A lack of money is the definition of being poor, but it isn't the cause of being poor. The causes lie elsewhere -- skills, intelligence, impulsiveness, criminality, addiction.
Instead of giving out money, it appears the better path will prove to be (1) education to provide skills, (2) treatment to curb bad behavior and (3) more law enforcement when incentives don't work.
Kelsey already pointed out that participants don't spend their extra money irresponsibly, so we should expect there to be a positive impact.
You can explain poverty however you like as a result of personal failing. But consider the marginal person - someone just unintelligent, ignorant, impulsive, lazy enough to be poor (in your view) but not so much of any of those things that they are destined to be poor in every circumstance. What do we expect there?
Why wouldn't money improve that person's life? And if it does not, why does your job training program have any better chance?
It's true that there are non-monetary things society can do to improve people's lives and certainly that people can do to improve their own lives. But nothing about that explains away these cash results.
TL;DR The best estimated study shows the control group increases income by 80% (!) The people in the study are already on the track which makes the cash less useful. This is why you wouldn't see them looking to upskill or change onto a new path with cash.
Let's focus on the best controlled study by Open Research: I read the same tables and see a story about upward mobility with and without cash.
1) Participants’ incomes rose a lot.
At baseline, individual income averaged about $21k and household income $30k. By year three, excluding the transfer itself, the control group reached $38.2k individual and $51.6k household—roughly +80% and +70% nominal gains in just three years. The treatment group landed at $36.3k and $45.7k without counting the transfer, and at $48.3k and $57.7k with it.
2) That counterfactual growth is a HIGH bar to beat.
When the “no-program” path delivers ~70–80% nominal income growth in three years, it’s unsurprising that short-run labor-market outcomes don’t show big incremental effects from a modest transfer. This is especially true given the period’s unusually hot job market.
3) The sample looks financially capable, not financially excluded.
Baseline total debt averages about $41k. Measured effects on credit categories are tiny (the only notable one is a modest uptick in auto loans). That’s consistent with borrowers already having access to standard credit markets—i.e., many investment/budget constraints were already relaxable via lenders. Cash is helpful, but it isn’t unlocking a closed door.
4) Education levels are relatively high; there’s no obvious “low-hanging fruit.”
Roughly 95% hold a HS diploma/GED or more. Combined with the income gains in the control group, this looks like a motivated, reasonably skilled cohort moving up the wage ladder as expected.
5) Put the magnitudes in global context.
The baseline individual incomes are already in the neighborhood of OECD GDP-per-capita levels like Portugal/Japan, and by year three the group sits near Germany/Canada/Australia ranges. Apples-to-oranges caveat noted—but it’s useful for calibrating expectations where these people are in a labor-leisure trade-off being richer than the average citizen of developed economies.
6) Poverty dynamics matter for interpretation.
Outside the study, U.S. poverty spells are often transitory: roughly half exit within a year, with exit probabilities falling the longer a spell lasts. That pattern fits what we see here: many people are on an upward trajectory already.
*In the spirit of the magazine I framed this as an argument
Kelsey, thanks for putting this article together! I’m curious if you reviewed or discussed with researchers the studies on the impact of the expanded/2021 Child Tax Credit? I gather that those are not RCTs, but is there anything to gain from the observational data on that large scale policy intervention (and withdrawal from it)?
Yeah, I was also curious about this. The CTC cut child poverty nearly in half in 2021, and it rebounded back up to prior levels in 2022 post-expiration — that seems like a pretty strong effect even if it’s not a RCT.
But there’s a difference between literally having more money (thus pushing you above the poverty line!) versus the other effects they’re evaluating — cognitive gains, maternal health, employment, stress, etc.
