Life Table for the Total Population: United States, 1979-81 wo Q) (3) @) 5) 6) Mm Proportion Average ane Remaining Proportion of —— Persons : Average Stationary Population Number of OF 100,000 Born Alive Years Lived Years of Life Age Interval Number in This Kemaining Period of Interval —_Livingat_ Number Years Lived _ and All Life between Dying during Beginning of Dying during inthe Age Subsequent Two Ages Interval Age Interval Age Interval Interval Age Intervals Interval xtox tt Me 1 a i T é Years 65-66 0.02059 7107 1587 763141273347 66-67 0x16 75.520 1674 74683 1197033 67-68 0.02389 73846 1764 72968 1122350 68-69 0.02585 72082 1864 71150 1049386 69-70 0.02806 70218 1970 69.233 978236 70-71 0.03052 68.248 2083 67206 909 003 oo3315 6 165 2193 65.069 841797 0.0359, 63.972 2299 62823 7716728 0.03882 61673 2304 60476 713.905 oats 39.279 2480 58.039 653.429 0.04507 36.799 2.560 35.500 5953901048 (104867 34239 2640 52919 539.870 995, 0.05274 51599 2721 50 23 486951 mm 005742 38878 2807 47 436 713, 893 0.06277 46071 2891 44626 339 238 845 0.06882 43180 2072 41694 sas 612 798 0.07552 40208 3.036 38 689 302918 753 0.08278 37172 3077 35 634 264 229 7 0.09041 34.095 3088 32553 228 595 670 0.09812 31012 3052 29 486 196 042 632 85-86 0.10725 27 960 2.999 26 461 166.556 86-87 ou 24961 2923 23.500 140 095 87-8 o1z717 22038 2503 20636 116595 88-89 0113708 19235 2637 7917 95 959 89-90 014728 16.598 2444 15376 78.042 90-91 0.15868, 14154 2046 13031 62.666 4s 91-92 017169 11908 2045 10.886 196. 47 92-93 0.18570 9.863 1831 8948 387 393 93-94 020023, 8032 1608 728 29 801 371 91 95 021495 624 1381 5733 22573 351 95-96 0.22976 5043 1159 4463 16 840 334 96-97 3884 915 3412 1237 319 97-98 2939 734 2562 8965 305 98-99 2185 387 1592 6.403, 295 99-100 1598 438 1374 4511 282 Stationary population is @ demographic concept treated in Chapter 19 a Section 3.3 Life TablesLife Table for the Total Population: United States, 1979-81 M a @) aw 6) ) ” Proportion Average ‘Dying Remaining Proportion of = Persons a Average fai ; Stationary Population? Number of at Beginning __Of 100,000 Born Alive Years Lived Years of Life Age Interval Ofage © Number in This ning Period of Interval Living at_ Number Years Lived _ and All Life between Dying during Beginning of Dying during inthe Age Subsequent of Age Two Ages Interval Age Interval Age Interval. Interval Age Intervals Interval xtox+t ae I dy he T, & Years 100-101 20120 1150 335 983 3137 273 101-102 030139 815 245 2 2154 dot 102-108 031089 570 7 $81 1462 257 109-108 o3197 393 126 330 381 250 104-105 032786 267 88 223 si a 105-106 033599 179 6 150 228 106-107 oaa0s 19 n % 28 407-108, 034870 78 ” & 178 106-108 osiss 31 18 2 109-110 035988 x B u ‘Stationary population is a demographic concopt treated in Chapter 18, Several observations about the 1979-81 U.S. Life Table are instructive. Observations: 1. Approximately 1% of a survivorship group of newborns would be expected to die in the first year of life 2. It would be expected that about 77% of a group of newborns would survive to age 65. The maximum number of deaths within a group would be expected to occur between ages 83 and 84 4. For human lives, there have been few observations of age-at-death beyond 110. Consequently, it is often assumed that there is an age « such that s(x) > 0 for x < ©, and s(x) = 0 for x = w. The age a, if assumed, is called the limiting age. ‘The limiting age for this table is not defined. It is clear that there is a positive probability of survival to age 110, but the table does not indicate the age w. 5. Local minimums in the expected number of deaths occur around ages 11 and 27 and a local maximum around age 24. 6. Although the values of I, have been rounded to integers, there is no compelling reason, according to (3.3.1), to do so. A display such as Table 3.3.1 is the conventional method for describing the dis- tribution of age-at-death. Alternatively, an s.f. can be described in analytic form such as s(x) = e-*, ¢ > 0, x = 0. However, most studies of human mortality for Chapter 3_ Survival Distributions and Life Tables @insurance purposes use the representation s(x) = I, /l, as illustrated in Table 3.3.1. Since 100,000 s(x) is displayed for only integer values of x, there is a need to inter- polate in evaluating s(x) for noninteger values. This is the subject of Section 3.6. kz ~LDZ = = (On the basis of Table 3.3.1, evaluate the probability that (20) will a. Live to 100 b. Die before 70 c. Die in the tenth decade of life. Solution: (100) _ op _ 1,150 _ 520) ~ Iny 97,741 ~ O18 {s(20) = s(70)1 68,248 _ ‘s(20) : 97,741 tute {80) = s(100)] _ (lo = hoo) _ (14,154 — 1,150) _ 3@0) 97,741 ae Insight into life table functions can be obtained by studying Figures 3.3.1, 33.2, and 3.3.3. These are drawn to be representative of current human mortality and are not taken directly from Table 3.3.1 Graph of p(x) oot 0.03 0.02 ao oO 10 40 Age ‘Section 3.3_Life TablesIn Figure 3.3.1 note two features: + The force of mortality is positive and the requirement f n(x) dx = 2 appears satisfied. (See Table 3.2.1.) + The force of mortality starts out rather large and then drops to a minimum around age 10. == Graph of J, w(x) a, 20 1500 1000 soo er «0 3% Age Graph of /, i 100 0 3 5 wo Z a ~ 40 Age ow 70 0 Chapter 3 Survival Distributions and Life Tables 65In Figures 3.3.2 and 3.3.3 note the following: + The function |, (x) is proportional to the p.df. of the age-at-death of a new- born. Since 1, (x) is the expected density of deaths at age x, under the random survivorship group idea, the graph of I, (x) is called the curve of deaths. ‘There is a local minimum of I, x(x) at about age 10. The mode of the distribution of deaths—the age at which the maximum of the curve of deaths occurs—is, around age 80. + The function I, is proportional to the survival function s(x). It can also be in- terpreted as the expected number living at age x out of an initial group of size ly + Local extreme points of 1, 