Between universalism and targeting: Exploring policy pathways for an Australian Basic Income

Abstract

Despite growing interest in proposals for a universal basic income, little advance has been made in implementation. Here we explore policy options for an Australian Basic Income. Our analysis responds to concerns that Basic Income is both too expensive and too radical a departure from existing welfare state structures to be a feasible policy option. Drawing on policy and Basic Income scholarship we identify changes to Australia’s current means-tested benefits structures that move substantially towards Basic Income while remaining consistent with historic policy norms, which we call ‘affluence testing’. Using microsimulation we explore fiscal and distributional trade-offs associated with the implementation of an affluence-tested Basic Income. Our results suggest Basic Income has the potential to significantly reduce inequality and poverty while also requiring taxes to rise substantially. Placing these trade-offs in international context we find the policy would reduce inequality to levels similar to Nordic welfare states while increasing overall taxation to approximately the OECD average.

Introduction

From Thomas Paine (2000) to Thomas Piketty (2020), Basic Income (BI) has attracted many prominent supporters. Increasing inequality, precarious employment and conditional forms of welfare have revived interest in BI in recent years. The COVID-19 pandemic, in particular, has underscored the need to enhance income security measures, including the idea of implementing ‘emergency’ and ‘recovery’ BIs (see The Independent, 2020; Torry, 2020).

While arguments in support of BI are diverse, (See Widerquist et al., 2013) concern over changes in the organisation and availability of work often feature prominently. Supporters argue that BI might help address insecurities experienced in relation to precarious employment and the processes of automation and digitisation, while strengthening the bargaining power of labour, and reducing economic inequality (Standing, 2017; Susskind, 2020). There is growing interest from policy scholars investigating how BI might operate and from governments and non-government organisations experimenting with trials (Adkins and Ylöstalo, 2020; De Wispelaere et al., 2018; Gentilini and Grosh, 2020; McFarland, 2018). However, there is yet to be any significant, ongoing implementation of BI as a policy model.

Political and institutional obstacles help explain the failure to implement BI. While its fiscal cost is often cited as central, policy scholars also identify challenges in integrating BI into existing institutional arrangements and prevailing political contexts (De Wispelaere and Noguera, 2012; Hoynes and Rothstein, 2019; Organisation for Economic Cooperation and Development (OECD), 2017). The diversity of ideological support for BI, together with its distinct characteristics as a policy model, is thus presented as an obstacle to practical implementation. Building on the emerging literature, we identify how a particular form of BI might address these challenges in the case of Australia. Australia’s post-War history of strong wage regulation, low unemployment and flat-rate benefits (with few non-income–related conditions up until the early 1980s) has increasingly given way to a labour market characterised by higher levels of non-standard employment and a transfer system more reliant on stigmatising workfare (Australian Bureau of Statistics (ABS), 1999; Campbell and Burgess, 2018; Carney and Stanford, 2018; McDonald and Marston, 2005; Peetz, 2006; Whiteford and Heron, 2018).

In this article, we explore the choices involved in moving towards a more universal social payment system. Acknowledging radical policy change is politically unlikely, we explore pragmatic policy frameworks and identify policy trade-offs. Drawing on the BI literature, we identify two important obstacles to BI implementation; managing trade-offs between cost, adequacy and efficiency, and challenges in adapting existing welfare state structures. Responding to these obstacles, our article has three aims. First, applying insights from historical institutionalism, we explore how a realistic policy trajectory can be built on existing policy structures and recent policy changes that already correspond to the principles of BI. Second, we use microsimulations to model the static fiscal and distributional consequences of moving policy in a more universal direction. Finally, we evaluate the merits of universalism by comparing these trade-offs against the range of existing policy frameworks offered across the OECD. In particular, we are interested in BI as a response to rising inequality given arguments that its lack of targeting may weaken its utility as a strategy in the Australian context, and more recent speculation that BI is a plausible policy alternative in the wake of the shutdown in response to COVID-19 (Dean, 2020).

The first section, ‘BI: Definitions and policy challenges’, describes commonly accepted criteria that define BI in general, before differentiating the principal models of this policy reform. We situate our ‘affluence-tested’ approach to BI along a policy continuum that runs the gamut from minimalist Negative Income Tax (NIT) proposals to radical Universal Basic Income (UBI) schemes. The second section, ‘Targeting within universalism’, explores the institutional context of Australia’s welfare and labour market institutions to develop our account, and to situate it in relation to international debates on targeting and redistribution. Next, ‘Designing an affluence-tested BI proposal’ enumerates the design principles of an affluence-tested approach to BI Australia. We detail two affluence-tested scenarios anchored to (1) the single rate Newstart Allowance (2019), and (2) the single rate Newstart Allowance plus AUD75 per week. ‘Method’ describes the method employed to model the static fiscal impact and distributional effects of these two scenarios using the Australian National University’s Centre for Social Research and Methods PolicyMod model of Australia’s tax and transfer system.

