Abstract
Targeting as a mechanism for allocating scarce public resources efficiently and equitably has recently come under sustained attack, for both pragmatic reasons – the apparent impossibility of achieving perfect targeting accuracy – and ethical reasons – social divisiveness and perceptions that excluding some people from benefits is socially unjust. Universalism is offered as a fairer and more inclusive alternative, although universal programmes are sometimes erroneously conflated with categorically targeted programmes (such as social pensions for all older persons). This article critically examines the ‘targeting’ versus ‘universalism’ debate, drawing on three principles of redistributive justice – equality, equity and need. A trifurcated approach to social policy is proposed: the equality principle (universal provision) should be applied to essential social services; equity (payments proportionate to contributions) are appropriate for social insurance schemes, but social safety nets or social welfare grants should be allocated on the basis of need (vulnerable groups or poverty).
Introduction
Targeting has long been acknowledged as one of the most intractable challenges in social policy. How to identify and register eligible individuals or households for social programmes, accurately and cost-effectively, without including some ineligible people and excluding some eligible people? In practice, 100% targeting accuracy is a chimera, virtually impossible to achieve, which raises questions about fiscal efficiency (how much public money can acceptably be wasted on ineligible beneficiaries?) and social justice (is it morally acceptable to ‘leave someone behind’?). This article examines the ethical and pragmatic case for and against targeting. This is a topical debate, both in wealthy countries with well-developed social security and comprehensive social welfare systems and in low-income countries with limited and fragmented social protection programmes.
Extensive literatures exist on alternative targeting mechanisms (Grosh et al., 2008; Samson et al., 2010; Sharp, 2001; Slater and Farrington, 2009), targeting errors (Coady et al., 2004; Cornia and Stewart, 1993; Devereux et al., 2015; Sabates-Wheeler et al., 2014) and administrative, social and political targeting costs (Besley and Kanbur, 1990; Grosh, 1995; Macauslan and Riemenschneider, 2011). A smaller literature addresses the question of whether targeting should be done at all or whether universal programmes should be implemented instead (Ellis, 2012; Hoddinott, 2007; Kabeer 2014; Mkandawire, 2005; Sepúlveda, 2014).
Many of the world’s wealthiest countries are currently facing systematic attacks on the guaranteed protections that were established during the 20th century, with cutbacks in universal access to essential services and means testing applied to previously untargeted social grants (e.g. child benefits in the United Kingdom), under the guise of enforced austerity responses to financial crisis and economic slowdown. These reversals of hard-won gains have been mirrored by a rapid expansion of social protection programmes, including free access to essential services (notably primary education and primary health care) and a range of unconditional or conditional cash transfers, across low- and middle-income countries of Latin America, Africa and Asia.
As a development policy agenda, the ‘much-celebrated popularity of social protection, particularly cash transfers’ (De Haan, 2014: 313) reached an apogee with a joint statement by the International Labour Organisation (ILO) and the World Bank in June 2015, which called for ‘Universal Social Protection’. The document’s repeated references to ‘universal coverage’ might seem like a call for untargeted social transfers to everybody, but this is not the case: ‘Universal social protection includes adequate cash transfers for all who need it’ (ILO and World Bank, 2015: 1).
Before proceeding, it is important to clarify what is meant by ‘universal’ programmes. Often the word ‘universalism’ is misleading, and its proponents actually mean ‘demogrants’ – programmes that target population cohorts on demographic characteristics. A social pension is often called universal if it is given to all persons aged over 60 years rather than being means-tested, but in targeting terminology, this is actually a ‘categorical’ programme that targets age-eligible individuals, instead of the entire population or the poor. Alternative terms include ‘targeted universalistic schemes’ (Standing, 2008: 4) and ‘targeting within universalism’ (Skocpol, 1991: 414). But surely a universal programme is one that reaches everybody, or at least all poor people? This is not merely a semantic issue. If we agree with Ravallion that targeting is ‘an instrument for reducing poverty’ (Ravallion, 2007: 2), then a pension that reaches everyone aged over 60 years and no-one aged under 60 years has high inclusion and exclusion errors, because all non-poor people aged over 60 years are included and all poor people aged under 60 years are excluded. Targeting older persons might be motivated by evidence that they are more likely to be poor than average, but age is a weak proxy indicator for poverty and a universal pension excludes large numbers of poor people who do not meet the age criterion. The only social programmes that are guaranteed to reach all poor people are those that are genuinely universal – such as free health care for all, a general food subsidy, or a ‘basic income grant’ (BIG) for all citizens.