4(2) correspond to points of inflection of I, since d d d eh mo) i(- 4u)=- 3.4 The Deterministic Survivorship Group ‘We proceed now to a second, and nonprobabilisti, interpretation of the life table. This is ronted mathematically in the concept of decrement (negative growth) rates. As such, it is related to growth-rate applications in biology and economics. It is deterministic in nature and leads to the concept of a deterministic survivorship group or cohort. A deterministic survivorship group, as represented by a life table, has the fol- lowing characteristics: + The group initially consists of J, lives age 0. + The members of the group are subject, at each age of their lives, to effective annual rates of mortality (decrement) specified by the values of q, in the life table. + The group is closed. No further entrants are allowed beyond the initial [,. The only decreases come as a result of the effective annual rates of mortality (dectement), From these characteristics it follows that the progress of the group is determined by 4 = WL = 9a) = Io ~ doy h=hd-4)=h hea da th- Sd, 24, =h\i-=—} = ha - 9) G41) 66 Section 3.4 The Deterministic Survivorship Groupwhere /, is the number of lives attaining age x in the survivorship group. This chain of equalities, generated by a value I, called the radix and a set of q, values, can be rewritten as 1 = by Por b= hp = (ly Po) Par (3.4.2) ‘There is an analogy between the deterministic survivorship group and the model for compound interest. Table 3.4.1 is designed to summarize some of this parallelism. Related Concepts of the Mathematics of Compound Interest and of Deterministic Survivorship Groups ee oe Al) = Size of fund at time f, time ize of group at age x, age {increment) (decrement) 7 A(t) 2 Effective n-year rate of interest, Effective n-year rate of mortality, 4. Alt + n) = AW) Lt Hi eeara nimi a Force of interest at time f Force of mortality at age x ‘A(t + At) — A(t) = besa os tm [ADA wt) — tm (ee Pan A(t) dt “There iso universally aocepied symbol for an effective r-yoar rate of interest, ‘The headings of the 7, and ,d, columns in Table 33.1 refer to the deterministic survivorship group interpretation. Although the mathematical foundations of the random survivorship group and the deterministic survivorship group are different, the resulting functions q,, ,, and d, have the same mathematical properties and subsequent analysis. The random survivorship group concept has the advantage of allowing for the full use of probability theory. The deterministic survivorship group Chapter 3_ Survival Distributions and Life Tables ais conceptually simple and easy to apply but does not take account of random variation in the number of survivors. 3.5 Other Life Table Characteristics In this section we derive expressions for some commonly used characteristics of the distributions of T(x) and K(x) and introduce a general method for computing several of these characteristics. 3.5.1 Characteristics The expected value of T(x), denoted by é,, is called the complete-expectation- of-life. By definition and an integration by parts, we have a= eireo= [tp mee +e at faces = tools + [Fede 651) The existence of E[T(x)] implies the lim t(—p,) = 0. Thus é, Coa 50 The complete-expectation-of life at various ages is often used to compare levels of public health among different populations. A similar integration by parts yields equivalent expressions for E[T(x)"} rire = [p,m +) at =2[ baat (53) This result is useful in the calculation of Var [T(x)] by Var[T(x)] = E[Tey"] — E[Te)P =2[tpa—@ 5.4) In these developments, we assume that E[7(x)] and E[T(x)?] exist. One can construct s.f/s such as s(x) = (1 + x)? where this would not be true. Other characteristics of the distribution of T(x) can be determined. The median future lifetime of (x), to be denoted by m(x), can be found by solving cy ‘Section 3.5. Other Life Table CharacteristicsPr[T(x) > m(x)] or sty + m(x)] 1 a) 2 (3.5.5) for m(x). In particular, m(0) is given by solving s[m(0)] = 1/2. We can also find the mode of the distribution of T(x) by locating the value of t that yields a maximum value of ,p, w(x + ) ‘The expected value of K(x) is denoted by e, and is called the curtate-expectation- of-life. By definition and use of summation by parts as described in Appendix 5, we have 6 = BLK] = 3 bapa dost = & kA.) = Kwa + & aks 5.6) Again, the existence of E[K(x)] implies the lim k(—,p,) = 0. Thus, with a change of the summation variable, ~ = > tp (57) Following the outline used for the continuous model and using summation by parts, we have EIK@y] = 3 Paps doe BF Ace) = ECs + D AMP. (58) ‘The existence of E[K(x)'] implies lim }°(—,p,) = 0. With a change of the summation variable, a (I= Fk + Dips = Dk Yip, 59) a Now, Var(K) = E[K?] E[KP = > (2k -1)p, - 2 (3.5.10) Chapter 3_ Survival Distributions and Life Tables ooTo complete the discussion of some of the entries in Table 3.3.1, we must define additional functions. The symbol L, denotes the total expected number of years lived between ages x and x + 1 by survivors of the initial group of Jy lives. We have -[ thar we + t) dt + hay (3.5.11) where the integral counts the years lived of those who die between ages x and x + 1, and the term I,., counts the years lived between ages x and x + 1 by those who survive to age x + 1. Integration by parts yields Ly -[ tdless + leat the + [eed + fa -f esy dt (35:12) f The function L, is also used in defining the central-death-rate over the interval from x to x + 1, denoted by mt, where fgeecoe Piaae ‘An application of this function is found in Chapter 10. (35.13) The definitions for m, and L, can be extended to age intervals of length other than one: = frame + de + ny (3.5.14) (3.5.15) For the random survivorship group, ,L, is the total expected number of years lived between ages x and x +» hy the survivors of the initial group of ly lives and ,m, is the average death rate experienced by this group over the interval (x, x + n). The