This report was published by the former Department of Families, Community Services (FaCS).
- Executive summary
- 1 Understanding population ageing
- 2 The birth rate
- 3 Measuring life expectancy
- 4 Indices of ageing
- 5 The birth rate, cohort size, population ageing
- 6 Natural increase and decrease; doubling and halving time
- 7 Is migration the answer?
- 8 Sub-population differences
- 9 Demographic compression
- 10 Age structure and the welfare state—a ‘social’ or ‘demographic’ contract?
- 11 Policy and population ageing
- 12 Population projections
- 13 Methodological implications—some useful techniques
- Useful web sites
Over the past two and a half centuries, the developed world has experienced what is arguably one of humanity’s greatest achievements: the demographic transition. This transition, denoted by a fall from high to low fertility and mortality, has taken place in every developed country, and is currently under way in every developing country. It has brought and continues to bring with it a number of momentous changes, most notably a shift from youthful age structures and expansive growth to ageing and stationary or declining populations.
The implications of the shifts are profound. More than any phenomenon in the recent past, they will challenge our social, economic, political and cultural structures, and the policymaking communities that must respond to these changes. However, while the implications of the transition are increasingly understood and recorded, a user-friendly explanation of the basic demography on which they are based is typically missing. The gap, usually a reflection of word constraints, causes problems for those who recognise the importance of understanding what is going on, yet are at a loss to know where to begin.
This ‘toolkit’, a compendium of concepts and specifically related to population ageing, is especially written for busy policy-makers, advisers and analysts. It does not purport to cover these concepts and issues comprehensively, but rather, to outline the key principles involved, and to indicate where further information may be found. The main points are summarised below.
- The demographic transition has one major outcome: a shift from youthful and growing populations to populations that are ageing and potentially declining. For policy purposes, a distinction needs to be made between structural and numerical ageing. The former refers to an increase in the proportion of aged in the population, and is primarily caused by falling fertility. Assuming a continuation of low fertility, the main effect of structural ageing will be to reduce the size of the working-age population/primary tax base in comparison with the increasing proportion of elderly. Numerical ageing, on the other hand, refers to an absolute increase in the number of aged, and is primarily caused by increasing life expectancy, first at the younger ages, then at older ages.
- The total fertility rate (TFR), which is used as a proxy for average family size, is a synthetic measure with many limitations. Most importantly, it conceals both the effect of delayed and recuperated fertility, and the proportion of women having no children at all. Over time, actual completed family size is typically higher than the lowest TFR, and lower than the highest TFR.
- Life expectancy specifies the additional number of years a person in a given birth cohort can expect to live beyond a reference age. Typically what is referred to is life expectancy at birth. This changes over the life cycle. When considering future demand for elder-oriented goods and services, it is important to be aware of measures of life expectancy at older ages, for example, age 65.
- A population is considered young when it has a median age of less than 20 years (or less than 5 per cent aged 65 years and over), and old when it has a median age of more than 30 years (or more than 10 per cent over the age of 65). Other useful indices of population ageing are the aged/child ratio and the familial support ratio. The more commonly used dependency and potential support ratios have many limitations. Uppermost among these is that they treat the working-age population (15–64 years) as if all its members of it were economically active.
- A cohort is a group of people connected by a similar event (for example, birth in a given year). The size of a birth cohort is the combined function of prevailing birth (and mortality) rates and the number of women at reproductive age (and actually giving birth). This caused Australia’s largest birth cohort to be born in 1971, rather than 1961 (the peak of the baby boom). It also means that most baby bust cohorts are larger than most baby boom cohorts. Changes in cohort size should not be confused with population ageing—the large cohorts born during the baby boom years initially made the population younger.
- A youthful age structure typically contains a momentum of population growth, while an older age structure contains a momentum of decline. The momentum effect is the unavoidable growth or decline potential contained with the age structure. For example, at the same time as the number of births are declining (causing structural ageing), the increased numbers of elderly (the result of numerical ageing) are causing an increase in the number of deaths. The two trends are on a seemingly unavoidable collision course that, in Australia, will see a shift from natural increase to natural decline around 2035.
- Net migration gains can have small reducing effects on structural ageing, but, in the long term, add to both structural and numerical ageing. Attempts to maintain either the size of the working-age population, or the ratio of working-age to elderly through replacement migration, would increase the size of host populations beyond what is believed to be socially or politically acceptable. (Replacement migration aimed solely at maintaining population size (in the context of intrinsic decline) is an exception.) A useful means of understanding the trade-off is McDonald and Kippen’s index of efficiency, which demonstrates the percentage reduction in structural ageing for each net million migrants gained. Fertility increase is argued to be a more efficient counter to population ageing than immigration. However, substantial fertility increase may now be unattainable, and in the short term would add to the total dependency ratio.
- Within the total population there are different age structures for different sub-population groups (sex, ethnic and regional). The Aboriginal and Torres Strait Islander population is considerably younger than the total Australian population, while the main immigrant groups of the 1940s and 50s are considerably older. Among the latter, males tend to outnumber females at older ages, which differs from the total population. Differences in age structure by State and Territory indicate that Tasmania and South Australia will begin intrinsic decline several decades before the remaining States and Territories.
- Demographic compression occurs when a number of key demographic events (such as age at child birth, age at which the last child leaves home, the length of the working life/ retirement, the ageing of parents), become compressed into a shorter space of the life cycle. They may also overlap with the demographic experiences of an individual’s own parents and offspring. Different cohorts may have different abilities to respond to inter-generational demands.
- The welfare states of most developed countries were developed at a time when the age structures of these countries were young and juvenescent. The ‘social contract’, pay-as-yougo type of welfare state may in fact require a more youthful demographic structure for long-term sustainability. If so, the Australian welfare state (and others like it) may have a built-in ‘use-by’ date.
- Policy has many dimensions, among which are explicit, implicit, direct, indirect, unintentional, and net effects. These sometimes conflicting dimensions mean that it is almost impossible to attribute a change in a social phenomenon to any single policy intervention. However, ostensibly non-demographic policies (such as higher education fees) can have demographic effects. Where possible, policies should be scrutinised for their potential anti-natal effects.
- Population projections are not predictions. They are computed on clearly specified and biennially revised sets of assumptions about birth, death and migration rates. Because birth and death rates change slowly, and migration into a country such as Australia can be reasonably well controlled and monitored, projections for the immediate years and decades are highly reliable. It is typical to use the medium variant assumptions for regional and international comparisons.
- When undertaking statistical analysis of social phenomena over time, it is important to distinguish between changes due to shifts in age structure (or changes due to other compositional factors, such as marital status), and actual changes in the variable(s) of interest (the ‘true’ or underlying change). The same applies when comparing data for two or more populations at a single point in time. The techniques of standardisation and decomposition are particularly suited to these tasks.
Almost every discussion of population ageing notes somewhere (either explicitly or implicitly) that the phenomenon is the inevitable outcome of the demographic transition. Seldom, however, is the latter itself explained. A simple overview of this phenomenon can assist in demystifying many of its consequences.
The demographic transition in a nutshell
The most succinct description of the demographic transition comes from Paul Demeny (1972), who stated that ‘in traditional societies, fertility and mortality are high. In modern societies, fertility and mortality are low. In between there is the demographic transition’.
Although many would take issue with his dichotomy of ‘traditional and modern’, Demeny’s description is very important for the way it draws attention to the period between the onset and end of the transition. Prior to the onset of the transition, births and deaths are not only high but are more or less in equilibrium—generally cancelling each other out—and population growth is either low or static, sometimes slightly negative (see Figure 1, Stage I). This was the case for most of human history. At the end of the transition, at least theoretically, low to zero, potentially slightly negative population growth is again reached (Stage III). But, during the transition (Stage II), populations grow in size, often explosively (Coale 1972a, b).
Figure 1: The demographic transition (classic or western model)
The growth occurs because, typically, the factor that heralds the onset of the transition is a decline in infant mortality. As infant mortality falls ahead of fertility, more babies survive, causing the population age structure (the numbers or proportions to be found at each age–see Figure 2) to expand at its base and become structurally younger. Within 15–30 years, typically before fertility has begun to fall significantly, most of these survivors have children themselves, causing further population ‘juvenescence’ and expansion.
Once fertility begins to fall, the rate of population growth slows, but the population continues to grow in size for several years, because the next typical occurrence is growth from a phenomenon known as the momentum effect (Keyfitz 1971). The momentum effect is the growth potential that remains contained within the age structure, after fertility has begun to fall. Even if fertility fell immediately to the levels required for the exact replacement of each generation (2.1 births per woman), populations generally continue to grow in size for at least one generation, because each successive cohort reaching reproductive age is typically larger than its predecessor. This is due to the past effects of the higher fertility and falling infant mortality. The higher the pre- or early-transitional fertility and the longer it takes to fall to replacement level, the longer the momentum effect continues.
Immediately fertility does begin to fall, however, the population age structure begins to mature, that is, to become structurally older. As fertility falls, the proportion of the population at the younger ages decreases; concomitantly, the proportion at the older ages increases. This is known as structural population ageing.
Theoretically, the reaching of replacement level fertility was supposed to herald the end of the demographic transition. It would bring with it a return to the situation of zero population growth noted above, or even incipient, intrinsic decline. The latter is a possibly temporary period of population decline caused by deaths outnumbering births, the outcome of increased numbers of elderly (see numerical ageing below) in relation to falling fertility. However, in most of the more developed countries, fertility has either fallen or is continuing to fall well below replacement level, and incipient decline is in danger of becoming population implosion. In the absence of substantial net migration gains, almost all industrialised countries are projected to decline in size over the next 50 years, some dramatically (United Nations 2000). For many this phase has already begun.
With reference to these dynamics, population ageing is best understood by considering it as having two technical dimensions: structural and numerical ageing.
- Structural ageing (an increase in the proportions of elderly) is primarily the result of falling fertility. Falling infant mortality and increasing life expectancy are also involved, in that they add to the numbers and thus proportions of elderly. However, they are not the primary cause of structural ageing: a population will not age structurally while it has high fertility. The latter reflects the situation that occurred during the baby boom (in Australia, between 1946–1965),1 when mortality was low but fertility increased. The result was a short-term juvenescence of the age structure, after several decades of ageing that had begun in the 1880s.
- Numerical ageingon the other hand, is primarily caused by falling mortality. As infant mortality declines, more babies survive, causing a spurt in population growth. Within a few decades these babies become reproducers themselves, causing a further spurt in population growth. Both cause an initial juvenescence of the age structure. As life expectancy improves among the adult population (later in the demographic transition), those who survive infancy and childhood have a high probability of reaching old and very old age. The high fertility over the baby boom years will shortly become a major contributor to numerical ageing (and to structural ageing), in that the numbers born then will begin to reach old age around 2010. However, this will not be the dominant cause. A population will not experience a significant increase in the numbers of elderly if mortality is high, even when fertility has been very high.
This distinction between structural and numerical ageing is very important for social and micro-economic policy. It is numerical ageing that is driving up the demand for and cost of income support, health-care services, and so on, while it is structural ageing that is the constraining factor. Structural ageing will soon2 mean a decline in the proportion of the population at workforce age (that is, the primary tax base), when compared with the increased numbers of dependent elderly, and a reduction in the ability of governments to fund these pensions and services.
The distinction is also important because structural ageing is essentially reversible (that is theoretically responsive to policy), while numerical ageing is not, at least in the short-term. A sustained increase in fertility would cause an immediate reduction in the proportion of the population at the older ages, and, after 18-20 years, an increase in the proportion at workforce age. However, the benefits to the working-age population/tax base would not be realised for those 18–20 years. On the other hand, a sudden or dramatic increase in fertility would create an age structure akin to an hour-glass, dramatically increasing the total dependency ratio (the ratio of 0–14 and 65+ year olds when compared with those aged 15–64 years (see Section 4 below). By contrast, the only way numerical ageing could be reversed would be via a net loss of people at the older ages. All things remaining equal this will not happen until most of the large baby boom and bust cohorts have died—around 2070–80.
