Revenue and distributional modelling for a UK wealth tax

In this paper, we model the revenue that could be raised from an annual and a one-off wealth tax of the design recommended by Advani, Chamberlain and Summers in the Wealth Tax Commission’s Final Report (2020). We examine the distributional effects of the tax, in terms of both wealth and other characteristics. We also estimate the share of taxpayers who would face liquidity constraints in meeting their tax liability. We ﬁnd that an annual wealth tax charging 0.17 per cent on wealth above £ 500,000 could generate £ 10 billion in revenue, before administrative costs. Alternatively, a one-off tax charging 4.8 per cent (effectively 0.95 per cent per year, paid over a ﬁve-year period) on wealth above the same threshold, would generate £ 250 billion in revenue. To put our revenue estimates into context, we present revenue estimates and costings for some commonly proposed reforms to the existing set of taxes on capital.


INTRODUCTION
The Wealth Tax Commission studied the possibility of a wealth tax for the UK, 1 and delivered a body of research into the desirability and deliverability of such a tax. A crucial aspect for politicians in deciding whether or not to support a tax on wealth is how much revenue it could raise. Meanwhile, public support will hinge at least partly on how much people could be asked to pay.
In this paper, we model the revenue that could be raised from an annual and a one-off wealth tax. We first consider an annual wealth tax, and calculate the tax rates that would be needed to raise £10 billion in revenue at various possible thresholds, taking into account likely behavioural effects. We analyse the distributional effect of these tax structures in terms of who pays and how much, both across the wealth distribution and across other characteristics. We also analyse who is likely to face liquidity constraints. For some of these tax structures, we look at how the revenue raised compares with the administrative burden, for both the government and the taxpayer.
We find that a wealth tax could raise a substantial amount of revenue at relatively modest tax rates. For an annual wealth tax, a flat tax of 0.17 per cent on wealth above £500,000 could generate £10 billion in revenue, but at a (proportionally) high ongoing administrative (admin) cost to government of £1.2 billion. The admin costs to taxpayers are even higher, at £7.8 billion per year, increasing substantially the effective tax rate inclusive of all taxpayer costs. At higher thresholds, higher tax rates are required to generate a similar amount of revenue, but aggregate admin costs are lower as there would be fewer taxpayers. For example, at £2 million, the ongoing admin costs to government fall to only 1 per cent of the revenue raised, comparable with other major taxes. 2 However, with a £10 billion revenue target, costs to the taxpayer amount to a quarter of revenue raised. This effectively adds a cost of 0.13 per cent of wealth to the headline rate.
We then consider a one-off wealth tax. Because this is a one-off event, certainly not something that would be seen for at least another generation, we consider the tax rate needed to raise £250 billion: the equivalent of raising an effective annual revenue of £10 billion per year over a 25-year period. We also consider a flat rate of 1 per cent per year for five years, as an alternative benchmark. We perform an analogous set of analyses, studying distributional effects, liquidity constraints and admin costs.
We find that a one-off wealth tax charging a tax rate of 4.8 per cent on wealth above £500,000 would generate £250 billion in revenue, before admin costs. This would come at a total cost of £1.7 billion to the government, and £7.8 billion to the taxpayer. Because this is a one-off event, it is possible to achieve a much higher ratio of revenue to cost than under an annual wealth tax. A higher threshold would reduce the admin costs further, though achieving the same amount of revenue would necessitate higher rates. We estimate that with a threshold of £1 million, a one-off tax of 8.5 per cent -or a fiveyear annualised rate of 1.7 per cent -would be required to raise £250 billion. Under this tax structure, the cost to the government would be 0.4 per cent of the revenue raised. Taxpayer costs would be 1.8 per cent of revenue raised, effectively adding a further 0.14 per cent to the headline rate.
Under a one-off wealth tax with an exemption threshold of £500,000 generating £250 billion in revenue, we estimate that 6.4 per cent of individuals would face liquidity constraints. This rises to 25 per cent with a threshold of £2 million, though the absolute number of liquidity-constrained taxpayers would fall, from 526,000 to 154,000. Fewer taxpayers would be liquidity-constrained under an annual wealth tax generating £10 billion in revenue.
Finally, we compare the revenue we could raise from a wealth tax with the revenue that might be raised by alternative reforms to the taxation of capital: in particular, reforms to capital gains tax (CGT), taxation on dividend income, inheritance tax (IHT) and council tax. Numerous reforms to these taxes have been proposed in recent years. 3 We focus on a few headline reforms that are common to almost all reform proposals.
A brief comparison shows that it would be possible to raise similar amounts of revenue to an annual wealth tax, or more, through some of the proposed reforms we examine. For example, raising tax rates on capital gains to be in line with those on income would raise an additional £12 billion, with little implied cost to tax authorities, and equalising rates on dividends would raise around £5 billion. However, a reform such as revaluing housing assets (and reforming rates) for council tax could raise even more substantial amounts, for around half the one-off cost to the government of implementing an annual wealth tax. These alternative reforms do not necessarily avoid some of the challenges inherent in implementing a wealth tax, such as the cost and difficulty of valuing assets.
The remainder of the paper is organised as follows. In Section 2, we describe the data we use and adjustments we make to better capture wealth held at the top. In Section 3, we present our revenue modelling for an annual wealth tax, explaining who would pay it and how many taxpayers would face liquidity constraints. We also discuss the effect of banding on our revenue estimates. In Section 4, we present similar analysis to Section 3, this time for a one-off tax on wealth. In Section 5, we provide a brief analysis of the revenue that could be raised from alternative reforms to the current tax system. We conclude in Section 6.

Wealth and Assets Survey
Our primary data source is the Wealth and Assets Survey (WAS) conducted by the Office for National Statistics, Social Survey Division (2020), which is the most comprehensive data source on wealth in the UK. We use the most recent wave of the data, collected in 2016-18, which covers around 40,000 individuals. The data exclude certain geographical regions, in particular Northern Ireland, and the area north of the Caledonian Canal. Individuals living in institutional settings, such as care homes, halls of residence and prisons, are also beyond the scope of the survey. As a result, we miss around 2 per cent of the UK population.

Sunday Times Rich List
A key caveat to relying on the survey data alone is that these data under-represent wealth held at the very top of the distribution. In response to this, we follow the approach set out in detail in Advani, Bangham and Leslie (2021), which supplements the WAS with information from the Sunday Times Rich List (STRL). We make a few adjustments to this methodology, as described below. The STRL captures, in theory, the 1,000 richest people or families in Britain. The compilation and measurement of wealth held by STRL individuals draws heavily on their observable business assets. More private forms of wealth, such as financial assets, are generally not captured. We proceed under the assumption that the wealth captured primarily reflects business wealth, and that the total wealth recorded is likely to be an underestimate of the wealth held by these individuals.
Though there are 1,000 entries in the STRL, some entries include multiple individuals, such as a husband and wife or other members of the same family. To be consistent with our use of individuallevel data in the WAS, we treat each individual named in the STRL as a separate unit. Where there are multiple named individuals per rich list entry, we divide wealth equally among them.
In contrast to , we use data from the 2020 STRL. We rescale this to match the aggregate wealth in the 2017 and 2018 (average of the two) lists, to be consistent with the time period in which the WAS data were collected. Our reason for doing this is to leverage information on the country of residence of STRL individuals, which we obtain using matched records from Companies House. Individuals need not have the UK as their main country of residence to be included in the STRL, and it is not clear that all individuals would qualify as resident for tax purposes. This is important for our revenue analysis, as we seek to establish how much revenue could be raised from those who are likely to be eligible to pay.
Though tax residence is not a readily observable characteristic, we can proxy for this using information on country of residence as recorded in the Companies House register. Most individuals in the STRL own or control part of a company registered with Companies House, and these companies are required to submit information on their directors or 'persons with significant control', including information on their usual country of residence. 4 We have matched individuals named in the STRL to Companies House records, using name and date of birth as our matching criteria. We were able to match 83 per cent of the 1,242 named individuals automatically. A further 149 were matched manually, having been missed often as a result of different variants of their name being used across the two data sources. For the remaining unmatched 5 per cent of individuals, we impute information on their country of residence based on the percentage of individuals in each five-percentile bin of the STRL who are UK resident. In total, we classify 84 per cent of individuals as UK resident, although they make up only 74 per cent of the wealth in the STRL.