symbol T, denotes the total number of years lived beyond age x by the survivorship group with ly initial members. We have 70 ‘Section 3.5 Other Life Table Characteristics[letate roa (3.5.16) The final expression can be interpreted as the integral of the total time lived be- tween ages x + t and x + + dt by the /,,, lives who survive to that age interval. We also recognize T, as the limit of ,L., as n goes to infinity. ‘The average number of years of future lifetime of the /, survivors of the group, at age x is given by tr. f eg dt Ie -f peat = 8 as determined in (3.5.1) and (3.5.2). We can express the average number of years lived between x and x + 1 by the |, survivors at age x as Lael ht = f Pot 5.17) This function is the n-year temporary complete life expectancy of (x) and is denoted by @,q. (See Exercise 3.16.) A final function, related to the interpretation of the life table developed in this section, is the average number of years lived between ages x and x + 1 by those of the survivorship group who die between those ages. This function is denoted by a(x) and is defined by Chapter 3 Survival Distributions and Life Tables vi1 i tls, we +) dt Sa ae (3.5.18) nQr + #) dt For the probabilistic view of the life table, we would have Jeane +o ae a(x) = E(T|T < 1]. f Pewee + Ot If we assume that Lame t thdt=ddt Ostet, that is, if deaths are uniformly distributed in the year of age, we have ; 1 alae This is the usual approximation for a(x), except for young and old years of age where Figure 3.3.2 shows that the assumption may be inappropriate. k= = ~~ Show that (x) Le + TL a be and Solution: From (3.5.11), (3.5.12), and (3.5.18), we have or The formula can be justified by using the trapezoidal rule for approximate integration on (35.12). v Key life table terminology, defined in Sections 3.3-3.5, is summarized as part of Table 3.9.1 in Section 3.9. R ‘Section 3.5 Other Life Table Characteristics3.5.2 Recursion Formulas Example 3.5.1 illustrates the use of a numerical analysis technique to evaluate a life table characteristic. The trapezoidal rule for approximate integration is used. The calculation of complete and curtate expectations-of-life can be used to illustrate another computational tool called recursion formulas. The application of recursion formulas in this book typically involves one of two forms: Backward Recursion Formula u(x) = e(x) + d(x) u(x + 1) 3.5.19) or Forward Recursion Formula -- 5 Met = =F + ay ue (3.5.20) The variable x is usually a non-negative integer. To evaluate a function u(x), for a domain of non-negative integer values of x, we need to have available values of c(x) and d(x) and a starting value of u(x). This procedure is used in subsequent chapters and is illustrated in Table 3.5.1 where backward recursion formulas are developed to compute e, and Ee Backward Recursion Formulas for e, and 2, 1. Basie equation a= San a= [pas 2 Separate te + Su, _ ds + _ ids *Gewabem = etm Sin ke Late Lama eee Pot Paes a f Peds + py Bess ry BTR woy=tea = fant 5. Starting value? u(o) (a "The integral 6) = J} ,p,ds ean be evaluated using the trapezoidal rule as tx) = (1 + p,)/2. From Secton 3.3.1 we have sx) = 0, x = w, and s(x) > 0, x Oc>1x=0 (1825) Makeham = A+BO —expl-Ax— mic!) B>0,A=-Be> 1x20 (1860), Weibull ke" exp(—ur"™) k>0,n>0,x20 (1939) Note + The special symbols are defined as B k ye 20) * Gompertz’s law is a special case of Makeham’s law with A = 0. + Ifc = 1 in Gompertz’s and Makeham’s laws. the exponential (constant force) distribution results. + In connection with Makeham’s law, the constant A has been interpreted as capturing the accident hazard, and the term Bc" as capturing the hazard of aging. ‘The entries in the s(x) column of Table 3.7.1 were obtained by substituting into (3.2.16). For example, for Makeham’s law, we have expl-Si(A + Beds] exp [=a -B «| s(x) exp[-Ax — m (= 1)] where m = B/log c ‘Two objectives governed the development of a mortality table for computational purposes in the examples and exercises of this book. One objective was to have mortality rates in the middle of the range of variation for groups, such variation caused by factors such as residence, gender, insured status, annuity status, marital status, and occupation. The second objective was to have a Makeham law at most ages to illustrate how calculations for multiple lives can be performed. The Ilustrative Life Table in Appendix 2A is based on the Makeham law for ages 13 and greater, 1,000 (x) = 0.7 + 0.05 (10° 7.1) The calculations of the basic functions q,, |,, and d, from (3.7.1) were all done directly from (3.7.1) instead of calculating 1, and d, from the truncated values of q,. 7 ‘Section 3.7 Some Analytical Laws of MortalityIt was found that the latter choice would make little difference in the applications. It should be kept in mind that the Illustrative Life Table, as its name implies, is for illustrative purposes only. ——————— 3.8 Select and Ultimate Tables In Section 3.2 we discussed how ,p, [the probability that (x) will survive to age x + #] might be interpreted in two ways. The first interpretation was that the probability can be evaluated by a survival function appropriate for newborns, un- der the single hypothesis that the newborn has survived to age x. This interpreta- tion has been the basis of the notation and development of the formulas. The second interpretation was that additional knowledge available about the life at age x might make the original survival function inappropriate for evaluating probability state- ments about the future lifetime of (x). For example, the life might have been un- derwritten and accepted for life insurance at age x. This information would lead us to believe that (2)'s future-lifetime distribution is different from what we might otherwise assume for lives age x. As a second example, the life might have become disabled at age x. This information would lead us to believe that the future-lifetime distribution for (2) is