Within this somewhat straightforward depiction of structural and numerical ageing is one further important point. Many people confuse population ageing, or more correctly, structural ageing, with the movement of baby boom cohorts through the age structure. The data presented in Figure 2 below clearly show how the baby boom (shaded dark), which initially created a triangular-shaped, youthful population pyramid, is moving upwards through the age structure, augmented, since its birth, by migrants. This upwards movement has indeed contributed to the slowly increasing median age of the population. From approximately 2009, it will contribute to both the proportions and numbers of elderly. However, it is the smaller, post-baby boom cohorts, with the significant exception of the larger, so-termed ‘baby bust’ cohorts born 1968–74, that are bringing about structural ageing, not the baby boom by itself (see also Section 5, below).
Figure 2: Age–sex structure of the Australian population, 1976, 1996 and 2016
Source: Compiled by the author. 1976, 1996: ABS Census of Population and Dwellings; 2016: ABS 2000, Catalogue 3222.0, Series IIa.
Notes: Dark shaded bands = baby boomers
The large cohorts born between 1968 and 1974, following the technical peak (1961) and end (1965) of the baby boom therefore need some explanation. The baby boom is defined in terms of the increase in the TFR or average family size that occurred between 1945 and 1965, not the size of the resulting birth cohort, which is what is shown at each age-point in Figure 2. Although average family size was indeed falling throughout the 1960s and ‘70s, the period often referred to as the baby bust, an increase was occurring in the numbers of women arriving at reproductive age. This reflected the first of the baby boom cohorts reaching this age. Because the size of each birth cohort reflects the combined effect of average number of births and the numbers of women at the key reproductive ages, these momentum effect dynamics caused Australia’s biggest birth cohort to be born in 1971, not 1961 (see section 5, below).
Given the importance of trends in fertility for structural ageing, the next most important concept to understand is the total fertility rate (TFR). The TFR is a synthetic estimation of the average number of children a woman would expect to bear during her lifetime if she were to experience all of the age-specific birth rates occurring in that year. This index, which is calculated for women aged 15–49 years, is also sometimes called a ‘period rate’ because it is based on births occurring during a given period (that is, a year). It contrasts with the completed fertility rate (CFR) which is sometimes called the cohort fertility rate). The CFR refers to the average number of children actually born to woman from a given cohort. Because the CFR requires longitudinal data, it can only be calculated for women who have reached their late forties. As a result of this time delay the TFR is used as an approximation of the CFR (see McDonald 2000 for a detailed description). Importantly, neither the TFR nor CFR permit identification of the number or proportion of women who are having no children, the implications of which are discussed below.
Source: ABS Births 3301.0
Understanding the synthetic construction of the TFR is especially important for understanding the limitations of the index. As shown in Table 1 above, the TFR is the sum of the age-specific fertility rates. These are ratios of the number of births at each age to the number of women at each age. When five year age groupings are used, as in Table 1, the result is multiplied by five, to account for the width of the age band (five years). When single year-of-age data are used, the age-specific rates are simply summed.
The major problem with the resulting index is that it can be heavily distorted by shifts in the timing of childbearing—the average age at which women give birth. As Figure 3 indicates for Australian women, this has changed dramatically across this century, falling from around 28.5 years in the 1930s, to 25.5 years around 1970, returning to nearly 30 years in 1999. An upward shift in age at childbearing tends to lower the TFR, while a downward shift raises it. De Beer et al. explain this:
If, at a certain point in time, increasing numbers of women decide to stop child-bearing… then the total number of births will decrease due to the loss of third or higher order births. If, at the same time, increasing numbers of women decide to postpone the arrival of a first and/or second child to a later age, then the total number of births drops even further… Summing up these age specific rates gives very low TFR values.
After a number of years the tide might turn. The women who postponed childbearing will have grown older and may decide to [begin childbearing] at age 27, 30, or even later [and] the fertility rates at these ages [will] start rising again. [If] at the same time, the youngest generations …prefer to have their first and or second child at young ages again, then fertility rates at these ages will also start rising. [T]heir sum, the TFR, ends up at a high level again. (De Beer et al. 1991, p. 40)
All women in this (hypothetical) example had two children. Their completed fertility didn’t change, only the age at which they had those children.
Figure 3: Median ages of mothers (all births), Australia 1921–98
Source: Compiled by the author. ABS Catalogue 3301.0, Various Years
Understanding the distinction between the TFR and CFR is especially important because when compared across time, the TFR is typically higher than the highest CFR, and lower than the lowest CFR (see also (ABS Births 1999; and Wilson 1985, p. 221). The disparity tells us that very high TFRs are likely to over-estimate average completed family size, while very low TFRs under-estimate it. As Figure 4 shows, the point is especially pertinent to the cohorts who gave birth to the baby boomers. The cohort born 1930, for example, experienced its peak childbearing years during the late 1950s to early 1960s, when the TFR was peaking around 3.6. According the CFR, the average completed family size for these women was fractionally above 3.0.
Figure 4: Total fertility rate 1921–99, and completed fertility rate for cohorts born 1905–60 lagged by 30 years, Australia
Source: Compiled by the author. ABS Catalogue 3301.0, Various Years
Notes: The latest completed fertility data available are for the cohort born 1950; data for cohorts born 1950-60 have beenestimated. The lagging of cohort data by 30 years permits comparison of the approximate TFR over the peak of these cohorts childbearing, against the actual average family size.
The disparity between the TFR and CFR has both micro- and macro-level policy implications. At the micro-level, the TFR under or overestimates such things as the number of children that each family will be supporting, and how many children each generation of parents will have to call on for support in their old age (see also the parent support ratio section 4, below). At the macro-level, the age-specific fertility rates (of which the TFR is comprised) are used to project the size and structure of the future population. Simply stated, current age-specific fertility rates are applied to (multiplied by) the number of women projected to be at each age, at each successive year (see section 12, below).
This calculation, when adjusted for mortality and migration, gives the number of new entrants (births) to the population. If the number of new entrants to the population age structure are over or underestimated, so too are the quantum and tempo of structural ageing. Indeed it should be noted that the most recent ABS 2000 high-range projections are based on a TFR of 1.75 through to 2051. For the lower-range projections this drops to 1.60 after 10 years, and remains constant across the projection period. Given that Australia’s TFR is already below the upper level, is close to the lower level in the Australian Capital Territory and Victoria, and that fertility is somewhat lower in many countries, the assumptions may be too high. If Australia’s fertility does in fact fall below the assumed levels, as many demographers expect (McDonald 2000), structural ageing will be more pronounced and will occur faster than anticipated.
As also noted, a further very important point is that neither the TFR nor the CFR give an indication of the number or proportion of women who are having no children at all. That is to say, the TFR is an average for all women of reproductive age (15–49 years), while the CFR is an average across women belonging to a specific cohort. Accordingly, in a context
where increasing numbers of women are remaining childless in Australia currently estimated at 20 per cent (Merlo & Rowland 2000), a TFR or CFR of 2.1 or less indicates that many of those who are still having children, are having more than two.3 As McDonald (1998) explains, Australia’s 1996 TFR of 1.8 was being held up by the relatively high proportion of women still having three or more children, around 25 per cent. However both age-specific and parity data for each successive cohort indicate that this proportion is falling sharply. Whether the fall will ultimately be mirrored in the CFR is open to conjecture, but, like the TFR, the CFR is likely to remain below replacement level (Bongaarts 1999; McDonald 2000). In that case it will have long-term implications for both structural ageing and population size (see also section 6, below).
As indicated in the previous sections, closely related to population ageing and its measurement is another important concept: life expectancy. Life expectancy concerns the probability of survival, and, similarly to the TFR (although quite differently constructed), is a synthetic measure based on the age-specific death rates occurring in a given year. Typically, the term life expectancy at birth (eo0 ). However, because the most dangerous days of life are the first and last, surviving the first days, weeks, months, and then year of life generally results in an increase in life expectancy. At birth, a person’s life expectancy may be 60 years. But if that person survives to age 60, their remaining life expectancy (eo60) will these days typically be another 20 or more years. Thus, life expectancy for any given age specifies the number of additional years the average person can expect to live. Table 2 illustrates this phenomenon for Australians born in 1932. Males born in 1932 had, at birth, a life expectancy of 63.5 years. For those who reached this age, a further 17.59 years had been added, giving an average minimum life span of 81.1 years. The data for females can be similarly interpreted.
|Life expectancy at birth||On reaching life expectancy at birth||Average minimum life span|
Source: ABS Deaths, various years
Notes: For males, on reaching 63 years; for females, on reaching 67 years
Life expectancy at birth is also often confused with average age at death. From the above discussion it can be inferred that the two measures relate to quite different populations and situations. Life expectancy at any age refers to average years of life remaining for those born in a given year (a birth cohort); average age at death refers to all those dying in a given year (thus from many cohorts). The distinction is important in a policy-making context, for example, in relation to projecting demand for Age Pension, because average age at death (say, 80 years) will always be lower than the average life expectancy remaining (say, nine years) for those reaching this age.
Life expectancy also differs substantially between men and women (typically between two and eight years), and between people of different socioeconomic and ethnic backgrounds. These understandings are critically important for policy makers endeavouring to determine future demand for and access to pensions and services, health care, and so on. Sex-specific (and possibly ethnic-specific) measures of life expectancy should be employed in, for example, the rationale for setting the age of eligibility for access to certain goods and services of the welfare state.
The most commonly used indicators of population ageing are the proportion of the population aged 65 and over, and the median age (the age above and below which half the population fall). Populations are considered young when less than 5 per cent of the population is aged 65 and over (or more than 35 per cent is aged less than 15 years), and oldwhen this proportion reaches 10 per cent, although in developing countries where mortality is still high, it is practical to take 60 years as the cut-off age. Similarly, a population is considered young when it has a median age of less than 20 years, and old when this index reaches 30 years. Populations in between these extremes are considered to be of intermediate age.
|Proportion aged 65+||Change
|Hong Kong S.A.R.||36.5||10.7||16.1||50.5|
|United States of America||35.9||12.6||16.5||31.0|
Source: United States Census International Database
Currently (2001), the median age of the Australian population is 35.4 years, and approximately 12.5 per cent are aged 65 and over. Australia is therefore considered to be an old population. However, as Table 3 shows, it is relatively young compared to the populations of several other developed countries.
Other common indices or proxies of population ageing are the ‘dependency’ or ‘support’ ratios. Conventionally, four such ratios are recognised:
- youth: 0–14 year olds in relation to those aged 15–64 years
- aged: 65+ in relation to those aged 15–64 years
- total: 0–14 and 65+ year olds in relation to those aged 15–64 years
- potential support ratio (PSR): 15–64 years in relation to those aged 65+
Whether as indices of population ageing, dependency, or support, these measures are extremely crude. They reflect a time when people (mainly males) entered the labour force at age 15, left it at age 65, and were employed full-time between those two ages. Today, the upper and lower boundaries delimiting the economically active population are much more fluid, while many of those aged 15–64 years are in fact ‘dependent’, for example, the unemployed and jobless, youth living at home, those people receiving illness, disability and other support pensions, and those people studying full-time, and/or caring for others (mainly women). More refined dependency ratios should be constructed depending on the uses to which they are being put. The upper and lower boundaries should reflect, for example, average age at labour force entry and exit, while for certain purposes the number receiving more or less full income support should be removed.
