3137.0 - Demography Working Paper 2003/5- Net Overseas Migration: Adjusting for Actual Duration of Stay or Absence, 2003  
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 11/12/2003   
   Page tools: Print Print Page Print all pages in this productPrint All

SUMMARY

1 This paper describes the method for estimating Australia's net overseas migration (NOM) through an adjustment for actual duration of stay either in or out of Australia. This paper describes the state of stay distribution that depends on the type of traveller and the movement category from which they are changing from. A preliminary estimation model is also presented in this paper.

INTRODUCTION

2 The Australian Bureau of Statistics (ABS) derives quarterly estimates of Australian population (known as Estimated Resident Population or ERP) by the cohort component method,

Pt = Pt-1 + B - D + NOM, where

Pt refers to the population at reference point t, Pt-1 existing population estimate at time point t-1, B number of births which occurred between t-1 and t, D number of deaths which occurred between t-1 and t and finally NOM number of net overseas migration between t-1 and t .

3 NOM is calculated from incoming and outgoing passenger movements at Australian ports maintained by the Department of Immigration and Multicultural and Indigenous Affairs (DIMIA). Overseas arrivals and departures are classified into three main categories of movements depending on stated duration of stay in Australia or overseas:

  • permanent movement,
  • long-term (one year or more) movement and
  • short-term (less than one year) movement.

4 Originally all long-term and permanent movements as stated on the first leg of travel were included in preliminary ERP. In revised ERP, calculations of NOM need to be adjusted for any changes in the duration of stay of travellers whose original intention was to move in or out of Australia's population for more or less than 12 months, but had changed prior to the next border crossing. For example, some visitors arriving may state that their intention is to stay in Australia for more than 12 months. However, they may change their travel plans and depart the country after an actual duration of only 6 months. Since migration figures are affected by this change in travel behaviour, an adjustment is required to revise the net overseas migration estimate and ERP.

5 As a legislative obligation, the ABS is required to provide a preliminary ERP for each December quarter by early June of the following year. The Australian government uses ERP to distribute funds across all states and territories, and to develop a wide range of government policies. As a component of ERP, the ABS has published two sets of NOM estimates, preliminary and final. Since final estimates can only be calculated up to 15 months after a reference quarter, preliminary estimates are calculated to meet immediate requirements.

6 Results from a new estimation method which ABS has recently used to derive revised estimates of NOM and ERP back to September quarter 2001 are presented in this working paper. This new technique updates the method documented in Demographic Estimates and Projections: Concepts Sources and Methods (ABS Cat 3228.0).

BACKGROUND

7 The estimating equation for NOM involves an adjustment which until recently was called a 'category jumping' adjustment by the ABS. This adjustment is necessary since some travellers choose to change their stated length of stay or absence, misinterpret the passenger card and other reasons. Any changes into or out of long-term migration impacts directly on the measurement of ERP. More details on the methodology of estimating NOM is presented in estimation of revised net overseas migration for 2001-02 section later in this paper.

8 In the past, ABS estimated category jumping as a net difference between two components, called overseas visitor arrival and Australian resident departure components.
  • The overseas visitor arrival component, which increases ERP, was estimated by taking a difference between the number of short-term visitors who entered Australia in a reference quarter, and the number of visitors who departed Australia over the year following the reference quarter but who had arrived in the reference quarter.
  • The resident component, which decreases ERP, was estimated by taking a difference between the number of short-term Australian residents departing in a reference quarter and the number of Australian residents who returned back to Australia over the year following the reference quarter but who had departed in the reference quarter.

9 This method had a few limitations. Firstly, it was based on gross flow of traveller movements rather than individual travellers. Secondly, and more importantly, until June 1998 the measurement of duration of stay or absence on the second leg of travel was based on passenger reporting on the arrival or departure card. This self reported duration, which itself was subject to reporting error, was used to determine the time at which a person arrived (for visitors) or left Australia (for Australian residents). However, from July 1998 onwards, implementation of a new passenger card design and preliminary systems enabled DIMIA to derive actual duration of stay or absence by matching both arrival and departure cards rather than relying on passengers reporting their duration of stay or absence.

10 Despite this improvement in the quality of actual duration of stay or absence data, the above estimation method appeared not capable of producing acceptable estimates of category jumping. The failure was primarily related to an increased number of international movements and their impact on the components of category jumping. For example, with increased numbers of visitor arrivals to Australia in recent years, a small change in visitor departure pattern can have a large impact on the size of category jumping estimate. Given that category jumping constituted only a small fraction of ERP and that the recent estimates produced by the above method seemed highly volatile, ABS decided to set category jumping estimate to zero from September quarter 1997 onwards until a better estimation technique was developed (more details on issues surrounding category jumping estimation are available in the Demography Working Paper 2003/1 - Estimated Resident Population and Measurement of Category Jumping).

