2010年-世界发展银行全球_Poverty_and_Inequality_Maps_for_Rural_Vietnam___An_Application_of_Small_Area_Estimation_38页_2mb

Summary of "Poverty and Inequality Maps for Rural Vietnam: An Application of Small Area Estimation"

Core Content

This paper presents updated small area estimates of poverty and inequality for rural Vietnam using data from the 2006 Vietnam Household Living Standard Survey (VHLSS) and the 2006 Rural Agriculture and Fishery Census (RAFC). The study aims to map poverty and inequality at the regional, provincial, and district levels, enabling more precise targeting of poverty reduction programs.

Main Objectives

• To update poverty and inequality estimates for rural Vietnam.
• To analyze poverty trends from 1999 to 2006.
• To apply the small area estimation (SAE) method developed by Elbers, Lanjouw, and Lanjouw (2003) to estimate poverty and inequality at finer geographical levels.
• To compare expenditure and income-based poverty measures and assess their accuracy against official statistics.

Key Findings

Poverty Trends (1999–2006)

• All provinces in Vietnam experienced a noticeable reduction in rural poverty.
• The largest reductions were observed in provinces with poverty rates close to the national average.
• The poorest provinces also saw reductions, but at a more modest pace.
• Provinces and districts with lower inequality in 2006 had above average poverty reductions.
• The use of the RAFC instead of the population census allows for more frequent poverty estimates due to its five-year interval, tripling the frequency of such estimates.

Poverty and Inequality Estimates

• Expenditure Poverty Line (2006): 2,560,000 VND/person/year.
• Income Poverty Line (2006): 3,288,000 VND/person/year (adjusted to match MOLISA's rural poverty rate of ~18.5%).
• Poverty estimates using both expenditure and income models are very similar, indicating consistency in the methods.
• Regional Poverty Incidence (2006):
  • Red River Delta: 11.3% (SE: 0.9%)
  • North East: 31.6% (SE: 1.6%)
  • North West: 57.3% (SE: 2.6%)
  • North Central Coast: 32.9% (SE: 1.7%)
  • South Central Coast: 17.8% (SE: 1.2%)
  • Central Highlands: 39.9% (SE: 2.0%)
  • North East South: 10.1% (SE: 0.9%)
  • Mekong River Delta: 12.6% (SE: 1.3%)
• The poorest region is the North West, with a poverty rate above 50%.
• The least poor regions are the Red River Delta and South Central Coast, with rates below 18%.
• The Central Highlands showed the largest increase in poverty incidence (from 34.4% to 39.9%), although the difference is not statistically significant.

Inequality (Gini Coefficient)

• Provincial Inequality (Expenditure):
  • Average: 0.27.
  • Lowest: Thai Binh (0.23).
  • Highest: Lam Dong (0.35).
• District Inequality (Expenditure):
  • Average: 0.25.
  • Lowest: Meo Vac (0.17).
  • Highest: Da Lat (0.47).
• Inequality tends to be lower in both the poorest and richest provinces, indicating a quadratic relationship between poverty and inequality, consistent with the Kuznets hypothesis.

Methodology

• The small area estimation (SAE) method, developed by Elbers et al. (2003), is used to estimate poverty and inequality at the district level.
• The method involves combining a household survey and a census through a regression model.
• The explanatory variables used include household characteristics, such as education, employment, asset ownership, and demographic data, as well as commune-level and GIS variables.
• The model selection is based on forward stepwise regression, using a large model (with more variables) for greater precision, though both large and small models yield similar poverty and inequality estimates.
• Standard errors are calculated using repeated Monte-Carlo simulations, ensuring the reliability of estimates.

Key Information

• Data Sources:

  • 2006 VHLSS (9189 households, 39071 individuals).
  • 2006 RAFC (50% sample of rural households).
  • Both datasets are collected by the General Statistic Office of Vietnam (GSO).

• Methodological Notes:

  • The ELL framework assumes the model is accurate and spatial correlation is properly accounted for.
  • The census is assumed to have complete coverage, and the survey is used to impute expenditure and income for the census.
  • Spatial correlation is addressed by including cluster-specific and household-specific random effects.

• Comparisons:

  • The VHLSS-based poverty rate for rural Vietnam is about 7.5%, significantly lower than the MOLISA reported rate of 19%.
  • This discrepancy highlights the complexity of MOLISA's poverty classification, which involves iterative procedures and subjective criteria.

Conclusion

The study successfully applies small area estimation to map poverty and inequality in rural Vietnam at multiple levels of aggregation. It finds that poverty has generally declined across all provinces, with the largest reductions in areas with moderate initial poverty levels. The method allows for more frequent and detailed poverty monitoring, supporting better policy design and resource allocation. The results are consistent across expenditure and income-based measures, and the findings align with the Kuznets hypothesis, suggesting a non-linear relationship between poverty and inequality.