The Cost of Basic Income in the United Kingdom: A Microsimulation Analysis

Despite many authors pointing out the error, misplaced focus on gross cost is common in the literature both within the UK and around the world.

Perhaps the most egregious example in the Portes et al. (2017). They claim that government-provided social services could deliver about ten times the “user-value” (as they call it) for the poor than a “cost-equivalent UBI”. While Portes et al. allow for second-order effects to be taken into account for their definition of “user-value”, their estimate of the “cost-equivalent UBI” solely rests on the gross cost of their scheme. Indeed, their stark conclusion can only be supported by such a lopsided comparison, which exaggerates the cost of UBI by 3-to-10 times.

Similarly, Piachaud (2016) claims that introducing a UBI would be prohibitively expensive. In a stylised example, he argues that a UBI of half the mean income (which would be close to the poverty line) would require an additional income tax of 50% over and above the current rate. In his example, current government spending on social security, which he estimates at one third of total government spending, would be completely abolished. This leads to a tax rate in his example of two-thirds of current taxes plus an additional income tax rate of 50%. He concludes that such a high tax rate would probably be untenable. However, his calculation greatly overstates his case by disregarding the amount of taxes citizens essentially pay to themselves in the form of UBI. His article demonstrates only how people can use faulty reasoning to arrive at prohibitively high cost figures that don’t reflect the benefits and burdens of financing a UBI scheme.

Miller (2017, chapter 14), avoids the gross-cost error without entering the debate over it. Although she recognises the importance of the net cost of a UBI scheme as the relevant cost measure (p. 149), the question that interests her is the required income tax rates necessary to finance them. One can answer this question without calculating either gross or net cost. She concludes that flat-rate income taxes necessary to maintain balanced-budget financing for a set of increasingly generous UBI schemes ranging from between 27% and 47%, compared to the progressive income tax rates of 20% to 45% currently in place. This information is useful and all Miller needs for her purposes, but it doesn’t answer the questions that are the focus of this article: who financially benefits from the UBI scheme; who financially contributes to it, and how much does that scheme financially benefit or cost them?

Regarding the use of microsimulation to evaluate UBI proposals, Malcolm Torry has extensively explored the feasibility of different proposals in the UK context over several years.⁵ His approach, however, differs substantially from the present paper, both in motivation and implementation. Torry investigates only partial UBI proposals (coalescing around an annual UBI grant of roughly £3,600 or approximately half of our own proposal). He also imposes strict constraints on financing sources and retains most of the current social security system. Additionally, Torry’s recent article focusses on UBI schemes which could be implemented on a revenue-neutral basis with only modest increases in income tax. In contrast, the present paper primarily illustrates the conceptual point that the relevant cost perspective for evaluating any UBI proposal is the net cost approach. Like Miller, Torry understands this issue but does not discuss it in any detail.

We use microsimulation to provide a more accurate consideration of UBI’s net costs. The present paper therefore extends the discussion on the effects of implementing a UBI scheme in the UK in a different direction than Torry’s work.

Martinelli (2017a; 2017b) uses microsimulation to analyse the effects of implementing different UBI schemes in the UK, but he neglects the net-cost issue. His most generous proposal has a gross cost similar to our own at £427 billion. He focusses on how much the elimination of different social security services might contribute to covering the gross cost of UBI, and explores different distributional as well as incentive effects. Martinelli therefore leaves the question of the actual cost which needs to be covered from outside the scheme open. Where appropriate, the present paper will compare Martinelli’s results regarding distributional effects to our own findings below. We omit discussing incentive effects in the present papers given our different focus.

Outside of the UK, Robert Greenstein writes, “There are over 300 million Americans today. Suppose UBI provided everyone with $10,000 a year. That would cost more than $3 trillion a year — and $30 trillion to $40 trillion over ten years.” Nowhere in the article does he mention the net cost or that these gross cost figures fail to reflect any real costs of implementing UBI. Canarie (2019) responds to Michael Howard argument that gross cost is not very meaningful (Howard, 2018) with the unsupported assertion, “[O]f course it is.”