Yep! I looked at some of the work there and I agree with you both that the ECTC worked well. And not just for 'the recipients had more money', which is pretty reliably true (the question is whether that's having any of the hoped for broader benefits) - the ECTC specifically worked well on a couple the metrics that I've been surprised to see not improve with cash. For example https://pmc.ncbi.nlm.nih.gov/articles/PMC12210083/ finds 'current on rent' and mental health effects - which is even more reason to be puzzled that we aren't seeing more results like that from transfers. A couple theories:
Some researchers raised with me the possibility that we should be shifting from monthly transfers to lump sums. One argument for this is the pattern in stress/wellbeing: often it initially gets better and then that benefit goes away (a hedonic treadmill effect, maybe?) and then towards the end of the study and after the end stress is actually higher (maybe related to the payment discontinuation). GiveDirectly said they mostly do lump sums in their non-US distributions now, and want to test more lump sums in the US. For stress in particular I'd hazard a guess (but note this is not that well supported by evidence yet) that a one time windfall is better than ongoing payments which will eventually terminate, since people habituate to them and then are extremely stressed when they end.
What can we conclude from the international evidence on the effects of child allowances on long-term outcomes (human capital development, criminality etc) for poor kids? Lots of countries have child allowances, after all, and surely many of them must have changed over time in ways that provide similar study fodder to the 2021-22 ECTC changes.
There are Canada and UK child allowance programs, but they're small enough I don't expect to see much effect and new enough I haven't seen any longitudinal research.
That said if I were to get way out ahead of the research and hazard a guess, I bet the connection between an allowance and reduced teen crime in particular is real. Teen crime seems disproportionately responsive to the labor market and honestly just disproportionately responsive to a lot of random things? It's more opportunistic and dumber and easier to deter.
There's an elegantly designed EITC study that makes use of the fact that a January baby gets bupkis for the prior year vs a December baby nets their family thousands of dollars. https://www.nber.org/papers/w30373
That study found detectable long term effects (including on the child's future earnings(?!)) which to be honest feels wildly implausible, but is a small point for "maybe these are compounding effects that turn up later" and/or "maybe you want to give money in a way that complements work vs replacing it."
It this one of those literal First World problems?
I don't mean it in a demeaning way in the sense that they are fake problems. On the contrary, poverty in the First World is the extremely hard part of the problem that is left to solve after you already fixed all the easy parts. As any engineer can tell you, the last 20% of the project takes 80% of the work.
More specifically, I think all the problems that poor people in the US have that can be solved with just a bit of cash have already been solved. Credit is cheap and easy to get. Hospitals have to treat you in an emergency even if you don't have money. Unlike in the Third World, a temporary lack of cash is pretty much never going to kill you.
The people who live in the most prosperous country in the world and nevertheless manage to remain persistently poor are those who have serious issues that probably require personalized attention. The ones who can be helped by cheap scalable interventions have already been helped and are not poor anymore.
I'm still curious as to whether these studies are really putting enough money down to say something meaningful enough. $1000/month is both "a lot" and "not really that much when you think about it".
The OpenResearch paper says that their "optimistic" estimate of the study's impact on people's net worth was $5000. That's not really a lot of growth over 3 years, right? What would happen if another (ludicrously expensive) study were done giving people $5000 a month? Or if a study were engineered to increase their net worth by 20k or 50k? Surely there has to be an amount at which giving money does indeed solve their problems.
Yeah I similarly believe that this is just not enough money to show the results researchers are looking for. I am pretty surprised by the Denver outcomes though.
But it's enough money that I expect to see a directional effect, even if a small one. $12k seems like it should be enough to forestall some downward spirals (e.g. car breaks, can't afford the fix, lose your job, etc)
I also wonder if the fact that the cash transfer was time limited affected the results. You might behave differently if the money was a government entitlement rather than a gift you’d receive for only a few years.
Especially with housing. A temporary grant isn’t very valuable as proof of income towards a lease or mortgage, whereas a government-backed UBI would be, as one example.
I also wonder how well landlords understood the program. If a tenant shows up and tries to use this program as proof of income, a lot of landlords are going to be skeptical or simply not understand what it is; they'd rather stick with something familiar.
One thing that surprises me is how cash transfers are so powerful in low-income economies yet don't do much in developed ones. I just wonder what the reason for this might be.
