2016年-世界发展银行全球_Poverty_Mapping_in_Tajikistan___Method_and_Key_Findings_34页_2mb

Summary of Poverty Mapping in Tajikistan: Method and Key Findings

Core Content

This report presents the methodology and key findings of poverty mapping in Tajikistan, focusing on both monetary poverty and a multidimensional poverty index (MPI). It is a joint effort by the World Bank Group and the Agency of Statistics under the President of Tajikistan (TajStat), supported by the UK Department for International Development (DfID).

The primary objective of poverty mapping is to provide detailed, reliable estimates of poverty at the sub-national level, particularly at the rayon (district) level, to support more effective policy targeting and resource allocation. Traditional household surveys are insufficient for small geographic units due to sampling limitations, while censuses provide complete coverage but lack detailed consumption data.

Main Methods

A. Monetary Poverty Mapping

• Methodology: The report uses the Elbers, Lanjouw, and Lanjouw (ELL) method, a form of small area estimation (SAE), to impute consumption and poverty indicators at the rayon level using data from the 2009 Tajikistan Living Standards Measurement Survey (TLSS) and the 2010 Census.
• Modeling Process:
  • A two-stage process is used to identify and compare candidate variables between the survey and census.
  • The ELL model estimates per capita household consumption using household and area-level characteristics.
  • The model is then applied to census data to simulate consumption and derive poverty estimates.
• Simulation Approach:
  • Poverty rates are estimated using simulation techniques to account for uncertainty and error.
  • The model parameters are estimated using OLS, and the residuals are used to generate variance estimates.
  • Variance components are modeled using a logistic transformation to account for heteroscedasticity.
  • A Generalized Least Squares (GLS) estimator is used to adjust for area and household-level variability.
• Standard Errors:
  • The method also estimates standard errors, ensuring the reliability of poverty indicators.
  • Parametric drawing and bootstrapping are considered, with bootstrapping preferred for smaller samples.

B. Multidimensional Poverty Index (MPI)

• Definition: The MPI is a non-monetary measure that combines indicators across three dimensions: demographic and labor, education, and services and infrastructure.
• Deprivation Criteria:
  • A household is considered multi-dimensionally poor if it is deprived on 33% or more of the weighted indicators.
  • Severe poverty is defined as deprivation on 50% or more of the weighted indicators.
  • Vulnerability is defined as deprivation on at least 20% but less than 33% of the weighted indicators.
• Indicators by Dimension:
  • Demographic and Labor:
    • Dependency ratio > 1
    • Both household heads not employed
  • Education:
    • Households with individuals (18+) who cannot read or write
    • At least one individual (20+) who has not completed secondary school
    • No household member has completed tertiary education
  • Services and Infrastructure:
    • No access to sewage system
    • No access to piped water
    • Heating from oven or absent
    • No garbage disposal system
    • No toilet inside the house
• Advantages of MPI:
  • Provides a more comprehensive understanding of the lived experience of poverty.
  • Can be relatively easy to collect and less prone to measurement error.
  • Does not require imputation as all necessary indicators are available in the 2010 Census.
• Limitations of MPI:
  • Not always directly comparable across regions or countries due to differing definitions of deprivations.
  • May not align with official monetary poverty measures used in Tajikistan.

Key Findings

• Data Sources:
  • The TLSS and the 2010 Census were the primary data sources for poverty mapping.
  • The HBS was not used due to instability in results and the preference for the TLSS due to its timing and geographic coverage.
• Geographic Coverage:
  • All 58 rayons of Tajikistan were included in the poverty mapping.
  • Dushanbe, a city, was treated as a single unit for estimation purposes.
• Modeling Considerations:
  • A national-level regression model was used due to the infeasibility of oblast-level models with a sample size below 300.
  • The model accounts for spatial heterogeneity by incorporating both household and area-level characteristics.
• Comparative Analysis:
  • Tables 1.1 and 1.2 compare the means of individual and household-level variables between the survey and the census.
  • These comparisons help validate the consistency of data and the appropriateness of variables used in the model.
• Poverty Estimates:
  • Table 2.1 presents poverty estimates from the census (using Povmap) and the survey (observed), highlighting the differences between the two data sources.
  • The poverty map (Figure 2.1) and the distribution of the poor population (Figure 2.2) illustrate the spatial variation in poverty levels.
• MPI Maps:
  • Figure 3.1 presents the MPI map of Tajikistan using the 2010 Census data.
  • Figure 3.2 compares MPI maps with monetary poverty measures, showing how they complement each other.

Conclusion

The report demonstrates the application of advanced statistical methods, including the ELL approach and the use of the MPI, to estimate poverty at the sub-national level in Tajikistan. These methods provide a more detailed and accurate understanding of poverty patterns, enabling better-informed policy decisions. While the ELL method is used for monetary poverty estimates, the MPI offers a complementary, non-monetary perspective that highlights the depth and nature of deprivation across different dimensions. Both approaches are valuable for poverty analysis and should be used in conjunction to provide a holistic view of poverty in Tajikistan.