Keywords

This chapter addresses claims that can be tested, but shows that they can only be tested partially, indirectly, or inconclusively. No claims from the list in Chap. 13 can be tested fully, directly, and reasonably conclusively in a small-scale experimental setting the way medicines can sometimes be tested. The central question is how to deal with the indirect and partial nature of the findings.

Experiments have some ability to examine the following claims:

  • The welfare claim

  • The economic-equality claim

  • The reduced -social-costs claim

  • The labor-effort claim

  • The affordability claim

  • The poverty-trap claim

  • The harm-to-workers claim

  • The benefit-to-workers claim

  • The widespread -benefit claim

  • The cost-effectiveness claim

1 The Welfare Claim

The welfare claim is probably the most important empirical claim in the UBI debate. The central reason to support a transfer payment is to make people better off. Although some past studies have underplayed the welfare question in favor of more easily measurable variables, the ability of UBI to achieve that goal is far more important than its potential side effects.

Welfare—an abstract concept about people’s inner state—is not directly observable. The best existing methods for determining welfare are self-assessments and observations of quality-of-life indicators. Welfare is at least partly subjective, and some quality-of-indicators can be morally loaded. Alcohol is clearly unhealthy and has many potentially damaging side effects, but if it has no ability to increase welfare, 70% of Americans don’t know what’s good for them.

Fortunately, many quality-of-life indicators are not as tricky. If you have more secure access to an adequate diet, more secure housing, fewer feelings of social isolation, and healthier, longer-lived children than otherwise, you are almost undoubtedly happier. People who are happier with an inadequate diet, ill-health, shorter-lived children, and so on probably suffer from a diagnosable mental disorder. And so, we can safely use many quality-of-life indicators.

Social scientists have developed reasonable welfare indexes based on well-researched indicators.Footnote 1 Researchers conducting UBI experiments can report on quality-of-life indicators in a nonjudgmental way and employ respected indexes to provide an overall measure of welfare. They can also conduct a survey asking people in the control and experimental groups about their well-being and about factors likely to affect it.

One important aspect of welfare that could be particularly important to UBI experiments is time use. UBI has the potential to free up people’s time. If so, will people spend more time in education, childcare, volunteering, positive social relationships, or various behaviors that might be labeled as “lazy” or “self-destructive?”

The need for welfare indicators means that the welfare claim is a host of claims and subclaims. I haven’t attempted to list each claim separately because there are too many of them, including effects on physical and mental health, homelessness, housing quality, infant mortality, education, food security and adequacy, nutrition, problems associated with the “ghettoization” of poverty, and many more.

Researchers could straightforwardly employ standard quality-of-life indicators and welfare indexes, but they might also consider addressing welfare issues that have particular importance to the UBI debate, such as those related to the freedom claim, the flexible-lifestyle claim, the consumerism claim, and the self-destruction claim. I’ve discussed the difficulty of dealing with these claims, but they do affect welfare and have particular importance to the UBI discussion in many countries.

The sheer volume of welfare indicators that one can put into an index distracts attention from how important each of them is. I’m guilty of that, leaving most of them out of the list of named claims. But UBI experiments must emphasize all quality-of-life indicators they can measure and explain the relationship between them and the ones they can’t.

The difficulty of observing, measuring, quantifying, and combining quality-of-life indicators into a good understanding of welfare discourages work on it. But it has to be the central focus of any attempt to find out whether UBI succeeds in achieving its central goal. By contrast, the labor-time comparison between the control and experimental groups, though far less important, attracts attention because it is a nice, neat, apparently-easy-to-understand number.

Community and long-term effects on welfare are likely to be substantial because there are so many channels by which UBI is likely to affect welfare: direct distribution, market effects on income and working conditions, reduced inequality, reduced ghettoization of poverty, improved education, and so on. Researchers will have to do a great deal of extrapolation to relate study findings to reasonably accurate predictions for a national program. Individual-level RCTs will underestimate the impact of UBI on quality-of-life indicators—both positive and negative. Saturation studies will do only slightly better. Most welfare effects are likely to accumulate slowly over the long term, to be larger for a policy expected to be permanent, and to involve national-level community effects.

One advantage of saturation studies is that some welfare-related community effects are local. A 5–10-year saturation study in an impoverished town—if feasible—could produce a great deal of information about the effects of ghettoized poverty, not just about UBI’s role in alleviating it.

