Canadian Group Annuitant Mortality Experience, Calendar Years

Transcription

1 Study Canadan Group Annutant Mortalty Experence, Calendar Years Penson and Group Annuty Experence Subcommttee Research Commttee Aprl 2017 Document Ce document est dsponble en franças 2017 Canadan Insttute of Actuares

2 1 Executve Summary Canadan Group Annutant Mortalty Experence, Calendar Years Ths s the naugural report of a study of experence under the group annutes of lfe nsurance companes. There are sx years of data ncluded, and the amount of experence s a lttle less than that of the ndvdual annutant experence study and exceeds that whch underles the constructon of CPM2014prv (but not CPM2014, the combned publc and prvate table). The experence s broadly consstent wth that of CPM2014prv, but there are varatons by age whch may be materal. The varaton n experence by sze s smlar to that observed n the CPM study. There s some suggeston that the slope by sze could be steeper for males n ths study than n CPM, but there s not enough data to be confdent. 2 Table of Contents 1 Executve Summary Table of Contents Introducton Overvew Analyses Sngle/Jont Defned Beneft or not Sze of ncome Duraton from retrement Supplementary Informaton Subcommttee References Introducton The CIA has, snce the 1980s, collected data on ndvdual annutant mortalty. In 2014, t publshed a mortalty table based on data collected from a number of penson plans. There s another sgnfcant group of annutants that s worthy of study, and that s the group annutant experence of lfe nsurance companes n Canada. Accordngly, the CIA Research Commttee (ResCo) ntated a study n There has now been suffcent data collected to warrant publshng a report. 2

3 Sx lfe nsurance companes are currently partcpatng n the ntercompany study. Table 1 shows the companes and the proporton of exposure, measured by ncome, contrbuted by each. Table 1. Proporton of exposure by ncome from contrbutng companes. Company Share Canada Lfe 18% Desjardns 7% Industral-Allance 16% Manulfe 13% Standard Lfe 21% Sun Lfe 24% Throughout ths report, expected mortalty s calculated on CPM2014prv wth projecton on CPM-B to the approprate calendar year, wthout sze adjustments unless expressly stated. (Ths mortalty bass s called qcpm n ths report for convenence.) CPM2014prv s used rather than the composte CPM2014 because t s unlkely that group annutant busness ncludes many annutes from the publc sector. The use of qcpm should not be nferred as an ndcaton of t beng approprate for ths block of busness. There s not yet any publshed table that s drectly relevant. Ths study wll observe how closely qcpm fts to group annutant mortalty experence. Several tables n ths report show standard devatons n the actual-to-expected ratos. These ratos are calculated on the assumpton that the exposure of each lfe to death n the next year s ndependent of the exposure for all other lves, that the number of deaths for any group of lves wth the same sex-age-year s bnomally dstrbuted, and that the mean of the dstrbuton s gven by qcpm. The formula for standard devaton s shown below, by ncome, where K s the annualzed ncome and n s the number of annutants wth that ncome and that sex-age-year. The sum s over all annutants under consderaton. The same formula may be used by count except that K s 1 n all cases. Standard Devaton of A/E by Income = 4 Overvew K 2 n p q K n q All sx companes have contrbuted data for the calendar years 2007 to The data were subjected to checks to ensure consstency from one year to the next, and correctons were made as needed. All companes have sgned off on ther data as suffcently accurate for the purpose of ths study. (Data s currently beng submtted for 2013 and 2014; a subsequent study s expected.) 0.5 3

