It’s hard not to marvel at Silicon Valley’s technological innovations. But I’m dumbfounded by the social science experiment they’ve been working on.

Last week, the National Bureau of Economic Research released two working papers, 248 pages all-up, with the first results from an experiment testing the effects of a Universal Basic Income – a UBI. 

OpenResearch, a spinout from Y Combinator, the Silicon Valley accelerator where Airbnb, Dropbox, Stripe and more got their start, coordinated the work. So it did not lack potential resources.

Resources matter if you are serious about testing a UBI in a rich country. If you don’t pay enough, it’s pointless. If you don’t have enough subjects, your tests will be statistically underpowered. You could confidently say that results could range from moderately positive to moderately negative, and that’s about it. But paying a lot of people a lot of money is expensive.

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.

Before we get to the results, perhaps a few basics.

Transfer programmes bring terrible trade-offs. New Zealand’s system targets assistance to those in most need and to older people regardless of need. Targeting means more assistance can be provided to those in most need, but also brings demeaning interactions with the administrative state. Targeting assistance requires checking that only those eligible are given the most help.  

A UBI is meant to avoid those kinds of problems. If everyone gets a basic income, nobody has to justify their existence to WINZ. Unfortunately, any UBI faces other hard trade-offs. Over a decade ago, the Treasury ran the numbers and found that even a flat income tax rate of over 50% would not be enough to cover a UBI large enough to replace most, but not all, other programs.

But perhaps it would be worth it if there were substantial benefits. Proponents argued that a UBI would help people find better jobs because they would not need to jump at the first available job. A UBI might also improve health outcomes either directly through access to services or indirectly through reduced stress. Meanwhile, critics argued that a UBI might discourage people from being engaged in work at all.

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 $5,000 less per year than the control group, not counting the transfer, or about $6,000 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.  

The work is remarkable. Serious Silicon Valley social science, partnered with academia.

Perhaps New Zealand’s targeted social assistance has had things right all along.

To read the full article on The Post website, click here.

The NBER working papers are available at https://www.openresearchlab.org/studies/unconditional-cash-study/documentation/category/category-b