| Description |
In this document the author provides details on a new set of survey weights which she created using entropy estimation to allow National Household Surveys produced by Statistics South Africa to be presented as a consistent series over time. These surveys include the October Household Survey, conducted from 1994-1999, the Labour Force Survey, initiated in 2000, and the General Household Survey undertaken from 2002. The data from these surveys is frequently stacked side-by-side to create time series data. The data is however, designed as cross sections with no emphasis on consistency in the series over time. As a result the series shows large fluctuations even at the aggregate level. In addition, until 2003, post-stratification was done at the person level which results in inconsistencies between the person and household files. The new set of weights are benchmarked to aggregate numbers from the ASSA AIDS model, e a series of mathematical models developed to assist the actuarial profession and the Actuarial Society of South Africa in assessing and addressing the impact of the AIDS epidemic in South Africa. They therefore present aggregates which are more consistent over time. In addition, these weights can be used at both the person and household level.
The new set of consistent cross-entropy weights and their supporting documentation and syntax files are publicly available with the Labour Force Survey 2007 September data files. |