One of the authors of this article, Widerquist (2017a; 2017b), has tried to correct the gross-cost error as he estimates the cost of a poverty-level UBI in the United States (US) using what he calls a “back-of-the-envelop” approach—i.e. a highly simplified methodology that makes it possible to estimate the cost of UBI directly from Census Bureau income tables without microsimulation. He finds that the US could implement a UBI at the cost of roughly $539 billion, 2.95% of GDP, about one-sixth (15.7%) of its gross cost. This figure “nets out” (accounts for) the amount people pay to themselves but not the costs and benefits of integrating UBI into the existing US tax-and-benefit system—i.e. the cost of “UBI in a vacuum,” so to speak.

Although we share the net-cost perspective and the goal of correcting the gross-cost error, our figures are an advance on Widerquist (2017a; 2017b) in three ways: they involve a more sophisticated microsimulation methodology, they net out the cost of existing programs that can be replaced by UBI, and they net out the tax replacement cost of UBI (more on these differences in sections 4.2 and 6.1).

Microsimulation and Tax-Benefit-Models (“TBM”) based on micro data have been used to assess costs and effects of social policy changes for at least 50 years (Sutherland and Figari, 2013).

The general idea of a TBM is simple: it represents all policy rules relevant for calculating social security payments as well as direct (income) taxes and social security contributions. In a first step, the TBM calculates the transfers between individuals and the state (e.g. social security payments and taxes) based on socio-economic data like age, income and household composition. The socio-economic data is usually collected at the individual or household level from national surveys such as the Family Resource Survey in the UK or the EU’s Statistics on Income and Living Conditions. By calibrating the model with observed transfer payments from the data, TBMs can accurately estimate the relevant taxes and benefits in a given year. In a second step, such a calibrated TBM can then be used to investigate the effects of changing individual policy rules like tax rates, benefit levels or allowances using the same set of micro-data as for the calibration. In this way, a TBM allows predictions about future effects of new policies.

Of course, TBMs have their limitations. They are, for example, usually only able to show the effects of policy reforms in a static framework without accounting for people’s behavioural changes caused by the policy change. They nevertheless represent one of the few tools to empirically investigate the effects of social policy proposals.

Considering the relevance of financial feasibility to the UBI debate and the value of TBMs in addressing the feasibility question, empirical works based on microsimulation are relatively rare but not absent in the UBI literature. Their use for evaluating UBI dates back at least as far as Atkinson’s work toward the end of the last century (Atkinson, 1995). In preparation of pilot projects carried out in recent years, national agencies also used microsimulation to explore results which could be expected from the pilot projects (e.g., Kela, 2016). In other cases, UBI schemes serve as a benchmark to evaluate other social policy proposals via TBMs (e.g., Jessen et al., 2015).

We specify our illustrative UBI at roughly the UK’s official poverty line, which uses the most widespread definition of poverty within OECD countries: a relative poverty line at 60% of median income (OECD, 2010). The equivalised median income for couples in the UK was £26,801 in the fiscal year 2018-2019 (Department for Work & Pensions (2022), table 2b.⁶ This yields a poverty line of £16,081 for couples.⁷

We choose a UBI for adults of exactly half that amount: £8,040 per person per year. Therefore, our illustrative UBI would provide a poverty-level income for two adults living together (£16,081), but it would be below the poverty line for one individual living alone (£10,774, see Table 1 below). We choose this level for the UBI because, on the one hand, it substantially reduces poverty and enables us to substitute most means-tested social security programs. On the other hand, it reduces the total cost of the scheme by taking advantage of household economies of scale—i.e. that the average living cost per household member decreases as the number of people in the household increases.

In the context of a UBI, the issue of household size is particularly relevant because the UBI itself, being unconditional, does not account for the number of household members. Because a couple living together can live affordably on less than double the income of a person living alone, an (unconditional) UBI ends up either leaving couples substantially above the poverty line or leaving adults living alone slightly below it.⁸ Therefore, it is impossible to set any individual UBI at exactly the poverty line. While our illustrative scheme will not eliminate poverty entirely, it will bring everyone much closer to the poverty line. Therefore, we say it is set at “roughly” the poverty line.