‘Universal social protection … for all who need it’ is an oxymoron that reflects this widespread confusion between targeting and universalism and between two diametrically opposed principles of redistributive justice: ‘equality’ and ‘need’. A third principle is ‘equity’ (Miller, 1999; Reeskens and van Oorschot, 2013).
Equity principle. People who contribute more should receive more. This principle underpins contributory social security systems, where payments are proportional to income tax and payments made into social insurance funds. ‘In practice, equity is applied by organizing social protection through social insurance … which is the dominant model of redistribution in conservative welfare states, as it reproduces social hierarchies and statuses’ (Reeskens and van Oorschot, 2013: 1176).
Equality principle. Every citizen or resident should receive the same benefits, irrespective of their level of need or their contributions to income tax and social security. This principle underpins universalistic approaches to social policy, such as free education or health services for all or an untargeted ‘BIG’.
Need principle. People who need more – low-income households, vulnerable individuals such as persons with severe disability and so on – should receive more support than the better-off, who can and should provide for themselves. This principle underpins social democratic welfare states with redistributive social welfare systems (Esping-Andersen, 1990), where the state targets social assistance to the poor and vulnerable, often by applying a means test to determine eligibility.
This article examines arguments for and against targeting of social programmes, especially social transfers, which amounts to a choice between the equality principle (universalism) and the need principle (poverty targeting). The article discusses the general principles of targeting versus universalism, but focuses specifically on targeting according to poverty-related criteria, which are the most contested, ideologically and pragmatically. It will be argued that social assistance should be targeted on the basis of need, while social insurance should follow the equity principle, whereas social services should be universally accessible following the equality principle.
The case for targeting
Two powerful arguments can be made for targeting social transfers, one normative or ethical and one pragmatic or operational. Both arguments support the ‘need’ principle of redistributive justice. The ideological argument is that targeting aims to transfer resources from those who have more than they need to those who have less than they need to sustain a minimum subsistence or a decent standard of living. Targeted transfers are a prerequisite for redistributive justice: if they are well designed and delivered, they will reduce inequality and poverty.
The pragmatic argument is that policy-makers operate in a real world of budget constraints. There is never enough money budgeted to distribute to all the poor – if there was, the problem of poverty could be solved simply through generous social assistance programmes. Given the reality of budget constraints, scarce public resources must be used optimally and allocated efficiently, where they can achieve the maximum impact. If poverty reduction is an objective of public policy, social spending should be directed towards the poor who need income support, not spread thinly over the entire population including to those who do not need it.
Of course, the line between ideological and pragmatic arguments for and against social protection programmes is often more blurred than real. Public spending decisions are always political choices, so a ‘pragmatic’ assertion by a government (‘We can’t afford a universal Basic Income Grant’) could instead be interpreted as a reflection of political priorities (‘We are choosing to spend public money that could have been allocated to a Basic Income Grant on the Ministry of Defence, which leaves only enough budget for a social grant that targets the poorest ten percent’). In high-income countries, public spending on social protection averages around 20% of gross domestic product (GDP), but in low-income countries, the average is only 2–4% (ILO, 2014: 126), so there is more fiscal space for redistributive social protection than politicians and policy-makers like to admit. Ultimately, any decision about the allocation of public resources has implications for social justice.
Targeting as a mechanism for redistribution
Redistribution is implicit in many definitions of social protection, specifically in its social assistance or ‘safety net’ component, as reflected in these three definitions from an academic, donor agency and government perspective, respectively.
‘Social protection is a means of providing short-term assistance to individuals and households to cope with shocks while they are temporarily finding new economic opportunities that will rapidly allow them to improve their situation’ (Barrientos et al., 2005: 4).
‘Social assistance involves non-contributory transfers to those deemed eligible by society on the basis of vulnerability or poverty’ (Department for International Development [DFID], 2005: 6).
‘Social protection refers to policies and practices that protect and promote the livelihoods and welfare of people suffering from critical levels of poverty and deprivation and/or are vulnerable to risks and shocks’ (Republic of Zambia, 2006: 210).