Keep in mind that the amount of money we're transferring as a share of income is generally much larger in developing countries. GiveDirectly's fantastic new study that found cash transfers cut infant mortality in half gave money in Kenya, where GiveDirectly's stats say more than a third of people live on less than $2.15 a day. I would guess this describes the typical recipient, since they target poor areas. They give them $1000, or the equivalent of more than a year's income. I'm not aware of any guaranteed income program in the US that transfers more than $30k per recipient, which is what it'd take to be close to equivalent.
We've also just already nabbed most of the low hanging fruit. The likely mechanism for that infant mortality cut in the GiveDirectly study is that women are much more likely to give birth in a hospital. Even the poorest women in America typically do give birth in a hospital. To a first approximation no one in America is starving to death. A lot of the most acute problems cash addresses, we've already solved - leaving things that are harder to fix.
My thesis is that a developed country has a multitude of options available to avoid poverty. The people who don't take advantage of those options likely have some underlying set of issues that are hard to change. But in a low-income country, there just aren't as many options to avoid poverty, even for those with talent and energy.
The poor in developed countries are still much better off than the poor in developing countries. So my guess is the things developed world poor need to meaningfully improve their lives are either much more expensive, or have to do with the conditions of their communities (public schools, healthcare availability, public safety, etc.). The latter can’t be solved by giving an individual more money, it would only be solved if the community as a whole got wealthier.
It’s discouraging to me how few people seem to be changing their mind as a result of this research. This seems like a pretty compelling case that cash transfers are a bad idea!
Fwiw there's some research into cash transfers that are more specific and contextual that I think is promising and that I certainly don't want to prematurely declare defeat on. Transfers to pregnant women seem to reduce the risk of serious complications (this saves money, because NICU stays are incredibly expensive); transfers to people just released from prison seems to reduce recidivism; transfers to domestic violence victims seem to make them somewhat likelier to safely leave. There's a ton of research underway and I wouldn't be surprised if we find some great use cases.
And then there's one reason that the world got excited about studying UBI - AI related mass unemployment - and I don't think these studies really tell us anything about how to handle that.
But I was indeed expecting stronger results for cash for homelessness and cash for chronic poverty, and this has changed my mind there.
If you want people to change their minds, you have to give them a better alternative. It can't just be "Well, cash benefits don't do anything, might as well give up"; it has to be "Cash benefits don't work, but here's a policy that will meaningfully improve outcomes for poor people". Otherwise, there's a strong suspicion that the research is just coming from the large community of people who are ideologically opposed to spending money on helping poor people. In a world where conservatives exist, liberals can't assume shared values and good faith.
My understanding is that both the test and control groups also received a fair amount of unplanned cash from the government during the pandemic relief, child tax credit, and other programs. I know that I'd totally ignore that if the results had come out the way I wanted them to (because, y'know, bias.) But I still think it's worth saying that the government actually did a fairly large amount of cash transfer in the middle of the test, which is ... hopefully a good thing?
How do you think about that? It probably makes for a robust comparison with the big lump sum group. But maybe less so with the other test groups? Or is that p-hacking myself to proving my priors?
I really appreciate this write up! I've been wrestling with these same findings in my last few posts, as I was also both surprised and disappointed by the results. But when I took a step back, I realized that these results jive really well with a finding from a big Megan Stevenson paper and Niskanen symposium from awhile ago. Essentially, what we fail to see in most social service interventions are "cascade effects" into other areas of people's lives, which is what we usually see when we do high quality studies like RCTs or try to replicate strong programs elsewhere. Giving people money is a good recipe for them to... have more money, which I agree is probably good! But the Stevenson article, which I wrote up a bit here, helped me understand why I should not have been as surprised by these findings as I was:
I really value this reflection on the limits of cash transfers. For me, it doesn’t mean unconditional basic income is a bad idea, but rather that we’re still asking money to do something it was never designed to do.
