did gender of the first two births have any role and (c) by focusing on first two births, did the design of the scheme in any way penalise the third child?

We answer these questions through a Difference in Difference (DID) framework using data from the National Family Health Survey-4 (NFHS-4) conducted in 2015–2016. IGMSY was first implemented as a pilot in December 2010 in 52 districts (Annexure 1) which were selected on the basis of parameters related to antenatal care provided to new births. It was expanded across the country in 2017. The identification strategy that we employ in this study exploits the staggered implementation of the scheme. Women who were not exposed to any scheme targeting maternity behaviour earlier and did not reside in the 52 IGMSY districts form our control group while those who were not beneficiaries of any pre-existing scheme and reside in those 52 districts form our treatment group. The analysis has been done on treatment and control districts matched on the basis of the indicators which were used to choose the treatment districts.

All births after 2011 in treatment districts have been defined as treated on and before 2011 as control. Since the scheme was implemented from December 2010, it is plausible that some beneficiaries get excluded from our empirical analysis because of our definition of pre and post treatment.¹ Given the data constraints, we are unable to provide causal estimates of increase in bias towards higher order child post IGMSY. This is because details on ante-natal and maternal care for births before 2010 are not available, and so, we are unable to estimate pre-treatment trends in treatment and control districts.

Our results broadly show that impact of the scheme on likelihood of birth is dependent on gender composition of the first two children. We find that there has been a decline in likelihood of giving birth only for the women who have two sons first. There is no change in the impact for women with any other child composition. However, we find no statistically significant impact on the probability of female birth for only those women who have two daughters. Finally, with regard to the impact on penalising of third and higher order births, although, we cannot provide causal evidence, our results do suggest that the scheme by incentivising pre-natal and maternal care for the first two children has penalised the third child in terms of ante-natal care given to him/her. We find that a scheme targeting the number of births could have unintended negative consequences and our results show that policies have to be designed with due regard to existing social norms and practices.

The article is organised as follows: We discuss the scheme in detail and conceptual pathways in Section II. Data and empirical framework are discussed in Section III. We discuss our results in Section IV, falsification tests in Section V and conclude in Section VI.

II. IGMSY Scheme and Conceptual Pathway

IGMSY is a maternity benefit scheme which was initially launched in 52 districts in India in December 2010 by the Ministry of Women and Child Development, Government of India. It is a conditional cash transfer scheme for pregnant and lactating mothers of 19 years of age or above for the first two live births. The objectives of the scheme include promoting appropriate prenatal/antenatal and postnatal practices, care and institutional service utilisation during pregnancy, delivery and lactation. The idea underlying the cash transfer is to compensate for being out of the job market. Eligible beneficiaries cannot be employees in the organised sector as they are entitled for a paid maternity leave. The cash benefits are given in three tranches once she fulfils the criteria for the payment (Table 1).

Table 1.

Eligibility Conditions for the Scheme.

	Instalment 1	Instalment 2	Instalment 3
Timing	Received at the end of second trimester	Received after 3 months of delivery	Infant is 6 months old
Cash benefits	₹1,500/-	₹1,500/-	₹1,000/-
Eligibility requirements	Pregnancy registered within 4 months	Childbirth registered	Child exclusively breastfed for 6 months
	Received at least one antenatal check (optimal 3)	Polio & BCG vaccination	Child introduced to complementary foods after 6 months
	Received IFA tablet	Polio & DPT-1 vaccination	Polio & DPT-3 vaccination
	At least 1 TT shot (optimal 2)	Polio & DPT-2 vaccination	Child weighed twice b/w the age of 3 and 6 months
	At least 1 counselling session at AWC	Child weighed at least twice	Mother attended at least 4 IYCF sessions at AWC/Home
		At least 2 IYCF sessions at AWC/home

Source: Ministry of Women and Child Development.

