1001.0 - Annual Report - ABS Annual Report, 2002-03  
Previous ISSUE Released at 11:30 AM (CANBERRA TIME) 10/10/2003   
   Page tools: Print Print Page RSS Feed RSS Bookmark and Share Search this Product  
Contents >> Section 2 - Special Articles >> Chapter 4 - Population Estimates: Importance, Complexity and Controversy

INTRODUCTION

One of the core functions of any national statistical agency is to provide regular estimates of the population, and the Australian Bureau of Statistics (ABS) is no exception. Indeed it is legislated for under the Census and Statistics Act 1905 which requires the Statistician to conduct a census every five years and to compile 'statistics of the number of people of each state as on the last day of March, June, September and December in each year'.

Population estimates are essential inputs to many aspects of decision making by the government, business and the community. At the Commonwealth level two key uses of the population estimates are the determination of the number of members for each state and territory in the House of Representatives and grants to the states and territories.

The critical importance of the population statistics can be illustrated by some of the media reports in response to the release of the preliminary September 2002 population estimates in February 2003. These estimates were used by the Australian Electoral Commissioner for determining the number of House of Representatives seats for each state and territory.

    'The Territory may lose one of its two seats in Federal Parliament tomorrow because it is just 267 people short of the required quota'
Northern Territory News, page 3, 19 February 2003
    'The Northern Territory’s shrinking population has cost it one of its two seats in the Federal Parliament’s lower house... South Australia will also lose a House of Representatives seat, while Queensland will gain one'
The Sydney Morning Herald, page 3, 21 February 2003

The loss of Northern Territory’s second seat in the House of Representatives has resulted in some public questioning of the accuracy of the ABS estimates.

This article provides background to the process and data sources for compiling the estimates of population for the states and territories and explains the limitations of those estimates in terms of potential sources of error. The article also explains the legislative background to the compilation of the population estimates particularly as they relate to the determination of electoral representation and grants distribution.


LEGISLATIVE REQUIREMENTS

Electoral representation

The Commonwealth Electoral Act 1918 requires the Electoral Commissioner to ascertain the numbers of the people of the Commonwealth and of the several states and territories in accordance with the latest statistics of the Commonwealth in the 13th month after the first sitting day of a new House of Representatives. Section 47 of the Commonwealth Electoral Act also places obligations on the Australian Statistician to provide statistical information to the Electoral Commissioner on request:

'47. The Australian Statistician shall, on request by the Electoral Commissioner, supply the Electoral Commissioner with all such statistical information as he or she requires for the purposes of this Division.'

These provisions are pursuant to section 24 of The Constitution which requires that the number of members chosen in the several states in the House of Representatives 'shall be in proportion to the respective numbers of their people'.

Grant distributions

The 31 December population estimates for each state and territory are required by the Commonwealth Treasury under A New Tax System (Commonwealth-State Financial Arrangements) Act 1999 for the purposes of determining the distribution to the states and territories of revenue from the goods and services tax using relativities prepared by the Commonwealth Grants Commission. These data are also used to distribute financial assistance grants to the states and territories for local government purposes under the Local Government (Financial Assistance) Act 1995.


ESTIMATED RESIDENT POPULATION

The conceptual base

The fundamental concept of measuring Australia’s population is the Estimated Resident Population or ERP. The ERP is the official measure of the population of Australia and is based on the concept of residence for a period of 12 months or more within Australia, regardless of nationality or citizenship, with the exception of foreign diplomatic personnel and their families. It excludes people who are overseas for more than 12 months and overseas visitors who are in Australia for less than 12 months.

The ERP concept was developed in the late 1970s reflecting the increasing mobility of the population, both internationally and within Australia. Prior to the introduction of the ERP, all overseas arrivals and departures were included in the estimation process regardless of duration of stay. However increasing volumes of international passenger movements (Australian residents as well as overseas visitors) introduced increasing volatility in post census estimates of the Australian population. Furthermore, with increased travel within Australia, the use of Census of Population and Housing counts on a place of enumeration basis meant that states and territories such as Queensland, the Northern Territory and Western Australia were advantaged while Victoria, South Australia and Tasmania were disadvantaged due to the fact that the census is traditionally conducted in winter. For electoral representation purposes it was important that population estimates changed to a place of usual residence basis.

