世界银行-主流多维贫困指标中固有的不正当激励的快速解决方案（英）-2025.5_18页_366kb

Summary of "A Quick-Fix for Perverse Incentives Inherent in Mainstream Multidimensional Poverty Measures"

Purpose: This paper addresses a critical issue with the adjusted headcount ratio (AHR) in multidimensional poverty measurement. It argues that AHR provides perverse incentives by prioritizing the alleviation of poverty among less intensely poor individuals rather than those who are worse off, contrary to prioritarian principles.

Key Findings:

• Perverse Incentives: The AHR discontinuously "jumps" when individuals escape poverty, creating incentives to target those closest to the poverty cutoff (k) rather than those with higher deprivation scores. This is demonstrated through examples and formalizes the violation of a strong transfer principle under regressive transfers.
• Proposed Quick-Fix: The solution involves adjusting the poverty contribution function to use the "multidimensional poverty gap" instead of the deprivation score. Specifically, the new index M′ is defined as M′(g0) = (1/n) ∑(s_i - s_{zi}) r_i, where s_{zi} is the "poverty line" derived from k and dimensionally-adjusted cutoffs. This tweak preserves the Alkire-Foster identification method but eliminates perverse incentives by ensuring that reductions in contributions do not create discontinuities when escaping poverty.
• Properties: The proposed M′ satisfies the same properties as AHR except for Dimensional Breakdown, meaning it cannot be decomposed across dimensions post-poverty identification.

Limitations of Dimensional Breakdown:

• Policy Optimization: When seeking policies to minimize poverty (e.g., under a budget constraint), Dimensional Breakdown does not provide the necessary information to identify optimal deprivations to target. This is because the decomposition isolates dimensions only after identification, ignoring how deprivations cumulate across dimensions for individuals.
• Monitoring Progress: This property can mislead policymakers when tracking changes over time. By breaking down the index into weighted censored headcount ratios (H_j), it may incorrectly attribute progress or deterioration to specific dimensions, obscuring true changes under identical poverty status adjustments.

Conclusion: The quick-fix solution M′ offers a practical adjustment to AHR, removing perverse incentives without altering individual poverty identification. However, it sacrifices Dimensional Breakdown, highlighting that while this property is often praised, it has impractical limitations in real-world policy applications. This underscores the need for policymakers to carefully interpret decomposition-based insights and prioritize solutions like M′ for better-aligned policy incentives.