By Shahra Razavi, Christina Behrendt, Valeria Nesterenko, Ian Orton, Celine Peyron Bista, Alvaro Ramos Chavez, Helmut Schwarzer, Maya Stern-Plaza and Veronika Wodsak | International Labour Organization (ILO)

 

The consensus on the need to build universal social protection systems to provide income security and health protection for all has been reinforced by the COVID-19 pandemic.  The crisis that shook the world over the past two years not only revealed large gaps in coverage, adequacy and comprehensiveness of social protection systems but also drove the message home that a universal social protection system reaching everyone is automatically primed to protect all those affected by a systemic shock.  In the face of complex and fast-moving crises, universalism is preferable to targeted approaches, especially where the administrative capacity to target is limited and a very high proportion of the population is vulnerable. Universalism makes more practical sense than ad hoc efforts to “effectively” target, the limitations of which are well documented.  

In the context of our shared commitment to universal social protection, exemplified by the Global Partnership for Universal Social Protection (USP2030), and given the growing urgency and commitment on the part of countries to build universal social protection systems, it is of utmost importance that there is clarity in international policy guidance. This would enable national policymakers to decipher where there is consensus on key policy issues, where differences may remain, and where we may simply not know enough. The recently released World Bank publication, Revisiting Targeting, provides an excellent opportunity to explore some of the key policy issues. This comes at a time when the COVID-19 pandemic and other unfolding crises have made the case for universal social protection more compelling than ever before, putting the spotlight on the need to identify financing options for extending social protection, and in so doing help re-build the social contract.

 

1. The need for conceptual clarity around the concept of targeting

 

While there are different ways of determining eligibility to provide differentiated benefits for different groups, the term “targeting” is usually used when referring to poverty targeting and is rarely used for other eligibility criteria, such as age or employment status. To do justice to the complex issues associated with poverty targeting, it is essential to treat it separately from other eligibility conditions. However, Revisiting Targeting blurs these distinctions.

Where eligibility for social protection schemes is differentiated based on age, employment status, or other characteristics, for example in the case of schemes for children (e.g. child benefits), older persons (e.g. pensions) or unemployed persons (e.g. unemployment benefits), this is not usually referred to as “targeting”.  Such eligibility conditions ensure that benefits are received by those who need them, in line with international social security standards. Such a differentiated approach ensures that everyone has access to comprehensive, adequate, and sustainable protection over the life cycle if and when they need it, for example when they are sick, unable to find work, or elderly. Such a differentiated approach is one way of achieving universal coverage, ensuring that everyone is protected throughout their life course, which does not require that everyone receive an equal benefit at every point in time (as would be the case with a Universal Basic Income) (ILO 2019). Such a differentiated provision also has other advantages compared to poverty-targeted provision. It significantly reduces exclusion errors and avoids stigma, and includes the middle class, thereby generating broader political support for social protection.

The more common understanding of “targeting” relates to poverty targeting, referring to a mechanism to identify those living in poverty who should be eligible for social assistance schemes and programmes.  Typically, a means test is used to assess whether the individual’s or household’s own resources (income and/or assets) are below a defined threshold (usually set out in national legislation), to determine whether the applicants are eligible for a benefit at all, and if so at what level that benefit will be provided. However, in some countries, proxy means tests (PMTs) are used; that is, eligibility is determined without assessing income or assets, and on the basis of other household characteristics deemed more easily observable, used as proxies, such as household composition, type of housing, productive assets, or level of education of household members. Other forms of poverty targeting, such as geographic targeting to channel benefits into areas with high levels of poverty, or community-based targeting, are also common in some countries, sometimes in combination with PMT mechanisms. Poverty-targeted schemes may also include entitlement conditions and obligations, such as work requirements, participation in health check-ups, or (for children) school attendance.

Research shows that targeted schemes that seek to reach the poorest, produce huge exclusions: out of 25 programmes or registries with coverage under 25 per cent, 12 had exclusion errors above 70 per cent and 5 had errors above 90 per cent (Kidd and Athias 2020).  There is also a strong relationship between higher coverage and lower exclusion of intended recipients, with universal schemes performing well and having exclusion errors that are below 10 per cent (ibid.).

