Cash,Culture and Constraints: Why Conditional Transfers May Fail to Delay Early Marriage

Introduction

Conditional cash transfers (henceforth, CCTs) have emerged as critical instruments in promoting gender equity and female empowerment by alleviating household credit constraints and increasing the opportunity cost of early marriage (Baird et al., 2013; Rubio-Codina, 2010). Early marriage, a key driver of the feminization of poverty (Otoo-Oyortey & Pobi, 2003), is inversely related to female educational attainment (Kalamar et al., 2016). This relationship appears bi-directional, as early marriage truncates schooling, which, in turn, raises the likelihood of early marriage (Delprato et al., 2015). Given this nexus, incentivizing female education through CCTs may augment welfare by improving girls’ ability to make autonomous life choices. CCTs are grounded in the collective household model, which departs from the unitary framework by allowing for intra-household bargaining among individuals with distinct preferences (Chiappori, 1992). This framework enables targeted transfers to enhance individual agency within households, particularly for adolescent girls.

In India, several girl-focused CCT programmes condition the final benefits on delaying marriage beyond the age of 18 (Sekher, 2010). While intended to promote education and postpone marriage, the effectiveness of these schemes remains context-sensitive. In poor households where daughters are seen as financial liabilities, transfers may fund marriage expenses rather than support long-term human capital investments, especially when education yields limited labour market returns. ¹ Consequently, families may time marriage to coincide with benefit eligibility, undermining programme objectives (Chae & Ngo, 2017; Nanda et al., 2022). Further, when educational infrastructure is weak, transfers often yield enrolment gains without commensurate improvements in learning. In such cases, complementary investments in school quality and teacher training may produce greater social returns. We examine the Kanyashree Prakalpa (henceforth, KP), a CCT initiative launched in West Bengal, India, in 2013, designed to promote girls’ education and delay marriage. Our contribution is twofold. First, we assess whether CCTs induce long-term behavioural changes. Our study leverages a cohort of beneficiaries who have reached the legal age of marriage since the programme’s 2013 launch, thus offering a more definitive assessment of its long-term impact on early marriage. Second, we explore heterogeneity in impact across household economic conditions, geographic contexts and social groups. The findings indicate that while KP increased school enrolment, it failed to prevent early marriage. To achieve transformative change, CCT programmes must be embedded in multidimensional strategies that strengthen educational quality, promote gender-equitable norms, and ensure effective programme implementation and targeting (Harper et al., 2020). Furthermore, effective on-the-ground implementation is crucial, with mechanisms to ensure proper administration and prevention of fund misuse and disbursements to ineligible recipients. These insights carry broader implications for the design of social protection policies aimed at improving gender outcomes in low-income settings.

The rest of the study is organized as follows: ‘Background and Programme Description’ section outlines the context and describes the programme under discussion in this article. The ‘Review of Literature’ section reviews the relevant literature. The ‘Data Source and Choice of Controls’ section discusses the data sources and study variables. The ‘Empirical Approach’ section describes the empirical framework. The ‘Results: Impact on Early Marriage’ section presents the results and subsequent discussion. The ‘Robustness Checks’ section elaborates on the robustness checks for the study. Lastly, the conclusion is contained in the ‘Conclusion’ section.

Background and Programme Description

Despite the Child Marriage Act of 2006, which set the legal minimum age of marriage for girls at 18, early marriage remains widespread in West Bengal. According to the District Level Household Survey-3 conducted in 2007–2008, the state ranked fifth nationally in child marriage prevalence, with 54.7% of girls married before 18. Educational attrition is also stark: female school attendance fell from 85% among 6–10-year-olds to just 33% among those aged 15–17, as reported by the National Family Health Survey-3 (NFHS-3), conducted in 2005–2006. The transition from free elementary to fee-based secondary education imposes financial burdens on poor households, incentivizing early marriage as a form of perceived economic security. This dynamic reinforces intergenerational poverty and limits women’s labour market participation, contributing to the feminization of poverty. Efforts by various policymaking units to prevent early marriage through legal prohibition and social communication remained ineffective.

