: ; etfiods of' Oed uDitig; ; ' . U. 0 rtal ity and ; W it hd ravvi' wrrn IHEiR APPLIDAtlOfj ;JD THE f XPERIFNDE AN.D, V/vLUAT!0N ' ■ : . ; X''':oT.. - t , : ■ (iLFRKS’ ASSOGrATIO'NS,:^ . ■ ■ ■ ^ ‘by;::-:;;.; 'i liOM/'.S G. ALKLAND, F.'I.A. ’ THE UNIVERSITY OF ILLINOIS LIBRARY 3GS.3. leoagtiies Digitized by the Internet Archive in 2017 with funding from University of Illinois Urbana-Champaign Alternates https://archive.org/details/iinvestigationofOOackl [Reprinted from '' T- r-e Jouenal of the Institute of Actuaeies”, Vol. XXXIII, pp. 68-96, 131-197.] 3 67.3 Ac. 51.L (!)• AN INVESTIGATION OP SOME OF THE METHODS FOR DEDUCING THE HATES OF MORTALITY, AND OF MMTHDRAM^AL, IN Y^EAUS OF DURATION ; with (II). THE APPLICATION OF SUCH METHODS TO THE COMPUTATION OF THE RATES EXPERIENCED, AND THE SPECIAL BENEFITS GRANTED, BY CLERKS' ASSOCIATIONS. BY THOMAS G. ACKLANI), F.I.A. IHE investigation of the experienee^ and valuation of the liabilities^ of what are known as Clerks^ Assoeiations, present some features of speeial interest; and I have thought that it might be useful to discuss and explain some of the methods that have been employed in analyzing the data^ and in deducing therefrom the tables of money values appropriate for the valuation of the risks. In dealing with this subject I have had occasion to investigate the most suitable methods of deducing the numbers exposed to risk; and the rates of mortality and of withdrawal^ as affected by the ages of the subscribing members, and by the duration of their membership respectively; and I propose, in the first part of the present paper, to discuss, in some detail, such of the methods suggested for dealing with the experience amongst assured lives, A /a 4 i O 9 as may seem to attain, in whole or in part, the desired objects. 1 shall investigate the formulie appro{)riate to give effect to these methods in a convenient tabular form ; and shall illustrate their practical working, and com])are their results, by an analysis of a ])ortion of the experience of a Clci-ks’ Association, the data of which 1 have recently investigated. I shall then seek to show the adaj)tability of the several methods and formulae suggested to an investigation of lives insured in an assurance office, and the distinctive characteristics of such an c\})erience; also, how the methods are modified in the case where the period of observation is limited by calendar years and by ])olicy years respectively. In the second portion of this ])aper, I ])ropose to state the general characteristics of Clerks^ Associations, as to contribution, benefit, and the like; and to investigate the methods of computing the values of their varying benefits at death; also of deducing the rate, and com])uting the value, of the s])ecial benefit during non- em])loyment, granted by such associations; with due allowance throughout for the effect of withdrawals or secessions upon the values ascertained, and generally upon the valuation reserves. (I). Investigation oe Methods for deducing the Hates OF Mortality and of Withdrawal in Years of Duration. (a). Period of Observation limited by Calendar Years.^ Illustrative Experience. It will be sufficient for the pi’csent to state (reserving further details for the second portion of this paper) that the data here employed for purposes of illustration formed part of the experience of a Clerks^ Association during the period of five complete calendar years, 1888 to 1892 inclusive; that the age at entry recorded on the cards constituting the experience was in all cases the ^^office age^^ upon which the member’s subscriptions at entry were based; that the subscriptions were throughout payable, at * From some remarks made in tlie discussion which followed the reading of this essay, it seems necessary to point out that the limitation of the period of observation within complete calendar yeai's is not to he confounded with the tabulation and scheduling of the facts according to the “calendar year’^ method. In both sections (A) and (B) of this paper the facts are tabulated according to “policy-years,” or years of duration; hut in section (A) the observations included are comprised within an integral number of calendar years, the period actually selected for illustratiou being from 1 January 1888, to 31 December 1892. 3 monthly intervals, upon the first day of eaeh calendar month; and that the number of withdrawals or secessions was heavy, especially in the early years of duration. With a view to investigating the rates of mortality and of withdrawal, as well as the rate of “non-employment’’, as affected by the age of the member, and by his duration, it was decided to tabulate the facts, and deduce the experience throughout, according to office ages at entry and years of duration. By this means, although at the expense of some additional labour, the material was readily available for ascertaining, in respect of each particular rate investigated, whether it was in the main a function of the age, or of the year of duration, or of both age and duration; and the material could be readily combined, in such a way as to give effect to the conclusions in this respect, as indicated by the experience. The data available, in the particular case here dealt with, were, however, evidently not sufficiently extensiye to permit of trust- worthy results, as tabulated for each successive age at entry, and for each year of duration; and it was therefore determined to group the entry ages quinquennially, and to consider each group as representing the experience of the central age at entry in the group. Thus, cases entering at office ages 18, 19, 20, 21, and 22, were all classed together as entrants at “Central Age at Entry (20)”, and so on; and, assuming that the numbers entering, and the rates obtaining, on either side of the central age, did not greatly differ, or so differed as to introduce compensating errors, there was evidently no material departure from the truth involved in this assumption. The formulae employed, and presently to be developed, are, however, applicable equally to the case of individual ages at entry; and where the experience is sufficiently extensive (as, for example, in the New Experience of the Institute of Actuaries), the natural and preferable course would be to tabulate the data, and deduce the results, for each separate age at entry. The individual years of duration were, in all cases, scheduled and dealt with separately, and the duration being entered upon each card, the material was available for ascertaining, in each quinquennial group of entry-ages, and in each separate year of duration, the precise incidence of deaths, withdrawals, and of the benefit during non-employment, with a view to deducing the true rates of mortality, withdrawal, and benefit, respectively. A 2 4 The several methods which have been proposed for dealing with a life experience according to policy ycars^^^ or years of duration, were all carefully referred to. These methods have been very usefully summarized and discussed by Mr. llyan [J.I.A., xxvi, 256), Mr. Chatham {J.I.A., xxix, 81), and Mr. MTiittall {J.I.A., xxxi, 161); while some of them have been illustrated in a practical form by Dr. Sprague {J.I.A., xxxi, 205) and Mr. Meikle [J.I.A., xxxi, 229). None of the methods suggested, as applied by the above writers, appeared, liowever, in all respects to meet the case here investigated; and 1 have thought that it would be of interest to deal fully with their special application to this particular case, in the hope that a further discussion of these several methods might elucidate more fully their comparative advantages. The methods which 1 have selected for illustration and com- ])arison arc — (1) the Exact Duration Method; (2) the Mean Duration Method; and (3) the Nearest Duration Method; each method being so applied as strictly to preserve the incidence of the cases throughout in their true years of duration. 