Using Moroccan survey and census data to estimate poverty and income inequality trends: Imputation techniques for building a continuous map from sporadic data points

Abstract

This paper proposes a novel methodology to estimate annual poverty and income inequality trends in Morocco using multiple imputation techniques adapted to sparse and irregularly spaced survey data. Traditional approaches to time series analysis are challenged in such contexts by limited observations and structural breaks, particularly in developing countries. Our contribution lies in applying multiple imputation to aggregate national indicators, an approach historically designed for micro-level data, and extending it to a single-country framework by incorporating temporal structures and inter-indicator relationships. The methodology integrates linear and quadratic time trends, as well as a structural break for the COVID-19 shock, and is calibrated using auxiliary socio-economic variables such as education, employment, and infrastructure access. This enables us to generate internally consistent, uncertainty-bounded annual series of poverty and inequality from 1999 to 2022, despite a sparse survey calendar. The proposed approach is aligned with recent empirical developments and offers a robust solution for monitoring welfare dynamics in data-scarce environments, with direct relevance for policy evaluation and SDG tracking.