Shove vs. Nudge vs. None

When I first read about this new paper by a slew of economists including Esther Duflo, it was presented as part of the wave of evidence that has recently come out saying that unconditional cash transfers are just as effective at changing behavior as conditional cash transfers. The primary difference being that monitoring costs were significantly smaller, needier households would be more likely to get assistance, and there would be more flexibility in what individuals will spend the money on, likely, as they’re not being asked to do any one particular activity or investment with it.

It may be a matter of semantics, but I don’t think that this paper is actually making that claim (nor do I think they authors are really making that claim). One of the problems with measuring effects of unconditional cash transfers is that flexibility. Because individuals can spend the money where they deem it most useful or necessary, aggregate effects, or averaged effects tend to be small or even zero. In the simplest of terms, if I give three people $100 and one spends it on new shoes to plow his fields in, one spends it on school fees, and one spends it on hospital bills, the first may have a better crop output, the second may spend more time in school, and the third may be healthier but on average, income, education, and health effects are small for the group. They may all be better off, or perceive themselves as better off, but as they are better off on different metrics, we can’t observe the effect.

This idea of targeting “labeled cash transfers” as opposed to conditional ones or unconditional ones is an attempt at getting to somewhere in the middle. If we label the transfer, it implies it’s for a specific purpose, which means that we should be able to see the effect on a single metric. We know that some individuals will use the cash transfer for something other than what is labeled, but likely compliance will be high without significant monitoring costs.

Overall, though, it’s hard to imagine that recipients don’t imagine they are in some way being monitored. I can’t imagine being handed money, told explicitly it was for school, and then going and spending it on something else. Even if I were up for that, I’d be afraid of getting caught, or not being eligible for a subsequent payment. So, while this program reduces monitoring costs, which are high and probably not very effective, I don’t think this paper shows that unconditional cash transfers are as good as conditional ones, but rather that labeled cash transfers are as good as conditional ones but with fewer costs associated with them.

I also don’t doubt the principle in theory that unconditional transfers are just as good. Basic economic theory says that individuals are rational and though I may doubt that in principle, I’d guess that on average individuals will use additional funds to make themselves better off as it makes sense to them. If someone wants new shoes and someone wants to send their kids to school, and someone needs to pay hospital bills, they likely know better what will increase their own welfare. But I do think an aggregate effect on welfare is more difficult to measure with the unconditional transfers.

Conditional Cash Transfers (CCTs) have been shown to increase human capital investments, but their standard features make them expensive. We use a large randomized experiment in Morocco to estimate an alternative government-run program, a “labeled cash transfer” (LCT): a small cash transfer made to fathers of school-aged children in poor rural communities, not conditional on school attendance but explicitly labeled as an education support program. We document large gains in school participation. Adding conditionality and targeting mothers make almost no difference. The program increased parents’ belief that education was a worthwhile investment, a likely pathway for the results.

Benhassine, Najy, Florencia Devoto, Esther Duflo, Pascaline Dupas, and Victor Pouliquen. Turning a Shove into a Nudge? A “Labeled Cash Transfer” for Education. NBER Working Paper 19227.

Advertisements

CCTs and Crime

The connection between poverty and crime is both well-established and notoriously difficult to distentangle. We know that high-crime areas are likely to be poorer than low-crime areas, and yet we don’t usually profess that crime causes poverty, although a certain blogger/writer team of economist and journalist is quick to remind you that crime doesn’t pay. We might expect poverty to cause crime for a number of reasons–idleness leads to thrill-seeking, social norms make stealing appear common or acceptable, families may not be able to feed their families without stealing–but separating one effect from the other is incredibly difficult.

In a careful and very well executed new paper by three economists at the PUC-Rio, crime is in fact lowered in the face of conditional cash transfers, or a directed attempt to put more money in the hands of low-income families while simultaneously requiring their kids to go school/not work during school hours. The authors exploit the expansion of the program–to pay benefits to families with older children–to causally identify the effect of additional income on crime.

The authors find that expanding the Bolsa Familia program to include 16- and 17-year olds did have a dramatic, causal effect on crime rates.

My primary question on the paper has to do with the expansion. Because the program had already been in place for some time, many families lost some income when their children turned 16 and thus were no longer eligible for benefits. Many of these same families would regain benefits with the expansion. So, did crime increase as these children aged out? Surely there’s some variation in average age and distribution of children in the program by school, so we should be able to at least speculate on whether there is something about turning 16 and 17 that makes one particularly prone to criminal behavior, or whether leaving the program leads to more behavior. Perhaps we can’t identify it the same way causally, but it’s an important dimension, I think.

The second problem I have is stylistic: a clear link to a number in a table with words such as “the program expansion lead to an average X% decrease in crime” would have helped make reading easier.

h/t: @franciscome

Cited: Laura Chioda, João MP de Mello and Rodrigo R. Soares. “Conditional Cash Transfer Programs: Bolsa Família and Crime in Urban Brazil.” PUC Working Paper No. 559.