1 should have liked also to include in this investigation the special method suggested by Mr. G. F. Hardy and the late Mr. Rothery {J.I.A., xxvii, 165), which may perhaps appropriately be described as the Mean Age Method^^; as well as that proposed by Mr. G. King {J.I.A., xxvii, 218), which might be styled the^^Nearest Age Method^^; as also the somewhat similar, but in some respects inferior method, called by Dr. Sprague the Final Age Method^^ {J.I.A., xxxi, 215); and especially as these methods are all very simple and facile in their operations. It appeared to me, however, that, although these methods were doubtless admirably adapted for the purposes designed by their respective authors, they certainly do not tabulate the cases ex])osed to risk, and deduce the rates experienced, both of mortality and withdrawal, strictly in years of duration. This has been shewn, as it appears to me, clearly, by Mr. Whittall {J.I.A., xxxi, 182-6); and I was, therefore, unable to include them as suitable for my present purpose. A further reason for their exclusion was, that they all involved data based upon the years of birth, or the birthdays, of the lives included; and, in the particular experience here selected for illustration of the selected methods, I had throughout no information whatever as to the dates of birth of the lives. 5 Exact Duration Method. The durations of the cases were entered on the cards constituting the experience (1) as at the commencement of the period of observation, where the case was then in force; (2) as at the close of the period of observation, where the case was then existing; (3) as at exit during the period of observation, by death or by withdrawal. The duration was obtained in class (1) by taking the difference between the date of entry and January 1888; in class (2) by taking the difference between the date of entry and January 1893; and in class (3) by taking the difference between the date of entry and the date of exit. As all subscriptions were due on the first day of a calendar month, the durations were thus truly stated in integral months, excepting only in the cases of death, where the duration was stated with a possible error not exceeding a month (and averaging a fortnight) in any particular case. It may be added that this error would have been reduced by one-half, if the date of death had been entered on the cards, or computed for the purpose of ascertaining the duration, as the first day of the month nearest to the actual date of death*. The durations were, however, recorded on the cards, not in years and months, but in years and equivalent decimals, correct to one decimal place. Thus, actual durations of 3 years and 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 months were entered as 3-1, 3-2, 3-3, 3-3, 3-4, 3-5, 3*6, 37, 37, 3-8, 3-9. Tlie fractional durations, thus expressed decimally, were found to be much more convenient, in practical use, than if expressed in years and months, being more easily cast and aggregated by inspection, and adapting themselves much more readily to the subsequent processes. Upon the assumption, which was found to be fully justified in this particular experience, that members are equally likely to enter under observation, or to withdraw, in any month of the year, the decimal expressions will be found to give' an average result in which the balance of errors is equal; the aggregate of the decimal expressions for the successive months of duration (0-(-‘l-|-’2 + '3 + "3 + '4-|-'5 + '6-f-*7 + ’7 -j-’8-f-'9), and of the actual months of duration (0+l-l-2-|-3-|-4+5-fG-l-7 + 8-f-9 + 10+ 11)^ being both equal to 5*5 years or 66 months. * If tliis modification lie introduced, enre must, liowevcr, lie taken that the cases of deatli are throughout located in their true years of duration. 6 While tins plan was found to work well, and to introduce no ])ei-ccptible error, in the particular experience here investigated, it is probable that, in the case of ordinary ])olicy investigations, where the assurances will have a strong tendency to terminate at or near to quarterly intervals in the year of duration, the current (juartcr at exit, expressed decimally as *25, *50, *75, or I’OO, would give, on the whole, more satisfactory results. The durations having thus been recorded upon the cards, they were first sorted into quinquennial groups of ages at entry, all cases entering at office ages 18 to 22 inclusive being grouped together as ‘^Central Age at Entry (20)^^; entrants between ages 23 to 27 inclusive as Central Age (25)^^; and so on, up to central age (45), which included the highest age at entry admissible under the rules. There were thus six groups of central ages at entry. Taking now the group of cases entering at office ages 18 to 22 inclusive, or at the central age (20), which group I shall throughout em])loy to illustrate tlie methods here followed, the cards constituting the group were sorted into survivors, in force on 1 January 1888, and ^^new entrants coming on the books during the five years 1888-92 inclusive. The new entrants were counted, and tbeir total number recorded. The survivors’^ were then sorted, according to the curtate duration at entry on observation, as recorded on the cards, and so that all cases, for cxanq)le, whose recorded duration at the commencement of the period ranged from 5T to 6'0 inclusive, were grouped together as of curtate” duration 5. The number of survivors in each year of duration was then counted, and recorded in a tabular form ; and the aggregate fractional exposure of the cases in each year of duration as recorded on the cards was cast bv inspection, and tabulated against the number of corresponding- cases. The operation of casting the decimal figures, disregarding the integers, is rapid and easy; the only point to note being that, according to the method of classification here suggested, an exact duration of G'O (for instance) is considered as of ‘^^curtate” duration 5 and fractional” exposure I'O; but this will be found in practice to give rise to no difficulty whatever. It may be added that, in respect of the survivors at the commencement, as well as of the cases existing at the close, of the observation, it seems to be quite immaterial whether an 7 integral duration of (say) G’O is classified among cases having six years duration and upwards (and counted as 0) or among cases having five years’ duration and upwards (and counted as 1). In the case, however, of withdrawals and deaths, it seems to be clear that they can only be properly treated by considering them as cases of withdrawal (or of death) occurring at the end of the sixth year of membership, and therefore classifying them among emergents having five years’ duration and upwards. This will be evident if consideration be given to the withdrawals, for instance, at the end of twelve months’ duration, which should clearly, as it appears to me, be tabulated with the withdrawals occurring in the first year, for the purpose of correctly deducing the rate of withdrawal in that year. The sorting would have been more symmetrical and a little more facile in this respect if all cases of fractional duration had been entered upon the cards at the next higher integer, or what may be called the current year of duration” (leaving, of course, integral durations unaltered); and in this case the fractional period would have had to be separately dealt with by way of deduction. This appears to have been the plan most frequently followed in published investigations; but, after careful con- sideration, I arrived at the conclusion that the disadvantages of this plan of deduction would on the whole be greater than the slight difficulty involved in the special treatment of the cases of integral duration. To provide against any risk of possible error in the tabulations, a sort of ^^danger-signal” may be put up, by underlining the *0 in red ink upon the cards, in all cases of integral duration, as a reminder to the clerk that they are each to be counted for a full unit. The number of original entrants, and the number and aggregate fractional duration at entry, of the survivors” in each year of duration, having been thus recorded, the cards were combined, and re-sorted into cases existing at the close of the period of observation; cases withdrawing during the period; and cases dying during the period. The cards in each of these three groups were then sorted according to their curtate” durations (for the existing) at the close of the observation, or (for withdrawals and deaths) at exit; and the number of cases, and their aggregate fractional exposures, were then tabulated in their appropriate years of duration. Tlie following Schedule (A) shews the form in which the 8 cases entering at ^^centi’al age at entry (20)^^, and their fractional durations at entry and at exit, Avere tabulated, as a result of the above processes of sorting and grouping. It will be remarked that the aggregate fractional exposures, as deduced from the cards, represents in the case of the survivors’^ that portion of the current year of duration already expired at the commencement of the period of observation; and in the case of the existing,” ^Svitlidrawals,’^ and “deaths,” that portion of the current year of duration over which the cases were actually at risk during the period of observation. It will also be observed that the fractional exposure of the death cases, as Avell as of the withdrawals, is, in this investigation, terminated at the date of exit, whether by death or withdrawal, and not continued until the end of the year of duration current at death. This was done advisedly, as the numbers exposed to risk, so arrived at, formed at the same time a suitable denominator for computing the rate of allowance during unemployment, and a eonvenient basis for deducing, by a simple moditication, the denominators appropriate for the calculation of the rates of mortality, and of withdrawal, respectively. The formulse adopted were also, throughout, more symmetrical and convenient upon this basis. I now proceed to state the formulre for deducing the number exposed to risk, and the rates of mortality and withdrawal. A detinitlon of the leading symbols employed throughout this paper may first be given. Let E^^j+^=the number exposed to risk, in respect of cases entering at “office age” x, during the (/H-I)th year of duration; the bar over the E indieating that the number exposed is eom])uted up to the actual cessation of the risk, whether by death, withdrawal, or elose of the period of observation;* and let E[-j;]+j; = the number exposed to the risk of death and (?/;^J)[j.]+^=:the number exposed to the risk of withdrawal during the (/ + I)th year of duration; where the cases of death and of Avithdrawal are respectively given a full year’s exposure in the year of duration eurrent at exit. * The function I’eally represents the number exposed [in the {t + l)th year] to risk of death or xoithdrawal ; and would be appropriately employed in calculations (1) of benefits the continuance of which depends upon the member’s being alive and in full membership (such as the annuity by which the members’ subscriptions would be valued) (2) of benefits which would necessarily cease on the occurrence of either death or withdrawal (such as an allowance during sickness or non-employment). 9 Schedule (A). — Central Age at Entry (20). Table of the Numbers Surviving, Existing, Withdrawing, and Dying, in Years of Duration ; ivith Fractional Exposure of the Survivor's, as at Entry on Observation, and of the Existing, Withdraivals and Deaths, as at Close of Observation, or Exit. Curtate Dura- tion* Survivors Existing Withdrawals Deaths Curtate Dura- tion* Cases Fractional Exposure Cases Fractional Exposure Cases Fractional Exposure Cases Fractional Exposure t d[x]+t d[x]+t <^'[x]+t t (1) (2) (3) (4) (3) (6) (7) (S) (9) (10) 0 137 81*8 82 44-2 51 24-4 1 0-5 0 1 101 56-5 83 43-3 47 27-1 5 3-6 1 2 132 73-2 7 6-1 44 25-5 3 1-3 2 3 120 65-8 72 40-1 45 20-2 1 0-5 3 4 129 73-9 75 43-1 55 25-3 3 1-3 4 5 101 56-3 69 41-1 44 25-3 3 1-1 5 6 77 37-8 62 37-9 33 18-7 6 7 89 53-6 74 43-4 23 13-5 *4 2-3 7 8 87 49-2 82 45-5 27 16-0 5 2-8 8 9 76 41-2 89 51-1 16 7-9 1 0-3 9 10 64 37-7 69 39-6 19 8-8 3 1-8 10 11 48 24-0 48 24-2 20 10-2 6 3-3 11 12 15 7-6 65 39-9 11 6-6 2 1-0 12 13 19 IIT 69 37-7 13 6*5 2 0-3 13 14 19 11-0 59 34-7 11 6-0 1 0-6 14 1 5 10 7-2 46 28-0 2 0-7 2 1-1 15 16 13 7-7 35 17-7 5 3-2 16 17 15 9-6 11 6-5 3 2-3 17 18 10 3-8 13 6-5 3 1-1 18 19 4 3-0 16 9-7 1 0-8 19 20 9 7-2 7 5-1 3 1-8 20 21 10 6-6 11 7-1 1 0-6 ’*i 0-2 21 22 15 8T 10 6-0 22 23 16 6-6 9 3 5 1 0-2 23 21. 10 6-5 4 3-0 24 25 7 30 9 7-2 1 0-1 i o’-’e 25 26 10 6-6 1 0-8 1 0-5 26 27 12 5-6 1 0-9 27 28 14 5-6 1 *6-9 28 29 8 5-8 2 0-3 29 30 5 2-6 30 1,333 753-0 1,225 698-4 4-81 1 254-5 >4^ OC 1 1 24-3 * Intofri’al durafions of {t + 1) years being treated throughout as of “curtate” duration {t), and “fractional exi^osure” (I’O). Note. — The above experience rein-esents a small portion only of the available data, the cases entering at the grouped entry ages 18 to 22 inclusive having been selected solely for the purpose of illustrating tlie different methods employed. 10 Also, let bo tlie annual rate of mortality; and {wq)^x-\-k-t tbc annual rate of withdrawal, in the (/H-l)th year of duration. liCt + ^ — the survivors in force at the commencement of the period of observation, having a duration exceeding /, but not exceeding (/ + 1) years; the new entrants at office age’^ x, during the ])eriod of observation; C[a-] + 1 — the cases existing at the close of the period of observation, having a duration exceeding t, but not exceeding (/ + 1) years; the during the period of observation, having a duration at exit exceeding /, but not exceeding (/ + 1) years; the deaths during the period of observation, having a duration at death exceeding t, but not exceeding (/+1) years. Also, let the exact fractional exposure of the 5[.,.]