With these limitations in mind, Figure 5 below, gives the crude dependency ratios for Australia for the 0–19 and 65+ year age groups, compared with those aged 20–64 years. (Note that international comparisons are typically based on a working-age population of 15–64 years.) According to these indices, Australia’s total dependency ratio will reach its lowest point in approximately 2009, after which time it will return quite rapidly to levels existing in the 1970s. However, by contrast with the 1970s, the driving forces of this change are, predictably, declining youth dependency and increasing aged dependency. This unprecedented change in the composition of the total dependency ratio is very important to understand because of the relatively greater costs associated with aged dependency, and because these costs are largely borne by government. It is argued that the cost to the government of support for the elderly is between two and four times that for children (Borowski & Hugo 1996, p. 49, who cite a number of studies).
Figure 5: Youth (0–19 years), aged (65+ years) and total (0–19 and 65+ years) dependency ratios, Australia 1971–2051
Source: Compiled by the author.
1971-1998: ABS Population Estimates; 1998-2051: ABS 2000, Catalogue 3222.0, Series IIa
Concealed within these indices is also the fact that between 2011 and 2051 the proportion of the Australian population aged 20–64 years, the primary tax base, is projected to decline from its peak of just over 61 per cent, to around 54.4 per cent. For the population aged 15–64 years, these figures are 68.1 per cent in 2009 and 59.6 per cent in 2051. The PSR, which is widely used in United Nations analyses, is illustrative of the impact. Currently sitting at 4.9 persons aged 15–64 years to each person aged 65 and over (having fallen from 6.5 in 1972), the ratio will fall rapidly to 3.0 by 2024 and 2.1 by 2051 (ABS Series IIa).
Two other very useful indices of population ageing are the aged/child and parent support ratios. The former measure directly compares the two age groups (0–14 and 65+ years) that undergo the most change during demographic transition. Because this measure is the most sensitive to changes in the age composition, it is conventionally considered the best index of ageing (Stockwell 1976). As Table 4 shows, for Australia this index will decline from 1.7 children per person aged 65+ in 2000 to 0.6 in 2050. According to these projections, the two will be briefly in balance around 2016–18; thereafter those aged 65+ will outnumber those aged 0–14 years.
|Aged/child Ratio (1)||Parent support Ratio (2)|
Source: Compiled by the author
ABS Population Projections 2000 Series IIa
Notes:(1) 0–14 years : 65+ years
(2) 45–54 : 75+ years
NB. When constructing the parental support index it is important to keep intergenerational shifts in the timing of family formation in mind, and also the fact that not all adults have children.
By contrast, the parent support ratio measures the relative size of offspring (for example, 45–54 years) and ‘parental’ (75+ years) cohorts to approximate potential family support available to the elderly. (The latter is also sometimes termed the parent/progeny ratio. It differs from the potential support ratio described above in that it is based on relational age groups.) This index will similarly decline from 2.4 in 2000 to 0.9 by 2050. Note that, like the TFR and CFR discussed earlier, this ratio implies universal childbearing, whereas in reality a proportion of adults never had children.
The concepts of birth rate, cohort size, and population ageing are often used interchangeably and incorrectly.
As outlined above, the birth rate, whether calculated as the TFR or the CFR, is an index used to approximate average family size. It has a number of limitations, not least that it conceals the extent to which an increasingly large proportion of people are not having children.
A cohort, on the other hand, is a group of people connected by a similar event. This may be birth in a given year (which derives a birth cohort), marriage in a given year (a marriage cohort), death (a death cohort), or even a war (those who were young adults between 1939 and 1945 are sometimes referred to as the war cohort). Cohort size in relation to a birth cohort refers to the number of people born in any given year, later augmented by immigration or reduced by emigration and death.
Despite the apparently clear distinction between the two concepts, they are often confused. For example, much attention has been directed towards the large cohorts born during and especially at the end of the baby boom. At its peak (in 1961), the TFR was 3.6 (and cohort size 239,986). However, as noted earlier, the cohort born in 1971 was considerably larger (n=276,361). This occurred because, although fertility had by then fallen to 2.9 births per woman, there were more women giving birth, the first of the baby boom generation having arrived at reproductive age (the momentum effect as outlined in section 1, above). In other words, cohort size (the number of births in any year) is the combined function of the birth rate and the number of women of reproductive age (and, of course, actually having children).4
The distinction between the two concepts (cohort size and the birth rate) is clearly illustrated in Figure 6, as is the momentum effect. The outcome of the momentum effect is that most of those born during the so-called ‘baby bust’ (1968–74) in fact belong to cohorts that were, and in most cases remain, larger than their baby boom parents and predecessors. Indeed ‘baby bust’ should be considered a misnomer.
Despite similarly clear technical distinctions, changes in cohort size are also often confused with population ageing. In particular, as noted earlier, the movement of the baby boom cohorts through the age structure is often referred to as population ageing. However, as explained, the changes in cohort size that occurred during the baby boom were part of a short-term shift to a younger population, not an older one. Also, seemingly paradoxically,
since most of the baby bust cohorts are larger than the baby boom cohorts, population ageing will not only continue once the baby boomers have reached very old age and begun to die, but may even accelerate. This will depend upon what happens with fertility in the meantime. Accordingly, structural ageing may be better conceptualised as a function of declining cohort size, than declining birth rate.
Figure 6: Total fertility rates and cohort size, Australia, 20th Century
Source: Compiled by the author
ABS Australian Demographic Trends 1997 Appendix 16; ABS Births, various years
Notes: Data exclude Aboriginal and Torres Strait Islander Population prior to 1966
According to demographic transition theory, significant fluctuations in cohort size are not expected to re-occur once the demographic transition is complete (Coale 1972 a, b). This is because, theoretically, the population age structure will reach the state of zero population growth noted earlier (births and deaths will be more or less equal) and become stable. The proportions at each age will not change appreciably from year to year. As could be seen in the panel for 1996 in Figure 2 (see section 1 above), significant fluctuations in cohort size at birth have already ceased to occur.
However, as was also implied, projections assuming zero growth and population stability are dependent on one very important factor—fertility returning to and remaining around a TFR of 2.1 births per woman, the theoretical replacement ratio. Currently, approximately 60 of the world’s populations have fertility lower than this. In Continental Europe, for example, the TFR ranges between 1.1 and 1.4 (United Nations 2000). If fertility fall to these levels in Australia, newly born cohorts will continue to decline in size, and structural ageing will accelerate.
These points aside, for both policy makers and analysts, changes in cohort size will remain very important for some time. They are, at this moment in Australia, more significant than population ageing. Two examples will suffice. First, the cohorts currently (2001) entering the elderly population are those born in the 1930s, and are smaller than either their predecessors or successors (see Figure 6). According to Borowski and Hugo (1996):
this group’s passage through the older ages will lead to a significant reduction in the pace of ageing in Australia in the 1990s and early twenty-first century …However, rapid growth of the elderly population will recommence and reach unprecedentedly high rates when the
post-war baby boom children begin to enter the retirement ages after 2011 (p. 27).
Figure 7 illustrates the situation using data for the 55–64, 65–74 and 75+ age groups. Over the next decade the population aged 55–64 years will grow at a considerably faster rate than the population aged 65–74 years. This is because the first of the baby boomers are now entering the former group. Thereafter, as they leave the first group and move into the second, the population aged 65–74 will grow at the faster rate. Finally, as the baby boomers reach the 75+ age groups, the latter population will grow at the fastest rate. They will outnumber both the 55–64 and 65–74 year age groups by between 2030 and 2035. The magnitude of the shift will be nothing short of remarkable, with the 75+ group growing from 1 million at present to more than 3.5 million by 2051. Within these broad age groups, trends for individual age groups are even more pronounced. At the older ages, significant differences between each sex should also be noted: at each successive age, women increasingly outnumber men.
Figure 7 clearly illustrates the importance of disaggregating the elderly population. Not only will there be successive waves of elderly, but each wave will differ from its predecessors (Mackay 1997). Indeed, when considering distinctions between cohort size and population ageing, one further distinction, that between the cohort and the age group, is also warranted. Over time, cohorts age (the people in a birth cohort grow older); age groups do not (people pass into and out of them). As a result, the waves of elderly age groups contain cohorts that have had very different life experiences (especially among women), including differences in education, income, savings behaviour, labour force attachment and childbearing. These differences pertain not only to level, but also to timing. As Easterlin (1988), Hagenaars (1990), MacKay (1997) and others have argued, each cohort encounters certain period events and circumstances (such as a depression, economic boom or restructuring) at a different age. This nexus has the potential to develop into cohort effects. For example, cohorts that encounter a situation of full employment around labour force entry age, such as the cohorts born in the
1930s, may experience higher lifetime levels of employment and savings potential than cohorts that experience the opposite. Such differentiated cohorts deal with each life stage in different ways, and are likely to require (and demand) quite different retirement experiences.
Figure 7: Projected increase in populations aged 55–64, 65–74 and 75+ years, Australia
Source: Compiled by the author
ABS 2000, Catalogue 3222.0, Series IIa
Each Australian cohort is also differentiated ethnically, with high proportions (around40 per cent) of the oldest cohorts born in the United Kingdom/Eire and Europe (Hugo 1988, see section 8, below). More recently born cohorts have higher proportions of, for example, people born in Asian countries. The implications of important information such as this are rendered invisible when trends in ageing are analysed by age group only.
The second example concerns cohorts currently at the younger end of the age spectrum. It has been argued (Easterlin 1988 and others) that large cohorts experience greater intra- and inter- cohort competition for available resources (such as education, jobs and income) than do small cohorts. As a result, large cohorts are likely to have a more negative labour market and earnings experience, and, subsequently, to have later and lower fertility, than small cohorts. Potentially substantiating the argument, both the extremely large cohort born in 1971, and those immediately surrounding it, have been strongly affected by unemployment, and have the lowest and latest fertility to date.
The implications of the situation are manyfold. For example, as this very large cohort leaves behind the high youth unemployment years, as it has recently done, the employment earnings situation is expected to improve for its successors. Described as a youth deficit by the American Central Intelligence Agency (CIA 1990),5 the situation of declining proportions of youth is expected to see an increase in global competition for the labour and skills of young people. Certainly as Figure 8 shows, labour market entry-exit ratios for Australia are now falling rapidly. Currently just on one 18–24 year old is at labour force entry age for each 55–64 year old reaching retirement age and beginning to leave; by 2018 this ratio will fall to 0.8; and by 2018, below 0.7. These factors need to be borne in mind when analysing or attributing findings to particular policy innovations that are attempting to reduce youth unemployment. At least part of the reason for a decline in unemployment could simply be a function of population ageing (see section 13 for methodological implications).
The positive implications of this trend notwithstanding, the earlier point that cohorts carry their accumulated experiences with them should be kept in mind. It should also be noted that the large cohorts born around 1971 are currently entering their main childbearing years. Despite low and still falling fertility, this shift could herald a small increase in the number of births, reflecting a momentum effect, and a concomitant increase in demand for child-related goods and services such as paediatricians, schooling, and family-related payments
Figure 8: Labour market entry-exit ratios (18–24:55–64 years), Australia 1971–2051
Source: Compiled by the author
1971–98 ABS Population Estimates; 1997–2051: ABS 2000, Catalogue 3222.0, Series IIa
Many people are familiar with the term ‘natural increase’ (technically called intrinsic growth because it occurs within a population, as opposed to externally from migration). This is simply the difference between births and deaths. Over the past two hundred and fifty years, that is, since the onset of Stage II of the demographic transition, the natural increase component of population change has become taken for granted.
Since the 1950s, when many of the developing countries began their transition, concerns with the global rate of growth in natural increase (RNI) have become associated with the concept of ‘doubling time’—the time it takes for a population to double in size. As a rule of thumb this index is estimated by dividing 69.3 years by the annual rate of growth (Weeks 1999, p. 11). Between 1950 and 1985 this gave a world population doubling time of about 35 years. As outlined in section 1 above, the main reason for the dramatic rate of growth was not, as many believe, high or increasing fertility in the developing countries, but falling infant mortality, which saw more babies survive and natural population growth compound.