At the same time, excessive targeting tends to increase the complexity of procedures, creating additional barriers and hurdles that beneficiaries have to navigate, and “transforms social workers and administrators into gatekeepers of the system, tasked with avoiding fraud by ‘undeserving’ applicants” (de Schutter 2022). Moving from strictly targeted, means-tested benefits to categorical or universal schemes “reduces the shame involved in having to prove that one is sufficiently poor’’ and can help address the problem of non-take-up of benefits by those who need them most (de Schutter 2022).

 

2. Targeting within a broader universal system

 

Targeted or means-tested social assistance schemes are usually one component of a more comprehensive social protection system: in well-developed social protection systems, they cover a small proportion of the population who are not sufficiently captured by non-means-tested schemes, and they are usually anchored in national legislation. Their relevance/usefulness is determined in each specific context against the problem they are trying to address.

It would be misleading to assume that there was no role for targeting within a broader universal system. However, its role needs to be carefully defined based on three main considerations.

First, targeted/means-tested social assistance schemes are part of a broader social protection system, which usually combines tax-financed schemes and social insurance to guarantee a social protection floor and provide higher-level benefits in line with international social security principles. This consideration is acknowledged in Essay 2 of Revisiting Targeting. However, the report tends to present the “system” as the juxtaposition of multiple means-tested tax-financed social assistance programmes and contributory social insurance schemes, aimed at different groups and based on differentiated needs.  Tax-financed and contributory schemes, however, are not mutually exclusive options, and a systems approach means combining both financing methods to create a multi-tiered social security system with a view to achieving universal life-cycle guarantees (Barrantes and McClanahan 2021).  Such a multi-tiered social security system allows the continuity of individuals’ social protection coverage even as their socio-economic situation changes through the option of subsidizing social insurance coverage during periods when incomes are low. This systemic approach also recognizes that everyone is vulnerable to shocks along the life cycle. Rather than intervening only after households have fallen into poverty, such comprehensive systems aim at preventing poverty in the first place and are thus better placed to respond to the dynamic nature of poverty. In that sense, both tax-financed programmes and social insurance should be fully embedded in national social protection policies or strategies, as well as the national legal framework, avoiding the fragmented delivery of social protection.  In this way, the implementation of means-tested/targeted programmes would be connected to policy deliberations on the extension of social security coverage.

Second, the policy question is not just about how to implement targeting, but also if, when, and where targeting is appropriate depending on the specificities of the national context. In countries where more than half of the population is considered to be income-poor, the merits of targeting a small share of the population is highly questionable (Brown, Ravallion, and van de Walle 2018). It may even put social cohesion at risk and increase mistrust of the State. In a context with a high prevalence of poverty, often coupled with no or limited availability of data and institutional capacity, even accurate and precise means-tested targeting, including with sufficient efforts and resources invested in implementing the selection criteria, will leave behind most of the poor and near-poor population who need social protection, not to mention others who are also vulnerable to both life-cycle risks and co-variate shocks. The idea of resorting to a public lottery as a “complementary tool for supporting the final selection of beneficiaries when there are more eligible people than can be covered” (Revisiting Targeting, p.273) is spurious and unsatisfactory.

Finally, it is encouraging that Revisiting Targeting acknowledges the relevance of human rights, pointing to the fact that the vast majority of countries are bound by international human rights treaties to recognize and ensure the realization of the human right to social security.  In fact, not only do most countries and United Nations bodies have an obligation to mainstream human rights, but the right to social security should also be taken into account to ensure that policies and practices of international financial institutions promote, not compromise, the right to social security (UN 2008, para. 83).