The ‘Results’ section describes our results, showing a fiscal impact of between 5.5 and 6.8 percentage points of gross domestic product (GDP), and a reduction in inequality of between 50 and 65 Gini points. Finally, in ‘Discussion’, we compare these findings to fiscal settings and distributional outcomes across the OECD. We argue our model presents a plausible, though, in the short-term, unlikely, pathway to significantly lower inequality in Australia consistent with existing institutions and policy legacies.

BI: Definitions and policy challenges

There is no universal consensus on the precise definition, normative justification or policy objectives of BI. However, the Basic Income Earth Network’s (BIEN)¹ official definition of BI highlights some commonly cited criteria. BIEN defines BI as: ‘a periodic cash payment unconditionally delivered to all on an individual basis, without means-test or work requirement’ (BIEN, 2019). The following important caveat regarding BI’s definition was included in a statement adopted by BIEN’s General Assembly in 2016:

A Basic Income that is stable in size and frequency and high enough to be, in combination with other social services, part of a policy strategy to eliminate material poverty and enable the social and cultural participation of every individual is often called a ‘full Basic Income’, and a lower one is often called a ‘partial Basic Income’. However, the definitions of ‘full’ and ‘partial’ are highly controversial, and BIEN has not attempted to define them officially. (BIEN, 2016)

BI can be further disaggregated into four basic models. These are: (1) Stakeholder Grants paid as a lump sum to individuals at a particular age; (2) Social Dividends paid to individuals from the proceeds of public ownership and publicly owned wealth funds; (3) UBI that pays the same regular amount to all individuals; and (4) NIT schemes that establish a universal income floor but reduce payment levels as market incomes rise (Ackerman et al., 2006; BIEN, 2019; Meade, 1993; Moffatt, 2004).

Our focus is on UBI and NIT models. The key difference between these models is fiscal churn. To achieve the same net distributional outcomes as the NIT, the UBI model generates substantial fiscal churn as all individuals receive BI through the transfer system and most individuals meet their liability to the government through the tax system (Harvey, 2006; Honkanen, 2014). As Gentilini and Grosh (2020) argue, UBI can be considered ‘conditional’ because it is ‘de facto targeted via taxes’ creating a sliding scale of net beneficiaries and net contributors for a given UBI scheme (p. 76). For some advocates, fiscal churn is a price worth paying to achieve the ethical and political goal of formal universality. For others, the NIT model offers a more feasible and efficient means of achieving the same, or at least similar, goals.

Fiscal churn clearly amplifies the fiscal impact of the UBI model, and cost is probably the most commonly cited objection to BI. The fiscal cost of a particular scheme depends on the BI model being implemented and rate at which payments are set (and withdrawn in the case of an NIT). The UBI model would have the largest fiscal impact for a given payment level. NIT models that establish an income floor have much lower gross fiscal impacts for a given BI payment. Here again, the fiscal cost of such NIT schemes would vary considerably based on the rate at which benefits are withdrawn as market incomes increase.

For illustrative purposes, Table 1 provides a summary of some recent research, focusing on the design, payment rates and gross fiscal cost of UBI-style proposals for the UK, US and Australia. The gross cost of the BI schemes in Table 1 falls within a range of 5%–25% of national income. Gross cost establishes the ‘scale of the challenge’ in relation to the policy trade-offs and choices that need to be made in order to achieve revenue neutrality. For example, the most generous Australian scheme (Cowan, 2017: 13–14) described in Table 1 would see all Australian adults receiving a payment at around the single Age Pension rate at a gross cost of close to $420 billion annually or over 20% of GDP. Subtracting transfer payments replaced by the UBI reduces this figure to a net cost of $230b. This is still more than the Federal government’s entire social security and welfare budget of $180b. The net cost of Cowan’s less generous schemes is a more modest, though still very substantial, $103b and $107b, respectively (Cowan, 2017: 13–14; Kesper et al., 2019).²

Table 1. Gross cost of pension model BI schemes for Australia, the UK and the US.