The trial will give some indication about the direction of UBI’s impact on various welfare indicators, but researchers will have to extrapolate using other evidence to estimate the welfare impact of a national UBI, including the feedback effects from employers and the community over the long run. Those predictions will be based largely on that other evidence, but experiments can provide useful information about the direction of change.

2 The Economic-Equality Claim

The economic-equality claim , as stated, needs no test because UBI necessarily reduces inequality through direct redistribution as long as it is set at a sustainable level. But the important issue is not whether but how much UBI reduces inequality. This question is partially testable because it depends on many market factors, some of which are observable. But experiments will only reveal the first step in a long chain of reactions that will determine UBI’s effect on economic equality. Experiments can compare the incomes of people in the control and experimental groups, but they will need to combine that with evidence from other sources for UBI’s likely effects on taxes paid by higher-income people and on employers’ wage response. Some kind of simulation will be necessary, and this estimate will be only the short-term effect of a temporary policy.

To get some idea of longer-term effects, researchers can observe the initial effects of UBI on education, health, safety, food security, and other factors that are correlated with economic mobility, but they cannot actually observe whether those factors do lead to greater economic equality for experimental participants. Researchers can use other evidence about how these variables are correlated to economic mobility to estimate their effect on economic equality, but experimental findings will make only a small contribution to that estimate and the effort becomes somewhat speculative.

3 The Poverty-Trap Claim

The poverty-trap claim implies that UBI will lead to greater labor effort for people eligible for full-time benefits under a conditional system. This can happen because many conditional programs (such as disability, public housing, unemployment insurance, and in the United States, free or subsidized medical care) require people to sacrifice all or most of their benefits if they accept employment or have private income above a certain level. This rule gives recipients a financial incentive to choose benefits over low-paid labor, discouraging them from taking steps toward economic mobility—hence the “trap.” Some conditional programs have effective marginal tax rates in excess of 100%, so that recipients are financially better off remaining on benefits than they would be taking a low-wage job.

UBI eliminates the poverty trap because people receive the grant regardless of income. Virtually all UBI proposals are structured so that people are always financially better off earning more than earning less, removing the trap.

A UBI experiment can test reasonably well whether people—in the short term—respond to the removal of the poverty trap at a given wage. But the long-run impact of permanently freeing people from the poverty trap is likely to be much larger. Experiments cannot determine whether improvements in health, education, housing, food security, market conditions, and similar variables increase people’s ability to get out of poverty in the long run. Additional theory and evidence will have to be combined with experimental findings to produce an estimate.

For this issue, it is extremely important to separate the effects of the size of transfer from the effects of the type of transfer . If a large UBI is tested against a small conditional program, some or all of the work-stimulating impact of removing the poverty trap will be counteracted by the creation of a more generous alternative to work.

4 The Reduced-Social-Costs Claim

Experiments can address the reduced -social-costs claim by examining the demand for social services among experimental subjects. Examples include UBI’s potential to alleviate the poverty trap or to improve health and reduce the demand for healthcare. Not all social costs are easily observable, and so the results will be only partial. Experiments cannot reveal the full impact of UBI on the demand for social services because that demand greatly depends on community and long-run effects. Researchers will have to rely on a large amount of nonexperimental evidence to estimate the effect of UBI on social costs.

This issue has been underemphasized in some past experiments because of its difficulty, but it is so important that it must not be ignored. For example, Michael McLaughlin and Mark Rank estimate that the annual cost of US child poverty alone is $1.0298 trillion or 5.4% of GDP,Footnote 2 not including the costs of adult poverty.

5 The Labor-Effort Claim

Experiments can provide some direct evidence about the labor-effort claim, but that evidence can be deceptive. Experiments will observe the difference between the average number of hours worked by the control and experimental groups, and that comparison is likely to attract a lot of attention not only because of the political importance of the labor-effort effect, but also simply because it is easily quantified. “What is the labor-effort response in the experiment?” “It is X%.” A simple number that took years of research to produce can be very satisfying, especially to an audience that doesn’t understand how far removed the raw comparison of control and experimental groups is from a prediction of the national labor-effort response to a fully implemented UBI system.