Pareto imputation
We account for wealth missing from the top of the distribution in two ways. First, we add individuals named in the STRL to our WAS data, removing the handful of individuals whose wealth overlaps with the STRL to ensure they are not accounted for twice. Second, as in , we implement a Pareto adjustment to estimate the amount of additional wealth that should be captured in the top tail of the wealth distribution. The first step increases our estimate of total wealth from £14.2 trillion to £14.6 trillion. The second step increases this to £14.8 trillion.
It is widely observed that the top tail of the wealth distribution approximates a Pareto distribution. 5 By fitting a Pareto distribution using data on wealthy individuals in the WAS combined with the STRL, we estimate the amount of excess wealth that is implied by the shape of the distribution. We refer the reader to  for full details of this approach.
We implement our Pareto adjustment to the distribution of business wealth (including shares), rather than using total wealth as recorded in the WAS. This is to ensure consistency with what is captured in the STRL, which we believe to be primarily business wealth. We choose a relatively low threshold of £500,000 in business wealth, though in practice the chosen threshold has little impact on the amount of additional wealth estimated. Using this approach, we estimate that there is an additional £280 billion in wealth in excess of the wealth recorded in the WAS and the STRL. This differs from the estimate in  for two reasons. First, their paper uses household, rather than individual-level, data to implement the Pareto adjustment. Second, they assign each entry in the STRL a single household weight of 1, whereas we assume each entry represents the number of individuals explicitly named in the STRL.
To estimate the amount of revenue that could be raised from individuals across the wealth distribution, we must allocate this additional Pareto wealth to observations in our data. We do this by assigning to each individual in our Pareto sample the amount of business wealth they would be expected to have according to their rank in the distribution. We then redefine each individual's total market value wealth, and total chargeable wealth, replacing their reported business wealth with the amount implied by the Pareto distribution.
For the purpose of analysing liquidity issues, it is essential to know how an individual's chargeable wealth compares to their income. However, by adjusting wealth at the top of the distribution, we have distorted this relationship. It is not clear how one could model a top wealth or income adjustment that accurately captures the relationship between these two variables at an individual level. Moreover, we do not wish to overstate the extent of liquidity issues by assuming that wealth has been under-reported while income is accurately captured. As a result, we have chosen to preserve the ratio of wealth to income as it is reported in the WAS. We do this by scaling net income by the ratio of an individual's adjusted to unadjusted wealth. At an individual level, the ratio of wealth to net income is therefore consistent with the liquidity analysis undertaken in Loutzenhiser and Mann (2021). This adjustment makes very little difference to our estimates: under each of our proposed tax structures, the percentage of taxpayers deemed to be liquidity-constrained is within half a percentage point of the result obtained when reported income is used.
In our revenue analysis, we use the WAS data augmented with the STRL and our Pareto adjustment. These data are also included in our analysis of the distribution of taxpayers by age and sex. 6 However, we do not have information on their income, asset composition or region of residence. STRL individuals are therefore excluded from our analysis of the distribution of taxpayers by region, liquidity constraints and the asset composition of taxpayers.

Approach
We model a wealth tax that is consistent with the recommendations outlined in Advani et al. (2020b) using the data on wealth above various thresholds described in the previous section. Our tax covers all adult individuals. Children are not taxed as separate tax units. Instead, their wealth is aggregated with the wealth of their parents. We assume that, in practice, parents would be able to choose who their children's wealth is allocated to, and would do so to minimise their joint tax liability. Accordingly, we allocate children's wealth reported in the WAS to the lower wealth parent, splitting any excess equally.
In this way, we preserve the wealth ranking between parents. Though data from the STRL are included in our analysis, in our main specification we exclude individuals who are classified as non-residents according to their Companies House records, targeting tax residence as the relevant connection criterion.
We also take into account behavioural responses. As outlined in Advani and Tarrant (2021), a net wealth tax is likely to elicit a number of avoidance and evasion responses, including under-reporting, offshore evasion, gifting and fragmentation, asset portfolio recomposition, saving responses, labour supply responses, and migration. Advani and Tarrant (2021) conclude that under a well-designed wealth tax covering all asset classes -as we assume ours will -the overall magnitude of behavioural responses could be limited to a 7-17 per cent reduction in wealth in response to a 1 per cent tax rate on wealth. In our revenue modelling, we take the upper bound as the 'high avoidance' scenario, and the lower bound as the 'low avoidance' scenario. We apply this response to the average tax rate faced by each individual under each tax structure. For example, for an individual facing an average tax rate of 0.5 per cent, we reduce their chargeable wealth by 3.5 per cent in the low avoidance scenario.
The figures 7 per cent and 17 per cent represent average behavioural responses, summarising the combined effect of each individual's response along the different margins. By applying this statistic in the way we do, we will miss heterogeneity in avoidance responses across individuals. In practice, some individuals will respond much more than others, and they will respond along different margins. For example, these statistics partly reflect migration responses. Rather than modelling who will choose to migrate and who will stay, we attribute the reduction in aggregate wealth arising from some migration, to a reduction in the wealth of all individuals. In doing so, we assume that behavioural responses are uniformly distributed across the wealth distribution. While this may not hold in practice, we currently have very little empirical evidence on the distribution of behavioural responses to a wealth tax. 7 If the wealth tax were at a flat rate (above some threshold), any heterogeneity would not affect our revenue estimates. If the tax were progressive, our revenue estimates could be affected; however, in the absence of information on how behavioural responses are distributed, it is not possible to even sign the direction of the effect. In our distributional analysis, we focus on who should pay the tax, and how much they should pay, rather than the amount they would pay after taking behavioural responses into account. This will be unaffected by our method of accounting for behavioural response in our revenue calculations.
Using household-level data, we can explicitly model the potential revenue effects of asset splitting between spouses. In addition to our core revenue estimates, which account for avoidance using the elasticities provided in Advani and Tarrant (2021), we estimate the amount of revenue that would be raised if all assets excluding pensions were split equally among the head of household and their spouse. We do not split pensions, as these cannot typically be transferred across individuals. We compare this to the revenue that would be raised if ownership among spouses were unaffected by the introduction of a wealth tax. In doing so, we assume no other avoidance responses. This is because our overall estimate of 7-17 per cent already accounts for asset splitting, and it is unclear what proportion of this average elasticity can be attributed to asset splitting relative to other responses.
To calculate net revenue, we estimate the admin costs that the taxpayer and the tax authority would face on an ongoing basis. Admin costs to the taxpayer partly reflect the valuation costs to taxpayers who are likely to need professional valuations for their asset portfolios. A range of cost estimates are calculated in Daly, Hughson and Loutzenhiser (2021); here, we estimate a 'central' scenario based on valuation costs amounting to 0.4 per cent of the value of hard-to-value assets (with the same overall cap of £50,000 per taxpayer).
Evidence presented by Burgherr (2021) suggests that there is also a fixed cost of filing, which, while generally dwarfed by the valuation cost, should be accounted for. We add a £2,000 cost for all taxpayers with hard-to-value assets, reflecting a central estimate from the range Burgherr establishes; we do not attempt to account for the opportunity cost of the taxpayer's time for those taxpayers who (we assume) do their own filing. These estimates also do not attempt to account for legal costs of disputes with the tax authority, as the scale and variation of these costs are much less predictable. The valuation cost estimates cover a single year of a tax and do not depend on the tax rate, nor do they vary between a one-off tax and an annual tax; in an annual tax scenario with valuation only necessary every few years, costs could be expected to be lower after the initial valuation.
We also calculate the one-off costs that the tax authority would incur in order to administrate the tax. Our ongoing admin costs are based on the cost to HMRC of auditing self-assessment (SA) income tax returns. A wealth tax would be administered in much the same fashion, with potential taxpayers having to submit a tax return, a certain percentage of which would be audited by the tax authority. We assume that the cost of auditing a wealth tax return will be the same as the cost of auditing a SA return, which is approximately £2,500 per audit. 8 We will assume that 5 per cent of wealth tax returns are audited, suggesting that the average cost per tax unit to HMRC from auditing is around £125. To calculate the total ongoing admin cost for each tax structure, we multiply this figure by the number of filers, assuming that anyone who thinks they are within 10 per cent of the tax threshold also has to file a tax return. This means that the population of returns that could potentially be audited will be slightly higher than the number of taxpayers.
There are two types of one-off cost we consider: first, the cost of revaluing residential property, which we assume is done centrally rather than by the taxpayer, as is the case for council tax; second, the cost of designing and developing an IT system for administering the tax.
For the cost of revaluing residential property, we draw on the estimated cost of revaluating properties for council tax in England, a project that began in 2001 but was never completed. 9 In 2005 it was estimated that the revaluation would cost the Valuation Office Agency (VOA), which was tasked with conducting the revaluation, £139.3 million in total. 10 An additional £38 million was expected to be incurred in the first year from the cost of appeals. At this point, £45 million had already been spent on bringing the VOA's systems up to date and digitising documents of paper records, an exercise that would not need to be fully repeated if the revaluation were resumed. However, because these old documents may now partly be out of date, we take a conservative approach by assuming that this cost would again be incurred in full. On this basis, the exercise would have cost £180 million according to estimates from 2005. Scaling this to 2018 (the final year in our wealth tax data) by the rate of wage inflation, which is the most relevant cost here, suggests that a present-day valuation would cost approximately £245 million.
We assume that this £245 million would be the cost of revaluing the entire housing stock. It is possible that some fraction of this cost would be avoided, as not all properties would need to be valued under a wealth tax with an exemption threshold. However, we do not know what this fraction would be. Moreover, valuing properties at the top end of the property distribution is likely to be much more costly than valuing a standard semi-detached house, and so we cannot assume that the cost of revaluing the housing stock is proportional to the number of houses valued. This is certainly not the case, as there will be fixed costs in producing a model for estimating house values that would serve as the primary basis of valuation for the majority of properties. Our cost estimate should therefore be thought of as an upper bound.
As a proxy for the cost of building a new system for administering the tax, we take the cost of designing and developing the Customs Declaration Service, a system that went live in 2018. The system will ultimately process over 250 million customs declarations, calculating the tariffs due on each. The most recent estimate of the total cost of the project is £334 million. 11 This is a comprehensive estimate, which includes the cost of planning, designing, construction and delivery, as well as ongoing maintenance costs.

