3228.0 - Demographic Estimates and Projections: Concepts, Sources and Methods, 1999
ARCHIVED ISSUE Released at 11:30 AM (CANBERRA TIME) 30/08/1999
Page tools: Print All
Contents >> Chapter 2. National and State population estimates >> Intercensal revision of population estimates

2.46. When the census year (ie. 30 June) population estimates become available for the States and Australia they can be compared with the alternative estimates for the same date already produced by updating the previous census date estimate (using births, deaths and migration data). The difference between the two estimates for each State and Australia is called the 'intercensal error'.

2.47. Of these two estimates, the first is customarily adopted as the true estimate (ie. the true estimate at the census year is the one based on that year's census, not the one carried forward from the previous census).

2.48. To overcome the break in continuity that this would entail, all population estimates updated from the previous census are then revised. In doing so, it is assumed that the discrepancy is accumulated by an equal number each quarter over the intercensal period. For example, in the case of a five-year intercensal period, (ie. 20 quarters), the population at the end of the first quarter is adjusted up or down by 1/20 of the intercensal discrepancy, the second quarter is adjusted by 2/20, the third by 3/20 etc.

2.49. Details of the intercensal error for the States and Australia at 30 June 1981, 1986, 1991 and 1996 are shown in Table 2.2.

 2.2: Intercensal Error (a) by State of Usual Residence at 30 June 1976-1981 1981-1986 1986-1991 1991-1996 Persons ('000) New South Wales 33.1 11.9 2.4 -15.8 Victoria 48.5 3.9 7.0 -20.9 Queensland -51.9 -32.0 11.1 15.9 South Australia 1.7 -9.4 10.4 4.5 Western Australia 1.0 -18.4 29.9 (d) -3.3 Tasmania 4.7 0.4 -6.3 -1.1 Northern Territory 4.3 -6.3 -6.7 -5.2 (c) Australian Capital Territory 9.5 5.5 4.2 -1.2 Australia 50.9 -44.4 51.9 -27.5 (b) Per cent of Population New South Wales 0.63 0.22 0.04 -0.25 Victoria 1.23 0.09 0.16 -0.46 Queensland -2.21 -1.22 0.34 0.48 South Australia 0.13 -0.68 0.72 0.31 Western Australia 0.08 -1.26 1.83 (d) -0.19 Tasmania 1.10 0.09 -1.35 -0.23 Northern Territory 3.51 -4.08 -4.05 -2.86 (c) Australian Capital Territory 4.17 2.12 1.45 -0.39 Australia 0.34 -0.28 0.30 -0.15 (b) (a) Positive figures indicate that the population estimate updated from the previous census was too high. (b) Includes a small component of intercensal error for Other Territories. (c) 1,045 persons of the intercensal error for the NT can be attributed to an under-estimate of the 1991 NT estimated resident population. (d) Approximately 9,000 persons of the intercensual error caused by Post-Enumeration Survey over-adjustment in 1986

2.50. The accuracy of Australia's post-censal population estimates has improved over time and compares favourably with that of other countries (see Table 2.3).
 2.3: International Comparison of Absolute Intercensal Discrepancy, 1981 to 1996 Country 1981 1986 1991 1996 Australia National 0.3 0.3 0.3 0.2 State (a) 1.6 1.2 1.2 0.4 Canada (b) National 0.4 0.4 0.1 0.6 Province (a) 1.8 1.2 1.2 0.9 USA (1980 & 1990) (b) National 2.1 - 0.5 - State (a) 2.5 - 1.5 - England and Wales England and Wales (a) 0.2 - 0.4 - Constituent Counties (a) 1.8 - 2.5 - New Zealand (b) National 0.4 0.3 0.6 0.3 (a) Average absolute error. (b) Equates with error of closure which is the difference between population estimates produced prior to a census and corresponding census counts. (No account is taken of variations in undercount between census)

2.51. There are two areas which contribute to the intercensal error:
• errors in the census-based estimates of the population at the current or previous census date; and/or
• errors in the estimates of any of the components of population change since the previous census.

An assessment of the accuracy of census date population estimates and the components of population change is given in Chapter 4.