Even as a measure of the initial response of workers, this comparison is likely to overstate the effect of a national UBI because, as earlier chapters explained, the sample will probably be drawn from a small segment of the income distribution, including people who are more likely to reduce their labor hours in response to UBI than other segments. Experiments drawing samples in this way will have to bring in nonexperimental evidence to connect their findings to the effect of a national UBI.

It is not certain that UBI experiments will find a correlation between UBI and decreased labor effort. As mentioned above, in less wealthy nations, UBI has been associated with an increase in labor hours, and it might be associated with an increase in labor effort if the sample focuses on people caught in a poverty trap. However, unless a nation has a very large number of people caught in a poverty trap or in extreme poverty, such as that experienced in poorer nations, a slight decline in labor effort is probable and its importance should not be overblown.

The observable reaction of laborers is not the full effect on labor effort even in the short run. As earlier chapters explained, supply and demand theory predicts that the market will react to a decline in labor hours by increasing wages and/or improving working conditions in the relevant sectors, and that each of these effects will cause labor hours to rebound, partially counteracting the initial decline. RCTs cannot directly observe the labor-demand response at all, although they can use a microsimulation model to estimate it. As always, that means that the experimental findings play a lesser role in determining the final estimate—much of which will come from the assumptions going into the model. Saturation studies can capture some demand response, but only at the local level, which is likely to be much smaller than the national demand response.

Even these simulations will produce incomplete results because the input data involves only the short-term response of workers to a temporary program. The long-term response of workers and employers cannot as easily be estimated with simulation techniques because it depends on unpredictable cumulative changes in variables, such as improved health, education, housing, cultural norms, bargaining power, food security, and so on.

Yet the simulations need to be run, and any possible unmeasurable long-term effects explained and perhaps predicted on an ad hoc basis, because of the central role labor effort has for many critics of UBI and because of its vulnerability to spin and misunderstanding. Recall from Chap. 6 that the labor-effort effect dominated the public discussion of the NIT experiments of the 1970s. The raw comparison of the control and experimental groups was discussed in the popular press as if it were a straightforward representation of the national response, when in fact the national response was estimated to be two-thirds smaller. This issue dominated the discussion and distracted attention from more important issues.Footnote 3 Anyone reporting or writing about future experiments should try to preempt a repeat of this misuse of experimental findings.

Writers can help by pointing out that the labor-effort claim is not merely the claim that UBI reduces labor hours; it is the claim that the fall in labor effort is “unacceptably high.” The definition of unacceptable is subjective and morally loaded. UBI supporters are likely to define “acceptability” synonymously with sustainability, connecting it with affordability (see discussion below). At least some opponents are likely to define it so strictly that they can present any decline in labor effort as unacceptable. In the absence of a shared understanding of the controversy over the acceptability criteria, many writers during the 1970s discussion tacitly assumed that any decline in labor hours was unacceptable —regardless of how large or small that decline was and seemingly all other factors.Footnote 4

UBI experiments in wealthy nations will probably find a decline that is “acceptable” by the sustainability standard and “unacceptable” by the no-decline-is-acceptable standard, giving each side the opportunity to spin the results their way. Researchers can help head off this kind of spin by recognizing that the controversy over acceptability criteria exists and by addressing it directly. They can discuss the relevance of the experimental results to people with each of these points of view and look for other standards that might be of interest to people with more moderate views.

Alternative standards of acceptability might involve other questions, such as: how much of the decline was composed by workers reducing their hours, by unemployed workers increasing their search time, or by people leaving the labor force? How do they spend their increased non labor time, as full-time caregivers, as students, as entrepreneurs, and so on? What costs and benefits are associated with this decline in average labor effort? Is the decline in labor effort something that can be counteracted by other factors, such as an increase in the number of available jobs that offer high wages and good working conditions?

6 The (Un)Affordability Claim and the Cost Issue in General

Experimental evidence can play a small but worthwhile role in addressing the (un)affordability claim and other issues relating to cost. For any given UBI scheme, cost can be assessed in terms of taxes and in terms of efficiency loss. Cost can be viewed in terms of taxes or in terms of efficiency, which is discussed above.

