The Evidence for Guaranteed Minimum Income

In the last few blog entries, I’ve been talking about Guaranteed Minimum Income. A reader asked the very reasonable question of what evidence was there that the GMI would actually “work”. So far, within North America, there have been several experiments to determine the impact of a GMI

These studies generally focused on similar dimensions, trying to understand both the positive and negative impacts of the GMI. In particular, they seem to focus on the labor market, health outcomes, and social outcomes. In this blog post, I’ll talk about the results.

Made in America

In the late 1960s and early 1970s, the US government carried out several experiments with a guaranteed minimum income. Oddly enough, this research was carried out under the direction of Donald Rumsfeld and his assistant Dick Cheney (providing more evidence that, regardless of what you think of Rumsfeld’s politics, he’s probably an interesting fellow to sit beside at a dinner party.)

That they were involved is not as mad as it seems. Milton Friedman, the patron saint of free markets and modern Republicans, suggested the GMI. Though Republicans today would be outraged at the idea of GMI, that outrage would typically stem more from ideology, polarized politics, and fervor incited by the mass media than from actual evidence or reasoning.

These experiments were carried out in New Jersey, Pennsylvania, North Carolina, Iowa, Seattle, and Denver. In all cases, they use a reasonable experimental framework, comparing a randomly selected experimental population to a similar control group.

The evidence indicated that the GMI resulted in reduction in work effort of 13%, the majority of which was secondary and tertiary earners. What’s more, part of the loss seems to have been attributed to adolescent males entering the workforce later, possibly spending more time at school. This hypothesis was supported by higher test scores in children in the experimental group, reduced dropout rates, and increases in adult continuing education. Little health data was analysed, though one study did show improved birth-weights in the most at-risk individuals.

Other than the small work effort reduction, the main negative was an increase in divorce rates of experimental families in one of the studies, though a subsequent analysis concluded that this discovery was a statistical error. To me, this result, even if it were true, isn’t actually all that negative. Is it better for society for a couple to be forced to stay together for economic reasons, or for an unhappy couple to be able to break up because they can afford to be apart? I tend to think the latter is preferable. Individual happiness seems more important than any moral imperative to maintain unhappy marriages.

In the cold, snowy north

A similar experiment was conducted in Manitoba, Canada, both in an urban Winnipeg location and a rural Dauphin location. The Winnipeg location had experimental and control groups, while the Dauphin location offered the program to everyone. The latter has the obvious disadvantage of no control group, while gaining the advantage of showing what would happen with a universal program, rather than a small sub-segment of a bigger community.

In Winnipeg, they only analysed the impact on the labor market. The results there were similar to the US studies–a relatively minor impact on the labor market, clustered in secondary and tertiary earners. This is an interesting result, but I think it’s frustratingly limited, since it can only be used to refute an argument for not providing a GMI, rather than actually supporting a GMI.

Dauphin, on the other hand, had broader data allowing a more detail analysis of other dimensions. Like the American study, the GMI led to fewer teenagers dropping out of school in Dauphin. Hospital use and mental health hospital use of Dauphin residents relative to non-Dauphin residents declined. There were no statistical differences between the incidents of low birth weights, the newborn death rates, or marriage dissolutions.

One other interesting hypothesis proposed by Evelyn Forget in her Dauphin analysis was that GMI may have a social multiplier. If GMI results in one individual deciding to stay in school longer, their friends may be more likely to stay in school as well. I imagine this effect could be both positive and negative–if someone decides to quit their job and just live off the minimum income, their friends might decide to do the same thing as well.

What I make of all this

This social multiplier hypothesis shows one of the problems with these sorts of studies. If this multiplier really exists, it could be highly problematic. It might not be visible in the standard limited-population studies where there is both an experimental group and a control group, but only in saturated studies, studies where every individual participates.

Of course, the saturated studies are closer to a real-world implementation. But in saturation studies, it’s much more difficult to form solid conclusions because it’s much harder to find a control group.

Nevertheless, these studies seem pretty useful to me, because I believe that we largely know already from other, non-GMI studies that higher incomes increase happiness and improve outcomes. And we know that we get the most leverage by helping the poorest. If you give a millionaire $10,000, they probably don’t care too much, but if you give it someone living at the poverty line, it matters a lot.

So to me, the primary argument against GMI is based on labor productivity, and I think the results here show that there is a relatively minor impact on the primary earners. If a bunch of parents stay home to take care of their kids and a bunch of teenagers complete school instead of working, that seems like it could be a good thing.

The bottom line

All that said, I still don’t think there’s great evidence either way, and I also think it’s basically impossible to get great evidence. But I believe the mediocre evidence largely supports the idea of a Guaranteed Minimum Income. At this point, we don’t have enough evidence to do it for an entire country. But perhaps we know enough to try it at a state or a province.