Revenue
In this subsection, we present estimates of combinations of rates and thresholds that would raise £10 billion in tax revenue under a low avoidance scenario, before accounting for admin costs (see Table 1). This revenue target is not chosen to be, in any way, optimal, or to reflect any kind of recommendation. Instead, it is selected as a useful benchmark, being roughly equivalent to increasing the basic rate of income tax by 2p. 12 With a threshold of £10 million, a tax charged at 1.1 per cent would raise £10 billion under a low avoidance scenario, or £8.8 billion with high avoidance. Ongoing admin costs to government here are essentially negligible, although there is some set-up cost that is needed.
Lowering the threshold to £2 million, the same amount could be collected with a tax rate of just 0.57 per cent under a low avoidance scenario. Now the number of taxpayers is substantially higher, increasing from 22,000 to 631,000. The aggregate admin costs of the tax would be much higher -the cost being largely borne by the taxpayer rather than the government. 13 Costs to the taxpayer amount to a quarter of revenue raised, effectively adding a cost of 0.13 per cent of taxable wealth to the headline rate. 14 At 22,000 taxpayers, the volume of taxpayers under an annual wealth tax beginning at £10 million is similar to IHT (which covers 24,000 taxpayers per year). Comparable amounts could be raised with 11  a progressive tax, covering the same number of taxpayers but at lower rates for those with less wealth, as shown in one particular example in the final row of Table 1. With a threshold of £500,000, 8 per cent of the revenue raised comes from individuals in the STRL. The importance of this group becomes more pronounced the higher the threshold: with a threshold of £5 million, these individuals account for 40 per cent of our revenue estimate. This reflects the concentration of wealth at the very top.
The amount of revenue raised by a tax with a given threshold can be varied by changing the rates. Figure 1 illustrates the rates that would be required to raise different revenue targets, net of ongoing admin costs to government. Evidently, the rates required to generate a given amount of revenue at a given threshold are higher when individuals are more responsive to the tax.
To understand the potential revenue effects from asset splitting, we model the revenue that would be raised if all assets, other than pensions, were split equally between the household head and his/her partner. Using the headline rates presented in Table 1, we first estimate the revenue that could be raised if assets were retained by the household member who owns them currently (see also Table G1 in the online Appendix). In the absence of any avoidance responses, revenue would be up to £0.9 billion higher than in our low avoidance scenario. However, if households were to respond to the tax by splitting their assets equally, revenue would be reduced by 1-7 per cent (up to £0.7 billion). Note that this revenue loss is smaller than the revenue loss obtained by moving to our low avoidance scenario. Our low avoidance measure is intended to be comprehensive of the range of behavioural responses available to individuals. It is reassuring to find that the revenue loss from asset splitting -even when modelled to the extreme -is lower than our more optimistic measure of avoidance responses, particularly as asset shifting is a known response to individual taxation. 15 F I G U R E 1 Rates and thresholds generating different revenue targets from an annual wealth tax, after admin costs 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 Threshold (£m) £10bn, high £10bn, low £5bn, high £5bn, low Note: Tax rates are those required to generate the revenue target after admin costs are taken into account. Source: WAS, 2016-18; STRL, 2020.

Distributional effects
If the UK were to introduce an annual wealth tax, who would pay it? And how would the amount of tax paid vary across individuals? In this subsection, we explore how tax liabilities would vary across the distribution of income and wealth under each of the annual tax structures presented in Table 1, assuming the rates required to generate £10 billion before admin costs under a low avoidance scenario. We then consider the characteristics of taxpayers, specifically considering age, sex and region. We include individuals in the STRL when looking at the distribution by wealth, age and sex. However, as we have no information on their income or on region of residence, this analysis is based on the WAS data only. Table 2 shows the amount of tax paid by a representative individual with different levels of wealth. A higher threshold does not necessarily mean that an individual who is still liable to pay the tax will face a smaller tax liability. Taking an individual with £7.5 million in wealth as an example, the tax liability that this individual faces is £20,150 under a flat tax starting at £1 million. If the threshold rises to £2 million, the rate required to generate the same amount of revenue as before means that the same individual would now face a tax liability of £31,350.
An exemption threshold of £250,000 would not charge any wealth tax to anyone in the bottom 70 per cent of the wealth distribution. Nevertheless, by international standards this would be a very low threshold: only Switzerland is lower. 16 With a relatively low exemption threshold of £250,000, the average tax rate faced across the wealth distribution would increase steadily, reaching 0.12 per cent (equal to the marginal tax rate) for those in the top 1 per cent. With a higher exemption threshold, the average tax rate increases more rapidly. Individuals in the top 1 per cent would face an average tax rate of 0.21 per cent with an exemption threshold of £2 million, for a tax generating £10 billion in revenue (see Figure 2). TA B L E 2 Amount of tax paid by a representative individual under an annual tax with different thresholds, (£) Percentile of net income £250,000 £500,000 £1,000,000 £2,000,000 £5,000,000 Note: All adult individuals are ranked according to their net income, and grouped into percentiles. The chart shows the percentage of adults in each percentile group who would pay the tax for different exemption thresholds. The distribution is independent of the rate chosen, for a given threshold. Individuals named in the STRL are excluded from this analysis, as we have no information on their income. We do not show the distribution of taxpayers for thresholds above £5 million due to small sample sizes. Source: WAS, 2016-18.
On the whole, individuals higher up the income distribution are more likely to pay a wealth tax ( Figure 3). However, at each level of income there is some variation in wealth, and not all high-income individuals have sufficient wealth to become taxpayers. Among those in the top 1 per cent of the income distribution, 91 per cent would pay a wealth tax with an exemption threshold of £250,000, compared with 25 per cent of the population. As the threshold rises to £2 million, 33 per cent would be liable to pay, and at a threshold of £5 million, this figure falls to just 8 per cent. Meanwhile, among those at the median of the income distribution, 10 per cent would be liable to pay a wealth tax with an exemption threshold of £500,000.
Older age groups are significantly over-represented among taxpayers for every threshold ( Figure 4). Despite accounting for just 39 per cent of the adult population, adults over the age of 55 represent 60 per cent of taxpayers when a £5 million threshold applies, rising to as much as 75 per cent with an exemption threshold of £2 million. This figure illustrates clearly that the majority of taxpayers would actually be of working age, with those in the 55-64 age category being the most heavily represented. Only 1-2 per cent of taxpayers would be under the age of 35.
The higher the threshold, the higher the percentage of taxpayers who are male ( Figure 5). For each threshold, female taxpayers are in the minority. The gender imbalance is most pronounced for a wealth tax starting at £5 million, under which 68 per cent of taxpayers are male. Note that this is assuming individuals do not adjust their wealth holdings in response to the tax. For a tax that defines the tax unit as the individual, we might expect some asset shifting within couples as a means of reducing their joint tax liability. This would make the gender imbalance less extreme in practice.
The geographical distribution of prospective taxpayers is skewed toward London and the South East, regardless of which threshold is chosen. Figure 6 illustrates the distribution of taxpayers for a tax starting at £500,000, under which London and the South East combined would account for 36 per F I G U R E 4 Age distribution of taxpayers under different exemption thresholds Note: The age distribution of taxpayers above different exemption thresholds is independent of the tax rate. Individuals from the STRL are included in this analysis. Source: WAS, 2016-18; STRL, 2020.

Liquidity issues
Specific solutions may be required for individuals who face high tax liabilities relative to their income, especially if much of their wealth is illiquid. In this subsection, we illustrate the extent of liquidity problems faced by individuals under the annual tax structures presented in Section 3.2. We ask how many individuals are liquidity-constrained under each of the tax structures raising £10 billion in revenue, and which groups of individuals are most affected.
In principle, we can define an individual as being liquidity-constrained if, as a result of their wealth tax liability, they would be forced either to reduce their consumption and standard of living, or to maintain their consumption by converting some of their illiquid wealth into cash. However, in practice, we do not have a comprehensive dataset that captures income, wealth and expenditure at an individual level. Rather than combining multiple datasets to approximate consumption levels at different points of the wealth distribution, we adopt a simpler approach. For each tax schedule, we classify an individual as being liquidity-constrained if their immediate tax liability exceeds 20 per cent of their net income and 10 per cent of their net income plus liquid assets. In online Appendix C, we illustrate the extent of liquidity issues using alternative cut-offs, to show how our estimates change when this definition becomes more or less generous.
We recognise that a specific solution is needed for the payment of taxes on pension wealth, as individuals below state pension age (SPA) generally do not have access to these funds. As recommended in Advani et al. (2020b), a solution to this would be to allow individuals below SPA to pay any tax due on their pension wealth out of their lump sum once they reach SPA. Accordingly, we assume that once an individual reaches SPA, all of their wealth is 'immediately taxable'. For individuals below the SPA, we define immediately taxable wealth as all non-pension wealth, plus the value of pensions that are already in payment, as this wealth has already been accessed. 18 We define 'liquid wealth' as financial wealth, plus certain forms of pension wealth depending on whether the individual is above or below SPA. If the individual is below SPA, we assume that all of their pension wealth is illiquid. 19 If the individual is above SPA, we assume that any remaining wealth in a defined contribution pension pot becomes liquid, plus any lump sums from defined benefit pensions that have not yet been claimed. However, wealth arising from the discounted stream of income from a defined benefit or annuitised pension pot, or any other form of regular pension income, is assumed to be illiquid. 20 In practice, it is difficult to distinguish between liquid and illiquid forms of wealth. We expect that some of our assumptions will classify too much pension wealth as illiquid, but that our classification of all financial wealth as liquid will have the opposite effect. It is not clear whether the net effect is positive or negative.
Of the annual tax structures raising £10 billion in revenue, a flat tax starting at £1 million generates the largest number of liquidity-constrained taxpayers, with over 48,000 taxpayers facing liquidity issues (Figure 7). Generally speaking, the lower the threshold is, the lower the share of taxpayers who are liquidity-constrained. A flat tax starting at £250,000 generates 32,000 liquidity-constrained taxpayers, representing just 0.2 per cent of taxpayers ( Figure 8). By contrast, though the number of liquidity-constrained taxpayers is much lower for a tax starting at £5 million, at 17,000, this accounts for 20.3 per cent of all taxpayers at this threshold. Note that for, each tax structure, we are adjusting the tax rates to target £10 billion in revenue. Therefore, the higher the threshold is, the higher the marginal F I G U R E 7 Numbers (in thousands) of taxpayers who would be liquidity-constrained under taxes raising £10 billion in revenue, by range of net wealth Note: An individual is liquidity-constrained if their immediate tax liability (defined in Section 3.4) exceeds more than 20 per cent of their net income and 10 per cent of their net income plus liquid wealth. Tax rates used are as per Table 1. Individuals in the STRL are not included in this analysis. For individuals at the top of the WAS, we use their Pareto-adjusted business wealth values, but adjust their net income to maintain the same ratio of wealth to income as reported in the WAS. We do not present liquidity analysis using thresholds above £5 million due to small sample sizes. The numbers underlying this graph are provided in online Appendix C. Source: WAS, 2016-18.

