Working Paper: NBER ID: w19863
Authors: Dean Karlan; Leigh L. Linden
Abstract: Commitment devices offer an opportunity to restrict future choices. However, strict commitments may deter participation. Using a school-based commitment savings program for children to save for educational expenses in Uganda, we compare an account fully committed to school expenses to an account with a weaker commitment (funds withdrawn in cash, rather than a voucher). Children save more in the weaker commitment treatment arm, and when combined with parental outreach spend more on educational supplies and score 0.10 standard deviations (standard error =0.04) on test scores. The fully committed account yields no such educational improvements, and neither account finds impacts on secondary or downstream outcomes such as attendance, enrollment, or non-cognitive skills.
Keywords: commitment devices; savings; education; Uganda
JEL Codes: D12; D91; I21; O12
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| weak commitment treatment (C78) | increased savings (D14) |
| increased savings (D14) | greater spending on educational supplies (H52) |
| greater spending on educational supplies (H52) | improved test scores (I24) |
| fully committed account (Y20) | no educational improvements (I29) |
| weak commitment treatment (C78) | improved test scores (I24) |
| weak commitment treatment (C78) | participation and engagement (I24) |
| strict commitments (D86) | negative effect on participation and engagement (I24) |
| neither account (Y70) | impacts on secondary or downstream outcomes (I14) |