2940.0 - Census of Population and Housing: Details of Overcount and Undercount, Australia, 2016
Quality Declaration
![](/AusStats/wmdata.nsf/activeimages/qualityicon/$File/qualityicon.gif)
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 23/02/2018
Page tools:
![]() ![]() | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
NON-CONTACT SECTOR
Descriptions of the three sub-components are provided below. Census Counts, PES Population Estimate and Net Undercount for the Non-contact Sector, 2016
(a) Percentages are calculated using the total Census count (23 397 296 persons) as the denominator. (b) A negative value indicates a net overcount. IMPUTED PERSONS FOR NON-RESPONDING CENSUS DWELLINGS Non-responding dwellings in the Census are dwellings where the Census never obtained a return, and the dwelling could not be established as having been unoccupied on Census night. During Census processing, imputation was used to create a number of people and their Age, Sex, Registered marital status and State/Territory of usual residence into these dwellings. Values of all other variables for these imputed people were set to not stated or not applicable, depending on the imputed value for Age. Missing values for Indigenous status and Country of birth were then imputed during PES processing. Inevitably, the imputed values differ from the true but unknown values to some extent. The number of person records imputed into Census non-responding dwellings was higher (5.1% of all Census person records) in 2016 compared with 2011 (3.6%). Imputed person records made up the majority of the Census non-contact sector in both 2016 and 2011 (67% and 87%, respectively), although the change in 2016 is due to the higher number of late returns and the addition of Census person records with insufficient personal identifier information. Imputed person records are also the primary driver for the net overcount in the non-contact sector, which is essentially over-imputation in the Census. LATE RETURNS Contact from the ABS following selection in the PES may act as a reminder to return a completed Census form for those people who have not already done so. These late returns, if not identified, would result in the PES sample having a higher proportion of Census response than in the overall population, resulting in a downward bias in the estimate of net undercount. To protect against this, and maintain the statistical independence of the PES from the Census, all Census forms received after the start of 2016 PES enumeration were deemed late. For the purpose of PES estimation, all dwellings from which late Census forms were received were treated as though they had not been contacted in the Census and classified to the Census non-contact sector. The number of Census late returns was higher in 2016 (2.4% of all Census person records) compared with 2011 (0.5%). This comes as a result of changes to the Census enumeration model in 2016, particularly the option for respondents to post back their Census paper form up to several weeks after the end of Census field operations, which was not available in 2011. For further information on changes to the 2016 Census, see Major Changes to the 2016 Census on the Explanatory Notes tab. CENSUS PERSONS WITH INSUFFICIENT PERSONAL IDENTIFIER INFORMATION A key requirement for successfully linking a PES person to their corresponding Census record is a sufficient level of data quality on both sides, including minimal-to-no missing data for responding persons. Without this, the ability to link to Census (i.e. confirming a PES person was counted) is substantially reduced. As part of the continuous improvements to PES methodology, an adjustment was introduced into the estimation method in 2016. This method identified Census records that had insufficient personal identifier information, required for linking to PES (e.g. where Census data was missing or imputed for multiple linking variables). To remove the potential for any upward bias on the PES population estimates (and level of net undercount) that would occur if these records were retained for estimation, they were moved to the Census non-contact sector and treated in a similar fashion to late returns. The adjustment moved 33,057 Census persons to the non-contact sector, or 0.1% of all persons counted in the Census. Analysis of the impact of this adjustment showed that it reduced the estimated net undercount rate for Australia by 0.10 percentage points. This is an indicator of the extent of the upward bias on the estimated net undercount that would have existed had these records not been identified and treated appropriately. CALCULATING NET UNDERCOUNT FOR THE NON-CONTACT SECTOR An estimate of net undercount in the non-contact sector for a category of person is equal to the difference between the PES estimate of the number of people who should have been counted in the non-contact sector in that category (as reported in the PES) and the Census count of people in that category in the non-contact sector. CALCULATING THE CENSUS IMPUTATION ERROR Imputed records are the primary driver for the net overcount in the non-contact sector, and will therefore impact the overall net undercount figure for Australia. For example, if the Census count increases due to a higher number of imputed persons but all else remains the same, net undercount will decrease (and vice versa). It is therefore important that PES corrects for the Census imputation error in the non-contact sector, and that the number is reported so it can be used to better understand estimates of net undercount. An estimate of the number of people who should have been imputed in the Census can be calculated by subtracting the PES estimate of late returns and for those person records with insufficient personal identifier information from the PES estimate of people who should have been counted in the non-contact sector. The Census imputation error is then the difference between this PES estimate and the actual Census count of imputed people. For 2016, the imputation error is estimated to be 650,856 persons that were over-imputed by Census during data processing. This over-imputation was likely the result of some non-responding dwellings being deemed occupied when they were in fact unoccupied, as well as too many people being imputed into non-responding occupied dwellings. Document Selection These documents will be presented in a new window.
|