Abstract

Evaluating the impact of public policies requires identifying differences in observed outcomes that can be causally attributed to specific interventions. This study examines the effects of the Bolsa Presença Program, an initiative of the Government of the State of Bahia (Brazil) implemented by the State Department of Education, which provides financial assistance to socioeconomically vulnerable families enrolled in the national social registry (CadÚnico), with the goal of reducing school dropout in the state public education system.
The primary objective is to estimate the program’s impact on reducing school dropout and improving students’ academic performance. To address the challenges posed by time-dependent interventions and dynamic treatment assignment, we employ causal inference methodologies based on propensity score weighting. Specifically, inverse probability weighting (IPW) and overlap weighting (OW) are used to adjust for systematic differences between treated and untreated students over time.
Given the potential for heterogeneous treatment effects across population subgroups, the analysis incorporates causal subgroup analysis within a marginal structural model (MSM) framework. This approach allows us to estimate both baseline and time-varying treatment effects, while simultaneously identifying heterogeneity across distinct student profiles. By integrating subgroup-specific characteristics with the temporal dynamics of the intervention, the proposed framework provides a nuanced understanding of how and for whom the program is most effective.
The empirical results indicate positive effects of the Bolsa Presença Program on students’ academic performance trajectories and school retention, while also revealing substantial heterogeneity in impacts across the analyzed subgroups. These findings contribute to the methodological literature on policy evaluation with time-dependent treatments and provide evidence-based insights to inform the design and targeting of education policies.