As Donella Meadows showed in her Leverage Points framework, the deepest way to change a system is by shifting the mental model behind it. If we keep defining value only as money, the outcomes will stay narrow. Life quality is much broader—safety, trust, understanding, awareness, relaxation, and tenderness. In our work at Me-We-World, we use this as a framework, but you can call them whatever resonates.
The bigger point is that we need a new paradigm of value—one linked to wellbeing, ecology, and community. Tools like digital identity, commons governance, and regenerative finance can help make these contributions visible and meaningful. That’s how economies can truly support flourishing.
In general I agree with this article, but I think it's tone is a little too pessimistic. Cash didn't have big, dramatic, positive results—but those types of results weren't the reason a lot of people were excited about giving people cash!
Mostly, people thought cash could (1) often help people more than in-kind aid, and (2) reduce poverty. The recent experiments didn't test (1). And they found that cash *did* work to produce (2).
Also, we still don't have data on the largest benefits predicted by the earlier research (long-term effects on cash recipients' kids).
This was really interesting, but it left me wondering about how external economic factors impact these kinds of results. In particular, if you conduct a trial like this in a tight labor market, could that meaningfully effect the outcome?
The scenario I'm imagining is this: in a tight labor market, I can choose to work a few less hours if I am given money, but that doesn't have a major impact on my well-being. If the labor market crashes, though, and I can't supplement my income as needed, does this external support play a much more important role of filling an otherwise-intractible gap?
I like this article a lot, thank you. The Denver results seem unsurprising to me. We already see poorer results than we would like giving people rent vouchers and services, and finding apartments for them.
I think the underlying theoretical mechanism here must be that people themselves are better positioned to know how to use a marginal dollar than social service agencies (especially when those agencies are dealing with many restrictions on the funding).
I bet that’s true in many cases, for example, the larger number of equally poor people who don’t become homeless seem to have this skill. Here though you have a group that’s already demonstrated they aren’t as good at this as others.
As media criticism a headline announcement that homeless people spend free money poorly seems like the least newsy story ever.
Like to go back to my journalism courses it has a bit of useful information not news quality to it and should lead activists to change tact not to get valuable space in American newspapers and websites.
Two things. First, how does this overview dovetail with, say, the Stevenson-Wolfers conclusion that increased income correlates to well-being? Is it that earning more money for labor performed should be distinguished from unconditional cash transfer? In that case, would the implication be that a job guarantee is better than a UBI, even if the guaranteed jobs amount to make work?
But the second part is that I'm not sure we should be evaluating the effectiveness of giving people money based on subjective factors. Like if the amount is used to pay off some debt, that may not show up in other metrics, but it's a good thing. Being in debt in my twenties wasn't a big stressor (and still isn't), but paying that down some did accelerate my ability to form a household and stuff like that. I'm not sure that would be measurable in any subjective survey (and I'm not suggesting that professional school graduates are a good candidate to receive these kinds of benefits), but I think you may be able to find some pretty good pro-social results on a longer time horizon, even if they don't show up in survey results in a two- or three-year window.
Regarding your first set of questions: I can definitely see a situation where *earning* more leads to a higher well-being. Likely the qualities that led you to earn more correlate to additional happiness and well-being metrics. But being *given* more does not say anything about the person's skills, work ethic or personal stability.
A lack of money is the definition of being poor, but it isn't the cause of being poor. The causes lie elsewhere -- skills, intelligence, impulsiveness, criminality, addiction.
Instead of giving out money, it appears the better path will prove to be (1) education to provide skills, (2) treatment to curb bad behavior and (3) more law enforcement when incentives don't work.
This seems like such an obvious point and yet one that is so commonly elided lest it come across as judgmental.
Of course, this can go too far; micromanaging people's grocery purchases with SNAP is almost certainly not worth the cost/effort involved.
Kelsey already pointed out that participants don't spend their extra money irresponsibly, so we should expect there to be a positive impact.
You can explain poverty however you like as a result of personal failing. But consider the marginal person - someone just unintelligent, ignorant, impulsive, lazy enough to be poor (in your view) but not so much of any of those things that they are destined to be poor in every circumstance. What do we expect there?
