Some factors influencing the comparability and reliability of poverty estimates across household surveys
Stellenbosch Working Paper Series No. WP03/2013
Author(s): Derek Yu
JEL Classification: I32
Keywords: poverty, income, expenditure, recall method, diary method, imputations, ASSA model, South Africa
Download: PDF (711 KB)
In order to evaluate the extent to which a country achieves the objectives of poverty and inequality reduction, up-to-date, reliable and comparable survey data is required. This paper critically reviews the factors which could affect the comparability and reliability of poverty estimates and trends across various household surveys. First, whether income or expenditure variable should be used for the analyses and whether the diary approach is associated with more reliable capture of income and expenditure information compared with the conventional recall method are looked at. If the respondents are asked to declare the income and expenditure in exact amounts, whether they are asked to report these as ‘one-shot’ amounts (single estimation approach) or aggregate amount derived from the sum of the amounts for sub-categories (aggregation approach) could affect the poverty estimates. If the respondents are asked to report income and expenditure in intervals, issues that could affect the reliability of this approach, such as the number and width of the intervals, the appropriate method used to approximate the income (expenditure) amount in each interval, as well as the possible methods to deal with households reporting zero or unspecified income (expenditure) are investigated. In addition, survey data is validated against external sources such as national accounts data to investigate if it would lead to improved reliability of the former data for the subsequent poverty analyses. Furthermore, since the survey data are, strictly speaking, not time-series data, the data are re-weighted by means of the cross entropy approach in order to be consistent with demographic and geographic numbers presented by the Actuarial Society of South Africa (ASSA) model and Census data so as to find out if the comparability and reliability of the poverty estimates and trends are improved.