Realizing the need for concrete actions for social transformation, the West Bengal Government launched KP in August 2013, with the primary objectives of ensuring girls’ retention in schools and postponing their marriages till they reach 18 years of age. The programme consists of two CCT elements: first, a yearly scholarship of ₹750 allocated to girls aged between 13 and 18, enrolled in Class VIII and equivalent or higher, subject to their being unmarried. Second, a one-time grant of ₹25,000 is to be paid when girls reach the age of 18, given that they are engaged in either academic or vocational institutes and remain unmarried. Comparing KP’s one-time grant with the average monthly household consumer expenditure in West Bengal emphasizes its financial significance. Data from the Household Consumer Expenditure Survey (2011–2012) conducted by the National Sample Survey Office reveal that the average monthly expenditure in rural areas of West Bengal is ₹5,295 and ₹9,844 in its urban areas. This grant amounts to approximately 4.7 months of expenditure for rural households and 2.5 months for urban ones. This amount, given when girls turn 18 years of age, aims to ease economic pressures that could otherwise drive early marriage. Moreover, a new component was introduced in July 2017 to further motivate girls towards higher education and self-reliance. Regardless of marital status or annual family income, those pursuing postgraduate degrees at any university would receive a monthly stipend of ₹2,500/- for the science stream and ₹2,000/- for the arts stream. However, prior enrolment in this programme’s first and second components and securing at least 45% marks in their undergraduate degrees are required to be eligible. The funds are directly deposited into the enrolled student’s bank account. The programme was initially meant for families with a yearly income of ₹120,000 and less; however, this income limit was removed in August 2018, enabling girls of all economic backgrounds to benefit from this programme. A review of the K1 and K2 sanction records from the Management Information System shows that programme uptake began increasing noticeably from 2014 to 2015, indicating an early and growing participation soon after the scheme’s launch. Figure 1 shows a non-decreasing trend in stipend claims till 2021. ²

OPEN IN VIEWER

Figure 1.

Source: Data have been obtained from the Kanyashree Prakalpa dashboard, Government of West Bengal.

KP’s national and international recognition reflects its scale and reach. National assessments support these findings: the Annual Status of Education Reports (ASER Centre, 2014–2022) exhibit a consistent increase in girls’ secondary school enrolment in West Bengal since KP’s launch in 2013. However, despite these educational gains, early marriage remains entrenched. Data from NFHS-4 (2015–2016) and NFHS-5 (2019–2021) show little change in early marriage prevalence, with West Bengal registering the highest incidence nationwide in NFHS-5. The Sample Registration System (SRS) Statistical Report (2020) corroborates this, showing a continued rise in the effective marriage rate among girls under 18 since 2018.

OPEN IN VIEWER

Figure 2.

Source: Data have been obtained from five rounds of National Family Health Survey (NFHS) (1992–1993, 1998–1999, 2005–2006, 2015–2016 and 2019–2021).

The comparative trends in the rate of early marriage in India and West Bengal, highlighting the state’s persistent underperformance in combating early marriage, are evident in Figure 2. With the sole exception occurring in 1998–1999, West Bengal’s rates have always remained higher than the national average. Remarkably, these findings show a discrepancy between the rising number of females enrolled in schools and the persistent prevalence of early marriage. The assumptions upon which KP was built are called into question by this inconsistency. It draws attention to the urgent need for a comprehensive assessment of the actual effectiveness of this programme in addressing the issue of early marriage in the state.

Review of Literature

Data Source and Choice of Controls

This study uses data from the NFHS, a nationally representative household survey conducted by India’s Ministry of Health and Family Welfare. Specifically, it draws on NFHS-3 (2005–2006) as the baseline and NFHS-5 (2019–2021) as the endline. The choice of these two rounds is motivated by the 2013 launch of KP, with NFHS-5 preferred over NFHS-4 to allow for a longer post-intervention window to assess long-term impacts on early marriage. NFHS-3 surveyed 124,385 women aged 15–49 across 109,041 households, while NFHS-5 included 699,686 women in the same age group from 724,115 households. Given the absence of a panel structure, the analysis treats these data sets as repeated cross-sections.