Why wouldn't money improve that person's life? And if it does not, why does your job training program have any better chance?
It's true that there are non-monetary things society can do to improve people's lives and certainly that people can do to improve their own lives. But nothing about that explains away these cash results.
TL;DR The best estimated study shows the control group increases income by 80% (!) The people in the study are already on the track which makes the cash less useful. This is why you wouldn't see them looking to upskill or change onto a new path with cash.
Let's focus on the best controlled study by Open Research: I read the same tables and see a story about upward mobility with and without cash.
1) Participants’ incomes rose a lot.
At baseline, individual income averaged about $21k and household income $30k. By year three, excluding the transfer itself, the control group reached $38.2k individual and $51.6k household—roughly +80% and +70% nominal gains in just three years. The treatment group landed at $36.3k and $45.7k without counting the transfer, and at $48.3k and $57.7k with it.
2) That counterfactual growth is a HIGH bar to beat.
When the “no-program” path delivers ~70–80% nominal income growth in three years, it’s unsurprising that short-run labor-market outcomes don’t show big incremental effects from a modest transfer. This is especially true given the period’s unusually hot job market.
3) The sample looks financially capable, not financially excluded.
Baseline total debt averages about $41k. Measured effects on credit categories are tiny (the only notable one is a modest uptick in auto loans). That’s consistent with borrowers already having access to standard credit markets—i.e., many investment/budget constraints were already relaxable via lenders. Cash is helpful, but it isn’t unlocking a closed door.
4) Education levels are relatively high; there’s no obvious “low-hanging fruit.”
Roughly 95% hold a HS diploma/GED or more. Combined with the income gains in the control group, this looks like a motivated, reasonably skilled cohort moving up the wage ladder as expected.
5) Put the magnitudes in global context.
The baseline individual incomes are already in the neighborhood of OECD GDP-per-capita levels like Portugal/Japan, and by year three the group sits near Germany/Canada/Australia ranges. Apples-to-oranges caveat noted—but it’s useful for calibrating expectations where these people are in a labor-leisure trade-off being richer than the average citizen of developed economies.
6) Poverty dynamics matter for interpretation.
Outside the study, U.S. poverty spells are often transitory: roughly half exit within a year, with exit probabilities falling the longer a spell lasts. That pattern fits what we see here: many people are on an upward trajectory already.
*In the spirit of the magazine I framed this as an argument
Those things, but also creating jobs that pay well that low-intelligence people can do.
I don't know how to do that. But I think it would help.
Kelsey, thanks for putting this article together! I’m curious if you reviewed or discussed with researchers the studies on the impact of the expanded/2021 Child Tax Credit? I gather that those are not RCTs, but is there anything to gain from the observational data on that large scale policy intervention (and withdrawal from it)?
Yeah, I was also curious about this. The CTC cut child poverty nearly in half in 2021, and it rebounded back up to prior levels in 2022 post-expiration — that seems like a pretty strong effect even if it’s not a RCT.
But there’s a difference between literally having more money (thus pushing you above the poverty line!) versus the other effects they’re evaluating — cognitive gains, maternal health, employment, stress, etc.
Yep! I looked at some of the work there and I agree with you both that the ECTC worked well. And not just for 'the recipients had more money', which is pretty reliably true (the question is whether that's having any of the hoped for broader benefits) - the ECTC specifically worked well on a couple the metrics that I've been surprised to see not improve with cash. For example https://pmc.ncbi.nlm.nih.gov/articles/PMC12210083/ finds 'current on rent' and mental health effects - which is even more reason to be puzzled that we aren't seeing more results like that from transfers. A couple theories:
Some researchers raised with me the possibility that we should be shifting from monthly transfers to lump sums. One argument for this is the pattern in stress/wellbeing: often it initially gets better and then that benefit goes away (a hedonic treadmill effect, maybe?) and then towards the end of the study and after the end stress is actually higher (maybe related to the payment discontinuation). GiveDirectly said they mostly do lump sums in their non-US distributions now, and want to test more lump sums in the US. For stress in particular I'd hazard a guess (but note this is not that well supported by evidence yet) that a one time windfall is better than ongoing payments which will eventually terminate, since people habituate to them and then are extremely stressed when they end.
