The sensitivity of poverty trends to dimensionality and distribution sensitivity in poverty measures—District level analysis for Pakistan

A key feature of recent poverty measurement in many developing countries is the transition from conventional (money metric) approaches to multi‐dimensional approaches. This change in the poverty measurement raises the question of whether the same poverty trends are apparent under the conventional and the multidimensional approaches. To answer this question I used six household‐level surveys for Pakistan fielded between 2004 and 2015. The analysis considers trends at the national, provincial, and district level with a particular focus on the variability in trends due to distribution sensitivity and insensitivity in poverty measures. The district‐level trend analysis leads to the results that the multidimensional measures show a smoother fall in national poverty rates while the conventional measures show rising poverty up until 2008 and then a sharper fall. Almost two‐thirds of all districts show opposite trends in poverty, if conventional rather than multidimensional measures are used, in at least two of the five inter‐survey spells, irrespective of whether distribution‐sensitivity is considered or not. Thus, apparent poverty trends are sensitive to the measurement approach used. Hence, when measurement methods evolve, policy analysts should be cautious in the conclusions they draw from poverty estimates. 许多发展中国家最近衡量贫困的一个关键特征是从传统（货币度量）方法向多维方法的转变。贫困衡量的这种变化提出了一个问题，即在传统方法和多维方法下，相同的贫困趋势是否明显。为了回答这个问题，我使用了 2004 年至 2015 年间针对巴基斯坦进行的六次家庭层面的调查。该分析考虑了国家、省和地区层面的趋势，特别关注由于分布敏感性和贫困衡量不敏感导致的趋势变化。区级趋势分析得出的结果是，多维指标显示全国贫困率下降较为平稳，而常规指标显示截至 2008 年贫困率上升，然后下降较快。如果使用传统而非多维衡量标准，则几乎三分之二的地区在贫困方面表现出相反的趋势，无论是否考虑分布敏感性，至少在五个跨调查期中的两个。因此，明显的贫困趋势对所使用的衡量方法很敏感。因此，当衡量方法发生变化时，政策分析人员在从贫困估计中得出的结论时应该谨慎。