For children below the age of 18, we set the UBI at 0.15 the median income, or £4,020 (half the adults’ UBI). This amount is between the poverty line for children age 14 and older (£5,086) and younger children (£3,083). For the average-sized UK family (1.88 adults, 0.51 children; see Office for National Statistics, 2018b, table 21), these specifications result in UBI grants totalling £17,166 per year.

The net-cost representation of a UBI scheme requires setting both the level of UBI and the net beneficiaries’ income tax rate. Only the combination of these two parameters fully specifies the gains a UBI scheme provides for net beneficiaries, and only the combination allows us to calculate the financial cost of providing these benefits. For our illustrative UBI scheme in this paper, we set the income tax rate for net beneficiaries at 50% flat. This figure is not an add-on but a top-up: the total income tax rate for net beneficiaries once the scheme is in place. Whatever net beneficiaries’ tax rate was before, at any income level in the net-beneficiary-range, it increases to 50%. From a political perspective, this marginal tax rate might be a difficult sell in the short-term (see e.g., Torry, 2019), but from an economic or distributional perspective, it is worth keeping in mind that net beneficiaries by definition benefit financially from the introduction of our scheme, most of them substantially so. This increase in their marginal tax rate comes along with a significant increase in their total income after taxes and transfers.

Although the 50% marginal tax rate is a top-up for income taxes, it is an add-on to the effective marginal rate at which some remaining means-tested benefits, mostly particularly, the Housing Benefit, taper off. This means that some beneficiaries will experience an effective rate in excess of 85% over a small range of income. We recommend reducing the rate at which means-tested benefits taper off, even folding them into our 50% rate. Unfortunately, data limitations and the complexity of the existing Housing Benefit system don’t allow us to make this calculation. Fixing this problem would increase the overall cost of the program, but not by very much, because the number of people affected is moderate and the high rate occurs over a small range of income.

Readers might ask the question what is the tax rate for net contributors. This article does not address that question: it focuses exclusively on the question how much does UBI cost, leaving the question how can or should the UK finance that cost for future research. We said above that the net cost of UBI is affected by the fact that nearly everyone both receives UBI and pays at least some taxes. To determine how much UBI costs, we need to calculate the difference between how much net beneficiaries pay in taxes and how much they receive in UBI, and so we have specified the marginal tax rate for net beneficiaries. But we do not need to know how much net contributors pay and receive in UBI, and so we do not need to specify the tax rate for net contributors. We know that they paid enough income taxes at the 50% rate to equal their UBI, but we do not need to know (for the purpose of determining how much UBI costs) whether they continue to pay at the 50% rate, whether it is higher or lower, whether their taxes are levied on some base other than income (such as wealth, resource-ownership, pollution, inheritance, and so on).

The scheme does require all net contributor’s income taxes to up just enough to match their UBI, but whether their income taxes go up any more than that depends on how the UK decides to finance this scheme. We briefly discuss financing options in section 6 and leave a more thorough discussion for future research.

The definition of an income tax schedule for net beneficiaries also specifies the “break-even point” for those beneficiaries: The break-even point indicates the income level at which net beneficiaries, given the composition of their household, become net contributors. Put differently, the “break-even point” is the income level at which a household’s UBI exactly matches the taxes they pay.

A tax rate of 50% implies a break-even point for a single-person household at a market income of twice the individual’s UBI grant. Because income taxes in the UK are collected at the individual level (a procedure that we retain), while UBI grants are likely shared to some extent between household members, break-even points for households change with their composition (see Table 1 below).⁹ See also Torry (2019) for a more extensive discussion of differences between collecting taxes on the basis of individuals or households.

