Australian Government, Australian Government Actuary

1. Mortality of the Australian population

1.1 Results for 2010-12

Figure 1 shows the mortality rates reported in the 2010-12 Life Tables on a logarithmic scale.

Figure 1: Mortality rates 2010-12

The pattern of mortality observed in Figure 1 is typical of developed countries. Mortality rates during the first year of life are relatively high for both males and females, primarily due to congenital abnormalities and perinatal conditions. After the first year of life, an increasing capacity to ward off disease and limited exposure to life threatening situations results in rapidly dropping mortality rates. Around age 10, mortality rates reach a minimum. At this point, the probability of dying within the year is less than 1 in 10,000.

Accidents are the single largest cause of death in childhood. With the developing autonomy of the teenage years, mortality attributable to accidental or self-inflicted causes increases steeply, particularly for males. This growth slows and is briefly, and almost imperceptibly, reversed in the early twenties before rates start to rise again as the falling mortality from accidents is more than offset by increasing rates of death due to disease. The shape of the curves around ages 18 to 21 has not changed much since the 1990-92 Tables, when the previous ‘accident hump’ flattened for the first time in several decades.

The shapes of the mortality curves for males and females are similar, but the absolute rates are different with female mortality being less than male mortality at all but the oldest ages. This difference is consistent with a number of factors, including:

  • • a greater level of risk‑taking behaviour by young males;
  • • the greater hazards associated with some occupations which have traditionally been dominated by men (such as mining and construction);
  • • the differences in the incidence of some diseases between men and women; and
  • • the differences in fatality from diseases which affect both genders.

The first two of these factors relate to behavioural differences, including gender stratification in the labour force, rather than physiological differences between men and women. Physiological differences may, however, in part explain the behavioural divergence. The latter two factors could be expected to be the result of both physiological and lifestyle differences.

1.2 Changes since 2005-07

Figure 2 charts the mortality rates from the current Tables together with those reported five years earlier. It shows that mortality rates have fallen at virtually all ages. The exception is at the very old ages where mortality rates have increased since 2005-07.

Figure 2: Mortality rates 2005-07 and 2010-12

Infant mortality continued to fall, as it has in every set of published Tables. Since the previous Tables, infant mortality has fallen by around 5 per cent per annum.

Mortality in the childhood years has also improved, more so for males than for females. However, the number of deaths observed at these ages is very small. This increases the risk that the shape of the smoothed mortality curve will be unduly affected by random variation in the number of deaths reported. We modified the graduation process for the 2005-07 Tables to address this issue and have maintained the approach, which is described in section 2.2, for the current Tables. This is, however, still the age range with the greatest volatility and limited significance should be attached to the changes in mortality at individual ages.

Male mortality has fallen over the teenage and young adult ages, most notably for those in their mid-twenties. Three decades ago, there was a clear peak in male mortality around age twenty, with mortality rates roughly comparable to those applying to males twenty years older. This phenomenon was known as the accident hump. While rates still increase substantially over the teenage years, the more rapid improvement in male mortality for those in their early twenties means there is no longer a distinct peak. This is, however, the age group with the greatest disparity between male and female rates and, as illustrated in Figure 4 overleaf, the gap remains significant.

There was also noticeable improvement in mortality rates over the age range from around the mid-fifties to the early-eighties.

Figure 3 shows the average percentage improvement in mortality rates over the five years following the 2005-07 Tables by gender for five year age bands.

Figure 3: Percentage improvement in mortality
since 2005-07 by gender

As noted above, mortality at the oldest ages has deteriorated slightly since 2005-07. For centenarians, part of the explanation is the additional effort which the Australian Bureau of Statistics (ABS) devoted to accurate reporting of age on the Census. This appears to have reduced the incidence of mis-statements and led to a reduction in the population counts at very old ages with a flow-on increase in mortality rates. In addition, our graduation choices were influenced by some analysis using the extinct generations methodology of estimating mortality rates at the oldest ages. However, as Figure 3 shows, the worsening mortality is apparent for those in their nineties, where there is a larger volume of reliable data. This feature has also been observed recently in other developed countries. Since the 1970s, mortality rates have improved noticeably among those aged in their sixties and above. This has led to an increasing proportion of the population surviving to the very old ages which may, in turn, have led to a decline in the average health status of this group. It will be interesting to see whether or not this trend towards higher mortality at the oldest ages persists.

