2940.0.55.002 - Information Paper: Measuring Overcount and Undercount in the 2016 Population Census, Jul 2016  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 01/07/2016  First Issue
   Page tools: Print Print Page Print all pages in this productPrint All

ADAPTING THE 2016 PES

PES ENUMERATION

2016 PES has two key changes to adapt to the new Census model:

  • the 2016 PES sample size has been increased by 20%, which is sufficient to cover the expected increase in sampling error due to population growth and new overcount and undercount effects; and
  • the 2016 Census follow-up phase starts and finishes several weeks later than the 2011 Census follow-up phase, and so the 2016 PES also starts and finishes several weeks later, immediately following the Census follow-up phase.

The 2016 PES has prepared a telephone interview (TI) option in addition to the traditional face to face interview. The TI option can be initiated by the respondent using contact information provided in the Primary Approach Letter. The availability of another mode of collection for PES is expected to improve response rates, especially from those who would be more likely to participate by telephone than face-to-face.

CONTACT SECTOR AND NON-SECTOR

For the purposes of the PES estimation, Census dwellings deemed occupied on Census night are divided into a Contact Sector (dwellings for which a Census form was received before the commencement of the PES enumeration) and a Non-Contact Sector (where no Census form was received, or a Census form was received after the commencement of the PES enumeration and is classified as a 'late return').

For the Contact Sector, it is usual for the PES to measure a net undercount of persons, with the number of persons missing from Census forms generally exceeding the number of persons counted twice on two Census forms. In 2011, the size of the Contact Sector was around 96.5% of dwellings, and the PES measured a net undercount of 635,000 persons missed from these dwellings.

For the Non-Contact Sector, it is usual for the PES to measure an overcount of persons, mainly due to imputation by Census for non-responding dwellings thought to be occupied. In 2011, the size of the Non-Contact Sector was around 3.5% of dwellings, and the PES measured an overcount of 260,000 persons in these dwellings.

International experience suggests that there will be new overcount and undercount mechanisms, because the Address Register is expected to have both over and undercoverage. A possible scenario under the new Census model is for the PES to report a net overcount if Address Register overcoverage leads to increased imputation in the Non-Contact Sector. Improvements in collection and linking of dwelling and address information have been made for the PES to ensure that new overcount and undercount mechanisms can be correctly measured and adjusted to give accurate population estimates.

LINKING AND ESTIMATION

There are no fundamental changes to how the PES selects the sample of dwellings, links records to Census and estimates the Census night population. While it is anticipated that the overcount and undercount may change with the introduction of the new Census model in 2016, it is expected that the PES will give an accurate estimate of the Census night population, as it did in 2011. In particular:
  • area-based PES sample selection will be used to complement and contrast the list-based Census model, and ensure statistical independence through sourcing dwelling lists with independent procedures;
  • Automated Data Linking (ADL) of persons will be comparable to the 2011 Census and the 2011 PES;
  • Prediction Regression Estimation (PREG) will be comparable to the 2011 Census and the 2011 PES. Reviews undertaken by the ABS have confirmed that the PREG methodology can adjust for both overcount and undercount when estimating the adjusted Census night population of Australia; and
  • the identification question for Aboriginal and Torres Strait Islander peoples in the 2016 PES will be the same as that used in the 2011 PES.