For the average UK family, the break-even point of our scheme is £34,333 which is well above the average family’s equivalised median market income of £28,474.¹⁰ The break-even point for the average family is also close to the 70^th percentile of the current income distribution.¹¹ Put differently, around 54% of households would, to different degrees, become net beneficiaries of the UBI scheme, assuming a uniform distribution of household sizes across the income spectrum.

Table 1 below shows UBI grants for different family compositions and how these compare to the poverty line of those model families.¹² The table also shows the break-even points for different family compositions.

We use UKMOD version 3.4.10 as our Tax-Benefit-Model. UKMOD is a Tax-Benefit Model covering the United Kingdom. It “simulates individual and household tax liabilities and benefit entitlements according to the policy rules in place in each member state” (Sutherland and Figari, 2013, p. 5).¹³

We use the 2019-2020 policy year as the basis for our analysis, because it is the last policy year that is largely unaffected by the COVID-19 pandemic, which had many transitory economic effects that are only now dissipating. The fiscal year 2019-2020 ended in March just as the Covid lockdown was beginning, and so our data is only very minimally affected by Covid disruptions. The micro data underlying the model is taken from the 2019-2020 UK Family Resource Survey.

Programs that might conceivably be replaced by UBI can be divided into three broad categories: means-tested benefits (MTB), contributory benefits, and non-means tested, non-contributory benefits.

• MTBs are dependent on a person’s current financial situation and usually require extensive oversight to monitor the eligibility for the benefit. Examples of such benefits in the UK are Social Assistance or Housing Benefits. The adverse effects of MTBs on privacy, self-worth, or saving rates are well documented (see e.g. Torry, 2018, pp. 59-63).

• Contributory benefits depend on previous contributions made by the individual. Such programs encompass most forms of pensions and unemployment benefits.

• Non-contributory, non-means-tested benefits depend on a set of eligibility conditions other than past monetary contribution. Such programs include, for example, most forms of disability benefits, student support, or child benefits.

Various authors argue that one of the central benefits of implementing a UBI is its ability to replace MTBs and hence reduce their adverse effects (see e.g. Miller, 2017, pp. 42-43). The major MTBs currently administered in the UK are listed in Table 3 below. Because our UBI scheme roughly equals the poverty line and is intended as an income floor enabling households to meet all basic necessities, it has the same general function as an MTB. Hence, most, but not all, MTBs should be replaced by our scheme.

The only MTBs we retain are housing benefits. Housing benefits pose a special case of MTBs because their primary role is to alleviate geographical income differences, providing more funds to low-income households living in high-rent environments, and the unconditional nature of a UBI is ill-equipped to deal with these spatial differences. The inability of UBI schemes to directly replace housing benefits is a topic which is recognised by many authors (e.g. Kela, 2016, p. 12; or Miller, 2017, pp. 90-91). Housing benefits constitute one of the substantial unresolved challenges when considering the implementation of a UBI: Housing benefits in their current form require strong oversight and exhibit most adverse aspects of MTBs. However, overcoming the challenges posed by housing benefits is outside the scope of this paper.

Regarding contributory programs, different approaches are imaginable. If one regards contributory programs as a service – albeit a mandatory one – by the government to its citizens, one could argue to retain them under a UBI scheme. There are certainly arguments for both keeping or abolishing them, and resolving this issue is again outside the scope of this paper. For the illustrative purposes of this paper, we abolish contributory programs, mostly because there is less need for them under our scheme: contributory social security programs are explicitly intended to alleviate the cost of MTBs in requiring able individuals to insure themselves from adverse events like unemployment or old age (Goodin, 1992). Pension schemes can especially help mitigate time-inconsistent saving preferences,¹⁴ or serve as a politically feasible redistribution program (e.g. Cremer and Pestieau, 2011). Because a UBI is designed to meet basic social needs, all of these justifications carry substantially less weight than before.

It is noteworthy, however, that contributory programs as defined here do not encompass healthcare contributions. While healthcare contributions are partly intended to counteract losses in earnings from illness, they mostly serve to cover the actual costs of treating the illness. As such, healthcare exists largely outside the work environment which we are considering here, and is thus retained. Because healthcare contributions are not part of UKMOD either, they do not enter any of our analysis below.

