Accurate and timely poverty measurement is central to development policy, yet the availability of up-to-date high-quality household survey data remains limited—particularly in countries where poverty is most concentrated. Survey-to-survey imputation has emerged as a practical response to this challenge, allowing practitioners to update poverty estimates using recent surveys that lack direct welfare measures by borrowing information from other comprehensive surveys. A critical review of the method is provided, revisiting its statistical underpinnings and testing its limitations through extensive model-based simulations. Through these simulations, the analysis demonstrates how violations of parameter stability, omitted variable bias, and shifts in survey design can introduce substantial errors—particularly when imputing across time or under economic and structural change. Results show that standard corrections such as re-weighting or covariate standardization may fail to eliminate these biases, especially when imputing across time or under structural change. The performance of alternative model specifications is also evaluated under various methods, including performance under heteroskedastic errors, non-normality. The findings offer practical guidance for practitioners on when survey-to-survey imputation is likely to succeed, when it should be reconsidered, and how to communicate its limitations transparently in the context of poverty monitoring and policy design.