Figure 4 compares the gender differential in mortality rates for the current and previous Tables. It shows that, at most ages, there has been little change, with male and female mortality rates improving in tandem. The main exception is the early adult years where the higher rates of improvement in male mortality have led to a noticeable narrowing of the gap relative to the 2005-07 Tables.

Figure 4: Ratio of male to female mortality rates — Ages 5 to 100,
2005-07 and 2010-12

1.3 Past improvements in mortality

The first official Life Tables for Australia were prepared based on data from the period 1881‑90 and there is now a history of 125 years of mortality data. Figure 5 plots the change in mortality rates over time expressed as a percentage of the rates reported in 1881‑90. The data for the four ages shown clearly illustrates the diversity of experience for different ages and genders.

Figure 5: Improvements in mortality at selected ages

Age 0

Age 20

Age 65

Age 85

Infant mortality has shown a sustained and substantial improvement over the entire period, with the improvement for males and females moving closely in parallel. The rates for both males and females are now around 3 per cent of their level in 1881‑90 and still do not appear to have reached an underlying minimum rate. While the rate of improvement had slowed somewhat over the decade between the 1995-97 and 2005-07 Tables, the current Tables have shown a return to a similar rate of improvement to that experienced in the 25 years prior to the 1995-97 Tables.

The picture at age 20 is quite different, with male rates initially improving more quickly than female rates but then deteriorating from about 1945 to 1970 before starting to decline again as the accident hump emerged, subsided and then disappeared. For females at this age, the biggest improvements occurred from the 1930s to the 1950s and probably reflected improved maternal mortality experience as medical procedures were reformed and became accessible to more of the population. Mortality rates for 20 year old females are about 5 per cent of the corresponding rates from 125 years ago. For males of the same age, the relativity is around 9 per cent.

At age 65, the rate of improvement was relatively slow for both males and females until around 1965. This is consistent with the benefits of medical advances up to that time primarily accruing to the young. Since the mid-1960s, however, mortality rates for 65 year olds have more than halved. Male rates for 65 year olds in the 2010-12 Tables are about a quarter of the corresponding rates from the original Tables, while for females the 2010-12 rates are less than 20 per cent of the corresponding rates.

The final chart shows the improvement in rates at age 85. Again, mortality rates at this age showed minimal improvement until the mid-1960s. Since then, there has been a steady improvement in mortality leading to mortality rates for males that are roughly half what they were 125 years ago. For females, the rates are now a third of what they were.

It is also interesting to see how the shape of the mortality curve has changed over time. Figure 6 shows the reported mortality rates for those aged between 10 and 40 as a three dimensional surface, plotted against both age and time. Note that the rates for years in between the Tables have been derived by linear interpolation. The enormous improvements in childhood mortality and the emergence and decline of the accident humps for males and, to a lesser extent, females are clearly visible.

Figure 6: Smoothed mortality rates from 1881‑90 to the present
— Ages 10 to 40

Males

Females

1.4 Longevity

One natural corollary of improving mortality is increasing longevity. Increased longevity has significant implications for both individuals trying to estimate the resources needed for retirement and governments dealing with rising pension and health and aged care obligations.

There are a number of measures of longevity. The most commonly used is life expectancy, which measures the average number of years that would be lived by a representative group of individuals of the same age if they experienced mortality at given rates.

Figure 7 shows how the improvements in mortality described in the previous section have translated into longer life expectancies as reported in the relevant Life Tables (Appendix A sets out the figures on which Figure 7 is based). Note that these reported life expectancies are known as period life expectancies and do not make allowance for any future improvements in mortality which might be experienced over a person’s lifetime. In other words, they are based on the assumption that the mortality rates reported in a particular set of Tables would continue unchanged into the future and, as such, represent a summary of mortality at a particular point in time rather than a projection of mortality over future periods. The impact of continuing mortality improvement is discussed in the next section.