The direct tax cost of UBI can be calculated fairly well with income statistics. That is, determine how much UBI costs assuming no one changes their behavior in response to it or to the tax increases that will accompany it. I’ve elsewhere estimated that a UBI of $12,000 has a net cost less than 3% of GDP, and a UBI of $20,000 has a net cost less than 10% of GDP.Footnote 5 The role of experiments is to help determine how changes in behavior affect that cost. A negative labor-effort effect increases cost. An increase in wages or a decline in the need for other social services (via desirable effects on health, education, crime, etc.) will decrease costs. The effects of social costs are too large to ignore, no matter how difficult they are to estimate. I’ve quoted figures showing that the annual cost of US child poverty is 5.4% of GDP.Footnote 6 That savings alone would more than pay for the $12,000 UBI and would relieve more than half of the cost of the $20,000 UBI.

Experiments estimate only the first step in the chains of reactions that lead to these results. Simulation models can help estimate some of the further steps.

The contribution of experimental data to the cost issue is so small that one could imagine using nonexperimental data to estimate labor-market responses in a microsimulation involving no experimental data at all, but microsimulations are also a highly imperfect method. Experimental findings need to be understood as an effort to improve estimates of some of the parameters that go into the model necessary to estimate cost.

The indirect effects on the cost of UBI through its effects on crime, education, health, nutrition, housing, and similar variables are so hard to estimate accurately that the best theoretical models will invariably leave some out and apply speculative estimates of others. But yet they’re extremely important. They are likely to have a major impact on the cost of a national UBI. These effects can’t be left out of the discussion without badly misinforming nonspecialists, most of whom will not grasp their importance without help.

The question “is UBI affordable?” is too vague to be meaningful. It requires two moral judgments to become meaningful. First, it requires an affordability criterion: how much is too much? Unfortunately, the affordability criterion is subjective and partly morally loaded. UBI supporters (and perhaps others who are positively inclined toward UBI) are likely to define the affordability synonymously with sustainability. That is, a program is unaffordable only if costs associated with it are so large that they collapse the program itself. Opponents (and others negatively inclined toward it) are likely to define the affordability criterion in such a way that any added cost is “unaffordable.” Many other criteria are possible, and many open-minded people might not have settled on an affordability criterion.

Second, the question is not simply whether UBI is affordable; it is whether the desired level of UBI is affordable. Some low level of UBI is clearly affordable (e.g. $1 per year), and some high level is clearly unaffordable (e.g. anything exceeding per capita income). We need to answer the question: how much is enough? Virtually all UBI supporters prefer a UBI high enough to live on—at least to live free from homelessness and economic destitution. That level is very likely to be sustainable in the context of universal education, healthcare, and other government services provided free-at-the-point-of-delivery as well as policies to ensure that affordable food, housing, and other basics are available in the market.

Eliminating destitution would be an important achievement, but it is not necessarily enough for all or most UBI supporters, most of whom want a UBI that frees everyone from the threat of poverty, ensures everyone a life in dignity, and protects them from significant social exclusion by lack of economic means. Whether that level of UBI is affordable depends both on the affordability criterion and on how generously these conditions are defined.

Researchers conducting experiments cannot hope to resolve these disputes, and they probably should not impose their own criteria on top of the controversy. But they can examine questions that are relevant to the different ways that people who are interested in the UBI discussion view cost and affordability. These might include: how much does a UBI at the official poverty level cost? Is it sustainable or affordable? How much does a significantly higher UBI cost? Is it sustainable or affordable? What is the highest sustainable UBI level? How much will UBI’s labor-market and welfare effects increase or decrease its overall cost? What is the efficiency cost of UBI? How do the tax and efficiency costs of UBI compare to the cost of other programs capable of achieving similar goals? What affordability criteria are relevant in the local discussion of UBI? What levels of UBI are part of the local discussion of UBI? How much do they cost and are they sustainable?

Existing evidence overwhelmingly indicates that a UBI high enough to eliminate absolute poverty is sustainable in high-income countries. It won’t hurt to double-check the sustainability, but the sustainability of absolute-poverty-level UBI is not a pressing source of serious disagreement in the debate. A sensational media headline saying “Study finds poverty elimination possible with UBI” would be true, but it would not report a groundbreaking finding. Such a headline would spin the discussion of research findings to the pro-UBI side. Yet, leaving UBI’s ability to eliminate poverty out of the discussion of the findings spins the issue to the anti-UBI side.