F I G U R E 8
Percentage of taxpayers who would be liquidity-constrained under taxes raising £10 billion in revenue, by range of net wealth Note: An individual is liquidity-constrained if their immediate tax liability (defined in Section 3.4) exceeds more than 20 per cent of their net income and 10 per cent of their net income plus liquid wealth. Tax rates used are as per Table 1. Individuals in the STRL are not included in this analysis. For individuals at the top of the WAS, we use their Pareto-adjusted business wealth values, but adjust their net income to maintain the same ratio of wealth to income as reported in the WAS. We do not present liquidity analysis using thresholds above £5 million due to small sample sizes. The numbers underlying this graph are provided in online Appendix C. Source: WAS, 2016-18. F I G U R E 9 Main asset among those who are liquidity-constrained under different annual tax structures generating £10 billion in revenue Note: An individual's main asset is the largest asset in their wealth portfolio after the exemption of low-value items (see Section 2.1 for details). Individuals in the STRL are not included in this analysis. For individuals at the top of the WAS, we use their Pareto-adjusted business wealth values. We do not present liquidity analysis using thresholds above £5 million due to small sample sizes. Source: WAS, 2016-18. tax rate faced by individuals at the top. If we did not adjust the rates, then raising the threshold would reduce the number of liquidity-constrained taxpayers, but this would also reduce revenue.
Under an annual wealth tax generating £10 billion before admin costs, the majority of liquidityconstrained taxpayers have a business as their main asset (Figure 9). The lower the threshold is, the more evenly spread the composition of assets among those who are liquidity-constrained. At a threshold of £250,000, 14 per cent of liquidity-constrained taxpayers have their main residence as their main asset. As the threshold rises to £1 million, this percentage falls to 4 per cent. At higher thresholds, business assets become much more important among those who are liquidity-constrained. At a threshold of £500,000, 57 per cent have a business asset as their main asset. With a threshold of £5 million, 94 per cent have a business as their main asset. Daly et al. (2021) discuss in detail the challenges of establishing the exact value of a person's total wealth at a given point in time -a difficult exercise, which is nonetheless necessary for all taxpayers captured in the flat (or progressive) tax regimes described above. One way to address this problem is to use a regime of tax bands, within each of which the tax charge is a fixed fee; this will obviate the need for exact valuations of wealth for many taxpayers. 21 In this subsection, we discuss how revenue raised changes if using a banded regime rather than one of the flat tax regimes as discussed above. Hughson (2020) addresses many of the issues and challenges in using such a regime as an alternative to a flat or progressive tax as described above. A key insight from this work is that a F I G U R E 1 0 Effective tax rates in a banded regime Note: Tax liability calculated with reference to the mid-point of the relevant band (£48m for top band); EATR is calculated dividing tax liability by total marketable wealth Source: Authors' calculations.

Banding
banding scheme is a blunt instrument that generates inequity: in a band covering wealth of £1-£2 million, someone with £1 million in wealth pays the same amount in tax as someone with almost twice as much wealth, and (perhaps substantially) more than someone with just under £1 million. There is a tension between limiting the extent of this inequity by setting bands narrow enough to effectively target wealth, and setting them wide enough to materially simplify the reporting burden of a significant proportion of taxpayers.
We demonstrate an example banding scheme with bands of increasing widths of total wealth: £500,000-£1 million, £1-2 million, £2-4 million, £4-8 million, £8-16 million, £16-32 million, and £32 million and over. We set the charge within bands based on the mid-point of the band (multiplied by a rate of 0.17 per cent, for comparability with a flat tax starting at £500,000). 22 The charge for the (open-ended) top of the band is set with reference to 150 per cent of the threshold. Figure 10 demonstrates what such a scheme would imply in terms of the effective average tax rate (EATR) paid -that is, the relevant banding charge divided by an individual's total wealth. The amount of tax paid under the banded regime is equal to the flat tax (only) at the mid-point of each band, and the band thresholds are clearly traced out at the points the EATRs jump higher. The vertical inequality created is clear: those at the bottom of each band pay a larger share of their wealth in tax than people at the top. The long tail at the right-hand side of the graph demonstrates a difficulty plaguing any banding regime: because the wealth distribution has such a long, thin tail, it is difficult to design a set of thresholds in which the very wealthiest members of society pay anything other than a tiny proportion of their total wealth in tax (especially compared with others at the bottom of the same band, who may pay extremely high rates).
The interaction between a banded regime and avoidance is also an important consideration. Relative to a flat or even a progressive tax regime, a banded regime creates considerably stronger incentives for avoidance for those at the bottom of each band, who have only to reduce their reported wealth by enough to fall into the lower band in order to significantly decrease their tax liability. As noted in 22 Hughson (2020) discusses in some detail the issues involved in the choice of the tax charge within each band. A charge based on median wealth in the band would imply lower tax rates throughout and revenues closer to the equivalent flat tax regime, but it is more difficult to justify for wider and wider bands, as well as being harder to implement in practice. F I G U R E 1 1 Effective tax rates and avoidance under a banding scheme Note: Tax liability calculated with reference to the mid-point of the relevant band (£48m for the top band); EATR is calculated dividing tax liability by total marketable wealth. Avoidance calculated as elasticity of 17 applied to the no-avoidance EATR. Source: Authors' calculations.
Section 3.1, by modelling an average avoidance response, we miss the heterogeneity in responses that is likely here: the incentives for avoidance are sharpened around band thresholds and dulled elsewhere. Figure 11 demonstrates the interaction between avoidance behaviour and a banding regime; the dark line repeats the line from Figure 10, showing what the EATR paid should be, with no avoidance. In our modelling, avoidance takes a very particular shape: everyone reduces their reported wealth. However, with bands, this only changes the tax liability of those who are close enough to the bottom of each band to fall into the lower tax band, while most others remain in the same band, so their tax liability is not affected. The result is that those who should be at the bottom of each band pay a lower EATR than those with slightly less wealth just the other side of the threshold. Because avoidance behaviour is being applied equally to all taxpayers and ignores heterogeneous responses, which in reality would probably see avoidance more focused around the band thresholds, the approach we have taken is likely to underestimate the overall impact of avoidance.
The impact of banding on revenue collected is notable. We compare the £9.8-10 billion raised by a 0.17 per cent flat tax above £500,000 with revenue that would be raised by the banded regime explored above (which has relatively low, dense bands), and with another possible banded regime, with fewer bands but extending much further up the wealth distribution: £500,000-£2 million, £2-5 million, £5-20 million, £20-50 million, £50-200 million, and £200 million and over. The former regime could be conceived of as an attempt to more evenly split numbers of individuals, while the latter will do a better job of ensuring that tax liabilities track wealth more closely at the very top. Table 3 provides a summary of estimated revenues using the two schemes, under the two avoidance scenarios outlined in Section 3.1. Unsurprisingly, more revenue is collected under a low avoidance scenario than under a high avoidance scenario. However, consistently across either banding regime and either avoidance scenario, the revenue collected exceeds that under a flat tax. This results from a combination of the pattern demonstrated in Figure 10 -where, as a result of the banding scheme, people at the bottom of each band are paying more than they would under an ad valorem scheme while those at the top pay less -and the positively skewed distribution of wealth, which means that there are many more people at the bottom of each band than there are at the top. The difference between the two banding schemes under either avoidance scenario highlights how sensitive revenues may be to the exact design of such a scheme.

Approach
As with an annual wealth tax, we use data from the WAS and the STRL on the amount of wealth above various thresholds to calculate the revenue that could be raised from a one-off wealth tax. Our approach is broadly similar to the one described in Section 3.1. The key difference between an annual and a one-off wealth tax is the behavioural response. We assume that a one-off wealth tax would be based on a predetermined date, providing no scope for real responses that reduce an individual's tax liability.
The kind of avoidance responses we allow for in our analysis of an annual wealth tax do not apply in this setting. However, we may still worry about non-compliance and genuine errors. Though individuals cannot reduce their liability by, for example, giving away some of their wealth, they could choose not to report it on their tax return, or unintentionally omit it. Troup, Barnett and Bullock (2020) estimate that 10 per cent of the revenue from a wealth tax would be lost due to some combination of non-compliance and errors in this scenario. The approach we adopt for a one-off wealth tax is to calculate the revenue raised based on wealth as it is reported in the WAS/STRL. We then reduce the resulting revenue estimate by 10 per cent to reflect the likely tax gap.
A one-off wealth tax is assumed to generate the same admin costs as an annual tax. The main difference is that, whereas for an annual tax it is the ongoing costs that inform the net revenue that would be raised going forwards, for a one-off tax there is no distinction between one-off and ongoing admin costs, as all costs are incurred just once.
There is no reason to assume that a one-off wealth tax would have to be paid in a single instalment -indeed this would be unrealistic. In the subsequent analysis, we often report the annualised rate of tax that individuals effectively pay, if payment was allowed over a five-year period. For example, an individual facing a flat tax at a rate of 5 per cent would pay 1 per cent per year over five years.

