Universal Basic Income vs. Unemployment Insurance: Which is the Better Safety Net?

A universal basic income (UBI) and unemployment insurance (UI) are two possible forms of social insurance for an economy in which job loss is a significant risk. Which works better? How generous should either program be? Would a combination of the two be best of all? These are the questions that Alice Fabre, Stéphane Pallage, and Christian Zimmermann (FPZ) address in a recent working paper from the Research Division of the St. Louis Fed.

The answers that the authors give to these questions will disappoint UBI supporters:

When compared head-to-head, UI is a better social safety net than a UBI.
In an economy with no unemployment insurance, a UBI would be better than nothing, but the optimal level would be quite low, about $2,000 per person per year for the United States.
No combination of UI and a UBI is superior to UI alone.

Skeptics are likely to seize on these findings, but in my view, they do not support a blanket rejection of a UBI. Instead, as I will explain, they highlight how important it is for UBI proponents to pay attention to details of financing and program design.

The FPZ model

Fabre, Pallage, and Zimmerman recognize that, a priori, UI and a UBI each has its advantages and disadvantages. The advantage of UI is that it is better targeted at people who need income support. Its disadvantages are substantial costs of monitoring and administration, despite which some people will manage to game the system in order to collect benefits to which they are not entitled. In contrast low administrative costs are a major advantage of a UBI, offset by the fact that a far higher share of benefits go to households that are not in need.

To determine the best balance of advantages and disadvantages, FPZ construct a mathematical model that allows them to turn a UBI and UI on and off, singly or in combination, and to adjust levels of benefits. The model allows for incentive effects of changes in taxes and personal income, macroeconomic effects, administrative costs, and the possibility that some people may collect benefits improperly. Parameters of the model are set to reflect recent conditions in the US economy. Outcomes are scored using a social welfare function that gives a positive weight to real GDP, but allows a decrease in GDP to be offset, at least up to a point, by the risk-reducing effects of social insurance.

Skipping over many technical details, I would like to comment on three features of the model that have particularly strong impacts on the authors’ findings:

First, the base conditions of the FPZ model, before introduction of a UBI or UI, include no other social safety net programs—no cash welfare benefits, no food stamps, no housing vouchers, no childcare studies, nothing. Sources from which households meet their living expenses are limited to some combination of wages, personal savings, unemployment benefits and UBI benefits.
Second, the government finances all expenditures, including UI and UBI benefits, with a proportional income tax. As the number of beneficiaries and benefit levels change, the tax rate goes up or down as needed to keep the budget in balance.
Third, the UI and UBI programs are additive. When both are in force, everyone gets the basic income and those who are unemployed get the unemployment benefit, too.

Let’s examine the importance of each of these features in turn.

UBI vs. UI or UBI vs. welfare?

Within the FPZ model, both a UBI and UI create work disincentives, but of different kinds.

The work disincentive of unemployment insurance comes from moral hazard, a term economists use to describe a situation in which people who are insured against the consequences of a risk make less effort to avoid it. In this case, since UI offers at least partial protection from the consequences of unemployment, it reduces the pressure for unemployed people to find new jobs. In some cases, that may be beneficial, as when longer search allows people to find jobs that better match their skills. However, other people may opportunistically go through the motions of looking for a job with no intention of actually taking one until their benefits run out. UI policies include measures intended to reduce moral hazard, but in practice, monitoring is expensive and safeguards are imperfect.

The work disincentive of a UBI, on the other hand, arises from the income effect—the economist’s term for the tendency of people to “spend” a part of any increase in their income on additional leisure, though shorter working hours, longer vacations, or earlier retirement. As I discussed at length in an earlier post, the income effect is not very strong. Standard estimates suggest that most people would be expected to reduce hours worked by about 1 percent or less in response to a 10 percent increase in income, other things being equal. Still, because a UBI would increase everyone’s income, the aggregate effect could be significant, especially if the UBI grant is large relative to the median income. Within the FPZ model, the relative merits of a UBI and UI are determined, in large part, by the balance between these two kinds of work disincentive.

