Most surveys suffer from non-response which can seriously affect the quality of the survey data. Declining survey response rates continue to affect survey research organisations worldwide and all sectors of the survey industry are suffering from falling response rates. However, just focusing on response rates can be misleading as a low response rate does not necessarily imply poor survey estimates. For example, if non-response is happening at random, estimates can still be unbiased.
This study uses representative indicators or ‘R-indicators’ as additional survey quality measures to investigate the possible risk of biased estimators in the latest Longitudinal Surveys of Australian Youth (LSAY) cohort. This cohort, which commenced the LSAY program in 2015, is referred to as the ‘Y15’ cohort. Using data from the first three waves of the cohort, R-indicators are used to measure how the response composition differs from that of the original sample. We will present R-indicators for each survey wave, providing a comparable measure to investigate the quality and level of representativeness of the data over time. We also compute partial R-indicators for a range of auxiliary variables including state, sector, location, gender, indigenous status, socio-economic status, mathematics and reading achievement scores and immigration status.
The effect of other activities undertaken to increase the sample size and improve the quality of the data are also explored. These include recruiting a top-up sample and re-engaging with non-respondents using additional efforts such as offering incentives. The presentation will conclude by identifying respondent subgroups that need to be targeted or prioritised for follow-up in future waves. Examples of strategies we have used to engage the identified subgroups will also be discussed.