Figure 7: Total life expectancy at selected ages

Period life expectancy at birth has shown dramatic improvement, increasing by over 30 years for both males and females since the inception of the Life Tables. Even at older ages, the substantial improvements in mortality rates for this group over the past forty years have flowed through into significantly increased life expectancies. For example, life expectancy at age 65 has increased by around eight years (more than 70 per cent) for males and ten years (more than 80 per cent) for females.

Figure 8 plots the gap between reported male and female life expectancies at birth and age 65. It shows that over the first third of the twentieth century, male and female life expectancies moved roughly in parallel, with the gap at birth steady at around four years, and the gap at age 65 only around one and a half years. From about 1930, the gap widened for both ages, reaching a maximum in the 1980‑82 Tables. Since then, the differential has been declining for both ages. At birth, the gap has declined by almost three years, falling to levels last seen around 70 years ago.

Figure 8: Gender differentials in life expectancy
at selected ages

Life expectancy at birth is a commonly used measure to describe population mortality. However, as a single summary statistic, it cannot provide information on the diversity of outcomes. For example, under the mortality rates reported in the current Tables, around 60 per cent of both males and females would be expected to survive beyond the reported life expectancy. This result is separate from the issue of mortality improvements that might occur over an individual’s life, which is discussed in the following section.

Figure 9 shows how the distribution of lifespan has changed over the past 125 years. The distributions shown here are based on the prevailing mortality rates and do not make allowance for future mortality improvement. The chart shows the period life expectancy, the median of the lifespan distribution and the interquartile and interdecile ranges.

Figure 9: Distribution of lifespan at birth

Males

Females

It can be seen that the reported life expectancy and median age of death have moved roughly in parallel. However, over the 125 years, the gap between the two measures has declined by around four years, reflecting the dramatic improvements in infant mortality that have had a greater impact on life expectancy than on median age at death.

While improving mortality at younger ages has tended to concentrate the age of death within a narrower range, outcomes for individuals can still vary widely. The interdecile range, for example, spans a range of over 30 years for males, from 62 to 94 and only slightly less for females, from 69 to 96.

Figure 10 reproduces the distribution of lifespan charts based on the expected outcomes at age 65 rather than birth.

Figure 10: Distribution of lifespan at age 65

Males

Females

A number of differences are apparent. Firstly, the lifespan distribution is relatively symmetrical at this age and as a result the mean and median are more closely aligned. Secondly, while the interdecile range is significantly less at age 65 than at birth, it has increased slightly over time rather than narrowing. In other words, outcomes at retirement are no more predictable today than they were 100 years ago.

1.5 Allowing for future improvements in mortality

The figures reported in section 1.4 are all based on the cross-sectional mortality rates from a single set of Life Tables. However, section 1.3 highlighted the substantial changes in mortality that could be expected to occur over an individual’s life time. By way of illustration, the life expectancy of a boy born in 1886, as reported in the 1881‑90 Tables, was 47.2 years, based on the rates in those Tables persisting throughout his life. However, his actual life expectancy would have been some six years higher. This estimate can be obtained by applying the rates reported in subsequent Tables that would be appropriate given his age and the year.

As a result, any realistic measure of longevity needs to consider the possible improvements in mortality that may occur in future. This section focuses on life expectancy in considering the impacts of future mortality improvement. However, the limitations outlined in the previous section of any summary measure such as life expectancy which obscure the diversity of outcomes should be borne in mind.

The issues associated with attempting to estimate more realistic life expectances by allowing for future mortality improvements were discussed in some detail in the 1995‑97 Tables. Those Tables included improvement factors derived from the ratio of the mortality rates in the Tables to those reported in the Tables from 25 and 100 years previously. The current Tables continue the practice of reporting two sets of factors, one based on experience over the last 25 years and the other using the full history of reported mortality. However, the methodology used to determine the factors has been modified slightly. Details on the methodology now being used are provided in Section 2.4.

Figure 11 presents the historical rates of improvement expressed as an annual percentage change in the probability of death at a given age. Note that the lower the value, the higher the improvement in mortality has been. It can be seen that the improvements over the 125 year period have generally been less than the improvements over the past 25 years. The main exception is for the ages around 30 where the rates of improvement over the past 25 years have been less than over the preceding 100 years, particularly for women.