Revisiting Targeting rightly notes that “human rights standards are not compromised by the use of targeted schemes as a form of prioritization of the most vulnerable and disadvantaged groups”.  Indeed, international standards have considered targeting as one of the means for achieving universal social protection for nearly a century, in so far as they conform to the essential elements of the right to social security and contribute to its realization (UN 2008, para. 5). As far back as 1944, the standards developed by the ILO acknowledge the necessary coexistence of social assistance schemes for persons not (sufficiently) covered by compulsory social insurance (Income Security Recommendation, 1944 (No. 67) and Social Security (Minimum Standards) Convention, 1952 (No. 102)), and their important role is also acknowledged by the Social Protection Floors Recommendation, 2012 (No. 202).

There is an important difference, however, between a system where means-tested benefits constitute the main strategy and a system where means-tested schemes play a secondary, residual role (as is the case in many high-income countries), designed to support individuals who, for some reason, are not reached by non-means-tested schemes (Kidd, 2022). In countries where universality is the norm, universal life cycle benefits constitute the lion’s share of social security spending, alongside some form of small-scale residual poor relief (ibid). Moreover, targeted schemes that are part of universal social protection systems are rights-based, that is, eligibility criteria, benefit levels and modalities are set out in the national legislation, which ensures transparency and accountability. This also implies that those who are eligible for benefits have a right to receive the benefits when needed (and not end up on waiting lists) – it is the government’s responsibility to make available the necessary financial resources.

 

3. Fiscal space expansion to make the right to social security a reality for all

 

The human rights perspective is not blind to the reality of financial constraints weighing on States in their quest for universal social protection. It does not, however, take at face value that the available resources are static.

Countries can use targeting mechanisms to make full use of the maximum available resources, while at the same time taking deliberate steps to progressively extend these resources (UN 2008). The available resources should be used, first and foremost, to provide minimum essential level of benefits to all individuals and families to secure effective access to basic goods and services (UN 2008)—in other words, a social protection floor in line with ILO Recommendation No. 202. In so doing, the Recommendation implicitly calls for the establishment of a national social protection expenditure floor which “would represent the core of social protection spending that should be guaranteed by the State at all times, and therefore protect against fiscal and economic austerity measures” (ILO 2019a).  If financial constraints impede States’ ability to provide the minimum essential level of benefits as a matter of priority, despite a careful appraisal of domestic and international resources, they can resort to selecting a group of social risks and contingencies, after a wide process of consultation (UN 2008). While priority can be given to individuals belonging to the most disadvantaged and marginalized groups, States should not lose track of their responsibility to progressively, but “as soon as possible”, reach the full implementation of the right to social security for everyone, by establishing long-term national strategies with clear targets and timeframes based on transparent and equitable budgeting and adequately funded long-term institutional frameworks (UN 2008, Social Protection Floors Recommendation, 2012 (No. 202), ILO 2019a).  

It is useful to recall that the shift to targeting in the 1980s was in part driven by macroeconomic policies of “fiscal constraint” which authorized the view that targeting was the most efficient and commonsensical thing to do under the circumstances.  As Mkandawire observed, ‘’Politically, it is much more convenient to deploy the language of cost containment and efficiency that comes along with budgetary constraints than to embark on a frontal attack on the legitimacy of universalism and its morally appealing language of rights and solidarity’’ (2005, p. 5). Revisiting Targeting notes that the COVID-19-related social protection policy responses entailing the expansion of coverage, benefit levels or both, ‘’may have heightened expectations for social protection going forward’’ (Grosh et al, p. 9). However, it warns that since many countries are likely to face ‘’increased deficits and debt loads … budget space to support larger social protection and meet the heightened expectations may not be easy to find’’ (ibid). 

But universal social protection systems are not built overnight. They require long-term planning and are built progressively, starting for example with employment injury protection or a universal social pension and gradually making the system more comprehensive by adding protection for other life cycle risks.  By building a universal social protection system, progressively covering everyone in the event of a life cycle risk or a co-variate shock, governments can avoid the significant exclusions that are rampant in targeted programmes. This is the true meaning of what constitutes a truly shock-responsive social protection system and how it should be built.

