2007年-世界发展银行全球_Yemen_Poverty_Assessment___Volume_2_Annexes_252页_17mb

Yemen Poverty Assessment Summary (Volume II: Annexes)

Core Content

This document is part of the Yemen Poverty Assessment (2005–2006), conducted by the Government of Yemen, the World Bank, and the United Nations Development Program. It focuses on data processing and analysis for the Household Budget Survey (HBS), including sampling design, outlier detection, and poverty line methodology.

Main Objectives

• To estimate poverty and its causes.
• To monitor poverty in Yemen as it implements its second Poverty Reduction Strategy Paper (PRSP) (2006–2010).
• To update National Accounts estimates for better economic planning.
• To collect and analyze data on living standards variation between urban and rural areas and among governorates.

Key Sections and Their Contributions

Annex 1: Sampling Design

• The HBS sample was based on the 2004 Population Census.
• Yemen was divided into 38 strata, with 19 urban and 19 rural strata.
• The sample size was 9,228 urban households and 5,172 rural households.
• Two-stage sampling was used:
  • First stage: Selection of Census Enumeration Areas (EAs) with probability proportional to size (PPS).
  • Second stage: Systematic equal probability sampling (SEPS) to select 12 households per EA.
• The final sample allocation was adjusted to ensure random distribution across 12 months and to make the number of households per governorate a multiple of 144.

Annex 2: Constructing the HBS Database

• Outlier detection and correction were performed using tools like:
  • WHO anthropometric tables to check consistency in children's measurements.
  • Food composition tables to detect abnormal food consumption levels.
  • Custom unit price tables to identify errors in quantity or price recording.
• Non-standard missing values (e.g., "999") were replaced with blanks to avoid distortion in analysis.
• Anthropometric data was cleaned by replacing values that were more than five standard deviations from the mean with blanks.
• Section 14 of the HBS, which contains food and non-food consumption data, was processed using a dedicated program over the Excel/VBA platform.
• The program used decimal logarithms to analyze weekly transactions and detect outliers based on standard deviations from the mean.
• If a unit price was more than 4 standard deviations away from the mean, the program corrected either the amount or the quantity, depending on which was further from its mean.
• Internal coefficients were calculated during the first pass, including:
  • Mean and standard deviation of logarithms of amounts, quantities, and unit prices.
  • Median unit price for each item.
• External coefficients were used to estimate energy values from recorded quantities, especially for items where only amounts spent were recorded (e.g., bread or spices).

Annex 4: Poverty Line Methodology

• Poverty measurement was based on consumption data, not income.
• Income vs. expenditure was considered, with a focus on per capita consumption.
• Units of measurement included kilo-calories, household size, and geographic distribution.
• Poverty lines were established using caloric requirements for different age groups:
  • Food poverty line for children up to 60 months.
  • Non-food poverty line for households with consumption patterns that fall below a certain threshold.
• The methodology included decomposition analysis to understand the distribution of poverty and targeting differentials between urban and rural areas.

Annex 5: Poverty Map

• The poverty map was constructed using census and survey data.
• A consumption model was developed to estimate household consumption per capita.
• Poverty indicators were derived from this model to identify areas of high and low poverty.
• The implementation involved:
  • Selecting common variables between the census and HBS.
  • Estimating consumption models using HBS data.
  • Predicting consumption in census data.
• Concerns were raised about the current results, suggesting potential issues with data accuracy and methodological limitations.

Annex 8: Public Expenditure Targeting in Yemen

• The report assesses whether public expenditure is pro-poor in Yemen.
• Background includes the context of fiscal decentralization and sub-national expenditures.
• Methodology involves:
  • Estimating household consumption using the HBS.
  • Comparing public expenditure per capita with poverty indicators.
• Data issues included inconsistencies and missing values.
• Findings indicated targeting differentials by governorate, suggesting that expenditure distribution may not be equally effective in all areas.
• Box A.8.4 explains the north-south and urban-rural dimensions of these differentials.

Summary of Key Findings

• The HBS 2005–2006 provided a detailed database for poverty analysis.
• Sampling design ensured representative data across urban and rural areas.
• Outlier detection was a critical step in ensuring data quality and accuracy.
• Poverty lines were established using caloric intake and expenditure data.
• Public expenditure targeting was found to have differentials between urban and rural areas, indicating uneven impact.
• The analysis was supported by technical notes, social accounting matrices, and input/output tables, which were updated for 2005.

Conclusion

This volume of the Yemen Poverty Assessment provides a comprehensive framework for data processing, outlier detection, and poverty analysis. It highlights the importance of accurate data collection and methodological rigor in understanding poverty dynamics and public expenditure effectiveness in Yemen.