Targeting the Marginal Social Valuation of Cash Transfers: Evidence from Indonesia

For decades, the bedrock of global anti-poverty policy has rested on a straightforward, almost mathematical premise: that government aid should be funneled directly to the poorest of the poor. Grounded in economic welfare theory, this approach implicitly assumes that households with the lowest levels of consumption derive the highest marginal social value from an additional dollar. In developing and developed nations alike, policymakers have designed vast administrative machinery—from proxy means tests to targeted cash transfer programs—to identify these low-consumption households, operating under the assumption that financial assistance does the most good where material deprivation is most acute.
However, a groundbreaking new working paper, designated as Working Paper 35984 with the DOI 10.3386/w35984 and officially issued in November 2026, challenges this foundational assumption. By surveying more than 5,500 villagers across 600 distinct rural communities in Indonesia, the researchers investigate a provocative question: Do the households that consume the least actually experience the highest social welfare gains from a modest cash transfer? The empirical findings suggest a resounding no. According to the study, traditional targeting strategies that rely purely on consumption metrics fail to reach the vast majority of individuals who stand to benefit the most from financial intervention, upending conventional wisdom in international development and social protection design.
Main Facts and Empirical Findings
The research project, which combines extensive community-level polling with a randomized controlled trial (RCT) framework, exposes a stark mismatch between standard economic metrics of poverty and the community’s localized understanding of need. In the 600 Indonesian villages sampled, over 5,500 participants were asked to evaluate their peers along two distinct dimensions: how well off other villagers were in absolute terms, and how much a modest, unexpected cash transfer would genuinely help improve their well-being. These peer-ranked assessments were then cross-referenced against traditional household consumption data and self-assessed welfare metrics.
The results reveal profound divergences in how different targeting methodologies operate in practice. When governments or development agencies rely strictly on traditional consumption-based targeting to identify beneficiaries, they miss an astonishing 77 percent of the households that actually possess the highest marginal social valuation—meaning those families who would experience the greatest overall welfare gain from an extra dollar. Even when programs pivot to target self-assessed welfare levels, the strategy still misses 57 percent of those high-margin households.
Even more striking is the comparison with community-level level rankings. When neighbors are asked to identify who is worst off in absolute levels rather than on the margins, community targeting misses 25 percent of the households with the true highest marginal social valuation. The study demonstrates that these divergences between various notions of neediness are not random noise; rather, they are systematically and statistically predictable.
Through the randomized controlled trial component, the research evaluates how different bureaucratic mechanisms perform against these divergent definitions of need. The data shows that proxy means tests (PMTs)—long favored by international financial institutions like the World Bank for their technocratic objectivity—are best suited for targeting consumption levels. Conversely, self-targeting ordeals, which require individuals to expend time or effort to receive benefits, are more effective at capturing self-assessed welfare levels. Most notably, community-based targeting approaches come the closest to capturing the elusive metric of marginal social valuation, aligning more closely with local dynamics of need and vulnerability.
Background Context and Evolution of Social Protection
To understand the weight of Working Paper 35984, one must examine the historical evolution of social safety nets over the past forty years. In the late 20th century, as structural adjustment programs and fiscal austerity reshaped public spending across the developing world, international development agencies sought efficient ways to allocate scarce public resources. Blanket subsidies, which benefited the rich and poor alike, fell out of favor. In their place arose targeted social safety nets, championed as a triumph of evidence-based policy.
Indonesia itself has served as a global laboratory for targeted social protection. Following the devastating 1997 Asian Financial Crisis, the Indonesian government, with technical assistance from international bodies, instituted massive subsidized rice programs (known as Raskin later transitioning to BPNT) and unconditional and conditional cash transfer initiatives like the Program Keluarga Harapan (PKH). These programs relied heavily on proxy means testing—scoring households based on observable assets such as flooring materials, wall construction, roof types, asset ownership, and household size.
While PMTs reduced administrative costs and minimized political patronage compared to manual allocation, they were frequently criticized by sociologists and local administrators for creating administrative exclusion errors. Families deemed wealthy on paper by an algorithm often struggled with localized shocks, hidden debts, or chronic illnesses that standard consumption surveys failed to capture. The November 2026 working paper provides rigorous empirical backing to these long-standing qualitative critiques, demonstrating that technocratic tools are fundamentally optimized for a metric—consumption—that may diverge sharply from social welfare maximization.
Chronology and Research Methodology
The empirical work underpinning Working Paper 35984 unfolded over a multi-year research timeline typical of large-scale development economics field experiments.
Phase one, conceptualization and survey design, began years prior to the November 2026 publication date. Researchers partnered with local institutions in Indonesia to select a representative sample of 600 villages spanning diverse geographic, cultural, and economic landscapes across the archipelago.
Phase two involved the extensive data collection exercise. Over 5,500 villagers were engaged in structured elicitation exercises. Unlike standard household surveys where enumerators ask isolated questions about household expenditures, this methodology leveraged local social networks. Villagers were asked to rank their peers, creating a dense matrix of relative economic standing and perceived vulnerability.
Phase three deployed a randomized controlled trial to test the operational efficacy of distinct targeting instruments. By randomizing the assignment of modest cash transfers and comparing the resulting welfare changes against baseline predictions from proxy means tests, self-assessment metrics, and community rankings, the researchers mapped out the exact margins of error inherent in each policy tool.
Broader Impact and Policy Implications
The release of Working Paper 35984 carries profound implications for global policymakers, international NGOs, and domestic ministries of finance. For decades, the technocratic consensus has treated targeting efficiency as a technical puzzle with a purely mathematical solution: build a better algorithm, collect more granular data, and reduce inclusion and exclusion errors.
The findings of this Indonesian study suggest that this framing is fundamentally incomplete. The authors argue that targeting strategies inescapably embed normative stances. Choosing to target consumption is not a neutral, objective act; it is a normative choice that prioritizes one definition of welfare—material resource command—over others, such as subjective well-being, community resilience, or marginal social returns.
If policymakers wish to maximize social welfare rather than simply equalize consumption, the current architecture of social protection may require a radical redesign. Proxy means tests, while effective at identifying households with low asset accumulation, may systematically bypass individuals who, despite appearing slightly better off on paper, operate under severe vulnerability or possess a much higher capacity to convert cash injections into meaningful welfare gains.
Furthermore, the study rehabilitates community-based targeting (CBT) to a degree, showing that local information networks capture dimensions of need that algorithmic models miss. However, development economists have long noted that CBT is susceptible to elite capture, local politics, and social friction. Balancing the superior alignment of community targeting with marginal social valuation against its political economy risks remains one of the foremost challenges for future program design.
As governments worldwide grapple with inflation, post-pandemic economic recovery, and the expansion of digital social registries, the debate over who receives aid and how they are chosen will only intensify. Working Paper 35984 serves as a timely reminder that the metrics we choose to measure poverty do not merely reflect reality—they actively construct it, determining which vulnerable citizens are lifted up and which are inadvertently left behind.