The view that societies should provide support to the ‘deserving poor’ draws on a long history of interventions by government agencies, such as the Elizabethan Poor Law of 1601 in England, which taxed wealthier citizens to provide assistance to ‘legitimately needy’ community members – but not to the ‘idle poor’ or ‘vagrants’, who instead faced imprisonment. In the late 19th century, this became codified in European countries as a ‘social contract’ which locates the responsibility for providing social protection with the state. In 1881, the German Emperor wrote to Parliament: ‘those who are disabled from work by age and invalidity have a well-grounded claim to care from the state’ (quoted in Liedtke, 2006: 1). In the United Kingdom, the Beveridge Report of 1942 elevated social welfare to a right of citizenship, which laid the foundations of the modern welfare state, including free health care, child benefits for all families with children and social grants to poor people who are sick, unemployed or retired.
The evolution of European-style social welfare systems is grounded in delivering targeted income support to those members of society who need it, as well as universal free or subsidised access to essential services (such as the National Health Service in the United Kingdom, which is free for all at the point of delivery) and benefits that are targeted not by income but to population categories (such as the Scandinavian welfare state model, which provides ‘universal’ pensions and child benefits (Blomqvist and Palme, 2015)).
The ‘need’ principle is analogous to Marx’s famous maxim, ‘From each according to his ability, to each according to his needs’. As Marx himself pointed out, the redistributive implications of this principle are that poor individuals will receive unequal rather than equal bundles of goods and services, since different people have different needs. This could be called a ‘differentiation principle’, and it applies to both the nature and level of needs. In terms of income poverty, some poor people are poorer than others – the poverty gap of a destitute person is larger, so they require more income to escape poverty than someone subsisting just below the income poverty line. In terms of the nature of individual needs, a hungry child needs a different basket of support (food) than a person with disability (a prosthetic limb). Ethical targeting is not only about reaching the right people but is also about delivering the appropriate amount and type of support.
Achieving an ethical distribution of wealth does not require transferring an equal amount of income or assets to every individual member of a society (equality of transfers); it requires transferring enough income or assets to those whose income is considered too low, such that everyone reaches a socially acceptable level of income (equality of outcomes). This is the redistributive function of social protection: ‘Social protection consists of the public actions taken in response to levels of vulnerability, risk and deprivation, which are deemed socially unacceptable within a given polity or society’ (Norton et al., 2001: 7).
It follows that achieving equitable outcomes requires delivering unequal transfers, following the need principle rather than equity or equality. ‘The right to an unequal share in the consumption of goods and services actually results in a higher form of equality – all people will be equal in the sense that the needs of all will be met’ (Pena, 2011). This is undeniably an ethical approach, but it requires targeting and it demands a lot of information not only to identify the poor but also to quantify precisely how poor each poor person is.
Targeting as an efficient use of public resources
Policy-makers working with a fixed budget for social transfers face a spectrum of options. A universal programme implies dividing the resource envelope by the total population. So if the population of Targetland is 1 million and the budget is 10 million dollars, each citizen or resident would get 10 dollars. There would be no administrative costs and no social costs of targeting – since everybody benefits, there will be broad popular support and no basis for resentment – but the transfer will be too small to make a discernible dent on poverty. At the other extreme, the policy-maker could transfer 20 times this amount to the poorest 5%, assuming these individuals could be accurately identified at zero cost (a heroic assumption, as discussed later). So 50,000 people would receive 200 dollars each, a significant amount of money which could have a poverty-reducing impact but would probably provoke resentment from excluded neighbours, especially those who are only slightly less poor than the targeted beneficiaries (Ellis, 2012).
This spectrum can be characterised as one ranging from wide coverage, low administrative and social costs but limited poverty impact to limited coverage, high administrative and social costs but significant poverty impact. From a poverty-reducing perspective, narrow targeting is more efficient as it reaches the poorest and excludes the non-poorest, while transferring enough resources to make a difference to poverty and inequality. This satisfies the need principle of redistributive justice. The challenges are in defining eligibility – where to draw the line between the poor and non-poor, or ‘ultra-poor’ and ‘moderately poor’ – and in identifying who is eligible and who is not, at relatively low cost with acceptably low errors.
