Title: Mortality improvements and Life Expectancy
1Mortality improvements and Life Expectancy
- OPA 25 October 2007
- Adrian Gallop
- ONS
2National Population Projections Mortality
- Historical data
- Regional comparisons
- Mortality projection methodology
- Assumptions
- Results from 2006-based population projections
- CMI
3Period expectation of life at birth, EW
4Period expectation of life at age 65, EW
5Proportion of persons surviving to successive
ages, historical or projected mortality, selected
years, EW
6Average annualised rates of improvement in
standardised EW mortality rates
Source calculations using English Life Tables
and 2003-05 ILT, standardised using 2001 pop ests
7Effective annual rates of improvement in
mortality, males, EW
Source calculations using English Life Tables
and 2003-05 ILT
8Effective annual rates of improvement in
mortality, females, EW
Source calculations using English Life Tables
and 2003-05 ILT
9The cohort effect
- Faster improvements have been observed for the UK
generation born 1925-1945 centred on 1931 - This feature has been explicitly allowed for in
GAD mortality projections since the early 1990s - The CMI (Continuous Mortality Investigation) have
described a similar effect in insurance and
pensioner data centred on 1926
10Annual improvement in smoothed mortality rates,
Males, UK 1961/2 to 2004/5
11Annual improvement in smoothed mortality rates,
Females, UK 1961/2 to 2004/5
12Possible causes of the UK cohort effect
- Patterns of cigarette consumption
- World War II
- Birth rates
- Diet
- Welfare State
13Potential drivers for future mortality change
- Reduced levels of deprivation, better housing etc
() - Govt support for increasing wealth, health and
incomes () - Public support for spending on medical advances
() - Decline in smoking prevalence ()
- Lifestyles ( and -)
- Obesity (-)
- Emergence of new infectious diseases (eg HIV,
SARS) (-) - Re-emergence of old diseases (eg TB) (-)
- Increased uncertainty at younger ages (- and )
- Differentials by social class
- Cohort effects
- Wide spread of opinion as to whether future
technical, medical and environmental changes will
have greater or lesser impact than in the past
14Male mortality by major cause, England Wales,
1911-2005
15Female mortality by major cause, England Wales,
1911-2005
16Medical advances - The pace of scientific
development
-
- Pace of scientific development accelerating
- Large element of current improvements driven by
advances in medicine - Improvements in heart disease mortality
partially caused by - - new medication, e.g. beta-blockers statins
- - new surgical interventions, e.g. angioplasty
- Improvements in cancer mortality partially
caused by - - advances in treatment
- - improvements in detection
17Medical advances v risk factors
- Reduction in EW CHD deaths recent study
suggests - 60 due to risk factors,
- 40 due to treatments
- Some risk factors produced small increases
(increases in diabetes and obesity, less physical
activity) - Main risk factor decrease due to smoking,
followed by chloresterol and lower blood pressure
levels
18Smoking Assured lives - Differences in period ex
19Infectious diseases - a growing threat?