Fifty-two districts were identified on the basis of a set of characteristics—percent literate female population; mothers registered in the first trimester when pregnant with last birth; mothers who had at least three ante-natal care visits during the last pregnancy; institutional births; children fully immunised and children breastfed within one hour of birth. An index was computed based on these six indicators and the entire country was divided into three groups—high, medium and low performing districts. Eleven districts each were randomly selected from high and low performing districts, 26 from medium performing districts and four from union territories. The condition was that two districts should be selected from each major state. Anganwadi workers were responsible for registering pregnant and lactating women and they would maintain a record if women were following all the conditions as required by the scheme. They would pass on the record to higher officials and finally the district office would transfer the amount to the woman’s savings account.

It was merged with the Food Security Act in 2013 and the cash transfer was increased to ‘6,000 and was paid in two rather than three instalments–one at 6 months of pregnancy and one after 6 months of delivery. Conditions of eligibility remained the same. In 2017, the programme was renamed Pradhan Mantri Matritva Vandana Yojana and expanded to all districts of India with a cash transfer of ‘5,000 per woman paid in three instalments, two during pregnancy and one conditional on fulfilling the conditions in the antenatal and post-natal period.

An important dimension here is related to the period before IGMSY. The Government of India had launched JSY in 2005 which was once again a conditional cash transfer programme. Women were given financial incentives on the condition of institutional delivery. It was implemented across the country, but the nature of the scheme was different across different groups of states. The country was divided in two groups of states—low performing states (LPS) and high Performing States (HPS). States like Uttar Pradesh, Uttarakhand, Bihar, Jharkhand, Madhya Pradesh, Chhattisgarh, Assam, Rajasthan, Orissa and Jammu & Kashmir with low level of institutional delivery were LPS while other states were HPS. In LPS, all pregnant women were eligible, and the benefits were paid regardless of whether women delivered in a government hospital or in a private accredited health centre. For HPS, women who held BPL cards or belonged to a scheduled caste (SC) or scheduled tribe (ST) were eligible for the scheme. Therefore, women who were not SC and ST, did not have a BPL card and resided in one of the HPS states did not receive any benefit from JSY.

Conceptual Pathway

The primary objective of the scheme was to encourage better maternal and neo-natal practices among women in India. As discussed earlier, it is likely that women would consider this scheme as a signal and restrict the number of births she would have in her childbearing years. However, the decision on restricting/altering fertility decisions is based on several factors in a country like India. Gender preferences play a key role in determining the desired and achieved family size (Anukriti et al., 2016, 2018; Clark, 2000; Javadekar & Saxena, 2019; Jayachandran & Pande, 2017, Jensen, 2012; Rosenblum, 2013). Birth of girls lead parents to increase their fertility in the hope of having more number of sons in subsequent births (Clark, 2000).

Data from NFHS-4 (2015–2016) shows that, on an average, the number of children born to a woman between the ages 15 and 49 is lower for households with first born sons compared to those with first daughters and gives evidence of stopping rule (Table 2). Given social norms related to son preference, it is possible that the extent of the impact of the scheme would be dependent on the extent of gender preference and gender composition of children in the households.

Table 2.

Number of Children and Gender Composition of Children.

Child Gender	1^st & 2^nd Daughter	1^st & 2^nd Son	1^st Daughter	1^st Son
Mean # children	2.76	2.29	2.02	1.77
Mean # (rest of the sample)	1.74	1.81	1.77	2.02
Diff—t-test	−1.02***	−0.47***	−0.255***	0.255***

Source: Authors’ calculations using NFHS-2015–2016 data.

Note: ***p < 0.01

Further, the scheme gives access to prenatal services which could have increased exposure of households to sex determination techniques and might have detrimental effects on sex selection and thereby influencing gender composition of the children. Studies (Anukriti, 2018; Bhalotra & Cochrane, 2010) have found that the advent of foetal sex determination technology increased the likelihood of households with strong preference for boys to abort their foetus if it was a girl. Because of mandatory antenatal visits and registration of pregnancy, it is likely that households have higher scope for sex determination and undergoing sex selective abortions. However, it has also become more difficult for them to undergo sex selective abortions (Pre-Conception and Pre-Natal Diagnostic Techniques (PCPNDT) Act 1994 (Amended 2003)). Therefore, a priori it is difficult to predict the direction of the impact of the scheme on births and sex ratio of children born after 2011.