The focus on targeting and delivering social assistance to a specified population below a given income level, recalls the early focus of Public Finance Management (PFM) on budgeting and the budget cycle. In the context of donor funding, where the fiscal envelope, that is, the level of resources that will be available, is fixed, it makes sense to emphasize budgeting skills and delivery capacity. However, more modern PFM now acknowledges the need to build state capacity in identifying funding sources, estimating revenue streams and their impact on economic activity so that budgets no longer start from the assumption of nearly fixed revenue levels (Gaspar et al. 2019).  Indeed, the key policy question for many developing countries is not just how to allocate or budget a given/fixed resource envelope, but also how to expand the fiscal space for social protection (Ortiz et al. 2019). This is the crux of the challenge as the coverage gaps in social protection systems—leaving 53 percent of the global population without any form of social protection—are associated with significant underinvestment, especially in low-income countries (ILO 2021).  Therefore, countries need support in the permanent struggle to expand the fiscal space for social protection and to do so in ways that are equitable and do not impose an undue burden on those with limited means.

This implies that countries need to worry about expanding both contributory capacities and the tax base. Both are closely intertwined with labour productivity and the formalization of enterprises and employment. Expanding general government revenue through taxation, preferably with innovative, progressive wealth and corporate taxes, is essential in the context of the very high levels of inequality found in developing countries (Chancel et al. 2022).  It is also important to consider other sources of additional revenues for social protection, such as the curbing of illicit financial flows and better management of debt. Building the capacities of tax and social security authorities, therefore, matters hugely, as governments cannot only rely on better delivery (see further below on state capacity).

One of the advantages of universal social protection systems is that it creates a virtuous cycle of revenue expansion/creation that is stronger than what narrowly targeted social assistance programmes alone can garner. For starters, additional revenues typically must be discussed in the context of national dialogues and ‘resource bargains’ (Hujo and Bangura 2020). As these revenues imply economic effort on the part of the stakeholders, they are also more likely to demand capacity building to deliver the services. The ability to deliver increases credibility and the willingness to pay on the part of stakeholders (on which see below). Furthermore, the fiscal space for universal social protection will likely contribute to greater and more inclusive economic growth, which will mean that government revenues will grow, even if levels of taxation and social security contributions remain constant. While this would not fully cover the cost of universal schemes, it will make a significant contribution towards realising/building such schemes. A study by the ITUC and Development Pathways (2021) across 8 countries has indicated that an investment in social security of one percent could result in cumulative increases in tax revenues over 10 years of between 2.1 and 10.4 percent.

 

4. The political economy of universalism

 

Compared to poverty-targeted schemes, universal social protection schemes receive more public funding, offer higher value transfers to recipients, enjoy higher quality implementation, with fewer people living in poverty being excluded. The political economy argument for universalism goes beyond the median voter theory and points to the level of trust in public institutions, different perceptions of who are “deserving” population groups and the overall levels of social cohesion.

There is a wide body of literature analysing whether welfare regimes that target benefits to people with low incomes produce better redistributive outcomes than welfare regimes grounded in universal approaches. While there is little, if any, evidence that poverty targeting produces more redistributive outcomes, the majority of studies seem to indicate that levels of inequality are lower in countries with universal approaches (Korpi and Palme 2000; Gugushvili and Laenen 2021). One explanation for this outcome is that more universalistic approaches are better able to mobilize support from the general public across all income levels and that as a result, redistributive budgets are larger in countries where universal approaches prevail. For example, Gaspar, Jaramillo, and Wingender (2017, p. 245) emphasize the importance of a sense of fairness: if citizens believe that sufficient public goods are being provided in return for the taxes they pay, and others also pay their fair share, they are more willing to comply.  They point to the importance of constitutive institutions, inclusive politics and credible leadership for the building of tax capacity, which would apply also to building universal social protection systems.

