Projection Model

Population projections from 2021 to 2051 were calculated using a custom-built cohort-component model designed for population groups defined by religious affiliation and birthplace. The model handles populations by sex and five-year age groups up to age 85+ years and moves forward in five-year time increments. International migration is modelled using immigration flows and emigration rates, whilst religious affiliation moves are handled with outward and inward rates (like a bi-regional internal migration model). There are, of course, no switches between birthplace categories. All births are allocated to Australian-born populations by religious affiliation irrespective of the birthplace of women, with babies assumed to have the same religious affiliation as their mothers.

Cohort populations are projected using the population accounting equation.

$$_^\left(t+5\right)=_^\left(t\right)- _^- _^- _^+_^+_^$$

(1)

where

\(P\):

denotes population

\(D\):

deaths

\(OR\):

outward religious movement

\(E\):

emigration

\(IR\):

inward religious movement

\(I\):

immigration

\(t\):

point in time

\(b\):

birthplace

\(r\):

religion

\(s\):

sex

\(a\):

age group

\(a\to a+5\):

period-cohort aged a at time t and aged a + 5 at time t + 5

All component flows occur in the t,t + 5 projection interval.

With the exception of immigration, the projected flows in Eq. 1 are calculated as rates multiplied by populations at risk. Deaths, emigration, and outward religious movement are projected as the product of rates multiplied by the population-at-risk of each birthplace-religious group. For example, outward religious movement is projected as:

$$_^=_^ \frac\left[_^\left(t\right)+_^(t+5)\right]$$

(2)

where

\(or\) denotes the outward religious movement rate.

Because religious movement is handled using a bi-regional arrangement in which movement occurs between (1) each religion and (2) all other religions, inward religious movement is calculated using a population-at-risk of all other religions. Thus:

$$_^=_^ \frac\left[_^\left(t\right)-P}_^\left(t\right)+_^\left(t+5\right)-_^(t+5)\right]$$

(3)

where

\(ir\):

is the inward religious movement rate

\(*\):

indicates a variable summed over all categories

A small adjustment is needed to ensure that religious outward movement by sex and period-cohort summed over all religious categories equals the equivalent sum of religious inward movement.

Births are projected as the product of age-specific fertility rates and female populations-at-risk, and then summed over all birthplace groups:

$$_^=\sum_\left(_^ \frac\left[_^\left(t\right)+_^(t+5)\right]\right)$$

(4)

where

\(B\):

refers to births

\(ASFR\):

age-specific fertility rate

\(f\):

female population

By definition, all babies born in Australia belong to the Australian-born birthplace category. We assume all babies have the same religious affiliation as their mothers. Births are then aggregated over age of mother and allocated to male and female populations using the sex ratio at birth. The newly born cohort of babies is then projected to age 0–4 years using accounting Eq. 1 except that the start-of-interval population is replaced by births.

Base Period Data Estimation

Demographic modelling was undertaken for population groups cross-classified by 18 birthplace categories and 12 religious affiliation groups (Table 1), giving 216 populations in total. The birthplace groups were aggregated from the Australian Bureau of Statistics (ABS) classification of countries (ABS, 2016) to reflect the countries and global regions from which Australia draws its migrants. The religious classification was based on Bouma and Halafoff’s approach (Bouma & Halafoff, 2017) for managing the many Christian denominations listed in the census. This involves collapsing multiple Christian categories into like groups (e.g., Congregational, Presbyterian and Reformed and Uniting have been combined to create the category CPRU). Similarly, to capture the wider Pentecostal community, the former has been combined with the Apostolic Church and Christians not elsewhere defined. Finally, both Anglicans and Catholics remain as stand-alone categories as they are the largest denominations, while the remaining Christian denominations that are below 1 percent of the population are aggregated in the classification ‘Other Christians’, which includes groups such as the Jehovah’s Witnesses, Lutherans, and Seventh Day Adventists. For the other major world religions such as Buddhism, Hinduism, Judaism, and Islam, there are stand-alone classifications in the census. ‘No religion’ and ‘Not stated and inadequately defined’ are also stand-alone categories, whilst all the remaining faiths (e.g. Australian Aboriginal Traditional Religions, Sikhism, Baha’i, and Nature religions) are aggregated in the ‘Other Religions’ Category as they are collectively less than 1% of the total population.

Table 1 Religious affiliation and birthplace categories

Population estimates by religious affiliation and birthplace were required for the jump-off year of the projections, 2021, and the earlier years of 2016 and 2011 for estimating base period demographic data. Estimated Resident Populations (ERPs) were available from the ABS for these years by birthplace, but not birthplace cross-classified by religious affiliation. We therefore disaggregated birthplace ERPs to religious groups using 2011, 2016, and 2021 Census counts obtained via ABS TableBuilder. This generated ERPs by sex and five-year age groups for 216 (12 × 18) religious affiliation by birthplace groups for the three years.