To evaluate the impact of KP on early marriage, we compare outcomes between a treatment group—girls from West Bengal and a control group comprising girls from neighbouring states with broadly similar socio-economic characteristics. ³ Five metrics were used to assess comparability: (a) percentage of women with 10+ years of schooling, (b) percentage of women aged 20–24 married before age 18, (c) sex ratio, (d) a women’s empowerment index, and (e) the Social Progress Index. ⁴ Sikkim was excluded due to its markedly higher performance across these key indicators. Odisha and Jharkhand were also excluded, as their NFHS-5 data were collected during the COVID-19 pandemic (Phase II: 2 January 2020–30 April 2021), potentially biasing key outcomes.

Tripura was included despite not sharing a direct border with West Bengal, due to its sociocultural proximity and linguistic similarity—approximately 65% of its population speaks Bengali (Census, 2011). Assam and Bihar, surveyed in Phase I of NFHS-5 (June 2019–January 2020), were also included. All control states were selected to ensure that their data predated the pandemic and shared structural similarities with West Bengal. Although Assam, Tripura and Bihar operate other CCT programmes targeting girls, these differ significantly from KP in design, timing and eligibility criteria. Given the 19–24 age cohort analyzed, these alternate schemes are unlikely to confound the results, thereby preserving the validity of the control group.

Empirical Approach

To identify the causal impact of KP on early marriage, we employ a difference-in-differences (DID) approach using repeated cross-sectional data from NFHS-3 (2005–2006) and NFHS-5 (2019–2021). This method compares changes in early marriage rates over time between the treated group (girls in West Bengal) and the control group (girls in comparable neighbouring states). The first difference captures the change in early marriage prevalence within each group over time, and the second difference isolates the programme effect by netting out the change observed in the control group from that in the treatment group.

DID accounts for unobserved, time-invariant heterogeneity between treatment and control states, as well as common time trends affecting both groups. This approach thus strengthens causal inference by controlling for potential confounders that are constant over time, thus attributing any differential change in early marriage rates to the KP intervention. The DID equation to be analyzed is as follows:

Y i t = β 0 + β 1 Treatment i + β 2 Post t + β 3 Treatment i × Post t + X i ' β 4 + ε i t

(1)

Where Y represents the dependent variable, that is, early marriage for girl i at time t. Here, t denotes the time periods in the data set, which correspond to two NFHS rounds—NFHS-3 (pre-treatment period) and NFHS-5 (post-treatment period). Treatment is a binary indicator that is coded as 1 for the treated group and 0 for the control group. The Post variable, which is also a binary indicator, equals 1 for data from the endline survey, that is, NFHS-5, and 0 for data from the baseline survey, that is, NFHS-3. Confounding individual and household-level factors that could affect the dependent variable are included in the matrix X i . ε i t indicates the error term in the model. The bootstrap method is employed to compute standard errors, as it provides more reliable inference, especially when the number of clusters in the treatment and control groups is small within the DID framework. ⁵ The Linear Probability Model is used in the study, enabling the interpretation of coefficients as percentage point (pp) changes in the likelihood of early marriage. Effect sizes are also reported as a percentage of the baseline mean of the outcome variable. Odds ratios from logistic regression models are presented alongside the linear probability model estimates to provide a complementary interpretation in terms of relative likelihood. All models incorporate survey weights to reflect the complex sampling design of the NFHS. The causal impact of KP on the probability of early marriage is measured by the interaction term Treatment × Post in the DID model. ⁶

Results: Impact on Early Marriage

Table 1 represents the DID regression results assessing the impact of KP on the prevalence of early marriage among girls, using three different specifications. The first specification addresses the main variable of interest, that is, early marriage, without any additional controls. The second specification includes additional individual and household-level control variables. The third specification, which is the preferred approach, uses the bootstrap method to compute standard errors. In all specifications, the coefficient for the time indicator (Post) is negative and statistically significant, reflecting a reduction in the rate of early marriage over time among the control states. In all the specifications, this interaction term is positive and highly significant, implying a relative increase in early marriages in the treated group post-intervention. The inclusion of individual and household-level controls in the second specification improves the accuracy of these estimates. With bootstrapped standard errors, the probability of early marriage in West Bengal rises by 8 pp during the post-KP period. Given a baseline mean of 54% in the control group, this corresponds to an effect size of 14% of the baseline mean. Considering that the scheme’s primary objective was to curb early marriage, this outcome shows substantially paradoxical results.

OPEN IN VIEWER

Table 1.

