International Child Sponsorship Impact on the Intended Choice of Acquiring a Higher Education Degree: The Case of Rural Mexico
Motivation
In Oaxaca and Chiapas, two of Mexico’s poorest states, the average adult has barely more than a primary school education: 7.5 and 7.2 years of schooling, respectively. This is striking, because the returns to higher education are positive and significant even for rural populations.
The standard explanation points to external constraints: low income, distance to schools, poor health. These are real, but a growing literature argues they are not the whole story. Internal constraints, aspirations, self-efficacy, self-esteem, grit, also hold children back, and may be at least as important in early development.
This paper investigates whether Compassion International (CI), a large holistic child sponsorship program, raises rural Mexican children’s aspiration to pursue a higher education degree. CI is unusual in that it combines material support (school supplies, healthcare, food) with an intensive socio-emotional program (~5–6 hours/week), including letter exchanges with international sponsors that broaden children’s sense of what futures are possible.
The central question is deceptively simple: does being sponsored raise a child’s aspiration to go to university?
The Identification Challenge
Answering this question is harder than it looks. Sponsored children are not randomly selected: CI targets the youngest and poorest households in the most deprived communities, and church affiliation plays a role. A naive comparison of sponsored vs. non-sponsored children would confound the effect of the program with these pre-existing differences.
in CI communities
Ages 12–15
N = 403
non-random
N = 163
✓ Aspiration Y1 observed
N = 240
✓ Aspiration Y0 observed
But observable differences are only part of the problem. More motivated families may seek out sponsorship; CI’s own targeting may select children with particular unobserved traits. This is selection on unobservables, and it means that even a standard instrumental variables approach is insufficient, because IV assumes the treatment effect is the same for everyone. If the children most likely to be sponsored are also those who benefit most from it (selection on gains), IV will give the wrong answer.
This paper uses a binary Roy-type model (Aakvik, Heckman, and Vytlacil, 2005) that explicitly models both the selection process and the potential for heterogeneous treatment effects, identifying three distinct treatment parameters: the Average Treatment Effect (ATE), the Average Treatment on the Treated (ATT), and the Marginal Treatment Effect (MTE).
Key Findings
ATT ≈ 17–20 percentage points. Sponsored children are estimated to be 17–20 pp more likely to aspire to higher education than they would have been without sponsorship. This is consistent with the broader CI literature. The effect is not statistically significant at conventional levels, partly due to the small sample and the complexity of the model.
No efficiency-equity trade-off. The MTE curve declines with the selection index, children most likely to be sponsored are also those who benefit most. CI does not face the typical dilemma between targeting the most vulnerable and targeting those with the highest gains.
Subjective income expectations are realistic but not predictive. Children aged 12–15 report income expectations that broadly align with census data, but perceived returns to education do not significantly predict aspirations. Other factors, internal constraints, social context, short-term thinking, may dominate at this age.
Study at a Glance
| Feature | Detail |
|---|---|
| Setting | Oaxaca and Chiapas, rural Mexico |
| Survey period | June–August 2017 |
| Communities | 8 total (4 with CI, 4 matched controls) |
| Age range | 12–15 years |
| Final sample | 403 children (163 sponsored, 240 non-sponsored) |
| Outcome | Binary: aspires to university degree (0/1) |
| Model | Binary Roy-type (Aakvik, Heckman, Vytlacil 2005) |
| Treatment effects | ATE, ATT, MTE |
| Journal | Education Economics (2025) |