The World Bank’s 'State of Social Protection Report 2025: The 2-Billion Person Challenge' shows that it is not learning its own lessons - Part 2

Part 1 of this blog - available at the Development Pathways website - argued that the World Bank’s long-term advocacy of poverty targeted schemes as the basis for the tax-financed side of social protection systems contradicts messages from its own research. This blog argues that the World Bank’s promotion of social registries also contradicts the organisation’s own research, this time into the COVID-19 pandemic.

Both Part 1 and Part 2 of this blog have been written on behalf of the organisations which are jointly campaigning for the Right to Social Security.

Download pdf version.

The rise of social registries

In the last decade, the World Bank has stepped up its advocacy of poverty targeting as the basis for tax-financed social protection schemes in low- and middle-income countries by promoting social registries. These are large data bases, in theory covering the whole population, which are used to poverty target multiple schemes, not just one. This is usually though not always through proxy means testing. The World Bank argues social registries are core parts of social protection systems. Social protection here means social security, or the benefit system, as it does in Part 1 of this blog, not its wider interpretations that include social care and other functions of government. And this is focussing on the tax-financed side of the system. Contributory schemes are also important of course, though as mentioned in Part 1, expanding their coverage is often restricted in low- and middle-income countries by high and persistent labour market informality. Social registries now exist in around 60 countries and are being promoted by the World Bank far and wide.

But this promotion runs against guidance in the World Bank’s 2022 book Revisiting Targeting in Social Assistance, that an open mind should be kept on whether to poverty target schemes or not.¹ The World Bank’s 2022 book acknowledges the challenges of poverty targeting, including through proxy means testing, which social registries face on a significantly expanded scale because they target multiple schemes. Among these challenges, it says, are high targeting errors, questions over stigma and human rights and an increased risk of community tensions. To this we could add the risk of a lack of political support and the creation of a ‘missing middle’ between the extreme poor and the relatively wealthy in receipt of contributory benefits. The book says, ‘Policy makers must decide whether, how broadly or narrowly, and how to target a program based on their appreciation of the magnitude of the benefits of concentrating resources where they are most needed versus the magnitude of the various errors and costs.’ In other words, don’t jump straight to poverty targeting.

However, the World Bank’s State of Social Protection Report 2025 gets things back ‘on message’ for poverty targeting and social registries. It asserts, ‘social assistance should be focused on those in the poorer quintiles, especially in LICs and MICs, where fiscal resources are limited and needs are high.’² It continues, ‘although some countries have made great strides in developing their social protection delivery systems over the past decade, others still lack basic elements such as social registries.’ So social registries, involving poverty targeting on a dramatically larger scale, have been given the status of ‘basic elements’ of any system – sitting at the core – despite the caution of the 2022 targeting book. But this just begs the question, if social registries really are ‘basic elements’, why are they so hard to find in high-income countries with well-developed systems which are commonly a mix of universal and means tested programmes?

The lessons from COVID-19

Part of the attraction of social registries is that they fit snugly into the World Bank’s and others’ promotion of ‘shock-responsive’ and ‘adaptive’ social protection, but this raises further major issues. Core to the concept of ‘shock-responsive’ schemes is that they respond to big shocks through vertical expansion (increasing benefit rates) or horizontal expansion (increasing coverage). In theory, social registries enable horizontal expansion because they cover everyone. But this doesn’t reflect reality: median coverage of the population is just 21 per cent according to the World Bank’s 2022 book on targeting. So then investment in social registries needs to increase, according to the World Bank. But this just gives us the horrendous but all too likely possibility of countries with very limited resources investing in complex and problematic social registries and drawing resources away from things like comprehensive birth registration, identity cards and other core systems that are still needed to increase the visibility of people to the state and also to enable universal, lifecycle social protection programmes. Nevertheless, the World Bank’s 2025 report argues, ‘Countries must continue to integrate adaptive and shock-responsive design features into existing programs by, for example, investing in early warning systems and dynamic social registries.’ ‘Dynamic’ social registries is the second major issue. Social registries need to be ‘dynamic’ because data from periodic surveys, on which proxy means tests depend, goes out of date. However, they are hard to achieve. They require households to proactively inform authorities of changes of circumstances. How likely is it that that will happen, especially in lower capacity environments and where people live remotely?

