2025-05-20-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）_18页_296kb

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

Abstract

This paper addresses a key limitation of multidimensional poverty measures based on the adjusted headcount ratio (e.g., the Global MPI). These measures provide perverse incentives to policymakers by encouraging the targeting of less intensely poor individuals over more intensely poor ones. The paper proposes a quick-fix solution by replacing the contribution of poor individuals based on their deprivation score with their "multidimensional poverty gap" — the minimum reduction in their deprivation score needed to lift them out of poverty.

Key Arguments

• Standard multidimensional poverty indices (e.g., Adjusted Headcount Ratio, Global MPI) prioritize targeting the least poor individuals, which violates prioritarianism — the principle that social policies should target the worst-off individuals first.
• This perverse incentive arises because the poverty contribution of an individual discontinuously "jumps" to zero when they escape poverty, irrespective of the magnitude of their previous deprivations.
• Such behavior contradicts Sen's (1976) critique of headcount ratios in unidimensional poverty measurement.

Proposed Solution: Multidimensional Poverty Gap

The quick-fix replaces the usual contribution function (based on the adjusted headcount ratio) with:

p(g₀) = (sᵢ - sᶻᵢ) × rᵢ

where:

• sᵢ is the individual’s deprivation score.
• sᶻᵢ is the largest deprivation score just below the poverty cutoff k.
• rᵢ is 1 if the individual is identified as poor, 0 otherwise.

This adjustment creates a new index (M') that:

• Preserves the original identification of poor individuals.
• Eliminates perverse incentives by ensuring contributions are continuous even as individuals escape poverty.
• Maintains the same general mathematical properties as the original index except for Dimensional Breakdown.

Limitations of Dimensional Breakdown

The paper highlights two key limitations of the Dimensional Breakdown property:

1. Policymaking: The decomposition into dimension-specific components does not provide the necessary information to find optimal policies (e.g., targeting specific deprivations). It only becomes useful post-poverty identification.
2. Monitoring: Dimensional Breakdown can mislead policymakers about the actual sources of progress. For instance, changes observed in each dimension may overstate or misattribute poverty reduction.

Conclusion

The quick-fix proposed by the paper eliminates perverse incentives while preserving the identification method used by mainstream multidimensional poverty measures. However, it sacrifices Dimensional Breakdown as a feature. The paper suggests that this trade-off may be acceptable for policy purposes because Dimensional Breakdown provides limited policy-relevant information and can distort progress monitoring.