Revenue
In this subsection, we present the amount of revenue that would be raised by a flat tax charging 5 per cent on wealth above various thresholds. We also show the combinations of rates and thresholds that would be required to generate £250 billion from a one-off wealth tax, before admin costs. This is effectively equivalent to raising £10 billion per year over a 25-year period. A tax rate of 5 per cent would generate a substantial amount of revenue if charged on wealth above a relatively low threshold -£391 billion with a threshold of £250,000. The higher the threshold is, the less revenue can be raised from a 5 per cent tax. If only wealth above £10 million were charged, this would raise £44 billion.
If the government wished to raise around £250 billion with a one-off wealth tax, it could do so with a relatively high threshold of £1 million, but this would imply taxing wealth above the threshold at a rate of 8.5 per cent (annualised rate of 1.7 per cent) (Table 4). This tax structure would imply admin costs to the government amounting to 0.4 per cent of revenue raised. The cost to taxpayers exceeds this, amounting to 1.8 per cent of revenue raised, effectively adding a further 0.14 per cent to the headline rate. 23 At lower thresholds, lower rates would be possible. Note, however, that raising £50 billion a year requires much higher rates than for the earlier annual wealth tax targeting £10 billion. This will have implications for the number and composition of those facing liquidity constraints under a one-off wealth tax, as discussed in Section 4.4. TA B L E 5 Amount of tax paid by a representative individual under a one-off tax with different thresholds (£) Note: Calculations of the tax liability of individuals at different points of the wealth distribution, under each tax schedule shown in Table 4. 'Number of individuals with similar wealth' shows the number of individuals whose net wealth is within 10 per cent of the representative individual, giving a rough indication of the number of individuals who would face that tax liability. Source: WAS, 2016-18.
Even under a progressive tax, raising the threshold does not necessarily mean that taxpayers higher up the wealth distribution pay less (Table 5). Under a one-off tax generating £250 billion with a £1 million exemption threshold, a taxpayer with £7.5 million in net wealth pays £116,000. By lowering the threshold to £500,000 while maintaining the same revenue target, the same individual will pay £84,000. Revenue is maintained by increasing the amount of revenue from taxpayers lower down the wealth distribution. Figure 12 shows the different combinations of rates and thresholds that would be required to generate different revenue targets from a one-off wealth tax, after admin costs. Naturally, the higher the revenue target is, the higher the rate needed for a given exemption threshold.

Distributional effects
Given that taxpayer status depends only on having wealth above the tax threshold, and not on the frequency of the tax or rates charged, much of the analysis presented in Section 3.3 also applies in the context of a one-off wealth tax. What will differ are the amounts of tax paid by different taxpayers.
F I G U R E 1 2 Rates and thresholds generating different revenue targets from a one-off wealth tax, after admin costs Note: Tax rates are those required to generate the revenue target after admin costs are taken into account. We assume that 10 per cent of tax revenue is lost to non-compliance. Source: WAS, 2016-18; STRL, 2020.
In Section 3.3, we have shown how the share of wealth taxpayers varies across the income distribution for a given threshold. This is the same for a one-off tax, as it does not depend on the tax rate charged. Figure 13 illustrates how the annualised average tax rate under a one-off tax varies across the wealth distribution. The annualised rates are noticeably higher than for an annual wealth tax. The average tax rate paid by someone in the top 1 per cent is 0.8 per cent in each of the five years, compared to 0.15 per cent under an annual wealth tax generating £10 billion in revenue with a threshold of £500,000. In online Appendix A, we show how the average tax rate varies by total wealth, rather than by percentile. Here, it is evident that as wealth increases, the average tax rate gradually converges to the headline marginal rate.

Liquidity issues
The tax rates required to raise £250 billion in tax revenue from a one-off wealth tax -effectively £10 billion per year over a 25-year period -are clearly higher than the rates required to generate £10 billion from an annual wealth tax. Under a flat tax starting at £500,000, taxpayers would face a tax rate of 4.8 per cent under a one-off tax, or 0.95 per cent per year over a five-year payment period. By contrast, the same individual would have to pay 0.17 per cent per year under an annual wealth tax. As a result, the number of taxpayers who are liquidity-constrained under a one-off wealth tax will far exceed the number constrained under an annual tax. Figure 14 shows the number of taxpayers who would be liquidity-constrained under a wealth tax generating £250 billion with a five-year payment period. An individual is liquidity-constrained if the amount of tax they have to pay in the first of the five years exceeds 20 per cent of their net income, and 10 per cent of their net income plus liquid wealth (see Section 3.4 for details). 24 In this setting, a flat tax starting at £250,000 generates the greatest number of liquidity-constrained taxpayers, at 528,000 (3.4 per cent). Fewer than 10 per cent of these individuals would be liquidity-constrained under an F I G U R E 1 3 Mean average (annualised) tax rate by percentile under different one-off tax structures Note: All adult individuals are ranked according to their total wealth measured at market value, and grouped into percentiles. For each tax structure, the tax rates are adjusted to target £250 billion in revenue, before admin costs, as per Table 4. The annualised average tax rate faced by individual is the amount they should pay in each year of the five-year payment period, and does not take behavioural responses into account. We take the democratic mean of average tax rates faced in each percentile. Online Appendix A shows the average tax rate by total chargeable wealth. Source: WAS, 2016-18; STRL, 2020.

F I G U R E 1 4 Numbers (in thousands) of taxpayers who would be liquidity-constrained under taxes raising £250 billion in revenue, by range of net wealth
Note: An individual is liquidity-constrained if their immediate tax liability (defined in Section 3.4) exceeds more than 20 per cent of their net income and 10 per cent of their net income plus liquid wealth. Tax rates used are as per Table 4, where we target £250 billion in revenue under each tax structure. Individuals named in the STRL are not included in this analysis. For individuals at the top of the WAS, we use their Paretoadjusted business wealth values, but adjust their net income to maintain the same ratio of wealth to income as reported in the WAS. We do not present liquidity analysis using thresholds above £5 million due to small sample sizes. The numbers underlying this graph are provided in online Appendix C. Source: WAS, 2016-18. F I G U R E 1 5 Main asset among those who are liquidity-constrained under different one-off tax structures generating £250 billion in revenue Note: An individual's main asset is the largest asset in their wealth portfolio after the exemption of low-value items (see Section 2.1 for details). Individuals in the STRL are not included in this analysis. For individuals at the top of the WAS, we use their Pareto-adjusted business wealth values. We do not present liquidity analysis using thresholds above £5 million due to small sample sizes Source: WAS, 2016-18. annual flat tax generating £10 billion. As a percentage of the number of taxpayers, a flat tax starting at £5 million produces the highest share facing liquidity constraints, at 38 per cent.
The composition of main assets among those who are liquidity-constrained under a one-off wealth tax is markedly different than for an annual wealth tax generating £10 billion in revenue (Section 3.4), as many more taxpayers are liquidity-constrained under a one-off tax generating £250 billion in revenue (see Figure 15). Business assets still feature prominently as a main asset among those who are constrained, and are the most common main asset among those constrained by a tax with an exemption threshold of at least £2 million. However, pension assets are a much more common main asset among those who are liquidity-constrained than under an annual tax. With a flat tax starting at £500,000, 53 per cent of those who are liquidity-constrained have illiquid pension wealth as their main asset. Most pension wealth can be considered illiquid. Individuals below SPA do not have access to pension pots that they are too young to claim. Defined benefit pensions, and annuitised pensions ('pension in payment') that provide a regular income stream are illiquid (regardless of age) as individuals cannot access future payments in advance. However, for individuals above SPA who have wealth remaining in DC pension pots, Additional Voluntary Contributions, and lump sums from DB pensions that have not been yet claimed, we classify this wealth as liquid -as these funds have not yet been annuitised, they could be accessed if the individual so wished.

Banding
As discussed in Section 3.5, a tax regime that utilises tax bands can alleviate the valuation challenges for taxpayers, but this can have a large impact on revenue; even with relatively tight bands, the revenue collected tends to be higher than that under a flat tax equivalent, because the bulk of taxpayers pay more.
Banding also has a marked impact on revenue in the context of a one-off wealth tax. Even if we assume the same response from taxpayers (i.e. reducing wealth above the threshold by 10 per cent to reflect non-compliance) as in a one-off flat or progressive tax, the estimated revenue raised is over £460 billion rather than £250 billion under a flat tax. 25 It is worthwhile explicitly pointing out that, in this context, more revenue is not necessarily better; here, the additional revenue comes from the distortion caused by the banded system, leading to many individuals paying far higher tax rates than they would under an ad valorem tax. The costs of valuation would need to be extremely high in this scenario to justify such a substantial distortion.
This could be attenuated to some extent by basing the band charge on a point lower in the band. For example, basing the charge on the value a quarter of the way into the band (as opposed to the halfway point, the mid-point) reduces the estimated revenue collected to around £390 billion.
That said, our modelling of avoidance behaviour in this scenario is quite mechanical, assuming that all taxpayers respond in the same way, by reducing their reported wealth by 10 per cent. Some taxpayers fall into a lower band as a result, but for most, their payment will remain the same. Clearly, this is unrealistic, as those whose wealth is close to a band lower boundary have stronger incentives to respond than those higher up in the band. On net, though, this will only partly work against the effect of the high EATRs being paid by individuals at the bottom of each band, and will not remove the large variation in EATRs across individuals.
While it may be desirable to introduce a banding scheme as a response to the difficulty of valuing wealth precisely, the invariable result is that some individuals will pay much higher taxes than they would under an equivalent flat tax. It is only possible to avoid large distortions in the tax burden by having very tight bands. Hughson (2020) shows that the trade-off for this choice is the better valuation accuracy needed to avoid large numbers of individuals accidentally mis-classifying their wealth. In the case of a one-off wealth tax as described above, this might not be unreasonable given the recommended design implies performing high-quality valuation across most of the tax base.

ALTERNATIVE REFORMS TO CAPITAL TAXES
To get a sense of scale for wealth tax revenues, it is helpful to consider the effects of alternative reforms that have been proposed. In this section, we provide some evidence on the amount of revenue that could be raised from alternative reforms to capital taxes, namely capital gains tax (CGT), taxation on dividends, inheritance tax (IHT) and council tax.