-
- Rapid global transport, especially air travel
(e.g. SARS) - Medical advances eg antibacterials,
xenotransplantation - Human behaviour
- Potential threat of bioterrorism
- So far, medical advances and international
networking limited effects of new diseases (eg
SARS) - HIV only infectious agent to emerge in recent
decades to have a dramatic impact on mortality - Cannot disregard potential threat of infectious
diseases - Relative impact of deaths from infectious
diseases may become more significant
20Variations in life expectancy
- Between local authorities and smaller areas
- Social class component of variation
- Between countries in the UK
- Between regions
- Social class trends in life expectancy
21Life expectancy of males in local authorities in
UK, 2003-5
22Life expectancy of males in Camden wards
23Trends in male period life expectancy at birth by
social class, 1972-2001
24Trends in male period life expectancy at age 65
by social class, 1972-2001
25Trends in male period life expectancy at age 65
Social classes I and V, 1972-2001
26Social class differences in male period life
expectancy at age 65, 1972-2001
27Mortality rates by social class
Source ONS Longitudinal Study
28Mortality rates - social class I
Source ONS Longitudinal Study
29Mortality rates - social class II
Source ONS Longitudinal Study
30EolB Males selected countries
31EolB Females selected countries
32Female Male EolB selected countries
33Period expectations of life at birth in 2005
34Broad classification of methodologies
- Process-based
- - model mortality rates from a bio-medical
perspective - - need to understand processes
- Explanatory-based
- - causal forecasting, e.g. econometric
techniques - - need to understand causal links, model
underlying factors and project over long term - Extrapolative
- - project historical trends into the future
35Extrapolative methodologies
- Parametric methods - e.g. fitting parameterised
curves to past data and projecting these forward - Targeting approach - interpolating between
current mortality rates and target rates assumed
to hold at a given future date - Trend projection - either deterministic or
stochastic
36Projections by cause of death
- Advantages
- provides insights into way mortality is changing
- appears to offer greater accuracy in forecasts
- of interest to those researching specific causes
- Disadvantages
- deaths from specific causes not always
independent - difficult to determine actual cause of death
- changes in classification of causes and diagnosis
practice - resulting aggregate rates may be implausible
- usually have catch all category where trends
difficult to project
37UK Population Mortality Projections
- Estimate current rates of mortality improvement
by age and gender - Set target rates of mortality improvement for
some future year (the target year) - Make assumptions on method and speed of
convergence of current improvement rates to
target rates and how improvement rates change
after target year
38UK Population Mortality Projections
- Target year is 25th year of projection
- (ie 2031 for 2006-based projections)
- Target improvements in 2031 assumed to be 1 pa
for most ages for both males and females - For those born between 1923 and 1940 target
improvement rises from 1 pa to 2.5 pa for those
born in 1931 then declines back to 1 - Applies to UK and constituent countries
39Evidence for target setting
- Historical data
- Expert opinion
- Trends in cause of death, and changes in medical
practices - Results of pure extrapolatory models
40Choice of target rate
- Rates of improvement at older ages most important
- Standardised average rate of improvement over
20th century 1.0 pa - Cohorts exhibiting greatest improvement will be
aged 85-105 in 2031 so likely to contribute less
to overall rate of improvement - Debate as to whether future technical, medical
and environmental changes will have greater or
lesser impact than in the past
41UK Population Mortality Projections
- Current improvements in mortality rates differ by
age and sex extrapolated from past trends - Current improvements assumed to converge to 1 pa
by 2031 for most ages, males and females - Convergence not linear more rapidly at first for
males, less rapidly for females - For those born before 1960, convergence assumed
along cohort - After 2031 rates of improvement assumed to remain
constant at the rate assumed for 2031 - Variants HLE target rate Principal 1
- LLE target rate Principal 1
42Projected smooth percentage changes in death
rates between 2006 and 2007 by age
UK Male/Female comparison
43Projected smooth percentage changes in death
rates between 2006 and 2007 by age
UK Male/Female comparison
44Projected smooth percentage changes in death
rates between 2006 and 2007 by age
UK Male/Female comparison
45Annual improvement in smoothed mortality rates,
Males, UK
46Annual improvement in smoothed mortality rates,
Females, UK
47Actual and assumed overall annual rates of
mortality improvement
Note Analysis relates to England Wales.
Historic estimates are based on comparison of
2003-05 Interim Life Tables with English Life
Tables for 1930-32, 1960-62 and 1980-82
48Period cohort life expectancy
2007 period and cohort life expectancy at various
ages, United Kingdom
Average number of additional years a person of
age x can expect to live a) according to the
mortality rates for 2007 b) according to
projected mortality rates
49Period and cohort expectations of life at birth,
United Kingdom
2006-based principal projections
50Period and cohort expectations of life at age 65,
United Kingdom
2006-based principal projections
51Projected period expectations of life at birth in
2050
- 2006-based projections
- Source latest published projections from
countrys national statistics website
52Period expectation of life at birth, UK
2006-based projections
53Period expectation of life at age 65, UK
54 Period expectation of life at birth, males, 1981
- 2044
2006-based principal projections
55 Period expectation of life at age 65, males,
1981 - 2044
- 2006-based principal projections
56 Period expectation of life at birth, females,
1981 - 2044
2006-based principal projections
57 Period expectation of life at age 65, females,
1981 - 2044
2006-based principal projections
58 Cohort expectation of life at age 65, males,
1981 - 2044
2006-based principal projections
59 Cohort expectation of life at age 65, females,
1981 - 2044
2006-based principal projections
60Continuous Mortality Investigation (CMI)
- Carries out research into mortality and morbidity
- Encompasses persons covered by long term risk
contracts issued by life assurance offices in the
United Kingdom and the Republic of Ireland. - Investigations cover all the main types of life
assurance, annuitant, pensioner, critical illness
and income protection insurance contracts offered
by the market. - The base data is supplied by life offices
covering the majority of the market. - All dealings with individual offices are
confidential.
