Statistical inference for poverty lines and measures: Theory and application

The measurement of monetary poverty is fundamentally based on the identification of a poverty line that distinguishes the poor from the non-poor. Although this threshold plays a central role in the analysis and comparisons of poverty levels over time and space, it is typically assumed to be known with certainty. This is for example the case with the concave poverty lines (CPL) developed by ESCWA, which facilitate international comparisons to be made. However, such an assumption overlooks the statistical uncertainty inherent in the estimation of poverty lines. In this article, we propose methods for approaching the uncertainty into both predicted poverty lines and the poverty measures derived from them. We develop four methods to address this by constructing confidence intervals for poverty lines and FGT poverty indexes. These methods are based on interval projection, optimization and/or bootstrap techniques. Using the international database developed by ESCWA and the estimated CPL model, we apply these methods to the case of Morocco, using five national households surveys conducted between 2001 and 2019. The results show that the uncertainty about poverty lines is not negligible, although its impact on FGT indices remains limited in a context of low poverty. This article makes an original contribution to enhancing the robustness of poverty assessments and addresses growing concerns regarding the international comparisons of poverty measures.