Methodology

In Australia, estimates of the population of each state and territory are calculated as at the last day of March, June, September and December. These estimates are compiled using:
  • data collected in the most recent Census of Population and Housing
  • birth and death statistics from state and territory registrars
  • overseas arrivals and departure data from the Department of Immigration and Multicultural and Indigenous Affairs
  • Medicare data on changes of address from the Health Insurance Commission which are used as the basis for modelling interstate migration
  • changes in state and territory defence force levels not accounted for in Medicare changes of address.

The five-yearly Census of Population and Housing provides the benchmark for population estimates. The results of the census are adjusted to take account of:
  • persons away from home on census night
  • the need to exclude overseas visitors
  • underenumeration of the population in the census
  • residents temporarily overseas on census night.

As the census is conducted in August the estimates are then backdated to 30 June using births, deaths, and net migration.

For subsequent periods the estimates are incremented from the census benchmark using estimates of natural increase (the excess of births over deaths), net overseas migration and net interstate migration in the intervening period.

Every five years new benchmark estimates are available from the census and it provides an opportunity to compare the estimates based on the previous census with the estimates from the most recent census. The difference in these two estimates is known as Intercensal Error. Table 1 illustrates the level and relative impact of the intercensal error for the last four intercensal periods. For 1996-2001, the intercensal error for Australia was minimal (0.05%), although the error varied from state to state with the Australian Capital Territory having the highest absolute percentage error (1.36%) and the Northern Territory the lowest (0.11%). The intercensal error provides the basis for revising the estimates for the period between the previous and current censuses.

Table 1: INTERCENSAL ERROR

1981-1986
1986-1991
1991-1996
1996-2001

PERSONS (’000)

New South Wales
11.9
2.4
-15.8
-27.0
Victoria
3.9
7.0
-20.9
35.6
Queensland
-32.0
11.1
15.9
5.8
South Australia
-9.4
10.4
4.5
-9.6
Western Australia
-18.4
29.9
-3.3
11.5
Tasmania
0.4
-6.3
-1.1
-1.8
Northern Territory
-6.3
-6.7
-5.2
-0.2
Australian Capital Territory
5.5
4.2
-1.2
-4.3
Australia
-44.4
51.9
-27.5
10.6

PER CENT OF POPULATION AT 30 JUNE OF END YEAR 

New South Wales
0.22
0.04
-0.25
-0.41
Victoria
0.09
0.16
-0.46
0.74
Queensland
-1.22
0.34
0.48
0.16
South Australia
-0.68
0.72
0.31
-0.63
Western Australia
-1.26
1.83
-0.19
0.60
Tasmania
0.09
-1.35
-0.23
-0.39
Northern Territory
-4.08
-4.05
-2.86
-0.11
Australian Capital Territory
2.12
1.45
-0.39
-1.36
Australia
-0.28
0.30
-0.15
0.05



DATA SOURCES AND LIMITATIONS

Like all statistical measures, the data sources used to compile ERP are subject to measurement error. Whilst the ABS population estimates are based on the best available data, understanding the limitations is important when using them for decision making.

The Population Census

The Population Census aims to count all people in Australia on census night. The planning and development for the census is a major exercise and commences some seven years before census night. The ABS recruits and trains census collectors who know the local area so as to minimise any errors that may occur during delivery and collection of forms. Procedures are in place to provide mail back envelopes when collectors are unable to make contact with residents in dwellings. Special enumeration strategies are developed and implemented for Indigenous Australians, the homeless, and persons in transit on overnight trains, buses, planes and shipping. Significant planning and processes are implemented to ensure a high quality Indigenous enumeration including recruitment of local community collectors, recognition and respect for cultural characteristics and multistage quality assurance processes.

A comprehensive media and advertising campaign is implemented to raise awareness of the census and keep the community informed during the enumeration cycle. A telephone census inquiry service, census booklet, web site and interpreter service are provided to help answer householders’ questions. Information packs are also provided to federal and state parliamentarians, local councils, community groups and schools.