different from that of those not disabled at age x. In these two illustrations, a special force of mortality that incorporates the particular infor- mation available at age x would be preferred. Without this particular information for (x), the form of mortality at duration f would be a function of only the attained age x + f, denoted in the previous sections by u(x + #). Given the additional formation at x, the force of mortality at x + t is a function of this information at x and duration f. Its notation will be 1,(t), showing separately the age, x, at which the additional information was available, and the duration, f. The additional infor- mation is usually not explicit in the notation but is conveyed by the context. In other words, the complete model for such lives is a set of survival functions in- cluding one for each age at which information is available on issue of insurance, disability, and so on. This set of survival functions can be thought of as a function of two variables. One variable is the age at selection (e.g., at policy issue or the onset of disability), [x], and the second variable is the duration since policy issue or duration since selection, t. Then each of the usual life table functions associated with this bivariate survival function is a two-dimensional array on [x] and . Note the bracket notation to indicate which variable identifies the age of selection. When the select status can be inferred from the force of mortality, the bracket notation will be suppressed to reduce the clutter of the symbols. The schematic diagram in Figure 3.8.1 illustrates these ideas. For instance, sup- pose some special information is available about a group of lives age 30. Perhaps they have been accepted for life insurance or pethaps they have become disabled. A special life table can be built for these lives. The conditional probability of death in each year of duration would be denoted by qr). i = 0, 1, 2,-.., and would be entered on the first row of Figure 3.8.1. The subscript reflects the bivariate nature of this function with the bracketed thirty, [30], denoting that the survival function in the first row is conditional on special information available at age 30. The second row of Figure 3.8.1 would contain the probabilities of death for lives on which the Chapter 3 Survival Distributions and Life Tables 73Select, Ultimate, and Aggregate Mortality, 15-Year Select Period Year Following Selection ky) Ist 2nd 15th 16th 17th 40 Cie Tpo+14 Fya0j+ 15) F304 + (30) oe eee oe or 44s ‘or 446 ¢ ou, init pauper Fouts Gin I ee cere Age fr 7 a Selection + A321 Fisaier Fister Fzj+15. Fr32}+10 [sq] |} -- -+|-----o-- -—t --¢-4 ore | Ate ; Apa F341 As3je18 yx} 15° Apois te [33] 7 F341 Fae Tsai14 ah 5 Apap 6: [34] ~-—-4-----4-- 0) -- 4 Time in Years since apecion, oo 1 2 4 1b 16 7 Path followed by a survivorship group selected at (x Links (age at selection and duration) cells after a 15-year select period into attained age groups. soeKK Alternative path followed by survivorship groups after 15-year select period. These probabilities constitute an ultimate mortality table. Notes: 1 In biostatistics the select table Indox [] may not be age. For example, in cancer research, [x] could be @ Classification index that depends on the size an location of the tumor, and time folowing selection would be ‘measured trom the time of diagnos, 2, Unimate moray, rouowing a TD-year select period, tor age [x] + 15, would be estimated by using observations ftom all cos identified by fe~ 1+ 15 + /, for] — 0, 1, 2... . Therefore, dyes = de.as 8 estimated by a Weighted average of mortality estimates from several diferent selection groupe lf the effect of selection is not smal, the resulting estimate wil be influenced by the amount of data from the various eels, 80 ‘Section 3.8 Select and Ultimate Tablesspecial information became available at age 31. In actuarial science such a two- dimensional life table is called a select life table. The impact of selection on the distribution of time-until-death, T, may diminish following selection. Beyond this time period the q's at equal attained ages would be essentially equal regardless of the ages at selection. More precisely, if there is a smallest integer r such that |fj.)., — 9je irs i8 less than some small positive con- stant for all ages of selection [x] and for all j > 0, it would be economical to construct a set of select-and-ultimate tables by truncation of the two-dimensional array after the (r + 1) column, For durations beyond r we would use Yo-fierri = Fare 7 > 0. The first r years of duration comprise the select period. The resulting array remains a set of life tables, one for each age at selection. For a single age at selection, the life table entries are horizontal during the select period and then vertical during the ultimate period. This is shown in Figure 3.8.1 by the arrows. The Society of Actuaries mortality studies of lives who were issued individual life insurance on a standard basis use a 15-year select period as illustrated in Figure 3.8.1; that is, it is accepted that Fe-nvrsey = Iurvs jf > 0. Beyond the select period, the probabilities of death are subscripted by attained age only; that is, qjy-yyy+1) is written as q,,,. For instance, with r= 15, qjoj.is ANd qsjsa0 would both be written as qys A life table in which the functions are given only for attained ages is called an aggregate table, Table 3.3.1, for instance. The last column in a select-and-ultimate lable is a special aggregate lable that is usually referred to as an ultimate table, to reflect the select table setting. Table 3.8.1 contains mortality probabilities and corresponding values of the I.)4. function, as given in the Permanent Assurances, Females, 1979-82, Table, published by the Institute of Actuaries and the Faculty of Actuaries; it is denoted as the Excerpt from the AF80 Select-and-Ultimate Table @ @ @ @ o 6 a fax] 1000 gy 1,000 gysy 1,000 gegen lg x42 3002220330 