Estimation Results for Early Marriage.

	Dependent Variable: Early Marriage
	LPM Coefficient			Odds Ratio (Logistic)
	(1)	(2)	(3)	(4)	(5)	(6)
Treatment×Post	.06*** (0.02)	.08*** (0.02)	.08*** (0.02)	1.30*** (0.12)	1.48*** (0.16)	1.48*** (0.16)
Treatment	−0.02 (0.02)	0.01 (0.02)	0.01 (0.02)	0.92 (0.08)	1.07 (0.11)	1.07 (0.11)
Post	−.18*** (0.01)	−.12*** (0.01)	−.12*** (0.01)	.48*** (0.29)	.54*** (0.04)	.54*** (0.04)
Controls	No	Yes	Yes	No	Yes	Yes
Bootstrapped standard errors	No	No	Yes	No	No	Yes
Constant	.54*** (0.01)	.71*** (0.02)	.71*** (0.02)	1.18*** (0.07)	2.47*** (0.26)	2.47*** (0.30)
Effect size (% of baseline mean)	11.62	14.21	14.21	–	–	–
No. of observations	25,061	21,364	21,364	25,061	21,364	21,364

Notes: Significance levels: ***p < .01, **p < .05, *p < .1. Columns (1) and (4) present basic specifications; columns (2) and (5) incorporate both individual and household characteristics. In columns (1), (2), (4) and (5), standard errors are robust. In columns (3) and (6), standard errors are computed by the bootstrap method (2,000 replications). Columns (1)–(3) report linear probability model (LPM) coefficients; columns (4)–(6) report odds ratios from logistic regression. Standard errors are presented in parentheses. All models are estimated using survey weights. Detailed results of the control variables corresponding to the analysis in columns (2), (3) and (5), and (6) are presented in Appendix A as Table A3.

Girls from richer households have a lower likelihood of marrying early, suggesting that families with economic stability and better financial conditions have alternatives to early marriage for their daughters, such as continued education and career opportunities (see Appendix A, Table A3). Girls who had the means and opportunity to obtain higher levels of education are significantly less likely to marry before 18. Further, girls residing in rural regions are more likely to marry early compared to those residing in urban regions. Differences also emerge across religious and social groups. The education level of the household head is also associated with early marriage outcomes. Girls from households with an educated head show a higher probability of delayed marriage compared to those from households with no formal education of the head.

To further support the main findings, an event study framework is used to assess the dynamics of treatment effects over time and to validate the assumption of parallel trends. This analysis utilized data from the first (1992–1993), second (1998–1999), third (2005–2006), fourth (2015–2016) and fifth (2019–2021) rounds of the NFHS, with the third round serving as the baseline period. Although the sample from NFHS-4 includes girls aged 19–24, most of whom (aged between 21 and 24) were not exposed to the programme, as they had already crossed the eligible age by the time of the scheme’s launch in 2013, this round is nonetheless included to capture transitional effects among partially exposed cohorts. The results, illustrated in Figure 3 through a coefficient plot, show no statistically significant effects during the pre-treatment periods, that is, NFHS-1 and NFHS-2, confirming that the parallel trends assumption holds. The post-treatment periods clearly show an increase in the prevalence of early marriage, indicating a paradoxical post-intervention pattern. Moreover, the higher estimated effect observed in NFHS-4 may be due to the age-based heterogeneity within the sample. In contrast, by NFHS-5, the 19–24 age group consisted entirely of cohorts exposed to the programme, leading to a uniform treatment effect.

OPEN IN VIEWER

Figure 3.

Note: The figure presents marginal effects with 95% confidence intervals. Individual and household characteristics are incorporated. The green vertical line marks the introduction of the Kanyashree Prakalpa (KP) in 2013.

The prevalence of early marriage increases by 13 pp when eligible girls from the poorer wealth quintiles are studied (see Appendix A, Table A4), indicating that the effects are concentrated among economically disadvantaged households. ⁷ This is likely because of a non-trivial effect of the associated one-time grant (component K2 of the scheme) of ₹25,000, which is paid to the girls upon turning 18 years of age. Possibly, this amount is used to fund the ‘dowry’ for the girls. ⁸ Although dowry practices are not confined to economically disadvantaged groups only, for households living at the margin, the transfer may provide a crucial means to meet dowry-related expenses. As a robustness check for the main analysis, we also examine the effects among girls from richer wealth quintiles, where we expect no or minimal impact of the programme, given that the dowry channel is less relevant and the one-time grant of KP is unlikely to substantially influence marriage decisions (see ‘Placebo Treatment Effect Test’ section).