BI: Basic Income; GDP: gross domestic product.

Sources: Cowan (2017: 13–14), Martinelli (2017: 38), Phillips (2018), Painter and Thoung (2015), Reed and Lansley (2016), Stern (2016), Yang (2019), ABS Cat. 5206001, 2019, Office of National Statistics (ONS, 2020), US Bureau of Economic Analysis (2020).

Note: $ = AUD$

Proposals not only differ in level, between ‘partial’ and ‘full’ BI, but also financing mechanisms and political orientation. For example, in Capitalism and Freedom, Milton Friedman proposed an NIT scheme funded through a cut in welfare spending in the United States by over 50% in the early 1960s (Friedman, 1982: 190–196). In contrast, James Tobin’s contemporaneous examination of NIT schemes proposed a net increase in spending of between US$7b and US$50b (see Tobin et al., 1967: 24).

Australia’s system of income support has been described by Ingles as ‘not too far removed from a NIT or [Guaranteed Minimum Income] GMI scheme, albeit of the categorical (and conditional) type’ that is ‘characterised by poor design, excessive complexity, and unwanted overlaps with the tax system and other income transfers’ (Ingles, 2000: 1). Since at least the First Report of the Henderson Poverty Review in 1975, there have been numerous efforts to explore the costs and benefits of reforming Australia’s transfer system based on NIT/GMI and UBI models. In Australia, and internationally, social scientists are increasingly using microsimulation techniques to estimate the fiscal impact and static distributional effects of different BI scenarios (see Arthur, 2016; Dawkins et al., 1998; Ingles, 2000; Ingles et al., 2019; Martinelli, 2017). We aim to contribute to this literature by modelling an affluence-tested BI for Australia. Our approach sits between an NIT and UBI model, as the payment is reduced with income, but some payment would be received by a large majority of people including most net contributors.

This discussion highlights the complexities of designing any real-world BI (see Martinelli, 2019; Standing, 2017). Putting dynamic effects to one side, Martinelli (2019) argues that ‘BI advocates face an irreconcilable trilemma in policy design, between a) affordability/controlling cost, b) adequacy/meeting need, and c) securing the advantages of a radically simplified welfare system’ (p. 3). The ‘three horns’ of Martinelli’s (2019) trilemma underscore the difficult policy trade-offs that must be considered in relation to any BI proposal (p. 20). Here, we are concerned with identifying the nature and scale of these trade-offs, taking current policy settings as a starting point, to assess the possibilities and implications of moving towards universalism.

Targeting within universalism

Policy scholars emphasise the challenges of incorporating BI into existing policy structures as much as technical feasibility. These approaches build on broader welfare state scholarship, which emphasises political contestation and institutional path dependence (Pierson, 1998). Influenced by power resource theory, many argue labour movement strength helps explain the development of welfare state institutions, which in turn shape overall levels of economic inequality (see Esping-Andersen, 1999). The most egalitarian welfare states, in Scandinavia, provide generous and extensive forms of social insurance, partly built around union-managed funds and union membership. In contrast, the liberal countries of the Anglosphere tend to have residual social policies, which provide greater space for market provision and only modest, flat-rate benefits. Flat-rate benefit systems are often means-tested and stigmatising. More recently, conditionality has been most extensive in the same liberal countries, where surveillance and penalties for non-compliance attach to flat-rate benefit schemes (see Peck, 2001).

Punitive forms of workfare reflect a divisive political dynamic, where welfare recipients are marginalised, undermining efforts to ensure payments are adequate. Earlier work by Korpi and Palme (1998) described a similar dynamic as a ‘paradox of redistribution’. Targeting, they argued, undermined political support for the welfare state because many citizens did not see themselves as beneficiaries. Instead efforts to focus social spending only on the poor produced less redistribution than more universal models of social insurance. Without political support, overall social spending was lower, reducing the aggregate redistributive effect. Thus, while social spending in the liberal welfare states is more redistributive per dollar, reflecting the targeting of benefits, aggregate social spending is lower than in Scandinavia (OECD, 2019a: 261, 259), and liberal regimes achieve much less redistribution overall than their Scandinavian counterparts (Ortiz-Ospina and Roser, 2016).