11 Through the provision of additional data from DIMIA, the ABS now has the ability to match traveller movements over time. This enables a movement history to be constructed for those arriving and departing in a particular quarter over the next 12 months and thus produce an actual duration of stay.

ESTIMATION OF REVISED NET OVERSEAS MIGRATION FOR 2001-02

12 The ABS now publishes two net overseas migration figures and is phasing out the use of category jumping. Revised NOM is produced up to 15 months after the reference quarter with an adjustment for the actual duration of stay or absence of a traveller. Preliminary NOM contains primarily traveller movements data modelled by an estimate of travellers who will change their intended duration of stay or absence.

13 Using matched passenger records this estimation method looks at the travel sequence (or movement history) of each person over a period of exactly one year and then calculates actual duration of stay in that one-year window. A person is added to ERP if they reside in Australia for more than 12 months (also known as the 12 month residency rule). Thus if a traveller changes their duration of stay, they may be added or subtracted from NOM depending on the circumstance. Generally a visitor who changes from a long-term to a short-term duration is subtracted from NOM. A visitor who changes from a short-term to a long-term duration is added to NOM. In the case of Australian residents, the opposite situation is needed.

14 There are four main groups of travellers who provide an intended duration on the passenger card who have the potential to change their duration of stay or absence:
  • Long-term overseas visitors stayed short-term in Australia;
  • Short-term overseas visitors who stayed long-term in Australia;
  • Long-term Australian residents who stayed short-term overseas; and,
  • Short-term Australian residents who stayed long-term overseas.

15 Table 1 provides estimates of the adjustment to NOM for actual duration of stay or absence during the 2001-02 financial year. An estimate with a negative sign indicates that these people are taken out of NOM. Similarly an estimate with a positive sign indicates that these people are added to NOM.


TABLE 1 Estimates of adjustment for actual duration by category of movement, 2001-02

Category of movement
Number of travellers
Adjustment for actual duration of stay or absence to NOM
Proportion of total travellers
no.
no.
%




Long-term visitor arrivals
166,763
-123,779
-74.2
Short-term visitor arrivals
4,317,763
182,494
4.2
Long-term resident departures
89,772
45,837
51.1
Short-term resident departures
2,981,179
-122,400
-4.1

16 People who arrive permanently as migrant and Australian residents who depart permanently for overseas are not required to provide any intended duration on the passenger card. However, in its adjustment to derive final estimates for NOM, ABS has considered two small groups of people, the first being taken out of ERP and the second being added back to ERP. These groups are:

  • Permanent migrants who arrived and left Australia in the same quarter without any return in a one-year period, and
  • Australian residents who left permanently but returned back in the same quarter without any further departure within a one-year period.

17 Table 2 shows the components of net overseas migration compared to the preliminary movements based data previously published. Apart from the adjustment for actual duration for the groups of travellers above, the NOM adjustment in the table also includes an element of adjustment for people being added to or taken out of the ERP multiple times in the same quarter. The total size of the adjustment is about -23,100 over four quarters. The revised NOM estimate is approximately 110,000 for the 2001-02 financial year, down from the previously published NOM of 134,000.

TABLE 2 Components of revised net overseas migration, 2001-02

Migration components
September 2001 quarter
December 2001 quarter
March 2002 quarter
June 2002 quarter
2001-02






Permanent arrivals
22,833
21,348
22,163
22,556
88,900
Permanent departures
11,632
11,222
14,449
10,938
48,241
Net permanent migration
11,201
10,126
7,714
11,618
40,659
Long-term arrivals
65,926
59,021
91,783
47,741
264,471
Long-term departures
41,477
41,695
48,909
39,365
171,446
Net long-term movements
24,449
17,326
42,874
8,376
93,025
Published NOM(a)
35,650
27,452
50,588
19,994
133,684
NOM adjustment
-7,938
3,737
-14,233
-4,694
-23,128
Revised NOM
27,712
31,189
36,355
15,300
110,556

% change between revised and published(a) NOM
-22.3
13.6
-28.1
-22.5
-17.3

(a) Previously published in March quarter 2003 Australian Demographic Statistics (cat. no. 3101.0)



STATE AND TERRITORY DISTRIBUTION OF REVISED NET OVERSEAS MIGRATION

18 State distribution of adjustment for revised NOM is complex. The 'Australian resident departing temporarily' box (or box E) of outgoing passenger cards provides the state of usual residence for both short-term and long-term Australian residents. The incoming passenger card provides the state of intended stay in Australia for both short and long-term visitors. Since long-term visitors are added to the population of the state of intended stay, and changes from first stated need to be taken out of the state and territory ERP to which they have been added in the first instance.