The perceived limited economic opportunities for women and the lack of return from education may influence household decisions regarding early marriage, even when conditional financial incentives are provided. Parents might view the KP grant less as a scholarship for education and more as financial support for their daughters’ eventual marriage, given the grim employment opportunities for women in West Bengal, where female labour market participation is noticeably low and declining. Further, the general deterioration in learning outcomes observed in West Bengal post-KP implementation (Das & Sarkhel, 2023) might reinforce the perception. Thus, with grim employment opportunities and faltering learning outcomes, positive returns from higher education are unlikely. Coupled with this, an increase in the age of the girls would only cause the dowry amount to go up. This highlights the possibility that financial incentives meant to encourage education and postpone marriage could instead be redirected to increase the affordability of dowries and weddings for the resource-constrained households. The potential misuse of the KP funds for facilitating marriages is not without precedent in similar schemes. For example, the ‘Bhagyalakshmi Scheme’ in Karnataka observed that 76.7% of its recipients used financial assistance for marriage-related expenditures (Prabhu, 2020).

Contrary to the scheme’s intended objective, the observed outcomes suggest limited success in preventing early marriage and coincide with an increase in these rates. As Anukriti (2018) suggests, the design of financial incentive schemes needs careful consideration to prevent unintended consequences, such as worsening existing gender disparities. The West Bengal scenario adds complexity to this narrative. Recent national reports corroborate a more nuanced picture. Data from NFHS-5 (2019–2021) and SRS (2020) reveal that West Bengal continues to have one of the highest rates of early marriage in the country, with no discernible decrease from NFHS-4 (2015–2016). In fact, reports from the SRS point to a gradual increase in underage marriages since 2018. Interestingly, a study by Das and Sarkhel (2023) found that KP has successfully increased school enrolment rates among girls, there has been no remarkable impact on the likelihood of attaining a secondary or higher secondary level of education. Higher educational attainment is traditionally considered a key factor in preventing early marriage by improving labour market outcomes and increasing the opportunity cost of marriage (Kirdar et al., 2018). According to the Unified District Information System for Education Plus Report (2018–2019), despite a high gross enrolment rate of 91.1% in secondary education among girls, enrolment in higher secondary education is relatively low at 56.6%, indicating a significant gap in the transition from secondary to higher secondary education. This trend is both alarming and surprising, given that KP requires girls to remain enrolled in either educational or vocational institutes and unmarried until they turn 18 years old.

This gap can have two possible explanations. First, a decline in learning outcomes among females following the implementation of KP (Das & Sarkhel, 2023). This decline can be explained by higher enrolment rates without corresponding improvements in school infrastructure. Parents may initially enrol their daughters in schools for financial incentives, but later withdraw them due to limited educational benefits. Second, the increase in school enrolment fails to translate into higher female workforce participation, given the poor quality of education and the lack of employment opportunities in the state. Based on data from the Employment and Unemployment Survey (2007–2008, 2011–2012) and the Periodic Labour Force Survey (2018–2019), West Bengal’s female worker population ratio (FWPR) remains lower than the national average (Panel a of Figure 4). In 2011–2012, West Bengal’s FWPR for females aged 15 and above was 24.6%, dropping to 21.7% by 2018–2019, when the national averages were 30.5% and 23.3%, respectively. Panel b of Figure 4, based on these surveys, confirms that this gap is more pronounced in rural West Bengal, where the FWPR for women with secondary or higher education is persistently below the national rural average. This pattern is consistent with the existing evidence, which suggests that educational attainment does not necessarily translate into labour force participation for women (Chatterjee et al., 2018). This implies that, without human capital accumulation through significant learning and a lack of subsequent employment opportunities, KP’s positive effect on school enrolment alone offers little incentive for delaying marriage. Thus, merely incentivizing school enrolment is insufficient to make the girls suitable for the labour market and, therefore, does nothing to influence their marriage decisions significantly.

