|Page tools: Print Page Print All|
Building a Predictive Model for the Geographic Spread of E-form Take-up
Given the widespread use of the Internet by Australian households, the ABS is moving towards providing an electronic form (e-form) as the preferred means for households to respond to ABS surveys and to the Population Census. In designing such a collection, fewer resources (such as interviews or Census collectors) can be assigned to geographic areas that are likely to have higher e-form take-up, thus reducing costs.
In this project, the geographic spread of e-form take-up in the 2011 Census has been modelled against known characteristics of small areas (Census workloads known as CLWs). Besides indicators of what state each CLW is in and how remote each CLW is, the predictors in our model included proportions of types of people (Aboriginal and Torres Strait Islanders, non-English speakers, school completers, women), proportions of people that fell into pre-defined income and age categories, and proportions of different types of dwellings (people living in buildings with 3 or more floors, private dwellings, different dwelling types, lone person households, older households) within each area.
Our final model contained 36 variables (plus an intercept), and we used the same set of variables to predict such things as e-form take-up for the entire population of a CLW, as well as e-form take-up just for people who have (or don’t have) broadband access.
This model is expected to be useful for planning a wide range of collections, even those with different e-form take-up rates than the 2011 Census. The model gives a way to distribute a specified high-level take-up rate across detailed geographic areas. Predictions from the model will be produced for Mesh Blocks (fine areas usually comprising about 50 dwellings) so that they can be aggregated to geographic areas used in planning surveys and the next Census.
These documents will be presented in a new window.