Note that the 25 year improvement factors for the oldest ages have been constrained to be zero. As can be seen from the chart, mortality rates have actually increased since 2005-07 for those at the oldest ages, from age 91 for males and age 92 for females. Over the 25 year period, the fitted mortality improvement factors also showed deteriorating mortality, from age 96 for males and age 98 for females. While there are reasons for assuming that this is a genuine feature related to the improving mortality for those in their seventies and eighties, I have decided to set the factors to zero until more data becomes available.

Figure 11: Historical mortality improvement factors derived from
the Australian Life Tables

Males

Females

There are two ways of taking account of mortality improvement in projecting future life expectances. The first is to apply the same number of years of improvement to the mortality rates at all ages, effectively estimating what future Life Tables might report as life expectancy. This measure, which is known as the period or cross‑sectional life expectancy, makes no allowance for improvements over an individual’s future lifetime and was discussed in the previous section. So, for example, in calculating a period life expectancy for the year 2020 based on the 2010-12 Tables, nine years of improvement would be allowed for at all ages. The following tables show the projected period life expectancies at ages 0, 30 and 65 using the 25 and 125 year improvement factors.

Projected period life expectancies at selected ages under two improvement scenarios
Males
  Age 0 Age 30 Age 65
  25 year 125 year 25 year 125 year 25 year 125 year
2011 80.1 80.1 81.0 81.0 84.2 84.2
2020 82.2 81.1 82.9 81.9 85.6 84.7
2030 84.3 82.1 84.8 82.8 87.1 85.3
2040 86.0 83.1 86.4 83.6 88.3 85.8
2050 87.6 83.9 87.9 84.4 89.4 86.3
2060 88.9 84.8 89.1 85.1 90.4 86.8
Projected period life expectancies at selected ages under two improvement scenarios
Females
  Age 0 Age 30 Age 65
  25 year 125 year 25 year 125 year 25 year 125 year
2011 84.3 84.3 85.0 85.0 87.0 87.0
2020 85.8 85.3 86.3 85.8 88.1 87.6
2030 87.3 86.3 87.6 86.7 89.2 88.3
2040 88.6 87.2 88.8 87.5 90.1 88.8
2050 89.7 88.0 89.9 88.2 91.0 89.4
2060 90.6 88.7 90.8 88.9 91.7 90.0

The 2005-07 Tables projected a period life expectancy at birth for a boy born in 2011 of 80.3 years under the 25 year improvement scenario and 79.7 years under the 100 year improvement scenario. The current Tables estimate a life expectancy roughly midway between these two figures. For a girl born in 2011, the equivalent figures from the 2005-07 Tables were 84.6 and 84.3 years. The current estimate is thus consistent with the 100 year improvement factors reported in 2005-07.

For the future years, the projected period life expectancies are a little lower than those projected in the previous Tables, most notably under the 25 year improvement scenario. This reflects that the rate of improvement in mortality over the five years to 2011 was, on average, slightly lower than over the five years to 1986.

Figure 12 shows how the period life expectancy at birth would change over time under these two improvement scenarios.

Figure 12: Actual and projected period life expectancy at birth —
1971 to 2061

The second measure of life expectancy is what is termed cohort life expectancy. This measure takes into account the improvements that could be experienced over the future lifetime of the individual. So, for example, in calculating the cohort life expectancy of a child born in 2020 based on the 2010-12 tables, nine years of mortality improvement will be applied to the mortality rate at age 0, ten years at age one and so on. In the example provided at the beginning of this section, the life expectancy for a child born in 1886 calculated using the mortality rates as they changed over his lifetime is a cohort life expectancy. Cohort life expectancies can be thought of as being a more realistic representation of the unfolding mortality experience of the Australian population, though the uncertainties around future rates of mortality improvement need to be kept in mind.

The following tables show the cohort life expectancies at ages 0, 30 and 65 using the 25 and 125 year improvement factors.