The third question we ask in this article is, given that the scheme provisions only for the first two births, one of the unintended impacts of the scheme might be lower ante-natal care for higher order births. Higher order births in India have been found to be stunted and have higher fatality and with poorer health outcomes compared to that of Sub-Saharan Africa (Jayachandran & Pande, 2017). Our hypothesis is that schemes like IGMSY might not have any impact on fertility choices given the norm of son preference, and low health outcomes for higher order births might be accentuated because of schemes targeting only the first two births.

III. Data and Empirical framework

We use data from the seventh round of the Demographic and Health Survey (IIPS and IFG 2018) conducted in India and commonly known as NFHS-4 for the analysis. NFHS is a nationally representative survey which collects detailed information on woman (ever married 15–49 years old), antenatal and postnatal care for all children born to her in the last 5 years. We construct a woman year panel using NFHS-4 data and drop all the women who became mothers before 2000. This has been done to ensure no significant age difference between the beneficiaries and non-beneficiaries.

Our objective is to arrive at a control and treatment group which was not subject to maternity benefit schemes or schemes targeting their fertility behaviour. We drop all women from our sample who were already exposed to JSY since 2005, and those residing in urban areas since the scheme targeted more women in rural areas and chances of availability of technology for sex selective abortions would have been higher in urban areas. We also drop women residing from a few states which already had similar existing schemes, such as Madhya Pradesh, Maharashtra, Odisha and Tamil Nadu. We match treated districts in the high performing states with control districts using nearest neighbour matching. Matching is done on the basis of a score based on the measures used to select 52 districts for the treatment. Table 3 gives a comparison of treatment and control districts on matched sub-sample. We use data on children born before 2012 for computing the index. This prevents our matching algorithm from having post-intervention spillovers.

Table 3.

Comparison of Mean for Treatment and Control Districts.

Variable	Treated-mean	Control-mean	% Bias	t-Test	t-Testp-value	V(T)/V(C)
Share literate	0.773	0.768	4	0.14	0.892	0.5
Share registration first trimester	0.443	0.433	5.9	0.24	0.812	1.47
Share ante natal visit (3)	0.455	0.438	10.1	0.41	0.686	1.31
Share institutional delivery	0.818	0.776	19.7	0.68	0.5	0.64
Share breastfed	0.269	0.263	4	0.16	0.876	1.79
Share BCG	0.079	0.069	19.1	0.63	0.533	0.98
Share polio3	0.057	0.081	−55	−1.54	0.129	0.42*
Share DPT3	0.061	0.081	−47.6	−1.35	0.182	0.48
Share measles	0.020	0.022	−5.6	−0.19	0.85	0.57

Source: Authors’ calculations from data using NFHS-4 (2015–2016).

We use a Difference in Difference (DID) framework to estimate the impact of IGMSY on the probability of women giving a birth after 2011 in treatment districts compared to control districts. The DID framework is often used in literature to control for time invariant unobserved heterogeneity (Angrist & Pischke, 2009, 2014; Gertler et al., 2016; Khandker et al., 2009; Lechner, 2011). The first equation we estimate is:

y_{i t d} = α_{0} + α_{1} {igmsy}_{i d} + δ {post}_{2001 - 2016, t} + β ({igmsy}_{i d} * {post}_{t}) + γ X_{i t} + θ {year}_{i t} + μ D_{d} + ε_{i t d}

(1)

Our outcome variable $y_{i t d}$ is a binary indicator variable and is equal to 1 if woman residing in district ‘d’ gave birth in year ‘t’ and is zero if not. In our second specification, our outcome variable is equal to one if the child born is female and zero if the child born is male. $p o s t$ is equal to 1 if year is greater than 2011 and zero if less than 2011. $i g m s y_{i d}$ is an indicator variable which is equal to one if the district in which the woman resides is a treatment district and zero otherwise. Our main explanatory variable is $(i g m s y_{i d} * {post}_{t})$ and coefficient β gives us the estimate of the impact of IGMSY on fertility decisions of women. We control for year effects to control for year specific un-observable factors which could have affected the likelihood of births or female births, district effects to account for district specific un-observable factors, other woman specific controls, such as age, dummies for Hindu, Muslim and Christian, wealth index and years of education. The standard errors are clustered at the district level. Table Annexure 2 in the Annexure reports the summary statistics of the variables used in the regression specifications.

