2013年-世界发展银行全球_Kyrgyz_Republic_-__Poverty_Mapping___Methodology_and_Key_Findings_33页_2mb

Summary of Poverty Mapping in the Kyrgyz Republic

Core Content

This document outlines the methodology and results of a poverty mapping exercise in the Kyrgyz Republic, conducted in 2013 using the Kyrgyz Integrated Household Survey (KIHS 2009) and the Population and Housing Census (2009). The objective was to estimate poverty incidence at the rayon (district) level, which is not achievable through standard household surveys due to high sampling errors. The exercise was a joint effort between the World Bank and the National Statistics Committee (NSC) of the Kyrgyz Republic.

Main Methodology

• Methodology Used: The Small Area Estimation (SAE) method developed by Elbers, Lanjouw, and others (ELL) was employed. This method combines census and survey data to estimate poverty at sub-national levels.
• Process:
  • A consumption model was estimated from the survey data.
  • The model includes household and individual characteristics that are available in both the survey and census.
  • Consumption expenditures of census households were imputed using the estimated model.
  • Poverty and inequality statistics were then calculated based on these imputed expenditures.
• Error Handling:
  • The method accounts for both cluster-specific and household-specific errors.
  • Simulated values of expenditures are generated by random draws from the estimated error distributions.
  • Standard errors of poverty estimates are calculated to assess their reliability.

Key Data Sources and Challenges

• Data Sources:
  • KIHS 2009: Approximately 4,500 households surveyed.
  • Population and Housing Census 2009: Covered around 1.2 million households.
• Challenges:
  • Inconsistencies between Census and Survey: Differences in household definitions (household vs. housing unit) led to discrepancies in variables like household size.
  • Over-statement of Precision: Tarozzi and Deaton (2009) warned that the ELL method may underestimate standard errors if higher-level error correlations are ignored.
  • Consumption Pattern Variations: Significant differences in consumption patterns across oblasts required the use of area dummies and interactions with other variables.

Technical Adjustments

• To improve comparability, the poverty mapping exercise used rayon-level aggregates of census variables, which were merged into the survey dataset.
• However, due to the large size of rayons, they were further disaggregated into villages, which significantly improved the model's ability to explain household expenditure variation.
• The exercise followed World Bank recommendations to use one national-level consumption model for all oblasts, but it was found that some oblasts had distinct consumption patterns, leading to the creation of five separate models:
  • Model #1: Issyk-Kul
  • Model #2: Jalalabad
  • Model #3: Chui
  • Model #4: Bishkek
  • Model #5: Naryn, Batken, Osh, Talas

Key Findings

• Model Performance:
  • Adjusted R-square values for the models were generally high, with 4 out of 5 models exceeding 43% and one reaching 55%.
  • This indicates that the models explain a significant portion of the variation in household expenditures.
• Poverty Estimates:
  • Poverty rates at the oblast level were estimated using the five models.
  • These estimates were well within the confidence intervals of the survey data and had low standard errors.
  • Estimated poverty rates for each oblast are as follows:
    • Issyk-Kul: 47%
    • Jalalabad: 33%
    • Naryn: 46%
    • Batken: 32%
    • Osh: 40%
    • Talas: 34%
    • Chui: 20%
    • Bishkek: 13%

Poverty Map Visualization

• The poverty map visually represents the distribution of poverty across rayons.
• It highlights spatial disparities in poverty levels, which may not be evident from aggregated data.
• The map uses shape-files and circles to depict towns and populated areas.
• It serves as a tool for policy makers to identify areas with higher poverty rates and allocate resources accordingly.

Conclusion

The poverty mapping exercise in the Kyrgyz Republic successfully generated reliable poverty estimates at the rayon level using the SAE method. It addressed key challenges such as data inconsistency, error correlation, and geographic variation in consumption patterns. The resulting poverty maps provide a detailed spatial understanding of poverty distribution, aiding in targeted policy interventions. The collaboration between the World Bank and the NSC ensured the methodological rigor and technical capacity to produce these maps, which are now available for use in poverty reduction strategies.