FEATURE ARTICLE 3: COMPARISON OF NET OVERSEAS MIGRATION AND ESTIMATES FROM THE 2011 CENSUS
In recent years, the change in net overseas migration (NOM), particularly that of temporary migration, has been the main contributor to variability in population growth over time. For the intercensal period 2006 to 2011, the ABS used an improved methodology for calculating the NOM component of population growth. The '12/16 month rule' method was introduced to better measure the changes in traveller behaviour and more accurately capture and measure temporary migration.
During consultation related to the rebasing cycle, the ABS identified some concern among some users about a perceived difference in the growth reflected in NOM and the growth reflected in the change between Census-based population estimates, particularly for the overseas-born population. In particular, there was concern over the contribution of any differences in the coverage of overseas-born population to intercensal error and intercensal discrepancy (for more information on these terms refer to the Feature Article: Final Rebasing of Australia's Population Estimates, September Quarter 2006 - June Quarter 2011, found earlier in this publication).
The purpose of this paper is to present the findings from three analyses that were undertaken by the ABS to determine whether there were any substantial differences between NOM estimates and Census-based estimates, including any contributions from changes in methodologies. These analyses each focus on the change in the overseas-born population, from 2006 to 2011, as reflected in NOM and the Census-based population estimates.
ANALYSIS 1 - INTERCENSAL DIFFERENCE FOR OVERSEAS-BORN PERSONS - FACTORING IN THE STATISTICAL IMPACT OF ADL
The first analysis focused on exploring whether there was a large intercensal discrepancy for the overseas-born population, compared to the Australian-born population. This analysis took the form of estimating the overseas-born and Australian-born populations at each point in the rebasing process for 2006 and 2011, and comparing the outcome over the intercensal period, as shown in Table 1.
The first section of Table 1 provides information on the components of population change for overseas-born and Australian-born persons. It shows that the ABS estimated that the overseas-born population changed by around 1,012,400 persons over the intercensal period - that is, 1,227,000 from net overseas migration, less the 214,600 deaths of overseas-born persons that occurred over the period.
The second section of the table provides ERP for the overseas-born and Australian-born persons as at 30 June 2006 and 30 June 2011. It includes an itemised account of the rebasing sources used to derive 30 June 2011 estimates, on a country of birth basis. This showed an increase of 934,000 overseas-born persons over the intercensal period, from 5,090,100 in 2006 to 6,024,000 in 2011.
The third section of the table provides a comparison of the increase for the overseas-born population observed in the components of population change (1,012,400) and the increase observed in the change between 30 June 2006 ERP and 30 June 2011 ERP (934,000). This revealed that the components had estimated 78,400 more overseas-born persons than was reflected in the change in ERP bases.
However, one major component of each rebasing process is an adjustment for undercount in the Census, based on the results of the Census Post Enumeration Survey (PES). For the 2011 PES, the ABS introduced a new linking and matching methodology known as Automated Data Linking (ADL). This improved methodology was estimated to have produced an undercount estimate that was around 247,000 persons less than what the previous methodology would have produced.
The ADL Statistical Impact Study also estimated that around 18% of this change related to the estimate of undercount for the overseas-born population, while most of the change (82%) related to the Australian-born population. When this impact is factored into the analysis, as was done in the recasting, where the previously published 30 June 2006 estimate was reduced by around 247,000 persons, the difference between the increase in the components and the change in the ERP bases is reduced by 45,100 (from 78,400 to 33,300 persons).
The ABS therefore, estimates that of the total 77,700 intercensal discrepancy (if there had been no improvement made in 2011 to measuring net undercount), only around 33,300 relates to overseas-born persons and around 44,300 to Australian-born persons. Neither of these estimates is large enough to suggest that there is a significant intercensal discrepancy that would warrant an additional rebasing adjustment. For instance, the intercensal discrepancy for overseas-born persons only represents around 3% of the increase reflected in the components and reflects a combination of error in the 2006 or 2011 Census-based population estimates and/or in the estimates of births, deaths and migration over the intercensal period. The 33,300 intercensal discrepancy also accounts for less than 0.6% of the estimate of overseas-born ERP at 30 June 2011.
ANALYSIS 2 - CONSIDERING THE IMPACT OF THE CHANGE IN NOM METHODOLOGY
Table 1. Intercensal discrepancy by country of birth - 2006 to 2011
|Section 1: Components of Population Change 2006-2011 |
|Net overseas migration (12/16 month rule) |
|Total increase in the components of population change 2006-2011 |
|Section 2: ERP at 30 June 2006 and ERP at 30 June 2011 |
|ERP at 30 June 2006 |
|ERP at 30 June 2006 (prior to recasting) |
|ERP at 30 June 2011 |
|Census count, place of usual residence as at 9 August 2011(a) |
|PES net undercount including gain from ADL(b) |
|Demographic adjustment |
|Residents temporarily overseas |
|Backdating components to 30 June 2011 |
|less - Births |
|plus - Deaths |
|less - Net overseas migration (12/16 month rule) |
|ERP at 30 June 2011 |
|Section 3: Comparing Census-based estimates with component estimates - 2006 to 2011 |
|Increase between the 30 June 2006 ERP and 30 June 2011 ERP |
|Total increase in the components of population change 2006-2011 |
|Intercensal discrepancy - with ADL gain included in undercount adjustment(c) |
|PES net undercount with gain from ADL removed(d) |
|Intercensal discrepancy - with ADL gain excluded from undercount adjustment(c) |
|- nil or rounded to zero (including null cells) |
|(a) These Census counts exclude the 1,195,728 persons for which Country of birth was not stated. |
|(b) Census net undercount from the 2011 Post Enumeration Survey (PES) which includes the improved method of automated data linking (ADL). The PES undercount adjustment includes an imputed value for 1,195,728 persons for which country of birth was not stated. An estimate of 419 persons for Other Territories is also included. |
|(c) Intercensal discrepancy equals the difference between Census-based ERP and component change ERP. |
|(d) Census net undercount from the 2011 Post Enumeration Survey (PES) with the the impact of automated data linking (ADL) removed. The PES undercount adjustment includes an imputed value for 1,195,728 persons for which country of birth was not stated. An estimate of 419 persons for Other Territories is also included. |
The second analysis explored the impact of the change in NOM methodology over the intercensal period, to consider what the intercensal discrepancy would have been, had the '12/12 month rule' continued to be used and not been replaced by the '12/16 month rule'.