+i5 survivors, computed from the commencement of the (/H-l)th year of duration, up to the commencement of the ])eriod of observation; the exact fractional exposure of the cases existing, com])uted from the commencement of (/ + l)th year of duration, up to the close of the period of observation ; the exact fractional exposure of the cases of withdrawal, computed from the commence- ment of the (/ + l)th year of duration, up to the date of withdrawal; the actual fractional exposure of the cases of death, computed from the commencement of the (^+l)th year of duration, up to the date of death; {e-\-w-\-d)— f, d\x\+t — Also, let that is, the total decrement in the year of duration, in respect of cases existing, withdrawing, and dying; and let that is, the net movement in the year of duration, among cases surviving, existing, withdrawing, and dying; and similarly, let {e -\-w d') = j' (*■'-/') = .!/' 11 where f' and 3'5 2‘3 59'2 8 87 82 27 5 114 _ 27 394 15*1 409-1 49‘2 45'5 i6'o 2-8 64'3 9 70 89 10 1 10() _ 30 304 379-1 44 -2 7 ‘9 o’3 S9’3 - iS*i 10 04 0.9 19 3 91 _ 27 337 I2‘5 349-5 37'7 39'6 8-8 1-8 50-2 11 48 48 20 () 74 _ 2(i 311 3-24-7 24*0 24-2 10*2 3’3 37'7 >3*7 1-2 15 ()5 11 2 78 _ 03 248 -287-9 7‘6 39 '9 6 6 i‘o 47'5 39*9 13 19 09 13 2 84 _ (i5 183 210-4 ii'i 37'7 6-5 o’3 44’S 33 4 14 19 59 11 1 71 _ 52 131 3^*3 101-3 II’O 34‘7 6‘o 0*6 4i‘3 ~ 15 10 4l5 2 2 50 _ 40 91 113-0 7’2 28‘0 o‘7 i‘i 29-8 22 6 1() 13 35 5 40 _ 27 04 77-2 7'7 17-7 3'2 20‘9 13*2 17 15 11 3 14 + i 05 + 04-2 9‘6 6'S 2’3 8-8 o‘8 18 10 13 3 10 _ () 59 3*3 02-8 3-8 6'S i‘i 76 53-5 19 4 10 1 17 _ 13 40 3'o 9'7 0-8 lo'S 7*5 ‘20 9 7 3 10 — 1 45 + 0*3 44-7 7'2 S'l 1-8 69 21 10 11 1 1 13 _ 3 42 43-3 6-6 7'i o‘6 0'2 7 '9 1*3 22 15 10 10 + 5 47 + 44 9 8-1 6‘o 6‘o 2‘I 23 10 9 1 10 + 0 53 50-1 6 6 3‘5 0'2 3'7 + 2*9 24 10 4 4 + 0 59 + 55-5 6-5 30 3’o 3*5 25 9 1 1 11 _ 4 55 59-9 3'o 7 '2 o‘i o‘6 7‘9 4*9 •20 10 1 1 12 — 12 43 50-9 6‘6 0-8 o‘S 7'9 7 9 27 12 1 13 _ 13 30 6*5 30-5 5’6 o‘9 65 21-5 28 14 1 15 _ 15 15 6*5 5‘6 o‘9 6-5 11-1 29 8 2 10 — 10 5 5-8 03 6‘i 6‘i 30 5 5 _ 5 2‘6 2-6 2‘6 26 T,333 1,225 *^8 1.754 — 421 0,210 0,434-2 7S3‘o 698-4 234'5 24‘3 977*2 1 — 224'2 13 Five Ca^lendar Years. — Exact Duration Method. — Schedule (B) . Bisk, and the Bates, of Mortality and of Withdrawal, in true years of Exposures. — Central Aye at Entry ( 20 ). Continuous Method Mortalitv Withdrawal Curtate duration Exposed Rate Exposed Rate 1 9M+t 2rxj+« d W J. ^9 [x]+i-l 1 - A/(x]+i-i 1 EUl+i ^ E = E -f- («• — iv') t (11) (12) (13) (14) (15) (16) (17) (18) »[xl = 421 411-3 411-8 -00243 437-9 •1165 0 + I2‘7 - 9*7 407-5 408-9 -01223 427-4 •1100 1 — 30"2 - 3*8 427-7 420-4 -00609 446-2 •0986 2 + 57’8 + 20*2 465-0 465-5 -00215 489-8 •0919 3 - 35‘3 + 37*3 461-8 463 -5 -00647 491-5 •1119 4 - 0-8 - 3*2 462-2 464-1 •00646 480-9 •0915 5 - iS‘4 + o’4 451-8 451-8 466-1 •0708 6 - 7-6 — 10*4 426-6 428-3 •00034 436-1 •0527 7 + 13-3 - 25*2 400-1 411-3 •01216 420-1 •0643 8 - 9'5 - I7'i 379-1 379-8 •00263 387-2 •0413 9 o — 30*0 340-5 350-7 •00856 359-7 •0528 10 + 2'6 — 29 '6 324-7 327-4 •01833 334-5 •0598 11 — 1'2 — 24’8 287-0 288-9 •00692 292-3 •0376 12 — 26’2 - 36-8 216-4 218-1 •00917 222 -9 •0583 13 + 6-s - 71*5 161-3 161-7 •00618 166 3 •0662 14 + 3'i - 55*1 113-6 114-5 •01747 114-9 •0174 15 + 7*7 - 47*7 77-2 77-2 79-0 •0633 16 + 9*4 - 36*4 64-2 64-2 64-9 •0462 17 + 14*0 - 13*0 62-8 62-8 64-7 •0464 18 - 4-6 - 1*4 53-5 53-5 53-7 •0186 19 - 3*7 - 9*3 44-7 44-7 45-9 •0654 20 + 7*3 — 8-8 43 3 44-1 •02268 43-7 •0229 21 — I "6 - 1-4 44-9 44 0 44-9 22 + 3*4 + I '6 50-1 50-1 50-9 •0197 23 + o‘8 + 5*2 55-5 55-5 55-5 24 + o’6 + 5*4 50-0 60-3 -01658 60-8 •0165 25 - 8*4 + 4*4 50-9 51-4 •01045 51-1 •0196 26 - 3*0 - 9*0 36-5 36-6 •02732 36-5 27 + 1*4 - 14-4 21-5 21-5 21-6 •0463 28 o - 15*0 11-1 12-8 •15625 11-1 29 + o’4 — io‘4 2 6 2 6 2-6 80 + 3*5 - 8-s i 6,434-2 (),457-9 6,()60-7 — 2'6 1 - 418-4 1 • • u Final! V; in columns (15) and (17) the rates of mortality and of withdi‘awal arc deduced by the respective formulai and (Rx]U — In ])racticc, one or other of the two methods set out in columns (8) to (10), and (11) to (13), would alone be adopted, thus reducing the labour and the number of columns involved; and it will ])robably be found preferable to adopt the continuous formula shown in columns (11) to (13), and to employ the summation formula in verification of the results at any stage, thus ^[20] +10— ^[20] "b 0 [20]+ 10 = 421-84+ 12-5 = 319-5 b^[20] + 20 — ^[20] + -0“"(i7) — //'[20J + 20 = 121-376-0-3 = 14-7 b^[20]+30 — ^^[20] "b 9 [20] + 30 = 421-421+2-6 = 2-6 I have, however, preferred to set out the full process under both methods, partly to illustrate the two operations, and partly to secure verification throughout. The rates of mortality as set out in column (15), being based upon only 48 deaths, are, of course, very irregular, and I need hardly say that they are only computed here for purposes of illustration and comparison, and are not presented as representing results practically available. The withdrawals are 481 in number, and the rates of withdrawal, as set out in column (17), show a smoother progression in years of duration. Bearing in mind that the Exact Duration Method gives effect to the precise exposures of the cases, and deduces the rates, both of mortality and of withdrawal, strictly as experienced in each successive year of duration, it may, I think, be considered as not unduly laborious for the valuable results obtained. Under this method no assumptions whatever are made, either as regards the ages attained, or as to the average epochs of entry or of exit, in the several years of duration. 15 In comparing tlie extent of the Schedule (B) with other tabular statements which have been published, it must be borne in mind that in the appended schedule the numbers exposed to risk and the rates experienced are deduced both for mortality and withdrawal. Mean Duration Method. If we consider that upon the average the fractional exposure of the survivors, in the years of duration current at the commencement of the period of observation, and of the cases existing, withdrawing, and dying, in the years of duration current at the close of the period, or at exit, is approximately equal to half a year, forniulie (1) and (3) will become [^see Appendix (A)] (5) a very simple and convenient formula for computing the numbers exposed to risk by a continuous operation. Similarly, formula (4) will become, upon the above assumptions, .... (7) a formula by which the number exposed to risk can be obtained by tabular summation, or which can be applied as a useful check upon the results obtained by formula (6). In this case it is not, of course, necessary to record the exact duration of the cases upon the cards, but only, in the case of the survivors^’, the curtate’"’ duration at entry; and in the case of the existing, withdrawing, and dying, the ^‘^curtate” duration at the close of the observation, or at exit; cases of integral duration of (^+ I) years being treated, as before, as of ^^curtate” duration (t). The cards of the survivors” must then be sorted and tabulated according to the duration at entry, and those of the cases existing, withdrawing, and dying, according to the duration at exit, as recorded upon the cards. The appended Schedule (C) shows the process followed in computing the number exposed to risk (E) by the two formulaj (6) and (7), also the values of E and (wE) deduced from the formulae — w and (wAE)f.Y]+< = E[.ri+^+ ^ Ajipendix C). IG Schedule (C). — OnsERVATiox Extending oyer Table showing alternative methods of deducing the Numbers Exposed to duration^ and with mean or average Fractional Curtato Dura- tion Survivors Existing AVith- (Irawals