Could poverty interventions in the US be cost-effective for global health?
How does US Poverty compare to poverty in LMIC?
In Matthew Desmond’s book Poverty, by America, he shares the finding that 5.3 million Americans live below the global poverty line. This is $4 a day in the US, equivalent in quality of life to the $1.9 a day used by the World Bank for defining poverty in low-income countries.
18 million live in deep poverty—making under 50% of $13,000, the bare subsistence poverty line. We can safely say that their low income significantly reduces their life expectancy:
The gap in life expectancy between the richest 1% and poorest 1% of individuals was 14.6 years (95% CI, 14.4 to 14.8 years) for men and 10.1 years (95% CI, 9.9 to 10.3 years) for women.
According to this research, the difference between being in the 4th income quartile ($17k/year) vs the 3rd ($47k/year) is roughly 5 years. Following this nonlinear trend, we might guess that moving from the 0-5% range to the 5-10% range would improve life expectancy about 2.5 years.
While charitable interventions that directly pay for interventions are probably not cost-effective compared to those in low-income countries, policy interventions may be. Why? Because the American economy is so large, and our existing (and possible) tax basis is enormous. In 2022, the federal government spent 1.19 trillion on welfare programs, three orders of magnitude higher than Givewell’s budget. State and Local governments spend another trillion on welfare programs and hospitals. Therefore, making existing programs more effective, or expanding their funding through advocacy, could highly cost-effective, on par with direct health interventions in LMIC.
Well, is it? Let’s look at a few napkin math examples.
Example Intervention: Improving an existing program through non-legislative means
~38 million Americans received SNAP (food assistance based on income), and the government estimated in 2021 that 1 in 6 eligible did not receive it (7.6M).
Let’s assume half of those are children, and that if they had food security in childhood, they would get a benefit of 3 QALYs due to better health, education, and home stability. That represents ~10 million QALYs. This money is already theirs by law, they just aren’t accessing it.
There seems to be little impact of food security on life expectancy for adults
I couldn’t find life expectancy numbers for children, but there seems to be a relationship between food insecurity and chronic health conditions, so I feel okay about this QALY estimate.
An intervention might be organizing volunteer outreach through schools, churches, or food shelters; or automatic enrollment through employers (eg, Walmart helps its employees sign up for the Earned Income Tax Credit).
Organizations like Code for America and New America specifically work on technical and process interventions with government agencies to make benefits easier to access.
I’d guess that interventions like these are in roughly the $1M-5M range for cost. If they can increase uptake by just 10%, that’s 1M QALYs. Valuing a QALY at 100k, the expected ROI is 20,000, or $5 per QALY.
Example Intervention: Increasing the coverage for an existing program through legislative means
~15 million Californians (40 percent of the population) are covered by MediCal, effectively free health care for people in poverty (The income limit is $30k for one person and $60k for a family of 4).
Suppose that the income limit was raised, such that 1% more Californians were covered under MediCal. Suppose each of them benefited from 2 QALYs due to coverage. I’m guessing the QALY improvement would come from direct and preventative care, but also ability to work and preventing financial catastrophe → 3M QALYs. (Rough guess based on this research.)
MediCal is roughly $15k per person per year, so this expansion represents a $2.2B cost (about 1% of the CA budget)
Advocacy Path: Suppose EA worked directly with a state congressperson to get this expansion written into law.
The goal would be to work with a congressperson on the appropriate committee, to find this money.
When hospitals treat the uninsured in an emergency, they often do not get fully reimbursed, and the government covers some of that cost. Preventative healthcare should reduce this expense.
One possibility to come up with additional funds would be to pass a tax on the wealthiest to pay for this expansion. Eg, a property tax on homes valued at more than $5M, or a capital gains tax on Retirement portfolios over $5M. Lower wealth limits will introduce more opposition from voters.
To achieve this end, EA might do relationship building with relevant legislators, providing research, and possibly campaign donations, to state legislators aligned with its goals. So suppose EA invested 5M in relationship building and legislative support.
The reelection campaign budget of a CA state legislator is around 1 million, so contributions in the 50-100k range go a long way.
If this had a 50% chance of working, the ROI would be 300,000 or $0.33 per QALY.
Ballot Path: Suppose EA worked to get a proposition on the ballot that would pass this funding.
Suppose EA funded a CA ballot to expand Medicaid by passing the tax that focuses on the top 1% of earners.
On average, getting enough ballot signatures costs about $11M. Advertising, depending on controversiality, might be around $60M.
Suppose that this investment of 70M has a 50% chance of succeeding. The ROI would be about 2000, or $46 / QALY.
This may not convince you that we should start writing checks today! But to me, this is suggests that compared to other near-term global health work, health and anti-poverty interventions in America is a research area worth looking at. It seems that this was an area of inquiry for OpenPhil 10 years ago, but has not been seriously pursued (with the exception of anti-incarceration work, which I haven’t been able to find a clear impact metric for).
