Supporters maintain that a UBI would not only provide income support to people in need, but would also increase work incentives. That is because, unlike the current welfare system, it would not claw back 50, 70, or even 100 percent of the earnings of low-income workers who make the effort to get a job. Opponents are more skeptical. They fear that if everyone were given a basic cash income with no requirement to work, people would quit their jobs in droves and we would end up with a nation of layabouts.
What the critics say
The fairest way to start is to let UBI critics speak for themselves. Consider, for example, the views of Jim Mazi, who contributed recently to a recent discussion of basic income proposals on Cato Unbound. Manzi notes, correctly, that “a major driver of welfare reform over decades has been political resistance to work disincentives.” Taking an astonished tone, he continues by saying, “It is fairly extraordinary to claim that the government could guarantee every adult in America an income even if they did zero work of any kind, and that somehow this would not reduce work effort.” He goes on to cite the results of a series of income maintenance experiments carried out between 1968 and 1980 in several US states. Those experiments, Manzi contends, “consistently found that the tested programs reduce the number of hours worked versus the existing welfare system.”
Pascal-Emmanuel Gobry is another UBI critic. Writing recently for The Week, he notes that although a UBI has become fashionable among progressives, it actually has a long history of support from conservatives. “Indeed,” he says, “Right-winger that I am, I was for a very long time a strong proponent of a UBI. But now I oppose it. . . What happened? I looked at the best science and changed my mind.”
What kind of science? “Science, properly understood, is the testing of hypotheses through rigorous experimentation,” says Gobry. He points to the same randomized field trials of alternative welfare systems as Manzi does. He shares Manzi’s conclusion that under a UBI, “Millions of people who could work won't, just listing away in socially destructive idleness (with the consequences of this lost productivity reverberating throughout the society in lower growth and, probably, lower employment, in a UBI-enabled vicious cycle).”
Conservative critics are not the only ones who see work disincentives as a built-in feature of a UBI. Some progressive supporters of the concept agree, with the twist that they, unlike the conservatives, see a reduction in work as a virtue. For example, writing on the website Liveable 4 All, C.A. L'Hirondelle and J.S. Larochelle give ten arguments in favor of their version of a UBI, which they call a Guaranteed Livable Income (GLI). Several of the reasons anticipate reduced work effort:
A guaranteed livable income would prevent the destruction of the environment due to the attempt to grow the world's economy to create full employment. The loss of clean air, clean water, forests, arable soil, all forms of "wild" life cannot be measured. . . GLI is the quickest way to say "no" to environmentally harmful and wasteful practices and to put an end to over-consumption.
Millions of jobs depend on war. . . Historically, when people lose jobs and income and they demand their governments do anything, including build their military, to give them jobs. Only with a guaranteed income could we afford to end the war industries.
With a guaranteed livable income, people will no longer be forced to 'make a living' regardless of the harm to themselves, others, or the environment. It also creates a way to for people to do beneficial (and essential) activities that are currently financially penalized for being 'unproductive.' For example: work done by volunteers and unpaid caregivers: people who take care of other people; who take care of their own health (for some that is a full time task); or who take care of their neighborhood, community or environment.
Based on these writers and on dozens of comments from both conservative and progressive readers of my own posts, I would go so far as to say that the default assumption among people who first hear of the idea is that a UBI would severely reduce work effort. In this series, I will take on the burden of proof in arguing that, to the contrary, replacing our current welfare system with a UBI would substantially increase incentives to work, especially among the low-income households that are the greatest cause for concern. Here goes—theory first, evidence next time.
The effects of means-tested income support
Let’s begin by looking at the work incentives of introducing a generic means-tested income support (MTIS) policy into a labor market that previously had no income support programs at all. Figure 1 illustrates that case. (This figure and those that follow take an approach used by Robert Moffitt, an economist cited favorably by UBI critic Manzi. I have modified the geometry of Moffit’s diagrams in a way that, I hope, makes them more intuitive. I should also mention that although Moffitt considers several varieties of income support policy, he does not explicitly examine a UBI.)