4 The data submtted and study method are very smlar to that used by the ndvdual annutant mortalty study. The excepton s that ths study has also requested ndustry codes for each group and postal codes for each annutant. Not all companes are able to provde ndustry codes or postal codes at ths tme, and accordngly reportng on these has been deferred untl enough companes can nclude these data. Ths study s able to dstngush experence based on sex, year of brth, year of experence, amount of annualzed ncome, whether the penson arses from a DB plan, and whether the annuty s sngle lfe or jont. Each contrbutng company provded factors to be appled to reported deaths to make an estmate of ncurred but not reported deaths (IBNR). All deaths and the annualzed ncome of deaths n ths report have been adjusted for IBNR as of the date of the data extract for The same factor s used for both count and ncome of deaths. Exposures are not adjusted. The IBNR factors of the contrbutng companes are consdered confdental. However, table 2 s ncluded n order to gve readers an dea of the magntude of the factors. The table shows the smple average of the IBNR factors for the sx contrbutng companes, combnng male and female. For example, deaths on sngle lfe annutes submtted three years earler are ncreased by 0.8% to allow for IBNR. Table 2. Smple average of IBNR factors Delay Sngle Prmary Secondary 1 5.0% 6.9% 76% 2 1.0% 2.3% 31% 3 0.8% 1.4% 15% 4 0.3% 0.7% 5.9% 5 0.1% 0.2% 1.5% 6 0.1% 0.1% 1.2% Table 3 compares the exposure and deaths ncluded n ths report to the totals from the studes that underle the constructon of the ndvdual mortalty table CIP2014 and the prvate sector penson table CPM2014prv. (Not all ages were actually used n table constructon.) The table shows that the group annutant data set s a szeable one. Note that the group data ncludes only sx years of experence, compared to 12 years for ndvdual and 10 years for pensoners. Table 3. Data n recent studes (annualzed ncome n thousands) Exposure Deaths Source Table Count Income Count Income Group Annutants none yet 2,287,560 12,517, , ,999 Indvdual Annutants CIP2014 3,649,413 14,845, , ,949 DB Pensoners (prv) CPM2014prv 962,899 10,519,535 46, ,876 4

5 Chart 1 shows the dstrbuton of exposure, by ncome, nto qunquennal age groups. The medan age s 75.8 for males and 74.4 for females. There s relatvely lttle exposure other than n the range Note that the age dstrbuton for females s slghtly flatter than for males. 25% Chart 1. Proporton of Exposure by Age 20% 15% 10% 5% 0% Male Female Chart 2 shows the dstrbuton of exposure, by ncome, over each of the sx calendar years ncluded n the study. The dstrbuton s farly flat, ncreasng slghtly toward the more recent years. The ncrease s greater for females than for males. 18.5% Chart 2. Proporton of Exposure by Year 18.0% 17.5% 17.0% 16.5% 16.0% 15.5% 15.0% 14.5% 14.0% Male Female 5

6 Chart 3 shows the dstrbuton of the annutes n the study by amount of monthly ncome, weghted by ncome. As the chart suggests, the average sze s n the frst ncome band. The average sze for males s $487 per month and for females, $423. (However, there may be some small annutes that represent adjustments to the ncome for a pensoner rather than the full amount of the penson; 14% of annutes are for less than $50 per month. If these are gnored, the averages ncrease to $560 and $491 for males and females, respectvely.) Although there are some large annutes ncluded, the chart shows that small annutes predomnate. 50% Chart 3. Dstrbuton of monthly ncome, weghted by ncome 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% 0-1k 1-2k 2-3k 3-4k 4-5k 5-6k 6-7k 7-8k 8-9k 9-10k 10k+ Male Female Table 4 shows the summarzed experence by qunquennal ages, and table 5 by years of experence. Both show data by ncome only. If qcpm were a good ft for the mortalty experence of ths data set, the actual-to-expected (A/E) ratos would be mostly wthn one or two standard devatons of 100%. In fact they are not, ether for males or females, but females are generally closer to 100% than males, although the males are close to 100% for many age ranges. 6

