Measures of poverty are usually estimated using data from national household surveys. The sample design of official household surveys is typically complex, involving multi-stage stratified cluster sampling. Complex sampling affects variance estimation and therefore standard error estimation. Accounting for the complexities of sampling is essential for reliable estimation and analysis. This study presents an overview and assessment of different approaches to poverty measurement in South Africa. The study researches and presents the estimation of different poverty measures and their standard errors in the case of complex multi-stage sampling, using the Income and Expenditure Survey conducted in 2010-2011 by Statistics South Africa.
The study presents background to poverty measurement in South Africa, the importance of measuring poverty, the general definition of poverty and its definition within the South African context, and theoretical concepts and methods for measuring poverty, in general and in the case of complex samples. The background confirms that there has been significant improvement in South Africa since the fall of Apartheid in addressing the inadequate information base for the measurement of poverty and inequality. The review of literature asserts that poverty measurement is essential in providing statistical standards and systematic approach to reporting on levels, contributes to evidence-based decision making in public policy, and also help in assessing the impact of poverty reduction programmes. The review of literature on the definition of poverty found that there is no single universally accepted definition of poverty as there are so many ways to think about what poverty means. Poverty in South Africa is defined and measured in both one-dimensional and multi-dimensional approaches in line with international practices.
Different poverty indices including the Poverty Headcount Index (??????), Poverty Gap (??????), Poverty Gap Index (????????), Squared Poverty Gap Index (??????????), Sen Index (????), Sen-Shorrocks-Thon Index (????????), Watts Index (????) and the time taken to exit poverty will be discussed in the general case and in the case of complex samples. The exact method and Jackknife method of estimating variances in the case of complex surveys are presented. An overview of different data sources for poverty measurement in South Africa will also presented.
The distributional properties of a variable used to measure poverty, in-kind consumption, suggest that in South Africa, it is critical that poverty is analysed at disaggregated level. The study conducts multiple comparisons using z-test and Bonferroni adjusted confidence intervals to test hypothesis of differences in estimated poverty by gender, population group, settlement type and province of the head of household. The test show significant differences in poverty between these subgroups. The study concludes that it is important to report poverty estimates with standard errors and confidence intervals as these allows inference when analysis of poverty is conducted over time.