The horizontal axis in Figure 1 represents earned income while the vertical axis shows disposable income, that is, earned income plus benefits. To keep things simple, we will assume no income or payroll taxes on earned income—an assumption that I will briefly return to near the end of the post. The dashed 45o line shows that earned and disposable income are the same when there are no taxes or income support. The solid red line shows the relationship between disposable and earned income with the MTIS policy.
This generic MTIS policy has three features:
- A minimum guaranteed income, G, that households receive if they have no earned income at all.
- A benefit reduction rate (or effective marginal tax rate), t, indicted by the angle between the 45o line and the red MTIS schedule. The fact that t is greater than zero is what we mean when we say that the program is means tested. As the figure is drawn, t = .75, that is, benefits are reduced by 75 cents for each dollar earned.
- A break-even income level, beyond which benefits stop. Past that point, earned income equals disposable income.
The figure is very general. It could represent a conventional cash welfare program like Temporary Assistance to Needy Families (TANF), an in-kind program like SNAP (food stamps), or a negative income tax, which would have cash benefits in the form of refundable tax credits administered through the Internal Revenue Service. The existing welfare system of the United States is, of course, more complicated. It is a combination of many state and federal programs, some with in-kind benefits and some with cash benefits; some with narrower eligibility requirements than others; some with benefit reduction rates of 100 percent or more and one, the earned income tax credit, with negative benefit reduction rates for some households; and some with income limits where benefits cut off abruptly, causing “notches” in the disposable income schedule. (For details, see this earlier post.)
Despite its simplicity, the generic program illustrated in Figure 1 is sufficient to support our first important proposition:
Proposition 1: Any means-tested income support program will unambiguously reduce average work effort.
The numbered arrows in Figure 1 show why. Begin with Arrow 1, which shows how the program affects your work incentives if your earned income is initially below the break-even level.
First, you can see that regardless of how much you work, you will receive some benefit from the government, so your disposable income will rise. Other things being equal, people tend to “spend” part of their increased income on increased leisure, in the form of shorter hours, longer vacations, a longer time in school, earlier retirement, or longer breaks between jobs. Economists call that the income effect of a higher income.
Second, the income support program changes the tradeoff you face between additional work and additional earned income. Suppose your best available job pays $10 per hour. Without the MTIS, one more hour of work brings you $10 of added disposable income. With the MTIS, one more hour of work raises your disposable income by only $2.50, because the $10 you earn is offset by a loss of $7.50 in benefits. In effect, the program reduces the “price” (more properly, the opportunity cost) of leisure, so you are more likely to substitute leisure for work and the disposable income it brings in. Economists call that the substitution effect of the program.
Below the break-even level of income, the income and substitution effects work in the same direction, so there is an unambiguous incentive to work less. Some people might stop working altogether. Some might reduce their hours, take longer vacations, or spend more time between jobs. Whatever the specifics, the average response will be a reduction in work hours. Even if some people have employers who allow no flexibility in hours worked, the average response for the whole population to whom the policy applies will be zero.
Next, suppose your initial earned income is higher than the break-even level. If it is far above B, the program will probably not affect your work effort at all. Suppose, though, that your initial income is only a little above the break-even level. In that case, you may be tempted to cut back your hours of work by enough to qualify MTIS benefits. Arrow 2 shows that possibility. If you follow the arrow, you give up a little bit of disposable income, but you gain quite a bit more leisure—a tradeoff between work and leisure that is better than you faced before the MTIS program came into force. Not everyone will take advantage of this opportunity, but if at least some do, while others do not change their work habits, then the average work effort of people whose initial incomes were higher than B will decrease.
In short, introducing a MTIS into an economy where previously there was no income support will tend to reduce the work effort of most people whose incomes were initially below the break-even point and of some with incomes just above break-even. Average work effort for the population as a whole will decrease.