7 Table 4. Summarzed experence by age. Expected on CPM2014prv wth CPM-B. (annualzed ncome n thousands) Exposure Deaths Actual/Expected Standard Devaton Ages Male Female Male Female Male Female Male Female , ,411 1,098 1, % 127.3% 11.4% 13.8% , ,779 4,245 2, % 98.1% 5.7% 7.5% , ,413 13,223 6, % 104.2% 3.5% 4.7% ,188, ,001 25,759 13, % 105.4% 2.5% 3.4% ,484,214 1,099,898 54,581 26, % 106.1% 1.8% 2.6% ,374, ,290 85,311 38, % 93.3% 1.5% 2.3% , ,182 84,722 42, % 92.9% 1.6% 2.1% , ,251 44,434 26, % 95.9% 2.4% 2.2% ,217 47,017 12,151 10, % 93.1% 2.9% 3.0% All 6,897,612 5,620, , , % 97.0% 0.8% 1.0% Table 5. Summarzed experence by year. Expected on CPM2014prv wth CPM-B. (annualzed ncome n thousands) Exposure Deaths Actual/Expected Standard Devaton Year Male Female Male Female Male Female Male Female ,106, ,463 49,747 24, % 97.3% 1.9% 2.6% ,121, ,396 52,277 25, % 95.5% 1.9% 2.5% ,145, ,778 53,822 26, % 94.4% 1.9% 2.5% ,163, ,648 56,320 29, % 98.9% 1.9% 2.4% ,167, ,883 55,133 30, % 97.1% 1.9% 2.4% ,192,076 1,004,074 58,225 32, % 98.3% 1.9% 2.4% All 6,897,612 5,620, , , % 97.0% 0.8% 1.0% Charts 4 (male) and 5 (female) show the actual-to-expected ratos by ncome for each age n the range The blue lne s for graduated ratos, the red damonds show the raw ratos, and the red tck marks show one standard devaton above and below the raw ratos. The graduaton was done wth Whttaker-Henderson; ts purpose s merely to hghlght the trend by age, not to suggest a new mortalty table. It s nterestng to note that these charts suggest a dfferent relatonshp than does table 3. The charts suggest that male A/E ratos have a pronounced downward trend wth ncreasng age. The female A/E ratos show more rapd varaton by age, partcularly n the neghbourhood of age 80. 7

8 110% Chart 4. Male Actual to Expected Ratos 105% 100% 95% 90% Graduated Raw Raw+1sd Raw-1sd 115% Chart 5. Female Actual to Expected Ratos 110% 105% 100% 95% 90% 85% Graduated Raw Raw+1sd Raw-1sd On the surface, t appears that mortalty experence for group annutants s not dramatcally dfferent from qcpm, but there are strong enough dfferences that qcpm 8

9 may not be an approprate bass for group annutes wthout some modfcaton. The analyses n the next secton attempt to look more deeply nto the group annutant experence to seek an explanaton for the dfferences from qcpm. 5 Analyses 5.1 Sngle/Jont The data dstngush three types of annutes: sngle lfe, jont lfe for the prmary annutant, and jont lfe for the secondary annutant. (After the death of the prmary annutant, the secondary annutant, f then lvng, s thereafter consdered the prmary annutant.) Not all companes were able to nclude data for a secondary annutant. It was not always possble to dstngush a sngle survvng annutant of a jont and last survvor annuty from a sngle lfe annutant, partcularly when the nsurance company took on the rsk after the frst death. However, there s no reason to beleve that there s enough msrepresentaton of the annuty type to nvaldate the analyss. Table 6 summarzes the experence by annuty type. Note that males n jont annutes are rarely the secondary annutant. The A/E ratos are lower for secondary annutants than for prmary, but not by enough that one can clam the dfference to be strongly sgnfcant statstcally. There may be an explanaton related to the delay n reportng deaths. The deaths of secondary annutants are often not reported untl the death of the prmary. The actual deaths have already been adjusted for an estmate of IBNR, and the adjustments are larger for secondary annutants than for other categores. However, t may be that the IBNR adjustment s not suffcent for secondary annutants. The contrbutng companes wll be able to refne the estmates over tme. Table 6. Summarzed experence by jont type. Expected on CPM2014prv wth CPM-B. (annualzed ncome n thousands) Exposure Deaths Actual/Expected Standard Devaton Type Male Female Male Female Male Female Male Female Sngle 2,758,820 1,649, ,076 78, % 103.7% 0.9% 1.1% Jont prmary 3,888, , ,026 35, % 95.3% 1.1% 2.6% Jont secondary 249,904 2,971,229 7,421 55, % 89.8% 9.9% 1.9% All 6,897,612 5,620, , , % 97.0% 0.8% 1.0% 5.2 Defned Beneft or not The data dstngushes three types of penson plans: defned beneft (DB), other types (not DB) such as defned contrbuton and group regstered retrement savngs plan (RRSP), and type unknown. Unknown was used by a company when ts admnstratve system dd not have the penson plan type. Table 7 summarzes mortalty experence by penson plan type. Unfortunately there s so much unknown that lttle can be nferred. 9