With fertility now also falling in most developing countries, the momentum effect described in section 1 is under way, resulting, for most, in massive population growth, but growth that is occurring at a decreasing rate. Indeed, the deceleration in the world’s population growth rate is nothing short of astonishing, from 2.0 per cent per annum in 1970, to 1.3 per cent in 2000, deriving a current doubling time of greater than 50 years.6In addition, in many developing countries HIV/AIDS is expected to cause an increase in mortality rates over the next two decades, as well as decimate reproductive age populations, with a loss of the children they would have borne (U.S. Bureau of the Census 1999). For these reasons, world population projections are being constantly revised downwards, with numbers in most developing countries expected to peak and begin to decline towards the end of this century (see also Lutz 1994, 1996).
As implied earlier, this opposite trend towards intrinsic (natural) decline, and potential concerns with population halving time, in the developed countries is well established. As McDonald (1998 p. 3) explains, its dynamics are simply the obverse of the above. Just as a young age structure contains a momentum of population increase, so too an old age structure contains a momentum of population decline:
If women, on average, have just one child …then the size of the generation will halve in one generation that, in demographic terms, is about 28 years. In 56 years, the generation size will only be a quarter of what it was two generations beforehand.
At the same time as the decline in fertility is driving down the number of births, the increasing numbers of elderly are driving an increase in the number of deaths.7 With the two trends on a collision course, the likely outcome is a cross over, and natural decline. Figure 9 illustrates the situation for Australia, where natural decline is projected to occur during the third decade (see section 8 below for regional differences; see also ABS Births 2000). It should be noted that these data include migration at the medium variant assumption.
Figure 9: Births and deaths, Australia, 20th Century, and projected
Source: Compiled by the author
ABS Catalogue 3301.0, Various Years; ABS 2000, Catalogue 3222.0, Series IIa
The extent to which individual populations will actually, rather than theoretically, decline, is difficult to determine, because as natural decline approaches it is likely that extra efforts will be directed at stabilising the birth rate (see section 11 below). Also, at least in the short term, increased migration is likely to be used to ameliorate the impact in countries such as Australia (see section 7 below for the feasibility of this option). However, what is singularly important to understand is that the shift to natural decline is not a cyclical trend. The one-off natural growth that accompanied the demographic transition is now over for the developed countries, and is expected to be over for the developing countries before the century’s end. Furthermore, if the birth rate continues to remain substantially below replacement level (2.1 births per woman) or declines further, intergenerational halving time has the potential to become total population halving time. An overall growth rate of -0.5 per cent would derive a halving time of 140 years; -1.0 per cent, 70 years, and so on. Such a situation would cause a further dramatic upward shift in the age structure (hyper-ageing), and, among other things, a concomitant incapacity to sustain a social security system of the type Australians currently enjoy. This latter is, of course, in the absence of social and economic changes that would, for example, increase productivity or delay retirement.
As will be elaborated in the following section, such scenarios are not merely conjecture. Current fertility levels in Germany (TFR 1.4) for example imply a negative rate of natural increase (in other words, natural decline) of –1.7 per cent. If maintained for 200 years, in the absence of a substantial increase in migration, such a rate would shrink the German population to one-thirtieth its current size (Demeny 1986, p. 153). Similarly, with reference to Italy (TFR 1.2), McDonald (1998, p. 3) explains that ‘once the impact of the crude birth rate on the current age structure has been wiped out (in about 40 years), the [Italian] population in the subsequent 100 year period would fall to just 14 per cent of its current level’.
Migration is often proposed as the answer to population ageing. That is, because migrants are typically concentrated at younger ages than the host population, a net gain from international migration is argued to assist in keeping a population young; or, more accurately, in keeping the labour force (and primary tax base) from declining in proportion to the elderly population. More recently, the emerging reality of natural decline has come to the forefront of the debate, resulting in an awareness that in the near future, replacement migration8 (United Nations 2000) will have to address three issues:
- maintenance of the size of the total population;
- maintenance of the size of the working-age population; and
- maintenance of the ratio of working-age to elderly.
The arguments have been broadly debated, but the general consensus is that migration will be hard pressed to solve the emerging problems (United Nations 2000).
First, the numbers required to offset structural ageing are enormous. Table 5 below, shows United Nations projections for a selected range of countries expected to undergo extreme ageing and intrinsic decline during the next 50 years. Even with the addition of sizeable numbers of migrants at the medium variant assumption level, these projections show Germany declining by just under 9 million (11 per cent) by 2050, Italy by 16 million (28 per cent), and Japan by 22 million (17 per cent). In order to keep the Italian population at its current size, Italy for example would have to take in a net 251 000 migrants per annum. This is many times greater than Italy’s historical experience (the medium variant for Italy is 6 000 a year). This level of net intake would total approximately 12.5 million migrants over the period. To maintain the Italian working-age population at its current size, that intake would have to be around 372 000 per year (a net of 19 million across the period); and to maintain the current ratio of working-age to elderly (the potential support ratio), the net number of migrants needed would be in the vicinity of 2.3 million per annum, or 113 million across the period. This amounts to twice the current population, and few of whom would be ‘Italians’.
|Current size 2000||Medium variation migration||Projected size 2050 medium variant*||Constant total population size||Constant age group 15-64 years||Constant age group 64/65+ years|
|Country/region||Thousands (per annum)|
Source: United Nations 2000, Tables 1, IV.14, IV.19, V.22,
Notes: * Includes migration at the medium variant assumption
Kippen (1999) illustrates the structural aspects of the argument for Australia. Currently, 12.5 per cent of the population is aged 65 years and over. Under conditions of zero net migration, and the TFR falling to 1.65 by 2008 and then remaining constant across the 21st Century, the percentage aged 65 and over would increase to 32.6 per cent. With the same fertility assumptions, and annual net migration gains of 80 000, the proportion aged 65 and over in 2098 would be reduced by a mere 4 percentage points, but 10.8 million would have been added to the population (over the zero net migration scenario).9 A net migration gain of 160 000 per year with similar fertility would reduce structural ageing in 2098 by a further 1.6 percentage points (to 26.9 per cent), but in total would add 15.4 million to the population.
Demonstrating these trade-offs, McDonald and Kippen (1999, p. 14) have developed a very useful index of efficiency (see box below), which shows the gain in numbers for each percentage point reduction in aged population.
McDonald and Kippen’s index of efficiency
‘The index of efficiency measures the population increase resulting from the migration changes required to reduce the proportion of the population aged 65 years and over by one percentage point. For example, a shift in annual net migration from zero to 50,000 would reduce the proportion aged 65 years and over by 3.05 percentage points by the year 2098. The same change would produce an increase in the total population over the same period of 6.72 million. Hence:
This means that, with this change in the level of migration, a one percentage point reduction in the aged population can be obtained at the cost of an addition to the size of the population of 2.2 million people. An efficient change would be one that minimised the increase in the population for each one percentage point reduction in the proportion of the population aged 65 years and over.’ (McDonald and Kippen 1999, p. 14)
Summarising, Kippen (1999, p. 22) argues, first, that ‘if we wish to minimise the proportion aged 65 plus and limit population growth, maintaining the birthrate is more efficient than increasing migration’ (see also McDonald & Kippen 1999). She shows that a scenario of zero net migration across the century and a rise to near replacement level fertility by 2008 would see the proportion over the age of 65 in 2098 being around 26.4 per cent (compared to 32.6 per cent with the TFR of 1.65), against a total population of around 21 million. It is interesting to observe that despite their growing concerns with natural decline, few of the European Union countries are as yet preferring the migration option, focussing their efforts instead on raising or maintaining their birth rates (United Nations 1999).
Second, it is unlikely that the Australian birthrate will be raised, or even maintained, at least in the short term, and especially through immigration. Not only is Australia’s TFR expected to fall towards that of similar countries within a decade, but increasingly, the births of Australian immigrants are also trending toward these patterns and levels. The fertility of several immigrant groups is already lower than that of the total population, thus adding to structural ageing (Abbasi-Shavazi 1998; Abbasi-Shavazi & McDonald 1997). Furthermore, a number of commentators have argued that trying to create a fertility increase through pro-natalist policies is less desirable than encouraging migration, because it takes many years for the effects of an increase in the birthrate to have an impact on the population of economically active young adults, while migration has an immediate effect (Heer 1986; Simon 1984; see also Höen 1987 on Europe).
Third, the problem with the latter argument, aside from the massive numbers that would be required, is that because migrants also age, they add to the problem of population ageing in the longer term (Young 1989, 1990; United Nations 2000). This point has been convincingly demonstrated for Australia. Kippen and McDonald (2000), for example, show that Australia’s current age structure is almost identical to what it would have been, had there been no net migration gain since 1945 (see also McDonald & Kippen 2000; see also Le Brass 1991).
Clearly, these arguments and their associated trends and patterns have significant implications for Australia’s future. With substantially higher fertility and per capita net migration gains than most of her counterpart countries, and natural decline not projected to begin until the third decade, Australia’s immediate situation is not as dire. From around 2030, however, replacement migration will need to be pursued in earnest, if Australia’s population size is to be held constant (see McNicoll 2000 for a critique of this ‘imperative’). However, Australia’s previous sending countries are those that are already or imminently anticipating intrinsic decline. Many have already become receiving countries, and others, such as Japan, which has had very little experience of immigration, are now faced with this option or with its economic consequences.
As the United Nations (2000, p. 22) points out, the European Union and the United States, currently the world’s two largest economic bloc, are projected to follow starkly contrasting demographic paths in the near future. By 2050, the population of the European Union will have declined in size by around 41 million, while that of the United States will have increased by around 82 million (however, it should be noted that it will also have peaked and be beginning to decline). The result will see the population of the United States, which in 1995 was 105 million smaller than that of the European Union, exceed the latter by 18 million. The economic and political implications of such divergence are large.
Thus, although migration is a poor counter to population ageing by itself, when considered in the concomitant context of intrinsic decline, it becomes obvious that it will be one of the major policy issues, if not the major issue, of the 21st Century. The feasibility of formulating and adopting suitable migration policies poses enormous challenges for governments that decide to pursue this option. Competition for migrants will be extreme. Moves to boost population growth will result in, among other things, massive and more rapid changes to the ethnic composition of host countries than previously experienced. Australia’s future migrants will almost certainly be ethnically different to those of the past. Along with such changes will come enormous cultural, social, economic and political changes to both host and donor countries, not least because the sought-after migrants are highly likely to be the educated young of the developing countries.
One often overlooked yet extremely important point concerning population ageing is that the extent and velocity of ageing may not be equal for all sub-populations, such as ethnic groups, or regions, within the total population.
Unfortunately, lack of appropriate data makes it impossible to construct true age structures for most ethnic groups.10 However, the significance of the phenomenon can be illustrated by comparing data for the indigenous Aboriginal and Torres Strait Islander and total Australian populations. As Figure 10 shows, Australia’s Indigenous population has a considerably younger age structure than the total population: the median ages of the two populations are, respectively, 20 and 35 years. These differences mean that as a proportion of each population there are almost two Indigenous children (0–14 years) for each non-Indigenous child, and at 15–24 years, 1.3.
By and large, the difference between the two populations reflects the higher fertility and more recently, though also more slowly, falling infant mortality of the Aboriginal and Torres Strait Islander population. But it is also partly classificational: according to ABS definitions, an Aboriginal or Torres Strait Islander is any person who claims descent from an Aboriginal or Torres Strait Islander, and is accepted as such by the Aboriginal or Torres Strait Islander community in which he or she lives. This definition means a potentially exponential growth in the number of births attributed to the Indigenous population (ABS 3230.0, 3231.0).