Overall, available empirical evidence confirms that targeting does benefit the poor. A review of benefit incidence on over 100 programmes found that targeted programmes ‘provide approximately 25% more resources to the poor than would random allocations’ (Coady et al., 2004: 2). The way targeting is done matters. Coady et al. (2004) ranked targeting mechanisms by their ability to reach the poor and found that self-targeting on public works was most accurate, followed by geographic targeting and means testing. Proxy means tests, community-based targeting and categorical targeting of children each achieved small but highly variable pro-poor benefits across more than 100 programmes reviewed. Categorical targeting of older persons, along with two other mechanisms, performed worst.
There are arguments in favour of social transfers that target all children or all persons aged over 60 years. Theoretically, these ‘demogrants’ should reduce child poverty and guarantee income security in old age; targeting costs are low because children and older persons are relatively easy to identify; and social costs are low because these are uncontroversial ‘vulnerable groups’ and no distinction is drawn between the poor and the non-poor. But these are not universal programmes, and demographic categories are not accurate proxies for poverty.
The case against targeting
Both the ideological and pragmatic arguments for targeting have been challenged by those who favour universal programmes over targeted transfers.
Ideological objections to targeting
By definition, targeting treats people differently. Targeted programmes divide a community into two groups: those who are eligible to receive benefits and those who are excluded from benefits. Critics point out that targeting the poor or the poorest is based on an essentially arbitrary poverty line, wherever it is set, and that there is often a negligible difference in terms of income and assets between the least poor eligible person and the poorest non-eligible person. Objections to poverty targeting also focus on the distinction drawn between the ‘deserving versus undeserving poor’, which Sepúlveda (2014) describes as a ‘poisonous paternalistic paradigm’: ‘the idea of reaching only the “deserving poor” continues to underpin many social protection programmes … From a rights perspective there are no “undeserving poor”’ (Sepúlveda, 2014: 7, 8).
Critics also argue that this segregation of the population stigmatises the poor: ‘policies that emphasize a distinction between the poor and the non-poor are more prone to carry stigma, social unacceptability and … contradict principles of solidarity’ (Ulriksen and Plagerson, 2014: 755, 760). One recipient of the Child Support Grant in South Africa said, ‘I don’t want to be called poor. Who wants to be called poor?’ and ‘Those who don’t have the grant look down on those who get the grant, they are saying they are lazy, they don’t want to work …’ (Hochfeld and Plagerson, 2011: 56, 57). But the challenge of targeting is not to draw a moral distinction among poor people; it is to identify the poor in order to alleviate their poverty. The conflation of a positive ambition of identifying the poor to deliver public assistance to them, with a negative outcome that separates out the poorest and stigmatises them, allows poverty targeting to be represented as morally reprehensible.
In most cases, stigma is an unintended consequence, but some interventions deliberately use stigma as a self-targeting device to restrict access to those who genuinely need support. Examples include the following: distributing yellow maize rather than white maize as food aid in cultures where yellow maize is regarded as animal feed; making public works participants perform menial manual labour and paying them with food rations (food-for-work) rather than with monetary wages (cash-for-work); and making beneficiaries queue instead of paying social grants directly into their bank accounts. Mkandawire (2005) describes these as ‘onerous and humiliating procedures often deliberately designed to discourage the “non-deserving”’ (p. 10).
Social and psycho-social costs associated with targeting – stigma, envy, resentment, shame, social exclusion and loss of dignity – are an inherent risk, especially if targeting is designed and implemented insensitively. Singling out some groups for special treatment has the merit of directing social assistance to them, but the disadvantage that they could lose the support of their social networks and their status in their community – not an ethically desirable outcome (Macauslan and Riemenschneider, 2011). On the other hand, older persons often gain status once they receive pensions, as one social pensioner in Lesotho affirmed: ‘Before, we were treated as if we were dead. People now respect me’ (Devereux et al., 2005: 32). Similarly, communities often appreciate external support that alleviates the burden of supporting poor community members. A cash transfer scheme that targeted the poorest 10% in rural Zambia had this positive ‘crowding out’ effect: ‘The community welcomes the scheme because it reduces the overwhelming social burdens and the incidence of begging’ (Devereux et al., 2005: 34).