To estimate the impact of IGMSY on probability of females born we replace our outcome variable in Equation (1) with females. It is once again a binary variable which equals one if the child born is a female and 0 if the child born is male. We control for the woman specific characteristics and once again β give us the estimate of the impact of the scheme on probability of females born.

IV. Results

Parallel Trend Assumptions

For a valid DID estimate, the first check is to ensure that there was no significant trend in difference in outcomes for treatment and control groups in the years prior to the treatment. These assumptions are known as the parallel trend assumption. To test the same, we estimate the following specification:

y_{i t d} = α_{0} + α_{1} {igmsy}_{i d} + δ t + β ({igmsy}_{i d} * {post}_{t}) + γ X_{i t} + + μ D_{d} + ε_{i t d}

(2)

Here, our outcome variables are birth and female, as discussed in the last section. We create a variable $t$ which is equal to the difference between 2012 and the year. We interact $t$ with IGMSY and our variable of interest is $({igmsy}_{i d} * {post}_{t})$ . If the coefficient β is statistically insignificant, we can conclude that there was no statistical difference in trend for outcome variables for treatment and control groups. The result for the estimation of Equation (2) is given in Table 4, and we find that there was no statistical difference in trends for both the outcomes.

Table 4.

Results of the Test of Parallel Trend Assumption.

	(1)	(2)
Variables	Birth	Female
t	0.001***	−3.79e-05
	(0.0003)	(0.001)
IGMSY*t	−0.002	0.001
	(0.001)	(0.001)
igmsy	−0.045***	0.075***
	(0.003)	(0.013)
Constant	0.079***	0.512***
	(0.005)	(0.022)
Observations	300,080	35,470

Source: The authors.

Note: Robust standard errors in parentheses ***p < 0.01, **p < 0.05, *p < 0.1. The other variables included in the regression are–district effects and other woman specific controls, such as age, dummies for Hindu, Muslim, Christian, wealth index and years of education. Bold values are the main coefficients of interest.

Impact of Scheme on Having a Child and Sex Ratio at Birth

Panel A in Table 5 gives the results of estimating the impact of IGMSY on probability of having a childbirth post 2011 (Equation (1)). Column 1 gives estimates of the impact for overall sample. Columns 2–5 report the results for the four sub-samples: column 2—first son and second son, column 3—first daughter and second daughter, column 4—first son and second daughter and column 5—first daughter and second son.

Table 5.

Impact of IGMSY on Birth and Female Birth.

	(1)	(2)	(3)	(4)	(5)
	did	SS	DD	SD	DS
Panel A: Outcome variable-birth
Did	−0.014	−0.017*	−0.017	−0.015	−0.007
	(0.008)	(0.010)	(0.016)	(0.012)	(0.011)
Constant	−10.29***	−7.355***	−19.45***	−10.01***	−9.633***
	(0.668)	(0.810)	(1.516)	(1.645)	(0.765)
Observations	300,080	111,264	43,760	44,512	100,544
Panel B: Outcome variable-female
did	−0.012	0.001	0.010	0.023	0.022
	(0.018)	(0.016)	(0.043)	(0.046)	(0.042)
Constant	−1.597	−1.302	−1.350	−0.646	−1.248
	(1.480)	(0.978)	(2.158)	(3.619)	(2.485)
Observations	35,470	10,625	7,570	6,594	10,681

Source: The authors.

Note: Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. The other variables included in the regression are— $p o s t$ , ${igmsy}_{i}$ , year effects, district effects and other woman specific controls, such as age, dummies for Hindu, Muslim, Christian, wealth index and years of education. Bold values are the main coefficients of interest.