Recent efforts to revisit the paradox of redistribution thesis have produced varied results. While many broadly confirm the thesis (Lefèbvre, 2007; McKnight, 2015), others have found contrary evidence (Brady and Bostic, 2015; Marx et al., 2013). One influential explanation of these contrary findings is that social policy itself has changed over time, as targeting has been combined with efforts to reduce poverty traps. Focusing on the comparison of child benefits Van Lancker and Van Mechelen (2015) argue that an emerging model of ‘targeting within universalism’ may achieve the greatest redistribution. This describes schemes where the poorest receive the greatest benefit while a broad section of society is able to access payments, ensuring programmes enjoy public support and are viewed as inclusive.

Some Australian social benefit systems are usefully understood through the logic of ‘targeting within universalism’, reflecting elements of an NIT approach. Developing from the late 19th century, Australian benefit schemes have long adopted a model of flat-rate payments funded through general revenue. While consistent with liberal welfare traditions, the Australian welfare state potentially differed in substance. Francis Castles (1985) argued the structure of benefits proved inclusive of the vast majority of workers when combined with commitments to full employment and wage arbitration. Australia’s ‘wage earner’ model, he argued, achieved equity by reducing the initial inequality of market incomes, rather than redistribution through social spending.

The conditions of Castles’ ‘wage earner’ welfare model, however, have transformed considerably since the 1980s. Wage arbitration has been substantially unwound, and increasingly acts as a safety net for the low paid. The breakdown of Keynesian institutions internationally, combined with liberalisation of policy settings within Australia, coincided internationally with the effective end of government commitments to full employment as an achievable policy goal in favour of greater emphasis on inflation targeting (Beggs, 2015; Langmore and Quiggin, 1994). Labour market liberalisation, in particular, has eroded the standard employment contract, coinciding with declines in union density. Social payments now sit alongside pervasive insecurities, despite decades of unbroken economic expansion and, until 2012, sustained increases in real wages (Bryan and Rafferty, 2018; Stanford, 2019; Wilson and Ebert, 2013).

Alongside labour market liberalisation, benefits have also changed. Changes followed a similar technical model of liberalisation. Almost all social payments are now means-tested; however, means-tests have become less strict, allowing beneficiaries to remain eligible for some payment while earning some market income. In addition, new forms of conditionality were introduced for many payments, reflecting the advance of workfare. Elsewhere, Spies-Butcher (2020) has argued that while these reforms involved similar technical changes across benefit systems, two distinct policy logics emerged.

Some payments, like age pensions and family payments, facilitated the combination of market and social incomes much more readily, by allowing recipients to earn some ‘free’ income without impacting benefit levels, or by reducing the rate at which benefits are withdrawn as market income rises. While these payments remain means-tested, there are few other forms of conditionality, they are received by a majority of the potential recipient population and enjoy broad public support. For example, the majority of families with children are eligible for the main family support payment, Family Tax Benefit Part A, although the highest payments are received by the lowest income families, and many middle-income families receive very small payments.

We call this type of payment structure ‘affluence testing’ (see Wilson et al., 2013: 633), as income-testing is primarily used to limit access by the better off. Alternatively, other payments, like the unemployment benefit, have become increasingly conditional, are the focus of divisive media coverage and political campaigns and have failed to increase³ in line with wages or other payments, like the age pension.

Affluence testing potentially aids universalism while also advancing competition by focusing attention on economic incentives. Analytically this is expressed through a focus on Effective Marginal Tax Rates (EMTRs), which incorporate changes in income received at the margin from additional market income by combining the effects of income tax scales and benefit withdrawal rates. By combining the effects of tax and welfare systems, EMTRs provide an analytic lens to examine the incentive and distributional impacts of change across both policy domains. This analytic choice also has political implications because systems of taxation and social spending have come to embody different administrative approaches and normative commitments.

Analysing the tax and welfare systems through the lens of EMTRs, we argue, highlights a tension between liberal and paternalist modes of liberalisation. Paternalist reforms to workfare treat market income and social income differently, managed through different bureaucracies, with strong forms of surveillance and conditionality applied to social income. These differences reflect a politics of ‘deservingness’ based on a construction of welfare ‘dependence’ (see Fraser and Gordon, 1994) as undesirable and caused by character and cultural flaws that require behavioural interventions (Mestan, 2014). Alternatively, EMTRs identify monetary incentive structures, reflecting a neoclassical economic assumption that all individuals are instrumentally rational.