Figure 10: Age-sex structures of the Aboriginal and Torres Strait Islander and total Australian populations, 1996
Source: Jackson 1999, Figure 2.2
Indigenous population numbers are also highly likely to be affected by the phenomenon of category jumping, whereby individuals of mixed descent identify differently, often inadvertently, between censuses and various data collections. According to Gardiner and Bourke (2000), a sizeable proportion of this unexplained growth can in fact be explained by reference to historical factors, such as the suppression of Aboriginal identity through the stolen generation and its subsequent reclaiming in recent years (see also Pool 1991 on the New Zealand Maori).
These and other identificational and classificational issues are very important for the policy maker and analyst to engage with. How the boundaries of a group are technically defined affects the size, structure, and growth rate of the group, with important implications for equitable resource allocation and so on. The rapidly increasing number of Indigenous children and young adults poses a significant social and economic policy challenge, in terms of resources to meet their educational, employment, family formation, housing, and health needs. If these needs are not met—if, for example, there is no recognition of the resource needs of a youthful population existing within the midst of a total ageing population—Indigenous marginalisation is likely to increase.
Also of importance is that such markedly differing age structures can inadvertently result in (or conceal) discrimination, through policies that may be ‘ethnically-neutral’ on the surface (Jackson 1994, 1998a). A policy that, for example, raises the age of eligibility for the adult rate of an unemployment-related benefit, is likely to have a disproportionately negative impact on a younger population. So too is a policy such as mandatory sentencing, given that a younger population is disproportionately exposed to the risk of the type of activities that see young people arrested. (These points are equally pertinent to regional differences in age structure, which are discussed below.)
Despite the difficulties in determining the age structures of Australia’s immigrant groups, country-of-birth data do provide an indication that is useful for policy purposes (see Table 6). The extremely high median ages of the European-born populations, which also comprise the largest ethnic groups among those aged 65+, should be especially noted.
A breakdown of these data by sex also indicates that, by contrast with the total population, some immigrant groups (particularly Italian, Polish, Greek, Dutch, and former Yugoslavian) have higher proportions of elderly males than females (currently affecting 75–84 year olds). This may reflect lower levels of marriage earlier on. These points are even more pertinent when English-speaking ability is considered. Approximately one in five older Australians was born in a non-English speaking country, and a significant proportion, which is known to increase as people age, is unable to communicate effectively in English. Hugo (1998) has shown that this phenomenon affects mainly female immigrants, because while most male immigrants of the time worked alongside English-speaking Australian’s and learned the language, their wives remained at home to raise children.
|Median Age||Percentage 65+|
|Former Yugoslav Republic||39||46||54||9.9||20.0||16.0|
Source: ABS 1981 and 1991 Censuses; 1999: ABS Migration Cat. No. 3412.0, pp. 83, 88.
With the non-English speaking background group currently increasing at a faster rate than the mainly English-speaking aged, it is important to reflect on the extent to which an aged care system developed by and for a primarily English-speaking population, can respond to the changing population’s needs (Hugo 1988, p. 33). Hugo cites Bertilli (1980, in Ware 1981, p. 95) as arguing: ‘it is of no use to an elderly person in need of constant supervision and care to be admitted into a nursing home where he or she cannot easily communicate with staff…psychologically and mentally it would be devastating: it would mean that the elderly person has entered a tomb before the time of death’.
Also, as noted earlier, just as the size of each birth cohort may differ and create waves of population, so too changes in the sending countries of migrants may create waves of ethnicity that are characterised by age and cohort (see Figure 11). The shifts have implications for the type of services that are and will in the future be needed by immigrants. Significantly, the data indicate that the early post-war migrant populations are needing aged-related assistance and resources now, not after 2010 when the ageing of the total population begins in earnest. By contrast, the elderly of the future (say 2030) will be disproportionately Asian.
Differences between regional age structures have equally significant policy (and economic) implications. As Figure 12 indicates, the populations of South Australia and Tasmania are substantially older than those of the Australian Capital Territory and the Northern Territory, and are projected to age at a faster rate. Tasmania will take over from Australia as the oldest State around 2016, and the gap between the two will slowly increase. By 2051, the Northern Territory will have a smaller proportion over the age of 65 than either South Australia, Tasmania, or the Australian Capital Territory have at present.
Figure 11: Percentage of each age group born overseas, by region of birth, 1998–99
Source: Compiled by the author
ABS Catalogue 3412.0, 1998-99, Table 6.3
Notes: *Asia = North-East, South East, Southern, Middle East/North Africa
Figure 12: Projected percentages aged 65+ years, selected States and Territories
Source: Compiled by the author
ABS Population Projections 1997-2051, Catalogue 3222.0, Series IIa
Figure 13 illustrates these differences in terms of the projected decline in the rate of natural increase (births minus deaths). Although Australia as a whole is projected to begin natural decline some time during the mid 2030s, it is very clear that Tasmania and South Australia will be experiencing this phenomenon much earlier (it should also be noted that these data are based on the ABS Series Ia projections, which is the ‘best case’ scenario (see section 12,
below). Natural decline is not expected to occur in Victoria and New South Wales until the late 2030s and 2040s respectively, while Queensland, Western Australia, the Australian Capital Territory, and the Northern Territory are not projected to go into natural decline before 2051.
Figure 13: Projected rate of natural increase and decline (per 1 000 persons), by State and Territory
Source: Compiled by the author
ABS Population Projections 1997-2051, Catalogue 3222.0, Series Ia
Demographic compression refers to the inter-generational phenomenon that occurs when a number of key demographic events are compressed into a shorter space of time, due to generational changes in the age at which women give birth; the age at which children become independent of their parents; trends in labour force entry and exit ages, and so on (Sceats; Young 1990; McPherson 1992; Jackson 1998b).
As a relatively simplistic example, imagine that one generation (B) begins having its children on average at age 22, and that those offspring (generation C) begin having their children on average at age 28. Assume that each generation has two children two years apart, and that the second child goes to university at age 20. When the second child of generation B parents reaches university age, the parents will be aged 44. When the second child of generation C parents reaches university age, the parents will be aged 50. Under this scenario, and assuming a retirement age of 65 years, generation B parents will have, on average, around 21 years in which to see their last child through university and concentrate on their own superannuation provision before retirement. Generation C parents will have around 15 years. Any further delay in the timing of childbearing or reduction in age at retirement would see the period available for savings decrease.
The analysis may be further complicated by the age at which generation B’s own parents (generation A) had its children. Until the mid-twentieth Century it was uncommon for retired people to have their own parents still living; today, the ‘sandwich generation’, wherein older cohorts have both dependent offspring and parents, is increasingly common (Young 1990).
As implied, such an analysis will also be complicated by inter-generational changes in the proportion of life spent in the labour force, compared with changes in life expectancy. Currently, people are living longer than ever before, but, at least for males, spending a shorter period in the formal workforce. Ruzicka (1986 p. 22) estimated that the average male aged 15 years in 1933 would spend approximately 44 years or 83 per cent of his life in the labour force; over the 1940s and 1950s this increased slightly to 84 per cent, but by 1981 the proportion had declined to 72 per cent (41 years), despite an increase in life expectancy of more than four years. These data have not been updated, but a comparison of age-specific labour force employment rates for males in 1947 and 1996 against a further increase in life expectancy at age 60 of 4.6 years over the same period, indicates substantial further compression. By 1996 only 47 per cent of males had entered the labour force by age 15–19, compared with 80 per cent in 1947. Only 51 per cent were still in the labour force at age 60–64, compared with 80 per cent in 1947.
Accordingly, the relative ability of the population to fully provide for its own education, health care, old age, and/or to care for others, may have a significant cohort-level dimension. Some cohorts may experience more or less difficulty than others, due to underlying demographic forces of which they themselves played only a small part. Failure to understand these constraints may see younger cohorts reduce their fertility still further, as they seek to maximise (or protect) their own material wellbeing.
Analysis of the phenomenon of demographic compression is extremely complex, and as yet relatively undeveloped. In the interim, it is increasingly important that policy makers and analysts think intergenerationally as well as longitudinally.
The Australian welfare state was officially established in 1943. Since its inception it has been based on the notion of the social contract, an implicit agreement between the state and the populace under which the economically active population is taxed, and these taxes are redistributed by the state as income support and services to the eligible dependent population (typically economically-inactive persons meeting specific criteria). Importantly, it is a pay-as-you-go form of welfare state, where all benefits are funded from current taxation. There are no vested funds for individual contributors.
Over the past three decades, a growing number of changes to Australia’s welfare state have been introduced. Uppermost among these is the increasingly strong state encouragement for and requirement of self-provision for post-compulsory education, health care, and superannuation. The emerging situation means that the young and middle-aged are simultaneously required to provide for their own education, health care and superannuation. They also pay taxation to support the currently old and those nearing retirement who do not now have time to self-provide. These changes are heralding a decisive shift from an internally coherent, universal, tax-based, flat-rate system (Castles 1994) to a more mixed or segmented self-funded, multi-tiered system, such as is found in the United States (Heidenheimer et al. 1990). As a result, the situation contains a serious challenge to the social contract, the legitimacy of which depends on equity and continuity of access between generations (Thomson 1992a, b).
Understanding how this tension is developing can be assisted by consideration of the changing demography. When the Australian welfare state came into being, the population was structurally young (approximately 5 per cent aged 65+ years). From the late 1930s it grew even younger, as fertility increased (it had been slowly declining since the 1870s) and gathered momentum with the baby boom. This trend continued for the next two and a half decades, until the peak of the baby boom in 1961. Thereafter, as fertility again fell, the long-term trend towards structural ageing resumed.
These dynamics meant that Australia’s welfare state (like the welfare states of much of the developed world) was therefore created during a period in which a particular age structure was extant—youthful and juvenescent. Ever-increasing numbers of young people were heading towards the labour force (or primary tax-base), and, while youth dependency was high, aged dependency, which typically costs two to four times as much (Borowski & Hugo 1996 p. 49) was low. From such a perspective, neither the manifestation nor implications of excessive structural and numerical ageing could easily have been foreseen. This remained true during the 1970s, when significant changes to welfare provision were enacted (see, for example, the Borrie Report). Although falling, fertility was still relatively high (the TFR averaging 2.4 across the decade), childbearing relatively universal, and there had been little improvement in life expectancy at older ages. At the time it seemed feasible to continue, even strengthen, welfare provision. Now, it is time to reflect that the development of the welfare state may not have depended upon the social contract as much as upon a youthful age structure, a demographic contract (Thomson 1992a, b).
Moreover, it has been postulated that continuity of the pay-as-you-go form of welfare state may actually accelerate structural ageing, via a ‘taxation-fertility’ spiral (Weaver 1986, p. 311). In what is probably a worst case scenario, it is argued that as the demand for Age-Pension and other elder-specific services (for example, expensive health procedures) increases as a result of numerical ageing, governments will have little alternative than to cut benefits and services, or access to these, or dramatically increase taxation levels. If the latter became the chosen option, higher taxation levels would conceivably see women undertake still higher levels of labour force participation than at present, as they sought to maintain current familial living standards. Such a situation would be expected to have a further depressing effect on fertility, and its outcome, a further increase in structural ageing. As structural ageing increased further, taxation would need to be further increased, creating a continuing downward pressure on fertility.
The results of such a scenario would not only be catastrophic for the welfare state; they would also have significant political ramifications. Seemingly, more likely scenarios will include a state-encouraged shift to later retirement (Bishop, in Access Economics 2001),11 and additional but incremental changes in access to benefits along the lines already being implemented, such as the currently occurring changes in the age of eligibility for Age Pension for females. Panels A and B of Figure 14 give an indication of the impact of increasing the age of eligibility for females from 60 to 65 years incrementally over a decade. The data assume no change to current age-specific rates (that is, uptake). The reduction in the component of change due to the changes in age structure (proportions at each age), and the increased numbers of elderly, is clear (see section 13 for standardisation methodology. The fiscal savings could be readily computed from these data).
