This report outlines the method used for producing population projections of Australian citizens aged 18 years and over for all Statistical Local Areas (SLAs) and Census Collection Districts (CDs) in Tasmania, spanning from 1997 to 2003.

The technique employed for the projections was the cohort-component method, widely accepted as the most accurate age/sex projection methodology. It involved applying fertility, mortality and interstate migration rates and overseas migration levels by age and sex to the base population to produce a projected population, which then became the base for projecting the next year. This cycle was repeated until the projection horizon was reached.

A three-tiered approach was taken in projecting resident population aged 18 years and over for all SLAs and CDs in Tasmania.

1. The Tasmanian population was projected by age and sex.
2. The population of all Tasmanian SLAs was projected by age and sex (constrained to 1).
3. The SLA projections were split into CDs.

The base population for the Tasmanian cohort-component projections was preliminary age/sex Estimated Resident Population (ERP) as at 30 June 1997. Assumptions for the projections were based on both short and long-term trends for each component of population change. The fertility, mortality and overseas migration assumptions were from the latest Population Projections of Australia, States and Territories 1997-2051 (ABS Cat. 3222.0), while the interstate migration assumption was based on the latest interstate migration data (and therefore differs a little from those in 3222.0).

The base population for the SLA cohort-component projections was preliminary 30 June 1997 SLA age/sex ERP. The fertility, mortality and migration assumptions were based on an assessment of SLA-specific levels and trends observed since 1991. The ABS regularly collects such demographic information down to the SLA level, which means that SLA projections (in contrast to smaller areas) are firmly based on a series of known data. At each yearly cycle in this process, the SLA projections were constrained to sum to the State projection, helping to produce more reliable SLA results.

The basis for calculating CD projections was an SLA to CD concordance derived from the 1996 Census of Population and Housing. Differing growth rates of CDs within SLAs were incorporated using 1991-1998 ABS CD building approval data, resulting in SLA to CD concordance split factors extrapolated through to the year 2003.

The lack of demographic data collected regularly at CD level made it necessary to use the conversion method outlined above. While the process is quite complex, it should be reiterated that the basic concept of splitting SLAs to CD level cannot be expected to give projections as reliable as those for SLAs. However, we understand that the end product will be aggregates of large numbers of CDs, so there is a high likelihood that any random errors or inconsistencies will be statistically offset in the aggregation process.

The SLA projections are on the most recent boundaries available, that is, those on which 30 June 1997 SLA Enrolment Resident Population (ERPs) have been/will be published. This corresponds to ASGC Version 1996. CDs are on 1996 boundaries.

It is important to recognise that the projection results given in this report essentially reflect the assumptions made about future fertility, mortality and migration trends. While these assumptions are formulated on the basis of an objective assessment of demographic trends over the past decade and their likely future dynamics, there can be no certainty that they will be realised.

ABS takes responsibility for the method employed, however in accordance with ABS policy regarding small area population projections, the assumptions used are the final responsibility of the client, and the projections are not official ABS population statistics.

The projections may be referred to as "...projections prepared by the ABS according to assumptions reflecting prevailing trends agreed to by the Tasmanian Legislative Council Electoral Boundaries Redistribution Committee...".

No liability will be accepted by the ABS for any damages arising from decisions or actions based upon this population projection consultancy service.

The three-tiered approach outlined in Appendix II has been further disaggregated in this accompanying paper. Apart from the births formulae all equations apply to both sexes, so sex has not been denoted.

This involved projecting the Tasmanian population by age and sex, 1997 - 2003.
The cohort component method used can be summarised in the formulae below:

x -> age
max -> highest age projected (100+ for State, 85+ for SLAs)
t -> base year
P -> population
F -> fertility rate
f -> females
B -> births
Q -> death probability
OM -> net overseas migration
IM -> net interstate migration
NM -> net migration (SLA projections only)

In Step 1 the following refer to interstate migration; in Step 2 they refer to overseas + inter-SLA migration.