Capital gains tax
The current CGT regime raised £8.8 billion in 2019-20, 26 but recent estimates suggest the amount raised could be almost tripled by equalising the tax rates charged on capital gains with that charged on income. Capital gains are afforded a significant tax discount compared with earned income, which incentivises those who can to take remuneration as gains instead of income. This means that the taxation system does not redistribute remuneration from gains nearly as effectively as it does remuneration from income. Advani and Summers (2020) show that, by taking advantage of the preferential tax rates afforded on capital gains, many of the highest-income earners pay lower rates of tax than those on below-average incomes. If everyone with total income and gains over £100,000 paid the headline average tax rates on earnings, this would raise a further £12 billion. The Office of Tax Simplification (OTS), in their review of CGT, suggested that the same reform could raise around £14 billion. 27 A second substantial reform would be the removal of 'death uplift'. Currently, any accrued capital gains are written off on death. The estate of the individual who dies does not have to pay CGT. The inheritor is treated as having a 'base cost' (original value of asset against which gain is calculated) equal to the value of the asset when they receive it, so when they sell they only pay tax on the gains that occur after receipt. This creates strong incentives to delay realising gains, so that the gains can be written off.
The OTS review of CGT proposes moving to a 'no-gain no-loss' basis. Here the estate of the individual continues not to have to pay the CGT, 28 but the inheritor now receives the asset at the base cost of the previous owner. No CGT is now written off; instead, the CGT owed by the deceased is paid by the inheritor when the inheritor sells (or otherwise 'disposes of') the asset. The OTS estimates that this reform will raise only £470-900 million in additional tax revenue in the short term.
The relatively modest sum raised here -at 5-10 per cent the current CGT revenue -is because many of the assets passed on will continue not to be sold for some time. The proposed reform also retains some lock-in effect for two reasons. First, for inheritors, CGT would have to be paid on the assets when sold, while the full value is otherwise available to borrow against; this is already a problem for the original owner, but the magnitude increases over time. Second, in the absence of new reliefs, it discourages inter vivos gift-giving (which would still require payment of CGT).
A better alternative would be to raise the money when assets are transferred, by treating this as a disposal. No-gain no-loss benefits those who have enough wealth that some assets can continue to be passed through multiple generations without ever being sold. In principle, the same revenue will eventually be owed. However, there is a risk that the tax owed on some of these gains will be wiped out by calls for 'rebasing' -essentially forgiveness of tax owed before some date. Such a proposal made its way in to the OTS review of CGT, on the basis that base costs for some assets cannot be found. 29 Moving to a disposal treatment at death is likely to raise £1.6 billion a year, almost 20 per cent of the current CGT revenue, and would increase proportionally with the tax rate, on a static calculation inter alia, OTS (2020) -showing the long-term equilibrium effects of removing uplift at death.
Removing uplift at death in favour of either of these alternatives would create some administrative costs. Although the estate must be valued on death anyway, base costs would now also be needed for assets. Assuming all assets are hard to value, and using the estimate from Burgherr (2021) that a central estimate of valuation costs is 0.5 per cent of the total value of the asset, this would cost an additional £80 million per year.

Income tax rate on dividends
As noted in Section 5.1, income from different sources is taxed differently, and individuals who can arrange their income in different forms can reap substantial tax benefits. For example, many individuals at the top of the income distribution receive substantial dividend income (and can also claim additional personal allowances on this income), which reduces their tax bill compared with receiving the same amount as earned income.
We consider the effect of raising the dividend rate such that it is equal to the income tax rate paid on earnings by individuals. As a very rough exercise, we use data from the Survey of Personal Incomes 27 OTS, 2020. 28 There are some complexities in how this is structured, including the implications for IHT, which we do not describe here. For full details see OTS (2020). 29 Given the work done by the Wealth Tax Commission to study valuation issues, we think that difficulty getting past asset values is unlikely to be a serious concern. Nevertheless, it appears to be a politically salient issue, which has led to a proposal that base values be rebased to a value in the year 2000 (OTS, 2020). We do not see any merit in such a recommendation, but were it implemented it would clearly affect the revenue that would be raised in the short term.
(SPI) in 2016-17, and estimate the additional tax revenue that would be raised if dividends were taxed at the same average tax rate as recipients pay on the rest of their income and the additional personal allowance removed. 30 The estimated effect of this reform would be to raise just over £5 billion. 31 Again, we note that these are static estimates, and we caution that a more sophisticated analysis taking account of behavioural responses may suggest a lower amount of revenue raised. However, these calculations do not include employee National Insurance contributions, which are currently due on employment income (but not on dividends); the inclusion of these would result in a larger amount of revenue.
We use statutory incidence in constructing these estimates, so taxes paid by firms, including corporation tax and employer National Insurance contributions, are not accounted for. Advani and Summers (2020) show that the opposite extreme -assuming full incidence of these taxes on employees -increases the overall level of tax paid, but does not change the incidence across the income distribution.
With no information on wealth available in the SPI, it is difficult to comment on the distributional impact of such a change with reference to wealth. However, in terms of income, the effect of the change would be most notable for additional rate taxpayers with dividend income, who would pay an average of £3,900 extra per year in tax, followed by higher rate taxpayers (£1,490 per year) and then basic rate (£475 per year).

Inheritance tax
Inheritance tax is a tax on wealth passing on death. The current tax base is far from comprehensive. Pension wealth is exempt from IHT, and reliefs are given on business assets (BPR) and agricultural property (APR). The exclusion of these assets from the tax base makes it easy for individuals who are flexible in how they hold their wealth to avoid the tax. Regarding pensions, the current system creates an unusual incentive for individuals to draw down all other wealth in order to pass on their pension pot tax-free to the next generation. This is particularly anomalous as pension savings already receive significant tax advantages when saving is being done. The tax is, in principle, levied on wealth in excess of £325,000, though exemptions also apply, notably for estates passing to a surviving spouse. 32 The exemption threshold ('nil rate band') each estate faces in practice depends on what assets it contains and who they are passing on to, as certain assets such as the main residence qualify for a higher exemption threshold if they are passed on to the direct descendants of the deceased. It also depends on whether or not the estate had been inherited from a spouse who did not use up their own nil rate band, as any excess can be transferred to the surviving spouse.
In sum, the multitude of exemptions and reliefs make the current IHT system complex, and mean that it is easy for wealthy and well-advised individuals to avoid the tax. This is evidenced by the fact that the effective average tax rate declines from 20 per cent among estates worth more than £8-9 million to 10 per cent for estates worth more than £10 million. 33 Numerous reforms to the existing tax system have been proposed. In this paper, we focus on two: the removal of APR and BPR, and the inclusion of pensions in the tax base. 34 We calculate the amount of revenue that could be raised from these reforms using data from the WAS for 2016-18, comparing 30 The dividend allowance was £5,000 in 2016-17 but has since been reduced to £2,000. 31 Comparable figures are found in Nanda (2019), who estimates that equalising tax on dividends with tax on earnings would raise £6.8 billion in 2020-21 (static estimate), or £6 billion after allowing for behavioural responses. this to official estimates of the charges levied on estates passing on death in 2017-18, which totalled £4.8 billion. 35