Moreover, social registries appear to be inconsistent with the World Bank’s own lesson drawn from COVID-19, that simple wide coverage programmes respond best to major shocks. COVID-19 was a huge and unique test of the world’s social protection systems. The World Bank’s own 2022 report on the pandemic, Cash Transfers in Pandemic Times: Evidence, Practices, and Implications from the Largest Scale Up in History, tells us that low- and middle-income countries were usually not up to the job of scaling up effectively.³ They were ‘often unable to reach people at the middle or so of the income distribution’. This takes us back to the ‘missing middle’ issue with narrow poverty targeting. The report adds that, in response to the pandemic, cash transfers ‘were mostly unconditional, and eligibility was simplified’. Also it appears that having wide initial coverage is important, because the report tells us that vertical expansion is faster than horizontal expansion. Wide initial coverage is also important of course because protection also comes from existing access to benefits, pre-shock. As Nick Freeland has pointed out, social protection, done properly, is innately shock-responsive, before developing specialist ‘shock-responsive’ schemes. ‘By any definition, social protection is there to respond to shocks, whether they be life-course vulnerabilities (being born; growing old; becoming disabled), individual misfortunes (loss of livelihood; illness or injury; death of a family member), or extensive crises (natural disasters; climate change; conflicts; pandemics).’⁴

The World Bank’s COVID-19 report points us away from across the board poverty targeting. With simplified eligibility and wide initial coverage being important, we are surely being pointed more to universal schemes rather than more complex poverty targeted schemes and social registries in responding to large shocks. Or at the very least, we are being pointed towards a mixture of the two approaches – to not excluding universal schemes. The report concludes that there is a ‘renewed interest in benchmarking programs against basic unconditional design’. The International Labour Organisation (ILO) puts it more explicitly, saying that as a result of COVID-19, ‘The weakness of limited safety-net approaches, typically characterized by narrow targeting and tightly monitored conditionalities, has become glaringly apparent.’⁵

Learning from countries

Many countries have gone their own way and introduced universal or near universal schemes, in line with the subsequent lessons from COVID-19. As mentioned in Part 1 of this blog, 52 low- and middle-income countries - more than a third of the world’s total - have chosen universal or near-universal schemes as part of the development of more comprehensive systems to address lifecycle risks.⁶ Yet in countries such as Mauritius, Mongolia, Nepal and Uganda, the World Bank has opposed the development of these schemes. In Kenya, I witnessed the World Bank opposing the government’s universal pension but supporting the poverty-targeted, shock-responsive Hunger Safety Net Programme (HSNP) there. HSNP is a programme with a targeting record ‘little better than random’ and coverage limited to 130 thousand households in just 8 out of Kenya’s 47 counties. It has the capacity to horizontally expand to three times that in response to shocks, but this is still far from a national programme. ⁷ ‘Shock-responsive’ schemes such as HSNP may be appropriate in specific circumstances, but promoting them ahead of core national coverage social protection schemes such as the universal pension is putting the cart before the horse, and it is not following the lessons from COVID-19. The Kenyan government co-funds HSNP the World Bank can argue, so supports it. Yes, but the government needs the World Bank’s resources, so, as mentioned in Part 1, we are not talking about a level playing field in discussions with the World Bank on programme design.

The World Bank needs to reflect on its own research into poverty targeting and COVID-19 and stop the widespread promotion of social registries. It needs to learn from those 52 national governments that are implementing universal and near universal schemes. It also needs to support genuine national decision making. Effective and comprehensive systems that address lifecycle risks and individual and wider crises are typically a mix of contributory and tax-financed schemes, and the latter are typically a mix of universal and means tested schemes. This is what developed systems show us. The World Bank argues, start with poverty targeting and think about universal schemes later as countries develop. But, apart from removing the possibility of universal schemes helping a countries development, such an approach can lead to path dependency, where precious resources are drawn disproportionately to poverty targeting and building social registries in a way that is hard to reverse. Universal and poverty-targeted schemes frequently require the development of different systems and infrastructure. The World Bank needs to change direction on its prescription for poverty targeting and social registries. And it need only look at its own research, including its research on COVID-19, to find the rationale for doing so.

To read more on why the World Bank does what it does on social protection and how governments are pushing back, see Matthew Greenslade’s book Beyond the World Bank: the Fight for Universal Social Protection in the Global South, published online - open access - in November 2025 and in hard copy December 2025. Key messages from the book can be found here and a recording of Matthew presenting them here.

Notes