Data of births by age group and birthplace of mother for 2016–21 were purchased from the ABS. Births data classified by mother’s religious affiliation are not available. Age-specific fertility rates by birthplace of mother, and for the Australian population as a whole, were then calculated. Differences between birthplace-specific Total Fertility Rates (TFR) and the national TFR were used in setting fertility assumptions. The national TFR assumption is formulated first, with birthplace-specific assumptions created as the national assumption plus specified a difference.

Data of deaths by birthplace, sex, and abridged life table age groups (0, 1–4, 5–9, 10–14, … 80–84, 85 + years) for 2016–21 were purchased from the ABS. Deaths data with a religious affiliation breakdown are not available. Age-specific death rates, and then abridged life tables, were calculated for each birthplace group, sex, and for Australia as a whole. Differences between birthplace-specific and national life expectancy were used in setting mortality assumptions.

International migration data was obtained from both the census and ABS overseas migration estimates (ABS, 2022b). Customised overseas migration tables for 2016–21 were obtained from the ABS by birthplace, age, and sex. A breakdown by religion is not available in this data source, though it is available with census immigration counts. Census data was extracted from TableBuilder on the number of people living overseas five years ago by birthplace, religion, age, and sex. Census immigration data are transition-type measures of migration (Rees, et al., 2010) which record net changes of address at five-year intervals; the migration values are smaller than movement-type measures of migration flows. The census immigration flows were therefore scaled up to match ABS immigration estimates for 2016–21. Emigration rates were estimated by birthplace, age, and sex using the ABS overseas migration estimates and ERPs by birthplace, age, and sex.

The estimation of religious movement rates involved multiple data sources and several steps. It required census data on people reporting different religious affiliations between the 2011 and 2016 censuses, obtained from the Australian Census Longitudinal Dataset (ACLD) via TableBuilder. The ACLD is a 5% sample of probabilistically-linked records between censuses (ABS, 2019). Due to small numbers, it was not possible to obtain a full matrix of 2011 religion/birthplace by 2016 religion/birthplace by sex and age group. We therefore extracted a two-category set of flows by age and sex for each religious group (but not cross-classified by birthplace) which consisted of outflows from the religion to all other religious affiliation categories, and inflows to the religious group in question from all other religious categories. These data were then used to calculate inward and outward religious change rates assuming just one religious movement in the five-year interval.

The religious change rates were then adjusted to be consistent with 2016–21 population change for religious populations. Population accounts by sex and age were created for each religious group to ensure that deaths, immigration, emigration, religious inflows, and religious outflows matched total cohort change over the 2016–21 period. For example, for the female cohort aged 25–29 years in 2016 and 30–34 years in 2021, the demographic components needed to match the change in size of this cohort between 2016 and 2021. Deaths, immigration, and emigration were assumed to be correct. Preliminary religious inflows and outflows were estimated by applying the 2011–16 rates to the 2016–21 populations-at-risk. Then the inflows and outflows were adjusted proportionally to be consistent with cohort population change. Religious change rates were then re-calculated using the newly estimated religious inflows and outflows.

Birthplace scaling factors were calculated to adjust age profiles of religious change rates up or down for each religious-birthplace group. The scaling factors are defined as the ratio of the religious-birthplace inward or outward movement rate to the religious-only rate. Scaling factors were estimated from an origin–destination matrix of ACLD movements to and from birthplace-religious categories for the period 2011–16. Because of the small numbers involved, the matrix had to be extracted without age-sex detail.

Demographic Projection Assumptions

A national Total Fertility Rate (TFR) of 1.65 was assumed for the whole projection horizon, with birthplace-specific TFRs maintaining the same difference with the national TFR recorded for 2016–21.

Mortality was assumed to continue increasing, with birthplace-specific life expectancy at birth maintaining the same difference with national life expectancy in 2016–21. National mortality projections were prepared using an extrapolative model of mortality (Ediev, 2008).

National immigration assumptions for 2021–26 and 2026–31 were taken from Commonwealth Treasury’s most recent projection assumptions (Centre for Population Projections, 2023). These assumptions comprise immigration of 2.76 million in 2021–26 and 2.89 million in 2026–31. Total immigration was then assumed to continue its long-run upward trend, although the share of immigration by birthplace was assumed to remain constant. The religious composition of immigration flows was assumed to change over time on the basis of immigration data recorded for 2011–16 and 2016–21 in the 2016 and 2021 Censuses. For each birthplace group, the proportion of people in each religious category was extrapolated into the future, with most birthplace-specific immigrant flows becoming less religious over time. The religious proportions were applied to the birthplace-specific immigration flows to obtain projections of migration by birthplace and religion. Emigration rates were assumed to remain constant.

Data for this study are freely available from the [ARC Centre of Excellence in Population Ageing Research Population Futures] Data Archive at [https://www.cepar.edu.au/cepar-population-ageing-projections].