Cohort and Period Mortality in Sweden A nearly 150 year perspective and projection strategies Hans Lundström Statistics Sweden Joint Eurostat/UNECE Work Session on Demographic Projections Lisbon, April 2010
Available mortality data 2 Mortality rates by sex in one year age classes 1861 to 2009 Age is calculated based on registered data Quality of data is high
Mortality rates by calendar year and year of birth 3 0
Cohort and period perspective 4
Female mortality Age Apart from stochastic fluctuations mortality rate is closely related to mortality rate for next higher/lower age and to mortality rate in preceding/following year. Stable pattern
Male mortality Age
Females Males Mortality show a smooth and gradual change over time and age.
Mortality projection 8 To sum it all up: The Lee-Carter model for the last 15 to 20 years fits nicely to male and female mortality
Problem 9 Male mortality is declining more rapidly then female mortality Lee-Carter method results in a cross-over after some years Is a situation with lower mortality for males than for females likely?
Male excess mortality for nearly all ages and years Age
Using Lee-Carter Age-Period model the projection results in lower mortality for males than for females in a 30 year perspective Likely?
A new model is needed To the two-factor model Age & Period we must add Sex to the model
An observation.. A more rapid mortality decline for females started in the 1950s. For males we had a 30 year lag before mortality started to decline. Present rate of progress for males similar to that for females in the 1950s
14 Male mortality shifted back 30 years in time for ages 40,50,60,70, 80 and 90
The cohort perspective Mortality show a smooth and gradual change over time and age in a cohort perspective too
For ages above 50 nearly parallell shift of mortality curve from one cohort to the next. Female mortality. Cohorts
Male mortality. Cohorts
A new model is needed So far we have used the Lee-Carter Age & Period model The Lee-Carter model has lately been extended to incorporate cohort effects too. This model is worth a closer look Probably we still have to add Sex as a fourth factor to the model
Alternative future approach We must gain better insight into the causes and predictors of mortality For that we must know the ”risk profile” for cohorts and know the relationship between”risk factors” and mortality Much remains to be done A first step is look into cohort cause-of-death data
Thank you for your attention 20
A cohort-based extension to the Lee–Carter model for mortality reduction factors. A.E. Renshaw, S. Haberman Cass Business School, City University, London, EC1Y 8TZ, UK 21 Abstract The Lee–Carter modelling framework is extended through the introduction of a wider class of generalised, parametric, non-linear models. This permits the modelling and extrapolation of age- specific cohort effects as well as the more familiar age-specific period effects. The choice of error distribution is generalised. Insurance: Mathematics and Economics 38 (2006) 556–570
Life expectancy
Remaining life expectancy at age 65, 75 and
Female mortality rate. Age 0-9
Male mortality rate. Age 0-9
Female mortality rate. Age 10-19
Male mortality rate. Age 10-19
Female mortality rate. Age 20-29
Male mortality rate. Age 20-29
Cohort mortality in Sweden - Mortality statistics since Cohort life tables in Excel format: The future population of Sweden publication The future population of Sweden publication