In the absence of reliable information and accurate targeting methods to identify eligible households, the Indonesian governments’ cash assistance program risked excluding poor households or diverting funds to richer households, or both.

Indonesia’s direct cash assistance program targeted poor households using a hybrid of two targeting methods: (1) community-based targeting, in which community leaders identified potentially eligible households based on their local knowledge of who qualified as poor, and (2) proxy-means testing (PMT), in which government surveyors visited these households and predicted a household’s income by collecting simple information on hard-to-hide assets to verify that they fell below the location-specific poverty line. Yet, in 2008, over half of all eligible households living below the poverty line were excluded from the program, and many community members voiced dissatisfaction with the program’s targeting.

Indonesian government officials worked with researchers to test the effectiveness of three different targeting methods: (1) PMT, (2) community-based targeting, and (3) a hybrid of the two, on poverty levels (as measured by per capita expenditures), community satisfaction, and legitimacy. Researchers carefully debated over the best option for collecting the large amount of administrative data needed to implement and test these targeting methods.

Following a baseline survey, the government operated a one-time cash transfer program, in which households identified under each of the three methods would receive 30,000 rupiah (about $3 USD). They randomly assigned 640 sub-villages across 3 provinces in Indonesia to the different targeting methods. They further randomized a subset of the sub-villages assigned to the community-based and hybrid methods to allow only elite community members to participate to test for elite capture, defined as elite community members using the targeting process to privilege their friends and relatives who live above the poverty line.

Researchers measured the accuracy of the three targeting methods by comparing the list of households below the location-specific poverty line identified in the baseline survey to those who received the cash transfer under each of the three methods. Researchers also collected data on community satisfaction after beneficiaries were selected. The community-based and hybrid methods performed somewhat worse than PMT, though not by enough to affect poverty outcomes. There was also no evidence of elite capture. Importantly, the community-based method resulted in much greater community satisfaction, and a selection of households that more closely aligned with those that self-identified as poor. This suggested that communities shared a concept of poverty that differed from the central government’s.

In efforts to further decentralize to the community and further improve targeting, researchers developed another targeting method using self-selection. This entailed developing a screening mechanism that would encourage the poor to self-enroll in the program. They also began to test whether a cash transfer with larger stakes would shift elite capture, expanding the conditional cash transfer program into 400 new villages in which beneficiaries receive about $100 USD per year for at least 6 years.

The results from this experiment also informed the creation of Indonesia’s Unified Database, which is one of the largest social assistance databases in the work that tracks welfare and socioeconomic indicators on about the lowest 40 percent of the population. The Unified Database better enables policymakers and program-designers to use high-quality data to inform policies that aim to improve targeting systems, increase satisfaction with social programs, and ultimately reduce the poverty head count in Indonesia. For example, the database was used to help inform the design of a COVID-19 response program in 2021 that relied on community-based targeting to quickly identify those in need.