Targeted social transfer programmes are susceptible to politicisation. Transfers might be directed not on the basis of need but on the basis of political interests – either to reward supporters (targeting the ruling party’s constituencies) or to buy support (pre-election handouts in opposition constituencies) (Devereux and White, 2010). Several countries (including Malawi, Rwanda and Zambia) have introduced social cash transfers in a few districts before scaling them up to national coverage, using geographic targeting. Districts are ranked by poverty incidence based on household budget surveys, and the poorest districts receive cash transfers first. This needs-based approach has been queried by politicians in some of these countries, either because they suspect the districts selected are biased towards opposition parties (which might indeed be the case, if the poorest parts of the country are those neglected and deprived of public spending by the ruling party) or because they are exerting pressure for their preferred politically favoured districts to be included.
Conversely, ‘positive politicisation’ of targeted social grants is equally possible. During Lesotho’s general election campaign of 2007, the main opposition party pledged to increase the Old Age Pension by over 300% if they were elected, which forced the governing party to raise the pension by 33% after they were re-elected with a reduced majority. A post-election survey revealed that many voters voted based on each party’s commitment to the pension, and in 2007, the Ministry of Finance conceded that ‘it would now be politically impossible to stop the Old Age Pension’ (Devereux and White, 2010: 71). In South Africa, a survey of 1200 adults in poor communities in 2013 found that most (76%) disagreed with the statement that ‘Giving social grants to people is a form of bribery so that they support the ruling party’. However, a majority (59%) agreed that they would vote for a political party because it provides social grants (Patel et al., 2014: 35).
Another political risk relates to the limited political support that targeted programmes might receive, in contrast to universal programmes. Amartya Sen once made the point that programmes that benefit only a minority of people, especially the politically marginalised poor, risk losing political support from the majority, especially the taxpaying middle class and elites, who have no direct interest in these programmes: The beneficiaries of thoroughly targeted poverty-alleviation programs are often quite weak politically and may lack the clout to sustain the programs and maintain the quality of the services offered. Benefits meant exclusively for the poor often end up being poor benefits. (Sen, 1995: 14)
Conversely, untargeted universal social programmes – such as food subsidies in the Middle East and the National Health Service in the United Kingdom – have proved extremely politically difficult to erode or remove, because they enjoy the support of the politically powerful middle classes who benefit directly from them. General subsidies that reduce the price of a commodity or service to all consumers or offer social services free of charge at the point of delivery, by definition do not exclude anyone. On the other hand, food subsidies tend to be captured disproportionately by the non-poor, who can afford to consume more of these essential commodities. In this case, applying the ‘equality’ principle leads to unequal outcomes – widening inequality. One argument for phasing out general subsidies and replacing them with targeted cash transfers is that judicious spending of the ‘subsidy dividend’ should achieve more pro-poor outcomes.
By contrast, poverty-targeted transfers will inevitably overlook some poor people, in design or implementation, and the benefits delivered might be inadequate or might erode over time due to limited or dwindling political enthusiasm. For this reason, programmes targeting the poor often need advocates with the ability and voice to influence those with decision-making power – such as non-governmental organisations (NGOs) and civil society – to campaign for them.
At a smaller scale, it has often been observed that programmes delivered to rural communities in Africa or South Asia perform better if local leaders also benefit directly from them because these local elites are also the gatekeepers to the community, and the smooth implementation of external interventions depends on their support. Indeed, even in participatory wealth-ranking exercises, the community sometimes ‘votes’ to include the village chief or headman among those who should receive benefits, because ‘even the chief must eat’. Again, this makes the case for allowing some inclusion error or even perhaps a geographically- rather than poverty-targeted programme.
At the political economy level, critics of targeting assert that it reinforces a neoliberal social policy agenda, whereby the poor receive palliative support only at the benevolence of the better-off, while leaving the structural determinants of poverty and inequality unchallenged (Hickey, 2014). This is an important concern, but instead of undermining the argument for targeted transfers it suggests a need for political activism at the macro-level to complement pro-poor social programmes at the micro-level.