The poverty claim is useful in framing research questions around the cost-effectiveness claim. The question “what is the cost of eliminating poverty with a UBI” is fairly neutral. But a noncomparative focus on cost creates a spin opportunity for the anti-UBI side.

However good the numbers might be, they are not likely to resolve the controversy because are likely to fall into a range where supporters (using a sustainability criterion) can declare UBI “affordable” and opponents (using a criterion putting UBI last on the list of priorities) can declare it “unaffordable.” Researchers and anyone else writing about the experiments can help head off spin by recognizing the controversy over the affordability criteria. For example, they can report that the cost of this UBI scheme is affordable by these controversial criteria and unaffordable by these other equally controversial criteria. They can also consider how UBI compares in affordability to other programs of similar size and/or effectiveness—that is, by connecting the affordability question to the cost-effectiveness question.

7 The Widespread-Benefit Claim

The widespread-benefit claim, as I use it, is distinct from the harm-to-workers and benefit-to-workers claims (discussed next). It is not simply the claim that UBI’s direct and indirect benefits are shared by many people (whether workers or not) at any given time, but also that a significantly greater portion of people will benefit from UBI at some time in their lives.

The spread of UBI’s direct financial benefits at any one time is determined largely by its structure. UBI proposals with feasible costs can be structured so that 40–60% of the population receive direct financial benefits.Footnote 7 This much is sufficient to say that a large portion of the population benefits at any one time. There are at least three ways in which UBI’s benefits might be spread more widely.

First, because of economic mobility, many more people can expect to benefit financially from UBI at some time in their lives than at any one time—that is, many more people’s incomes will go below the break-even point at some point in time. Simply counting contributors and beneficiaries can give the impression that these categories are fixed. Presumably the UBI system is a net benefit to people at the times when they need it most—that is, when they have the least. The question of how many people can expect to benefit at some time during the course of their lives is clearly as important as the question of how many people benefit at any given time.

Second, UBI might create more favorable market or social conditions that directly benefit net financial contributors. (See the benefit-to-works claim below.) For example, the psychological impact of permanently removing the fear of poverty and destitution could benefit everyone.Footnote 8

Third, positive community effects of UBI might benefit net contributors enough to counteract the loss of the taxes they pay. Although it’s overly ambitious to hope everyone will benefit all-things-considered, there is evidence that more equal societies are in many ways better for everyone. Lower crime, more stable communities, less group antagonism, healthier environments, and so on can lead to better outcomes for people across the income spectrum.Footnote 9

Unfortunately, RCTs are unable to provide any direct evidence about the community or psychological impact on net (financial) contributors. A saturation study will do only slightly better. Direct observation of the widespread -benefit claim would require an extremely long-term study involving subjects at all levels of income. Researchers can use historical evidence about economic mobility to estimate how many people will fall into the net recipient range at some point in their lives. Experiments can make two small contributions toward understanding this claim by observing the labor-effort effect and UBI’s impact on welfare factors likely to improve economic mobility, safety, health, education, and so on. Of course, these are only the first steps in a chain that might benefit net contributors over time.

Again, UBI experiments can only contribute a small piece of evidence to the effort to make these estimates, but a focus on how people benefit throughout their lives is essential to a good public understanding of UBI’s likely effects.

8 The Harm-to-Workers Claim and the Benefit-to-Workers Claim

The harm-to-workers and benefit -to-workers claims—as stated—are oversimplified. Any UBI system financially benefits some workers and harm others. The relevant questions seem to be: which workers benefit and how much? Which workers are harmed and how much? Is there evidence that a group of people will abandon all “work” (however defined); if so, how many will, and how will this group affect workers? Of course, not everyone agrees that the existence of such a group is ethically problematic, and research should avoid giving off the impression that it necessarily is.

These claims also present at least two difficult subjective definitional issues. First, what do we mean by harm and benefit? Financial harm and benefit are easier to observe and quantify than overall benefit, but they aren’t as important. And so, it is best to consider both.

