Principal investigator: Omar Barroso Khodr
Authors: Katy Bergstrom and William Dodds
Original title: The targeting benefit of conditional cash transfers
Location of the Intervention: Mexico
Sample Size: The survey covers all households in the evaluation sample (i.e., all 506 localities) and contains extensive information about the households (including eligibility for the Progresa program), as well as information about each child, including age, sex, education, job availability, income, and school enrollment. In this way, the survey covers approximately 24.000 households throughout 1997, 1998, and 1999.
Primary Variable of Interest: School Enrollment.
Type of Intervention: Social Assistance Programs; Conditional and Unconditional Money Transfer.
Methodology: Study of elasticity, OLS and Instrumental Variables.
Summary
According to Bergstrom and Dodds (2021), Conditional Cash Transfers (CCTs) are a popular type of social assistance program that makes payments to families conditioned on investments in human capital for children. From a targeting perspective, compared to Unconditional Cash Transfers (ICTs), CCTs are costly because they exclude some low-income families, since access is linked to normal investments in children. However, we argue that the conditionalities on children's school enrollment offer an untapped targeting benefit compared to ICTs: CCTs direct money to families who forgo a specific amount of child income. The authors show that the magnitude of the targeting benefit relative to the targeting cost of CCTs is directly related to consumption differences between families with and without schooling and to two elasticities already popular in the literature: the income effect of an ICT and the price effect of a CCT. Additionally, the authors estimate these elasticities for a large TCR program in rural Mexico, Progresa, using the variation in transfers to younger siblings to identify income effects. In this context, Begstrom and Dodds find that the benefit of targeting is of a magnitude similar to the cost of excluding some low-income families; this implies that 33% of the Progresa budget should be allocated to a CCT instead of a UCT based solely on targeting criteria.
- Policy Problem
The study by Bergstrom and Dodds (2021) seeks to identify and solve a central problem in social welfare policies: how to optimally allocate a fixed budget between Conditional Cash Transfers (CCTs) and Unconditional Cash Transfers (ICTs). In particular, the authors analyze the so-called "targeting trade-off." CCTs, by requiring specific behaviors—such as school attendance—can better direct resources to the most needy families (the "benefit of targeting"), but also impose a welfare cost by restricting the choices of these families (the "cost of targeting"). Thus, the fundamental problem consists of determining whether the benefit outweighs the cost and, based on this, quantifying what portion of the budget should be conditional or unconditional, considering only the efficiency logic of targeting.
To address this issue, the authors develop a theoretical framework analogous to the economic literature on unemployment insurance, expressing the commitment to focusing in terms of estimable “sufficient statistics.” These elements include the average marginal utility of consumption among households that send and do not send their children to school, the school enrollment rate, and crucial behavioral elasticities—specifically, the income effect and the price effect of a child welfare program. These elasticities allow inferences on how changes in the design of transfers affect both school enrollment and the government's budget constraint.
The empirical application focuses on Progresa, a large-scale TCR program in rural areas of Mexico. To estimate the necessary elasticities, the study uses an innovative identification strategy: it explores the random introduction of the program to identify the price effect and uses transfers intended for younger siblings—whose enrollment is practically universal—as an exogenous source of variation in family income to identify the income effect. Among children of high school age, the results indicate that the income effect corresponds to 48% of the price effect.
Combining these elasticities with household consumption data, the study quantifies the cost-benefit ratio of targeting in the Progresa program. The authors conclude that, in the program's current design, the benefit of targeting is substantial—equivalent to 79% of the cost of targeting. Furthermore, they estimate that, considering only the improvement in resource allocation to families with higher marginal utility (i.e., greater need), 33% of the Progresa budget should be allocated to CRT (Conditional Remuneration Rate), and the remainder to Universal Income Transfers (UI). This result stems from the fact that families with school-age children tend to be poorer—given that working children contribute significantly to family income—and the low initial inequality among eligible families, which makes the conditionality mechanism particularly effective in identifying needs.
In summary, the study rigorously addresses the problem of how to design income transfer programs rationally, going beyond the simplistic debate between TCR and TUR. It offers a measurable framework that allows policymakers to determine the ideal combination of conditionalities within a fixed budget, balancing the goal of poverty reduction with the cost of imposing behavioral requirements.
- Policy Implementation Context
The context of this evaluation is Progresa (later renamed Oportunidades), a pioneering program in Mexico and one of the first large-scale Conditional Cash Transfer (CCT) initiatives in the world, launched in 1997. Progresa was designed with a dual objective: to alleviate immediate poverty through cash transfers and, simultaneously, to break the intergenerational cycle of poverty by promoting the human capital of children. The program's main component—and the focus of this analysis—consisted of scholarships paid directly to mothers, conditional on their children's regular school attendance (at least 85% of school days) from the 3rd to the 9th grade. The transfers were substantial, often representing a significant portion of family income.