Pragmatic objections to targeting
Is targeting efficient? Critics point out that identifying eligible individuals within a population group is not costless but could absorb a significant proportion of the budget. The validity of this complaint depends on variables such as targeting costs (which vary by targeting mechanism and the capacity of implementing agencies), targeting errors and the distribution of income in the population. In our hypothetical example above, even if targeting costs 10% of the programme budget, the poorest 5% of people in Targetland would still receive 180 dollars each. A low-cost mechanism such as geographic or categorical targeting would cost less than 1 million dollars but would incur high inclusion and exclusion errors, while a more expensive mechanism such as means testing could cost more than 1 million dollars if done well, but it would probably be more accurate. This trade-off between targeting errors and targeting costs is unavoidable: the level of ‘acceptable error’ is a political choice.
Concerns about the difficulties of accurately identifying those in need of redistributive transfers were raised in the early 1990s by Giovanni Andrea Cornia and Frances Stewart, who identified ‘two errors of targeting’ (Cornia and Stewart, 1993). The first error is ‘inclusion’, a financial inefficiency caused by giving transfers to non-poor people who do not need this assistance. The second error is ‘exclusion’, a humanitarian cost with ethical consequences, caused by failing to identify someone as needing assistance and/or failing to deliver assistance to them. Although the two errors cannot be strictly compared, Cornia and Stewart proposed weighting exclusion error three times higher than inclusion error.
Two specific pragmatic objections that undermine the claims of targeting to be efficient and equitable are administrative costs and incentive effects.
Targeting the poor requires knowing who the poor are, so they are included in the social transfer programme, and knowing who the non-poor are, so they can be excluded from the programme. Obtaining this information is expensive. As a rule of thumb, the more accurate the targeting, the more information is needed and the more expensive it is to collect. Besley and Kanbur (1990) asserted that the costs of targeting rise asymptotically as targeting accuracy rises. In effect, there is an unavoidable trade-off between accuracy and cost.
Crude mechanisms such as geographic targeting (blanket coverage of poor or disaster-affected districts) or categorical targeting (‘vulnerable groups’ such as older persons or persons with disability) are cheap to administer because they require relatively little information, but they are susceptible to high inclusion and exclusion errors – not everyone aged over 60 years is poor, while many people aged under 60 years are poor. So these mechanisms tend to be cheap but inaccurate. Conversely, more rigorous mechanisms such as means testing (detailed assessment of each claimant’s income, assets, savings and support networks) are costly to administer because they require a lot of data that must be verified and validated (claimants have obvious incentives to under-report) and updated regularly (claimants’ circumstances can change – a recipient of unemployment benefits could find a job). So these mechanisms can be accurate (if done well), but they are expensive.
The poverty context matters. If poverty in a country or community is widespread, then universal programmes make more sense than spending a high proportion of the budget on identifying the relatively small number of non-poor people who should not receive benefits. If the poverty headcount is 80%, for example, then rather than spending, say, 30% of the budget to exclude the one in five non-poor people, a universal programme should be implemented and 20% inclusion error should be accepted. This might also have the advantage of ameliorating some of the social and political costs of targeting.
Delivering free social transfers to some members of a community, but not to everyone, introduces incentives to access these benefits. One immediate effect is that people might under-report their true wealth status if they are aware that eligibility for receiving free transfers depends on being poor or convincing administrators that one is poor. This is what makes means testing expensive – information provided by claimants has to be validated – and motivates the use of alternatives such as proxy means testing, which does not require detailed information about individual income and assets.
Even worse, targeting can introduce perverse behavioural incentives. For instance, a public works programme in Argentina that targeted impoverished households with dependents resulted in children being ‘shared’ across households to meet the demographic eligibility criteria (Slater and Farrington, 2009). In South Africa, popular wisdom asserts that the Child Support Grant causes increased teenage pregnancy, but an empirical analysis found ‘no grounds to believe that young South African girls are deliberately having children in order to access welfare benefits’ (Makiwane and Udjo, 2006: 2). On the other hand, there is qualitative evidence that ‘some people with AIDS may intentionally “yo-yo” on and off treatment in order to maintain access to disability grants’ in South Africa (Simchowitz, 2004: 13).
These challenges to targeting might seem insurmountable, but before examining the case for and against universal programmes, it should be noted that empirical reviews of targeted programmes have found that they do benefit the poor, on average (as noted above), and they can be relatively accurate and cost-effective. However, often targeting is done badly, leading to high inclusion and/or exclusion errors and high social costs. Coady et al. (2004) found that effective implementation is critical. Targeting accuracy is positively correlated with implementation capacity (proxied by national income), government accountability, levels of inequality and the use of multiple targeting methods.