61CMI Mortality Tables
- Standard period life tables
- Base tables graduated using data for 4
consecutive years - Extrapolated for oldest ages with ad hoc
adjustments of rates or parameters - Assured lives
- Select/ultimate
- Smoker/non-smoker
- Pensioners
- Lives/amounts
-
- For pensioners and annuitants, tables of
projected mortality rates - Standard tables (92 tables) have base
projections and three further variants allowing
for cohort effects for those born around 1926
(short, medium and long cohort) based on members
of insured pension schemes
62Pensions
- Different mortality bases may be used, on
actuarial advice - Valuation funding exercise assuming scheme
ongoing - responsibility of trustees prudent scheme
specific - Accounting purposes FRS17
- responsibility of directors best estimate
- Buy out costs
- prudent margins
- Costing options
- may have margins to protect scheme
- Similar principles apply to the financing of
(unfunded) public sector pension schemes
63Pensions
- In advising on a mortality basis actuaries will
usually analyse past mortality experience of the
scheme and compare to standard tables - Often adopt rates from standard tables with
adjustment, as thought appropriate - If very large scheme may use own experience
- Projections based on standard tables, with
adjustment - May also consider mortality rates or rates of
improvement in population projections - Bases likely to vary by occupations covered, type
of work, regions
64Cohort expectations of life at age 65, males
65Comparison of 92 tables and 2006-based pop
projections cohort eol65 UK males
66CMI 00 Series Tables
- New tables issued 1 Sept 06 based on
- 1999-2002 data 00 series
- Data at oldest ages sparse and unreliable
- Two methodologies proposed for projections,
emphasis now on providing measures of uncertainty
67SAPS investigation
- Data from consultants to pension schemes
- 354 validated submissions with 3.8m records to
March 2007 - Just over 300 schemes remainder resubmissions
- 2000-04 data analysis published 1 Oct
- Possible draft graduations for consultation
end-2007 - Possible paper on mortality improvements in 2008
- Results published as CMI Working Papers (WP29)
- http//www.actuaries.org.uk/Display_Page.cgi?url/
library/cmi.xml
68CMI Mortality Tables
- Insured lives mortality lower than general
population - Mortality by amount lower than by lives
- SAPS mortality higher than for those in insured
pension schemes - SAPS mortality differential by amount
- Differential highest at younger ages
- Differential by industry
- Financials lowest
- Basic Industries highest
69Projection methodologies
- Many different methodologies
- Predicting changes in mortality by cause
- Extrapolating past trends
- Aiming for target mortality in some future year
- Key feature for risk management is estimation of
uncertainty in projected mortality - CMI investigated stochastic methodologies
- Cohort effect considered important
70Probabilistic/stochastic methodologies
- Many candidate methodologies
- Regression/extrapolation/smoothing (e.g.
P-spline) - Time-series (e.g. Lee-Carter)
- Projection of future life tables (c.f. term
structures) - What has the CMI done so far?
- Explored P-spline models and Lee-Carter models in
detail - Issues with both both dependent on improvements
in past data - Published a library of around 50 different
projections - The CMI will contribute to research but does not
expect to recommend particular models
71Projections sources of uncertainty
- Model uncertainty
- Parameter uncertainty
- Stochastic uncertainty
- Measurement error
- Heterogeneity
- Past experience may not be good guide
- (e.g. change in business mix)
72Challenges
- Can we understand past drivers of mortality
change and quantify their effects? - Can better predictive models of mortality risk be
developed? - Where are the dangerous concentrations of
longevity risk? - Can mortality/longevity risk be securitised?
73Mortality improvements and Life Expectancy
- OPA 25 October 2007
- Adrian Gallop
- ONS