DEP -> departures
ARR -> arrivals
DEPRATE -> per capita departure rate
ARRRATE -> per capita arrival rate

For ages 0 to maximum age - 1:

Births were then calculated:

These were then used to calculate age 0 in the projected year:

Interstate migration was calculated by applying departure rates to the Tasmanian population and arrival rates to the mainland population (to obtain numbers departing other States to reside in Tasmania). These rates were derived from 1996 Census data and were held constant for the duration of the projection.

The resulting total arrivals and departures were then scaled to net-out to a predetermined interstate migration net level. Finally, the arrivals and departures by age and sex were scaled to the new arrival and departure totals, then combined to give net interstate migration.

Then add the interstate migration:

Year t+1 then became the base for projecting the next year and the cycle was repeated until the final projection year was reached.

This used the cohort component method to project all Tasmanian SLAs. The formulae in Step 1 also apply to the SLA projections, except that Net Migration (overseas + inter-SLA) was used instead of overseas and inter-SLA separately.

This slightly simpler approach was warranted as the overseas component is negligible in most SLAs in comparison with inter-SLA migration. Furthermore an annual historical time-series only exists at the SLA level for net migration, any overseas/inter-SLA split must approximated using past Census data.

For ages 0 to maximum age - 1:

Births were then calculated:

These were then used to calculate age 0 in the projected year:

SLA migration departures were calculated by applying 1996 Census-derived departure rates to the population:

Total SLA arrivals were then derived using the pre-set net migration assumptions:

The age-specific arrival levels were then re-scaled:

Then apply arrivals and departures to the population:

Year t+1 then became the base for projecting the next year and the cycle was repeated until the final projection year was reached.

However, before P_x(t+1) became the new base, the projected SLAs were constrained to sum to the State-level projection. This involved an iterative proportional fitting process, the year is t+1:

SLA -> Statistical Local Area
T -> Tasmania
a -> first SLA
z -> last SLA
r -> SLA number

Scale the SLA totals to the State total.

For each SLA, scale all ages to sum to the new SLA total.

For each age, scale all SLAs to sum to the State total.

Stages (xviii) and (xix) were then iterated several times before the resulting matrix was rounded.

This involved splitting the completed SLA projections into Census Collection Districts.

From 1996 Census data an SLA to CD concordance was constructed giving the proportion of the SLA total population in each CD. This concordance was then updated for each year up to 2003 to allow for differing rates of CD population growth.

CD building approval data was used to do this, with 1999-2003 building approvals being modelled on the observed 1992-98 data. CDs significantly redrawn for the 1996 Census were assumed to have 1992-96 levels similar to those in 1997 and 1998. The shape of each CD’s 1992-98 building approval distribution determined where/when on the model building approval curve each CD lay, thereby allowing future levels to be extrapolated.

The model curve was originally calculated using thousands of CDs across Australia and is defined as: %BAP_CD(t) = 7 + (29 / (1+t²))

t -> year
CD -> Census Collection District
F -> CD split factor (as a proportion of its SLA)
P₁₉₉₆ -> Population in 1996 Census count

where 2.5 is an approximate occupancy ratio for new dwellings, and SLA is the SLA to which the CD belongs.

The CD split factors were then applied to the SLA projections to give projected CD populations:

The Australian Bureau of Statistics (ABS) have provided projections of the population of Australian residents aged 18 years and over for each Census Collection District (CCD) at 30 June 1997 and annually to 30 June 2003. To allow baseline comparison with electoral roll counts, interpolation was used to derive 25 September 1998 population.

It has been assumed that electoral enrolment will grow by the same amount as the population of Australian residents 18+. The projections have been converted to enrolment projections as follows:

For example, Census Collection District 6032202:

Where a CCD crosses existing Legislative Council Boundaries, the projected enrolment has been allocated to Legislative Council Divisions in the same proportion as current enrolment.

For example (from the previous redistribution) Census Collection District 6051407:

Thus, E_{2000(Derwent)} = 26/321*335