OPEN IN VIEWER

Figure 4.

Source: Data have been obtained from the Employment and Unemployment Survey (EUS) 2007–2008, 2011–2012, and the Periodic Labour Force Survey (PLFS) 2018–2019.

In resource-constrained settings, where the design of CCTs often favours immediate gains from tangible assets over long-term investments, the difficulties faced by the KP scheme highlight the larger implications for development policies. Due to the expensive screening processes required to ensure educational quality, this preference may result in underinvestment in quality education, thereby subverting the intended goals (Das & Sarkhel, 2023). As Ghatak and Muralidharan (2019) noted, the issue lies not in welfare spending per se, but rather in the poor targeting and implementation of such spending, which leads to major distortions. This creates a vicious cycle of low state capacity and poor spending quality, requiring more funds to gain any impact, and this ultimately leads to a state that is chronically underinvested. This cycle can force the government to opt for visible transfer mechanisms, like CCTs, to minimize the political costs of restructuring the complicated systems controlled by unions and political leaders, which often lead to poor human capital formation, threatening broader developmental objectives (Reimers et al., 2006).

The unintended outcome strongly suggests a high possibility that the conditions set for the scheme are being circumvented on a large scale. The local officials might engage in malpractices such as the disbursement of grants to girls who do not meet the eligibility criteria to reach the annual targets set for each district by the District Project Management Unit, which are based on the Unified Districts Information System for Education Plus reports. These malpractices include providing grants to girls who are already married or not enrolled in any educational or vocational institute, thus undermining the objectives of the programme. This is consistent with Lipsky’s (2010) theory of street-level bureaucracy, which holds that the discretion and pressure faced by local officials might lead to the prioritization of quantity over quality when awarding one-time grants. This interpretation is supported by evidence from the sanction records of K2, obtained from the management information system data. With 53,701 applications approved under the K2 category of the scheme, the district of Murshidabad had the highest number of sanctioned applications for the one-time grant of ₹25,000 during the 2019–2020 fiscal year. This was followed by South 24-Parganas and North 24-Parganas, with 44,511 and 44,022 approved applications, respectively. The annual KP report (2019–2020) has recognized the districts of North 24 Parganas, Murshidabad, South 24 Parganas and West Midnapore as high achievers in terms of overall performance since the inception of KP. However, NFHS-5 (2019–2021) shows that the percentage of women aged between 20 and 24 who were married before turning 18 years of age in Murshidabad and West Midnapore exceeded 55%, an increase from NFHS-4 (2015–2016). This rate is even higher than the state average of 41.6%. Additionally, in a district like Hooghly, despite the approval of over 25,000 K2 applications, the number of early marriages has significantly increased from 31.9% in NFHS-4 to 40.8% in NFHS-5. These statistics indicate a likely correlation between high rates of sanctioned KP applications and increased rates of early marriage. This issue is further illustrated by Putnam’s (1994) Social Capital Theory, which highlights the importance of norms, networks and social trust in promoting cooperation for mutual gain. A high level of social capital can ensure appropriate implementation of the KP programme by enhancing community involvement and accountability. There is a greater possibility of funds being misused in places with poorer community networks.

Again, in the absence of any educational schemes for boys, parents may be reluctant to continue their daughters’ education, fearing it will be difficult to find suitable grooms, especially in places with low literacy rates (Sen & Dutta, 2018). Given that the amount of dowry tends to rise with age (Maertens & Chari, 2020), these dynamics pose a troubling possibility that parents marry off their daughters before the age of 18 and claim the one-time KP grant for purposes contrary to the objectives of the programme. ⁹ The scheme does not fully address the underlying structural concerns, which are indicated by the possible misuse of KP funds. This underscores the programme’s limitations and reflects the need for complementary measures, such as better education quality, awareness campaigns and sex education, to address these emerging challenges (Dutta & Sen, 2020). Further, a secular decline in employment opportunities in the state (PLFS, 2021) disincentivizes investment in higher education of daughters, leaving open the only alternative of marriage, as reduced labour market returns are known to prepone women’s marriage and fertility (Jensen, 2012). These facets alter the effectiveness of monetary incentives designed to discourage early marriage and imply that KP, while well-intentioned, functions within a complex socio-economic framework that is limiting its impact.