In conceptual terms, NOM is based on an international travellers' duration of stay being in or out of Australia for 12 months or more. With the introduction of the '12/16 month rule' method for estimating NOM, this 12 months does not have to be continuous and is measured over a 16 month reference period. That is, whether a traveller is in, or out of, the population is determined by their exact duration of stay in, or away from, Australia over the subsequent 16 months after arrival or departure.
Prior to 1 July 2006, NOM estimation methods used a '12/12 month rule' to determine if a traveller contributed to ERP. This meant that in order for a person to contribute to NOM and therefore ERP they must stay in, or be absent from, Australia for a continuous period of 12 out of 12 months.
A statistical impact analysis undertaken by the ABS compared the two methods over a three year period (December quarter 2003 to September quarter 2006). It showed the current method estimate to be on average 25% higher than the previous estimate (or, conversely, the previous method to be on average 20% lower than the current method). For further information on the '12/16 month rule' methodology, see:
Using the results from the aforementioned statistical impact analysis it is possible to derive an illustrative estimate of what the previous method would have produced - a NOM estimate of 949,100. As shown in Table 2, had this alternative NOM estimate been used to estimate the change in the population over the 2006-2011 intercensal period, the intercensal discrepancy would have been much larger, in absolute terms (159,600, compared to 77,700 under the current '12/16 month rule'). That is, the intercensal discrepancy would have been considerably further from a zero value, which is the optimum outcome for any accounting framework.
The difference between the two methods is particularly noticeable in the intercensal discrepancy for the overseas-born population (212,100 and 33,300 in absolute terms). This demonstrates that the NOM estimates produced by the '12/16 month rule' align much closer with the increase in the Census-based ERP from 2006 and 2011 than the '12/12 month rule), further reinforcing the value in the methodological change.
ANALYSIS 3 - COMPARISON OF RECENT MIGRANTS IN NOM AND CENSUS DATA
Table 2. Intercensal Discrepancy by Country of Birth, Using assumptions for NOM 12/12 month rule - 2006 to 2011
|Net Overseas Migration |
|12/12 month rule(a) |
|12/16 month rule |
|Intercensal discrepancy(b) |
|12/12 month rule(a) |
|12/16 month rule |
|(a) Based on previous analysis, an assumption is being made to produce a synthetic estimate of the '12/12 month rule' method for NOM (i.e. 20% less than the current '12/16 month rule' method). |
|(b) Intercensal discrepancy equals the difference between Census-based ERP and component change ERP. This measure excludes the ADL gain to net undercount adjustment, for comparability with Table 1. |
While the first and second analyses compared changes in the levels of overseas-born persons, the sub-population of most interest are recent migrants to Australia. It is difficult to make a direct comparison with how recent migration is captured by the Census-based estimates and how it is measured in NOM, since Census data does not collect the cross-border movement information required to establish whether a person satisfies the '12/16 month rule'.
However, while noting this particular limitation, the ABS undertook a comparison using Census year of arrival (where year of arrival information indicated that migration to Australia had occurred during the July 2006 to June 2011 period) and country of birth information.
Where year of arrival was not stated in the Census, a prorated adjustment was made, with a related prorated adjustment also made for Census counts with a not stated country of birth. In addition, the ABS was able to estimate recent migrants (i.e. persons arriving from 1 July 2006) who were temporarily overseas on Census night and include them in the data. No adjustment was made for Census undercount for recent migrants, given it would have been small and there was no direct estimate available.
The difference between the Census-based estimate of recent migrants and the equivalent NOM estimate was 19,300 persons. While there were more persons in the Census-based estimate than in NOM at the total level, this varied by country of birth, though not to any great extent, as is evident in Graph 1. This should also provide further reassurance that the population is being consistently measured in both data collections.
Graph 1. Recent Migrants,
by country of birth(a) -
The results of the three analyses that are summarised in this article all suggest that there is a relatively close alignment between Census-based population estimates and NOM estimates under the '12/16 month rule', for both the total overseas-born population, as well as those who had recently migrated to Australia. As such, there was no identified requirement for any additional adjustment to be made during the rebasing of population for the 2006 to 2011 intercensal period.
The ABS will continue to monitor the quality of NOM estimation on a regular basis and identify further methodological improvements, particularly at the state and territory level. The analyses summarised in this article will be repeated during the 2016 rebasing cycle, together with investigations involving additional data sources, as they continue to become available.