Table 5 shows that the scheme had no impact on the overall fertility choices made by women in our sample. However, we find that the impact is dependent on the gender composition of the children. Among all the sub-groups, women with two sons are the only ones who have a lower likelihood of giving birth again. Our results show that the programme reduced the likelihood of a third order child among households eligible for the treatment by 1.7% among those who had two sons. On the contrary, we do not find any statistically significant impact on any other sub-groups of women in our sample. The results point towards gender preference in India and gives evidence to the theory of son-preference based stopping rule (Clark, 2000) where households stop having more children after reaching their desired number of sons. Jain (2019) while evaluating the impact of LLY in Madhya Pradesh also found that the change in fertility behaviour was dependent on the gender composition of previous births. The results for the estimation with female as the outcome variable are given in Panel B. We did not find any impact of the scheme on the likelihood of a female birth in the household.

Broadly these results show that the scheme’s impact was driven by gender preference of the household. The indirect signal of discouragement worked only for households who already had two sons first. Our results show that the norm of son preference might be a bottleneck in the success of any such scheme which targets the number of births.

Is the Scheme Penalising Higher Order Births?

In this sub-section, we explore some unintended impacts of the scheme on maternal and neo-natal health care for third or higher order births. We use data from child questionnaire of NFHS-4 for this analysis. The data is in the form of child year panel. As discussed earlier, data on these aspects are only available for births in the last 5 years that is, children who are born after 2011–2012. Therefore, we are unable to present causal estimates of the scheme on these aspects and validity of pre-trend parallel trend assumptions. We estimate the following equation:

y_{i t d} = α_{0} + α_{1} {igmsy}_{i d} + δ {post}_{2001 - 2016, t} + β ({igmsy}_{i d} * {post}_{t}) + γ ({igmsy}_{i d} * post_{2001 - 2016, t} * {order}_{i t d}) + θ {order}_{i t d} + π X_{i t} + ρ y e a r_{i t} + μ D_{d} + ε_{i t d}

(3)

Here our outcome variable is $y_{i t d}$ which is equal to one if the response to the variable is yes and zero if no. The outcome variables that we use are all related to visits to anganwadi centres, ante natal care visits, consumption of iron tablets, institutional delivery, vaccinations and child ICDS benefits. In particular, the outcome variables reported in Table 6 are—during pregnancy: received supplementary nutrition from anganwadi centre (supnut_awc), during pregnancy: given or bought iron tablets/syrup (ifa), during pregnancy: received benefits from anganwadi/ICDS centre (ante_iycf), number of antenatal visits during pregnancy (anc), number of tetanus injections before birth (tetanus) and place of delivery (insti). The outcome variables reported in Table 7 are: birth weight in kilograms less than 2.5 kgs (lbw), polio3, dpt3, measles are one if the respondent indicated that the child got these vaccines individually, vaccine is yes if child got combination of DPT3+Polio3+BCG+measles shot and childicds is one if child had received benefits from anganwadi/ICDS centre, in the last 12 months.

Table 6.

Health During Pregnancy Period—Child Questionnaire

	(1)	(2)	(3)	(4)	(5)	(6)
	supnut_awc	ifa	ante_iycf	anc	Tetanus	insti
igmsy*post	0.012	0.001	0.027**	0.009	0.004	−0.003
	(0.021)	(0.019)	(0.012)	(0.021)	(0.022)	(0.018)
igmsypostorder	−0.074***	−0.122***	−0.005	−0.127***	−0.112***	−0.018
	(0.027)	(0.034)	(0.019)	(0.032)	(0.039)	(0.022)
Constant	−189.7***	−234.5***	−3.120	−249.4***	−259.4***	−38.89***
	(11.30)	(10.85)	(8.997)	(10.16)	(10.58)	(7.606)
Observations	41,290	41,290	41,290	41,290	41,290	41,290

Source: The authors.

Note: Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Bold values are the main coefficients of interest.

Table 7.

Vaccination and Some Early Antenatal Care–Child Questionnaire

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	lbw	Polio3	DPT3	BCG	Measles	Vaccine	childicds
igmsy*post	0.008	0.026	0.032	0.022	0.038*	0.035	0.033**
	(0.020)	(0.016)	(0.022)	(0.015)	(0.020)	(0.023)	(0.014)
igmsypostorder	0.003	−0.066***	−0.054**	−0.066***	−0.055***	−0.039*	−0.043***
	(0.026)	(0.020)	(0.024)	(0.023)	(0.020)	(0.020)	(0.016)
Constant	−8.206	−56.23***	−17.65**	−26.94***	−10.34	−42.17***	−54.18***
	(8.944)	(9.592)	(8.066)	(7.118)	(7.323)	(9.345)	(9.894)
Observations	33,005	41,290	41,290	41,290	41,290	41,290	41,290

Source: The authors.