Lastly, the third category of non-means-tested, non-contributory social security programs is designed to meet special needs arising from special circumstances. Because the UBI is not well equipped to account for individual differences in beneficiaries, most of these programs should be retained under a UBI. The only major program of this category which can naturally be replaced by a UBI is the Child Benefit. We don’t retain it separately because our illustrative UBI scheme also encompasses a payment to everyone under the age of 18 years which is more generous than the Child Benefit.

Our replacement strategy has at least three drawbacks. First, even if UBI becomes politically viable, the question of which programs should or should not be replaced by it will probably remain controversial. Second, some contributory benefits might be motivated by market failures, such as under-rewards for some types of behaviour. Third, the goal of almost any UBI supporter is not to redistribute money among low-income people but to redistribute money from high-income or wealthy people to low-income people in general.

We have addressed this third problem by choosing to “hold” the transition “harmless” for people in the bottom 20% of the income distribution. Simply substituting our UBI scheme for all the above programs would cause a small number of households in the bottom 20% of the income distribution to lose out financially, and we don’t want to aggravate the situation of the most vulnerable.

The households losing under the UBI scheme are usually single people who receive a large amount of social security payments under the current system, or otherwise disproportionally benefit from the current system. It is difficult to identify the specific circumstances of every such household and the reasons they lose under this UBI. The impact of this intervention on general financing is very small, however.

Therefore, we impose a “hold-harmless” condition in our model which raises the income of all households in the bottom 20% losing from the introduction of the UBI to their pre-UBI income level. This intervention costs about £15 million; a negligible share of the total cost stated above. Most net beneficiaries in the next two quintiles gain on average, but a minority loses more than they gain by the replacement of the selected programs with our UBI scheme. It is possible to extend the hold-harmless condition to people with higher incomes, but we have not estimated the cost.

Table 3 below summarizes our UBI scheme’s treatment of all major social security programs currently in place in the UK. Italics denote programs retained under the UBI scheme.

We ask readers to bear in mind that our proposal is illustrative. We are not committed to it as necessarily the “best” UBI scheme. It gets people an idea of the cost of UBI by focusing on one specific scheme. We suggest readers make up their own mind whether it might be worth spending a little more than our estimated cost and/or accepting a little more complexity than our greatly simplified social security system to retain such-and-such a policy. Readers might equally well decide to save a little money by cutting a few more programs.

The net cost of this illustrative UBI scheme is far lower than the often-mentioned but not very useful concept of “gross cost.” Netting out only the amount people pay to themselves (and ignoring the additional savings generated by integrating it into the existing tax-and-benefit system) shows that the cost is £139 billion per year, about one-third (39.5%) of gross cost (£472 billion per year). Once we net out the cost of integrating this UBI scheme into the existing tax-and-benefits system, the net cost comes to only about 15.4% of its gross cost.

One clear ramification of our findings is how much more useful net cost is than gross cost. Not only do gross-cost figures wildly exaggerate the cost of UBI, they are also not useful in determining who benefits and by how much; nor are gross-cost figures useful in determining who bears the financial burden and how much it costs them. These are exactly the things one wants to learn from a cost study. Only with net-cost analysis was this study able to determine that 54% of the UK population benefits from this UBI scheme; that the benefit would be about £45 billion (2.0% of GDP); and that that benefit would be a financial cost to the remaining 46% of the population.

We’ve argued that the gross cost of UBI is not merely less useful than net cost but that it is meaningless. Now we can further explain why. To determine that 54% of UK families would benefit by a total of £45 billion per year, we had to specify not only the size of the UBI but also a wider UBI scheme including the tax rate on net beneficiaries (50%). Had we made the tax rate on net beneficiaries higher, people would have reached the break-even point sooner; there would have been fewer net beneficiaries; the cost of UBI would have been lower; and more net contributors would have been available to share that cost. Therefore, the financial burden on each one would have been lower. The gross cost of UBI is independent of the tax rate on net beneficiaries, and therefore, it could not capture any of those changes.

