Imputing Poverty Indicators Without Consumption Data: An Exploratory Analysis
Introduction
Accurate poverty measurement is essential for effective poverty reduction policies. However, available household survey data often suffer from inadequacies, such as lack of representativeness or timeliness. To address these challenges, researchers have developed alternative methods relying on data imputation instead of direct data collection through surveys. This paper explores the use of survey-to-survey imputation to estimate several poverty indicators, including headcount poverty, extreme poverty, poverty gap, near-poverty rates, and mean consumption levels.
Data and Methods
The study analyzed 22 multi-topic household surveys conducted over the past decade in Bangladesh, Ethiopia, Malawi, Nigeria, Tanzania, and Vietnam. The analysis focused on adding household utility expenditures or food expenditures to basic imputation models, which include household-level demographic, employment, and asset variables. Geospatial data was also incorporated to further enhance the accuracy of imputation models.
Key Findings
- Improvement in Accuracy: Adding household utility expenditures or food expenditures to basic imputation models improved the probability of imputation accuracy by 0.1 to 0.4.
- Geospatial Data: Incorporating geospatial data increased the accuracy of imputation models.
- Time Interval Between Surveys: Larger intervals between surveys were associated with a lower probability of accurately predicting some poverty indicators.
- Model Fit: A better fit of the imputation model (as measured by R²) did not always translate into higher accuracy.
Implications
These findings suggest that survey-to-survey imputation can provide cost-effective estimates for future surveys, particularly in countries with limited statistical capacity. The method offers a practical solution for generating reliable poverty estimates when direct consumption data is unavailable or insufficient.
Conclusion
This paper contributes to the literature on survey-to-survey imputation by extending its application to various poverty indicators, including near-poverty status, extreme poverty, poverty gap, and FGT poverty indices. The study also examines the performance of imputed consumption distributions against actual household consumption data. These insights are valuable for policymakers and researchers seeking to improve the accuracy and reliability of poverty measurements in developing countries.