Approach
The WAS represents a snapshot of the wealth held by the living population, rather than the population of estates passing on death. To model the value of estates passing on death, we apply age-sex-specific mortality rates to individual survey weights in the WAS. That is, within each age-sex cell, we scale the weights of individuals in that cell to match the number of deaths among that group recorded in the Office for National Statistics (ONS) official UK death statistics for 2018. The aim of this exercise is to produce a sample that is representative of those who would die in the year following the survey. The resulting sample is not fully representative, as we do not adjust mortality rates to reflect the fact that wealthier individuals are likely to live longer than other individuals of the same age and sex.
Modelling the removal of APR and BPR can only be approximated using these survey data, and our estimates are only intended to be illustrative. 36 The questionnaire does not allow for the construction of wealth totals that map neatly into the eligibility criteria for these reliefs. First, agricultural business assets are combined with other businesses in the WAS, making it difficult to distinguish between the two reliefs. 37 Second, listed and unlisted shares are combined into a single category, though they are treated differently under BPR. 38 We approximate this by classifying all shares owned by individuals who only own unlisted shares as unlisted, and all other shares as listed. We assume that the former are eligible for BPR, while the latter are not. This could lead us to overstate the revenue gains from abolishing BPR, as some shares classed as 'listed' would have already qualified for at least partial relief. However, it is possible that some business assets categorised under the 'own businesses' section of the WAS, which we assume are eligible for 100 per cent relief, would not qualify currently.
Our revenue analysis excludes individuals from the STRL. This is because, though we believe their recorded wealth primarily reflects business assets, we are not confident in assigning the full value of this wealth as qualifying for 100 per cent relief under BPR. Including this wealth in the wealth added through the removal of BPR and APR would overstate the revenue implications of this reform. We do, however, take into account the adjustment made to business wealth owned by WAS individuals through our top wealth adjustment. 39 In modelling the revenue implications of taxing inherited pension wealth, we account for the fact that the value of pensions that are inherited differs from the total value of pension wealth held by an individual while they are alive. While the value of a defined contribution pension pot generally remains intact when it is inherited, defined benefit pensions and the value of pensions that have been annuitised can be worth less to the individual on to whom the pension is passed, as they do not usually receive the full amount of income paid to the original recipient. We assume that the value of a defined benefit pension when passed on is worth 50 per cent of what it was worth to the original recipient. 35 Tax due on estates passing on death in 2017-18, Inheritance Tax Statistics Table 12.2 (HMRC, 2020c). 36 As we explain, the data we use do not directly map on to the categories available for relief. We also do not have the necessary information for the very top of the wealth distribution (which we elsewhere impute using the STRL), nor do we have information on the remaining nil rate band that a given individual has. Therefore, our revenue estimates here have more uncertainty than when modelling a wealth tax. A more in-depth analysis of IHT reform is beyond the scope of this work.
Pensions that have already been inherited from a former spouse or partner are assumed to cease when the individual who inherited them dies. We cannot accurately model the exemption threshold that each individual in our dataset would face, not least because we cannot identify individuals who have inherited some nil rate band from a spouse, nor do we know who would inherit the estate. This makes it impossible to say whether exemptions such as the residence nil rate band, which depends on the relationship between donor and donee, should apply. Regarding the spousal exemption, we can assume that all individuals who are married when they die would pass their estate to the surviving spouse, but we cannot confirm this. As a result of these limitations, the revenue model we implement will necessarily be stylised.
Our approach makes use of the fact that the current IHT system taxes 4 per cent of all estates passing on death. We take this as our target tax-paying population total, N. We then exclude individuals who are married when they die, under the assumption that these estates qualify for spousal exemption. Finally, we assume that the wealthiest N of the remaining estates are taxed, at a rate of 40 per cent (as per the current system), on wealth in excess of the minimum wealth required to be among the taxpaying population. This final step amounts to assuming that all individuals face the same 'effective threshold', when in practice some individuals would face a higher threshold and some a lower one.
As well as calculating the revenue that could be raised from these reforms, it is important to consider their administrative cost. We will assume that the average admin cost per taxpaying estate, currently £1,450, 40 is the same when pensions are included as when they are not. While implementing the reform is likely to come at some admin cost, it is possible that this will be partially offset by the average complexity of estates brought into IHT being lower than the complexity of estates that already face the tax. Overall, we assume that these effects offset one another.
For the removal of APR and BPR, we assume that there is an admin cost to valuing businesses. Following our method for calculating taxpayer costs for a wealth tax, we assume that this is a fixed percentage of the business value reported in the WAS. We take 0.8 per cent of business wealth -the upper bound applied for valuing hard-to-value assets under a wealth tax -for all taxpayers with business wealth in excess of £30,000, but assume that the total cost of valuing a business cannot exceed £25,000, which Burgherr (2021) suggests is the maximum cost faced in practice. This is then added to the current average cost of administering IHT per taxpayer. This is likely to be an upper bound, as the cost of valuing businesses will be partially offset by no longer having to decide whether or not a business is eligible for reliefs. In this exercise, we are assuming that the additional cost of valuing businesses is borne by HMRC. This may not be the case. If individuals are required to obtain valuations, then this cost would be borne by the taxpayer instead, as a cost to the executor that would be able to be taken out of the estate.
Our main specification does not take avoidance responses into account. It is not clear how individuals would respond to these reforms, given that they shut off some of the channels that can currently be exploited to avoid IHT, such as passing wealth on through a pension pot or business. In online Appendix D, we present a specification which assumes avoidance responses that are likely to be larger than we would observe in reality, for comparison. The effect of these responses on our revenue estimates is small, and does not change our interpretation of the results.

Revenue
Our stylised approach to modelling the IHT system predicts that with the tax base that exists currently (i.e. excluding pensions and business assets) IHT ought to have raised £4.4 billion in revenue from estates passed on in the year following the WAS survey. This is slightly lower than the £4. Note: 'Current IHT tax base' shows our stylised model of the current IHT system. The 'effective threshold' is the amount of wealth required to be among the taxpaying population. The revenue calculation is the rate applied to total wealth above the effective threshold. 'Adding pension wealth' adds pensions to the current tax base, according to their inherited value, keeping the effective threshold fixed. 'Adding business wealth' adds businesses and unlisted shares to the current tax base, keeping the effective threshold fixed. 'Adding pensions and business wealth' combines the previous two reforms. 'Raising the threshold' takes the tax base as inclusive of all assets (pensions and business wealth included) and raises the threshold to maintain the same number of taxpayers as the current IHT system. 'Reducing the rate' calculates the rate required to generate the same amount of revenue as the 'current' IHT system (our model), from the same number of taxpayers, with a comprehensive tax base. Source: Authors' calculations based on the WAS, 2016-18. explanation for this shortfall is that our data under-represent wealth held at the top of the distribution. While we adjust the data for our wealth tax analysis by including the STRL and adjusting business wealth, this is not taken into account in our IHT modelling as business wealth is excluded from the tax base and we do not include the STRL. A second factor is that we do not observe lifetime gifts made in the seven years prior to death, which are also subject to IHT. As shown in Table 6, we estimate that including pension wealth in the tax base, keeping the threshold the same, could increase revenue by 41 per cent, or £1.8 billion. We estimate an additional admin cost (as a result of additional taxpayers needing to file) of £18 million, around 0.1 per cent of the additional revenue. It is important to note that this, and all other estimates here, are static revenue estimates; they do not account for how individuals might respond. We provide some suggestive modelling of this in online Appendix D.
An alternative reform that adds businesses (agricultural and other) to the current tax base, keeping the threshold the same and not changing the treatment of pensions, would increase revenue by 20 per cent, or £900 million, at an additional cost of £5 million.
Combining the two reforms would yield 61 per cent more revenue, and would cost £23 million more than the current system, bringing an additional 12,000 estates into IHT. In proportional terms, these reforms raise substantial revenue, and at very low admin cost. They also have the benefit of not distorting choices of which assets to hold in order to reduce tax liabilities on death. However, relative to the amounts that a wealth tax could raise, the revenue gains from these reforms alone are small. 42 Implementing these reforms would involve many of the valuation challenges faced under a wealth tax, though -importantly -for much lower volumes of taxpayers in any given year. By being paid once (at most) per individual, rather than annually, the admin costs are much lower, though the principle of needing to value additional asset classes for IHT is not changed. 43 The removal of APR, BPR and pensions relief would reduce distortions. Whether a government wants to then also have more taxpayers and more revenue is a political choice, rather than something that can be judged objectively. We therefore consider two further reforms. 42 Corlett (2018) estimates the revenue that could be raised from a different alternative: completely replacing IHT with a lifetime receipts tax. Under a tax structure that features a £3,000 annual allowance per recipient, a £125,000 lifetime allowance, and a flat rate of 15 per cent on lifetime gifts in excess of this, it is estimated that £6.9 billion could be raised in the first year of the tax. Over time, the amount of tax revenue would increase as some individuals who do not reach their lifetime allowance in year one receive additional gifts and inheritances. 43 As Advani et al. (2020b) note, all these asset classes are already valued else where in the tax system. TA B L E 7 Tax due under a reformed IHT, by range of total wealth on death First, we target the same number of taxpayers as the current system after including pensions and businesses. Implicitly, this raises the 'effective threshold' needed to be a taxpayer. This reduces the revenue raised by the reform from £7.1 billion to £5.7 billion, but this is still an increase of £1.3 billion on the status quo.
Second, we estimate the rate that would be needed to raise the same amount of revenue as the 'current' scenario, after a reform that includes all assets in the tax base. The tax rate suggested by our model in this scenario is 25 per cent. Table 7 illustrates how these reforms shift IHT liabilities toward the top of the wealth distribution. Taxing all wealth on death at 40 per cent, while maintaining the same number of taxpayers by raising the threshold, reduces the average tax paid among taxpaying estates worth £650,000-£1 million by more than a third. By contrast, it increases the average tax paid by taxpaying estates by 9 per cent for those worth £2-5 million and 13 per cent for those worth more than £5 million.
If we lowered the rate to make the reform revenue neutral, but kept the current threshold, the average tax liability would be lower than the current system in all wealth bands. However, total tax revenue would remain unchanged as the tax liability would be spread across a greater number of taxpaying estates. This reform would create greater horizontal equity by taxing estates with the same total net worth equally.

Council tax
Another tax on wealth that regularly receives proposals for reform is council tax. Council tax is a tax levied on residential property. The statutory incidence of the tax is on the person occupying the property, rather than the owner, although it is important to note that part of this cost may be passed on through lower rents. The tax is administered at a local level, with some elements of the tax structure set centrally. The charge is based on banded property values, with the bands fixed across local authorities. However, local authorities have some freedom in setting the tax liability that is paid in each band. Residential properties in England have not been revalued for council tax since 1991, and the current tax bands are based on prices from this period. 44 This creates horizontal inequities in the current system, with properties that have the same value today being taxed different amounts depending on how their value has changed since 1991. A revaluation exercise is almost certainly needed, to ensure tax liabilities are based on what the property is actually worth. As discussed in Section 3.1, the cost of this exercise to government would be around £245 million. However, it is worth noting that the alternative -sticking to 1991 property values -is not without cost: currently all newly built properties have to be assessed as if they had been built in 1991, which is more likely to be open to dispute than the current value of a property.
Revaluation need not affect the revenue raised from council tax, nor the average tax liability households face. If the bands were adjusted to current values, maintaining the same proportion of properties in each band, and rates were set the same, the amount of revenue would not change (at a national level). All that would change is who pays: properties that have appreciated considerably since 1991 would attract a higher liability, while properties experiencing more modest growth would see a decline in their tax liability.
However, there are good reasons why we might want to change the rates paid across different tax bands. The current system is highly regressive. The average tax rate paid on a house at the mid-point of the lowest council tax band is 1.65 per cent. 45 By contrast, a house at the bottom of the top band attracts an average tax rate of just 0.2 per cent. The regressivity with respect to wealth of the current system is evident in Figure 16. This plots council tax paid as a share of wealth across the wealth distribution, based on data from the WAS. A quarter of households in the 25 th percentile of the wealth distribution pay more than 2.9 per cent of their wealth in council tax. At the 75 th percentile, the average tax rate is 0.23 per cent, with no household paying more than 0.45 per cent.
A more progressive council tax could be achieved by making tax liabilities proportional to property values. 46 Again, this need not affect the total amount of revenue raised, as rates could be set to increase the tax paid at the top of the property distribution and reduce the amount paid by those at the bottom. However, bringing the tax rate charged on valuable properties in line with rates currently paid on properties at the bottom could raise a significant amount of revenue. We model such a reform in the following subsection.