Deaths Total De- creiiieiit Net Movement Summation Method t •Vj+i fM+t ^[x]+t fu]+t fflx]+t «rxi + V(5') 2 (1) (-0 ( 3 ) (. (^0 ( 6 ) {7) ( S ) «[x] = 421 (i>) (10) 0 137 82 51 1 131 + 3 424 — 1*5 422-5 1 101 83 47 5 135 -34 390 + 17-0 407-0 2 132 7 44 3 54 + 7S 468 -390 429-0 3 120 72 45 1 118 + 2 470 - 1-0 469-0 4 129 75 55 3 133 - 4 466 - i - 2'0 468-0 5 101 69 41 3 116 -15 451 + 7-5 458-5 6 77 62 33 95 -18 433 -f 9-0 442-0 7 89 74 23 4 101 -12 421 + 6-0 427-0 8 87 82 27 5 114 -27 394 + 13-5 407-5 9 76 89 16 1 106 -30 364 -f 15-0 379-0 10 64 69 19 3 91 -27 337 + 13-5 350-5 11 48 48 20 6 74 -26 311 + 13-0 324-0 12 15 65 11 2 78 -63 248 - f - 31-5 279-5 13 19 69 13 2 84 — 65 183 -f 32-5 215-5 14 19 59 11 1 71 -52 131 + 26-0 157-0 15 10 46 2 2 50 -40 91 + 20-0 111-0 16 13 35 5 40 -27 64 - i - 13-5 77-5 17 15 11 3 14 + 1 65 - 0-5 64-5 18 10 13 3 16 - 6 59 -f 3-0 62-0 19 4 16 1 17 -13 46 + 6-5 52-5 20 9 7 3 10 - 1 45 + O'o 45-5 21 10 11 1 1 13 - 3 42 + 1-5 43-5 22 15 10 10 + 5 47 - 2-5 44-5 23 16 9 i 10 -f 6 53 - 3-0 50-0 21 10 4 4 + 6 59 - 30 56-0 25 7 9 1 i 11 - 4 55 + 2-0 57-0 26 10 1 1 12 -12 43 + 6-0 49-0 27 12 1 13 -13 30 -f 6-5 36-5 28 14 1 15 — 15 15 - t - 7-5 22-5 29 8 ”2 10 -10 5 + 5-0 10-0 30 5 5 - 5 4 - 2’5 2-5 1,333 1,225 481 48 1,754 -421 6,210 210-5 6420-5 The values of q[x]+t and (w-'(7)[a']+6 as deduced from these num- bers exposed to risk, are also appended. Here, again, the columns (8) to (10) can be dispensed with, and the. summation formula employed simply for purposes of verification at suitable intervals. This method has the advantages of being very simple and rapid in working, of avoiding, to a great extent, the employment of fractions, and, at the same time, of preserving the incidence of the cases strictly in their appropriate years of duration. It treats all cases of entry, of emergence, and of existence, as occurring in the middle of the year of duration: and one effect of this is that surviving^’ cases entering and emerging in the same year of 17 Five Calendae Yeaes. — Mean Dueation Method. — Schedule (C). Risk, and the Rates, of Mortality and of Withdrawal, in true years of Exposures. — Central Age at Entry (20). Continuous Method Mortality Withdrawal Curtate Dura- Exposed Eate Exposed Eate tion 2 ^Lxl+t SM+i ^x]+t - d = E+g d “ E — w to toE t (11) (12) Wfa:] = 421 (13) (14) (15) (16) (17) + 1-5 422-5 423-0 •00237 448-0 •1138 0 - 15-5 407-0 409-5 •01221 430-5 •1092 1 + 22-0 429-0 430-5 •00697 451-0 •0982 2 + 40-0 469-0 469-5 •002 L3 491-5 •0917 3 - 1-0 468-0 469*5 •00639 495-5 •1110 4 - 9-5 458-5 460-0 •00652 480-5 •0915 5 - 16-5 442-0 442-0 458-5 •0719 6 - 15-0 4270 429-0 •00932 438-5 •0524 7 - 19-5 407-5 410-0 •01220 421-0 •0641 8 - 28-5 379-0 379-5 •00264 387-0 •0413 9 - 28-5 350-5 352-0 •00852 360-0 •0528 10 - 26-5 324-0 327-0 •01835 334-0 •0599 11 - 44-5 279-5 280-5 •00713 285-0 •0386 12 - 64-0 215-5 216-5 •00924 222-0 •0586 13 - 58-5 157-0 157-5 •00635 162-5 •0679 14 - 46 0 111-0 112-0 •01786 112-0 •0179 15 - 33-5 77-5 77-5 80-0 •0625 16 - 130 64-5 64-5 66-0 •0455 17 - 2-5 62-0 62-0 63-5 •0472 18 - 9-5 52-5 52-5 53-0 •0189 19 - 7-0 45-5 45-5 47-0 •0638 20 - 20 43-5 44-0 •02273 44-0 •0227 21 + 1-0 44-5 44-5 44-5 22 + 5'5 50-0 50-0 50-5 •0198 23 + 6-0 56-0 56-0 56-0 24 + I'O 57-0 57-5 •01739 57*5 •0l’74 25 - 8-0 49-0 49-5 •02020 49-5 •0202 26 - 12-5 36-5 37-0 •02703 36-5 27 - 14-0 22-5 22-5 23-0 •0435 28 - 12-5 100 11-0 •18T82 10-0 29 — 7*5 - 2-5 2-5 2-5 2-5 30 -421 6,420-5 6,444-5 6,661-0 duration are altogether eliminated from the experience. Thus^ a case entering upon the period of observation at a duration of 2*1 years^ and emerging at a duration of 2’9 years, is considered as entering upon observation at 2 5 years, and emerging at 2 5 years. As some compensation for this, a case entering (for instance) at 2*9 years and emerging at 3*1 years is considered as entering at 2*5 and emerging at 3*5, and as under observation for a full year. Upon the whole, however, the method deals with the fractional exposures fairly at average values ; and in the case of an experience such as that here investigated, gives, as will be seen, values for the numbers exposed to risk, and for the rates of B 18 mortality and of withdrawal^ in successive years of duration^ which agree very closely with those deduced by the Exact Duration Method. Nearest Duration Method. This is the method which has been illustrated by Dr. Sprague (J.I.A., xxxi^ 208-12) in schedules arranged aecording to ages at entry and years of duration. It has^ I tliink, been somewhat hastily assumed that this method necessarily involves a mixing-up of the ])olicy years^^ (J.I.A., xxxi, 309-315), and that, therefore, although possessing manifest and acknowledged advantages by way of simplicity of arrangement and facility of computation, it is unsuitable for an investigation (such as the ])resent) which aims at deducing the true experience of each year of duration. It is mainly with a view to a further discussion of this question that I here introduce this method; and I shall illustrate its application by the same ])artial experience which has served to illustrate the Exact Duration ]\Iethod and the Mean Duration ^lethod. The Nearest Duration Method ])roeeeds upon the assumption that all cases entering on the ])eriod of observation do so either at the commencement or at the end of the year of duration current at entry; that all cases emerging during the period do so cither at the commencement or at the end of the year of duration current at exit; and, consequently, that all cases existing at the close of the ])cr’.od have then completed integral years of duration. These assuinjitions involve the reference of all cases of fractional exjiosure to the beginning or to the end of the year of duration current at entiy, or at exit; and in carrying this into effect, the nearest boundary of the year of duration is in all cases adojited. Thus, cases having a duration at entry (or at exit) of 6-1 to 6‘4 years inclusive, would be considered as entering (or emerging) at an integral duration of 6 years; cases having durations of 6'6 to 7*0 inclusive would be considered as entering (or emerging) at an integral duration of 7 years; and cases entering (or emerging) precisely mid-way, or at 6‘5 years, would be alternately classed as entrants (or as emergents) at integral durations of 6, and of 7 years. The operation, so far, deals solely with the amount of fractional exposure within each separate year of duration, and adopts convenient and average assumptions of equivalent integral durations, hut always so as strictly to 'preserve the incidence of the cases in their true years of duration. 