We also considered the possibility of families migrating to West Bengal to access the benefits of KP. However, the programme has stringent residency requirements, limiting eligibility to only households residing in West Bengal. The beneficiaries are also required to be enrolled in academic or vocational institutes located within the state. Additionally, the social and economic costs associated with migration make it unlikely that households from neighbouring states would relocate to access the programme. These factors substantially limit the potential for migration to confound the findings of the study.

Robustness Checks

Birth-year Cohort Specific Effects

To examine cohort-specific effects of KP, we divide birth cohorts based on the scheme’s launch year (2013) and the NFHS survey years. The pre-treatment cohort includes girls born between 1968 and 1994 (with 1968 marking the birth year of women aged 24 in the first NFHS round), who were 19–24 years old during the NFHS rounds before KP’s implementation. All five rounds of NFHS are used to capture cohort variations across periods. The post-treatment cohort includes those born between 1995 and 2000, who were exposed to the scheme by the time they reached eligibility age. The interaction term between the treatment group and the post-KP cohort is positive and statistically significant, indicating a 5 pp increase in the probability of early marriage (Appendix A, Table A6). This corresponds to an effect size of 8.26% relative to the baseline mean. This result reinforces the core finding that the KP scheme was ineffective in delaying the age at marriage for the targeted cohorts.

OPEN IN VIEWER

Figure 5.

Source: Authors’ calculation using National Family Health Survey (NFHS) data sets (all rounds).

Figure 5 presents the predicted probabilities of early marriage by each birth year for West Bengal (treated group) and control states. The green vertical line marks the birth year from which onward girls in West Bengal were exposed to the treatment. Prior to the programme, a continuous high prevalence of early marriage was not observed among the treatment group compared to the control group. However, after the introduction of KP, the probability of early marriage for the treatment group not only remains continuously higher but also shows an upward trend. In contrast, the probability for the control group remains stable or declines slightly. This widening gap post-intervention suggests that exposure to KP did not lessen the likelihood of early marriage among targeted girls; rather, the probability increased relative to the control group. These birth-year specific patterns, observed across all five rounds of NFHS data, align with the main finding of higher early marriage prevalence among its intended beneficiaries.

Conclusion

This study evaluates the impact of KP, a CCT programme designed to delay early marriage in West Bengal. Using a DID regression analysis, we find that KP has not significantly reduced early marriage rates. While programmes like the ‘Dhanlakshmi Scheme’ in Punjab successfully increased the number of girl childbirths (Biswas et al., 2023), the effectiveness of some initiatives may be influenced by complex socio-economic dynamics and entrenched cultural practices, much like the case with KP. Though KP has effectively increased girls’ school enrolment, its ability to delay marriage seems constrained by entrenched sociocultural norms and economic pressures that prioritize marriage over career opportunities for young women. These findings suggest that the programme’s design needs re-evaluation to address broader socio-economic factors and improve the quality of education to better align with labour market demands. Moreover, stricter compliance monitoring and financial oversight are essential to prevent misuse of funds in regions where early marriage remains culturally entrenched.

This study, however, has a few limitations. First, our reliance on nationally representative survey data introduces potential biases, such as recall errors and respondent ambiguity. Additionally, the lack of unique identifiers for KP beneficiaries in the NFHS data necessitated using age as a proxy for potential recipients, which may not fully capture programme participation. A limitation of the analysis is that the NFHS does not distinguish between government and private school attendance. Consequently, it is not possible to directly verify exposure to the schooling-related eligibility conditions of the programme. Our analysis also includes both never-married and ever-married women, but for the latter, we cannot capture the conditions of the households deciding to marry off young girls. This is a significant limitation, as early marriage decisions typically involve both the bride’s and groom’s families. Furthermore, the absence of district-level identifiers in NFHS-3 restricts our ability to control for district-specific unobservables through fixed effects. Finally, unobserved socio-economic and cultural factors may be influencing the prevalence of early marriage and the effectiveness of KP, pointing to the need for further research that integrates these factors into the analysis.

Data Availability Statement

Publicly available data sets were used in this study. These data can be found in the following link: https://dhsprogram.com/data/available-datasets.cfm, Accessed on 20 June 2024.