Unfortunately, the model completely neglects one whole source of work disincentives that play a major role in the literature supportive of a UBI. Those are the disincentives that are built into conventional means-tested welfare policies, such as TANF, food stamps, housing vouchers, and child care subsidies. Some programs reduce benefits by a certain percentage for each dollar earned. Others have income cutoffs above which benefits cease altogether. As I have explained elsewhere, the benefit reduction rates of various programs are additive. For example, if a family loses 20 cents in food stamp benefits for each added dollar earned and 15 cents in housing subsidies, then the total benefit reduction rate is 35 percent. If we sum all of the benefit reduction rates and add payroll taxes, , we get what economists call the effective marginal tax rate (EMTR) facing poor families.

According to a 2012 study published in the National Tax Journal, effective marginal tax rates for poor families vary greatly from state to state and by level of income. In many states, EMTRs are low for the poorest families, largely because the earned income tax credit provides a positive wage subsidy that offsets the benefit reductions of other programs. For families from half the poverty level to twice the poverty level, however, EMTRs can be quite high, often in the range of 50 to 70 percent. For some individuals, especially second earners in households at or just above the poverty line, they can be more than 100 percent.

These work disincentives become important when we examine a UBI not as an alternative to unemployment insurance, as in the FPZ study, but as an alternative to means-tested welfare. It then becomes plausible to argue that replacing conventional welfare with a UBI would provide a net increase in work incentives for the population as a whole. The moderate disincentive for upper income households arising from the income effect would be more than outweighed by the strong positive work incentives for the poor that would come from lower benefit reduction rates.

How to finance a UBI?

A second key feature of the FPZ model is its assumption that the government uses a proportional income tax to finance all expenditures, so that each time the benefits of a UBI or UI are increased, tax rates must rise to keep the budget in balance. Higher tax rates create a further work disincentive over and above the moral hazard and income effects already discussed. The feedback from social benefits to taxes holds down the estimated optimal levels of UBI and UI benefits.

I agree with the FPZ team on this point. UBI supporters should be wary of financing a basic income by raising marginal tax rates, whether uniformly for everyone, as in the FPZ model, or selectively on high earners, as in the versions of a UBI favored by some progressives. Instead, I have argued that any UBI could and should find sources of financing that do not raise anyone’s marginal tax rates.

In yet another earlier post, I explained how this would be possible. First, replacing existing federal welfare programs would allow a shift of about $500 billion annually to a UBI. Second, UBI benefits could also replace “middle-class welfare” in such forms as tax preferences for home ownership, retirement, and charitable contributions, freeing up another $600 billion or so for the UBI. Middle-class families would be free to allocate their UBI benefits to those purposes if they chose, or they could spend them on something else. Third, Social Security retirement and disability beneficiaries could be given the option of choosing between the benefits they now get or the new UBI benefits, taking whichever were greater, but not both. Such a prohibition on “double dipping” would raise a few additional billions for the UBI. More importantly, it would reduce the number of UBI claimants by more than 50 million, permitting higher benefits for the remainder of the population.

Elimination of double dipping by the unemployed

A third feature of the FPZ model that deserves comment is the way it treats double dipping by the unemployed. Equation (2) of the model allows unemployed persons to draw both unemployment benefits and UBI benefits when both programs are in force. In my view, that would be bad policy, for three reasons.

First, prohibition of double dipping would make it possible to finance a more generous UBI without raising marginal tax rates. The logic for UI is similar to that for Social Security retirement and disability, although the case is made more complicated by the fact that currently, federal taxes are not the primary source of finance for UI.

Second, it seems to me that double dipping is inconsistent with the whole concept of social insurance. Under our current system, for example, we offer support for people who are disabled and for those who are unemployed, but we don’t allow anyone to collect both forms of benefits. When we consider adding UBI to the mix, we should think of it as a form of broad, multi-peril social insurance that fills gaps in the present system and mitigates some of its perverse incentives rather than as a program that doubles up on the protections that are already on offer.

Third, a prohibition on double dipping would significantly mitigate the moral hazard of unemployment insurance. Consider a simple numerical example: Imagine Jane, a hotel worker, who has lost a $400-a-week job and is drawing unemployment benefits of $160 per week, consistent with the US average replacement rate of about 40 percent. Assume that a UBI, if there is one, caries a benefit of $100 per week, about the level I have suggested in earlier writings.