0422 9:906.7380 9 904.5387 9.901.2702 32 310234 = 0352 0459 9.902.8941 9 900.5769 9 8970019 33 3 0250 «0377 0500 9.898.757 9 896.2800 9 892549134 330269 = 0407 0545 9 89429039 891.6287 9 887.6028 35 340291 441 01596 9 889.4519 9 886.5741 9 8822141 36 Chapter 3 Survival Distributions and Life Tables 81AF80 Table. This table has a 2-year select period and is easier to use for illustrative purposes than tables with a 15-year select period such as the Basic Tables, pub- lished by the Society of Actuaries. In Table 3.8.1 we observe three mortality probabilities for age 32, namely, joa) = 0.000250 < qynysr = 0.000852 < qyy = 0.000422. The order among these probabilities is plausible since mortality should be lower for lives immediately after acceptance for life insurance. Column (3) can be viewed as providing ultimate mortality probabilities. Given the 1-year mortality rates of a select-and-ultimate table, the construction of the corresponding select-and-ultimate life table (survival functions) is started with the ultimate portion. Formulas such as (3.4.1) can be used, which would yield a set of values of I,,, = yj. Where ris the length of the select period. We would then complete the select segments by using the relation | Ngee = TERE k wr Plater working from duration r — 1 down to 0 _—L = =——ti‘_‘OSC;COT.. Use Table 3.8.1 to evaluate a Prob. Pray © disney Solution: Formulas developed earlier in this chapter can be adapted to select-and-ultimate tables yielding a Pow) = a T= Sere — 0.99045 DB. sProo = : SaReSe — 0.99807 © a= k= = SSO = 9857009 ayo stir = Tore a 2200S = oer ins = 0.00131. vy Section 3.8 Select and Ultimate TablesEe Chapter 3 Concepts Symbol ‘Name or Description of the Concept @ Notation for a life age x te] Age, or other status, at selection x ‘Age at death, a random variable Te) Future lifetime of (x), equals X ~ x Kay CCurtate-future-lifetime of (x), equals the integer part of T(x) 5G) Future lifetime of (x) within the year of death, equals T(x) ~ K(x) st) Survival function, equal to the probability that a newborn will live to at least x na Force of mortality at age x in an aggregate life table nO Force of mortality at attained age x + f given selection at age x a Probability that (x) dies within t years Pe Probability that (x) survives at least t years ads Probability that (x) dies between t and t + w years z Complete expectation of life for (x), equals E[T(2)] 6 Curtate expectation of life for (x), equals E[K(x)] Fe) Cohort’s number of survivors to age x, a random variable 2 Colrt’s number of deaths between ages x ad x + 1 Expected number of survivors at age x, equals E£(x)] Expectedl number of deaths between ages x and x +n, equals E],2,] Expected number of years lived between ages x and x +n by survi- vvors to age x of the initial group of fy lives T, Expected number of years lived beyond age x by the survivors to age x of the initial group of fy lives Central death rate over the interval (x, x + 1) ® Omega, the limiting age of a life table 3.9 Notes and References Table 3.9.1 summarizes this chapter's new concepts with their names, symbols, and descriptions. Life tables are a cornerstone of actuarial science. Consequently they are extensively discussed in several English-language textbooks on life contingencies: + King (1902) + Spurgeon (1932) + Jordan (1967) + Hooker and Longley-Cook (1953) + Neill (1977). These have been used in actuarial education. In addition, life tables are used by biostatisticians. An exposition of this latter approach is given by Chiang (1968) and Elandt-Johnson and Johnson (1980). The deterministic rate function interpretation is discussed by Allen (1907). London (1988) summarizes several methods for es mating life tables from data. Chapter 3 Survival Distributions and Life TablesThe historically important analytic forms for survival functions are referred to in Table 3.6. Brillinger (1961) provides an argument for certain analytic forms from the viewpoint of statistical life testing. Tenenbein and Vanderhoof (1980) restate the case for analytic laws of mortality and develop formulas for select mortality Balducci’s (1921) contribution was preceded by a remarkable set of papers by Witt- stein (1873). Wittstein’s papers were published first in German and translated into English by T. B. Sprague. Some of the methods for evaluating probabilities for fractional ages are reviewed by Mereu (1961) and in Batten’s textbook on mortality estimation (1978) (see alsu Seal’s 1977 historical review). Discussions of the length of the select period for various types of insurance selection procedures have a long history, for example, Williamson (1942), Thompson (1934), and Jenkins (1943). The Society of Actuaries 1975-80 Basic Tables use a 15-year select period and are pub- lished in TSA Reports 1982. International Actuarial Notation is outlined in TASA 48 (0947). We planned to use the 1989-91 USS. Life Table for illustrative purposes in Table 3.2.1, but this plan was not realized because the life tables based on the 1990 US. Census were not completed when this chapter was revised. Exercises Section 3.2 3.1. Using the ideas summarized in Table 3.2.1, complete the entries below. s@) _ =) fel) rr) tan x, 0 1 Lox=0 3.2. Confirm that each of the following functions can serve as a force of mortality. Show the corresponding survival function. In each case x = 0. a Be B>0 — c>1 (Gompertz) b. kx n>0 — k>0 (Weibull) cabtay! a>0 — b>0 (Pareto) 3.3. Confirm that the following can serve as a survival function. Show the corre- sponding 44(x), fi(), and F(x). sa)se® x20. 3.4. State why each of the following functions cannot serve in the role indicated by the symbol: aua)=( +x? x20 84 Exercises:35. — x/100, 0 = x = 100, calculate a. w(x) b. F(x) © fe) d. Pr (10 < X < 40). 3.6. Given the survival function of Exercise 3.5, determine the survival function, force of mortality, and p.d.f. of the future lifetime of (40). 