What can we conclude from the international evidence on the effects of child allowances on long-term outcomes (human capital development, criminality etc) for poor kids? Lots of countries have child allowances, after all, and surely many of them must have changed over time in ways that provide similar study fodder to the 2021-22 ECTC changes.
It's possible I've missed some of the literature but the studies I'm familiar with were in Latin America (eg https://www.nber.org/papers/w29056 and (https://www.iza.org/publications/dp/6371/spillovers-from-conditional-cash-transfer-programs-bolsa-familia-and-crime-in-urban-brazil), so they run into the same dynamic discussed below in the comments - in countries that are poorer to start it's both easier to give a large % of household income and there's a lot of low hanging fruit in terms of things we know produce good life outcomes.
There are Canada and UK child allowance programs, but they're small enough I don't expect to see much effect and new enough I haven't seen any longitudinal research.
That said if I were to get way out ahead of the research and hazard a guess, I bet the connection between an allowance and reduced teen crime in particular is real. Teen crime seems disproportionately responsive to the labor market and honestly just disproportionately responsive to a lot of random things? It's more opportunistic and dumber and easier to deter.
Thanks. Is there not more research (maybe not in English) on the effects of changes to universal child benefits in continental Europe or Japan?
There's an elegantly designed EITC study that makes use of the fact that a January baby gets bupkis for the prior year vs a December baby nets their family thousands of dollars. https://www.nber.org/papers/w30373
That study found detectable long term effects (including on the child's future earnings(?!)) which to be honest feels wildly implausible, but is a small point for "maybe these are compounding effects that turn up later" and/or "maybe you want to give money in a way that complements work vs replacing it."
That study had a sample of around 300,000 December babies who got the money.
That's about 100x the sample size of a typical recent cash transfer study, which makes many more effects detectable.
It this one of those literal First World problems?
I don't mean it in a demeaning way in the sense that they are fake problems. On the contrary, poverty in the First World is the extremely hard part of the problem that is left to solve after you already fixed all the easy parts. As any engineer can tell you, the last 20% of the project takes 80% of the work.
More specifically, I think all the problems that poor people in the US have that can be solved with just a bit of cash have already been solved. Credit is cheap and easy to get. Hospitals have to treat you in an emergency even if you don't have money. Unlike in the Third World, a temporary lack of cash is pretty much never going to kill you.
The people who live in the most prosperous country in the world and nevertheless manage to remain persistently poor are those who have serious issues that probably require personalized attention. The ones who can be helped by cheap scalable interventions have already been helped and are not poor anymore.
This was well-written and thoughtful.
I'm still curious as to whether these studies are really putting enough money down to say something meaningful enough. $1000/month is both "a lot" and "not really that much when you think about it".
The OpenResearch paper says that their "optimistic" estimate of the study's impact on people's net worth was $5000. That's not really a lot of growth over 3 years, right? What would happen if another (ludicrously expensive) study were done giving people $5000 a month? Or if a study were engineered to increase their net worth by 20k or 50k? Surely there has to be an amount at which giving money does indeed solve their problems.
Yeah I similarly believe that this is just not enough money to show the results researchers are looking for. I am pretty surprised by the Denver outcomes though.
But it's enough money that I expect to see a directional effect, even if a small one. $12k seems like it should be enough to forestall some downward spirals (e.g. car breaks, can't afford the fix, lose your job, etc)
I also wonder if the fact that the cash transfer was time limited affected the results. You might behave differently if the money was a government entitlement rather than a gift you’d receive for only a few years.
Especially with housing. A temporary grant isn’t very valuable as proof of income towards a lease or mortgage, whereas a government-backed UBI would be, as one example.
