DEMOGRAPHIC ADJUSTMENT

OVERVIEW

'Demographic adjustment' is an all-inclusive term for a range of adjustments made to the population estimates during the rebasing process. It is in essence a data coherence exercise. A number of checks are made to address known and anticipated issues in the new population base, such as implausible sex ratios (the number of males per 100 females) and the effects of misreported ages on the Census form. Adjustments also correct demographically implausible features that may have been introduced by the sample survey nature of the PES.

The data coherence checks are usually made at the five year age group level, but for the very young or upper ages, single year of age is also checked. For these ages, population levels and sex ratios for the rebased ERP are analysed. All adjustments are applied at the single year of age level. Included in these checks are:

• comparisons with other sources of population data
• consistency against other Census years' undercount adjustments;
• imputed ages for Census age non-response; and
• upper age misreporting on the Census form, including malicious age misreporting.

The data coherence process compares the estimate of the new population base with the following sources of data:

• ABS demographic databank: The ABS maintains a population account independent of the Census, based on administrative data: births and death registrations, and net overseas migration (NOM) derived from overseas arrivals and departures data;
• 2016 Census counts: The 2016 Census counts on a usual residence basis are useful for analysing demographics that are not influenced by undercount adjustments;
• Unrebased 30 June 2016 ERP: This is based on the 2011 Census, which has been updated with components of population change between 2011 and 2016. The difference between the unrebased 2016 ERP and the new population base is called the intercensal difference. For ages 0–4, this source is considered to be of good quality as it is based on recent births registrations and less influenced by migration than other ages;
• 30 June 2011 ERP, aged five years: This data excludes components of population change between 2011 and 2016;
• Medicare enrolments: Almost all Australians are covered under the Medicare system, and as such, Medicare enrolment data is close in scope to the Australian resident population. Medicare data is considered to be particularly reliable for children as incentives to enrol children are strong, both for medical treatment and access to government benefits and rebates;
• Australian Electoral Commission (AEC) enrolments: AEC enrolments data represents enrolled citizens aged 17 years and over and is maintained through a continuous roll update process; and
• School enrolments: Data from ABS' publication Schools Australia, 2016 (cat. no. 4221.0) is based on enrolments in all government and non-government schools in Australia during 2016.

Additionally, a 'composite method' approach was adopted for determining the sex ratio adjustments using the following two populations:

• the unadjusted rebased population as at Census night 2016; and
• an unrebased population for Census night 2016, derived by updating the rebased ERP as Census night 2011 with population change components.

More emphasis was placed on the 2016 Census-based population for the composite estimate, using a 60% weighting of the 2016 Census-based sex ratio, and 40% of the updated 2011 Census-based sex ratio.

The data coherence process and the composite method were considered for ages 0–74. Populations for ages 75 and over were considered too small for robust data coherence and composite method analyses to be made.


Adjustments to population levels

For three age groups, the intercensal difference was relatively large compared to neighbouring age groups and to previous Censuses for these cohorts. The unadjusted rebased 2016 ERP was considered too high for the 50–54 and 65–69 year age groups, and too low for the 70–74 year age group. As such, adjustments were made to these age groups so that the intercensal differences were halved. This is an extension of the smoothing treatment which is the basis of the composite method.


Adjustments to sex ratios

For the young age groups (0–19 years), the sex ratios derived from composite method did not align well with the data coherence sources. Consequently, sex ratios were adjusted based on the data coherence analysis, or in conjunction with the composite method. Adjustments made for each of these age group are detailed below.

0–4 years: The sex ratio of 106.2 for the unadjusted 2016 population base was too high compared with data coherence sources. The sex ratio for the unrebased 2016 ERP was 105.7, for Medicare enrolments was 105.3, and for the 2016 Census count was 105.5. The sex ratio for this age group was therefore adjusted to 105.5, by reducing the male population and increasing females by the same amount.

5–9 years: The sex ratio of 106.4 for the unadjusted 2016 population base was too high compared with data coherence sources. The sex ratio for school enrolments (aged 6 to 9) was 105.4, for the demographic databank was 105.5 and for the 2016 Census count was 105.4. The sex ratio for this age group was therefore adjusted to 105.4, by reducing males and increasing females by the same amount.

10–14 years: The sex ratio of 106.1 for the unadjusted 2016 population base was too high compared with data coherence sources. School enrolments, Medicare enrolments and 2016 Census counts all had a sex ratio between 105.6 and 105.7. The sex ratio for this age group was therefore adjusted to 105.7, by reducing males and increasing females by the same amount.

15–19 years: The sex ratio of 104.8 for the unadjusted 2016 population base was too low compared with data coherence sources. The sex ratio for Medicare enrolments was 105.2, and for the demographic databank was 105.1. The sex ratio derived from the composite estimate of 105.5 was considered too high compared with the data coherence sources. The sex ratio derived from data confrontation alone did not raise the sex ratio to a value that was plausible, and so a middle value between the two options was chosen and the sex ratio adjusted to 105.3 by increasing males and reducing females by the same amount.

For the age groups 20–24 to 70–74, the composite method was applied to arrive at a consistently-derived demographic adjustment to the sex ratio, with consequential balanced adjustments to the male and female populations. The consequential adjustments to male and female populations were mostly within one standard error of the PES net undercount adjustment, and were all within two standard errors. For more information on PES net undercount and standard errors, refer to Census of Population and Housing: Details of Overcount and Undercount, Australia, 2016 (cat. no. 2940.0).