Similarly, had we made the tax rate lower, people would have reached the break-even point later; there would have been more net beneficiaries; the cost of UBI would have been higher; and fewer net contributors would have been available to share that cost. As before, the gross cost being independent of the tax rates would not reflect any of these changes. Cost figures for distributional programs are meaningless if they cannot show what the financial burden is or who remains to bear that burden after beneficiaries are accounted for. Gross cost figures can’t do any of that or contribute to a discussion of it.

One might ask, what if we introduced a UBI with no increase in any taxes of any kind. Then the gross cost would be relevant. That is true, but it’s only true in that case because the gross cost happens to be equal to the net cost. There is nothing inherently useful about gross cost if and when it differs from net cost.

Although a net cost of £45 billion (2.0% of GDP) is almost certainly sustainable and affordable, we do not recommend financing it solely by increasing the income tax on net contributors.

Partially financing our scheme through higher income tax rates might be feasible and desirable: empirical evidence suggests that welfare-maximising marginal tax rates for top-income earners are well above current levels (Bach et al., 2012). However, financing our entire scheme from income taxes would still require extremely high marginal tax rates on upper-income earners, and other options exist that tax the same group but don’t involve high marginal tax rates. Demonstrating why a purely income-tax financed UBI would involve high marginal tax rates is beyond the scope of this article.

It’s important to remember though that high marginal tax rates do not imply a high overall tax burden. For example, the net beneficiaries in our study face a relatively high marginal income tax rate of 50% when overall, they effectively pay negative taxes on their incomes (by receiving more in UBI than paying in income taxes). Even if the overall tax burden would not be excessive, high marginal rates can have negative effects, and therefore, we should consider options that involve lower marginal rates.

Such options include the value added tax or VAT, which we do not recommend because of its regressive aspects. Better options include resource and rent taxes, wealth taxes, and financial or technology taxes. A thorough discussion of the merits of each of these taxes is beyond the scope of this article, but it also is a promising area for further research.

The most important aspect of the financing issue is that 2.0% of UK GDP can be easily raised out of taxes on the wealthiest 46% of UK households. The UK is a highly unequal country (relative to most industrialized countries although not to the US), and inequality has been rising in the UK for decades, meaning that higher earners can pay higher income, wealth, resource-use taxes, and so on without eliminating the disproportionate benefits they’ve received from the past 50 years of unequally experienced income growth. The money is available, if the UK chooses to use it to eliminate poverty and create a more equal society.

Our analysis is entirely static. While TBMs are valuable tools for exploring static effects, they are silent on changes in the behaviour of both net beneficiaries and contributors over time. Given the strong effects of introducing a UBI on income distribution and, at least in our scheme, tax rates, we expect dynamic effects to change the results of our analysis substantially. These dynamic effects have at least three layers: first, direct changes to individual behaviour with respect to labour supply, saving and other factors; and second, the likely response of firms to these changes in individuals’ behaviour encompassing, for example, the adjustment of wages, prices or output. These first two layers might have considerable effects on the cost of UBI. The expected initial effect of people working fewer hours will shift the tax burden of UBI; the labour demand response is likely to increase wages and improve working conditions overtime, at least partly counteracting the decreased labour supply effect.

The third layer of dynamic effects—probably the hardest one to measure—will also have substantial effects. It is expensive to be poor; it is expensive to live in a country that has poor people. Poverty is correlated with ill health, accidents, crime, violence, low education, and many other problems all of which generate enormous costs for society. Children who grow up in poverty suffer long-term negative effects that are costly for them and for society as a whole for the rest of their lives. Research shows that reducing poverty with cash transfers can have positive effects on these and many other problems with the potential that a UBI scheme like this one could save enormous amounts of money over time, maybe shifting the cost-benefit analysis dramatically (Forget, 2011; Lewis, 2005; Pereira, 2017; Widerquist and Lewis, 2006; Wilkinson and Pickett, 2009).