The linked google document and its companion spreadsheet were my entries to the GiveWell Change our Mind Contest. I argue that the cost-effectiveness of GiveDirectly’s (GD’s) cash transfer program in GiveWell’s current analysis is undervalued because the health, wellbeing, and general equilibrium benefits are not explicitly modeled. The intro section I have included below within this post - see the link for the full argument.
I show that GiveDirectly’s cost-effectiveness could increase 6.7x if health, wellbeing, and general equilibrium effects are explicitly modeled. The majority of the increase comes from reductions to child mortality in the health section. The values I calculate are intended as a proof of concept — I know that I have not been able to use the same level of rigor GiveWell would, and in particular my moral weights have not been calibrated among a group of people and so would need to be further refined. However, I believe they are good enough to demonstrate the value of explicitly modeling health, wellbeing, and general equilibrium effects within GiveDirectly’s cost effectiveness analysis.
A summary of the major sections follows:
- Health: The adjustments to capture health benefits of cash transfers in the current model are too small, but should be explicitly modeled and considered one of cash’s primary outcomes. Cash prevents child and adult deaths from malnutrition, both near-term, as recipient families purchase food, and longer-term, as a result of increases to income that allow them to continue buying food. It also prevents stunting, wasting, and other health issues that result from malnutrition, and may impact other pathways to better health outcomes.
I explicitly model the moral weight of reductions in mortality for children under five due to a transfer, in the immediate term only, using the effect size from GiveDirectly’s study in Rwanda. This has the largest impact of all changes that I model, because of the large moral weight associated with reductions to child mortality. I also model reductions to child stunting and assign them a moral weight. This modeling increases GD’s units of value by 654, for a new total 4.5x higher than the original of 172. The same logic could be followed to explicitly model other health benefits, which I believe would increase the cost-effectiveness further. - Wellbeing: There is no moral weight for improvements to mental health or wellbeing in the current cost-effectiveness model, but these improvements are substantial for large, unconditional cash transfers. I argue that there are wellbeing benefits independent of the benefits to increased consumption that are not being captured. I suggest one way of defining a moral weight for wellbeing and quantifying the wellbeing impacts independent of consumption that results in 118 additional units of value, or a 70% increase on the original value.
- General Equilibrium: The general equilibrium (GE) study conducted on GiveDirectly’s program found an increase to consumption among non-recipient households that was roughly equivalent to the increase to consumption found among recipient households after 18 months. I model this for an additional 211 units of value.
When taken together, explicitly modeling the health, wellbeing, and general equilibrium impacts of GiveDirectly’s cash transfers could increase GD’s cost-effectiveness by 6.7x or more, depending upon the moral weights assigned. I recommend following along in the spreadsheet as you are reading the narrative, as it makes it easier to follow the math. All changes I have made to the original cost effectiveness analysis are highlighted in yellow.
I then close in section D with a discussion of other benefits associated with incorporating wellbeing into GiveWell's moral weights. While these additional considerations do not impact the cost-effectiveness of any charity in the near-term, I decided to take the opportunity provided by this contest to include them because I believe they are important in the long term for GiveWell. The relatively smaller change to GiveDirectly’s cost-effectiveness analysis may not merit the effort of incorporating wellbeing into your moral weights, but I believe these other factors may.
It's worth noting that the fiscal multiplier in the GE cash transfer study has a point estimate of 2.4, but their confidence intervals are wide and they can't reject that the fiscal multiplier is 1 (ie no positive spillovers). The study is also puzzling in finding no effects on prices even though the local money supply increases dramatically. The authors do not have a clear explanation for why there is no price increase from cash transfers, and I think we should have a pretty skeptical prior about that result; if we do then we might conclude that our best guess of the welfare from large scale cash transfers is much lower than estimated in that study.
Thanks for this thoughtful post Carolina! I would second Karthik's note here - I think there have also been a few other GE studies which show contradictory results, so it's not clear that the spillover effects would be positive once inflation and exchange rates effects are taken into account. Others have also raised concerns about possible negative pyschological spillovers, though from memory I think GiveDirectly typically provides cash to everyone in a village, which may mitigate this issue.