Table 1 summarises the demographic adjustments made as a result of the data coherence and composite estimate analyses.

Age Group	Sex ratio		Demographic Adjustment
(years)	Before Demographic Adjustment	After Demographic Adjustment	Males	Females
0-4	106.2	105.5	–2 391	2 391
5-9	106.4	105.4	–3 700	3 700
10-14	106.1	105.7	–1 582	1 582
15-19	104.8	105.3	1 799	–1 799
20-24	105.6	105.0	–2 168	2 168
25-29	101.5	100.5	–4 542	4 542
30-34	97.0	98.8	7 905	–7 905
35-39	98.7	99.2	1 749	–1 749
40-44	98.0	98.2	1 128	–1 128
45-49	95.8	95.7	–767	767
50-54	96.5	97.2	–1 528	–7 222
55-59	97.2	96.6	–2 248	2 248
60-64	96.7	95.9	–2 815	2 815
65-69	96.8	97.4	–3 133	–6 367
70-74	95.3	95.9	5 607	3 193
Total for population	98.5	98.4	–6 686	–2 764

Census upper age misreporting

The Census upper age misreporting adjustment is a targeted adjustment for known intentional and unintentional misreporting of age and/or date of birth in higher age groups.

The upper age structure is particularly distorted by age misreporting due to the small population in these very high ages. Even a small number of misreported ages can have a large impact on sex ratios, death rates, and life expectancy calculations.

Centenarians in the 2016 Census results were first assessed by comparing with centenarian estimates created indirectly from death counts. This analysis showed that Census data for centenarians was of good quality. Consequently, a method that was used to identify the overall likelihood of upper age misreporting in the Census for 2011 rebasing was not applied for 2016 rebasing. Instead, Census forms with the highest ages were individually assessed for age misreporting, and those with implausible combinations of age with other characteristics were assigned different ages. The total number of respondents adjusted due to Census upper age misreporting was 119.


Summary of adjustments

Table 2 summarises the demographic adjustments made during the cohort analysis process.

Age Group (years)	Males	Females	Persons
0-4	–2 391	2 391	0
5-9	–3 700	3 700	0
10-14	–1 582	1 582	0
15-19	1 799	–1 799	0
20-24	–2 168	2 168	0
25-29	–4 542	4 542	0
30-34	7 905	–7 905	0
35-39	1 749	–1 749	0
40-44	1 128	–1 128	0
45-49	–767	767	0
50-54	–1 528	–7 222	–8 750
55-59	–2 248	2 248	0
60-64	–2 815	2 815	0
65-69	–3 133	–6 367	–9 500
70-74	5 607	3 193	8 800
75+	-54	–65	–119
Total	–6 740	–2 835	–9 575

The overall net adjustment from the 2016 cohort analysis was -9,575 persons, which is less than 0.05% of the total 2016 population base. The net adjustment for males was a decrease of 6,740 persons and for females was a decrease of 2,835 persons.

Figure 2 and figure 3 show the effect of the 2016 demographic adjustment on the new 2016 population base for males and females. The graphs show that the adjustments had minimal impact on the overall population level and sex ratio.



State/territory level

The national net demographic adjustment for preliminary 2016 rebasing was distributed between the states and territories on a pro-rata basis, as shown in Table 3.

	NSW	Vic.	Qld	SA	WA	Tas.	NT	ACT	Aust.
Males	–2 059	–1 754	–1 523	–424	–626	–172	–80	–101	–6 740
Females	–982	–771	–449	–256	–242	–63	–7	–64	–2 835
Persons	–3 041	–2 525	–1 972	–680	–868	–235	–87	–165	–9 575

Sub-state level

Below the state and territory level, an additional demographic adjustment process was performed at the part of state (greater capital city / rest of state region) level. This was done to account for relatively large variations in the PES undercount adjustments at this level between Censuses. The disparities in undercount adjustment for some parts of state showed an implausible proportional split of population between the greater capital city and rest of state regions within each state/territory in 2016, compared with the 2011 results.

Following detailed analysis, a composite method was adopted using the following two populations to arrive at the new (adjusted) 2016 rebased ERP for each part of state:

• the original (unadjusted) 2016 rebased part of state ERP, and
• the 2016 unrebased part of state ERP (as published in March 2017 in Regional Population Growth, 2016 (cat. no 3218.0)).

A weighting of 60% was applied to the unadjusted 2016 rebased ERP split, and 40% to the 2016 unrebased ERP split. The population for each part of state region was then adjusted to arrive at this composite split (with balancing adjustments made so that the net adjustment at the state/territory level was zero), as shown in Table 4.

	NSW	Vic.	Qld	SA	WA	Tas.	NT	Total
Greater capital city (no.)	–10 715	90	–5 160	3 869	–898	–1 510	–1 285	–15 609
Rest of state (no.)	10 715	–90	5 160	–3 869	898	1 510	1 285	15 609
Greater capital city (%)	–0.21	0.00	–0.22	0.29	–0.04	–0.67	–0.87	–0.10
Rest of state (%)	0.40	–0.01	0.21	–0.99	0.17	0.52	1.30	0.20

ERP for all regions below each part of state were then apportioned to the new part of state ERP (by age and sex) on a pro-rata basis.