In recent decades, dozens of countries around the world have adopted cash transfer programs to combat poverty and prevent it from being passed from generation to generation. While many of these programs have been highly successful at tackling long-term poverty, incomes fluctuate due to economic shocks and the target population of social programs move. During economic crises, such as the COVID-19 pandemic, many households that were not previously in poverty find themselves slipping into poverty, in desperate need of government transfers to pull them back out. Governments, meanwhile, need to know how to adjust the social safety net without breaking the budget.
The challenge of selecting beneficiaries for cash transfers during crises thus becomes critical, especially in Latin America and the Caribbean where more than half the population works in the informal sector without unemployment insurance and even people who work in the formal sector may lack adequate protection.
The Virtues of Dynamic Targeting
In a recent paper we evaluate four methods of selecting beneficiaries for a cash transfer program. We show that while no panacea exists, a dynamic and flexible method that incorporates high frequency data on economic shocks can improve targeting at a reasonable cost during an economic crisis. At the systemic level, shocks can be due to economic crisis, pandemics, and natural disasters. At the family level, they may cause job losses, deaths or illness. Incorporating information on shocks into traditional methods of selecting beneficiaries allows cash transfer programs to serve the purposes of traditional anti-poverty programs and provide a kind of unemployment insurance for low-income families.
We document the advantages and disadvantages of four different approaches to targeting for a hypothetical program that aims to provide cash transfers to households with income below the extreme poverty line. We use panel data for a random sample of households in the Colombian government’s social registry. The social registry covers close to 50% of the Colombian population and is based on detailed household surveys that collect information about asset ownership, dwelling quality and demographic characteristics –that are combined in a statistical model, called a proxy means test (PMT), to produce an estimate of household income.
Our results show the virtues and shortcomings of different methods. Many governments use a static PMT, meaning they establish a strict threshold for the PMT under which people are either selected or rejected from social assistance. Our results show that such a static approach would have left many people with very low incomes without assistance as job and income losses piled up during the pandemic. This is reflected in the exclusion error (i.e., the percentage of eligible people who are excluded from social assistance), which climbed from 30% in 2019 to 35% in 2020. An alternative approach relies on the same data but simply increases the eligibility threshold to account for widespread loss of income and include more people. We simulated this approach by shifting the threshold of eligibility to 1.3 times the extreme poverty line. This reduces the exclusion error in our model. But it also comes at the cost of a large inclusion error, meaning that assistance is delivered to people who are above the extreme poverty line.
A third approach mimics one used by the Colombian government, allowing households to request an update of their asset ownership — a proxy for their income. This method may reflect changes in long-term income, but it comes up short during crises because families may find it difficult to liquidate their assets in such emergencies, preventing the method from accurately reflecting income fluctuations.
Incorporating Income Fluctuations into Social Programs
A fourth approach, which we call dynamic, incorporates higher frequency data. It takes an individual’s baseline poverty indicators, or PMT, and updates it monthly with new information on jobs losses and gains, as well as the previously mentioned non-labor shocks like a natural disaster or illness in the family, to predict income variation. This method, we find, incorporates more people in need and leads to social welfare gains for society, even accounting for moral hazard — i.e. the misrepresentation of employment status by some individuals. This dynamic method increases overall social welfare by 13%, compared to the method of using the traditional static PMT (with the original eligibility threshold) and does so by increasing the budget by only 8%.
Policymakers, of course, will decide what methods work best for them, depending on their flexibility in terms of transfer amounts and overall budget constraints. For example, expanding the safety net by increasing the eligibility threshold to 1.3 times the extreme poverty line increases social welfare by the 32%, but increases the budget by 37%. Other methods have different tradeoffs.
A Valuable Tool for Latin America and the Caribbean
A critical challenge in Latin America and the Caribbean, with its high informality rates, is the large number of people who lack any kind of insurance to protect them in cases of a severe income shock. At risk of falling into poverty, they need governmental help and need it quickly before their poverty becomes chronic. The region has achieved significant results in the fight against structural poverty. It now needs to make its social protection systems more flexible, to deal with shocks that affect vulnerable people who are normally above the poverty line. A more dynamic targeting approach offers such an option. It could prove essential during widespread disruptions ranging from pandemics to natural disasters.
Leave a Reply