2085.0 - Microdata: Australian Census Longitudinal Dataset with Social Security and Related Information, experimental statistics, 2006-2011 Quality Declaration 
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 23/08/2017  First Issue
   Page tools: Print Print Page Print all pages in this productPrint All

LIMITATIONS

While every effort has been made to assure the quality of the statistics presented in this publication, they should be considered experimental and treated with caution.

While the linkage rate of SSRI records to the ACLD is believed to be relatively high, there remain a number of records where no link was able to be made. Certain sub-groups (such as highly mobile populations) are known to have a lower propensity to link, meaning they may be under-represented in the linked dataset.

The use of weights aims to adjust for the fact that the linked records may not be representative of the whole population and correct for any under-representation of particular population sub-groups. This aims to ensure that valid inferences can be made from the linked dataset.

The weights on the ACLD-SSRI dataset reflect the benchmarking of the ACLD records only to estimates of the 2006-2011 longitudinal Australian population, that is, an estimate of the number of people resident in Australia at the times of both the 2006 and 2011 Censuses. While this approach largely accounts for the missed links between the ACLD 2006 and 2011 samples, it does not account for missed links between the ACLD and SSRI. As the overlap of ACLD and SSRI datasets is unknown, it is difficult to calculate accurate weights to account for these missed links.

As a result, caution should be used in drawing inferences from analysis of the ACLD-SSRI dataset. As the weights do not account for missed links between ACLD and SSRI records, estimates of the number of people with particular characteristics using SSRI data items may be underestimates of the actual 'real world' number of people with that characteristic. The extent of the underestimation is likely to differ for different population groups.

The dataset should therefore not be used to generate estimates of the number, or proportion, of people on a particular payment.

For example, using unlinked SSRI data we know that approximately 2.2 million Australians received the Age Pension around the time of the 2011 Census. However, the weighted estimate using the ACLD-SSRI dataset shows 2.1 million Age Pension recipients at the same time point. The difference could be due to either (1) missed links between ACLD and SSRI records, (2) differences between the longitudinal Australian population and the point in time SSRI population, or (3) a combination of (1) and (2). Similarly, using unlinked SSRI data we know that approximately 510,000 Australians received Newstart Allowance around the time of the 2011 Census, compared with a weighted longitudinal estimate of only 360,000 using the ACLD-SSRI dataset.

Analysis showing the proportion of people on a payment is likely to be an underestimate, and the extent of this underestimate may also vary between different groups.

However, the dataset does provide valuable information on the characteristics of those people receiving a benefit, both at the time of receipt of the benefit and five years previously. While it should not be used to generate proportion of people on a particular payment, the dataset should be used to examine the distribution of people on a particular payment with various characteristics. For example, the proportion of people on Newstart Allowance that volunteered for an organisation or group, or the proportion of people on the Age Pension living in a nursing home. It should be noted though that these statistics may also be subject to some bias due to missed links.

While the majority of people resident in Australia in 2011 were also resident in 2006, as the SSRI data was linked to a dataset reflecting the 2006–2011 longitudinal Australian population, results may not be representative of the point in time population in 2011. In particular, recent migrants to Australia and new births since the 2006 Census are not in scope of the ACLD. Cross-sectional estimates should be interpreted with caution.

Due to the missed links, one cannot know with certainty for any given record whether that person was not in receipt of a benefit, or if they were in receipt of a benefit but just were not able to link. Consequently, one should not conclude that these records did not receive a payment. Any comparisons should be to the total population rather than the residual category (i.e. persons on a payment compared with the total population, rather than those on a payment compared with those not on a payment).

These same limitations apply to the use of migrant-related data linked to the ACLD. Therefore, analysis of those records that have both migrant-related information and SSRI information should be performed with particular caution.