Intercensal Revision 1991-96

2.52. For the 1991-96 State intercensal estimates, the 1996 Census allowed an assessment of how much of the intercensal error was due to inaccuracies in estimates of interstate migration. The 1996 census data which allowed this assessment were from the two census questions on usual residence 1 year ago and 5 years ago. Estimates of interstate migration based on census data were used to revise the interstate migration component of population updates for the intercensal period, before the intercensal discrepancy was calculated.

2.53. After the intercensal error is adjusted for the intercensal interstate migration revision, the remaining (unattributable) portion is known as the intercensal discrepancy. Thus the intercensal discrepancy acts as a balancing item, that when combined with births, deaths and migration equals the change in the intercensal population estimates.

2.54. The procedures used to revise 1986-91 intercensal interstate migration can be found in the previous edition of this publication, Population Estimates: Concepts, Sources and Methods (3228.0), published in June 1995.

(a) Revisions of interstate migration

2.55. Comparisons of the Medicare-based estimates of interstate migration (see Appendix 6) for the 1991-96 period, with 1996 census counts of 1991-96 interstate migration (adjusted for undercount and to include an estimate for ages 0 to 4), are shown in Table 2.4.

 2.4: Net Interstate Migration, comparison between medicare-based estimates and census-based (a) estimates 1991-96 Interstate migration 1991-96 Preliminary intercensal error (c) Estimated (b) Census (a) Difference 30 June 1996 New South Wales -78,478 -63,675 -14,803 -13,738 Victoria -122,775 -84,167 -38,608 -19,814 Queensland 223,337 156,045 67,292 15,566 South Australia -21,016 -19,211 -1,805 4,764 Western Australia 10,660 18,420 -7,760 -3,005 Tasmania -9,498 -6,523 -2,975 -1,213 Northern Territory -4,733 -478 -4,255 -4,201 Australian Capital Territory 2,503 -411 2,914 -514 Australia 0 0 0 -22,155 (d) (a) Census counts have been adjusted for net undercount. (b) Medicare-based estimates cumulated quarterly. (c) Difference between 1996 census-based population estimates and preliminary estimates at 30 June 1996 as carried forward from the 1991 Census. (d) Includes a small component of intercensal error for Other Territories.

2.56. The approach used for revising population estimates for the intercensal period assumes that intercensal error is primarily due to errors in estimating interstate migration (as it is the most difficult component to estimate), to the extent that this is supported by the census-based migration data. If the difference between census-based and Medicare-based migration estimates was the same sign (positive or negative) as a State's intercensal error, then that migration difference was interpreted as accounting for the intercensal error.

2.57. For example, NSW's preliminary intercensal error of -13,738 could be entirely attributed to interstate migration being 14,803 lower in Medicare-based migration than census-based migration. As an adjustment of only 13,738 is required to explain the intercensal error, that was the amount the Medicare-based estimates were altered by (giving -64,740 in Table 2.5), rather than the full Medicare-based migration difference of 14,803.

2.58. The intercensal errors for South Australia and the Australian Capital Territory were not explained at all by the census-based migration data, so their Medicare-based migration estimates were not altered. For instance, South Australia had an intercensal error of -4,764 (ie. the 1991-based 30 June 1996 population estimate was 4,764 too high), but census-based migration indicated that, in net terms, fewer people departed the State. As such, if South Australia's migration estimate was adjusted according to census-based migration, its intercensal error would increase.

2.59. As the intercensal error sums to 26,405 rather than zero, using it to help quantify the extent of interstate migration revisions means that a further adjustment is required to ensure interstate migration sums to zero for Australia, as shown in Table 2.5. These 26,405 persons were redistributed across all States in proportion to each State's share of total interstate migration flows, according to 1991-96 Medicare-based interstate arrivals and departures.
 2.5: Revisions to 1991-96 Net Interstate Migration Interstate migration 1991-96 Intercensal discrepancy 30 June 1996 After adjustment to minimise intercensal error After zero-sum adjustment (final) Preliminary (a) Final (b) New South Wales -64,740 -71,770 -7,030 -9,062 Victoria -102,961 -107,830 -4,869 -5,984 Queensland 207,771 201,035 -6,736 -6,440 South Australia -21,016 -23,107 2,673 2,434 Western Australia 13,665 11,526 -2,139 -2,473 Tasmania -8,285 -9,136 -851 -761 Northern Territory -532 -1,831 -1,299 -2,368 Australian Capital Territory 2,503 1,113 -1,904 -2,518 Australia 26,405 0 -22,155 (c) -27,345 (c) (a) Preliminary intercensal error adjusted for revised 1991-96 interstate migration. (b) Preliminary intercensal discrepancy adjusted for finalised 1991-96 birth, death and category jumping estimates. (c) Includes a small component of intercensal discrepancy for Other Territories