10 Table 7. Summarzed experence by penson plan type. Expected on CPM2014prv wth CPM-B. (annualzed ncome n thousands) Exposure Deaths Actual/Expected Standard Devaton Plan Male Female Male Female Male Female Male Female DB 2,999,981 2,566, ,600 68, % 96.5% 1.2% 1.6% not DB 915, ,644 47,447 27, % 100.4% 1.7% 2.2% Unknown 2,982,103 2,253, ,476 74, % 96.2% 1.2% 1.6% All 6,897,612 5,620, , , % 97.0% 0.8% 1.0% 5.3 Sze of Income The data underlyng CPM2014 showed a strong correlaton between the level of mortalty and the sze of the penson, so much so that sze adjustment factors were publshed along wth the CPM2014 tables. It s mportant to dscern f the group annuty data show a smlar relatonshp. Table 8 shows the experence by ncome band. Each band s shown as $500 wde n monthly ncome as of Lke the CPM report, the bands are adjusted approxmately based on the average weekly earnngs (AWE). The CPM report suggests that AWE was about 90% durng of what t was n Accordngly, the frst ncome band s actually of monthly amounts of 0 $449, the second $ , etc. Ths adjustment to the bands was used for all years of experence. It was not convenent to reflect the varaton n AWE year by year as was done n CPM. The expected mortalty n table 6, lke all pror tables and charts, s on qcpm wth no sze adjustment. As ndcated by the standard devatons, lttle can be nferred for the hgher bands, but the lower bands show a strong decreasng trend wth ncreasng ncome. 10

11 Table 8. Summarzed experence by monthly ncome range. Expected on CPM2014prv wth CPM-B wthout sze adjustments. (annualzed ncome n thousands) Exposure Deaths Actual/Expected Standard Devaton Income Male Female Male Female Male Female Male Female ,488,235 1,480,484 85,218 57, % 104.1% 0.6% 0.7% ,529,604 1,300,029 82,415 42, % 102.0% 0.9% 1.3% ,321, ,499 56,367 24, % 96.3% 1.5% 2.2% , ,270 32,601 14, % 93.0% 2.2% 3.3% , ,157 22,686 9, % 86.1% 2.8% 4.5% , ,314 11,368 5, % 90.1% 4.7% 6.8% , ,429 7,631 3, % 89.8% 6.2% 8.9% , ,089 4,893 2, % 88.1% 7.3% 11.8% ,790 80,263 3,868 1, % 71.5% 8.8% 13.4% ,691 54,796 3,205 1, % 88.4% 11.2% 17.2% ,045 43,372 3,371 1, % 80.2% 12.4% 18.4% ,089 29,135 1,911 1, % 112.1% 15.3% 23.5% , ,404 9,989 3, % 60.4% 10.3% 17.0% All 6,897,612 5,620, , , % 97.0% 0.8% 1.0% Chart 6 shows the A/E ratos graphcally. The downward slope s qute evdent. The slope for males appears to be slghtly steeper than for females, but the dfference may not be statstcally sgnfcant. 120% Chart 6. A/E ratos by sze band. No sze adjustment. 110% 100% 90% 80% 70% 60% Lower bound of monthly ncome band Male Female 11