Second, what is a “worker?” Is a full-time parent or caregiver a worker? Are other unpaid workers “workers?” Is a person living off financial investments a worker? How many hours per week does a part-time laborer have to be employed to count as a worker? How many weeks can someone be unable to find a job and still count as a worker? Is a person who uses UBI for a 1-year sabbatical from a 40-year working life a worker? Do children, the retired, and the disabled count as “workers?” And so on. If we define any of these groups as workers, the number of workers UBI benefits will be much higher than if we don’t. And even if we don’t, we might judge the financial harm these groups create for workers differently than the harm other nonworkers create for workers. This ambiguity is why most of this book avoids the term “worker” altogether in favor of the clearer term “laborer” (meaning a person working for pay). But this section uses “worker” because the ambiguous idea is what matters for the discussion of these claims.

Experiments can say something about these claims, but researchers need to approach them cautiously because what they can say is very limited, easily misinterpreted, and connected to contentious ethical disagreements, such as the exploitation debate. Researchers can’t ignore them because experimental findings might be misunderstood or spun as showing much more about these claims than they actually do.

As with the affordability claim, experimental evidence plays only a small role in calculating the harm and benefit to workers. Most of the financial harm and benefit of a UBI system is determined by its structure and does not need a test. If UBI is largely income tax financed, anyone making less than the break-even point financially benefits and anyone making more is financially harmed. Other ways of financing UBI make the break-even point more difficult to calculate, but all financing methods create winners and losers.

The last section mentioned that a UBI system can be structured to directly benefit 40–60% of the population (including a lot of workers) at any given time. The direct financial harm to workers in the low end of the net contributory range will be small and might be overridden by positive community effects. Many workers will be in the net beneficiary range at some point in their lives. Also, not all net contributors will be workers. Some will be people living off investment income.

Researchers can help avoid misunderstanding by presenting findings for various demographic groups and various definitions of workers. What percentage of workers are financially harmed? What percentage are financially helped? What is the average net benefit to the average net beneficiary worker? What is the average net harm to the average net contributory worker? What are the average before-and-after -tax-and-transfer incomes to the average net beneficiary worker and the average net contributory worker? What percentage of UBI net benefits go to people in other demographic categories of interest to the discussion, who might not be expected to be laborers? These might include children, caregivers, retirees, students, and so on.

Researchers will understandably reject making the controversial judgment of identifying a group of people as those who could work, should work, and don’t work under UBI. But they can better help improve the public understanding by trying to find some nonjudgmental way to report numbers that usefully inform people who have different ethical positions on these issues. One way might be to report the percentage of the cost caused by the benefits to people in the various demographic categories relevant to the national discussion.

Most of the experimental contribution to the understanding of financial harm to net contributory workers is determined by its contribution to our understanding of the total cost of UBI. Policymakers can choose to spread that burden in many different ways, some of which would put most of the burden on rent-paying assets rather than on labor income. This difference will have different implications for people with different moral positions.

Workers working less is the first step both in the story ending in worker harm and in the story ending in worker benefit. The ability to work fewer hours or take more time to search for the right job if one happens to become unemployed is a direct benefit to workers, but this also increases the tax cost of UBI, some of which might be borne by workers. Theory predicts that employers respond to initial reductions in labor effort by improving pay and working conditions, possible even for net contributory workers. Even if increased wages only go to net recipient workers, it (and any positive response in labor time) will mitigate some of the tax cost of the initial decline in labor time. Estimating the extent to which these factors are both benefits and costs to workers can help avoid misunderstanding.

Trials will contribute to the understanding of the costs and benefits to workers through possible reductions in social costs and through possibly improved worker productivity (see above).

If labor-market response of workers is small, the financial harm and benefit to workers will be pretty much dictated by the structure of the program. If not, other evidence will be required to estimate whether those changes increase or decrease the benefit to net contributory workers. Researchers would need to run a simulation model using nonexperimental estimates of the elasticity of supply and demand in various labor markets. And of course, the outcome of any such model will be somewhat speculative, driven largely by the assumptions of the model. But experimental data is still useful, potentially indicating which segments of the labor market (in terms of occupation, income level, etc.) will be most affected.

9 The Cost-Effectiveness Claim

Although the cost-effectiveness claim is the bottom line, it requires little additional discussion because it is examined by putting together the evidence discussed above. Each variable discussed above can be looked at individually in cost-effectiveness terms, and all the variables of interest can be indexed into one overall cost-effectiveness estimate. Combining experimental, historical, and theoretical information to address the cost-effectiveness question makes the results one step less direct and conclusive, but it is more important to report less conclusive answers to meaningful questions than more conclusive answers to less meaningful or misleading questions.