Note: Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1. Bold values are the main coefficients of interest.

Table 6 gives the results of the impact of the scheme on maternal care while Table 7 gives the impact of the scheme on neo-natal care. The controls and model specification remain the same as Equation (1). Order is an indicator variable which is equal to one if the child is third or higher order and zero otherwise. Here, our primary variables of interest are $({igmsy}_{i d} * {post}_{t})$ and $({igmsy}_{i d} * {post}_{2001 - 2016, t} * {order}_{i t d})$ and size of β and γ tells us about the impact of the scheme on maternal and neo-natal health care outcomes for overall child birth and outcomes for third and higher order child birth respectively.

From Table 6 we find statistically significant impact of IGMSY on only one indicator of maternal care related variables, (ante_iycf). This implies that after the implementation of the scheme, there was an increase in the likelihood of receiving benefits from anganwadi/ICDS centre during pregnancy. However, we find no impact on number of visits for antenatal care or other indicators of maternal care. It is interesting to note that there has been a statistically significant decline in likelihood of several indicators of maternal care for third and higher order births. In particular, they are less likely to get ante natal care, iron tablets and tetanus shots. But we do not find any change in likelihood of institutional delivery, even for the higher order births. The results point towards an evidence of penalty for third or higher order pregnancy.

On similar lines, results reported in Table 7 show that kids are more likely to get measles vaccine, or total vaccination of DPT3+Polio3+BCG+measles shot and child benefits from ICDS in his/her early years. However, we find that the likelihood of all these have reduced for the third and higher order births. Given that we do not have data on child born before 2011 from NFHS-4 dataset, we cannot conclude causality from our estimates. The results point towards an unintended negative impact of the scheme.

Overall, we find that the scheme could not reduce the likelihood of births except for sub-samples comprising of households with two sons in India. The heterogeneity in the impact of the scheme across different sub-samples points towards the strong gender preference in the country. However, the scheme unfortunately declined the maternal and neo-natal care received by mothers for higher order births and children of higher birth order.

V. Falsification Tests

It is plausible that evidence on the impact of IGMSY on fertility choice and sex ratio at birth that we reported in Section IV are driven by some year specific effect. Therefore, we test the robustness of our results by first altering the year of intervention. Our hypothesis is that there should be no statistically significant estimate of our impact if we alter the years of intervention. Table 8 gives the estimates of the impact if the scheme had been implemented earlier than 2011. We do not find any impact of the scheme for most of the years. It is plausible that the year 2010 gives significant impact since the scheme was implemented in December 2010 and hence some births post 2010 could have been registered for the scheme and hence exposed to the treatment.

Table 8.

False Years.

	(1)	(2)	(3)	(4)	(5)
	DID2006	DID2007	DID2008	DID2009	DID2010
did2006	−0.013
	(0.009)
did2007		−0.011
		(0.009)
did2008			−0.013
			(0.009)
did2009				−0.011
				(0.009)
did2010					−0.014*
					(0.008)
Constant	9.047***	6.946***	1.750***	−4.282***	−8.054***
	(0.706)	(0.643)	(0.609)	(0.614)	(0.613)
Observations	300,080	300,080	300,080	300,080	300,080
	DIDf2006	DIDf2007	DIDf2008	DIDf2009	DIDf2010
did2006	0.006
	(0.015)
did2007		0.011
		(0.015)
did2008			−0.001
			(0.015)
did2009				0.005
				(0.017)
did2010					−0.002
					(0.019)
Constant	0.773	0.953	−2.733	−2.045	−1.303
	(1.902)	(1.927)	(1.849)	(1.751)	(1.653)
observations	35,470	35,470	35,470	35,470	35,470

Source: The authors.

Note: Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1.