2.60. The preliminary intercensal discrepancy is the preliminary intercensal error adjusted for the revisions made to interstate migration (see Table 2.6).

2.61. The final intercensal discrepancy is the preliminary intercensal discrepancy adjusted for finalised estimates of births, deaths and category jumping which occurred between 1991 and 1996. As there is a lag before finalisation of these components of population change, final intercensal discrepancy is calculated considerably later than the preliminary levels.

2.62. The final intercensal error, as shown in Table 2.2, is equivalent to the final intercensal discrepancy but without the interstate migration revisions introduced to account for intercensal error. They are therefore identical at the Australian, but not State level.
 2.6: Definitions of Intercensal Error and Intercensal Discrepancy Definition 1991-based estimate 1996-based estimate Preliminary intercensal error Preliminary Preliminary Final intercensal error Revised Final Final intercensal discrepancy Final (a) Final (a) Revisions to interstate migration on the basis of the 1996 Census

2.63. The 1991-96 revised interstate migration estimates were then divided into two parts:
(1)  Net movement for 1995-96. This used data from the census question on usual address one year ago (adjusted for undercount and to include an estimate for age zero).
(2)  Net movement for 1991-95. These data were obtained by subtracting the figures for 1995-96 from the totals for 1991-96.

2.64. The 1995-96 revised net interstate migration levels [2.63(i)] were derived by:
Step 1.
Calculating the difference between the original 1991-96 Medicare-based estimates and the final revised figures.
Step 2.
Expressing the result of Step 1 as a proportion of the difference between the original 1991-96 Medicare-based estimates and the 1991-96 census migration estimates.
Step 3.
Applying the proportions from Step 2 to the gap between the original 1995-96 Medicare-based estimates and the 1995-96 census migration estimates.
Step 4.
Adjusting the revised net interstate migration levels in Step 3 to ensure they sum to zero.
Steps 1 to 3 can be summarised as:

2.65. Quarterly interstate arrival and departure estimates for 1995-96 and for the four years 1991-95 were calculated by spreading the total net gain/loss for each State over the 4 quarters for 1995-96 and the 16 quarters for 1991-95 proportionally according to the original quarterly Medicare-based estimates.

(b) Distribution of the intercensal discrepancy

2.66. The 1991-96 intercensal discrepancy was distributed evenly over the intercensal period for each State, sex and birth cohort as follows:

2.67. For cohorts born prior to the 30 June 1991, ie. the start of the intercensal period,

 IDq = ID 20 for each quarter q of the intercensal period (20 quarters) where: IDq = intercensal discrepancy for quarter q ID = total intercensal discrepancy
2.68. For cohorts born between the two censuses, the intercensal discrepancy was split over those quarters including and following the financial year of birth, that is,
 IDq = ID 4n for each quarter q of the intercensal period (20 quarters) where: n = the number of financial years that the cohort has been represented in the population including the financial year of birth of the cohort. For example the discrepancy for the cohort born in the 1995-96 financial year can only be distributed over the 1995-96 period, while the discrepancy for those born in 1994-96 can only be distributed over the two financial years (ie 8 quarters) 1994-95 and 1995-96.

2.69. These formulae were also used for distributing the intercensal discrepancy over the quarterly population estimates by marital status and country of birth.

(c) Final revised intercensal population estimates

2.70. The revised estimates of the population for the intercensal period were calculated by the quarterly adjustment of the population, starting with the 30 June 1991 (Census-based) population, for subsequent births, deaths, overseas migration, interstate migration and intercensal discrepancy. This was the same procedure as outlined in paragraph 2.15 for post-censal estimation, except that finalised components of change were available and the intercensal discrepancy component was also included.

 Previous Page Next Page