I also wonder how well landlords understood the program. If a tenant shows up and tries to use this program as proof of income, a lot of landlords are going to be skeptical or simply not understand what it is; they'd rather stick with something familiar.
Yeah, one-time transfers vs a known 3-year thing vs permanent.
Still plenty of experimenting to be done IMO.
One thing that surprises me is how cash transfers are so powerful in low-income economies yet don't do much in developed ones. I just wonder what the reason for this might be.
Keep in mind that the amount of money we're transferring as a share of income is generally much larger in developing countries. GiveDirectly's fantastic new study that found cash transfers cut infant mortality in half gave money in Kenya, where GiveDirectly's stats say more than a third of people live on less than $2.15 a day. I would guess this describes the typical recipient, since they target poor areas. They give them $1000, or the equivalent of more than a year's income. I'm not aware of any guaranteed income program in the US that transfers more than $30k per recipient, which is what it'd take to be close to equivalent.
We've also just already nabbed most of the low hanging fruit. The likely mechanism for that infant mortality cut in the GiveDirectly study is that women are much more likely to give birth in a hospital. Even the poorest women in America typically do give birth in a hospital. To a first approximation no one in America is starving to death. A lot of the most acute problems cash addresses, we've already solved - leaving things that are harder to fix.
My thesis is that a developed country has a multitude of options available to avoid poverty. The people who don't take advantage of those options likely have some underlying set of issues that are hard to change. But in a low-income country, there just aren't as many options to avoid poverty, even for those with talent and energy.
The poor in developed countries are still much better off than the poor in developing countries. So my guess is the things developed world poor need to meaningfully improve their lives are either much more expensive, or have to do with the conditions of their communities (public schools, healthcare availability, public safety, etc.). The latter can’t be solved by giving an individual more money, it would only be solved if the community as a whole got wealthier.
Fantastic first article from Kelsey here. Really looking forward to more from the team!
It’s discouraging to me how few people seem to be changing their mind as a result of this research. This seems like a pretty compelling case that cash transfers are a bad idea!
Fwiw there's some research into cash transfers that are more specific and contextual that I think is promising and that I certainly don't want to prematurely declare defeat on. Transfers to pregnant women seem to reduce the risk of serious complications (this saves money, because NICU stays are incredibly expensive); transfers to people just released from prison seems to reduce recidivism; transfers to domestic violence victims seem to make them somewhat likelier to safely leave. There's a ton of research underway and I wouldn't be surprised if we find some great use cases.
And then there's one reason that the world got excited about studying UBI - AI related mass unemployment - and I don't think these studies really tell us anything about how to handle that.
But I was indeed expecting stronger results for cash for homelessness and cash for chronic poverty, and this has changed my mind there.
If you want people to change their minds, you have to give them a better alternative. It can't just be "Well, cash benefits don't do anything, might as well give up"; it has to be "Cash benefits don't work, but here's a policy that will meaningfully improve outcomes for poor people". Otherwise, there's a strong suspicion that the research is just coming from the large community of people who are ideologically opposed to spending money on helping poor people. In a world where conservatives exist, liberals can't assume shared values and good faith.
"We're not doomed. We just have a very long to-do list."
Love this! Thank you Kelsey!
My understanding is that both the test and control groups also received a fair amount of unplanned cash from the government during the pandemic relief, child tax credit, and other programs. I know that I'd totally ignore that if the results had come out the way I wanted them to (because, y'know, bias.) But I still think it's worth saying that the government actually did a fairly large amount of cash transfer in the middle of the test, which is ... hopefully a good thing?
How do you think about that? It probably makes for a robust comparison with the big lump sum group. But maybe less so with the other test groups? Or is that p-hacking myself to proving my priors?