Approach
To model the revenue that could be raised from a more progressive council tax based on current property values, we draw on evidence from Adam et al. (2020), who model the distributional effects of six council tax reforms. The first, which we shall use in our analysis, is a revaluation reform whereby properties are revalued and placed in one of the current council tax bands (in England), with the band thresholds adjusted to maintain the same proportion of properties in each band. Their estimates suggest that a council tax based on revalued property values would have the structure presented in Table 8. Revaluation alone does nothing to address the regressivity in the system. Under this system, a property worth £250,000 is charged 0.62 per cent, while a property worth £10 million is charged just 0.0035 per cent. We can use these revalued tax bands as a basis for considering a more progressive tax. Achieving a more progressive tax does not necessarily mean abolishing the banded system. One F I G U R E 1 6 Council tax paid as a share of wealth, by wealth percentile Note: 'Mean' shows the average (mean) share of wealth paid in council tax at different percentiles of the wealth distribution. 'Median' shows the median share of wealth paid in council tax at different percentiles of the wealth distribution. 'XXth percentile' shows the XX percentile share wealth paid in council tax at different percentiles of the wealth distribution. All households are ranked according to total household wealth measured by market value and divided into percentile groups. Based on Wave 6 (July 2016 to March 2018) of the WAS. Weights are scaled to match the full population total. Source: Authors' calculations based on the WAS, 2016-18. option for making council tax more proportional to property values would be to increase the tax rate paid in higher bands, perhaps by setting this with reference to the median or mid-point in each band. The use of additional bands would also provide greater capacity for charging a higher tax liability on high value properties than lower valued properties.

TA B L E 8
In this paper, we estimate the amount of revenue that could be raised by reforming council tax into a continuous progressive tax, under which the tax liability would be set by reference to the exact value of the property, rather than by reference to bands. For the purpose of estimating the revenue that could TA B L E 9 Revenue raised from council tax reform be raised from council tax reform, this exercise is informative even if there is no desire for a system that does away with banding. This is because the amount of revenue raised from a fully continuous tax provides a lower bound on the revenue that would be raised from a banded system, which sets the charge by reference to the median or mid-point of the band. Hughson (2020) explains how and why this is also true in the context of a wealth tax. The key intuition is that, because the distribution of property values is skewed to the right, properties at the lower end of each band attract a much higher average tax rate than the property at the median or mid-point.
To estimate the revenue that could be raised from a continuous and proportional council tax based on current property values, we simply multiply aggregate gross property wealth by the desired tax rate. 47 We estimate the revenue that would be raised under two alternative reforms. The first brings the tax rate in line with the rate paid at the 24 th percentile of the currently property value distribution under the revalued band system, which is approximately the top of band A. A property at this point, worth £142,000, would pay an average tax rate of roughly 0.83 per cent. The second reform brings the tax rate in line with the average tax rate at the 45 th percentile of the distribution -the bottom of band C with a property value of around £250,000 -which is 0.76 per cent.

Revenue
Bringing the tax rate faced by all properties in line with the tax rate currently faced by a property at the 24 th percentile could raise an additional £17.6 billion, at a one-off admin cost of £245 million (Table 9). Alternatively, charging the rate currently faced by a property at the 45 th percentile of the distribution could result in a revenue gain of £13.4 billion.
More than a quarter of households would see a reduction in their council tax liability under these reforms (Figures 17 and 18). Reductions in council tax liabilities would be observed at the bottom of the property value distribution, for properties currently in bands A and B. The top 10 per cent of properties would see a significant increase in their council tax charge, with a minimum additional annual charge of £3,000. For properties in the top 1 per cent of the distribution, the minimum additional charge would be £10,000 annually. Under this system, households at the top would be paying the same tax rate as households at the bottom.
Note, however, that we have not yet accounted for the effect of the reform on property prices. We would expect a reform that lowers the tax charged on low value properties, and increases the tax charged on high value properties, to lead to a compression in house prices. High value properties will be worth less, as any buyer now faces a higher council tax charge. Meanwhile, the lower council tax levied at the bottom of the property distribution would, to some extent, be capitalised into higher property values. F I G U R E 1 7 Council tax liability under reform 1, by tax band Note: 'Current tax' is the average tax liability faced in the tax band across local authorities in England. 'Reform 1' is a council tax reform that charges all properties the rate currently charged at the 24 th percentile of the property value distribution, which is 0.83 per cent. The maximum (minimum) shows the tax that would be paid by a property at the upper (lower) threshold of the band, where applicable. Source: Authors' calculations based on the WAS, 2016-18; Adam et al. (2020).

Revenues from other taxes
In this subsection, we provide a brief summary of revenues raised by existing taxes to put our revenue estimates and alternative reforms into perspective. We also provide estimates of wealth tax receipts in other countries; however, we caution that revenue depends heavily on the specific structure of the tax (as this paper makes clear), so cross-country comparisons should be interpreted with care. The UK government raised £828 billion in taxes in the 2019-20 tax year. 48 An additional £250 billion in revenue (or £50 billion per year over five years) from a one-off wealth tax would have increased tax receipts by 30 per cent (6 per cent) in that year; an annual wealth tax raising £10 billion would have represented a 1.2 per cent increase in total tax receipts.
The biggest sources of taxation are taxes on income: income tax receipts accounted for £194 billion in 2019-20, and National Insurance contributions added a further £145 billion. Value added tax (VAT) is the third big source of tax revenue, and raised around £134 billion from taxing general expenditure. Council tax contributed £36 billion to total public receipts, while (as noted above) CGT raised £8.8 billion; IHT raised just over £5 billion in the same year. 49,50 Current wealth taxes in other countries raise between 0.5 per cent and 4 per cent of total tax revenues. 51 In 2018, the wealth tax raised €1.1 billion in Spain, around 0.5 per cent of total taxation. The Norwegian wealth tax raised €1.6 billion in 2020, approximately 1.1 per cent of total tax revenue, and a similar share is raised in France by the IFI (€2.1 billion in 2019). The Swiss wealth tax generates 3.8 per cent of total tax revenue; Perret (2021) notes that Switzerland has always been an outlier in terms of the relative importance of their wealth tax. 48 Keep, 2021 These revenue figures differ slightly from those presented in Sections 5.3 (IHT) and 5.4 (council tax), as those are for different tax years, to be comparable to the data sources used in those sections. 51 Chamberlain, 2020; Perret, 2021. F I G U R E 1 8 Council tax liability under reform 2, by tax band Note: 'Current tax' is the average tax liability faced in the tax band across local authorities in England. 'Reform 2' is a council tax reform that charges all properties the rate currently charged at the 45 th percentile of the property value distribution, which is 0.76 per cent. The maximum (minimum) shows the tax that would be paid by a property at the upper (lower) threshold of the band, where applicable. Source: Authors' calculations based on the WAS, 2016-18; Adam et al. (2020).

CONCLUSION
We find that a tax rate of 0.17 per cent on wealth above £500,000 would be needed to raise £10 billion with a flat annual wealth tax. The same revenue could be generated at higher thresholds with a higher tax rate, or by a progressive tax in which those towards the lower end of the wealth distribution pay a lower rate, and those at the top end pay a higher rate. Under this tax structure, just 0.5 per cent of taxpayers would face liquidity constraints. Alternatively, a one-off wealth tax could raise £250 billion in revenue by charging 4.8 per cent on wealth above £500,000 (effectively, 0.95 per cent per year, paid over a five-year period), with similar possible alternative rates and thresholds. Taxpayers would be more likely to be male, of working age, and residents of London and the South East. Under this tax structure, 6.4 per cent of taxpayers -526,000 individuals -would face liquidity constraints.
We also estimate administrative costs to the taxpayer and to the government. Administrative costs to taxpayers are estimated at £7.8 billion per year under a tax with a £500,000 threshold, and would decrease in aggregate (but increase on a per-taxpayer basis) at higher thresholds, as there would be fewer taxpayers but those at the top of the wealth distribution typically hold more hard-to-value assets. The cost to the government of setting up the tax would be around £580 million, with ongoing costs for a £500,000 exemption threshold estimated at around £1.2 billion, again decreasing at higher thresholds.
A brief comparison shows that it would be possible to raise similar amounts of revenue to an annual wealth tax, or more, through reforming existing taxes on capital. These reforms would also come at a cost, and would not necessarily avoid some of the challenges imposed by a wealth tax, including a need to re-value housing, and high valuation costs for other assets. However, in many cases, the volume of taxpayers would be lower, making these costs per taxpayer more tolerable.

A C K N O W L E D G E M E N T S
The authors thank Andy Summers, Emma Chamberlain, David Burgherr, Daniel Butler, John Barnett and Edward Troup for helpful comments. The authors are also grateful to staff at HMRC, HMT, ONS and the VOA for providing information that served as input to the analysis. The authors also thank Shaun Fu, Louis Goddard, Sam Leon and Aruna Popescu for assistance matching Companies House data to the Sunday Times Rich List. This does not signify any endorsement or agreement by any of the above of the conclusions in this paper.
This work contains statistical data from ONS, which are Crown Copyright. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research datasets that may not exactly reproduce National Statistics aggregates.
A previous version of this work circulated as Wealth Tax Commission Evidence Paper No. 13. The Wealth Tax Commission acknowledges funding from the Economic and Social Research Council (ESRC) through CAGE at Warwick (ES/L011719/1) and a COVID-19 Rapid Response Grant (ES/V012657/1), and from the LSE International Inequalities Institute AFSEE COVID-19 fund.