Projected cohort expectation of life at selected ages under two improvement scenarios
Males
  Age 0 Age 30 Age 65
  25 year 125 year 25 year 125 year 25 year 125 year
2011 90.5 85.6 88.1 84.2 86.0 84.9
2020 91.4 86.4 89.2 84.9 87.4 85.4
2030 92.2 87.2 90.3 85.7 88.7 85.9
2040 92.9 87.9 91.2 86.4 89.8 86.5
2050 93.5 88.5 92.0 87.1 90.7 87.0
2060 93.9 89.1 92.7 87.7 91.6 87.5
Projected cohort expectation of life at selected ages under two improvement scenarios
Females
  Age 0 Age 30 Age 65
  25 year 125 year 25 year 125 year 25 year 125 year
2011 92.2 90.0 90.3 88.4 88.6 87.9
2020 92.9 90.6 91.1 89.1 89.5 88.5
2030 93.5 91.3 92.0 89.8 90.5 89.1
2040 94.1 91.9 92.7 90.5 91.4 89.7
2050 94.5 92.5 93.3 91.1 92.1 90.3
2060 95.0 93.0 93.8 91.6 92.7 90.8

A comparison with the cohort life expectancies reported in the 2005-07 Tables shows that the reduced mortality improvement factors have led to a decline in life expectancies under almost every scenario, with the greatest differences of two to three years reported under the 25 year improvement factors for life expectancy at birth. This is the result of the small changes in the mortality improvement factors being magnified when mortality is being projected many years into the future. The one exception is for males aged 65 under the 125 year improvement scenario. The improvement factors themselves are marginally less than those reported in 2005-07. This is in part due to a change in the period from 100 years (used in the 2005-07 Tables) to 125 years used here. Despite this, since the rate of improvement in male mortality over the five years to 2010-12 was faster than the 125 year average, there has been a very small increase in the cohort life expectancy in the short term. Further into the future, the effect of the lower improvement factors more than offsets the lighter initial mortality rates.

Figure 13 shows the cohort life expectancies for those currently alive in the Australian population. It highlights the considerable gap between the period life expectancies reported in these Tables and the outcomes that would arise if the rates of mortality improvement observed in the past are maintained in the future. The additional life expectancy (the gap between the ‘no improvement’ curve and the other two curves) reduces with increasing age, reflecting the shorter period for improvements to have an impact. At very old ages, the gap has disappeared but the curve rises, reflecting the fact that these people have already reached an advanced age.

Figure 13: Cohort life expectancies by current age

Males

Females

The period and cohort life expectancies set out above illustrate what would occur if mortality continued to improve at the rates observed in the past. Measured mortality improvement can change appreciably between successive Tables, particularly for the factors derived from the most recent 25 years of experience where the earliest period is removed from the calculation and the experience from the most recent five years incorporated. So, for example, at age 45 the 25 year improvement factor has reduced from 2.35 per cent per annum to 1.72 per cent per annum, reflecting the fact that mortality at this age dropped by almost a quarter between the 1980-82 and the 1985-87 Tables but by only around 5 per cent between 2005-07 and 2010-12 Tables.

Furthermore, the effects of these movements are magnified because the projections assume that mortality improvement will be constant for a particular age. This is not a major issue in the short term. One year into the future, for example, the difference in mortality rates at age 45 under the two assumptions is less than 1 per cent. However, in considering cohort life expectancy at birth, the projected mortality rate to be used at age 45 will include 45 years of mortality improvement and the mortality rate under the 2010-12 assumption is a third higher than it would have been under the 2005-07 assumption.

The sensitivity to changes in mortality improvement is also evident in the projected distribution of deaths, as illustrated in Figure 14. The two improvement scenarios presented are both based on 25 year improvement factors.

Figure 14: Distribution of deaths at age 65 allowing for cohort mortality improvement

Males

Females

This chart also suggests that the range of lifespans under credible mortality improvement scenarios is at least as wide as the range where no allowance for mortality improvement was made. In other words, making an allowance for future improvements in mortality does not decrease the challenges individuals face in dealing with longevity risk in retirement.

History demonstrates that mortality improvement is not constant at a particular age and, indeed, can vary within a quite considerable range. The choice of the period over which mortality is measured will also affect the estimates of mortality improvement. Thus, the estimates of cohort mortality included here must be accepted as projections of outcomes under assumptions which have a certain historical basis. They should be regarded as indicative rather than firm forecasts of life expectancy.