2015年-世界发展银行全球_Measuring_Poverty_in_Lebanon_Using_2011_HBS_34页_1mb

Summary of the Technical Report: Measuring Poverty in Lebanon Using 2011 HBS

Core Content

This technical report, authored jointly by the Central Administration of Statistics (CAS) of Lebanon and the World Bank, outlines the methodology for constructing a welfare aggregate and poverty line using the 2011 Household Budget Survey (HBS). The report aims to provide a reliable and comprehensive measure of poverty in Lebanon, taking into account regional disparities and data limitations.

Main Methodological Issues

• Welfare Aggregate Construction:

  • The welfare aggregate is constructed using consumption data from the HBS, which includes both food and non-food items.
  • The aggregate must be comprehensive yet avoid measurement errors.
  • It is necessary to distinguish between goods and services consumed and those that are not (e.g., lumpy infrequent expenditures, investment, public goods).
  • The use of COICOP (Classification of Individual Consumption by Purpose) groups helps categorize consumption data.

• Non-Response and Imputation:

  • The survey faced significant non-response issues, with only 57% of the intended sample collected.
  • Imputation methods were used to address missing data, particularly in the individual diary and household forms.
  • Two imputation methods were considered: Tobit regression and predictive mean matching.
  • The Tobit model was ultimately chosen for its better alignment with the observed consumption patterns and lower imputation bias.

• Adjustments:

  • Spatial and inter-temporal deflation: Adjustments are made to account for price differences across regions and over time.
  • Household size adjustment: To ensure fair comparisons across households of different sizes, adjustments are made using adult equivalence and economies of scale.
  • Rent adjustment: Rent is included in the welfare aggregate as a proxy for housing services. The report discusses the challenges of using reported actual rent due to small sample sizes and rent control, and concludes that estimated rent is more appropriate.

Key Components of Welfare Aggregate

3.1 Data Issues

• The 2011 HBS was conducted over a period of more than a year due to security concerns, introducing potential seasonality bias.
• The response rate varied significantly across regions, with lower rates in Beirut and Mount Lebanon and higher rates in South Lebanon and Nabatieh.
• Non-response was more common among poorer households, leading to overrepresentation in the data.
• Rescaling of sample weights was applied to mitigate non-response bias.

3.2 Food Items

• Food consumption is derived from individual diaries, which are filled by adults over 15 and children.
• About 27% of individuals did not complete the diary, leading to missing data.
• Food items include both purchased and home-produced goods, as well as transfers.
• Imputation methods were used to fill in missing food data, with the Tobit model preferred due to its alignment with consumption patterns.

3.3 Non-Food Items

• Non-food components include transportation, utilities, housing services, and other services.
• Data on non-food items were collected through the household characteristics and living conditions form.
• Rent is a critical component, but data collection was problematic due to small sample sizes and rent control.
• Durables are included in the welfare aggregate, but limited data on their current value, age, or condition prevents accurate estimation of their use value.
• Education and health are considered, but there is no consensus on whether health expenditures should be included in the aggregate. The report suggests that health expenditures may be a reflection of preventive care and thus should be included if they are correlated with overall consumption.

3.4 Adjustments to Welfare Aggregate

• Spatial adjustment: Used to account for regional price differences.
• Inter-temporal adjustment: Adjusts for price changes over time.
• Household size adjustment: Ensures that poverty measures are comparable across households of different sizes.

Poverty Line Construction

• The cost-of-basic-needs approach (CBN) is used to construct the poverty line.
• The first step is to estimate the minimal caloric requirement for a healthy life.
• The second step involves estimating the cost of meeting this requirement based on a diet typical of households near the poverty line.
• The non-food component of the poverty line is added using various methods, including the use of estimated rent and other services.
• The final poverty line is adjusted for spatial and temporal price variations.

Supplemental Information

• Robustness tests show that poverty estimates based on consumption per capita are sensitive to imputation methods and non-response.
• Supplementary poverty rates are calculated using consumption per adult equivalent, which provides a more equitable measure across household sizes.
• The report highlights the importance of triangulation and sensitivity analysis in ensuring the reliability of poverty estimates.

Concluding Remarks

• The 2011 HBS data, while valuable, presents several challenges in terms of non-response, seasonality, and data completeness.
• Adjustments to the welfare aggregate and poverty line are essential to ensure accurate and comparable poverty estimates.
• The report serves as a foundation for future poverty monitoring and social protection strategies in Lebanon.
• Collaboration between CAS and the World Bank is crucial for improving data quality and ensuring the robustness of poverty measurement.

Key Information

• Survey Period: September 2011 to November 2012.
• Sample Size: 2,746 households (43% non-response).
• Regions: The survey was stratified across nine regions, including Beirut, Mount Lebanon, North Lebanon, Bekaa, South, and Nabatieh.
• COICOP Groups: Used to categorize consumption data; key groups include food, non-alcoholic beverages, transport, and miscellaneous goods.
• Imputation Methods: Tobit regression and predictive mean matching were tested, with Tobit preferred for its lower imputation bias.
• Non-Food Adjustments: Includes estimated rent, durables, and other services, with careful consideration of their use value.
• Poverty Estimates: Based on the CBN approach, with regional variations noted in both consumption and poverty rates.

Figures and Tables

• Figure 1: Compares consumption per capita using rescaled and raw weights, showing a 10% increase with rescaling.
• Figure 2: Shows the timing of the survey in different regions, highlighting seasonal bias.
• Figure 3: Displays daily purchases of selected products in three regions, indicating seasonal variations.
• Figure 4: Compares cumulative distribution functions for imputed and actual purchases, showing differences in imputation methods.
• Figure 5: Compares nominal annual consumption per capita using different imputation techniques.
• Figure 7: Demonstrates the impact of excluding rent on the welfare aggregate.
• Table 1: Presents OLS regression results for different rent values in logarithms, showing regional and housing characteristic variations.
• Table 5: Compares daily per capita calorie intake and costs in 2004 and 2011.
• Table 7: Provides poverty rates in Lebanon for 2011/2012, showing regional disparities.
• Table 8: Shows the source of income across quintiles, indicating income distribution.

References and Annex

• The report references studies by Haughton and Khandker (2014), Deaton and Zaidi (2002), and Ravallion (1998).
• The annex includes detailed information on COICOP items and adjustments made to the welfare aggregate.