The case for universalism
Accurate targeting on the basis of need is so difficult and expensive that it raises questions about whether it is worthwhile doing it at all. The ethical principle ‘leave no-one behind’ places a premium on not excluding anyone from assistance who needs it. A strict adherence to this principle would imply that, if there is a non-zero possibility of failing to reach even one person who needs social transfers, targeting should be abandoned. The only way to guarantee reaching every person in need is with an untargeted universal intervention. Universal programmes, it is argued, are inclusive and non-discriminatory, they do not stigmatise or patronise the poor. The pragmatic consideration, that social budgets are constrained, is dismissed by proponents of universalism as spurious: there are no binding budget constraints, only political choices. Advocacy is simply needed to campaign for increased allocations of public resources to social programmes: ‘more encompassing social protection policies are more likely to build a stronger sense of solidarity and to be more politically sustainable’ (Ulriksen and Plagerson, 2014: 755).
So the policy implication of targeting scepticism is not to target at all. This argument was taken to its logical conclusion by Mkandawire (2005), and this line of thinking also led to a movement for a ‘basic income grant’ in many countries (Standing, n.d.). Civil society activists in South Africa have consistently lobbied the government to introduce a universal, non–means tested BIG, which would be recovered from wealthier South Africans through a ‘solidarity tax’. However, ‘the Department of Finance stonewalled the proposed BIG on account of financial sustainability considerations’ (Barchiesi, 2007: 569), and it has never been implemented. In 2008, the Namibian Basic Income Grant Coalition implemented the world’s first universal cash transfer pilot project, disbursing N$100 per month (about US$14) for 2 years to all 930 residents in one poor peri-urban community. Positive impacts were recorded across a range of outcomes, including community mobilisation, women’s empowerment, child nutrition status, (self-)employment, local economic growth, school attendance, access to health services – even crime rates fell (Haarmann et al., 2008). Despite this evidence of success and estimates that the BIG is affordable – ‘a national, universal BIG of N$ 200 per person per month will be about 2–3% of GDP annually … equivalent to 5–6% of the national budget’ (BIG Coalition, 2015: 2) – it has not been implemented at scale by the government of Namibia.
At the political level, targeting is often derided by its critics as neoliberal and socially divisive, whereas universalism is reified as rights-based and upholding social solidarity. In Europe, ‘the principle of universality in social protection’ is upheld as fundamental ‘to preserving and fostering social cohesion’ (Council of Europe, 2004: 125). At the personal level, this implies that better-off taxpayers will only feel a sense of social solidarity if their government gives them money that they do not need – but many middle-class citizens derive a sense of solidarity from paying taxes that are redistributed to the poor as social grants. Studies of subjective well-being have found that inequality aversion is high in Western Europe (Delhey and Dragolov, 2013), where the welfare state originated as a mechanism for supporting the poor through redistributive transfers targeted on the basis of need.
These polarised positions will be exacerbated if public attitudes towards claimants are negative – for instance, if there are widespread perceptions that social grants are misused (spent on alcohol, etc.), create dependency (make people lazy) or cause perverse incentives (e.g. that child grants increase teenage pregnancies). It is unfortunate but true that a well-functioning social programme can be undermined by a few anecdotes of ‘welfare cheats’ in the media. Political support for poverty-targeted programmes is likely to dissipate if beneficiaries are seen as ‘undeserving’.
At the pragmatic level, the argument that universal programmes are not redistributive and only reduce relative inequality has been challenged by the argument that this may be true at the first-round level, but second-round effects through the fiscal system are redistributive in absolute terms. Since those who pay for social transfer programmes tend to be taxpayers while the beneficiaries tend to be non-payers of income tax, it follows that some degree of redistribution is inherent even to universal programmes – some of the income transferred to the non-poor will be recovered in tax. As an extreme example, Hoddinott (2007) mentions a programme in Canada that taxed a universal transfer at 100% for individuals above a certain income level, thereby ensuring that a universal programme in design and first-round implementation became a means-tested programme in intention and second-round incidence.
