The SAINT mortality model: theory and application Quant Congress USA New York, 9 July 2008 Tryk Alt+F8 og Afspil auto_open for at vise værktøjslinien til.

Slides:



Advertisements
Similar presentations
Healthy life expectancy in the EU 15 Carol Jagger EHEMU team Europe Blanche XXVI Living Longer but Healthier lives Budapest November 2005.
Advertisements

June 23, 2003AVW International Mortality Comparisons Richard MacMinn Richard MacMinn Edmondson-Miller Chair Katie School College of Business Illinois.
Backtesting of Stochastic Mortality Models: Kevin Dowd (CRIS, NUBS) Andrew J. G. Cairns (Heriot-Watt) David Blake (Pensions Institute, Cass Business School)
1. Which of the following statements about conventional life tables is NOT true? They were first used by life insurance companies to estimate human survival.
Katja Hanewald - Lee-Carter and the Macroeconomy H U M B O L D T – U N I V E R S I T Ä T Z U B E R L I N Mortality Modeling: Lee-Carter and the Macroeconomy.
The Danish Labour Market Social security Active labour market policies Life long learning Dynamic labour market Social partners Public authorities The.
Lectures 2, 3 Variance in Death and Mortality Decline Shripad Tuljapurkar Ryan D. Edwards Queens College & Grad Center CUNY.
Which is the Optimal Portfolio in Retirement? FOR DUE DILIGENCE ATTENDEES Van Harlow 19 May 2006.
V. Statistical Demography
Giving and Receiving Aid: Does Conflict Count? Eliana Balla Gina Yannitell Reinhardt.
OECD Short-Term Economic Statistics Working PartyJune Impact and timing of revisions for seasonally adjusted series relative to those for the.
© The Treasury 2009 Job Summit John Whitehead, Secretary to the Treasury.
Hunter centre for strathclyde Global Entrepreneurship Monitor Scotland 2002 Jonathan Levie Wendy Brown Laura Galloway.
Part 2 – US Social Security System from an International Perspective How similar or different is the Social Security system to that of other developed.
Poverty & Human Capability 101 Introductory Class.
Section 4.1 Exploring Associations between Two Quantitative Variables?
New approaches to study historical evolution of mortality (with implications for forecasting) Dr. Natalia S. Gavrilova, Ph.D. Dr. Leonid A. Gavrilov, Ph.D.
International Survey of Adult Skills (ISAS) Policy Results 14th October 2013.
Swedish Health Care in Transition Swedish Health Care in Transition Resources and Results with International Comparisons.
Overview of U.S. Results: Digital Problem Solving PIAAC results tell a story about the systemic nature of the skills deficit among U.S. adults.
1. Measuring the Impact of Universal Preschool Education and Care on Literacy Performance Scores. Tarek Mostafa Institute of Education – University of.
PIAAC results tell a story about the systemic nature of the skills deficit among U.S. adults. Overview of U.S. Results: Focus on Numeracy.
New Skills for New Jobs: Action Now Professor Mike Campbell OBE Director of Research and Policy ETUC Conference International Trade Union House, Brussels.
1 Disability trends among elderly people in 12 OECD countries, and the implications for projections of long-term care spending Comments on Work Package.
Solar Physics Board Meeting Rio de Janeiro July, 2009.
Global Science Forum OECD Global Science Forum Study on Declining interest in science studies Preliminary Report on the Quantitative Analysis Prof. Jean-Jacques.
New approaches to study historical evolution of mortality (with implications for forecasting) Leonid A. Gavrilov, Ph.D. Natalia S. Gavrilova, Ph.D. Center.
Reversing the reversal? The cross-country correlation between female labour market participation and fertility revisited Anna Matysiak and Tomáš Sobotka.
1 Guide to the PISA Data Analysis ManualPISA Data Analysis Manual Computation of Standard Errors for Multistage Samples.
University of New South Wales
Overview of U.S. Results: Focus on Literacy PIAAC results tell a story about the systemic nature of the skills deficit among U.S. adults.
