INTRODUCTION
In 2009, the Australian Bureau of Statistics (ABS) was asked by the Council of Australian Governments (COAG) to look for ways of improving the accuracy of estimates from the Survey of Education and Work (SEW). The issue of particular concern was the ability to measure annual change in performance measures/indicators, especially in the smaller jurisdictions. The SEW is one of the major data sources used for performance measures in the National Education Agreement (NEA) and National Agreement on Skills and Workforce Development (NASWD). SEW is run as a supplement to the monthly Labour Force Survey (LFS) in May each year with a sample size of about 40,000 persons in 2010.
The ABS has investigated combining data collected from more than one survey, known as data pooling, to improve the accuracy of estimates. Data pooling was seen as potentially a cost effective way of increasing the sample size since it was based on making further use of data already collected rather than embarking on additional collection, with its consequent impact on cost and provider load.
This report presents the results of data pooling of the SEW over two and four years between 2007 and 2010. That is, of doubling and quadrupling the sample over the respective time periods. It examines the impact of data pooling on the reliability of selected performance measures and the capacity of the pooled data to detect statistically significant change over time.
Combining data from different surveys to improve estimates may have applicability in other situations, and as such, this report presents detailed technical material on the rationale and methodology chosen for data pooling in this instance.
Print Page
Print All