Figure 14: Projected changes in numbers of females receiving Age Pension under different eligibility criteria
Source: Compiled from Jackson 1999 (Department of Social Security unpublished data and ABS Population Projections 1998, Series II)
Importantly, changes such as these should be clearly related to the context of improved life expectancy. The issue of Age Pension is illustrative. Prior to the establishment of the welfare state in 1943, discrete benefits such as the Age Pension (1909) had been introduced. At the time, life expectancy at birth (55 years for males and 59 years for females) was lower than the age of eligibility (65 years for males and 60 for females). For those who reached the age of eligibility (in the 1970s, when the Borrie Report was received), a further 9.4 years (on average) could be expected for males; a further 16 years for females (see section 3 on life expectancy). Currently, a male reaching age 65 can expect to live on average a further 14.6 years; a female reaching age 60, a further 24 years. There are many indications that this increase will continue. Trends such as these, positive though they are, necessitate what must be understandable changes in eligibility criteria.
The relationship between policy and demographic change in general, and population ageing in particular, is easier to understand if the term policy itself is first paid some analytical attention. Demographers make useful distinctions between ‘explicit’, ‘implicit’, ‘direct’ and ‘indirect’ policy (Lucas 1994). Also in the demographic lexicon are ‘unintended’ and ‘net’ policy effects.
- Explicit policies are those where the objective is formally stated, written down, acted upon by a specific set of bureaucrats, and so on. A classic example would be Australia’s migration policy.
- Implicit policies are those that are not formally stated, written down, necessarily acted upon by a specific set of bureaucrats, and so on. They do, however, typically have intended effects. An example would be the sale of contraceptive devices. This is a policy which implicitly encourages fertility limitation, but which is not made explicit in a country like Australia because of its near-universal acceptance.
- Direct policies are those that are developed with the objective of directly altering the phenomenon or situation in mind. An example would be raising the age of eligibility for Age Pension in order to reduce, at least in the short-term, the cost of Age Pension.
- Indirect policies are those that are developed with the objective of altering the phenomenon or situation in mind via an indirect mechanism. An example would be the payment of child allowance in the hope of raising fertility (or reducing structural ageing).
- Unintended policy effects are those that arise as an unintended consequence of the above. An example would be a further fall in fertility and an increase in structural ageing as a result of the introduction of user-pays fees for education (such as Australia’s Higher Education Contribution Scheme (HECS). The accumulation of large education-related debts could be expected to cause individuals and couples to delay family formation and/or to have less children than they may have otherwise wished (Jackson, forthcoming).
Self-provision for health care and retirement may have similar effects.
- Net policy effects are similar to unintended effects, but are the manifestation of two or more policies that contain conflicting or mutually compensating elements (Johansson 1991). An example would be a reduction in fertility and an increase in structural ageing if there was a reduction in financial support for child care (policy objective: fiscal saving) at the same time as there was an increase in the number of women working to pay off their HECS debt (policy objective: fiscal saving).
When disaggregated in this manner, it can be understood how policies that are developed to respond to, for example, numerical ageing (for example, self-provision for health care and retirement), or even apparently unrelated factors (fiscal savings in education; industrial and labour market policy) may unintentionally exacerbate structural ageing (Chesnais 1996; Esping Anderson 1996; McDonald 1997, 1999, 2000). Similarly, other policies, such as those facilitating the casualisation of the labour force, may inadvertently stimulate fertility, thereby adding to the dependency ratio. For policymakers, who often work in terms of explicit and/or direct policy effects, the following is a memorable quote:
If policy is acknowledged to exist in diverse, and even invisible incentive-like forms (which are not necessarily written down, or enforced by a specific set of bureaucrats, or even related to the consciously articulated thoughts of a governing elite), one can begin to coherently argue that, ‘theoretically’, policy is always efficiently enforced, and is always an active determinant of fertility, indeed the most important one in virtually all cases (Johansson 1991, p. 383).
In short, it is essential to understand that policies that have no demographic objectives often have demographic effects, yet also that it is almost impossible to determine precisely which factor delivered (or did not deliver) which effect. Some observers believe that the impact of indirect political action on fertility (for example) is much stronger than that of policies designed explicitly to affect fertility (Höhn 1986, 1987). This is especially so in respect of efforts to increase fertility (pro-natal policies).
Much literature pertaining to the vexed question of how to bring about an increase in fertility (and/or whether this is desirable) exists, and is beyond the scope and interests of this publication to review in detail. Indeed, before venturing into that sphere it would be necessary to review explanations for low fertility as such, a huge task that this paper is purposely not attempting (for an excellent overview see McDonald 2000. See also Birrell & Rapson 1998 for an implicit explanation related to declining levels of partnering).
However, it can be recorded that the effects of explicit and/or direct pro-natal policies have typically been found to be nil or negligible (Demeny 1986, p. 350; Höhn 1987). The three main exceptions: Germany’s rise in the birth rate of the 1930s as a result of eugenic policies; Romania’s increase following a ban on abortion in 1966; and Singapore’s early 1990s increase as a result of giving tax exemptions for higher numbers of children to the higher
socioeconomic strata. These were all temporary effects only, and are not examples likely to be pursued by Australian policy makers. On the other hand, since the issue is likely to receive much more attention in the near future, a brief review of tried and proposed measures is given below.
- Höhn (1986, 1987) and Hugo (2000) provide an overview of measures attempted in several European countries, many of which have also been implemented in Australia at various times. Pro-natal policies include: child allowances, birth grants and loans, income tax relief and incentives, income splitting, paid and unpaid child-rearing leave with re-employment guarantees, childcare facilities, mutual responsibilities of families and societies (children not seen as a private good only), access to subsidised housing, monthly salaries at the birth of a second or subsequent child, free education, restricted sale of contraceptives (mainly Eastern European countries), and (in Romania only) a taxation on childlessness. Generally, expenditure for the more directly subsidised measures is considerable, and the effects short-lived. Greater success appears to come from the more social measures that reduce role incompatibility (between family and work) and opportunity costs (foregone earnings and seniority, superannuation contributions, risk of re-employment). In other words, policies that alter the environment in which people make decisions about having children are likely to be the most successful.
- In Sweden, for example, not only is paid parental leave institutionalised, but it is mandatory that one month of that leave be taken by the father (Chesnais 1996 p. 733). These measures reduce the immediate opportunity costs of childbearing and rearing, and contribute to gender equity. According to many commentators, empowerment of women ensures against a very low birthrate.
- Reflecting these arguments, McDonald (1997, 2000; see also Chesnais 1996; and Esping- Anderson 1996) argues for more ‘family-friendly’ workplaces. The very low levels of fertility experienced in developed countries today are largely ascribed to an incoherence between the levels of gender equity applying in different social institutions, such as the family and the market place. Where gender equity in these institutions is low, or differs markedly between institutions, fertility is very low (that is, considerably below the TFR required for generational replacement); where it is higher, as judged, for example, in Sweden, fertility is higher (around or closer to replacement level). As McDonald (1997, p. 1) explains, when women have access to the same educational and employment opportunities as men, but these opportunities are severely curtailed by having children, then women will restrict the number of children that they have. Inflexible workplace arrangements that penalise, rather than encourage, those who have children, are particularly correlated with low fertility. It is at this juncture that policy interventions might most usefully be directed.
- Demeny (1986) proposes formal incorporation of the (nuclear) family. Revenues, however acquired (and presumably taxation liabilities), would accrue to the corporation, becoming equally vested in spouses. This would enhance the economic security of women and provide for greater choice in matters pertaining to labour force participation, household production, and child rearing. Problems would be experienced in defining the family unit, while the underlying assumption of equal sharing and reciprocity within the family could
not be taken uncritically.
- Demeny (1986, 1987) also proposes linking old-age economic security with prior fertility behaviour. The aged, in aggregate, have raised the subsequent generation of taxpayers who make the system viable (whether for pensions or investment returns, funds come primarily from the productive efforts of the current generation of workers). Individual demographic contributions to the aggregate should be recognised through differential access to the resources eventually generated. Women who have taken time out of the labour force to
raise the future taxpayers are especially disadvantaged in situations where self-provision for retirement is required. (However, so too are those who have experienced long-term unemployment and who also may not have had children.)
Most discussions concerning population ageing are based on population projections. It is common to see criticisms of these projections. Most typically the criticism will include the term predictions. Population projections are not predictions. They are based on clearly specified assumptions about the three demographic factors that together cause population change: births, deaths and migration. Past and present levels of these factors are used to develop
several sets of assumptions (variants). For example, a combination of higher fertility, lower mortality, and higher net migration than is currently extant usually comprises the high variant assumption. Similarly a combination of lower fertility, higher mortality and lower net migration than is currently extant usually comprises the low variant. The various assumptions used by the ABS are always published along with the projections themselves (see ABS 3222.0).
Projections are calculated using the cohort component method:
P1 = P0 + B - D + NM
Where P1 = the ‘new’ population
P0 = population at the present time
B = Births
D = Deaths
NM = net migration (the difference between in migration and out migration).
The analyst begins with a census-derived base population by sex and single year of age (such as appears graphically in a population pyramid). The birth rate assumption is applied to the number of women at each single reproductive age (15–49 years) and the resulting projected number of births is added to the base of the population age structure. The death rate assumption for each single age and sex group is then applied to the resulting age structure. Finally the migration assumption for each single age and sex group is applied, the resulting numbers being either added to, or subtracted from, the numbers at each age. The population is then ‘aged’ by one year to become the new base population, and the process is repeated for each successive year. The calculations are made separately for each statistical local area, with different fertility, mortality and migration assumptions being used for urban and rural areas. These data are then aggregated to provide total and State/Territory level data.
The resulting projections, which derive both age structures, and total numbers, indicate what the outcome will be if (and only if) the specified assumptions have been met. As such, they provide a useful benchmark against which actual trends can be plotted.
Currently, the ABS produces 24 sets of projections; typically only three are published: Series Ia (the high outcome variant), Series IIa (the medium outcome variant), and Series IIIa (the low outcome variant), sometimes referred to as the ‘best case’, ‘medium case’ and ‘worst case’ scenarios. Conventionally, where data from only one set of projections are presented, they reflect the medium variant. This is especially so with international data comparisons.
Because birth and death rates typically change quite slowly, and international migration into an island nation such as Australia is reasonably well controlled and monitored, projections for the immediate years and decades can be considered highly reliable approximations. However, it is important to note that all measures of migration are somewhat less reliable than births and deaths data, which are derived from Vital Registrations. In particular, internal migration data, which are based on Medicare ‘change of address’ registrations, are subject to many limitations. Because of these shortcomings, longer-range projections (to 2051 or longer), should always be viewed as educated guesses.
The ABS issues a new set of projections every second year. They are of course based on revised sets of assumptions that have taken account of demographic changes during the previous two years.
One other type of population projections deserves a brief mention. These are intercensal projections, which, as their name suggests, are short-term projections undertaken between censuses, which are themselves usually undertaken every five years. A very similar process to that described above is carried out, with the outcomes being revised after the following census.
The factors outlined in this paper have a number of methodological implications for policy makers and analysts. Among these is the need to control for compositional changes in the phenomenon being studied. For example, if the proportion of a population receiving Age Pension increased over time, we would want to know what the proportion would have been if the age structure had not changed. This can be established via a simple technique called direct standardisation. Using a slightly more refined technique called decomposition analysis, we can also show (a) what proportion of that increase was due to an increase in the numbers of elderly, and/or (b) what proportion was due to an increase in uptake (those applying for Age Pension who previously would not have). The former (a) would reflect the effect of population ageing (that is, it would have a demographic explanation) while the latter (b) would reflect a true increase (that is, it would have a social or economic explanation).