19 In carrying this method out in ju’actice^ the cards of the survivors at the commencement of the ])criod of observation would be entered up with the nearest integral duration at entry, according to the above plan; the cards of the eases ^^existing^^ would similarly be eiitei-ed up with the nearest integral duration at the close of the period; and the cards of the cases of with- drawal would be entered up with the nearest integral duration at withdi’awal. As regards the deaths, a different course would be followed, in an investigation intended to deduce the rate of mortality; for, in order to give each case of death a full yearns exposure in the year of death, the duration entered upon the cards must be the year of duration then current (that is, the curtate duration + 1) and not the nearest integral duration. The cards are then sorted into original entrants, and survivors’"’; the latter being then sorted, counted, and tabulated, according to the integral durations at entry, as recorded upon the cards; and the cards then combined, and re-sorted into cases existing, withdrawals, and deaths; and these again sorted, counted, and tabulated according to their recorded integral durations at exit. The tabulation takes the form set out in Schedule (D). I shall throughout employ the convenient symbols {as), (ae), (aw), and (ad) to indicate the cases surviving, existing, withdrawing, and dying, which are, by the Nearest Duration Method, referred to the beginning of the year of duration current at entry or at exit; and the symbols {hs), {he), {bw), and {bd) to indicate the cases referred to the end of the year of duration current at entry or at exit; the sums of these quantities being, of course, equal to s, e, w, and d in any given year of duration. Thus, by the Nearest Duration IMethod, the number of survivors’^ tabulated against duration t, which I shall call is equal to and similarly with the cases existing (e), and withdrawing (w) ; as shewn in the headings of columns (2), (3), and (4), of Schedule (D). 2 20 ScjiEDULE (D). — Observation Extending over Table showing methods of deducing the Numbers Exposed to Eisk, and the Fractional Exposures being taken to the nearest integer. (^This method the Rate of Withdrawal). — Central Age at Entry (20). 1 Dura- tion Survivors Existing Witlidrawals Deaths Total Decrement Net Move- Mortality ment Exposed Rate 1 ! ^ i = Slx]-l-< (M)[x]\t-l \ + i t > (^»’)rx]-(-<-i ) -f ) = "■lx]+/ These rates, as well as the numbers exposed to risk in columns (9) and (12), are identical throughout with those given in Schedule (D). It will be remarked that the experience dealt with in both schedules differs from that investigated by Dr. Sprague [J.I.A., xxxi, 212) by including a body of survivors (s) in force at the commencement of the period of observation, instead of tracing the assurances from their original entry. The introduction of the survivors presents no special feature of difficulty. It must, however, be noted that the survivors who have, at entry on the period of observation, a duration of less than six months, will, by the rule of nearest duration, be classed as of Duration 0^’ at entry, and will thus practically be included among the cases entering as new assurances during the period of observation. Thus, there were, in the experience here dealt with, 56 cases of survivors of less than six months^ duration at the commencement of the period of observation. These 56 cases were included with the 421 cases of new entrants during the period, thus raising the total to 477. On the other hand, there were, at the close of the period of observation, 40 cases existing with a duration of less than six months, and therefore classed as existing at ^^Duration 0’^ : these 40 cases were altogether eliminated from the ex})erience, thus 2G Schedule (E). — Observation Extending over Table shewing methods of deducing the Numbers Exposed to Risk, and the Eractional Exposures being taken to the nearest integer. {Cases of Central Age atEntry (20) . Dura- tion Survivors Existing Withdrawals Deaths Total Decrement Net Movement »'fxl + t (M[x]+t-i| + {ae)[x]+tj = <'[x]4-< (a[hs) ; [ae) = ?>[be) ] and (J)w) = d {aw) ; these values being computed to the nearest integer. The death cases were tabulated throughout in half-years, as actually experienced. To illustrate more clearly the methods followed, and the effects of the assumptions made as to the distribution of the cases, I have tabulated the cases surviving, existing, withdrawing and dying, in half-years of duration throughout. It will be understood, however, that as regards the surviving” and existing,” in columns (2) and (3), it is not necessary to sort or tabulate the cases in half-years, for the purpose of computing the numbers exposed to risk, or the rates of mortality and withdrawal. The number exposed to risk in each year of duration (E[j:]+i) is computed by formula (11), and the values of E[a’]+i; and of 37 and of q[x]+t and {wq\x]j^t, ai’c deduced as in Schedule (E), pp. 26, 27. If now the numbers exposed to risk of death and of with- drawals, as arrived at in Schedule (F), are compared with those deduced in Schedule (E) upon the basis of a (practically) uniform distribution of the entrants and emergents throughout the year, it will be seen that the aggregate results are as follows ; — Nearest Duration Method. 2(E) '^{'wJE) Uniform Distribution (Schedule E) 6,440 6,667 Assumed Distrihution (Schedule E) 6,664 6,683 Difference .... -h 224 -hl6 Thus the numbers exposed to the risk of death are, in this particular experience, considerably increased when the special assumptions as to distribution are given effect to ; while the numbers exposed to the risk of withdrawal, are (in this particular case) not substantially varied. It will be readily seen that the difference between the numbers exposed to the risk of death, as deduced (by the Nearest Duration Method) from the assumed distribution of entrants and with- drawals, and as deduced (by the same method) from their uniform distribution, may be expressed as equal to \j(e-s)- [(be) - (6s)] } + | ^’ - (*«’) | which expression holds good in any given year of duration, or in the aggregate. Taking aggregate values, we have, {K1225-1333)-(668-711)*} + |^ x4.81-225| = (16 + 207-9) =223-9, which agrees closely with the actual difference of the aggregate numbers. Similarly, the difference in the numbers exposed to the risk of withdrawal is equal to i(e-s)-[(Ae)-(6s)] = i(1225-1333)-(6G8-711)* =16. Here the distribution of the death cases, in the year of death, does not affect the numbers exposed to the risk of death ; and similarly the distribution of the withdrawals, in the year of * The values of {be) and {hs) are not sejiaratoly stated in Schedule (E); hut I have taken them from working- sheets, which include the values tliroughout in half-years of duration. 38 SciiEDULK (F). — Oesertation Kxtendtxg over Assumed Distimijution of Table showing methods of deducing the Numbers Exposed to Risk, and the Fractional Exposures being taken to the Dura- tion Survivors Existing Withdrawals Deaths Total Decreinent Net Movement «|x] + 2c/(^) ■^5.vix)+<-i ^ ‘25eix]+(-i / ■9'qx]-l-(-l ( |(w)w+<-il (e f s — "F ^ t ■75eix]+< ) («^)rxi+( ) -fw -f d) x; = Slx|+< = ^lz]+i = "’[xJ-K = ‘^|xH-( = f[xl-t( = g^[xJ-t-( (1) (•-') (3) (4) (5) (0) (7) (3) «lx] = -J21 0 103 01 5 ' 1 07 4- 3t) 457 34 1 |21 40 o» 1 70 f to2 5 1 135 - 25 432 251 |21 42 1 n •2 99 ' t 5 5 2 f 79 -I- 45 477 33 1 39 ' U 3 IK)-' ^54 5 1 / 102 + 21 498 30 \ (18 40 01 4 97 ^ ^ 50 0 3 / 123 -1- 4 502 32 1 / 10 49 0 70 / F5-> 4 1 2 I 120 - 18 484 25 1 T' 40 ' n 0 58 Uo 3 1 .. / 107 - 24 400 191 jio 30 7 07 > ’-55 2 ■ 1 ( 104 - 18 442 221 / 19 21 1 8 05 > ^01 3 3 / 110 - 23 419 22 1 (21 24 1 n 9 57 / ^07 ■> i 1 / 117 - 38 381 191 (22 14 ; 01 10 48 1 l52 2 91 - 24 357 10) (17 17 ! n 11 30 f Uo 2 4 < 78 - 2() 331 121 (12 18 “1 12 111 t 49 1 2 f 84 - 01 270 (10 10 13 14 1 152 1 2 / 81 - 03 207 12 01 14 14 1 l44 1 i 74 - 55 152 |15 10 15 7 I \35 0 1 ) 02 - 50 102 3) 2 11 K) 10 f l20 1 41 - 28 74 f 4 •1 17 11 1 1 8 0 21 - 7 07 ■11 1 ^ 3 •• l 18 rf lio 0 10 - 5 02 3 •• 1 19 3 / i 12 0 .. / 18 - 12 50 1 1 I ^ 1 ■ i 20 7 / 1 5 0 10 _ 2 48 3 •• l 21 8 / 1 8 0 1 / 14 - 4 44 21 1 22 11 / { 8 12 + 1 45 “"I • 1 23 12 f i 7 0 .. f 9 + 7 52 1 • • 1 24 8/ { 3 .. f 0 + 6 58 21 ■ 1 25 5 i { 7 0 of 8 - 1 57 1 i 11 2(> { 3 ll 13 - 11 40 1 ! 