The design was intended to provide a national stratified sample of Year 9 students which would permit a sample of some 10,000 young people to be interviewed by phone in late 2000. The major stratum considered in the design was State of schooling. Students from smaller states were to be over-sampled and, correspondingly, students from larger states were under-sampled. However, in the 1998 sample, the over-sampling of the smaller states and the territories was undertaken at a lower rate than for the 1995 sample. There were two reasons for this: 1. Continued over-sampling of schools in the smaller states would place a greater burden on schools within those States. As the proportion of schools required approaches 100 per cent of all the schools (allowing for refusal, non-return of materials, and so on), sampling becomes more problematic and school and system-level resistance increases. 2. National estimates are produced by weighting the sample. Greater variation in sampling fractions between States (that is, higher levels of over-sampling) requires greater variation in those weights -- over-sampled States require smaller weights and under-sampled States require smaller weights. Disparity in the weights can contribute to instability in analyses over the years of the survey. Reducing the extent of over-sampling of the smaller States would address these problems somewhat. It is, however, a question of balance between these concerns and the desire to provide stable State estimates from the survey.
Long, M., & Fleming, N. (2002). The Designed and Achieved Sample of the 1998 LSAY Sample Technical Paper No. 16. https://research.acer.edu.au/lsay_technical/30