In technical terms, the problem is defined in the following way. Any summary measure (for example, the percentage of a population receiving an income support payment) is the product of at least two things. These are: (i) the underlying level or incidence of the phenomenon of interest, and (ii) the composition of the population for which the calculation is being made; that is, the extent to which the population of interest is concentrated in the compositional categories where the phenomenon of interest is likely to occur (for example, age group, sex, marital status group, educational or employment group). If the effects of (ii) are not controlled, any ratio-type measure used to make comparisons either within or between populations, at either a single point in time or over time, is at risk of yielding distorted comparisons (Carmichael 1995).
Standardisation: With simple (direct) standardisation, the age-specific (or category-specific) measures for one population are applied to the age structure (or category structure) of another population (the standard population), and then summed. The algorithm is:
Μs(i) = Σc mi(c).ps(c)
Μs(i) = the summary measure for population i standardised to the composition of population s
c = the compositional categories for the variable(s) being standardised (age, age-sex category etc.)
mi(c) = the specific measure equivalent to M(i) for compositional category c for population I
ps(c)= the proportion of the standard population s in compositional category c
Interpretation of these results proceeds by comparing the summary measure for the standardised population with either its own non-standardised equivalent, or with the measure for the standard population. Interpretation rests on one important axiom—that standardised measures are hypothetical. That is to say, the resulting values are values we would expect the summary measure in question to take on if it had the composition of the standard population.
Importantly, the standard population must match precisely the denominator for the summary measure. That is to say, if the summary measure pertains to the proportion of the population aged 65+ years receiving Age Pension, the standard population must cover the exact same age groups.
Decomposition: Two-way decomposition is a refined form of standardisation that splits the differences between two summary measures into components that are attributable to two phenomena, for example as above, to changes in age structure and changes in uptake. The algorithm for component analysis (Carmichael 1995 p. 51) is:
Csm = 0.5[M(1) – Ms1(2) + Ms2(1) – M(2)]
Cc = 0.5[M(1) - Ms2(1) + Ms1(2) – M(2)]
Csm = Component due to differences in underlying characteristics
Cc = Component due to differences in population composition
M(1) = Summary measure 1, relates to population 1
M(2) = Summary measure 2, relates to population2
Ms1(2) = Summary measure 2 directly standardised to population 1
Ms1(2) = Summary measure 1 directly standardised to population 2
This algorithm standardises the summary measures for each population against the age composition of the other, deriving alternative expressions for Csm and Cc. The two values are then summed and averaged. Interpretation then proceeds in a manner similar to that for direct standardisation, only in the case of component analysis it is the sign (+ or -) on each component that is important, and how this sign compares with that on the overall differences between the two original (unstandardised) summary measures (Carmichael 1995). If the sign on the component is the same as that on the overall difference, that component helped produce the overall difference. If the sign on the component is opposite to that on the overall difference, that component has partially offset, or moderated the overall difference (that is, made it less substantial than it otherwise would have been).
Figure 15 shows the effect of decomposition analysis on the proportion of the Australian male population receiving the Disability Support Pension (DSP) between 1971 and 1997. The substantial growth in numbers receiving this pension has, in the past, been superficially attributed to population ageing. However, as Figure 15 shows, the effect of changes in the age structure have been negligible. For most of the period shown, population ageing (or more correctly, changes in cohort size (see section 5) had an offsetting effect, becoming additive only in 1997, and then only fractionally. This finding is explained by the fact that the first of the baby boomers have only just passed age 50 and entered the key DSP age group. Thus, the growth in the numbers receiving DSP has been real in the sense that it cannot be attributed to population ageing.
Figure 15: Components of change in disability support pension (percentage point change over 1971), males 1971–97
Source: Compiled from Jackson 1999 (Department of Social Security Unpublished Data, and ABS Population Projections (1998)Series II)
Notes: The age structure effect is barely visible, showing just below the line denoting zero growth, over some of the years 1979–1994
Similarly, decomposition analysis of several other Commonwealth income support categories identifies that population ageing has as yet had very little effect on any payment category other than Age Pension, and then only for females (Whiteford & Jackson 1998; Jackson 1999, Figure 3.4, Figure 3.5 and Figure 3.6). By contrast, in a manner almost identical to that for the DSP, population ageing, or more accurately, changes in cohort size, has partially contained the demand for, or growth in, spending on unemployment related allowances.
These findings and their technical underpinnings are very important for policy makers, advisers and analysts working in such areas as income support and services, because if changes in the numbers (or proportions) of the population receiving certain payments and benefits are erroneously attributed to population ageing, the resulting policy interventions may fail. Moreover, failure to specify appropriately the ‘problem’ can also be highly detrimental to
those people who comprise the affected groups. For example, since the early 1980s changes in cohort size have had a small additive effect on the numbers of females receiving the
Supporting Parent/Sole Parent Pension, (SPP) the reason being that the age group with the highest incidence of SPP receipt (30–39 year olds) has also been the largest age group in the population because it contains the peak baby boomers. In fact the age-effect is very small (in the late 1990s accounting for less than 4 per cent of growth in numbers since 1975), but it serves as a useful illustration. Not all growth in SPP numbers is due to an increase in uptake; nor is it due to population ageing. Rather, at least some of it is due to changes in cohort size.
Finally, another factor demanding the use of standardisation and/or decomposition analysis is change in the family and the household. Among other forces, population ageing is a significant driver of the widely reported decline in the couple with children (or two-parent) household and a concomitant increase in couple only (no children) and sole person households. Age- standardisation of such data readily identifies what might be termed a cascading effect. A quantifiable proportion of the decline in the couple with children household is simply due to the shift to later family formation and thus later entry into this household type, while there is a corresponding increase at these ages in the proportions residing in couple only families (Jackson & Pool 1996, pp. 163–64). The trend is further compounded by smaller average family sizes than in the past, which mean that the ‘empty nest’ phase is reached earlier. This results in reduced proportions in couple with children households at the middle to early old ages, and again, an increase in couple only households at these ages. Finally, longer life expectancy is further extending the period spent in the couple only household, while the higher life expectancy of females than males, coupled with numerical ageing, is causing a similar increase in the sole person household at older ages.
The overall effect is a reduction in the proportion of the total population residing in a couple with children household, against an overall increase in the number of households, and a decline in number of persons per household. These trends, which the momentum of ageing contained within the age structure ensures will now accelerate, are often attributed (by the media) to the increase in sole parenting and/or the number of elderly living alone. Certainly the latter are contributing factors, but the changes fall far short of accounting for the decline in the couple-with-children household as such. Age-standardised analyses would contribute substantially to the debate.
1 It is difficult to define precisely the beginning and end of the baby boom (it differs slightly in each developed country), but the Australian Bureau of Statistics recognises the period 1946–65 because in 1965 the TFR had fallen below its 1946 level of 2.98.
5 A youth deficit is defined as occurring when the proportion of the population aged 15–24 years falls below 15 per cent. In 1980, no countries had recorded this phenomenon. By 1985, it was apparent in seven countries, and currently (2001) it can be observed in 54 countries with many others close behind.
7 It is here that the distinction between structural and numerical ageing is again useful. The declining number of births is the cause of structural ageing; the increasing number of deaths, the result of numerical ageing.
10 With the exception of 1986, the Australian census has historically collected ‘ethnic’ data by ‘country of birth’ (for example, New Zealand). These data do not determine the ethnicity of migrants, which is related to cultural affiliation. An approximation of an ethnic group age structure could possibly be achieved by combining these data with data for Australian-born people with parents born in that birthplace, but the result would still not reflect actual ethnic or cultural affiliation.
Abbasi-Shavazi, M. 1998, the fertility patterns of selected Australian immigrant groups, 1977–1991, unpublished Ph.D. thesis, The Australian National University, Canberra.
Abbasi-Shavazi, M. & McDonald, P. 1997, Fertility and multiculturalism: immigrant fertility in Australia 1977–1991, paper presented to the International Population Conference, Beijing.
Access Economics. 2001, Population Ageing and The Economy, Melbourne, Access Economics Pty. Ltd. Alvarado, J. & Creedy, J. 1997, Migration, Population Ageing and Social Expenditure in Australia, Commonwealth of Australia.
Australian Bureau of Statistics, Australian Demographic Trends, Cat. no. 3102.0, various years.
Australian Bureau of Statistics, Births Australia, Cat. no. 3301.0, various years.
Australian Bureau of Statistics, Census Australia, various years.
Australian Bureau of Statistics, Deaths Australia, Cat. no. 3302.0, various years.
Australian Bureau of Statistics, Experimental Projections of the Aboriginal and Torres Strait Islander Populations, Cat. nos. 3230.0 and 3231.0.
Australian Bureau of Statistics, Migration Australia, Cat. no. 3412.0, various years
Australian Bureau of Statistics, Population Projections 1997–2051, Cat. no. 3222.0.
Birrell, B. & V. Rapson. 1998, A Not So Perfect Match. The Growing Male/Female Divide, 1986–1996, Melbourne: Centre for Population and Urban Research, Monash University.
Bishop, B. 1999, The National Strategy for an Ageing Australia, Independence and Self-provision Discussion Paper, Minister for Aged Care, Commonwealth of Australia
Bongaarts, J. 1999, ‘Fertility decline in the developed world: where will it end?’, American Economic Association Papers and Proceedings, May, pp. 256–260.
Borowski, A. & Hugo, G. 1996, ‘Demographic trends and policy implications’, chapter 2 in A. Borowski, S. Encel, and E. Ozanne (eds) Ageing and Social Policy in Australia, Cambridge University Press, Melbourne.
Burnely, I. 1996, Atlas of the Australian People 1996, Australian Government Publishing Service, Canberra.
Caldwell, J., Caldwell, P., & McDonald, P. 1998, Challenges of changing age structures, paper presented to International Symposium on Population and Development Policies in Low Fertility Countries, Kihasa, May 7 to 12.
Carmichael, G. 1995, ‘Comparison: standardisation and decomposition’, chapter 2 in G. Carmichael, Methods of Demography, Course Text, The Australian National University, Canberra.
Castles, F. 1994, ‘The wage earner’s welfare state revisited: refurbishing the established model of Australian social protection, 1983–1993’, Discussion Paper No. 39, Public Policy Program, The Australian National University, Canberra.
Chesnais, J-C. 1996, ‘Fertility, family and social policy in contemporary Western Europe’, Population and
Development Review, 22(4), pp. 729–739.
CIA (Central Intelligence Agency) 1990 ‘The CIA on youth deficits’, Population and Development Review,
16(4), pp. 801–807.
Coale, A. 1972, The Growth and Structure of Human Populations, Princeton University Press, Princeton, New Jersey.
Coale, A. 1972b, ‘How a population ages or grows younger’, chapter 3 in R. Freedman (ed.) Population: The Vital Revolution, Doubleday–Anchor, New York, pp. 47–58.
Cowgill, D. & Holmes, L. 1970, ‘The demography of ageing’ in A. Hoffman (ed.) The Daily Needs and
Interests of Older People, Charles C. Thomas, Springfield.
Davis, K., Bernstam, M.S., Ricardo-Campbell, R., (eds) 1986, ‘Below replacement fertility in industrial Societies. Causes, consequences, policies’, Population and Development Review, Supplement to Vol. 12, Cambridge University Press, Cambridge.
De Beer, J., Beets, G., Bosman, E., & Willems, P. 1991, ‘Chronicle: trends in population and family in the low countries’, in G. Beets, R. Cliquet, G. Dooghe, J. Gierveld, & J. de Jong (eds) Population and Family in the Low Countries, Swets and Zeitlinger, Amsterdam, pp. 22, 133–215.
Demeny, P. 1972, ‘Early fertility decline in Austria-Hungary: a lesson in demographic transition’, in D. Glass & R. Revelle (eds) Population and Social Change, Edward Arnold, London.