The rest of this series will explore:
A framework for assessing tractability and impact, and how to consider neglectedness in prioritization.
The pros and cons of choosing poverty advocacy in America as an EA cause area.
What sorts of interventions might be researched, with some rough prioritization.
Thanks for reading, and looking forward to any thoughts on this topic!
The next post, on prioritization, is here.
Unless I’ve misread it I don’t think the linked article shows this? It shows they are correlated, but not how the causation goes, and there are many clear candidates for common causes. For example, having a serious medical condition might make it harder for you to work, reducing your income, and cause you to die sooner, reducing your life expectancy.
I would expect lower income to reduce life expectancy somewhat but for things like health, pollution, IQ, drug addiction, conscientiousness etc. to be common factors, resulting in a causal influence lower than suggested by the correlation alone.
Very true, the study definitely does not provide a causal result, and the language I used about moving people out up an income tier definitely implied a causal attitude.
I also agree that health and pollution are major factors, and would argue that people with more income have more ability to improve these factors or mitigate their impacts. Eg, more income allows better access to health care, nutrition, moving to less polluted areas or being able to pay for Asthma treatment.
Thanks for calling this out!
Thanks for sharing this! You’ve convinced me that policy advocacy in the US could be really cost-effective, simply by leveraging the huge resources at stake. My main objection is to the ballpark 2-3 QALY gain for each person who receives food aid or health insurance. I understand that these are just simple calculations for illustration, but I think that added complexities will tend to make this estimate go down a lot:
I’d be surprised if, among Americans, food security alone is worth this much per person. Health outcomes in the US are horrendously unequal because of a nexus of strongly-correlated factors that also include poverty, unhealthy lifestyles, healthcare, drugs, etc. Removing a single one may not do much.
You’re assuming that any improvements in benefit-claiming lasts a long time (many years of food aid are required to ensure a child is never food insecure). In reality, counterfactual effects will decay over time. Less so for structural changes that make it easier for people to claim benefits.
Another point is that diverting public funds comes at a counterfactual cost. What would have happened to that MediCal money that would have gone unclaimed until your expansion advocacy? Maybe something less cost-effective (like schools, which form a third of CA’s budget), but with a non-negligible impact that should be subtracted from the effect of MediCal expansion.
Agree that this would be great and agree that it would face opposition.
QALY Assumptions
I agree, I was reflecting on the arguments and also felt that the QALY assumptions I made were very uncertain. The reasoning you point it is very sound; the 3 year estimate could easily be two orders of magnitude off. That would still leave it in a range worth looking at (I think Givewell targets something like $70 per QALY), but there’s a big question about how we could get more certainty on the impact.
Most studies on the subject are observational, and I think this points to the need for government programs to actually do some experimental analysis. But since these are programs passed by law, it might be ethically tricky to provide access to a benefit for one county and not for another. Looking at families who utilize a benefit vs not is also not experimentally sound.
Counterfactual impact
That’s a great point. I’m actually not sure how those funds get handled. I think they are typically earmarked in the budget, so they might not be automatically diverted. On the other hand, when legislators are creating a budget, they might factor in that 15% of eligible people won’t use a benefit.
Increasing budget with taxes on the ultra-rich
Yeah, I’m really curious about how hard this would be. Theoretically, a very rich person and their friends might band together to spend against this sort of proposition, but I’m not sure they would? I think a lot of progressive taxes fail because they their income limits are too low, and so there’s a significant voting block which is loud and well-organized, who will call their representative in opposition. Since there is a lot of concentrated wealth at the top, it makes sense to me to focus on going after that money. However, I have no idea how much a tax like the one proposed would actually raise.
Thanks so much for the great points!
Yes for me the 2 QALY benefit seems really unlikely, I would be instinctively thinking an order of magnitude less than that.
I also find a 50 percent chance of success in any kind of campaign like this pretty unlikely too, given the history of policy like this not getting through and opposing powers. Would be guessing more 5 to 15 percent range.
Nice article interesting one! Am a big fan of advocacy and think it can be super cost effective!
Executive summary: Poverty interventions in the US could be highly cost-effective and comparable to global health interventions in low- and middle-income countries, based on preliminary analysis.
Key points:
5.3 million Americans live below the global poverty line of $4 per day, with deep impacts on health and life expectancy.
Existing US poverty programs could be improved through better access and uptake at low cost, with extremely high estimated cost-effectiveness.
Legislative expansions of programs may also be feasible and have high estimated cost-effectiveness, albeit with more uncertainty.
More research is warranted on the tractability and impact of US poverty interventions.
Key considerations include assessing neglectedness, uncertainty over estimates, and comparing to other global health interventions.
This comment was auto-generated by the EA Forum Team. Feel free to point out issues with this summary by replying to the comment, and contact us if you have feedback.