This week's column for the Stuff papers covered the excellent new US work testing the effects of a UBI.
From November 2020, 3000 low-income people were randomly assigned into two groups for three years. One thousand people each received $1000 per month in unconditional funds for three years. Two thousand people each received $50 per month.
Both groups filled in detailed surveys on how they spent their time, on their purchases, their health experience and more. Participants had blood tests to check health outcomes. Government administrative records were combined with the survey data to provide more detail.
The UBI amounted to about a 40% increase in recipients’ income – large enough to matter.
The researchers pre-registered their study design to guard against, well, fiddling. If you have dozens of potential outcomes that might be affected by the cash transfers, there are many ways for studies to accidentally find effects that are not there, and even more ways to put a thumb on the scales. Trial pre-registration says in advance how statistical testing will be run.
...
What did the experiment find?
Households receiving the UBI were able to enjoy more leisure than other households. Leisure is good and should not be underrated. But that was the largest actual effect. Their overall earnings, including the transfer, were higher – but labour force participation and hours worked fell. Households receiving the UBI earned about $5000 less per year than the control group, not counting the transfer, or about $6000 more than the control group when counting the transfer.
People receiving the transfer spent more time unemployed if they became unemployed, but more time for job search did not help. There was no effect on job quality – and a good study design meant they could rule out even small effects.
Remember that the funding for the study came from donors. Any real-world UBI would need to be funded by taxation that would have its own pernicious effects on work incentives. And a permanent UBI would have larger effects on work choices than a three-year programme.
What about health? In the short term, people receiving the transfer enjoyed reduced stress and greater food security. But those effects quickly faded. There was increased uptake of health services and some healthy behaviours. We can be more confident that a UBI benefit would not be blown on drugs and alcohol. But the study found not even minimal effects on physical health. And initial improvements in mental health disappeared after the first year.
While poverty is certainly associated with worse health outcomes, very large and sustained cash transfers did not improve health. Worth remembering when reading the next public health study asserting that more income redistribution would improve health outcomes.
They put a heck of a lot of work into this trial and its evaluation. To the extent that they got a law change so that the payments wouldn't be considered taxable or affect eligibility for other benefits - so they knew it was the same $1000/month increase for everybody.
The disemployment effects were a bit larger than I'd expected, so I've updated my expectations on that. My priors on health effects were reinforced; in richer countries, the income-health gradient is going to be an artefact of an underlying correlate of both.
I was a bit more surprised by zero effects on job quality. I'd put some weight on that less time-pressure to match with a new job when unemployed could yield better job matches. But the 1.1 extra months' duration of unemployment spells relative to the control group cashed out into precisely estimated nil effects on a whole big range of job quality items.
You can catch a twitter thread by Eva Vivalt, of the researchers on it, here.
But UBI-stans seems to be some of the stranger beasts out there. I mean, look at this.
Eric Crampton must have been reading another study from the report he describes in his hostile critique of basic income (‘Putting a UBI to the test’, July 29). The US cash transfer project he eulogised was not a test of basic income.
There have been over 100 experiments which show positive results. The one he cites is not one of them. By definition, a basic income is a modest amount paid regularly to all usual residents, individually, without means-tests or behavioural conditions, regardless of income, gender, marital status or work status.
The study Crampton cites does not pretend to respect that definition. Several limitations make the results irrelevant for assessing basic income.
It was a means-tested benefit paid to 1000 individuals spread across 19 counties in Texas and Illinois, about 50 per county. That is hardly universal. The individuals self-selected. They only received the cash if they could prove they were poor. It was only paid to individuals aged 20-40. It was only paid to one individual per household, if nobody in the household was receiving disability benefits and they were not in publicly-subsidised housing. It deliberately over-sampled those from minority groups. Those features invalidate any claim to randomness, unconditionality or universality.
Ok. Where to start. If you want to run a UBI trial, you can try saturation where you enroll an entire town. In that case you get the second-round effects from the spending of UBI income and the like and interactions among people who all get the UBI, but you really need to set the thing so it's funded by the community receiving the benefits so you get the effects of imposing the taxes necessary to pay for the thing. That would get you more of a total effect.
Or, you can do what this group did and test just the effects of the transfer on outcomes for those receiving it, leaving out both potential community-wide benefits from everyone getting it (whether social stuff or spending effects) but also the incentive effects of the taxes necessary to pay for it.
They targeted the group whose responses are of most interest for this policy agenda: lower-income people of prime working age. Randomisation was achieved by randomly putting people from that group either into the treatment or control groups. Payment when assigned to the treatment group was unconditional. You didn't have to do anything to get the money. If you filled in the time use surveys you got extra payments for that - as did people in the control group. And payment was universal within the treatment group.
I suppose you could take the position that the only real trial of a UBI is to actually implement it across an entire community, but we quickly get into No True Scotsman issues. Did the trial encourage inward or outbound migration? Well, gonna have to apply it to the whole country aren't we for it to be a real trial.
I think we can pretty confidently say that a UBI providing an after-tax transfer on the order of 40% of non-UBI income for low income people will have effects like the ones found in this experiment, but very likely with larger labour market effects both because of the permanent income hypothesis (a permanent transfer will have larger effects than a known-to-be-time-limited trial), and because of the effects of tax rates required to fund the transfer. And that there won't be the hoped-for improvements in health or job quality. But that there won't be increases in problem drinking because of it, and that people choosing more leisure probably consider themselves better off. Anyway, here's one of the authors on this:
Leisure is a normal good, people.Since this study captures the pure income effect of giving cash, without any of the substitution effects that might occur with a means-tested program, it tells us how much people prefer more leisure on the margin, even without disincentives to work. 37/
— Eva Vivalt (@evavivalt) July 22, 2024
The experiment began at the height of Covid, so it is surely a marvel that a cash transfer only resulted in a drop of just over an hour a week in paid labour.
Covid affected both treatment and control groups. The control group wasn't in some no-Covid place. Maybe the argument is that the transfers gave people the ability to better hide from Covid if they wanted? Remember too though that there were all kinds of work-from-home and other changes that will have hit both groups, as well as Covid payments. And that the trial ran for three years, finishing in the second half of 2023.
Moving towards a basic income for every resident citizen is a matter of common justice, freedom and basic security. Economists have shown it is affordable without raising income taxes.
Guy Standing, Co-president, Basic Income Earth Network; Professorial Research Fellow, SOAS University of London
The first sentence here is a values assertion.
The second one, well, Treasury in NZ showed that we'd need a flat income tax of more than 50% to fund it and it still wouldn't be enough to replace all other benefits. A small one could be done without raising taxes - say if you abolished NZ super and split that money equally regardless of age. The transfer would be pretty small though. Or maybe he's expecting it to be funded out of some other tax he's not mentioning.
The basic tradeoffs in a UBI are the same as I'd pointed out in the Spinoff ages back. We now have some better numbers on the likely effects.
Read the NBER papers for yourself though. They're here.
Also fun to compare the study's results, with what people had hoped would be the results. I guess those whose expectations were most dashed and who were most committed to specific beliefs about outcomes are most upset about it?
But you can also check Table 16 of the paper comparing expert predictions of what the effects would be and the eventual effects - which is just so neat as method.I asked GPT4-o to complete the UBI experiment abstract based on the first part: pic.twitter.com/shLDQDYMeG
— John Horton (@johnjhorton) July 22, 2024
No comments:
Post a Comment