In spite of these efforts, experience has shown that a small percentage of the population are missed and an even smaller percentage of the population are counted more than once. A household based Post Enumeration Survey (PES) conducted 3 weeks after the census is the main source for assessing the level and characteristics of people undercounted and overcounted. The PES does not include dwellings in very sparsely populated areas due to the high cost of enumeration. The PES also does not include Indigenous communities, as the close involvement of the Indigenous community organisations in the census enumeration process make it impractical to effectively conduct an independent PES for these communities. Alternative methodologies are used for estimating the extent of undercount for these population groups. The level of net undercount in each state and territory is then used to add to the census counts on a usual residence basis. Results of the 2001 Census PES are presented in table 2. Overall the net undercount for Australia was 1.8%, with the Northern Territory having the highest (4.0%) net undercount and the Australian Capital Territory the lowest (1.0%).

Table 2: NET UNDERCOUNT 2001 CENSUS

Per cent

New South Wales
2.0
Victoria
1.4
Queensland
1.9
South Australia
1.6
Western Australia
2.0
Tasmania
1.6
Northern Territory
4.0
Australian Capital Territory
1.0
Australia
1.8



As the PES is a sample of dwellings, estimates derived from the survey are subject to error. For 2001, the net undercount rate for Australia of 1.8% has a standard error of 0.1 percentage points. This means that there are 19 chances in 20 (95%) that the net undercount rate is in the range of 1.6% to 2.0%.

Natural increase

Birth and death statistics are provided by state and territory registrars. For preliminary estimates, births and deaths on a date of registration are used as a substitute for date of occurrence basis.

Australia is regarded as having a sound civil registration process. Although most births and deaths are registered promptly, there is an interval between the date of occurrence of a birth or death and the date of registration. There is a detectable delay in a small proportion of registrations, more so for births than deaths. Accordingly, estimates of natural increase for each financial year are revised 15 months after the end of the year. The revisions take account of subsequently reported registrations, converted to a date of occurrence basis and an adjustment applied for anticipated further late registrations based on historical trends.

Overseas Migration

Net overseas migration is measured using passenger cards completed by passengers crossing Australia’s borders. Data on permanent and long-term arrivals and departures are used for population estimation purposes, together with an adjustment factor to take account of passengers’ change in travel intentions.

Because of the significant impact of overseas migration on population growth for Australia and in particular states and cities, the data available from the Department of Immigration and Multicultural and Indigenous Affairs incoming and outgoing passenger cards are critical to the accurate measurement of the Australian population.

Overseas migration estimates can be affected by changes between intended and actual duration of stay of travellers to or from Australia. As a duration of stay or absence of 12 months or more determines inclusion or removal from the population, it will not be until over 12 months after the initial crossing of the Australian border that confirmation is obtained that the person was indeed a permanent or long-term arrival or departure. Any resultant change in category of overseas movement is known as category jumping. In recent years with increased numbers of overseas students and temporary business entrants to Australia, the volume of long-term arrivals to Australia has increased substantially. However, many of these people depart Australia for short absences, which presents difficulties in determining their residential status. In 2003, the ABS is seeking to better understand and develop a new method of measuring the extent of category jumping and its impact on overseas migration statistics.

Interstate Migration

Australia does not have a comprehensive and timely source of statistics on interstate migration. Some European countries have population registers which permit regular updates of the population based on the place of residence. Unlike these population registers, Australia’s lists of residents such as electoral roles, tax and rate payers and drivers license registers do not cover the complete population and are not always updated in a timely manner.

In the absence of such data for Australia, interstate migration estimates are compiled using a model constructed from the internal migration results of the most recent census, changes of address as advised to Medicare/Health Insurance Commission each quarter and changes in the level of Defence Forces in each state and territory. The model is calibrated on each subsequent census ensuring that it is relevant and of acceptable accuracy. Estimates of intercensal interstate migration are evaluated and revised with the availability of the results from the next census. The intercensal interstate migration model for the next intercensal period is then recalculated using the new census results. It does include some specific adjustments to take account of the special characteristics of some states and territories (e.g. Northern Territory which has a proportionally higher itinerant population than other jurisdictions).

Due to the lack of hard data, interstate migration is the most challenging component of population change to measure. The ABS periodically reviews alternative sources of data to estimate interstate migration. Although Medicare address changes are not without their limitations as a data source, no other alternative data source has been found to be superior for quarterly population estimates.


COMPOSITIONAL CHANGE

Data in Attachment A shows the composition of population change for each state/territory from 1997-2002.

