• Summary
• Downloads
• Explanatory Notes
• Related Information
• Past Releases

Predictive Modelling to Support 2016 Census Data Collection

The Operations Research and Process Improvement (ORPI) section is developing a predictive modelling framework for the 2016 Australian Census of Population and Housing. The 2016 Census is introducing several significant changes to the data collection operation which aim to provide a more efficient process while maintaining the coverage of the Census. This new enumeration model removes the need for Census field officers to visit every dwelling in Australia. In a large proportion of areas across Australia, secure access codes for the online form will be mailed to households in the first instance. These will be followed by reminders sent by mail. Householders wishing to complete their Census form on paper will have a form mailed to them and can mail back their completed form. In the remaining areas of Australia, a more traditional delivery approach using Census field officers will be retained.

The primary objective of the predictive modelling framework is to predict the amount of field staff resources needed for the follow-up phase of collection. In the follow-up phase of collection field officers will travel to households which are still to respond. The predicted proportion of households requiring follow-up in each fine geographic region is a key input. In addition to modelling the proportion which respond prior to follow-up, the number of return follow-up visits and field officer travel time are being modelled to predict field staff requirements. During the enumeration period the predictions made during planning will be updated using information on the collection process. These updated predictions will ensure the allocation of staff resources across areas is responsive to collection progress.

The extensive changes to the enumeration model presents a significant challenge to predicting respondent behaviour. The modelling framework combines data from the 2011 Census and from field tests of the new enumeration model. The rates of internet response and follow-up effort required in different regions in the 2011 Census will be used to identify demographic and geographic characteristics associated with response prior to follow-up. These factors will be included in a model predicting the relative level of response prior to follow-up across regions. Data from the tests will inform assumptions about how respondent behaviour will be different under the new enumeration model compared with 2011. For example, the test data will inform assumptions about how various prompts impact the amount of response received during each enumeration phase.

The components of the modelling framework have already been used to inform decisions about of the size of the workload areas for the 2016 Census. The modelling framework and underlying assumptions will be refined following analysis of the August 2014 Major Test, and the refined model will guide the recruitment strategy for the 2016 Census.

For more information, please contact Julian Whiting (08 8237 7362, Julian.whiting@abs.gov.au)

The ABS Privacy Policy outlines how the ABS will handle any personal information that you provide to us.