A focus on incentive structures can also be found in Milton Friedman’s (1968) well-known proposal for a NIT, which is sometimes understood as a form of BI. Friedman’s model takes the logic of EMTRs as its organising principle, by applying a flat ‘EMTR’ that is structured entirely as a benefit withdrawal rate for low- and middle-income earners until no entitlement remains and then taxing income above this threshold at the same marginal rate. Friedman’s proposal envisaged funding the NIT through the abolition of other social programmes, effectively replacing in-kind provision with cash. Similar ‘flat tax’ BI models have also been proposed by free market advocates in Australia (Humphreys, 2005). In practice, affluence testing involves a series of smaller policy changes in this direction, although with greater overlap between social payment recipients and income taxpayers, and it is funded by expanding social spending rather than contracting other forms of social provision.

Following this argument, we suggest a series of principles that would represent an affluence-tested model of BI that conceptually integrates tax rates and benefit withdrawal rates,⁴ which in principle is identical in distributional and incentives terms to an equivalent NIT model or a full BI funded through income tax. In developing an affluence-tested model of BI, we adopt four principles (also see Spies-Butcher and Henderson, 2019) that reflect the logic of integrating tax and welfare systems via EMTRs and acknowledge existing institutional arrangements as the starting point for reform. They are as follows:

1. 1. EMTRs should only increase with income (progressivity of incentives);⁵

2. 2. High-income earners should receive no net benefit from moves towards universalism (fiscal efficiency);

3. 3. No below median income earner should be left worse off (i.e. have a lower net income for any given market income) (equity);

4. 4. The current tax scale should be largely taken as given (path dependency).

The first of these principles amounts to a significant relaxation of current means-testing arrangements, but it is broadly consistent with the underlying principles of current policy structures. The second reflects existing policy norms that prioritise public spending based on need. The third principle requires that those already on benefits are not left worse off. And the final principle allows us to build a model that is analytically separate to the financing measures introduced to fund it.

Designing an affluence-tested BI proposal

Taken by themselves, these principles under-determine a specific BI proposal. To facilitate microsimulation modelling, we develop specific payment structures presented in Tables 2 and 3. We are conscious that any specific model is unlikely to constitute a real-world policy prescription. Instead, our aim is to better understand the fiscal and distributional dynamics of moves towards a BI structured according to our four principles, which we argue reflect the logic of affluence testing built into aspects of Australia’s existing social policy architecture.

Table 2. Affluence-tested Basic Income payment Model 1, Newstart.

Source: Authors’ calculations.

1. This Column differs from the marginal tax rate by taking account of a tax offset, which provides a rebate for some taxpayers (earning between $19,543 and $48,000), and then unwinding the rebate for those earning over $90,000. The initial rebate effectively reduces the tax paid per additional dollar earned, while unwinding the rebate effectively increases the amount of tax paid per dollar earned. Structured as a rebate, however, these impacts are not recorded in the headline tax rates.

Note: $ = AUD$

Table 3. Affluence-tested Basic Income payment Model 2, Newstart +$75pw.

Source: Authors’ calculations.

Note: $ = AUD$

We model two versions of BI. Our analysis is based on payment and tax systems prior to the changes made in response to COVID-19. Table 2 describes the tax and benefit structure of our affluence-tested BI Model 1 with a base payment rate approximately equivalent to Newstart (the main unemployment benefit in 2018, now renamed JobSeeker). This echoes our fourth principle, applied to benefit levels.

Table 3 describes the tax and benefit structure of our affluence-tested BI Model 2 with a base payment set at approximately $75pw above Newstart,⁶ reflecting the campaign to ‘raise the rate’, which has now gained broad stakeholder consensus, although not permanent policy change (Raise the Rate, 2019).

Withdrawal rates for our models were derived by taking tax rates and benefit levels as given and withdrawing benefits to approximately zero for incomes above $180,000 (the top marginal income tax threshold).⁷We note that one implication of this method is that the higher base payment in Table 3 is part-funded by higher EMTRs (tighter targeting) compared to Table 2. This trade-off is common to a number of recent tax and transfer system changes such as part funding an increase to the base pension rate via an increase in the withdrawal rate under the Rudd Labor Government (Yeend, 2009), suggesting our model follows an existing institutional logic.

