Loose Knots: Strong versus Weak Commitments to Save for Education in Uganda

East Africa

Uganda

Labour Market

Microcredit

D Karlan, L Linden

2016

Karlan, Dean, and Leigh L Linden. “Loose Knots: Strong Versus Weak Commitments to Save for Education in Uganda.” (2014).
View publication

N. A.

2016

Supersavers

Late 2009-Early 2011 (baseline and enline dates)

The intervention aimed at facilitating and encouraging the practice of children saving for education and, through saving, improving overall academic performance and supporting students’ continued enrollment.

Students in grades five, six, and seven, i.e., the last three years of primary school (higher risk of dropout).

A total of 136 primary schools were randomly assigned to one of three groups: a strong commitment savings account (funds could be withdrawn no earlier than the end of the term, and had to be spent on educational items through a voucher that we provided), a weak commitment savings account (funds could be withdrawn no earlier than the end of the term, but were available in cash, to be spent as individuals wished), or control. For both treatments, students could deposit cash into an account. At the end of each trimester, they were able to use their cash or vouchers to purchase school supplies at a fair.

• Saves money
• School supply
• Parental involvement; School Absence.

RCT

Study design: there were four treatment variations, a 2×2 design: “cash” or “voucher” for the withdrawals, and “Parent Outreach” or “No Parent Outreach”. For the cash treatment arm, students received, in cash, their savings from one trimester at the beginning of the next trimester. They could then spend the funds at their discretion—at the markets provided on the disbursement day (thus “making it easy” to spend on school supplies) or elsewhere. The voucher treatment arm, on the other hand, employed a stronger commitment – students had to buy educational products or services at the market on the disbursement day. In both variants, children could also re-deposit savings for the next trimester.
Sample: Two samples of students were selected, as well as data at both the classroom and student level. The first sample for analysis includes all students present in class at baseline and then all students present in class at the endline. This provides a sample frame of all students attending school. Second, the authors created a representative, longitudinal sample of students identified before treatment assignment. If the intervention affected attendance or enrollment, the first sample frame will be subject to sample selection biases, both on entry and exit. Thus the second sample provides information on a smaller subset of students that were tracked regardless of whether or not they continued to be enrolled in the original schools.
Survey Instrument: the authors administer a baseline and an aptitude test during the final trimester of 2009. The endline survey and an exam were applied at the beginning of the third trimester of 2011.

The comparison of an account fully-committed to educational expenses and an account with a weaker commitment (funds withdrawn in cash, rather than a voucher) leads to the conclusion that the weaker commitment generates higher account savings, and when combined with a parental outreach generates higher educational supplies expenditures and 0.11 standard deviation higher math and language test scores. No effect is found from the fully-committed account, and no effect for either account on attendance, enrollment, or non-cognitive skills.

Child Labour

Individuals

Local

There is no impact from the program on child labour, either hours worked or total wages.

The comparison of an account fully-committed to educational expenses and an account with a weaker commitment (funds withdrawn in cash, rather than a voucher) leads to the conclusion that the weaker commitment generates higher account savings, and when combined with a parental outreach generates higher educational supplies expenditures and 0.11 standard deviation higher math and language test scores. No effect is found from the fully-committed account, and no effect for either account on attendance, enrollment, or non-cognitive skills.

Paper has similarities with intervention applied in Ghana in 2018 (paper # 565). One characteristic of the study is that although the accounts are framed as the children’s accounts, the authors could not rule out that some of the funds were parents’ funds.