Poverty lines, household economies of scale and urban poverty in Malaysia

Abstract

This thesis presents three essays on Malaysia’s poverty profile based on the Household Expenditure Survey (HES). The first and second studies were motivated by the shortcomings of the official poverty lines and poverty measurements. There are several conceptual and measurement problems related to evaluating the extent of poverty in Malaysia. The first study offers several alternative regional poverty analyses based on the consumption expenditure approach with varying underlying assumptions. The poverty lines are estimated using Ravallion-Bidani and Kakwani-Sajaia approaches and the consumption pattern of the 10th and 20th percentile per capita expenditure (PCE) households. Regional poverty lines based on Kakwani-Sajaia and Ravallion-Bidani lower bounds produced robust poverty measurement rankings across regions in the country for both the 10th and 20th percentile PCE households. However, for the 10th percentile PCE, Ravallion’s upper bound poverty lines do not produce robust poverty rankings. In relation to the shortcomings of the official poverty measurements, the second study analyses the economies of scale in consumption, specifically amongst poor households. Using the 10th and 20th percentile PCE households, the household size economies are estimated using specifications proposed by Deaton-Paxson and Kakwani-Son. The findings show that the economies of scale indices are sensitive to the selection of methods and sample groups. Economies of scale in poor household consumption are present for food, housing, clothing, furnishing, personal goods and miscellaneous goods. This study further suggests that these indices be used as complementary to the existing national poverty measurements. The final study provides new insights into the limited urban poverty studies and to the new dimension of urban poverty. Using logistic regression, the determinants are analysed using the new poverty lines estimated in the earlier essay. The test of robustness of the determinants is conducted through re-estimating the logistic regression using a range of poverty lines. The findings show that education, locational dimension, foreign migrant workers and household size are significant determinants of poverty in the urban areas.