The conclusion of Essay 9 of Revisiting Targeting focuses mostly on median voter theory when arguing that there was little empirical evidence to back the claim that redistribution can be more easily achieved through universal compared to poverty-targeted approaches. However, this is a selective review of the literature. Empirical evidence from a wide range of countries (both institutionalized welfare states and developing countries) provides support for the argument that compared to poverty-targeted schemes, inclusive social security transfers receive more public funding, offer higher value transfers to recipients, and enjoy much higher quality implementation, with fewer people living in poverty being excluded. They are also much more effective in their impacts on poverty and inequality. For example, Nelson (2004)  presents empirical evidence that non-means-tested benefits lifted much larger proportions of people out of poverty.  Gugushvili and Laenen (2021) likewise conclude that overall “universalist countries tend to have better redistributional outcomes”, while adding that “targeting, however, seems to work best when embedded within the overall framework of universalism. The required balance between universalism and selectivism will inevitably vary across countries but in any context, it seems important that selective policies are not too narrowly targeted at the very poor”. 

Overall, the political economy analysis of poverty targeting versus universal (categorical) provision must consider many more variables than voting behaviour and budget allocation for social protection expenditure, including factors that are inherently difficult to measure but important for the overall fabric of society, such as the level of trust in public institutions, different perceptions of who are “deserving” population groups, and the overall levels of social cohesion. Public preferences or support for certain type of scheme design or for a certain type of welfare paradigm may not necessarily be reflected in voting behaviour. As Revisiting Targeting rightly points out, voters cannot vote on individual policies but rather on policy packages as presented by the manifestos of parties or candidates, and they may vote not only on the basis of their enlightened self-interest, but their value-system may also influence their attitude towards re-distribution and questions of targeting.

Rather than limiting politics to the periodic casting of a vote, political economy analysis should take into account the overall social and political legitimacy that public institutions enjoy, often described as the level of “trust” that the population places in the public administration beyond party politics. In light of their greater simplicity and transparency with regard to eligibility criteria, categorical benefit schemes, for example, perform better in this regard than poverty-targeted programmes that select beneficiary households on the basis of PMT scores which are often difficult to understand. To build trust and ensure legitimacy, it is not only the factual outcomes (levels of poverty or inequality) that matter, but also the processes through which these are achieved. The principles of reliability, transparency, accountability and predictability enshrined in the Social Protection Floors Recommendation No. 202 are important not just because they allow beneficiaries to regain control over their own lives but also because they increase trust in public institutions and strengthen the social contract between citizens and the state.

Where Revisiting Targeting discusses public opinion in more general terms, beyond the casting of votes, it looks into attitudes towards redistribution but not poverty targeting of social protection programmes per se. However, public support for redistribution should not be equated with support for poverty-targeted benefits. People may well be in favour of more redistribution but not through poverty-targeted benefits. Many other factors impact redistribution and public attitudes towards it, including progressive taxes, levels of poverty and inequality before taxes and transfers, the demographic structure, labour force participation rates, unemployment levels, wage policies, and the overall structure of the economy.

 

5. Building high-capacity states: Beyond the delivery of social protection

 

Deficits in state capacity need to be addressed in a more concerted fashion that goes beyond a focus on delivery systems alone. It is equally urgent to build state capacity to regulate labour markets, formalize employment and enterprises, build effective and well-governed social insurance systems, and strengthen the capacity of tax authorities to collect taxes.

Historically, crises have prompted the progressive reconfiguration of the role of the State and existing social arrangements. Arguably, the COVID-19 pandemic has reshaped thinking about the role of the state and reasserted the recognition of the need for strengthened state capacity. The stress test applied by COVID‑19 has shown that states are not powerless to act in the interest of all their citizens, and massively expanded the scope of policy action that can be taken (Mazzucato 2013; 2021). Only the state could act decisively enough to protect health, income, jobs and enterprises on the scale that was required, and to ensure macroeconomic and social stability.

Moreover, some States acted with determination to assert their authority over practices that were considered not to be in the public interest. For example, Denmark barred companies operating in tax havens from access to employment retention benefits (Australia Institute, Nordic Policy Centre 2020). Similarly, the war in Ukraine has shown the power of states, especially when they act in concert, to combat illicit finance and to accommodate millions of displaced persons into their social protection systems in the near blink of an eye.

The pandemic has shown that, when prompted into action, States are not without choices, and that they have both the potential and the requisite instruments to combat major challenges. However, it also showed that there are major deficits in state capacity in too many countries with respect to states’ ability to protect their citizens from the adverse impacts of the pandemic (ILO 2021).