37. If s(x) = [1 — (/100)}"?, 0 = x = 100, evaluate a Ps be iste © ssa6 4. n@6) —e. E[T(36)] 3.8. Confirm that qo = ~As(k), and that > yiqo 39, If w(x) = 0.001 for 20 = x = 25, evaluate a243y Sections 3.3, 3.4 3.10. If the survival times of 10 lives in a survivorship group are independent with survival defined in Table 3.3.1, exhibit the pf. of £(65) and the mean and variance of £(65). 3.1, If s(x) = 1 — x/12,0 4 x = 12, ly = 9, and the survival times are independent, then (ys 32 39) is known to have a multinomial distribution. Calculate a. The expected value of each random variable b. The variance of each random variable c. The coefficient of correlation between each pair of random variables. 3.12. On the basis of Table 3.3.1, a. Compare the values of 4g and b. Evaluate the probability that (25) will die between ages 80 and 85. 3.13. Given that [,,, is strictly decreasing in the interval 0 = f = 1, show that a. If I,,, is concave down, then 4, > u(x) b. If L,., is concave up, then q, < w(x). 3.14. Show that a da a. Fle ws) <0 when w(a) < uA) d d bo Fhe mee) = 0 when 7 w(a) = w7(9) a d 7 ©. Fk m@) > 0 when 5 w(x) > w7(0), Chapter 3 Survival Distributions and Life Tables 353.15. Consider a random survivorship group consisting of two subgroups: (1) the survivors of 1,600 persons joining at birth; (2) the survivors of 540 persons joining at age 10. An excerpt from the appropriate mortality table for both subgroups follows: x 1 0 40 10 39 70 26 If Y, and Y; are the numbers of survivors to age 70 out of subgroups (1) and 2), respectively, estimate a number ¢ such that Pr(Y, + Y, > c) = 0.05. As- sume the lives are independent and ignore half-unit corrections. Section 3.5 3.16. Let the random variable T(x) @) 0 0, calculate a. 8 = EIT] b. Var(T) c. median (T) d. The mode of the distribution of T. If w(x + th = a. peux +8) bv &. [Hint: Recall, from the study of probability, that (1 /-V2m) e~®/? is the p.df. for the standard normal distribution.| f= 0, calculate If the random variable T(x) has df. given by t o0. Calculate ane) b8 Section 3.8 3.36. Using Table 3.8.1, calculate a aMfsayea Be Posuyst 3.37. The quantity 1 ~ feet = Kx, has been called the index of selection. When it is close to 0, the indication that selection has worn off. From Table 3.8.1, calculate the index for x = 32, k=0,1 3.38. The force of mortality for a life selected at age (2) is given by wx(t) = YOR), 1 > 0. In this formula (2) is the standard force of mortality. The symbol x denotes a vector of numerical information about the life at the time of selec- tion. This information would include the age and other classification infor- mation. It is required that (x) > 0 and ¥(x,) = 1, where x) denotes standard information. Show that the select survival function is Poa = (raid and the p.df. of T(x), the random variable time-until-death given the infor- mation x, is —¥(x) ,Pfay(Ptq)""” |, where ;pgy is the derivative with respect to £ of Pig): This is called a proportional hazard model. Miscellaneous 3.39. A life at age 50 is subject to an extra hazard during the year of age 50 to 51. If the standard probability of death from age 50 to 51 is 0.006, and if the extra risk may be expressed by an addition to the standard force of mortality that uniformly from 0.03 at the beginning of year to 0 at the end of the year, calculate the probability that the life will survive to age 51 decreas 3.40. If the force of mortality y(t), 0 = t = 1, changes to H(t) ~ ¢ where ¢ is a positive constant, find the value of c for which the probability that (x) will die within a year will be halved. Express the answer in terms of jy 341. From a standard mortality table, a second table is prepared by doubling the force of mortality of the standard table. Is the rate of mortality, qi, at any Chapter 3 Survival Distributions and Life Tablesgiven age under the new table, more than double, exactly double, or less than. double the mortality rate, q,, of the standard table? 3.42. If w(x) = Bc’, ¢ > 1, show that the function J, (x) has its maximum at age Xy where u(x.) = log c. [Hint: This exercise makes use of Exercise 3.14.] 3.43. Assume p(x) 4. Calculate the survival fanction, s() b. Verify that the mode of the distribution of X, the age-at-death, is given by 1, ~ Walog.o) = log A a loge 344. If w(x) = 9 10 for 40 < x < 100, calculate 00-x 20-zx 3 auP a0 b. The mode of the distribution of X, the age-at-death 3.45. a. Show that, under the uniform distribution of deaths assumption, Ie My ™ Tray, 4 TEC /Dm, b. Calculate m, in terms of q, under the constant force assumption. c. Calculate m, in terms of q, under the hyperbolic assumption d. If l, = 100 — x for 0 = x = 100, calculate yey) where i heey wo + 1) dt I 3.46, Show that K and S are independent if and only if the expression a = at fine Teor does not depend on k for 0 = 5 =1 Computing Exercises: These are the first in a series of exercises that involve sufficient computation to make it worthwhile to use a computer. The series will continue in the following chapters, and in each exercise it is assumed that the results of previous exercises are available. For example, in Exercise 3.47, you are asked to set up a life table that will then be used in risk analysis in Chapters 4 and 5. 3.47. Using spreadsheet or other mathematical software, set up an object that will accept input values for the Makeham law parameters and then calculate 90 Brercses3.48, 3.49, 3.50. 351, 3.52. 3.53. 