Sweetening the program
Next, consider the effects of sweetening the MTIS by increasing the minimum guaranteed income or reducing the benefit reduction rate. Figure 2 shows the effects of both changes together. The old program is the same as shown in Figure 1. The new program has a higher minimum guarantee G2 and a lower benefit reduction rate of 0.5, shown by t2.
The arrows show three possible cases.
First, suppose your initial earned income is below B1, the break-even level for the old program. You now have an increased incentive to work because you get to keep 50 cents of each dollar you earn, rather than 25 cents as before. At the same time, however, your income will be higher, which by itself, will give you an incentive to work less. Since the two effects work in opposite directions, whether you work more or less will depend on whether the income effect or the substitution effect is stronger. As we will see in the next post, evidence from the experiments of the 1970s and 1980s suggests that for most people in this category, the income effect will prevail, so that average work effort will increase, as shown by Arrow 1. However, it is likely that some individuals will work less, and it is theoretically possible that average work effort for people in this group would be less.
Second, suppose that your initial earned income is slightly higher than the new break-even level B2. In that case, as in our first example, you might decide to cut back your work hours by enough to qualify for benefits, as shown by Arrow 2. Not everyone will do so, especially if their incomes are far higher than the break-even level. Still, the average effect for this group will be at least a slight reduction in work effort.
Third, suppose your initial earned income is between B1 and B2—too high to qualify for the old program but low enough to qualify for the new one. In that case, the new, more generous, MTIS will increase your disposable income, which, by itself, would cause your work effort to decrease. At the same time, you now face the 50 percent benefit reduction rate of the new program, whereas before you faced no benefit reductions at all. The substitution effect of the benefit reductions provides second disincentive to work. With both effects working in the same direction, you have a double reason to reduce your desired level of work, as shown by Arrow 3.
Taking the three cases together, we can see that formally speaking, the effects of sweetening the MTIS are ambiguous. Some people (Arrow 1) will increase their work effort, while some (Arrows 2 and 3) will reduce theirs. It would not be at all surprising if the negative effects on work to outweighed the positive effects, but we can’t be sure on theoretical grounds alone. Ultimately, the average response for the whole population will depend on the number of people in each group and the relative strength of the income and substitution effects. We can formalize this result as follows:
Proposition 2: The effects on work effort of “sweetening” a means-tested income support program by raising the minimum income guarantee, lowering the benefit reduction rate, or both, are ambiguous.
Replacing a MTIS with a UBI
Now we come to the heart of the issue we set out to discuss. What will be the effect on work incentives if we replace a pre-existing means-tested income support program with a universal basic income that is not means tested at all? Figure 3 illustrates this possibility. The MTIS is the version from Figure 1, before sweetening. The UBI grant, Gu, is set at a level slightly below the minimum guaranteed income under the original MTIS, Gm. (We will consider the effects of a change in the relative levels of Gu and Gm shortly.) The benefit reduction rate for the UBI is zero, so the UBI schedule is parallel to the original earned income line and lies above it by the amount of the UBI grant, Gu.
The arrows show four possible responses, according to the original level of income.
First, suppose that your original earned income is below the crossover point of the UBI and the MTIS schedules. Now you have two incentives to work more: you can now keep more of each added dollar you earn, and at the same time, unless you increase your work effort, you face a decrease in your disposable income. For this group, the substitution and income effects work together to encourage work effort, as shown by Arrow 1.
Second, suppose your original earned income is a little above the crossover point of the MTIS and the UBI schedules. As in the first case, the substitution effect is strong because you are no longer subject to the 75 percent benefit reduction rate of the MTIS. At the same time, you face a negative work incentive from the income effect, since your income will rise a little even if you don’t work more. In practice, though, the income effect is likely to be weak, because it applies only to the difference between the UBI and the MTIS benefits, not to the full value of the UBI. On balance, you will probably want to increase your work effort, as shown by Arrow 2.