Demeny, P. 1986, ‘Pronatalist policies in low-fertility countries: patterns, performance, and prospects’, in K. Davis, M. Bernstam, & R. Ricardo-Campbell (eds) ‘Below Replacement Fertility in Industrial Societies. Causes, Consequences, Policies’, Population and Development Review, Supplement to Vol. 12,
Cambridge University Press, Cambridge.
Demeny, P. 1987, ‘Relinking fertility behaviour and economic security in old age: a pro-natalist reform’,
Population and Development Review, 13(1), 128–132., pp. 22, 128–132.
Easterlin, R. 1988, Birth and Fortune. 2nd Edition., University of Chicago Press, New York.
Esping-Anderson, G. 1996, ‘Welfare states without work: the impasse of labour shedding and familialism in Continental European social policy’, in G. Esping Anderson (ed) Welfare States in Transition: National Adaptations in Global Economies, Sage Publications, London, pp. 66–87.
Gardiner, G. & Bourke, E. 2000, ‘Indigenous populations, ‘mixed’ discourses and identities’, People and Place 8(2), 43–52.
Hagenaars, J. 1990, Categorical Longitudinal Data, Sage Publications, Newbury Park.
Heer, D. 1986, ‘Immigration as a counter to below-replacement fertility in the United States’, in K. Davis, M. Bernstam, and R. Ricardo-Campbell (eds) ‘Below Replacement Fertility in Industrial Societies. Causes, Consequences, Policies’, Population and Development Review, Supplement to Vol. 12, Cambridge University Press, Cambridge pp. 262–269.
Heidenheimer, A.J., Heclo, H., and Adams, C.T. (1990) Comparative Public Policy. The Politics of Social Choice in America, Europe and Japan (Third Edition), New York, St. Martins Press.
Höhn, C. 1986, Demographische Wirkungen Politischen Handelns, Zeitschrift für Bevölkerungswissenschaft 12, 1, pp. 3–51, and 2, pp. 185–219.
Höhn, C. 1987, ‘Population policies in advanced societies: pronatalist and migration strategies’,European Journal of Population, 3, pp. 459–481.
Hugo, G. 1998, ‘South Australia’s ageing population and its increasingly multicultural nature’, paper presented to the annual general meeting of Multicultural Aged Care, Adelaide.
Hugo, G. 1999, ‘Regional development through immigration? The reality behind the rhetoric’, Research Paper 9, Commonwealth of Australia, Parliamentary Library, Canberra.
Hugo, G. 2000, ‘Declining fertility and policy intervention in Europe: Some lessons for Australia? Journal of Population Research, 17(2), 175–197.
Jackson, N. 1994, Youth unemployment and the ‘core family’. population, policy and political economy, unpublished Masters thesis, University of Waikato.
Jackson, N. 1998a, Ethnic stratification in New Zealand. a total social production perspective, unpublished Ph.D. thesis, The Australian National University, Canberra.
Jackson, N. 1998b, ‘Demographic compression and its implications for familial self-reliance’, Paper presented to the Australian Institute of Family Studies Conference, Melbourne, 25–27 November.
Jackson, N. 1999 ‘Understanding population ageing: a background’, Australian Social Policy (1), pp. 203–224.
Jackson, N. (forthcoming) ‘The Higher Education Contribution Scheme. A HECS on ‘the family’?, in G. Carmichael & D. Dharmalingham, (Eds.)Australia and New Zealand at the Millennium, proceedings of the Australian and New Zealand Population Association’s joint conference, Wellington.
Jackson, N. & Pool, I. 1996, ‘Will the real New Zealand family please stand up? Substantive and methodological factors affecting research and policy on families and households’, Social Policy Journal of New Zealand, 6, pp. 148–176.
Johansson, S. 1991, ‘‘Implicit’ policy and fertility during development’, Population and Development Review, 17(3), pp. 377–414.
Keyfitz, N. 1971, ‘On the momentum of population growth’, Demography, 8, pp. 71–80.
Kippen, R. 1999, ‘A note on ageing, immigration and the birthrate’, People and Place, 7(2), pp. 18–22. Kippen, R. & McDonald, P. 2000, ‘Australia’s population in 2000: the way we were and the ways we might have been’, People and Place, 8(3), pp. 10–17.
Le Bras, H. 1991, ‘Demographic impact of post-war migration in selected OECD countries’, in Migration. The Demographic Aspects, OECD, Paris, pp. 15–26.
Lucas, D. 1994, ‘Population policies’, chapter 13 in D. Lucas and P. Meyer (eds) Beginning Population Studies, 2nd Edition, The Australian National University, Canberra.
Lutz, W. (ed.) 1994, The FuturePopulation of the World. What Can We Assume Today? London, Earthscan.
Lutz, W. (ed) 1996, The FuturePopulation of the World. What Can We Assume Today?, rev ed, London, Earthscan. (www.iiasa.ac.at/Research/POP) (30–05–01)
Mackay, H. 1997, Generations: Baby Boomers, Their Parents and Their Children, Macmillan, Sydney.
McDonald, P. 1997, ‘Gender equity, social institutions and the future of fertility’, Research School of Social Sciences Working Papers in Demography No 69, The Australian National University, Canberra.
McDonald, P. 1998, ‘Contemporary fertility patterns in Australia: First data from the 1996 Census’, People and Place, 6(1), pp. 1–17.
McDonald, P. 2000, ‘Low fertility in Australia: evidence, causes and policy responses’, People and Place, 8(2), pp. 6–20.
McDonald, P. & Kippen, R. 1999, The impact of immigration on the ageing of Australia’spopulation, Department of Immigration and Multicultural Affairs, Canberra.
McDonald, P. & Kippen, R. 2000, Strategies for labour supply in sixteen developed countries, 2000–2050, paper presented to the 2000 Annual Meeting of the Population Association of America, Los Angeles, California, 23–25 March.
McNicoll, G. 2000, ‘Reflections on ‘replacement migration’’, People and Place, 8(4), pp.1–13.
McPherson, M. 1992, Cohort vulnerability to lack of support in old age: a theoretical and analytical exploration, unpublished Master of Social Science Thesis, Hamilton, University of Waikato.
Merlo, R. and Rowland, D. 2000, ‘The prevalence of childlessness in Australia’, People and Place, 8(2), pp. 21–32.
Petersen, G. 2000, ‘Gray dawn. The global aging crisis’, in F. Moulder (ed.), Social Problems of the Modern World, Wadsworth/Thomson, Australia, pp. 126–134.
Pool, D. 1987, Implications of changes in the cohort/age structure of the New Zealand population, paper presented to the New Zealand Demographic Society, Auckland, 10.July.
Pool, I. 1991, Te Iwi Maori. A New Zealand Population. Past, Present and Projected. Auckland University Press, Auckland.
Ruzicka, L. 1986, The Length of Working Life of Australian Males, 1933–1981, Monograph Series No.15, Bureau of Labour Market Research, Australian Government Publishing Service, Canberra.
Sceats, J. 1992, Competing generations: the New Zealand family meets the free market, Keynote Address, National Conference, Parent–Centres of New Zealand, Rotorua, New Zealand.
Shryock, H. and Siegel, J. 1971, The Methods and Materials of Demography, U.S. Bureau of the Census, Washington D.C.
Simon, J. 1984, ‘Immigrants, taxes and welfare in the United States’, Population and Development Review, 10(1) pp. 55–69.
Stockwell, E.G. 1976, The Methods and materials of Demography, San Diego-London, Academic Press. Thomson, D. 1992a, Selfish Generations. The Ageing of New Zealand’sWelfare State, Bridget Williams Books, Wellington.
Thomson, D. 1992b, ‘Welfare states and the problem of the common’, CIS Occasional Papers 43, Social Welfare Research Program, New South Wales, Centre for Independent Studies, Sydney.
Thompson, D. 1998, A World Without Welfare. New Zealand’sColonial Experiment, Auckland University Press, Bridget Williams Books, Auckland.
United Nations 2000, ‘Replacement Migration, Population Division, Department of Economic and Social Affairs, United Nations Secretariat.
United States Bureau of the Census 1999, ‘Report WP/98’, World Population Profile: 1998, U.S. Government Printing Office, Washington, DC.
United States Bureau of the Census, International Database, www.census.gov/ipc/www/idbpyr.html. (20-05-01).
Van de Kaa, D. 1998, ‘Postmodern fertility preferences: from changing value orientation to new behaviour’, Research School of Social Sciences Working Paper in Demography No. 74, The Australian National University, Canberra.
Ware, H. 1981, A Profile of the Italian Community in Australia, Melbourne, Australian Institute of Multicultural Affairs.
Weaver, C. (1986) Social security in ageing societies, in K. Davis, M. S. Bernstam and R. Ricardo-Campbell (eds.) Below Replacement Fertility in Industrial Societies. Causes, Consequences, Policies, Population and Development Review Supplement to Vol. 12, Cambridge, Cambridge University Press, pp. 273–94.
Weeks, J.R. 1999, Population. An Introduction to Concepts and Issues, 7th Edition, Wadsworth Publishing, Belmont, CA.
Whiteford, P. and Jackson, N.O. 1998, Demographic influences on social security spending, paper presented to the 10th Biennial Conference of the Australian Population Association, University of Queensland, Wadsworth Publishing, Brisbane.
Wilson, C. (ed.) 1985, The Dictionary of Demography, Basil Blackwell Ltd, Oxford.
Wilson, C. (2001) Implications of global demographic convergence for fertility theory, paper presented to the International Union for the Scientific Study of Population Working Group on Low Fertility, International Perspectives on Low Fertility: Trends, Theories and Policies, 21–23 March.
Young, C. 1989 ‘Australia’s population. A long-term view’, Current Affairs Bulletin, 65(12), pp. 4–11.
Young, C. 1990, ‘The Impact of Demographic Change on Australia’sLabour Force with Reference to the Special Role of Women’, Working Papers in Demography, No.19, The Australian National University, Canberra.
Australian Bureau of Statistics’—contains downloadable demographic and socioeconomic data, including information on concepts, projections and so on.
Statistics New Zealand —contains downloadable demographic and socioeconomic data for New Zealand. A significant feature of the New Zealand data is the attention paid to ethnic differentials.
Australian Department of Health and Aged Care—with a link to the National Strategy for an ageing
Australian Institute of Family Studies—with links to databases and publications.
Australian National University demography program—has links to hundreds of leading information facilities of value and/or significance to researchers in the field of demography.
Australian Department of Immigration and Multicultural Affairs.
Australian Population Association—includes a Population Facts booklet and downloadable, related information.
United States. Bureau of the Census International Data Base— includes a computerised data bank containing statistical tables of demographic and socioeconomic data for 227 countries and regions; can generate tables and pyramids for the present and future; and has a dynamic option via which projected changes in these population age structures can be observed for the next fifty years.
United States. Bureau of the Census—contains demographic information and data on population ageing in the United States.
United States. Social Security System—contains reports for 1999 of special interest in relation to population ageing.
Demographic research on population ageing by the United Nations—includes links to many other age-related web sites.
United Nations Economic Commission for Europe—contains details of work by the United Nations on population ageing.
United Nations—contains a number of other United Nations’ resources.
United Nations—includes references on fertility data for a large number of countries and regions.
National Council on Science and the Environment— access to an wide range of country briefing books. Abstracts allow quick identification of the resources that are most useful.
Social Science—information gateway to a huge range of sites, including for the 1970 British cohort study.
Economic consequences of population ageing.
Global trends in population, health, economic factors etc.
Easy-to-use, two-step database enabling comparison of statistical information for United Nations’countries.
Set of links to selected, evaluated and annotated Internet resources relevant to demography.
Contains approximately 24 000 time-series for 196 countries and geographical areas covering population, exchange rates, fund accounts, international liquidity, international banking, money and banking, interest rates, prices, wages, production and employment, international transactions, government finance, national accounts. Some sites require authorisation from the Data Archive before access is permitted.