An attempt to correct the underestimation of inequality measures in cross-survey imputation through generalized additive models for location, scale and shape

This paper contributes to the debate on ways to improve the calculation of inequality measures in developing countries experiencing severe budget constraints. Linear regression-based survey-to-survey imputation techniques (SSITs) are most frequently discussed in the literature. These are effective at estimating predictions of poverty indicators but are much less accurate with inequality indicators. To demonstrate this limited accuracy, the first part of the paper review and discuss the SSITs. The paper proposes a method for overcoming these limitations based on a Generalized Additive Models for Location, Scale and Shape (GAMLSS). Before to apply this method to Moroccan data with the aim to analyze the relation between poverty and climate changes a simulation is carried out to compare classical SSIT and SSIT based on GAMLSS.