354, and display the values of p, and q, for ages 0 to 140. As a check on your output, input the parameter values given in (3.7.1) and compare your 4, values with those for x = 13, 14, . . . in the Illustrative Life Table in Ap- pendix 2A. We will refer to this computing object as your Illustrative Life Table. When the Makeham parameter values are not stated, those of (3.7.1) are implied. [Remark: With a Makeham Table, s(x) > 0 for all x > 0, 50 o does not exist as defined in Section 3.3.1. For the parameter values of the Illustrative Life Table, q,.) is zero to eight decimal places; thus we choose © ~ 140 for our Illustrative Life Table, that is, Table 2A.] In your Illustrative Life Table use the forward recursion formula I... = (p,)(0,) and initial value /,, = 96,807.88 to calculate the /, values of Table 2A. [Remark: ‘The Makeham law was not realistic for ages less than 13, so the Illustrative Life Table is a blend of some ad hoc values from 0 through 12 and the Makeham law table from age 13 up.] Illustrate the result of Exercise 3.41 by doubling the A and B parameter values in your Illustrative Life Table. Use the backward recursion formula of Table 3.5.1 to calculate values of en your Illustrative Life Table for ages 13 to 140. Compare the values of e, at x = 20, 40, 60, 80, and 100 in your Illustrative Life Table with those found when the force of mortality is doubled. Use the backward recursion formula of Table 3.5.1 and the trapezoidal rule to calculate values of @, in your Illustrative Life Table for ages 13 to 110. Verify the following backward recursion formula for the temporary curtate life expectancy to age y: Cxgm=a = Px t Px rigmaoy forx=0,1,..-,y— 1 Determine an appropriate starting value for use with this formula. For your Illustrative Life Table calculate the curtate temporary life expectancy up to age 45 for ages 13 to 44. Verify the following backward recursion formula for the 1-year temporary curtate life expectancy: cA = Px (1 ~ yPasa) + Pr @rvray forx =0,1,...,0— 1. Determine an appropriate starting value for use with this formula. For your Illustrative Life Table calculate the 10-year temporary curtate life expectancy for ages 13 to 139. . “Look up” eys5 in your Illustrative Life Table. [Hint: Since the c(x) term in the relation in Exercise 3.53 does not depend on 1, it may be more efficient to view e;s3 as a curtate temporary life expectancy to age 40 for (15).] Chapter 3 Survival Distributions and Life TablesLIFE INSURANCE 4.1 Introduction We have stated that insurance systems are established to reduce the adverse financial impact of some types of random events. Within these systems individuals and organizations adopt utility models to represent preferences, stochastic models to represent uncertain financial impact, and economic principles to guide pricing Agreements are reached after analyses of these models. In Chapter 2 we developed an elementary model for the financial impact of random events in which the occurrence and the size of impact are both uncertain. In that model, the policy term is assumed to be sufficiently short so the uncertainty of investment income from a random payment time could be ignored. In this chapter we develop models for life insurances designed to reduce the financial impact of the random event of untimely death. Due to the long-term nature of these insurances, the amount of investment earnings, up to the time of payment, provides a significant element of uncertainty. This uncertainty lias (wo causes: the unknown rate of earnings over, and the unknown length of, the in- vestment period. A probability distribution is used to model the uncertainty in regards to the investment period throughout this book. In this chapter a determin- istic model is used for the unknown investment earnings, and in Chapter 21 sto- chastic models for this uncertainty are discussed. In other words, our model will be built in terms of functions of T, the insured’s future-lifetime random variable. While everything in this chapter will be stated as insurances on human lives, the ideas would be the same for other objects such as equipment, machines, loans, and business ventures. In fact, the general model is useful in any situation where the size and time of a financial impact can be expressed solely in terms of the time of the random event. Chapter 4 Life Insurance4.2 Insurances Payable at the Moment of Death In this chapter, the amount and the time of payment of a life insurance benefit depend only on the length of the interval from the issue of the insurance to the death of the insured, Our model will be developed with a benefit function, b,, and a discount function, v,. In our model, v, is the interest discount factor from the time of payment back to the time of policy issue, and t is the length of the interval from issue to death. In the case of endowments, covered in this section, f can be greater than or equal to the length of the interval trom issue to payment For the discount function we assume that the underlying force of interest is deterministic; that is, the model does not include a probability distribution for the force of interest. Moreover, we usually show the simple formulas resulting from the assumption of a constant, as well as a deterministic, force of interest. We define the present-value function, z,, by 2 = by, (4.2.1) Thus, 2, is the present value, at policy issue, of the benefit payment. The elapsed time from policy issue to the death of the insured is the insured’s future-lifetime random variable, T = T(x), defined in Section 3.2.2. Thus, the present value, at policy issue, of the benefit payment is the random variable zy. Unless the context requires a more elaborate symbol, we denote this random variable by Z and base the model for the insurance on the equation Z = byw, (4.2.2) The random variable Z is an example of a claim random variable and, as such, of an X, term in the sum of the individual risk model, as defined by (2.1.1). This model is used in later sections when we consider applications involving portfolios. We now tum to the development of the probability model for 7. The first step in our analysis of a life insurance will be to define b, and v,. The next step is to determine some characteristics of the probability distribution of Z that are consequences of an assumed distribution for T, and we work through these steps for several conventional insurances. A summary is provided in Table 4.2.1 on page 109. —— 4.2.1 Level Benefit Insurance An t-year term