Widespread deficits in state capacity in many countries require a more concerted strengthening of state capacities beyond delivery systems alone. The extensive consideration of how to ensure effective social protection delivery systems (chapter 4 of Revisiting Targeting) is useful in its examination of how to ensure entitlement-holders can best access their rights across the delivery chain, especially in the case of the most vulnerable. In this discussion, however, there is still a reluctance to concede that certain targeting methodologies are problematic. For instance, it is asserted that “a great deal of the bad reputation of targeting with respect to human rights is earned through insufficient delivery systems rather than inherent in the process of eligibility determination” (Grosh et al, p. 206).

While there is clear merit in ensuring an effective delivery chain, a singular focus on delivery obscures and sidelines pivotal questions about the capacity of states to effectively regulate labour markets, formalize employment and enterprises, build well-governed social insurance systems and administrations, put in place tax policies and strengthen the capacity of tax authorities to actually collect taxes to be able to finance social protection and public services. The capacity of states to mobilize resources improves policy space and the ability to set agendas and influence employment and social protection strategies to achieve their development objectives. The analysis in Revisiting Targeting does allude to the need to build basic state capacity when it comes to official and effective citizen ID systems, which are weak in many countries. Nevertheless, the focus on delivery and IDs, though welcome, tends to consider social protection in isolation from the wider state architecture, or lack thereof, which needs strengthening

Building up state capacity will ensure that the State can effectively fulfil its role and realise its primary responsibility for establishing the legal and administrative architecture and sustainable financing of social security, and be the final guarantor of its proper administration and good governance. This would ensure the State is answerable to citizens as social protection rights-holders and would help forge a reinvigorated social contract.

 

6. The use of new technology for better targeting: Applying machine learning algorithms to save PMT models?

 

Digital technology can certainly help to improve the administration and delivery of social protection systems, yet it needs to be applied with caution.

Chapters 6 and 8 of Revisiting Targeting provide an overview of the evolution of targeting methods, as improvements in data availability and technology have enabled more accurate modelling for targeting, potentially reducing both inclusion and exclusion errors as well as human error. Along with the progress made, however, the authors acknowledge some of the major drawbacks of targeting methodologies that remain, especially when it comes to PMTs, including significant errors, considerable data requirements, lack of transparency and opacity of methods, reproduction of bias and discrimination, and difficult trade-offs between targeting accuracy and administrative costs.

In addition to these weaknesses, it is worth noting the high costs of establishing and maintaining social registries required for PMTs, especially in countries where data collection tends to be irregular. Household surveys for PMTs must be as up-to-date as possible and use large sample size, including the majority of the households in the country. However, no country has yet managed to achieve full coverage of households within its social registries, which undermines the argument that social registries are “inclusion” systems. Additionally, social registries can only be up-to-date if governments conduct frequent surveys with large sample sizes. It is estimated that the cost of surveying all households could be as high as US$215 million to US$450 million in a large country like Nigeria (Kidd, Athias, and Mohamud 2021). Whether this represents value for money is a moot point.

Given the high financial costs of social registries, it is understandable that countries are reluctant to undertake social registry surveys frequently, and the information used for PMTs often remains out of date, thereby creating a perception of unfairness that feeds mistrust towards the government. Therefore, those who promote PMTs try to use technology to ease the task of governments, improve the accuracy of the data, and ideally, reduce the costs of targeting.

Chapters 6 and 8 of Revisiting Targeting consider the potential improvements in targeting methods, including of PMTs, through two recent IT-related developments: Big data and machine learning. The Novissi cash transfer programme in Togo (Grosh et al, p. 422-424) is presented as a case in point, where big data was used to accelerate the social protection response and to reduce the administrative costs in the challenging context of the COVID-19 pandemic.

