APPENDIX 2 MISSED MOVERS INDEX (MMI)


Missed intra-postcode moves
Using the Missed Movers Index to adjust for data loss


Missed intra-postcode moves

Medicare change of address data is not collected for regional internal migration purposes which presents challenges with using this data source. For historical estimates, the main challenge is that the addresses recorded are coded to postcode and need to be converted by the ABS to Australia's offical statistical geography, the Australian Standard Geographical Classification (ASGC).

Consider the case of a user registering a change of address with Medicare. In this case, the only geographical information provided to the ABS is user postcode. When a Medicare user moves residence but remains within the same postcode, no change of address information is provided to the ABS. Thus the move is not accounted for, even if the user moves from one SLA into another.

The extent to which moves between SLAs are not perfectly captured by Medicare records depends on the extent of the overlap between postcodes and SLAs:

1) When two or more SLAs are completely contained with the same postcode all moves made between these SLAs will not be represented in the converted data (ie. will be missed).

2) When two or more SLAs are partially contained within the same postcode, a proportion of the moves made between these SLAs (those to/from areas of the SLAs which share the same postcode) will not be represented in the converted data.

3) When two or more SLAs contain no overlapping postcodes then all moves made between these SLAs should be represented in the converted data.

A quantitative indicator, the Missed Movers Index (MMI), has been produced which summarises the extent of this data loss. It is an estimate of the percentage of moves missed between any two SLAs as a result of using a postcode to SLA correspondence.

The index has been created so that each SLA pairing has a value between 0 and 100%. An index of 100% indicates that both SLAs lie entirely within the same postcode, and therefore 100% of moves between these SLAs will be missed. An index of 0% indicates that the two SLAs contain no shared postcodes and therefore 0% of moves between these SLAs are missed when using the postcode-based data. In all other cases the index will fall somewhere in between.

The MMI between a pair of SLAs can be calculated based on an SLA to postcode correspondence, combined with a population weighting for each postcode.

For each pairing of SLAs, consider movements from each postcode component of the first SLA into each postcode component of the second SLA. The contribution of these two postcode components to the total number of moves between the SLAs is equal to the product of their proportions within the two SLAs. When the postcode components are the same, these moves will be uncounted. The MMI is therefore the sum of these uncounted proportions, expressed as a percentage.

In other words, the MMI for a pair of SLAs is equal to the sum of the products of the percentage of each postcode within each SLA. The formula below summarises this relationship.


An example is now worked through to explain how this index is calculated for two SLAs.

Table 1. Proportions of Movements from SLA1 (split into postcode components) into SLA 2
Move from SLA1	Proportion (%)	Move to SLA2	Proportion (%)	Proportion of total moves (%)	Move counted
PC1	82	PC1	67	55	No
PC1	82	PC2	33	27	Yes
PC1	82	PC3	0	0	Yes
PC2	18	PC1	67	12	Yes
PC2	18	PC2	33	6	No
PC2	18	PC3	0	0	Yes
PC3	0	PC1	67	0	Yes
PC3	0	PC2	33	0	Yes
PC3	0	PC3	0	0	No
Total missed movers					61%

This example shows that an estimated 39% (27 + 12) of moves between SLA1 and SLA2 will be counted, or 61% (100 – 39) of moves between these SLAs will be missed, due to these SLAs sharing parts of postcodes.

As the index is symmetrical, it provides the same value for estimated moves in the other direction, ie. from SLA2 to SLA1.

Using the Missed Movers Index to adjust for data loss

The creation of a MMI means that the data loss due to intra-postcode moves can be quantified and thus rectified to some extent. The simplest way to do this is to use the MMI to upwardly adjust the number of movers predicted by Medicare. The following steps outline this method.

Step 1: Extract Medicare movement data by arrival and departure postcode.

Step 2: Convert the data to Medicare movement by arrival and departure SLA.

Step 3: For each SLA pairing, upwardly adjust the number of arrivals and departures by the corresponding MMI, i.e. adjusted arrivals = (original arrivals)/(1-MMI), and adjusted departures = (original departures)/(1-MMI).

However if the MMI for an SLA pairing move is high (say above 90%) then the small proportion of moves accounted for in the postcode-based data may not be a reliable base for adjusting to account for the unknown moves between these SLAs. Furthermore, if the MMI between two SLAs is 100% then there will be no Medicare movers recorded and another data source will have to be used to estimate movers between these SLAs. One source is population movement data from the most recent Census (data on place of usual residence 1 year ago).