The evaluation utilizes a unique Randomized Controlled Trial (RCT) design, incorporated into the initial implementation of Progresa. In 1998, 506 eligible locations were randomly assigned: 320 received the program immediately (treatment group) and 186 were scheduled to receive it two years later (control group). This randomization constitutes the main source of experimental variation. The study employs detailed data from household panel surveys conducted in 1997 (baseline), 1998, and 1999, encompassing approximately 24.000 households in the treatment and control locations. This data includes comprehensive information on demographics, consumption, child school enrollment, and labor supply.
According to Bergstrom and Dodds, the main empirical challenge of the study is isolating the separate “price” and “income” effects of the transfers. Although the randomized introduction of Conditional Cash Transfer Programs (CCTPs) clearly identifies the total effect of the program—which combines both effects—Progresa did not explicitly offer Unconditional Cash Transfers (ICTs). To circumvent this limitation, the authors develop an innovative identification strategy that exploits the near-universal school enrollment among children under 12 years old. Since the attendance of these children is already around 98%, the grants intended for them do not alter school behavior and, therefore, function in practice as unconditional transfers to the family.
Thus, for a focal child aged 12 to 15, the variation in unconditional income stems from the number and eligibility of their younger siblings in the treatment locations. The “price” (incentive) effect of TCR is identified by comparing the enrollment of older children in treatment locations with that of children in control locations whose younger siblings are ineligible.
Finally, the combined effect of “price + income” is identified by comparing older children in treatment settings with those in control settings whose younger siblings are eligible. The pure income effect is then obtained by the difference between these two estimates. This approach allows us to estimate, within the same experimental context, two fundamental behavioral elasticities: the enrollment response to a pure monetary transfer (income effect) and the response to financial incentives for schooling (price effect).
- Evaluation Details
The study's evaluation is based on a formal theoretical framework that models the fundamental relationship between conditional and unconditional cash transfers. Bergstrom and Dodds develop a two-generation model in which parents decide whether to send their child to school or to work. Families are heterogeneous in both parental income and child ability (or parental altruism). A utilitarian social planner has a fixed budget to distribute among a pre-identified group of eligible poor families. This planner does not consider family income or child ability, but rather the binary decision of schooling. Thus, their policy tools consist of a uniform Unconditional Cash Transfer (ICT) for all eligible families and a uniform Conditional Cash Transfer (CCT), paid only to families who send their children to school.
The core of the analysis stems from the planner's first-order condition, which balances a "benefit of targeting" with a "cost of targeting." The benefit of targeting corresponds to the welfare gain resulting from directing additional resources (the CRT) to families with school-age children, who forgo income from child labor and therefore may have higher marginal utility of consumption. The cost of targeting, in turn, arises from the need to reduce the universal IRT to finance the CRT, which disadvantages families without school-age children. The main theoretical result of the model is that a pure IRT can be suboptimal. If, in a scheme based solely on the IRT, the average marginal utility of consumption is higher among families with school-age children than among those without children in that age group, then the introduction of a CRT improves the targeting of resources to the most needy families. This condition is more likely when the income lost from child labor is high and income inequality among eligible poor parents is relatively small.
Finally, the authors translate this theory into an empirically implementable framework based on sufficient statistics. They demonstrate that the benefit-to-cost ratio of targeting—and therefore the ideal combination of Conditional Cash Transfer (CCT) and Universal Cash Transfer (UCT)—can be calculated from five observable or estimable quantities: (1) the consumption distributions of households with and without schooling; (2) an assumed curvature of the utility function (risk aversion coefficient); (3) the school enrollment rate; (4) the average income effect, which measures how enrollment responds to a purely monetary transfer; and (5) the average price effect, which measures how enrollment responds to the financial incentive of the CCT. Crucially, the model shows that the budgetary trade-off—that is, how much the UCT must be reduced when the CCT is expanded—is a function of the enrollment rate and these two elasticities. This elegant formulation allows for the calculation of the optimal policy without estimating the complete structural model, relying only on essential behavioral parameters and distributional data.
