3405.0.55.001 - Discussion Paper: Assessment of Methods for Developing Experimental Historical Estimates for Regional Internal Migration, Dec 2011  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 22/12/2011  First Issue
   Page tools: Print Print Page Print all pages in this productPrint All

APPENDIX 3 CONCORDANCE CONFIDENCE INDEX (CCI)


Assigning a postcode-ASGC confidence index
Apportion Medicare data across movers using 2006 Census data


Assigning a postcode-ASGC confidence index

For all ASGC regions (including SLAs) that concord only from whole postcode/s, there is complete certainty that the postcode-based data aligns with the SLA. However, for SLAs where only a part of at least one postcode feeds into it, there is only part-certainty about the converted data. The Concordance Confidence Index is structured so that SLAs created from complete postcodes are assigned an index of 100% (complete confidence in the converted data), while SLAs that map on to small proportions of postcodes are assigned a low index (to an absolute minimum of 0%). Indexes close to 100% imply greater confidence in the converted data. The correspondence is based on total population estimates because the movement data contain much smaller numbers.

Postcode based data can be converted with greater confidence by grouping SLAs with neighbouring SLAs. The postcode-ASGC confidence index provides an indication as to which SLAs need to be considered for combining. Combined SLAs should consist of more complete (or close-to-complete) postcodes and ideally SLAs should be combined within LGAs, SSDs and SDs. By combining confidence index values for each region with the region's population size, an overall level of confidence in converted data can be calculated. This can be done by assigning weights to confidence index values for a group of regions based on the population of these regions. These weights are based on Medicare enrolments in each region.

Examples:

SLA A has a confidence index of 50%
SLA B has a confidence index of 100%
SLA B is nine times the size of SLA A (or 'nine times as important', because there are nine times as many people in SLA B [weight=0.9] than in SLA A [weight=0.1])
overall confidence for SLAs A & B : 0.1*50% + 0.9*100% = 95%

SLA A has a confidence index of 50%
SLA B has a confidence index of 100%
SLA A is nine times the size of SLA B (or 'nine times as important', because there are nine times as many people in SLA A [weight=0.9] than in SLA B [weight=0.1])
overall confidence for SLAs A & B : 0.9*50% + 0.1*100% = 55%

Assessing the overall level of confidence may be a good performance indicator of how far to go with 'SLA groupings'. For instance, if an overall confidence level of 90% was chosen as a minimum requirement to convert data for a state/territory, then SLAs could be grouped up until the overall confidence index for that state/territory reaches 90%. However, it is more difficult to achieve this level of confidence for the smaller states and territories due to the variation in alignment of postcode boundaries with SLA boundaries.

Apportion Medicare data across movers using 2006 Census data

After the creation of SLA groupings, one method of estimating more accurate SLA level data is to apportion the data across movers (departures and arrivals) using 2006 Census population counts. Departures in the Census data refer to people who have recorded a SLA of usual residence one year ago and arrivals refer to their current SLA of usual residence. Census is the most recent information available with the arrivals/departures breakdown. The Medicare data refer to the 2005-06 year as a logical comparison.

The table below illustrates this process. In this table, we have low confidence in the way that the 2005-06 Medicare estimate for arrivals/departures has been allocated to SLAs using the postcode-ASGC correspondence. SLA-level arrivals/departures data from the 2006 Census is used to allocate the data for this postcode to these SLAs with more confidence.

Apportion Medicare-based data across movers using 2006 Census data

Arrivals
Departures
2005-06 Medicare
2006 Census
2006 Census
New Medicare arrivals (a)
2005-06 Medicare
2006 Census
2006 Census
New Medicare departures (a)
SLA group in Qld.
no.
no.
%
no.
no.
no.
%
no.

Highgate Hill
1 066
961
32
977
1 007
908
33
951
South Brisbane
825
1 009
33
1 024
777
686
25
720
West End
1 202
1 074
35
1 092
1 132
1 188
43
1 245
Total
3 093
3 044
100
3 093
2 916
2 782
100
2 916

- nil or rounded to zero (including null cells)
(a) Total Medicare multiplied by proportion of census movers.