Article

Threshold Effects and Classification Bias in Poverty Measurement: Evidence from Alternative Poverty Lines 

---

Abstract

This study examines threshold effects and classification bias in poverty measurement by analyzing how incremental changes in poverty lines affect cumulative poverty headcounts. Using income-band distribution data and applying elasticity and reclassification bias metrics, the study evaluates the sensitivity of poverty measurement across alternative thresholds. A detailed case analysis of Albania is complemented by cross-country comparisons with Brazil, India, and Indonesia. The findings reveal substantial non-linearity in poverty responsiveness. Elasticity estimates exceed 3.5 within the $5–$10 per day range, indicating that small proportional increases in the poverty line generate disproportionately large increases in measured poverty. Classification bias is similarly concentrated within these lower-middle thresholds, with reclassification effects exceeding 150 percent in some cases. A clear structural break emerges around the $10 threshold, separating a high-instability regime from a relatively stable upper-income regime. Cross-country analysis further demonstrates that threshold sensitivity varies according to national income distribution structures, suggesting that poverty measurement is context-dependent rather than universally stable. The results indicate that poverty statistics are not threshold-neutral and that binary poverty classification may generate substantial volatility when poverty lines intersect regions of income clustering. These findings highlight the importance of incorporating distributional density considerations into poverty line selection and policy interpretation.

Keywords: Threshold Effects, Poverty Measurement, Classification Bias, Income Distribution, Elasticity Analysis

How to Cite:

Harimanto, B., (2026) “Threshold Effects and Classification Bias in Poverty Measurement: Evidence from Alternative Poverty Lines ”, Journal of Digital Society 2(1). doi: https://doi.org//JDS.148