01 27 0 / 12 - 12 34 / ^ U 28 111 0 ' 15 - 15 19 1 • 1 29 { i’> 1 2f 12 - 12 7 i 30 { 4 1 . 1 6 - 6 1 ( 1 1 •• 1 31 i 1 - 1 1,000 919 48 1 29 1 333 300 433 ' 19 ! 1,754 -421 6,635 39 Five Calendaii Yeaes. — Neaiiest Duration Method. — Schedule (F). Entrants and Withdrawals. Hates, of Mortality and of Withdrawal, in true years of duration; the nearest integer . — Central Aye at Entry ( 20 ). Mortality Withdrawal Dura- Exposed Deaths Rate Exposed Witlidrawals ; Rate tion ^dx]+< E + {ad) d[x]+t 2Lx]+« d ~ E (?rE')[a;i+e = E + {aio) W {loE) t (9) (10) (11) (12) (13) (14) (15) 458 1 •00218 462 51 •1104 0 433 5 •01155 437 47 •1075 1 479 3 •00020 482 44 •0913 2 499 1 •00200 503 45 •0895 3 505 3 •00594 508 55 •1083 4 480 3 •00017 488 44 •0902 5 400 463 33 •0713 6 443- 4 •00903 444 23 •0518 7 422 5 •01185 422 27 •0640 8 382 1 •00262 383 16 •0418 9 353 3 •00837 359 19 •0529 10 335 0 01719 333 20 •0600 11 272 2 •00735 271 11 •0400 12 209 2 •00957 208 13 •0625 13 152 1 •00658 153 11 •0719 14 103 2 •01942 102 2 •0196 15 74 75 5 •0607 16 07 67 3 •0448 17 62 62 3 •0484 18 50 50 1 •0200 19 48 48 3 •0625 20 45 1 •02222 44 1 •0227 21 45 45 22 52 52 1 •0192 23 58 58 24 57 1 •01754 57 1 •0175 25 47 1 •02128 46 1 •0217 26 34 1 •0-2941 34 27 19 19 1 •0526 28 9 2 •22222 7 29 1 1 30 31 6,(i04 48 ! 0,083 isi 1 40 withdrawal, docs not affect the ninnbcrs exposed to the risk of withdrawal. This would of course be anticipated, and evidently arises from the fact that the deaths and withdrawals respectively arc given a full year’s exposure in their year of exit. It will also be remarked that the difference in the aggregate numbers exposed to the risk of death and of withdrawal, is considerably affected by the fact that the numbers existing (e) and surviving (s) are approximately equal, and nearly balance one another. If the experience were such that the value of {s — e) was relatively large (as, for exam])le, where the cases are under obser- vation from original entry up to a fixed date, and the ^^survivors” wholly disappear), the results above shown would be materially varied. A comparison of the values of q and of (wq), in columns (10) and (12) of Schedule (F), with those given in columns (11) and (14) of Schedule (E) will show the effect, upon the rates of mortality and withdrawal, of the assumed variation in the distributions of the entrants and emergents. It will be seen that the rate of mortality is, under the assumed distribution, lower in the first 12 years of duration, and throughout higher, from the 13th year of duration to the end of the table; while the rates of withdrawal, although somewhat lower in the first six years of duration, do not, upon the whole, show any very marked deviations. Let us now consider the effect of the distribution of entrants and emergents, in the case where the Mean Duration Method is applied. It is evident that this method, being based upon the assumption of an equal distribution of the entrants and emergents, will give ])rccisely the same results, whatever the actual or assumed distribution may be; and that those results will, in the case of this ])articular ex])erience, be those shown in Schedule (C), pp. 10, 17. Comparing, then, the numbers exposed to risk, as given in Schedules (C) and (F), we find that the aggregate values are as follow: — Assumed Distribution of Entrants and Emergents. 2(E) 2(w^) Mean Duration Method (Schedule C) . 6,444‘5 6,661‘0 Nearest Duration Method (Schedule E) 6,664* 6,683* Difference ..... +219*5 +22*0 Here, the difference in the numbers exposed to the risk of death is evidently equal to 41 = 27+192-4 = 219-4 and the difFcrence in the numbers exposed to the risk of withdrawal is equal to = 27+ [19-24] =22 The remarks made above as to the counter-balancing effects upon these results of the nearly equal values of (e) and {s) will also apply here. It is thus evident that the Mean Duration Method gives^ as might have been anticipated, erroneous results as regards the numbers exposed to risk, and the resulting rates of mortality and of withdrawal, in the case supposed where the average fractional exposures of the entrants and emergents materially differs from half a year. Special Suitability of Nearest Duration Method for THE Experience of Assured Lives generally. The Nearest Duration Method, as here applied (in Schedule F) to the case of an assumed distribution of entrants and emergents, cannot be compared with the Exact Duration Method, as in this illustrative case the exact durations have not been ascertained, or rather have been assumed to be faithfully reproduced by the Nearest Duration Method. There can, however, be no doubt of the special suitability of this method to the experience of assured lives; for this method gives effect, to a very great extent, to the actual average durations of the assurances in their several years of duration, and that by a sort of automatic reference to the beginning or end of those years. As compared with the Mean Duration Method, which gives an average of half-a-year^s exposure to all entrants and emergents (and thus deduces a result which can hardly fail to be erroneous in the case of assured lives) the Nearest Duration Method has manifest advantages. This method also seems to me to be greatly su])erior to the method [suggested and illustrated by Mr. Ryan {J.l.A., xxvi, 259-264)], of taking a constant ratio of the entrants and emergents, to re])resent their assumed fractional exposure in all years of duration. The Nearest Duration Method, by giving approximate 42 effect to tlic true fractional exposures in cacli year of duration, will allow for any marked deviation from the general average ratio in individual years; and thus possesses an elasticity which compares very favourably with the rigidity of the Constant Ratio method. Let us now consider briefly the probable deviations of the Nearest Duration Method from the true exposures, as regards the several cases of withdrawals (1) by lapse, (2) by surrender, (3) by miscellaneous causes; also of cases ^^surviving^^ and ^^existing^^; and cases of death. Taking first the cases of withdrawal (1) by lapse, it will be borne in mind that cases of non-renewal of ])remium, occurring at the end of the year of duration (whether in respect of yearh^, half- yearly, or quarterly premiums), will, in all cases, be given their true integral exposures; also, that cases of non-renewal occurring in the middle of the year of duration (whether in respect of half- yearly or quarterly premiums) will, by the system of alternate reference to the beginning and end of the year of exit, be also given their true ex])osures, taken one with another. There only remain the quarterly cases of non-renewal at the first and third quarters of the year of duration; and those occurring at the first quarter (and referred by the Nearest Duration Method to the beginning of the year) may fairly be considered to be balanced by those occurring in the third quarter (and referred to the end of the year). As regards (2) surrenders, there will be a certain error in the estimated exposures in respect of the period interv^ening between the date of