To impute or not to impute, and how? A review of poverty‐estimation methods in the absence of consumption data

Motivation A major challenge with poverty measurement is that household consumption (or income) data are often unavailable or not comparable across survey rounds. Furthermore, panel data are even rarer, thus making it difficult—if not impossible—to track the dynamics of these households’ movements into or out of poverty in different periods. Purpose We review imputation methods that have been employed to provide poverty estimates in such data‐scarce contexts. We provide a concise and introductory synthesis, which focuses on intuition and nuanced practical insights rather than technical details. Approach and methods We start first with each method’s motivation, a brief description, some recent application examples, and the remaining challenges. This format offers a self‐contained treatment and facilitates comparison between the various methods and highlight their nuanced differences. Findings The growing demand for more frequent and accurate poverty estimates is not satisfied by current data availability, at least in the short run. Imputation methods offer a promising solution and have received increasing attention. This review helps remedy the dearth of research analysing how to bridge the gap between typical development practitioners and the latest advances in the field. Policy Implications Poverty‐imputation methods offer several policy‐relevant advantages, including In the immediate term (when micro‐survey data are unavailable for all countries). Survey costs or implementation pose challenges. Back‐casting consumption from a more recent survey for better comparison with older surveys. Bypassing thorny issues of obtaining appropriate intertemporal/intraregional price deflators. Furthermore, poverty‐imputation methods can also be used in other fields.