life insurance provides for a payment only if the insured dies within the 1-year term of an insurance commencing at issue. If a unit is payable at the moment of death of (x), then oa ‘Section 4.2 Tnsurances Payable at the Moment of DeathT=n T>n. These definitions use three conventions. First, since the future lifetime is a non- negative variable, we define b,, v,, and Z only on non-negative values, Second, for a value where b; is 0, the value of v; is irrelevant. At these values of f, we adopt definitions of v, by convenience. Third, unless stated otherwise, the force of interest is assumed to be constant, The expectation of the present-value random variable, Z, is called the actuarial present value of the insurance. The reader will find that the expectation of the present value of a set of payments contingent on the occurrence of a set of events is referred to by different terms in different actuarial contexts. In Chapter 1, the expected loss was called the pure premium. This vocabulary is commonly used in property-liability insurance. A more exact term, but more cumbersome, would be expectation of the present value of the payments, We denote actuarial present val- ues by their symbols according to the International Actuarial Notation (see Appen- dix 4). The principal symbol for the actuarial present value of an insurance paying a unit benefit is A. The subscript includes the age of the insured life at the time of the calculation. How this age is displayed depends upon the form of the mortality assumption. For the actuarial present value of an insurance on (40), the age might be displayed as [40], 40, or [20] + 20, for example. As in Section 3.8, the bracket indicates selection at that age and hence the use of a select table commencing at that age. The unbracketed age indicates the use of an aggregate or ultimate table. Thus [20] + 20 indicates the calculation for a 40 year old on the basis of a select table commencing at age 20. The actuarial present value for the 1-year term insurance with a unit payable at the moment of death of (x), E{ZJ, is denoted by Al;. This can be calculated by recognizing Z as a function of T so that E[Z] = Elz;]. Then we use the pdf. of T to obtain E[Z] = Elz,] = f 2% frld) dt = f vt pe mt) dt (423) The j-th moment of the distribution of Z can be found by Bz = [ey nna = fey, mato at Chapter 4 Life insurance 5The second integral shows that the j-th moment of Z is equal to the actuarial present value for an 1-year term insurance for a unit amount payable at the moment of death of (x), calculated at a force of interest equal to j times the given force of interest, or j6, This property, which we call the rule of moments, holds generally for insurances paying only a unit amount when the force of interest is deterministic, constant or not. More precisely, FZ In addition to the existence of the moments, the sufficient condition for the rule of moments is b/ = b, for all t = 0, that is, for each f the benefit amount is 0 or 1. Demonstration that this is sufficient is left to Exercise 4.30. @&, = F[Z] @j8, (4.2.4) It follows from the rule of moments that Var(Z) = *Alg — (Aba? (4.2.5) where *Al3 is the actuarial present value for an 1-year term insurance for a unit amount calculated at force of interest 28. Whole life insurance provides for a payment following the death of the insured at any time in the future. If the payment is to be a unit amount at the moment of death of (x), then 120, +=0, T=0. The actuarial present value i I. ezi= | For a life selected at x and now age x + hh, the expression would be Ps walt) dt (4.2.6) Aus = [0 aaa mea Whole life insurance is the limiting case of n-year term insurance as n + “, Example 4.2.1 The p.d.f. of the future lifetime, T, for (x) is assumed to be 1/80 0 =1 =80 i {i elsewhere. ‘Section 4.2 Insurances Payable at the Moment of DeathAt a force of interest, 8, calculate for Z, the present-value random variable for a whole life insurance of unit amount issued to (x): a. The actuarial present value b. The variance . The 90th percentile, &°. Solution: A “yt ey Ly = a. A, = EIZ] = i vt fxlt) dt i et eat b. By the rule of moments, Tae fy pal? vara) = 15" _ (Loe) 840 c. For the continuous random variable, Z, we have Pr(Z = &°) = 0.9. 840, 808 Since we have the p.d.f. for T and not for Z, we proceed by finding the event for T which corresponds to Z = &. From Figure 4.2.1, which shows the general relationship between the sample space of T (on the horizontal axis) and the sample space of Z (on the vertical axis), we see that &° = v'', Because Z is a strictly decreasing function of T for whole life insurance, the percentile from T's distribu tion that is related to 90th percentile of Z’s distribution is at the complementary probability level, 0.1. In this example T is uniformly distributed over the interval (0, 80), so & = 8.0 and thus &° = v®°. v The graph in Figure 4.2.1 can be used to establish relationships between the df. and p.d.f. of Z and those of T: Relationship of Z to T for Whole Life Insurance Chapter 4_Life Insurance 7For z = 0, {Z =z) is the null event For 0<2<1,(Z52)={ log z/log v}, and For z= 1, (Z is the certain event. Therefore, 0 z=0 F,(2) 1-F,(logz/logv) O 80} = 0.0, so Pr{0 < Z < v™ = 0.0. Therefore, from (4.2.7) 0 z< 0% Faz) = [ — [log 2)/(log v)|/80 uo <2 <1 1 221 ion of the df in part (a), fete) = \siaseu wo n, T=n T>n. It follows that (42.11) and by taking expectations of both sides Ara = Ala + A. (4.2.12) We can also find the Var(Z,) by using (4.2.11), Var(Z;) = Var(Z,) + Var(Z;) + 2 Cov(Zy, Z,) (4213) By use of the formula Cov(Z,, Z:) = E[Z,Z:] — EIZ,1EIZ,1 (42.14) and the observation that Z,Z,=0 for all T, we have Cov(Z,, Z;) = -EIZJEIZ:] = ~Al aA. (42.15) ‘Substituting (4.2.5), (4.2.9), and (4.2.15) into (4.2.13) produces a formula for Var(Z;) in terms of actuarial present values for an s-year term insurance and a pure endowment. 02 ‘Section 4.2 Insurances Payable at the Moment of Death