Queensland has the largest population increases in percentage terms over the period with the natural increase, net overseas migration and net interstate migration all being contributing factors. Tasmania, on the other hand, has had population decline over the period. Although natural increase and net overseas migration contribute positively to Tasmania’s population change, it is the net interstate migration which impacts heavily resulting in the population decrease.

The large positive net interstate migration figures for Queensland are predominantly as a result of people shifting to the warmer climate from Victoria, Tasmania and South Australia in particular. This is mirrored by the negative net interstate migration figures which these states have for the most part.

We know that many overseas migrants settle in Sydney and Melbourne. This is confirmed by figures in Attachment A which show higher net overseas migration rates for New South Wales and Victoria than most other states/territories. The Northern Territory recorded the largest natural increase contribution to population change of all states/ territories from 1997-2002, while South Australia had the lowest.


CONCLUSION

The compiling of ERP’s is a difficult and complex process reliant on data derived from a number of different sources. While sound statistical and objective processes are used to assemble the ERPs, and we believe they are the best estimates given available source data, ultimately the ABS ERPs are estimates and reflect any errors in the source data. The ERPs are also subject to revision, as updated data become available.

Attachment A.
POPULATION CHANGE RATES

Year Ended 30 June
Natural increase
(%)
Net overseas migration
(%)
Net interstate migration
(%)
Total
(%)

New South Wales
1997
0.69
0.60
-0.17
1.16
1998
0.63
0.51
-0.20
0.99
1999
0.64
0.65
-0.21
1.14
2000
0.64
0.68
-0.22
1.17
2001
0.61
0.90
-0.25
1.37
2002
0.57
0.78
-0.36
0.99
Victoria
1997
0.63
0.46
-0.14
0.81
1998
0.60
0.42
-0.01
0.88
1999
0.58
0.53
0.05
1.05
2000
0.59
0.58
0.11
1.17
2001
0.56
0.75
0.11
1.34
2002
0.57
0.71
0.13
1.41
Queensland
1997
0.77
0.38
0.59
1.68
1998
0.73
0.37
0.51
1.56
1999
0.71
0.40
0.48
1.56
2000
0.70
0.50
0.53
1.72
2001
0.71
0.59
0.56
1.89
2002
0.67
0.69
0.80
2.16
South Australia
1997
0.47
0.21
-0.23
0.48
1998
0.45
0.21
-0.13
0.55
1999
0.45
0.18
-0.11
0.55
2000
0.42
0.26
-0.24
0.48
2001
0.37
0.18
-0.16
0.44
2002
0.39
0.30
-0.12
0.56
Western Australia
1997
0.80
0.70
0.26
1.68
1998
0.76
0.67
0.18
1.54
1999
0.80
0.73
0.02
1.48
2000
0.75
0.76
-0.12
1.34
2001
0.75
0.87
-0.17
1.42
2002
0.69
0.91
-0.22
1.38
Tasmania
1997
0.52
0.05
-0.70
-0.18
1998
0.44
0.01
-0.77
-0.35
1999
0.56
0.04
-0.70
-0.11
2000
0.44
0.09
-0.56
-
2001
0.43
0.02
-0.45
0.08
2002
0.45
0.10
-0.36
0.20
Northern Territory
1997
1.50
0.30
0.96
2.79
1998
1.51
0.30
-0.25
1.59
1999
1.45
0.53
-0.50
1.50
2000
1.41
0.49
-0.47
1.47
2001
1.46
0.45
-0.81
1.13
2002
1.46
0.07
-1.41
0.12
Australian Capital Territory
1997
0.99
-0.02
-0.80
0.26
1998
0.92
-0.08
-0.64
0.27
1999
0.95
-0.07
-0.16
0.79
2000
0.89
-0.03
-0.03
0.92
2001
0.85
0.23
0.13
1.30
2002
0.83
0.26
-0.31
0.78
Australia(a)
1997
0.69
0.48
. .
1.13
1998
0.65
0.43
. .
1.05
1999
0.65
0.52
. .
1.15
2000
0.64
0.57
. .
1.20
2001
0.62
0.71
. .
1.36
2002
0.60
0.69
. .
1.29

(a) includes Other Territories.
Source: Australian Demographic Statistics, September quarter 2002 (cat. no. 3101.0).


Previous PageNext Page