世界银行-利用调查对调查的推断以低成本填补贫困数据缺口：来自随机调查实验的证据（英）-2024.3-78页_1mb

Summary of "Using Survey-to-Survey Imputation to Fill Poverty Data Gaps at a Low Cost: Evidence from a Randomized Survey Experiment"

This report introduces a randomized survey experiment conducted in Tanzania to determine how different survey designs affect poverty imputation accuracy. It fills a gap in the literature by providing experimental evidence on the effects of survey length and questionnaire design on poverty estimates derived from survey-to-survey imputation.

Key Findings

1. Survey Design and Imputation Accuracy:

   • Using standard, comprehensive forms yields highly accurate imputation across most models.
   • Simplified questionnaires show improved accuracy when predictor variables are standardized with base survey data.
   • Demand modules (food and non-food consumption) designed to reduce costs did not sufficiently predict poverty rates without standardization.

2. Role of Standardization:

   • Standardizing predictors (matching base survey distributions) greatly enhances准确性 for simplified modules.
   • Using variables transformed to normality via Box-Cox method and standardization procedures (subtracting means and scaling variances) proved effective.

3. Sample Size Requirements:

   • Minimum 1,000 households recommended for both base and target surveys to achieve reliable results.
   • Larger samples marginally improve accuracy but stabilize estimates similarly across many models.

4. Comparative Performance:

   • Proposed im
   • Standardization techniques help compensate for design variations.
   • Assumptions of representative and reliable predictor variables are crucial for accurate imputation.

Recommendations

• Design surveys with care, balancing burden and data quality.
• Standardize key predictors to improve cross-survey imputation.
• Use sufficiently large sample to avoid precision loss due to sample-size effects.

In conclusion, survey-to-survey imputation remains a valuable low-cost approach for poverty measurement, especially when proper comparisons and adjustments for sample variability are conducted. This research provides robust empirical evidence to inform survey design and implementation.