19 The short-term visitors are not added into ERP. Therefore, addition of short-term visitors whose actual duration of stay is more than 12 months to the state or territory in which they spend most of their time in Australia seems to be a sensible solution. However, since this group of travellers did not make any subsequent movement in the one-year window of reference, state in which they spent most of their time is not known. Assuming that they would behave similarly as those who have left before the end of a 12 month stay, a state and territory of actual stay distribution of short-term visitors leaving Australia between six months and one year arrival was applied to this component of the adjustment. The NOM adjustment by quarter and state or territory are presented in table 3.

20 As apparent in table 3, Victoria had the highest loss in population through the adjustment for actual duration, which was followed by New South Wales. Victoria's preliminary ERP is reduced by about 14,000 and NSW by about 7,000. Queensland and Northern Territory are the only two state/territory which gained more people through the adjustment for actual duration.

TABLE 3 Adjustment to net overseas migration, state/territory distribution, 2001-02

NSW
VIC
QLD
SA
WA
TAS
NT
ACT
Australia(a)









September quarter 2001
-2,701
-5,136
1,239
-610
-496
-104
108
-238
-7,938
December quarter 2001
1,987
-781
1,766
283
11
148
198
125
3,737
March quarter 2002
-3,703
-6,067
-2,028
-1,166
-1,276
-98
206
-100
-14,233
June quarter 2002
-2,512
-2,013
645
-190
-576
-121
-1
74
-4,694
2001-02
-6,929
-13,997
1,622
-1,683
-2,337
-175
511
-139
-23,128

(a) includes Other Territories


PRELIMINARY ESTIMATE OF NET OVERSEAS MIGRATION, 2002-2003

21 The preliminary estimates of NOM are required to derive timely preliminary ERP. During preliminary estimation, not all data required for a final estimate of NOM are available. A model based approach seems to be a suitable option for the preliminary estimation of NOM. Analysis of six quarters of data, beginning from March quarter 2001 to June quarter 2002, have shown a somewhat stable relationship between the number of travellers in each of the four major movement categories and the number who changed their actual duration of stay or absence during the revised NOM estimate (more detail of these relationships are in the case for no preliminary adjustment to net overseas migration section below). Under the assumption of similar patterns, these relationships were used to derive preliminary NOM estimates each of the quarters in 2002-03 (table 4).

22 The overall preliminary estimate of the adjustment is approximately -29,000 for the 2002-03 financial year. As more data becomes available, the relationships used in the model are subject to revision and so are the preliminary estimates of NOM. Only the four major groups of travellers have been considered to derive the preliminary estimates, excluding minor changes to permanent migration (as previously mentioned above).
TABLE 4 Preliminary estimates of net overseas migration, 2002-03

Category of movement
Estimate of adjustment for actual duration
Gross traveller movement
Proportion of gross movements
no.
no.
%




Long-term visitor arrival
-128,942
184,095
70.0
Short-term visitor arrival
177,595
4,655,803
3.8
Long-term resident departure
42,664
86,211
49.5
Short-term resident departure
-120,247
3,293,337
3.7



CASE FOR NO PRELIMINARY ADJUSTMENT TO NET OVERSEAS MIGRATION

23 One possible option for ABS is to continue setting the preliminary adjustment to NOM as zero and then make an adjustment only during the usual revision of ERP up to 15 months from a reference quarter. However, table 5 shows that the proportion of travellers who had changed their duration of stated stay or absence have had small variations between quarters for each traveller category. For example, about 70% of all long-term visitors were taken out of preliminary NOM and about 50% of long-term Australian residents were added back to preliminary NOM. Such a stable relationship can easily be used to derive a modelled estimate of an adjustment to NOM.

24 The preliminary estimates of adjustment for actual duration in table 4 are based on the average relationships observed in table 5 for the most recent four quarters. The number of quarterly gross traveller movements in each movement category was multiplied by corresponding average proportion in table 5 to derive a quarterly estimate of an adjustment for actual duration of stay or absence. These quarterly estimates have then added to NOM for the whole 2002-03 financial year. It is a fair assumption that a model based preliminary estimate based on recent and real data is more likely to be a better option than setting a zero estimate considering the gross number of movements each financial year.

