Learning from 'Where’s Wally?': measuring missing populations and families caring for children with disability

Annemarie Ashton-Wyatt, Australian National University

The popular illustrated children’s books ‘Where’s Wally?’ present a puzzle of finding a single character within a busy scene populated by many hundreds of other characters. Researchers interested in missing populations face similar challenges. The three key issues for data relating to missing populations are conceptualization, coverage and reporting. Researchers have to identify and accurately measure small groups of critical interest while filtering out the ‘noise’ created by the background population. This paper discusses challenges experienced in identifying this missing population and constructing appropriate analytical models using the examples of a research project into families caring for a child or young person aged 0 to 19 years with disability in Australia. Findings indicate these families are more likely to experience economic disadvantage and are more likely to live in areas of relative economic disadvantage. Children with disability are also more likely to live in larger families and are 5 times more likely to have moved house than children with no disability. Triangulation of analysis from two datasets adds strength to the overall findings. Conceptualization problems occur from self-identification barriers and marginalisation of minority groups. Coverage is also an issue. Small numbers of cases and low participation rates make these groups hard to find in population data. Researchers need to be creative in their strategies to uncover the parameters of their population, while also avoiding the ‘red herrings’ of similar groups. When data is found, inflexibility in reporting often prohibits the use of many statistical tests. This is especially an issue when working with small sub-samples as measures to protect confidentiality, such as aggregated tables and random adjustment, can render the data unusable. Constant tension is created by the need to consolidate categories to have sufficient sample sizes to attain statistical significance, and the associated sacrifice of loss of detail.