A nice new paper  by Dean Karlan and Leigh Linden looks at different commitment devices to get primary school kids (yes kids) to save in Uganda. The core part of all of their interventions consists of giving participating schools metal lock boxes. An implementation team visits schools weekly to check in and to "encourage savings." Security is tight; each box has two padlocks -- with one key held by a parent representative and the other the bank they were working with. Each kid who is participating gets a passbook and can deposit weekly or daily, depending on the school choice. At the end of the trimester, the box is opened and the money goes to the bank and into an account. One interesting fact: this account bears no interest, and inflation is running about 10 percent during this time, so there is negative real interest. At the start of the next trimester, the bank representative and the implementation team return to the schools for payouts. And, at the same time, to up the nudge, the implementation team organizes a mini-market where kids can buy school supplies or things like practice exams and tutoring sessions.
Layered on top of this core are four variants of treatment (in a 2X2 configuration). Schools are randomly assigned to cash or voucher treatments -- the cash is obvious, the voucher is another nudge: it's only good for stuff sold at the school mini-market. The other set of treatment gets the parents involved. In a random subset of schools, parents of sixth and seventh grade kids were invited into meetings (with the obligatory soda and snack) which talk about how parents can support their kids' education, with a particular mention of the school savings program.
So what happens? Kids in the cash program save significantly more than kids in the voucher program. So for these kids, the additional tying of the hands is a nudge too far. Average savings for the kids across the two years ranges from 111 to 289 shillings (that's about 5-10 cents) and about 44 percent of kids actually participate. The crossover parental outreach intervention had no impact on savings levels.
Karlan and Linden then trace through the impacts of these savings. They look at impacts on school supplies (measured through classroom visits as well as surveys), parental involvement, savings attitudes and school fees payment. And this next stage of impacts is only significant for an index of school supplies (an index made up of the fraction of kids in the classroom with uniforms, notebooks, math sets, and shoes). Interestingly, this factor is only significant for the treatment group that gets the cash treatment plus the parent outreach. So it looks like if mom and dad know more about this savings program, there might be a little nudge at home to spend it on school supplies.
So do the kids learn more? Karlan and Linden & co. run these kids through a one hour test at baseline and then at follow-up. And they do learn more: grammar goes up by 0.15 sd and reading by 0.11 sd. Ironically, despite this being a savings program math doesn't move (maybe if they paid interest?). And in good comparative fashion, Karlan and Linden stack up the cost effectiveness of this intervention ($2.24 per student per year) and it looks pretty good relative to a number of other studies.
As with a lot of behavioral papers, this study raises a bunch of interesting questions for follow up. What matters most here – is it the markets in the schools that make school supplies super easy to buy? How much does the encouragement of the implementation team matter? What's going on with the parents? And what's going on with how they are interacting with their kids? Are the kids leveraging parental money with their savings when it comes time to buy the school supplies? Food for thought -- and for further work.