Abstract
Low-error identification of the poor relative to the non-poor is both an ideal and an essential assumption of poverty measurements. As classification deteriorates, inferences about the size of poverty, the profile of the poor population, and decisions on public policy and resources become clouded. This article uses latent variable modelling to estimate the evolution of second-order classification error based on the components of the official Mexican index, 2016-2024. The results show that the classification error is around 25%, mostly false negatives. This has important consequences for the eventual evaluation of the official index and the conclusions drawn from it.

This publication has 54 references indexed in Scilit: