Published on Voices

Is there a best way to target social assistance?

This page in:
Hafia village, prefecture of Dalaba, Guinea. Villagers await to receive cash as part of a social net project  Photo: Vincent Tremeau /World Bank Hafia village, prefecture of Dalaba, Guinea. Villagers await to receive cash as part of a social net project Photo: Vincent Tremeau /World Bank

Social protection programs help individuals and families escape poverty, mitigate and manage risks, and improve resilience and opportunity. Yet, most countries have limited resources for social protection interventions.  Prioritizing populations most in need therefore becomes necessary and important.

Take Nepal as an example, where the government uses categorical and geographical targeting to provide cash transfers to their vulnerable population, including senior citizens, single women, those with disability, endangered ethnicity, and children under 5.  In contrast, in Senegal the national registry system combines geographic targeting, community-based targeting and proxy means testing to determine who is eligible the country’s social programs.  Countries can switch among methods as they develop capacity, both Yemen and Chile have done so in the last decade.

These are just two examples and there are almost as many as there are social assistance programs.  However, bring up the theme of targeting with a group of practitioners or academics and the debate will heat up. This is because in the world of constrained resources assisting the neediest first and most is both morally compelling and fiscally responsible.  It is also hard to execute, posing challenges and costs, and errors in targeting can be significant.  Moreover, there is no single targeting method that fits every situation; hence it is crucial to take context and policy objectives into account when deciding on a method.

Taking a hard look at targeting methods

Revisiting Targeting in Social Assistance: A New Look at Old Dilemmas discusses how targeting has been used over past years, what principles and practices were applied. The work began before COVID-19 and continued during the pandemic that catalyzed a dramatic expansion of social protection. This makes our work very timely as governments figure out how to move from crisis response to a ‘new normal’, hopefully one that is adaptive and reinvigorated toward Universal Social Protection (USP). 

Targeting approaches need to be adjusted to the context

One key message of our analysis is that targeting and USP are not mutually exclusive.  Nearly all countries espouse USP as a goal, and nearly all have at least one targeted social protection program.  What’s important is that targeting programs are well designed and implemented, and have the support of society.

We also found that there is no single targeting method that is universally preferred or equally effective across regional and country differences. Methods always must be adjusted to the purpose of the program, availability of data, capacity of institutions,  and also depend on factors like economic recessions, health emergencies or natural disasters.  But in each context, there are a range of potential options.  

Delivery systems matter – a lot

No matter the targeting method chosen, delivery systems are also critical.  Deficient delivery systems can exclude people or be stigmatizing and also contribute to high errors of inclusion.  Improving delivery systems with more adequate staffing, information systems, operating budgets and rules of operation can be quite an important and viable way to improve targeting outcomes.

Good data are at the core of good targeting

Data are key for more accurate targeting: they either directly measure people’s means or help to predict them. Advances in information technology -- including big data, artificial intelligence and machine learning—show promise in improving the accuracy in targeting but demand supportive policies to help capture their potential. 

Traditional government-held data such as social security contributions and land registration are key to assess welfare and target social protection programs. Making systems talk to each other and improving the data quality are important for making decisions based on real observations of welfare.  

But often such data don’t cover poorer people, especially in poorer countries. That’s why many countries have developed social registries based on interviews with potentially eligible people looking at  age, education, livelihoods of family members, location and quality of housing, what assets they own, and much more. It’s important, though, to keep social registries data updated through continuous recertification, data interoperability and data integration.   

More recently, the use of “big data” is growing such as call records details or use of social media, which are faster and cheaper to update.   It is still unclear, though, how accurate estimates of welfare based on such data are, how comfortable people are with its use, and what incentives to change behavior that would entail.

What is clear, though, is that all use of data for social protection calls for strong protocols for data use and sharing and for data privacy and protection.

Targeting should be dynamic

How countries target is a dynamic story.  At the level of individuals and families, who needs benefits or services can change from day to day and year to year. At the level of programs and governments, new administrative capacities, new data sources and new computer power have moved practice forward. We also see a few cases of stagnation where efforts at improvement have stalled and outcomes deteriorated, reminding us that progress is always the result of political will and administrative effort.


The editors of the report are Margaret Grosh, Phillippe Leite, Matt Wai-Poi and Emil Tesliuc


Michal Rutkowski

World Bank Regional Director for Human Development, Europe and Central Asia

Margaret Grosh

Senior Advisor of Social Protection and Jobs

Join the Conversation

The content of this field is kept private and will not be shown publicly
Remaining characters: 1000