Changes to the tax system applied retrospectively to the 2017/2018 tax year complicate our models by creating tax offsets. Offsets are logically equivalent to a reduction in the marginal income tax rate (a tax cut) for low- and middle-income earners. However, the offset itself is income tested. Withdrawing the offset for higher income earners is logically identical to an increase in marginal income tax rates (a tax rise). These effects are reported in each Table under the column ‘Effective tax rate including tax offset’.⁸

We note that a number of other benefits are paid at a higher rate than Newstart or Newstart plus $75. We follow other BI scholarship and modelling⁹ in retaining these more generous categorical benefits on the basis that their respective payment levels reflect non universal needs, such as parenting or disability. We therefore assume supplements for those on higher payments so that no current beneficiary is left worse off as a consequence of introducing BI.¹⁰

While reliable distributional estimates require BI proposals be net fiscally neutral, our aim here is to explore potential benefit structures, not to propose specific financing mechanisms. We note there are a wide range of potential options for financing BI schemes through the tax system. Some, such as the elimination of existing tax concessions, are strongly progressive, others, such as an increase in consumption taxes, are regressive (for alternative financing options based that include wealth taxation, see Ingles et al., 2019). In practice, trade-offs between policy objectives, expenditure requirements and levels of tax resistance would likely necessitate a combination of financing measures for any BI scheme.

While the optimal tax mix is not our concern here, any real-world change is very likely to require proposed funding to be at least proportionate overall, reflecting a ‘compensation politics’ observed in previous Australian tax reforms (Wilson et al., 2013: 633). For the purposes of estimating distributional impacts of our fiscally neutral Models 1 and 2, we simulate a proportionate tax increase by increasing marginal tax rates evenly across the entire income distribution sufficiently to ensure fiscal neutrality.¹¹

Method

Australian National University’s (ANU) Centre for Social Research and Methods has developed a microsimulation model of the Australian Tax and Transfer system, PolicyMod (2019). This model is capable of modelling most aspects of the Australian personal income taxation and welfare system including the current and alternative policy settings. The model is based on ABS survey data and uses a range of other data from the Australian Bureau of Statistics and other government departments to improve the accuracy and timeliness of the model. The major data source used by the model is the ABS Survey of Income and Housing 2015–2016. This data source has around 27,000 adults aged 15 plus living in around 14,000 households. This data set is updated to the latest ABS survey usually every 2 years. The model is initially a static model in that it considers the impact of policy on the ‘day after’ meaning that PolicyMod does not attempt to model behaviour changes that may result from policy changes.

PolicyMod is well suited to modelling the household and individual impacts of the existing policy and alternative policy for both the current financial year and over the forward estimates. The impacts modelled can be viewed either through the distributional model or the hypothetical (cameo) model. The distributional model considers the entire Australian population and is capable of estimating the impacts of policy for a vast array of variables such as family type, income deciles, housing tenure and social security payment type. The cameo model provides policy impact analysis for selected families such as single parents with a range of private incomes. The cameo model details the disposable income and effective tax rates schedule as private income or hours worked change.

In this analysis, we have focussed on the distributional model for analysis of the fiscal impacts and the impacts on different household types. This is a similar approach to that taken in several of the microsimulations of BI proposals detailed in Table 1. The standard outputs from the model include the ‘winners’ and ‘losers’ from policy change and an estimate of the overall fiscal impact on the Commonwealth Budget. The model is flexible enough to enable this analysis over the current year or any set of years into the future.

In modelling the various BI options in PolicyMod, we provided a BI to each adult under pension age. This amount is tapered away as described in Tables 2 and 3. This amount replaces any government pensions or allowances a person may be entitled to. Family benefits, childcare payments and any rent assistance amounts remain unchanged from the current entitlement. The BI amount is treated as income for childcare and family payment purposes but is not considered taxable income.

Results

To assess the distributional and fiscal impacts of our proposals, we present four microsimulations of our affluence-tested BI proposals. We present a range of data for both unfunded and fiscally neutral versions of Models 1 and 2. The estimated net cost of Model 1 is $103.45b, and the net cost for the more generous Model 2 is $126.1b. To achieve fiscal neutrality, all marginal income tax rates would need to be increased by 12 percentage points for Model 1 and by 14.5 percentage points for Model 2.¹² To clarify, we do not suggest this specific tax change is a likely financing mechanism, rather we use it as a means to approximate the distributional effects of our affluence-tested BI. We do not consider any reduction in spending on compliance involved in administering the current systems. Thus, our estimates likely overstate the tax change required.

Table 4 presents the static distributional results for both unfunded and fiscally neutral microsimulations of Models 1 and 2. We focus on the net distributional impacts of our fiscally neutral models compared to the existing tax and social security system. Table 4 shows the percentage (and dollar) changes in annual equivalised disposable income by quintile for our two revenue neutral simulations. We also provide distributional breakdowns by household type for these models in Appendix 1. These modelling results are based on the relevant demographic, taxation and social security data for the financial year 2017–2018.

