The Continuous Mortality Investigation Bureau Chris Daykin CMI
The Continuous Mortality Investigation Bureau Chris Daykin, CMI Executive Committee
The CMIB • • • History Role Structure Funding Investigations Reporting results NB “Office” = “company”
History • actuaries produced Mortality table - 1843 – “Seventeen Offices’ Table” – assured lives – experience up to 1837 • further tables during 19 th century • investigation into annuitants 1900 -20 • continuous collection of data started in 1924 – emergence of the CMI Bureau
Features • • • sponsored by the actuarial profession continuous investigations independent confidentiality is paramount production of standard mortality tables actuarial profession provides expertise
Standard Tables Period Assured Lives Annuitants Pensioners 1924 -29 (males) 1947 -48 1949 -52 (males) 1967 -70 (males) 1975 -78 (females) 1979 -82 1991 -94
Comparison of the mortality of male assured lives
Role of CMI • Research – Mortality, IP and CI. – Methodologies – Graduation – Models • Data collection • Analysis & reporting – Industry experience – Contributing offices • Standard Tables • Projecting future experience
Structure Life Companies and Profession CMIB Life Companies
Structure of the CMIB Executive Committee Management Committee Mortality IP Secretariat/Bureau CI
Who serves on the Committees? • • • life office actuaries reinsurance actuaries consultants government actuaries academics
Role of the Secretariat § Servicing committees § organising Meetings § drafting standard reports § printing and distribution of CMI Reports § Day to day operations § collecting data § corresponding with offices § producing results § collecting money & accounts
Funding Each office bears their own data contribution cost + Contributions based on premium income Change to risk-based approach?
Investigations • Mortality – life contracts issued at standard rates – impaired lives – annuitants – individual pension arrangements – group pension arrangements
Investigations • Income Protection – individual policies – group policies • Critical Illness
Data Timetable • • • Collect data as at each 31 December Wait until 30 June July October: collect and process data Nov Dec: final chasing & checking December: run & distribute “all office” results
Reporting results Own Office Results – As soon as data is clean – Data summary – A/E comparison with standard tables – Special requests
Confidentiality • • taken extremely seriously only Secretariat & office sees results office numbers can be restrictive
Reporting results All Office pooled results – annual – quadrennial – available to members first – interim results – available to all member offices
Reporting results To the Actuarial Profession – CMI Reports (CMIRs) – the profession’s magazine & internet site – conferences – sessional meetings
Data Collection Methodologies
Main methods • What are we doing? – What are we measuring? – Definitions • Census • Policy data
Example of census data Age 20 21 22 23 24 25 26 27 In force at 31/12/t IF 20. t IF 21. t IF 22. t IF 23. t IF 24. t IF 25. t IF 26. t IF 27. t Deaths in year t D 20. t D 21. t D 22. t D 23. t D 24. t D 25. t D 26. t D 27. t
Census - calculations • • Exposure = ½ (IFx, t + IFx, t+1) + ½ Dx. t correspondence between in force and deaths Expected deaths = Exposure * q compare Actual & Expected deaths – 100 A/E
Census method • approximate • currently used by CMIB in mortality investigation for historical reasons • offices provide schedules showing number of policies at each age in force at 1 January and deaths during year • ongoing: start in force = previous year end in force • care with age definitions
Census - drawbacks • approximate, so reduced accuracy • limited checking of underlying data possible • limited scope for analysis of subgroups – durations – policy types • cannot analyse “amounts” properly • policy alterations hard to spot • duplicates
Census - advantages • less data (can be handled manually) • less work to check data • cheaper
Policy data • Data on per policy basis at each 31/12/t – date of birth (avoids defn. problems) – sex – start date of policy – date of death/claim/exit – type of exit – policy type – amount of benefit – identifier
Policy data method • IP & CI investigations use this method • exposure calculated exactly for each policy by counting days • calculation of expected deaths & 100 A/E as with census method
Policy data – features • advantages over census method – – greater accuracy more checking possible better data quality more control over data included in investigations more detailed analyses possible • should be easier for offices to supply But • increased storage requirements • more complex to process - hence expensive
Observations (1) • • • need for detailed rules consistent interpretation across offices must check to make sure data is sensible will have delays in data collection offices “come and go” office mergers
Observations (2) • staff who produce data are not the same as staff who use the results • sometimes difficult to get offices to pay attention • speedy turn around helps data quality • data audits
Common data problems • • • policy alterations (e. g. amounts) duplicates What is a claim? (claim date in IP) multiple claims (IP) matching data across periods consistency - over time - between offices
Sub-population differences
Questions to be investigated • Do differences justify a standard table? • if not, how to adjust current table? – pricing – valuation • trends in sub population
Categories investigated Main categories • age • male / female • policy type • duration • smoker / non-smoker • impairment Other possible ( but only have insurance data) • regional variation • social variation
Variations by age Plot of AM 92 qx by age qx Age x
Variation by sex
Assured lives - Variation by duration
Variation by smoker status
Variation by policy type
Sub-population comments • must collect data! • data collection follows market • companies that innovate via sub-population differences are exposed • getting credible data sometimes difficult • takes time for investigations to get established
- Slides: 41