It can also be argued that the impact of the scheme as in Section IV was driven by the definition of treatment and control districts. To ensure that our results are not obtained by chance, we re-define the treatment and control groups by randomly selecting districts into control and treatment groups and re-estimate Equation (1). Our hypothesis is that if our estimates in Section IV are robust, we should not find any statistically significant impact of the scheme in our re-defined treatment and control groups. Table 9 gives the results of estimations using randomly generated treatment and control groups and we do not find any statistically significant impact.

Table 9.

Randomly Generated Treatment and Control Groups.

	(1)	(2)
	Birth	Female
Treatment*t	0.002	0.004
	(0.002)	(0.012)
Constant	−10.29***	−1.606
	(0.667)	(1.480)
Observations	300,080	35,470

Source: The authors

Note: Robust standard errors in parentheses. ***p < 0.01, **p < 0.05, *p < 0.1.

VI. Conclusion

Indira Gandhi Matritva Sahyog Yojana (IGMSY) is the largest maternal and child conditional cash transfer programme in the world and has been implemented in India since May 2010. As discussed in the earlier sections, the scheme provides financial incentives to women above 19 years of age for their first two births and does not provide for higher order births and does not, in any way, target the gender of the child born. The scheme was initially piloted in 52 districts which were chosen based on a score computed using parameters related to maternal and neo-natal care. This staggered implementation of the scheme allows us to estimate the impact of the scheme by comparing treatment districts (52 districts) with other control districts (those who did not receive any treatment) after matching them on the basis of the parameters which were used to choose the 52 districts in the first place.

Using data from the National Family Health Survey (NFHS-4)—2015–2016 and a Difference in Difference framework, we explore three questions in this study—(a) did IGMSY reduce the probability of women having a child post 2011 and did son preference have any role in that, (b) did the likelihood of female birth change after the scheme and once again did gender of the first two births have any role and (c) by focusing on first two births, did the design of the scheme in any way penalise the third child?

Our results show that as expected, there was a reduction in likelihood of birth but only in those households where the first two born were sons. We find that there was a 1.7% decline in the likelihood of birth of a third child for such households. Our results point towards the role of son preference norm driving the impact of the scheme. These findings are in line with other studies in literature, such as Anukriti (2018), Javadekar and Saxena (2019) and Jain (2019) which point towards the significant role of gender norms and son preference in influencing the impact of other schemes/interventions in the past in India which targeted fertility and sex ratio at birth. With regards to penalising the third child, we are unable to provide causal estimates given the paucity of data. However, we do find evidence of some penalty for indicators of maternal and ante-natal care.

The findings from the study direct us towards the design of the scheme in terms of targeting first two births. Given the context of son preference in India, it is likely that the scheme would not be able to successfully alter fertility choices of households and unfortunately might penalise higher order births.

Annexure

Annexure 1.

List of 52 Districts where IGMSY was Implemented.

Sr. No	State	District
Sr. No	State	District
1	Andaman and Nicobar Island	South Andaman
2	Andhra Pradesh	West Godavari, Nalgonda
3	Arunachal Pradesh	Papum Pare
4	Assam	Kamrup, Goalpara
5	Bihar	Vaishali, Saharsa
6	Chandigarh	Chandigarh
7	Chhattisgarh	Dhamtari, Bastar
8	Dadra and Nagar Haveli	Dadra & Nagar Haveli
9	Daman and Diu	Diu
10	Delhi	West, North West
11	Goa	North Goa
12	Gujarat	Bharuch, Patan
13	Haryana	Panchkula
14	Himachal Pradesh	Hamirpur
15	Jammu & Kashmir	Kathua, Anantnag
16	Jharkhand	East Singh Bhumi, Simdega
17	Karnataka	Kolar, Dharwad
18	Kerala	Palakkad
19	Lakshadweep	Lakshadweep
20	Madhya Pradesh	Chindwara, Sagar
21	Maharashtra	Bhandara, Amravati
22	Manipur	Tamenglong
23	Meghalaya	E. Garo Hills
24	Mizoram	Lawngtlai
25	Nagaland	Kohima
26	Orissa	Bargarh, Sundargarh
27	Pondicherry	Yanam
28	Punjab	Amritsar, Kapurthala
29	Rajasthan	Bhilwara, Udaipur
30	Sikkim	West Sikkim
31	Tamil Nadu	Cuddalore, Erode
32	Tripura	Dhalai
33	Uttar Pradesh	Mahoba, Sultanpur*
34	Uttarakhand	Dehradun
35	West Bengal	Jalpaiguri, Bankura

Source: The authors.