Despite this positive example, the chapters do not see the new technologies as a “panacea” (Grosh et al, p. 349). Citing the results of the biggest comparative test for the use of machine learning algorithms to improve PMT targeting, they show that improvements were indeed very modest, at most around 5 per cent improvement in accuracy. More importantly, the Revisiting Targeting clarifies that the application of machine learning and big data do not solve the basic concerns about PMT methodology, and may in fact accentuate them: lack of transparency, opacity of methods to the  general public and political constituencies, high levels of statistical error, reproduction of real world biases and discriminations, added risks of data breaches, violations of privacy rights, surveillance and manipulation by States, hackers and enterprises, among others (Ibid. p. 417, 429-431). Though the authors admit that there still are “some technical questions” to answer, as well as a need to undertake more research and gain greater experience, the conclusions of chapter 6 read as if policymakers have no other choice than to continue with targeting, opting perhaps among the different methods of targeting and expecting some improvements in methods.

We see, however, a number of risks and threats that the adoption of new technologies may entail. First, if the inclusion of new IT technologies aims to accelerate data up-dates used for targeting and reduce inclusion errors (i.e. hasten the exclusion of persons from the programme who are ‘’incorrectly’’ included), this can have quite unsavoury outcomes, keeping in mind that movements of individuals and households above or below the poverty line can oscillate wildly, and significant numbers may hover just above the threshold. And those who may be lifted out of poverty are likely to remain very vulnerable and can easily fall into poverty again.  In fact, data from Indonesia and Uganda shows considerable movement of households across the 2nd, 3rd and 4th quintiles because, in most countries, there is little difference in the relative consumption (or income) of households below the wealthiest 10 per cent of the population (Kidd and Athias 2020).

It is also necessary to ensure that PMTs that resort to machine learning methods and/or big data do not exclude from the programmes those whose situation improves thanks to the social benefits they receive, and who will likely fall back into poverty if those benefits are withdrawn. These are all arguments for searching for a broad-based approach, rather than sharpening the “precision” and “performance” of algorithms that aim to exclude beneficiaries.

The chapters do not say much about the monetary costs of targeting, which can be considerable. The new IT technologies are very data-intensive, the cost of obtaining data has not been resolved (since big data depends on information generated from private companies, especially tech giants), and the IT capabilities and resources required may be expensive too. There are also legal questions that remain unresolved. Not only is there a need for a legal framework to access data for the new IT applications, but issues of privacy, human rights standards, and constitutional principles are yet to be addressed. In fact, these topics are more contentious today than ever before, especially if artificial intelligence (in machine learning processes - coupled or not with big data) is to make decisions on behalf of public servants on public policies affecting the standard of living for some of the most vulnerable and disadvantaged groups who are entitled to social assistance (UN 2019).

 

7. Focusing on the overarching objective: Building universal social protection systems

 

The discussion around targeting, and the objectives and design of social assistance schemes is very timely, as these constitute a key component of most social protection systems. Further improving them is critical for realizing universal social protection and achieving the Sustainable Development Goals.

While Revisiting Targeting is focused on a specific policy tool for implementing social protection benefits, a discussion around the objectives and design of social assistance schemes is crucial from a human rights perspective. The main objective of social protection policies and programmes should ultimately be to realize the human right to social security. In practice, this requires that the system, entitlements, and eligibility criteria are established under domestic law, and public authorities take responsibility for the administration or supervision of the system, including by ensuring that those who are eligible for the benefits have a right to receive them when needed by making the financial resources available and by establishing effective and accessible complaints and redress mechanisms. These are core elements of a rights-based universal social protection system, which guarantee that everyone has access to comprehensive, adequate and sustainable protection over the life cycle, as highlighted by the International Labour Conference in 2021.

 

References

Social Protection Building Blocks: 
  • Programme design
    • Targeting
  • Programme implementation
    • Benefits payment / delivery
    • Informations Systems (MIS, Social Registry, Integrated Registry)
Social Protection Approaches: 
  • Digital social protection
  • Political economy
  • Social protection systems
  • Universal Social Protection
Cross-Cutting Areas: 
  • Emergency response and Disaster Risk Management
  • Human rights
  • Labour market / employment
  • Poverty reduction
Regions: 
  • Global
The views presented here are the author's and not socialprotection.org's