Table 4. Distributional impacts of affluence-tested Basic Income proposals. Percentage change in annual equivalised disposable income, (change in dollars).

Source: Authors’ calculations, PolicyMod (2019).

Note: $ = AUD$

The results show all proposals would achieve a substantial redistribution of income. For Models 1 and 2, both unfunded and fiscally neutral, the biggest proportional beneficiaries are the bottom quintile. This is perhaps surprising given Australia’s current social payments system is one of the most targeted in the world, providing the least to high-income groups and targeting payments to those on the lowest incomes (Whiteford, 2017). Both our proposed payments wind back existing targeting and, in Model 1, provides no additional payment to beneficiaries currently receiving the full Newstart allowance or other higher payment.

We offer three explanations of the distributional outcomes. First, not all those potentially currently eligible for categorical payments receive them, a narrowing of effective eligibility that is likely reinforced through conditionality. Our findings suggest the non-payment of benefits to those on low incomes increases inequality, which is addressed through more universal provision, although we also note the bottom quintile likely includes some with more resources due to underreporting of business income, which might be addressed through imputing income from capital. Second, individualisation of payments likely aids redistribution. Rules restricting payments based on a partner’s or parent’s income are more likely to impact low-income earners than high-income earners (Fisher and Zhu, 2019). This conclusion is reinforced by detailed household breakdowns provided in Appendix 1, which report the largest beneficiaries are from households with multiple adults.

Finally, we note that in our unfunded models the highest absolute increases in income are in the top three quintiles, a group usually described as ‘middle Australia’. This highlights a key feature of the current targeted Australian model. While aiding technical redistributive efficiency, the current model potentially disadvantages middle-income workers who have not enjoyed proportionate increases in market incomes following liberalisation, but who are also excluded from compensation via social payments. Even so, all payment models prove progressive, because the relative increases in income are much higher at the bottom of the distribution, highlighting how loosening targeting can reduce inequality.

In addition to the direct impacts of introducing additional payments, the fully funded models achieve additional redistribution through taxation. This has the most impact at the top of the distribution, where even proportionate increases in taxation yield large absolute increases in tax contributions, and our model of ‘affluence testing’ limits access to the new benefit. This impact would obviously be larger (smaller) if the tax changes were progressive (regressive).

Next, we examine trade-offs between social goals of redistribution and poverty reduction against fiscal costs. While we acknowledge that the fiscal costs involved in either model are substantial, and outside usual short-run political constraints, we also note the wide variety in levels of taxation, inequality and poverty across the OECD. We suggest this real-world variety implies a medium- to long-term political choice, albeit one more likely to be achieved through a series of incremental moves (something we explore in Spies-Butcher and Henderson, 2019) rather than a single policy change.¹³

Table 5 presents key fiscal and distributional information for the two fiscally neutral microsimulations of Models 1 and 2. Data for the Gini coefficient, poverty rate and median income were produced by PolicyMod. The tax as a percentage of GDP is derived by adding the cost of each model to current Commonwealth tax receipts, assuming no other change.¹⁴ We then compare these results internationally in Table 6. Note the different tax-to-GDP ratios for Australia in Tables 5 and 6. Table 5 includes Commonwealth taxation only, while Table 6 is a measure of total government tax-to-GDP ratios that includes taxes levied at the federal, state and local levels. The centralised structure of Australia’s tax and transfer system makes the Commonwealth by far the most important level of government in relation to evaluating any BI proposal.

Table 5. Fiscal and distributional impacts of Basic Income models.

GDP: gross domestic product.

Source: Authors’ calculations, PolicyMod (2019).

Note: $ = AUD$

Table 6. Fiscal and distributional impacts compared internationally.

GDP: gross domestic product.

Source: Organisation for Economic Cooperation and Development (OECD) (2019b).

It is notable that the poverty rate for Model 1 is below that for Model 2, even though the latter involves a more generous payment and lower inequality. This is explained by the method of calculating poverty as a proportion of median income. The structure of our BI payment has relatively larger impacts around the middle of the distribution, lifting the median income and thus increasing the relative poverty line. We suggest it is more reasonable to interpret these changes as indicating that the current poverty rate may underestimate need (based on a deflated median income), but also that caution should be exercised in comparing poverty rates across time and place.

