Mortality Report. Individual Life Experience Subcommittee Research Committee. April Document

Transcription

1 Mortality Report Canadian Standard Ordinary Individual Life Mortality Experience Between the 2004 and 2010 Anniversaries Using and Tables: Databases from Seriatim Data Individual Life Experience Subcommittee Research Committee April 2013 Document Ce document est disponible en français 2013 Canadian Institute of Actuaries

2 CANADIAN STANDARD ORDINARY INDIVIDUAL LIFE MORTALITY EXPERIENCE BETWEEN THE 2004 AND 2010 ANNIVERSARIES USING AND TABLES: DATABASES FROM SERIATIM DATA The CIA Individual Life Experience Subcommittee of the Research Committee publishes industry mortality experience reports on an annual basis. In the past, each report provided tables 1 to 10 as an appendix in PDF format. Those wishing to do their own analyses needed to scan or retype the data. In the study, the appendix included Excel versions of the tables, but even those hindered analysis by being one-dimensional with no complete picture of the interactions among risk factors. For example, table 8 has select experience A/E ratios by duration for all issue ages combined and by issue age for all durations combined, but there is no way to break down duration by issue age. Early in 2011, the subcommittee decided to proceed with the Mortality Database Project in order to create detailed data files that represent the data underlying the recent annual mortality experience studies. The following database files were created: 1. IndLifeMDB.0405.zip 2. IndLifeMDB.0506.zip 3. IndLifeMDB.0607.zip 4. IndLifeMDB.0708.zip 5. IndLifeMDB.0809.zip 6. IndLifeMDB.0910.zip These are available for download from the CIA website. Mortality Database Format This is a description of the format of mortality databases for each study year from through We used the CIA and CIA mortality tables to separately calculate the expected death claims for males and females and for smoker/non-smoker distinctions. We calculated exposure using the Actuarial Exposure method, in which a full year of exposure is credited in the year of death. Ages are calculated on an age-nearest birthday basis. Those companies that have a preferred business indicator also designated whether the business is standard or preferred, smoker, non-smoker, or cigar-smoker, and the class. For data submitted without classification by sex, we used the male table. For data with attained age less than 16, we assigned mortality rates from the aggregate expected table. The following fields are included in the six comma-delimited text files. There is one text file per policy year. 2

3 Year Sex Smoker Type of Ins Evidence Face Size: 2005 = ; 2006 = , etc. 1 = Male; 2 = Female; 0 = Unknown. 1 = Smoker; 2 = Non-smoker; 3 = Smoking status unknown. Policy Type: 1. Whole Life 2. Renewable Term 3. Universal Life 4. Term Rider Underwriting Type: 1. Medical 2. Non-medical 1. < $10, $10,000 $49, $50,000 $99, $100,000 $249, Term to Other Permanent 7. Other Term 8. Other. 3. Unknown 4. Paramedical. Preferred Class: (Not available for experience study ) 5. $250,000 $499, $500,000 $999, $1,000,000 and over. Policy experience is divided between Preferred and Standard underwriting types. If the underwriting type is Preferred, the policy experience is further divided by Preferred Class as either Preferred (better than standard) or Standard (last) class: 01 = standard underwriting and N/A for preferred class 02 = preferred underwriting and preferred class = standard 03 = preferred underwriting and preferred class = preferred DB Dur Policy Duration DB Issue Age Issue age Sum of # Exposed Sum of $ Exposed Sum of # Deaths Sum of $ Claims 8692 # Exp Deaths 8692 QX Sum of # Exposed / 1, $ Exp Claims 8692 QX Sum of $ Exposed / 1, # Exp Deaths 9704 QX Sum of # Exposed / 1, $ Exp Claims 9704 QX Sum of $ Exposed / 1,000 Starting with the study, each mortality experience report will include a database in the format outlined above. Reconciliation of the Published Results Against the Mortality Database A significant effort was made to reconcile the published results against summaries generated by using the Mortality Database. The results of this investigation are outlined in the attached files: 3

4 C8692 Summary.xlsx C9704 Summary.xlsx The main focus was on table 1 for each annual study and for two of the five-year studies. Note that table 1 was published on an annual basis beginning with the study. There are some small differences between the published reports and the mortality database; these are explained below. Records with Unknown Face Amount Information The database files include only records with known policy face amount information. In the past, all experience study data were provided on an aggregate basis. Those submissions included exposure and actual claim information for each combination of duration, issue age, attained age, sex, smoking, and other factors. This structure allowed calculation of expected values and production of standard reports with the exception of tables by policy size. The most recent studies were mainly based on seriatim data with a small part still on an aggregated basis. This small portion of the original data had to be excluded from the database because of missing policy face amounts on individual policies. Therefore, prior to the experience study, the database does not match the published actual and expected exposures and deaths. However, the table 1 A/E ratios in the published reports are similar to those in the Mortality Database. CIA Expected Rates The published results are based on a version of the tables that differs in several minor respects from the original tables published by the CIA and/or widely used by Canadian actuaries in the GGY Axis modelling software: In the annual studies, juvenile rates are smoking distinguished, unlike the unismoke rates implemented in Axis. Table extensions for issue ages 81 through 99 were provided with the introduction of the rates but were never officially reviewed by the Individual Life Experience Subcommittee. These extensions, used in the annual studies, differ from those in the Axis tables that use select rates equal to the rates for the previous issue age shifted up one year. In the annual studies, ultimate rates for attained ages greater than 99 were equal to the mortality rate for attained age 99. In order to eliminate any misinterpretation of the Mortality Database and future study results, the subcommittee decided to replace the original version of the tables with the widely used tables in the Axis actuarial software. The recent reconciliation analysis confirmed that the published numbers are reasonably close to the results generated from the Mortality Database. CIA Expected Rates The published tables are also limited to issue ages less or equal to 80. There is no official extension for issue ages greater than 80. In order to calculate expected values for the experience study, ultimate rates were used for issue ages greater than 80. The subcommittee returned to this issue and implemented an extension to the tables (for issue ages 81 through 99) using the same approach described above for the tables, i.e., the select rates are equal to the rates for the previous issue age shifted up one year. The overall impact of these 4

5 changes was minimal. The decision was made not to correct the published study results, but to implement these extensions in the Mortality Database and study. Rounding The published dollar amounts were original dollar amounts divided by 1,000 and rounded to the nearest whole number. The Mortality Database generates dollar amounts without rounding, which is another reason for small discrepancies between the two sources, particularly for cells with small dollar amounts. Outliers The Mortality Database includes all the underlying data from the last six published studies except for records with unknown face amounts. There are two areas where the user should be careful when interpreting results: ultimate results for attained ages greater than 94 and results for issue ages greater than 89. It appears that not all deaths were reported to insurance companies for attained ages greater than 94 based on the observed actual mortality rates. That the observed mortality rate decreases for these ages is counterintuitive. There are also a few active records with attained ages greater than 119 or issue ages greater than 89. The subcommittee discussed filtering these outliers but decided to include all data that had been included in the studies. This report was approved by the Individual Life Experience Subcommittee of the Research Committee. Marc-André Belzil (Research Committee Chair) Dave Dickson (Research Committee Vice-chair) Nikolai Serykh (subcommittee chair) Mark Andrews Blake Hill Marie-Josée Blanchet Simon Bélanger Jean-Pierre Cormier Scott Spencer Rhys DeGrave Lisa Zwicker Kim Girard This report was prepared by: Barbara Thomson Thomson Data Analysis Toronto, ON We give a special thanks to the University of Waterloo actuarial students who helped to validate and reconcile underlying data: Chris Cheng, Curtis Chim, and Kelly Zhan. 5