Representativeness of the HILDA Survey sample

The challenge​

Given the HILDA Survey’s broad applicability, it is important to systematically review its representativeness. This project compares a range of HILDA Survey estimates with those from the Australian Bureau of Statistics. We examine:

  • Geographic mobility
  • Child sample (including new births over time)
  • Working-age population
  • Older age Australians
  • Indigenous Australians
  • Recent immigrants to Australia

Careful comparisons are made to several ABS sources, such as the Census, Australian Census Longitudinal Dataset, Labour Force Survey, General Population Survey, and the National Health Survey.

The research​

Non-response and attrition reduce the overall sample size but only introduces bias if the respondents and non-respondents are significantly different from one another. Previous research on non-response and attrition typically find associations with demographic, economic, health and geographic mobility.

Further, there is one obvious area where the sample will struggle to adequately represent the changing cross-sectional population: recent immigrants. Without a refreshment sample, the HILDA sample can only incorporate recent immigrants into the sample if they live with existing sample members, which results in a biased sample of all immigrants. A general population sample was added in 2011 and immigrant specific samples have been added in 2023-2025. These refreshment samples allow immigrants arriving since 2001 to be represented.

The impact​

This project provides valuable information for the users of the HILDA Survey data by indicating areas where there is good comparability with external sources and where there is not. The careful comparison of the survey design and questions used in each source help identify if the difference is likely due to mode, proxy reporting, question differences or other design choices, or if it is due to non-response, attrition and under-coverage of new members of the population. It also shows how any differences may change over the life of the panel and demonstrates how refreshment samples help correct particular biases.

Our researchers

Melbourne Institute - Nicole Watson, Mossamet Kamrun Nesa, Taylor Ey

Publications