|Page tools: Print Page Print All|
2. For the June 2000 estimated resident population (ERP), net overseas migration in 1999-2000 contributed 0.5% of the total population.
3. Persons arriving in, or departing from, Australia provide information in the form of incoming and outgoing passenger cards. Incoming persons also provide information in visa applications, apart from people travelling as Australian and New Zealand citizens. These and other information available to the Department of Immigration and Multicultural Affairs (DIMA) serve as a source of statistics on overseas migration.
4. DIMA is in the process of implementing an automated passenger card processing system which will be used for all movements after July 2000. However, data from the automated system will not be provided in time for the Australian Statistician's "before 10 June" determination of the 31 December 2000 population as required under the A New Tax System (Commnonwealth-State Financial Arrangements) Act 1999.
5. Given the unavailability of final data, special arrangements will be put in place with the cooperation of DIMA to estimate net overseas migration for the period July to December 2000. This information will then be used to determine the estimated resident population (ERP) at 31 December 2000 in accordance with ABS' statutory requirements. The special arrangements proposed by ABS are discussed in this document. Comments are invited before 22 April 2001 and should be forwarded to Matthew Berger at firstname.lastname@example.org or telephone (02) 6252 6639.
2. ABS DATA REQUIREMENTS
6. Due to the lack of final overseas arrivals and departures data for August and September 2000, ABS has not yet been able to produce quarterly population estimates at 30 September 2000. Therefore, in order to generate population estimates at 31 December 2000, overseas arrivals and departures data for both the September and December quarters 2000 are required.
3. SAMPLING OF PASSENGER CARDS
7. As a result of the delays in the new DIMA passenger card processing system, contingency plans for the manual data entry of passenger cards for August through December 2000 have been considered. The volume of data to be processed (around 8 million cards) and the short timeframe mean that data entry from every card is expensive and impractical. ABS considered a variety of sampling and/or projection methods to obtain the minimum subset of data required to determine net overseas migration, and thereby determine national and State ERP. Following advice from DIMA in February 2001 that final data for August through December 2000 would not be available before July 2001, these options were presented to DIMA to determine feasibility. The selected option is discussed in Section 5. Planned Method. Since passenger card data was provided to ABS under the old manual system for July 2000, overseas arrivals and departures data for this month would be added to that compiled for August through December 2000 to derive estimates of net overseas migration for the September and December quarters. In the absence of appropriate data, category jumping for these quarters will be assumed to be zero.
8. ABS plans to use information obtained from the DIMA Travel and Immigration Processing System (TRIPS) and samples of incoming and outgoing passenger cards for each month August through December 2000. The sample design and the resulting sample error will allow inferences about the population to be made from the estimated overseas migration figures obtained.
9. Estimates of net overseas migration for August through December 2000 based on a sample of data rather than complete enumeration will involve sampling error. Sampling designs can be made that set the maximum permissible error, expressed in terms of Standard Error (SE) or Relative Standard Error (RSE). The RSE of a sample is the standard error expressed as a percentage of the quantity being estimated. Due to the importance of the data and its intended use, a maximum RSE of 1% is planned.
Concept to be measured
10. The concept of interest is net overseas migration. For the purposes the current exercise, this is defined as the difference between permanent and long-term arrivals and departures and can be shown in the following equation:
= Net overseas migration;
= permanent and long-term arrivals;
= permanent and long-term departures; and
= category jumping (assumed nil).
11. Permanent and long-term arrivals and departures in each month would be considered as separate populations, with independent samples taken from each. Passenger card populations contain all movements, not just permanent and long-term. In this context, the variables of interest are a rare population, as illustrated in the table below. Between August and December 1999, only 3.4% of overseas arrivals and 2.3% of overseas departures were permanent or long-term movements.
12. The ABS also requires estimates of net overseas migration at the State level. This is a particular concern for the smaller States/Territories, such as the ACT and the Northern Territory. For example, only 0.1% of all arrivals in Australia in the period August through December 1999 were permanent and long-term movements to the ACT.