TABLE 5 Percent of travellers who changed from stated duration of stay or absence, by category of movement and quarter

Category of movement

Long-term visitor arrival(%)
Short-term visitor arrival(%)
Long-term resident departure(%)
Short-term resident departure(%)




March quarter 2001
67.3
5.1
51.8
4.2
June quarter 2001
67.0
4.4
49.2
3.7
September quarter 2001
70.6
4.1
49.2
3.7
December quarter 2001
69.3
3.6
47.8
3.7
March quarter 2002
71.3
4.1
52.3
4.1
June quarter 2002
69.0
3.5
48.7
3.2
Average (4 most recent quarters)
70.0
3.8
49.5
3.7



METHODOLOGICAL NOTES

25 NOM is simply the difference between two quantities: permanent and long-term arrivals, and permanent and long-term departures. In simplest mathematical form, it can be written as

NOM = A - D .... (i),
A = Arrivals
D = Departures

26 ABS uses monthly passenger card data supplied by DIMIA to estimate NOM. On the incoming passenger card, each overseas visitor is asked to state a duration he/she intends to stay in Australia. Similarly, on the outgoing passenger card, each Australian resident or citizen is asked to state a duration they intend to be overseas. This intended duration of stay in Australia (for visitors) or absence from Australia (for residents) affect the size of both A and D through their third component mentioned below.

27 There are three elements comprising A (Arrivals):
  • Persons arriving to Australia with a permanent resident visa;
  • Australian residents returning home after one year or more since last departure (these people were taken out of the ERP at their departure); and,
  • Visitors who state in the passenger card that their intention to stay in Australia is for one year or more.

28 Similarly, three elements comprise D (Departures):
  • Australian residents or citizens departing permanently to live overseas;
  • Long-term visitors departing after a stay of one year or more (these people had been added to ERP on their arrival); and,
  • Australian residents departing who state on the passenger card that their intention to live in overseas (or to be away from Australia) is one year or more.

29 At the time a person crosses Australian boarder, it is not empirically known how long the person would actually spend in Australia or overseas. The person's actual behaviour, which can only be determined after a waiting of at least one year, could be different from what he/she intended at the first leg of the movement. Equation (i) above therefore needs an extra component which would reflect a change in intended behaviour from actual behaviour. NOM has therefore been estimated by the equation

NOM = A - D + (Adjustment for actual duration of stay or absence).... (ii).

DISCUSSION

30 The estimation method employed here is a major improvement to the method previously used by the ABS (which primarily was a residual approach). The new method is based on:

  • a direct calculation of the time spent by each person;
  • persons rather than movements; and,
  • population of movement records rather than a sample of records (for short-term travellers only).

31 The method also enables an adjustment for people who have been added to or taken out of the ERP multiple times. However, there are two specific concerns in this analysis.

  • The method removes, through its adjustment for actual duration, long-term visitors with student or temporary business entry visa from the ERP who in fact are residing in Australia but are making overseas travel for short periods during their overall period of residency in Australia. Although there could be strong argument to retain these people into the ERP, this estimation technique does not allow to do so unless a change in duration of residence is adopted.
  • About 2% of travellers were found to have implausible movement sequence, for example, no arrival record between two departures or no departure record between two arrivals. These people could not be included in this analysis.

32 In this analysis, the ABS has proposed a NOM adjustment of -23,100 for 2001-02. While an independent analysis conducted by McDonald et al. (2003) proposed an adjustment of -37,700 for the same period, the two estimation methods are not strictly comparable.

33 McDonald et al. combine DIMIA stock data (based on student visas, business visas and a number of other small, temporary long-term visa types) with preliminary data on net overseas migration published by the ABS. While this approach is likely to provide a better estimate of some category jumping components (e.g. long-term students or business people on long-term temporary visas) than the ABS method, it does not take into account the short-term resident component of the category jumping adjustment.

34 The ABS is continuing to develop methods to adjust estimates of NOM for actual durations of stay or absence. The estimation methods described in this paper will be used until better alternatives are developed. Preliminary estimates of NOM will be revised annually, on a financial year basis.


BIBLIOGRAPHY

Australian Bureau of Statistics, Demographic Estimates and Projections: Concepts Sources and Methods (cat. no. 3228.0), 1995.

Demography Working Paper 2003/1 - Estimated Resident Population and Measurement of Category Jumping.

McDonald, P., Khoo, S-E and Kippen, R. An Alternative Net Migration Estimates for Australia: Exploding the Myth of a Rapid Increase in Numbers, Working Papers in Demography No. 89, The Australian National University, Canberra, 2003.


For more information please refer to Australian Demographic Statistics Quarterly (cat. no. 3101.0), June quarter 2003 released on 11 December 2003.