Note: ***p < 0.01, **p < 0.05, *p < 0.1

Annexure 2.

Summary Statistics of Explanatory Variables.

	Control	Treatment	Difference
Age in years	29.225	29.785	−0.56***
Education in years	8.595	8.807	−0.211***
Hindu	0.693	0.705	−0.012***
Muslim	0.132	0.103	0.029***
Rural	2	2	0***
Sex hhs head	1.113	1.155	−0.043***
Age hhs head	48.19	48.725	−0.535
Newspaper	0.984	0.934	0.049
Radio	0.325	0.248	0.077*
Television	2.429	2.553	−0.124***
Wealth index	3.667	3.783	−0.116***
Number of observations	17,081	1,674

Source: The authors.

Note: ***p < 0.01, **p < 0.05, *p < 0.1

Footnotes

Acknowledgements

We thank the editor and the reviewers of the journal for their insightful comments and suggestions.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

Note

ORCID iDs

Tirtha Chatterjee

Ritika Jain

References

Angrist

J. D.

, & Pischke

J. S.

(2009). Mostly harmless econometrics: An empiricist’s companion Princeton university press.

Angrist

J. D.

, & Pischke

J. S.

(2014). Mastering’metrics: The path from cause to effect . Princeton university press.

Anukriti

, Bhalotra

S. R.

, & Tam

E. H. F.

(2016). On the quantity and quality of girls: New evidence on abortion, fertility, and parental investments.

Anukriti

(2018). Financial incentives and the fertility-sex ratio trade-off. American Economic Journal: Applied Economics , 10(2), 673–688.

Bhalotra

S. R.

, & Cochrane

(2010). Where have all the young girls gone? Identification of sex selection in India .

Clark

(2000). Son preference and sex composition of children: Evidence from India. Demography , 37(1), 95–108.

Gertler

P. J.

, Martinez

, Premand

, Rawlings

L. B.

, & Vermeersch

C. M.

(2016). Impact evaluation in practice . World Bank Publications.

Ghosh

, & Kochar

(2018): Do welfare programs work in weak states? Why? Evidence from a maternity support program in India. Journal of Development Economics , 134(C), 191–208.

Government of India, Ministry of Women and Child Development, Implementation guidelines for state governments/U.T. Administrations, F.No.9-5/2010-IGMSY. October 1. https://wcd.nic.in/sites/default/files/IGMSYImpGuidelinesApr11.pdf

10.

Jain

(2019). Do Financial Incentives Affect Fertility and Sex Ratio at Birth? An Impact Evaluation of Madhya Pradesh’s Ladli Laxmi Yojana .

11.

Javadekar

, & Saxena

(2019). The impact of a conditional cash transfer program on household’s fertility choices in India. SSRN 3341448 .

12.

Jayachandran

, & Pande

(2017) Why are Indian children so short? the role of birth order and son preference. American Economic Review , 107(9), 2600–2629.

13.

Jensen

(2012) Another mouth to feed? the effects of (in) fertility on malnutrition. CESifo Economic Studies , 58(2), 322–347.

14.

Khandker

S. R.

, Koolwal

G. B.

, & Samad

H. A.

(2009). Handbook on impact evaluation: quantitative methods and practices . World Bank Publications.

15.

Lechner

(2011). The estimation of causal effects by difference-in-difference methods. Foundations and Trends® in Econometrics , 4(3), 165–224.

16.

Rosenblum

(2013). The effect of fertility decisions on excess female mortality in India. Journal of Population Economics , 26(1), 147–180.

17.

von Haaren

, & Klonner

(2020). Maternal cash for better child health? The impacts of India’s IGMSY/PMMVY maternity benefit scheme .