Australia is currently a low-tax economy, ranking well below the OECD average and most comparable countries. While Australia has a history of cultural egalitarianism, it is now a relatively unequal country by OECD standards. Given these conditions, we note the impact of introducing our models would potentially involve convergence to international norms. Results from both payment models would reduce inequality to levels between Scandinavia and Germany, yet involve an overall tax take well below the levels of taxation in these countries, and closer to the OECD average. Of course, policy change towards BI is likely to be accompanied by other policy changes with fiscal implications. We simply note such a model appears to offer a policy pathway towards greater egalitarianism consistent with Australia’s existing policy institutions and consistent with international fiscal norms.

Discussion

Our article began with three aims. First, we drew on welfare state and institutionalist literature to analyse how Australia’s benefit system already incorporated elements of a BI and dynamics that might realistically move it further towards universalism. Through this analysis, we identified the pattern of ‘affluence testing’. We then proposed four principles consistent with affluence testing to design BI models consonant with existing Australian policy institutions. Finally, we explored the distributional and fiscal trade-offs associated with these models through PolicyMod.

The results facilitate a number of useful reflections for policy scholars on the possibilities for an Australian BI. First, our results are consistent with previous research indicating that any reasonable BI payment is likely to have a large fiscal impact, even when incorporating elements of a NIT model. However, the model we present is less costly than what might be expected from other recent analysis. For example, the 2017 report by the Centre for Independent Studies (see Table 1) found a BI model of only $10,000pa that also provided supplements to all those currently on higher payments to ensure they were not worse off, similar to our models, would cost $102.7b (Cowan, 2017: 13–14). This is virtually identical to the cost of our $14,647 payment in Model 1 ($103.45b). Including some degree of ‘affluence testing’ is thus an effective mechanism for reducing gross fiscal cost by moving closer to an NIT model with a gradual taper.

Second, an affluence-tested model of BI appears consistent with the redistributive logic of the current tax and welfare system. In all the models presented, the largest proportionate gains accrue to lower income households compared to higher income households. Indeed, our model unwinds aggressive forms of targeting, by eliminating conditionality, individualising payments and significantly loosening benefit withdrawal rates. Furthermore, it substantially lowers inequality and poverty, even if the base payment of Newstart were not increased. This conclusion is broadly consistent with the ‘paradox of redistribution’ thesis, and with recent research on ‘targeting within universalism’.

Third, our models help to situate medium- to long-term political choices in framing trade-offs between levels of taxation and inequality. Given these models build on the institutional architecture of the existing tax and payment systems, substantially reduce inequality and do so without exceeding international fiscal norms, we argue an affluence-tested BI is a viable potential pathway for Australia, reflecting a welfare model consistent with both the egalitarian norms of social democracies and the institutional norms of liberal regimes. We suggest it therefore represents a realistic medium-term political choice, which might be gradually realised by advancing the relaxation of benefit withdrawal rates over tax cuts, and individualising payments.

Finally, we reflect on the relationship between payments, inequality and work. We make three observations based on the distributional data. First, moves towards universalism are likely to significantly benefit those on the margins of the labour market, with very low or no market income. Those in the lowest quintile stand to make the largest proportional gains across both affluence-tested models evaluated in this article. Second, universalism is likely to substantially benefit middle-income workers. Unwinding targeting expands access to social support to much of the ‘middle’. Third, we note the likely dynamic effects of these models. Affluence testing implies a partial buffer to the loss of income at the margin, while improving work incentives for low- and middle-income workers. As many casual and contract workers face irregular incomes, this partially shifts the risk of unstable incomes from workers to the state, mitigating what Jakob Hacker (2006) describes as the ‘great risk shift’. In shifting risk to the state’s fiscal capacities, affluence testing aids redistribution in general, but also facilitates intra-national redistribution towards poorer regions.

Conclusion

BI has re-emerged as a potential policy option in the context of growing inequality and labour market insecurity. Yet it remains unrealised. Previous research identified ideological diversity among proponents and the policy’s radical departure from existing welfare state structures as important explanations of its failure. Our aim was to explore policy possibilities in the light of these conclusions and to highlight the fiscal/distributional trade-offs. Affluence testing, we argue, reflects a commitment to universalism, while also reflecting political and policy legacies of the Australian welfare state. Our models take current policy settings as a starting point, and move towards universalism by individualising payments, removing non-income–based conditionality and reducing benefit withdrawal rates. While full implementation of these principles remains beyond short-term political possibilities, we argue our findings reflect a surprising compatibility between Australia’s existing benefit system and a modified BI that can inform debate over future efforts to reduce inequality and insecurity.