TABLE 1 - OVERSEAS ARRIVALS AND DEPARTURES BY STATE/TERRITORY, August-December 1999
Sampling methods considered
13. Random sampling and targetted random sampling of passenger cards were considered to determine net overseas migration at the State level. Random sampling methods considered included simple random sampling, stratified random sampling and cluster sampling. Targetted random sampling involves specifically targetting the cards to be sampled using existing information from the DIMA's TRIPS system.
14. To achieve a 1% RSE, very large sample sizes would be required to determine State level estimates under each of the random sampling methods and the targetted sampling method. A random sampling design close to complete enumeration (around 7.7 million cards) would be required, and a sample of around 300,000 passenger cards could be required under a targetted random sampling design. DIMA advised that the scale of either of these programs within the permitted timeframe was not feasible.
15. A third method was proposed that again relies on specifically targetting the passenger cards to be sampled, but aims for reliable Australia level estimates as opposed to State level estimates. At a 1% RSE level, this would require a sample of around 45,000 cards and is the option that has been adopted.
4. ASSUMPTIONS OF NET OVERSEAS MIGRATION
16. The ABS has also considered the possibility of using an assumed level of net overseas migration for each State and Territory to determine the 31 December 2000 population. Since it is possible to obtain actual estimates based on sampled data, the use of assumed levels remains a contingency measure. If used, these assumed levels would only apply to the five months August through December 2000.
5. PLANNED METHOD
17. It is proposed that ABS produce estimates of net overseas migration for August to December 2000 based on a targetted random sample of passenger cards to produce Australia level estimates of net overseas migration.
18. The numbers of first arrivals on migrant visas, Australian residents returning after an absence of 12 or more months overseas and visitors departing after a stay of 12 or more months in Australia will be obtained from the TRIPS system and become direct components of estimated net overseas migration.
19. Visitor arrivals and resident departures that have not yet been cancelled out by a return movement on TRIPS will be sampled. Passenger cards will be examined for the sampled arrivals and departures to determine the number of movements relating to an intended period of 12 months or more.
20. Because a feasible sample size will only allow national estimates, State level estimates are proposed to be based on each State's share of the national total for the August-December period over the last three years.
21. ABS investigations have shown that, based on the number of cards received between August and December 1999, a sample size of around 45,000 cards would be required. DIMA has indicated that this option is feasible, albeit difficult to achieve within the required timeframe.
Australian Bureau of Statistics
22 March 2001
APPENDIX A Results of sampling project - what was achieved
6. RESULTS OF PASSENGER CARD SAMPLING
22. The sampling project described in this document was implemented in April 2001. The sampling scheme achieved the following results.
TABLE 2 - ESTIMATED COMPONENTS OF NET OVERSEAS MIGRATION FROM PASSENGER CARD SAMPLING SCHEME, August-December 2000
7. ISSUES WITH IMPLEMENTATION
23. During implementation it became apparent that several modifications needed to be made to the original design. There were two areas where the design needed to be modified to reflect availability of data prior to sampling. Firstly, account needed to be taken of non-resident Australian citizens - that is, people living overseas but travelling to Australia on an Australian passport. Under the original design specification it was assumed that all Australian citizens arriving in Australia after an absence of 12 or more months would be residents returning. Analysis of previous years' data showed that around 11,000 people a month arrived in Australia on an Australian passport after an absence of 12 or more months who intended to visit Australia, rather than take up residence. The sampling frame needed modification to enable passenger cards from these people to be sampled and classified according to their intended duration of stay.
24. The second modification that was required concerned New Zealanders. Since it is possible for New Zealand citizens to reside permanently in Australia without applying for a visa, it was not possible to classify movements of New Zealand citizens as settlers, residents or visitors prior to the sampling of passenger cards. The sampling framework needed to be modified so that movements of New Zealanders were considered on a separate basis, and they were sampled at a different rate from other international movements.
25. The key consequence of these changes to the proposed sample design meant that the maximum sample of 43,400 cards would no longer achieve a design standard error of 1%. This is reflected in the sample errors presented in Table 2 above. The sample errors achieved at the national level were considered suitable for the purpose of deriving preliminary estimates of the resident population for December 2000.
Australian Bureau of Statistics
7 June 2001
These documents will be presented in a new window.