I really appreciate this write up! I've been wrestling with these same findings in my last few posts, as I was also both surprised and disappointed by the results. But when I took a step back, I realized that these results jive really well with a finding from a big Megan Stevenson paper and Niskanen symposium from awhile ago. Essentially, what we fail to see in most social service interventions are "cascade effects" into other areas of people's lives, which is what we usually see when we do high quality studies like RCTs or try to replicate strong programs elsewhere. Giving people money is a good recipe for them to... have more money, which I agree is probably good! But the Stevenson article, which I wrote up a bit here, helped me understand why I should not have been as surprised by these findings as I was:
https://flyovertakes.substack.com/p/what-do-we-do-if-nothing-works
I really value this reflection on the limits of cash transfers. For me, it doesn’t mean unconditional basic income is a bad idea, but rather that we’re still asking money to do something it was never designed to do.
As Donella Meadows showed in her Leverage Points framework, the deepest way to change a system is by shifting the mental model behind it. If we keep defining value only as money, the outcomes will stay narrow. Life quality is much broader—safety, trust, understanding, awareness, relaxation, and tenderness. In our work at Me-We-World, we use this as a framework, but you can call them whatever resonates.
The bigger point is that we need a new paradigm of value—one linked to wellbeing, ecology, and community. Tools like digital identity, commons governance, and regenerative finance can help make these contributions visible and meaningful. That’s how economies can truly support flourishing.
In general I agree with this article, but I think it's tone is a little too pessimistic. Cash didn't have big, dramatic, positive results—but those types of results weren't the reason a lot of people were excited about giving people cash!
Mostly, people thought cash could (1) often help people more than in-kind aid, and (2) reduce poverty. The recent experiments didn't test (1). And they found that cash *did* work to produce (2).
Also, we still don't have data on the largest benefits predicted by the earlier research (long-term effects on cash recipients' kids).
I've got a paper coming out on this later this fall (https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5397609) and this article inspired me to do a write-up on it on substack (https://substack.com/home/post/p-171407588).
This was really interesting, but it left me wondering about how external economic factors impact these kinds of results. In particular, if you conduct a trial like this in a tight labor market, could that meaningfully effect the outcome?
The scenario I'm imagining is this: in a tight labor market, I can choose to work a few less hours if I am given money, but that doesn't have a major impact on my well-being. If the labor market crashes, though, and I can't supplement my income as needed, does this external support play a much more important role of filling an otherwise-intractible gap?
I like this article a lot, thank you. The Denver results seem unsurprising to me. We already see poorer results than we would like giving people rent vouchers and services, and finding apartments for them.
I think the underlying theoretical mechanism here must be that people themselves are better positioned to know how to use a marginal dollar than social service agencies (especially when those agencies are dealing with many restrictions on the funding).
I bet that’s true in many cases, for example, the larger number of equally poor people who don’t become homeless seem to have this skill. Here though you have a group that’s already demonstrated they aren’t as good at this as others.
As media criticism a headline announcement that homeless people spend free money poorly seems like the least newsy story ever.
Like to go back to my journalism courses it has a bit of useful information not news quality to it and should lead activists to change tact not to get valuable space in American newspapers and websites.
Two things. First, how does this overview dovetail with, say, the Stevenson-Wolfers conclusion that increased income correlates to well-being? Is it that earning more money for labor performed should be distinguished from unconditional cash transfer? In that case, would the implication be that a job guarantee is better than a UBI, even if the guaranteed jobs amount to make work?
But the second part is that I'm not sure we should be evaluating the effectiveness of giving people money based on subjective factors. Like if the amount is used to pay off some debt, that may not show up in other metrics, but it's a good thing. Being in debt in my twenties wasn't a big stressor (and still isn't), but paying that down some did accelerate my ability to form a household and stuff like that. I'm not sure that would be measurable in any subjective survey (and I'm not suggesting that professional school graduates are a good candidate to receive these kinds of benefits), but I think you may be able to find some pretty good pro-social results on a longer time horizon, even if they don't show up in survey results in a two- or three-year window.
Regarding your first set of questions: I can definitely see a situation where *earning* more leads to a higher well-being. Likely the qualities that led you to earn more correlate to additional happiness and well-being metrics. But being *given* more does not say anything about the person's skills, work ethic or personal stability.