- Method
The study's methodology is based on an ingenious identification strategy to separate the pure effect of income transfer from the effect on the "price" of school enrollment within the Progresa program in Mexico. Since the program offered only Conditional Cash Transfers (CCTs), the authors creatively explore the very design of Progresa and the near-universal enrollment among young children to generate exogenous variation in transfers that, in practice, function as unconditional transfers. The main finding is that, for children under 12 years old, enrollment is practically universal, regardless of the transfer. Thus, the CCTs intended for this age group effectively operate as Unconditional Cash Transfers (ICTs) for families.
For the target sample—eligible children aged 12 to 15—the relevant variation in the TCRs stems from two elements: (i) the random implementation of Progresa (treatment versus control locations) and (ii) the transfer schedule, which varies by school year and child's sex. The variation in the IRRs, however, is generated by the presence and number of younger siblings (under 12 years old), since the transfers intended for these siblings constitute an exogenous income shock, unrelated to the older sibling's decision to attend school.
In this context, the core of the identification is a triple difference approach, illustrated by the authors with a simple example: comparing 13-year-olds who have a 7-year-old sibling (ineligible for the benefit) with 13-year-olds who have an 8-year-old sibling (eligible). By comparing the enrollment differences between treatment and control groups for these two types of families, before and after the introduction of the program, the authors isolate: (i) the price effect (the direct incentive offered to the 13-year-old) and (ii) the combined price + income effect (the incentive to the 13-year-old plus the transfer received by the younger sibling). The difference between these two effects provides the isolated income effect.
Empirically, the authors estimate a regression model on an unbalanced panel of children aged 12 to 15 years, from 1997 to 1999, with fixed effects of child and year. The dependent variable is school enrollment, and the main regressors are: (i) the per capita value of the TCR offered to the child. (tc), and (ii) the per capita sum of transfers intended for siblings under 12 years of age. (you), which functions as a measure of unconditional transfer. The coefficients of these variables capture, respectively, the price and income effects. To mitigate measurement errors and possible endogeneity in the declared paternal income — included as a control — the authors instrumentalize it with the median hourly wage of the locality.
Finally, estimates using instrumental variables reveal positive and statistically significant effects: a weighted increase of 1 in the per capita CTR raises school enrollment by 0,8 percentage points (price effect), while a weighted increase of 1 in the per capita IRR raises enrollment by 0,4 percentage points (income effect). The authors demonstrate that these results are robust to several alternative specifications, including controls for sibling composition and for transfers intended for older siblings.
- Main results
This study challenges the prevailing political consensus that Unconditional Cash Transfers (UCTs) are inherently superior to Conditional Cash Transfers (CCTs) in efficiently allocating resources to the families that would value them most. Theoretically, the authors demonstrate that there may be a "targeting benefit" associated with CCTs, capable of outweighing the "targeting cost" they impose. This cost stems from the fact that conditioning transfers on certain behaviors (such as school attendance) can exclude some needy families who do not meet the condition, even if they highly value the resources. The main theoretical conclusion is that when families who meet the condition—for example, by sending their children to school—systematically exhibit lower average consumption than those who do not, the CCT directs resources to a poorer subset of the population. This targeting benefit may be strong enough to justify allocating part of the budget to a CCT, even considering only targeting criteria, contradicting the conventional view.
Empirically, the analysis of the Progresa program in Mexico quantifies this mechanism. Using consumption data, utility parameters, and behavioral estimates, the study calculates that, for children of secondary school age, the targeting arguments alone justify allocating approximately one-third of the Progresa budget to a Conditional Cash Transfer (CCT) instead of a Universal Cash Transfer (UCT). This result is driven by the observed pattern that families with school-age adolescents had lower average consumption than families without adolescents in that age group—implying that the school attendance condition functioned, in practice, as a mechanism for directing resources to the poorest families.
- Lessons in Public Policy
Finally, the study argues that the main public policy lesson is that the design of income transfer programs should be guided by a careful assessment of local conditions, and not by a generalized preference for Conditional Cash Transfers (CCTs). The authors show that a CCT tends to be ideal when beneficiary families are, on average, poorer and when there is substantial underinvestment by parents in education—that is, when there is a relevant gap between the social and private returns of schooling. In contrast, a Universal Cash Transfer (UCT) is preferable when eligible families are relatively wealthier and educational underinvestment is minimal. In intermediate contexts, a combination of CCT and UCT may be the most efficient solution.
Thus, the authors emphasize that policymakers should carefully evaluate specific factors, such as the difference in consumption between beneficiary and non-beneficiary families, the magnitude of educational underinvestment, the curvature of the utility function, the intensity of behavioral responses, and the government's distributional objectives. The study demonstrates that the targeting benefit provided by TCRs can be quantitatively relevant and, therefore, should be explicitly incorporated into the design and evaluation of social programs.