Third, suppose you are one of those people who would have earned an income above the break-even level B if there had been no MTIS, but who cut back your work effort to qualify for the MTIS--the case shown by Arrow 2 in Figure 1. Now that the old program is gone, you will increase your hours of work again. This variant is shown by Arrow 3 in Figure 3.
Fourth, suppose your earned income is far above the break-even level for the MTIS. In that case, the UBI has no substitution effect at all, so you have no incentive from that source to work more. Since your income will rise by the amount of the UBI, you have a small incentive from the income effect to work less, as shown by Arrow 4. Again, though, the effect on your work effort is likely to be slight, because the higher your original income, the smaller the percentage increase from the UBI. If you have a middle-class income to begin with, a few thousand dollars extra from the UBI is unlikely to have a big effect on your work habits.
Keep in mind, though, that the effects of replacing a MTIS with a UBI depend on the relationship of the UBI grant Gu to the MTIS minimum Gm, and also on the benefit reduction rate of the MTIS. The more generous the minimum guarantee of the MTIS relative to the amount of the UBI and the higher its benefit reduction rate, the more likely it is that replacing an MTIS with a UBI will increase work effort.
We can summarize these effects as follows:
Proposition 3: Replacing a means-tested income support program with a UBI would substantially increase work incentives for low-income households while having small disincentive effects, if any, for middle- and upper-income households.
The most important implication of the preceding analysis is that we can intelligently discuss the incentive effects of a UBI only if we specify what we are comparing it to.
If we were to introduce a UBI where, previously, there was no income support program at all, the UBI would presumably decrease average work effort. In that case, UBI critic Manzi would be right to think it “extraordinary to claim that the government could guarantee every adult in America an income even if they did zero work of any kind, and that somehow this would not reduce work effort.”
Similarly, we could expect a negative effect on work effort if we were to “sweeten” the existing welfare system by adding a UBI on top of all the programs we now have. The work disincentives of the current high benefit reduction rates would remain in force. Adding the UBI would have only an income effect, which, by itself, would tend to reduce work effort.
However, the policy change proposed by most UBI advocates—especially those who approach the idea from a conservative or libertarian perspective—is to replace the means-tested welfare system we now have with a UBI. In that case, the favorable substitution effect of a UBI would be strong and the unfavorable income effects would be weak. In that situation, it would “extraordinary” if the UBI did not increase work effort.
The preceding discussion comes with several caveats.
First, theory alone is not conclusive. We also need to look at the empirical evidence provided by the income maintenance experiments of the 1970s and 1980s and by other sources.
Second, the work incentive effects of a UBI would depend, in part, on how it was financed. In my view, we should avoid financing a UBI by raising marginal tax rates on middle- and upper-income households, since doing so would produce negative work incentives for those groups that would, at least in part, offset the positive incentive effects for the poor. Also, to minimize negative work incentives, a UBI should replace not just existing antipoverty programs, but also “middle class welfare” in the form of tax loopholes, and it should be properly integrated with income support programs aimed at the general population, such as Social Security retirement and disability benefits . This previous post outlines an approach to financing a UBI that would incorporate these principles.
Third, as noted above, the effects of a UBI on work incentives depend, in part, on how generous it would be. My own preference would be for a modest UBI that falls well short of offering a life of middle-class ease and comfort. Instead, I see it as providing a minimal but secure platform on which lower income households could build a better life for themselves through work.
Finally, I would like to emphasize that I argue only that replacing the current welfare system with a UBI would increase work efforts on average. There is no guarantee that every individual would react by working more. Many critics are troubled that a UBI does nothing to exclude people who are able to work but who choose not to. That concern raises the important issue of the inevitable tradeoffs in the design of any income support program between its effectiveness in aiding those who are seen as “deserving” and its effectiveness in barring aid to the “undeserving.” A thorough exploration of those tradeoffs will have to wait for a future post.
Continue to UBI and Work Incentives. Part 2 Evidence