OECD Short-Term Economic Statistics Working PartyJune Establishing guidelines for creating long time series for short-term economic statistics.
Statistics about unknown primary tumors Riccardo Capocaccia National Centre for Epidemiology, Surveillance and Health Promotion Istituto Superiore di Sanità,
Countries of Europe France Spain Italy Germany Which country is this?
Population Mortality and Morbidity in Ireland n April 2001.
1 Modeling Coherent Mortality Forecasts using the Framework of Lee-Carter Model Presenter: Jack C. Yue /National Chengchi University, Taiwan Co-author:
Abcd AGEING POPULATION - Burden or Benefit? Demographic Trends Adrian Gallop Edinburgh 21 January 2002.
1 Mortality Compression and Longevity Risk Jack C. Yue National Chengchi Univ. Sept. 26, 2009.
INSTITUTE OF OCCUPATIONAL MEDICINE EDINBURGH, EH8 9SU, UK Potential magnitude of chronic mortality effects of air pollution J Fintan Hurley & Brian G Miller.
Helmholtzstraße 22 D Ulm phone+49 (0) 731/ fax +49 (0) 731/ It Takes Two: Why Mortality Trend Modeling is more.
Demographic Uncertainty and the Sustainability of Social Welfare Systems Jukka Lassila ETLA Finland.
May 21, 2013 Lina Xu Methods to model Mortality Improvement 2. Lee Carter Model 3. Model fit/Analysis/Result 4. Industry Mortality Improvement.
The Cairns-Blake-Dowd Model Andrew Cairns Maxwell Institute & Heriot-Watt University David Blake Pensions Institute & Cass Business School Kevin Dowd Nottingham.
Peterson-Kaiser Health System Tracker How does health spending in the U.S. compare to other countries?
The Ambiguous Crisis of Global Economic Inequality: Contradictory National and International Trends? WUN Horizons in Human Geography Seminar Series November.
International Comparison of Health Care Gene Chang.
MORTALITY CHANGE THE DETAILS ARE MESSY Year to year decline irregular Persistent, puzzling differentials Cause of death structure difficult to understand.
At each period of the history of mankind, since the most primitive times, a small number of individuals were able to live up to 100 years and to thus carry.
Was slowing postponement really the engine for TFR rises in European countries? Marion Burkimsher Affiliate researcher University of Lausanne, Switzerland.
The Simple Linear Regression Model. Estimators in Simple Linear Regression and.
2014-based National Population Projections Paul Vickers Office for National Statistics 2 December 2015.
INTERNATIONAL MIGRATION DATA as input for population projections Anne HERM and Michel POULAIN Estonian Interuniversity Population Research Centre, Estonia.
Dan Kašpar, Klára Hulíková Charles University in Prague, Faculty of Science, Department of Demography and Geodemography
USAGE OF DRUGS IN EUROPE LSD CANNABIS. ALL ADULTS (15-64) USAGE OF LSD IN EUROPE All adults (15-64) Usage of LSD in Europe datesample sizemalefemaletotal.
Life expectancy Stuart Harris Public Health Intelligence Analyst Course – Day 3.
Cohort religiosity: does it stay at a stable level everywhere and across all cohorts? Marion Burkimsher University of Lausanne.
Linear Regression Hypothesis testing and Estimation.
1 Main achievement outcomes continued.... Performance on mathematics and reading (minor domains) in PISA 2006, including performance by gender Performance.
Stochastic Sensitivity Testing for the Canada Pension Plan Actuarial Reports Assia Billig, Office of the Superintendant of Financial Institutions, Canada.
EXPOSURE TO TOBACCO SMOKE IN THE EUROPEAN UNION 2nd Working Meeting on Adult Premature Mortality in the European Union October 2006, Warsaw, Poland.
Improvements in Life expectancy and sustainability of social security schemes – A Danish perspective ChrestenDengsoe Chresten Dengsoe.
PISA 2015 results in England
John Jerrim UCL Institute of Education
Which is the Optimal Portfolio in Retirement?
Mohammad Jalal Abbasi-Shavazi
Presentation for Session VI.
Census and forecast, Mexico from 1940 to 2050.
Bayesian Inference for Small Population Longevity Risk Modelling
NORC and The University of Chicago
2006 Rank Adjusted for Purchasing Power
Presentation transcript:

The SAINT mortality model: theory and application Quant Congress USA New York, 9 July 2008 Tryk Alt+F8 og Afspil auto_open for at vise værktøjslinien til opdatering af automatisk indsat tekst (forfatterinfo og præs.overskrift 2 på masterdias) Søren Fiig Jarner Chief Analyst

The SAINT mortality model Agenda  Motivating example: Danish mortality -data highly volatile, but with underlying structure -Danish vs. international mortality  The SAINT framework -short-term deviations from long-term trend  Population dynamics and frailty  The model -illustrative example  Forecasts and uncertainty

The SAINT mortality model Evolution of Danish female mortality Δlife expectancy = 21 yrs Life expectancy 40 yrs (1835) Life expectancy 80 yrs (2006) Δlife expectancy = 6 yrs Δlife expectancy = 13 yrs See Jarner et. al (2008) for the life expectancy decomposition

The SAINT mortality model A more detailed look at the recent development Danish female mortality Age Sharp decline in young age mortality Very little improvement at the highest ages Stagnation/increase from 1980 to 1995 High annual rates of improvement

The SAINT mortality model Simple projections very sensitive to estimation period! Age Reasonable short-term projections Implausible long-term projections lacking (biological) structure Danish female mortality

The SAINT mortality model Data characteristics and modelling challenge  General pattern -age-specific mortality rates declining over time -rates of improvement decreasing with age (rectangularization)  Substantial deviations from the general pattern -even periods with increasing mortality for some age groups  Challenge: Produce plausible, long-term forecasts reflecting both the underlying trend and the ”wildness” seen in data  Idea: Estimate the underlying trend from less volatile reference data

The SAINT mortality model Data and terminology  Human Mortality Database (  Danish and international female mortality from 1935 to countries in the international dataset: USA, Japan, Germany, UK, France, Italy, Spain, Australia, Canada, Holland, Portugal, Austria, Belgium, Switzerland, Sweden, Norway, Finland & Iceland  Death counts and exposures for each year and each age group D(t,x) = number of deaths E(t,x) = exposure (”years lived”) Death rate, D(t,x)/E(t,x), is an estimate of (the average of) underlying intensity, μ(t,x) Death probability, q(t,x) = 1-e -∫μ(t,x) ≈ ∫μ(t,x) t t+1 time x x+1 age

The SAINT mortality model Danish fluctuations around stable international trend Danish and international female mortality Age Danish life expectancy among the highest in the world Similar development at the highest ages Denmark falling behind the international trend Is this the beginning of a catch up period?

The SAINT mortality model SAINT (Spread Adjusted InterNational Trend) framework Parsimonious parametric model for long-term trend Time-series model for short-term deviations (spread) : Family of intensity surfaces (gender specific) : MLE based on Poisson-model; : Age-dependent vector of regressors (fixed) : Time-dependent spread parameters (estimated); Fit multivariate time-series model for

The SAINT mortality model Trend modelling concepts  Population dynamics -Ensure consistent intensity surfaces over time and ages by aggregating individual intensities to population level -Individuals living in the same period of time are influenced by common as well as individual factors -Some factors have a cumulative effect on mortality  Frailty -People are genetically different. Only the more robust individuals will attain very high ages -Lack of historic improvements among the very old may be due to selection effects. In the future the frailty composition at old ages will change

The SAINT mortality model From individual to population intensity  Mortality intensity for an individual with frailty  Individual survival function  Survival function for population with frailty distribution  Population intensity

The SAINT mortality model Selection effects within a cohort Intensity (μ) (x) Individual: Cohort:

The SAINT mortality model Selection when mortality is time-varying Average frailty in population Individual:

The SAINT mortality model Trend model  Underlying individual intensities  Population intensity (mean 1 and variance σ 2 Γ-distributed frailties) Previous values of κ are ”remembered” by the population ”treatment” level ”wear-out” rate ”accident” rate

The SAINT mortality model Trend – fit and forecast International female mortality Age General, long-term rate of improvement = 1.8% Early, young are rate of improvement = 9.1% Increasing old age rate of improvement

The SAINT mortality model Spread model  Model of Danish mortality  The spread is assumed to fluctuate around zero -that is, no mean term included in the model  The spread controls the length and magnitude of deviations -and provides information about projection uncertainty Mean zero, orthogonal regressors normalized to (about) 1 at age 20 and 100

The SAINT mortality model Illustration of spread adjustment Estimates a 2004 = 21% b 2004 = 5% c 2004 =-19% International trend Danish data Danish fit Female mortality in 2004

The SAINT mortality model Long recovery period Fitted a t Fitted b t Fitted c t Forecast Estimated and forecasted spread

The SAINT mortality model Danish mortality – fit and forecast Danish female mortality and international trend Age Denmark falling behind … and catching up again Similar development in old age mortality

The SAINT mortality model Forecast uncertainty  Analytical methods -only feasible for very few quantities of interest, e.g. the spread itself  Monte Carlo -simulate N spread series and calculate mortality forecasts for each -calculate quantity of interest, e.g. life expectancy, for each forecast -compute uncertainty measures, e.g. 95%-confidence intervals … … Females aged 60 in 2005

The SAINT mortality model Summing up  Model for small population mortalities showing irregular patterns of improvement  Parsimonious trend model -estimated from reference population -biologically plausible mortality projections -future improvements in high age mortality as frailty composition changes  Time series model for deviations from trend -spread controls length and size of excursions from trend  Projection uncertainty calculated by Monte Carlo methods

The SAINT mortality model Selected readings  Lee & Carter (1992). Modelling and forecasting U.S. mortality. JASA,  De Jong & Tickle (2006). Extending Lee-Carter mortality forecasting. Mathematical Population Studies,  Cairns et al. (2007). A quantitative comparison of stochastic mortality models using data from England & Wales and the United States.  Vaupel et al. (1979). The impact of Heterogeneity in Individual Frailty on the Dynamics of Mortality. Demography,  Thatcher (1999). The Long-Term Pattern of Adult Mortality and the Highest Attained Age. JRSS A,  Jarner, Kryger & Dengsøe (2008). The evolution of death rates and life expectancy in Denmark. To appear in Scandinavian Actuarial Journal.