Suppose now that Jane has uncovered a job that pays $400 per week, but that she thinks that because of imperfect monitoring she could get away with continuing her UI benefits even if she passes the job up. How strong is her incentive to take the job?

Under the current system (UI but no UBI), she will gain $240 per week by taking the job, which may or may not be enough to compensate her for the loss of leisure time and any work-related expenses.
If we add a UBI and allow double dipping (as in the FPZ model), she will get an additional $100 per week whether she takes the job or not. Her marginal gain from taking the job is still $240 per week, less loss of leisure and work expenses. However, she will be slightly less inclined than before to sacrifice the leisure time because of the income effect of the UBI.
If we add a UBI but prohibit double dipping (my suggestion), she will have $500 a week if she takes the job but just $160 if she does not, giving her a $340 gain from taking the job. By any reasonable estimate, the increased marginal gain from taking the job will more than offset the deterrent to work arising from the income effect of the UBI. (For a detailed explanation of why that is likely to be the case, see the discussion of income and substitution elasticities in this earlier post.)

In short, if we allow double dipping, as in the FPZ model, introducing a UBI exacerbates the problem of moral hazard that is inherent in an imperfectly monitored system of unemployment insurance. In contrast, if we disallow double dipping, adding a UBI mitigates the moral hazard problem.

The bottom line

FPZ recognize that the economy they study is highly simplified, but they do not think that changing details of the model would significantly affect their findings:

The model economy we consider here is obviously a very crude approximation and could beneﬁt from some bells and whistles. We want to argue that such additions are not necessary, as they would only reinforce our results that UI generally dominates UBI.

I disagree. As I have tried to explain, their pessimistic findings regarding a UBI arise in substantial part from three specific simplifying assumptions:

Using a baseline for comparison that includes no social programs except for the UBI and UI magnifies the income effect of a UBI. The situation changes if, instead, we introduce the UBI is as a replacement for our existing means-tested welfare system, as I and many other UBI proponents recommend. In that case, the positive effect on work incentives from the reduction of the high effective marginal tax rates built into the current welfare system would at least partly offset, and very likely more than offset, the negative income effect of a UBI.
If we were to finance a UBI with a proportional income tax, as in the FPZ model, then every increase in UBI benefits would add to work disincentives through increased marginal tax rates on everyone with a job. If, instead, the UBI were financed by shifting funds not used for means-tested welfare for the poor and “middle-class welfare” in the form of tax preferences, then a substantial benefit (on the order of $4,000 to $5,000 per year, by my estimate) would be possible without any increase in marginal tax rates.
If we allow double dipping for UBI and UI benefits, as in the FPZ model, the moral hazard effect of UI becomes stronger. If instead, as I recommend, we disallow double dipping, introducing a UBI reduces moral hazard.

As far as I can see, the FPZ model could be modified to include means-tested welfare in the baseline assumptions, to change the source of financing for the UBI, and to prohibit double dipping. I am not equipped to rerun the model in order to obtain accurate quantitative results. However, it is evident from the qualitative arguments made above that doing so would improve the scoring of a UBI relative to UI, would raise the estimated optimal level of the UBI, and would (possibly) indicate that a combination of the two policies, without double dipping, would be better than either one alone. Perhaps the original authors or other researchers will undertake the necessary modeling. Absent a revised set of model results that incorporate the modifications I have suggested, I caution UBI skeptics against using the FPZ results as “proof” that “a UBI would not work.”

At the same time, however, I recommend that UBI proponents take the FPZ study as a word of caution. First, their paper is right to point out that the work disincentives of a UBI become more pronounced the more generous the level of benefits. Although I think a modest UBI is workable, I would have misgivings about one that allowed people to live in middle class comfort on their basic income grant alone. Second, the disincentive effects highlighted by the FPZ model would pose a real risk if a UBI were simply added on top of our current welfare system, rather than replacing large parts of it. Finally, I agree with FPZ that financing a UBI by increasing marginal tax rates would risk significant disincentive effects.

Could We Afford a Universal Basic Income?

A Universal Basic Income: Conservative, Progressive, and Libertarian Perspectives

A Universal Basic Income and Work Incentives. Part 1: Theory

A Universal Basic Income and Work Incentives. Part 2: Evidence