The case against universalism
If targeting is costly, universalism is even more expensive. Universal programmes are cost-ineffective because they do not allocate public resources efficiently to achieve the greatest impact. Even if the administrative costs of targeting the poor absorb 15% of a programme’s budget, opting for a universal social transfer instead will be cost-effective only if more than 85% of the population is poor. As Skocpol (1991) pointed out, ‘universal programs, whether benefits or services, give most of the available resources to those who need them least’ (p. 413).
Surprisingly few social programmes are genuinely universal in their incidence – three examples are a basic income grant, a general food subsidy and a health service that is free for all at the point of delivery. Even fee-free education is restricted to school-going children.
The ‘social protection floor’ (ILO, 2012) sounds like a set of universal interventions. It advocates rights-based, guaranteed access to health care and income security for all. However, income security for all is not necessarily the same as free income transfers for all. In June 2015, as noted above, the World Bank and ILO called on world leaders to adopt ‘universal social protection’, which they define as ‘adequate cash transfers for all who need them, especially children; benefits and support for people of working age in case of maternity, disability, work injury or for those without jobs; and pensions for all older persons’ (ILO, 2015).
The Universal Social Protection Initiative sounds like free cash for all – but it is not: ‘cash transfers for all who need them’ implies targeting the poor, not a BIG for everybody. Many countries in the Middle East implement universal food subsidies, which ensure 100% coverage but give wealthier people access to more food than they need but poorer people access to less food than they need. Subsidy reform processes now underway across the region, strongly supported by the World Bank, involve abolishing universal subsidies and replacing them with targeted cash transfers. Is this compatible with universal social protection? The World Bank has also successfully promoted the adoption of conditional cash transfer programmes throughout Latin America and beyond, whereby social protection is withheld from families who fail to comply with conditions on their behaviour. But how are conditional cash transfers compatible with ‘cash transfers for all who need them’?
The Social Protection Floor aims to ensure universal coverage, but not through providing universal social grants. Wealthy people who can afford private insurance do not need social assistance. Salaried workers who have employment-related social security do not need social assistance. Only those who are poor and do not have access to social security or private insurance should qualify for social assistance under a national Social Protection Floor, which requires identifying and targeting the poor. A right to social protection does not imply a right to free cash transfers from the state; rights can be realised through targeted rather than universal programmes.
Conclusion
A good society is one where, inter alia, everyone enjoys access to essential services and enough income to live in decency and with dignity. The only way that universal access to essential services can be guaranteed is by providing these free of charge at the point of delivery, while recognising that most users will be contributing towards the cost of these services through various taxes collected by the state. Income security can be guaranteed by giving money to people who need it – it is illogical and wasteful for the state to give a living allowance to every person in the country. So the two arguments in favour of targeting – redistributive equity and efficient allocation of social budgets – remain persuasive, at least for social welfare, if not for social services.
The pragmatic arguments against targeting – that targeting is so expensive and inaccurate that it can be counter-productive – are real challenges, but do not invalidate the equity and efficiency arguments in favour of targeting. The costs can be managed, and errors can be reduced. Administrative costs can be contained by selecting a targeting mechanism that is reasonably cheap and setting an acceptable level of inclusion error. A purist ‘leave no-one behind’ view implies zero exclusion error, which is an insurmountable challenge without universal programmes. But is it better to give everybody so little that it makes no real difference to poverty, in the interests of reaching everyone, or to give more to the poor so some can escape poverty and the majority are made less poor? Social costs of targeting can also be managed through sensitive design and implementation. For example, geographical targeting has been found to be more effective than targeting older persons or children and is less socially divisive than means tests or proxy means tests.
Two sources of confusion have contributed to the polarisation of the debate over targeted versus universal interventions. One is lack of clarity over what is meant by universal, and the other is lack of clarity over which dimensions of social policy are being discussed. Programmes that target children or older persons are not ‘universal’, they are applications of categorical targeting. And while there is a strong ethical case for providing universal free health care, it is not appropriate for the state to hand out universal free cash transfers.
In conclusion, therefore, this article argues for a ‘trifurcated’ approach to targeting social programmes, with the equality, equity and need principles being applied to social services, social insurance and social assistance, respectively.
Equality principle – for essential
Equity principle – for contributory
Need principle –
Footnotes
Funding
The author(s) received no financial support for the research, authorship and/or publication of this article.