Our policy question is: when is it optimal for an organization such as Give Directly to transfer money to recipients as soon as possible, and when is it better to delay giving some of the money until the recipients experience an adverse event causing a loss of income.
We’ve found this paper, but it only has aggregated mean/variance quantities across households and only one type of shock. Any pointers would be appreciated!
(Some quick thoughts hastily written based off some class papers I wrote a while back.)
One dataset that pops to mind is the India Human Development Survey. This is a rich household-level dataset that includes total household monthly income (disaggregated by source) and if I recall right, also tells you what month it is. These are time-intensive to work with, but I imagine a few others datasets like this exist in the world. And you can estimate “income” per month with them.
My guess is you’ll get obvious insights from this, like income dropping during cold / dry seasons in more agricultural-dependent villages.
That said, my gut feeling with the policy question is that sending cash transfers sooner is better. A few reasons:
The book Portfolios of the Poor suggests rural households, especially poor ones, are good at managing their own finances and spreading out their resources over time (or consumption smoothing as economists call it).
Household debt (not captured in income) may accumulate interest, and interest rates can be exorbitantly high.
Idiosyncratic shocks (like a family illness) are hard to surveil and predict from afar. So it’s hard to implement strategic delays
Extreme systemic shocks like very harsh droughts / floods may temporarily constrain food / fuel supply reducing effectiveness of cash transfers at that time.
Households may want to stock up on durable foods like grains or oils ahead of time
Counterintuitively, giving money while households have high-income may push them over a wealth-threshold that lets them make durable investments (roofs, goats, farm equipment). This may be welfare-enhancing in the long run. I think this goes by the “lump-sum” effect in the cash transfers literature.
Hope this helps. Happy to chat about this more.
Thank you for the comment, I learned a lot from it. Would appreciate to hear what you think about my responses.
I think the first point about consumption smoothing is critical. My reading of the literature is almost the opposite—that although the poor find ingenious ways to save, their ability to smooth consumption is very limited. I wonder why that is. Maybe it’s because Portfolios of the Poor focuses on Bangladesh, India and South Africa which are much more developer financially than some other countries. But we will need to think about this some more.
The point about debt is very important and well taken. We will need to consider not just income but wealth more generally. I’ve corrected the question.
Regarding idiosyncratic shocks, our thought was that family level income/consumption/wealth data allows us to measure the frequency of idiosyncratic shocks. In addition, an Organization like GiveDirectly must in any case have direct communication with potential recipients to be able to transfer the money, which implies family illness may be observed.
Regarding systematic shocks, out thought was that it is easier for the donor to convert cash to goods than for the affected people. So if what they need in a drought is grain and not cash, the donor could accommodate this quite easily.
Hi Niki, glad to hear it helped. Here’s some more thoughts. Can’t promise they’re any good.
Yes, I agree the consumption smoothing point is critical. I could have worded my answer a bit better. What I meant to say is that rural households are good at trying to smooth consumption given their situation. That can still be a low overall ability given how sporadic income can be. The crux, I suppose, is whether we trust the households to smooth their own consumption or if we should make the decision for them. If we think the households are better able to make the decision, we send the transfer now. If we think they’re worse (perhaps for lack of willpower reasons), then we delay.
From the dataset I linked you, it’s tough to decompose the idiosyncratic shock from just family-level income / consumption / wealth data. You’d want to correct for village-level shocks and seasonality and so on. I believe it’s do-able but noisy. And it’s usually cleaner to simply ask the families whether they suffered from X adverse event in the past 12 months (which the India Human Development Survey does ask).
To do this more cleanly, you’d want a dataset that interviews each family multiple times in a single year. Then you could see the variability in income / consumption / wealth over time per family.
I’m not familiar with how GiveDirectly does their operations, but I’d be surprised if the communication was frequent enough that GiveDirectly staff could observe and react to every family illness in real-time.
On systemic shocks, you might be right. I was envisioning a bad natural disaster where the logistics of shipping more food out to villages is temporarily clogged up. But perhaps NGOs are more crafty that I think.