South African annuitant mortality

1
South African annuitant mortality

• CSI Committee
• Prepared by
• (Rob Dorrington and Suleiman Tootla)

2
CSI Committee

• Bernard Ross (Convenor)
• Rob Dorrington
• Judy Faure
• John Graham
• Stephen Jurisich
• Mike McDougall
• Richard Montgomery
• Gerhard Potgieter
• Johan Potgieter
• Salmien Symeonidis
• Colin van Zyl
• Philip van Zyl
• Frans Vergeest

3
Overview

• Background
• Data
• Results
• Graduation
• Time trend

4
Background

• Intial requests for participation in 1994
• Professor Rob Thomson
• Progress slow and data problematic
• By April 1999 Commercial Union (data for 1995),
  Liberty (1995), Metropolitan (none), Momentum
  (1995 and 1996?), Old Mutual (1995), Sanlam
  (1995-1997), SAAC (none)
• Many data problems, in particular implausibly low
  rates for those in guarantee period

5
Data

• Period 1996 2000
• Companies Liberty, Sanlam, Old Mutual, Momentum,
  Fedsure and Southern
• One company only 1997, 1999 and 2000 and another
  1999 and 2000 only

6
Data Deaths and exposure above age 60
7
UK CMI Immediate Annuitant Mortality
Investigations
8
Ages with significant data
9
Average amount per life year (Rand p.a.)
10
Data collected

• Voluntary/Compulsory purchase (V/C)
• Guaranteed/Non-Guaranteed period (G/N)
• Male/Female (M/F)
• Amount of annual annuity (at year end) 6 bands
  (ltR1,000 R3,000 R10,000 R30,000 R100,000
  R100,000)
• Age
• Central Exposure to Risk
• Number of Deaths
• Amount of Annuities Exposed to Risk
• Amount of Annuities in Respect to Death

11
Data excluded

• A back-to-back annuity arrangements
• Deferred, reversionary and contingent annuities
• Annuities issued in a territory outside the
  Republic of South Africa
• Annuities payable in a currency other than South
  African currency
• Annuities paid to or on behalf of a pension or
  provident fund
• Annuities with missing data (date of birth, sex,
  whether voluntary or compulsory purchase, whether
  the annuity commenced during the year, whether
  the annuity ceased during the year)

12
Some data problems

• Negative exposure and deaths in data from 2
  offices
• Rounding down of exposure by 4 offices
  overstating mortality, possibly significantly at
  older ages
• Inconsistency in age definition in data from 1
  office
• Rates from 2 offices implausibly low

13
Results
14
Male CIs vs standard tables
15
Female CIs vs standard tables
16
Amount vs lives males
17
Amount vs lives females
18
Mortality rates by class
19
Actual vs. Expected for Class

• Overall Rates taken as standard
• Age range 50- 94

20
Males by company
21
Male outliers vs UK rates
22
Female outliers vs UK rates
23
Male rates excluding outliers
24
Female rates excluding outliers
25
Graduation

• Two methods
• Gompertz-Makeham fitted using Generalized
  Non-Linear Models
• Reference to a standard table, the standard being
  a(90) less 1 year, and

26
Chi-squared (vs 5 critical) males
27
Crude vs graduated rates males
28
Chi-squared (vs 5 critical) females
29
Crude vs graduated rates females
30
Graduated vs Assured 85-90 rates
31
Trend over time males
32
Trend over time females
33
Trend over time males individual years
34
Trend over time femalesindividual years
35
Actual vs. Expected for Year

• Overall Rates taken as standard
• Age range 50- 94

36
Actual vs. Expected for Year
37
Conclusions

• No difference in rates between classes
  (voluntary/compulsory, guaranteed/non-guaranteed,
  by amount) except sex
• Any difference is in the wrong direction!
• Apparent time trend but needs more data and
  investigation
• Data problems need to be addressed what does it
  say for our science with these sorts of
  problems (particularly companies 22 and 25)

38
Conclusions

• Should collect data on duration
• Pattern similar to a(90) suggesting our rates lag
  those in UK by 5-10 years however, need to
  understand trend in the UK rates
• Change in class means that annuitant rates no
  longer lower than those with permanent assurance
  (could also be the case in SA?)
• Should we produce a standard table?
• Separate tables for voluntary and compulsory?