12 Table 9 s smlar to table 8 except that the expected mortalty rates have been sze adjusted usng the factors publshed wth CPM2014prv. If the sze adjustment factors were approprate for the group annutant data, the A/E ratos would be very flat n table 9. Table 9. Summarzed experence by monthly ncome range. Expected on CPM2014prv wth CPM-B wth sze adjustments. (annualzed ncome n thousands) Exposure Deaths Actual/Expected Standard Devaton Income Male Female Male Female Male Female Male Female ,488,235 1,480,484 85,218 57, % 97.2% 0.5% 0.7% ,529,604 1,300,029 82,415 42, % 98.1% 0.9% 1.3% ,321, ,499 56,367 24, % 95.7% 1.4% 2.2% , ,270 32,601 14, % 95.8% 2.1% 3.4% , ,157 22,686 9, % 91.5% 2.8% 4.6% , ,314 11,368 5, % 98.1% 4.7% 7.1% , ,429 7,631 3, % 99.5% 6.3% 9.4% , ,089 4,893 2, % 98.7% 7.6% 12.5% ,790 80,263 3,868 1, % 80.1% 9.2% 14.2% ,691 54,796 3,205 1, % 98.8% 11.9% 18.2% ,045 43,372 3,371 1, % 89.4% 13.2% 19.5% ,089 29,135 1,911 1, % 125.5% 16.5% 25.0% , ,404 9,989 3, % 67.2% 11.2% 17.9% All 6,897,612 5,620, , , % 95.8% 0.7% 1.0% Chart 7 shows the A/E ratos of table 9 graphcally. It s clear that the lne s farly flat for females under the $ band. The trend s less clear for males, but a downward slope s suggested. A downward slope would mply that the varaton by sze s greater than that observed n the CPM study. 12

13 120% Chart 7. A/E ratos by sze band. Expected s sze adjusted. 110% 100% 90% 80% 70% 60% Lower bound of monthly ncome band Male Female It s mportant to state that the relatonshp between the level of mortalty and the amount of penson ncome s one of correlaton not causalty. There are many objectons that can be rased to sze adjustments. For example, one would not expect dfferent mortalty for two men who are otherwse very smlar but one worked for 40 years under one penson plan and the other spent 10 years under each of four dfferent plans, but sze adjustments suggest heaver mortalty for the latter. Ths study does not allow us any nsght nto the ndvdual crcumstances of pensoners beyond the demographc data submtted. Nonetheless, t s clearly n the data that mortalty rates go down dramatcally as ncome goes up, at least over the range for whch there s an abundance of data. Actuares may be advsed to be alert to possble applcatons of the sze relatonshp, but whether a sze adjustment s approprate n a partcular case s beyond the scope of ths study. 5.4 Duraton from Retrement In most cases, the contrbutng companes are not able to determne the date of retrement. The recorded date of ssue relates to when the company took on the rsk; n the case of a de-rskng strategy by the plan trustees, that date may be many years after retrement. Accordngly, duraton from the ssue of the annutant s not relevant and s not studed n ths report. 6 Supplementary Informaton For those who wsh to study the data further on ther own, a tool used n developng ths report s avalable for download. That tool s an Excel workbook and a bnary fle 13

14 contanng the summarzed ntercompany data. These two fles and a text fle wth nstallaton nstructons are combned nto a.zp fle whch s avalable here. 7 Subcommttee The Penson and Group Annuty Experence Subcommttee of the Research Commttee s responsble for the content of ths report. The members of the subcommttee that partcpated n the revew of ths report are lsted below. Bob Howard was engaged by the CIA to comple the data for the report on behalf of the subcommttee. Clare Blodeau Steve Bockng Paul Burnell Andrew Kugler Scott McManus (Char) Dana Psanu Myram Roux André Velleux Sylvan Velleux 8 References Annutant Mortalty Subcommttee. Canadan Indvdual Annutant Mortalty Experence Polcy Years to Canadan Insttute of Actuares, February Howard, R.C.W. Canadan Insured Payout Mortalty Table 2014 (CIP2014). Canadan Insttute of Actuares, February Howard, R.C.W. Whttaker-Henderson-Lowre Graduaton, London, D. Graduaton: The Revson of Estmates. Wnsted and Abngdon, CT: Actex Publcatons, Penson Experence Subcommttee. Fnal Report: Canadan Pensoners Mortalty. Canadan Insttute of Actuares, February Penson Experence Subcommttee. CPM2014 Mortalty Tables. Canadan Insttute of Actuares, February