surrender and the next following renewal date. This period is, in practice, usually found to be very short; and any error will be much reduced by the fact that the intervening period is understated, if the surrender take place in the first half of the year of exit, and overstated if it occur in the second half. If there be (a) yearly cases, {b) half-yearly cases, and (c) quarterly cases, surrendered in any year of duration ; if represent the period intervening between surrender and the next renewal date; and if it be assumed that a surrender is equally likely to take place before any of the four quarterly renewal dates, or before either of the half-yearly renewal dates; it can readily be shown that the aggregate exposures of the yearly cases surrendering are overstated by — years; and that the aggregate exposures of the 43 lialf-ycarly and quarterly cases arc respectively understated by — i^years and by J years. The aggregate error would thus amount to -h m U mJ V B mJ and the average error would be found by dividing this expression by (« + Z> + c). If we assume (as before) that the cases are in the proportions of 250 yearly, 130 half-yearly, and 20 quarterly, and that the average interval between surrender and the next renewal date is one month, the above expression becomes 250 130 20 _ 10 T2 6 24 ~ 6 showing an aggregate error (for 400 cases) of 1| years’ exposures; and an average error of g^^th of a year, or say 1 ^ days, in each case. The remarkable power of the Nearest Duration Method, in adapting itself to the conditions of actual practice, could, perhaps, hardly be better exemplified. The miscellaneous cases of withdrawal (3) include maturity of endowment assurances; expiration of term assurances; cancel- ment by forfeiture, &c. These cases would not be relatively numerous; and while the majority would usually fall at the end of the year current at exit, the remainder may fairly be considered as likely to be equally distributed over that year. As regards cases of surviving entrants, they will, as already stated, tend to enter upon observation in the first half of the year of duration current at entry; at least, in the usual case, where the commencement of the period of observation is also the commencement of a financial year of the office. These cases will, by the Nearest Duration Method, be given an exposure for the whole of the year of duration current at entry, which will be in excess of the true exposures. Upon the other hand, the cases existing at the close of the period of observation will, on the whole, tend to pass out of observation in the first half of the year of duration then current; and such cases will, by the Nearest Duration Method, be given no exposure in that year of duration, thus involving an under- statement of the true exposures. There is thus a compensating infiuence at work, in reduction of the balance of error; and where the number of cases of surviving entrants does not greatly differ from that of the eases ^^existing^^, the ultimate error may be cxj)cctcd to be relatively small. In the case, however, of an exj)erience where there are no ^^survivors^^, there would be no such compensating influence; and the cases “existing^^ would, as a whole, have their exposures understated by the Nearest Duration Method. Finally, as regards the cases of death. Upon the usual assumption of an equal distribution of the deaths over the year of duration current at death, the average exposure would be half a year; and this would be correctly tabulated, taking one case with another, both by the Mean Duration Method, and the Nearest Duration Method. (See, however. Appendix D). (B) Ferioi) of Obseuvation Limited by Years of Duration. Illustrative Experience. AVc have hitherto investigated the case where the period of observation extends over an integral number of calendar years, so that survivors enter at fractional durations at the commence- ment, and cases exist, also at fractional durations, at the close. If now the experience is so investigated that the period of observation runs concurrently with years of duration, that is, so that the survivors” come under observation from the com- mencement, and the cases existing” are traced up to the close, of a year of duration, the formulaj and the tabular operations are throughout very much simjilified. As these are precisely the conditions under which the New Institute Experience is being investigated, the cases being traced from their policy anniversary in the calendar year 1863 (or from subsequent entry) to their ])olicy anniversary in 1893 (or previous exit), I have deemed it useful to investigate this condition of things somewhat in detail. For this purpose I eliminated, from the illustrative experience previously emjiloyed (1) all cases emerging (by death or with- drawal) ])rior to their policy anniversaries in 1888; (2) all cases entering in the year 1892; while cases emerging (by death or withdrawal) during the year 1892, and subsequently to their policy anniversaries in that year, were treated as existing.” There were thus eliminated 65 survivors” and 89 original entrants, leaving 1,268 and 332 respectively; of which 1,184 existed at their policy anniversaries in 1892, 385 withdrew during the period of observation, and 31 died. The period of observation thus extended over four years of duration, from the policy anniversary in 1888 (in the ease of survivors”) to the poliey anniversary in 1892 (in the case of 45 ^^cxisting^’) ; and included all eases of entry, withdrawal, and death falling between those anniversaries. I have, in Appendix (B), investigated the modifications introduced, in the formulse already given, to meet the particular case of an experience limited by years of duration, in the several cases where the Exact Duration Method, the Mean Duration Method, and the Nearest Duration Method, are adopted. Exact Duration Method. I have com})uted, in the appended Schedule (G), the numbers exposed to risk, and deduced the rates of mortality and with- drawal, by the formulje E[a;]+< = So^(G) + + + « . . . (18) and — • . (21) where G[x]+t= [«’ — + ^)] M+i and s’ and e’ indicate the numbers respectively surviving and existing at the integral duration of t years; the withdrawals (iv) and the deaths [d) being tabulated as before, according to their curtate durations at exit. Upon comparing Schedule (G) with Schedule (B)* it will be seen that the tabular operations are much simplified, in the case here investigated, where the period of observation is limited by years of duration. — In Schedule (G), fractional expressions enter only into column (4) of withdrawals, and column (5) of deaths ; and by the formulse given above, the values of E and of [wE) are deduced by a direct process, and one not involving the calculation of the function E. It will also be noted that the new entrants (332 in number) during the period of observation here con- veniently figure as surviving entrants at precise age {x) ; and that the table, as a whole, is thus rendered more symmetrical and convenient. This Schedule (G) appears to me to present, in a very simple form, the whole of the tabular work involved in deducing the numbers exposed to risk of death and of withdrawal, and the resulting rates, in true years of duration, and with exact fractional exposures. It need hardly be added that the values of q and of [wq) deduced in columns (II) and (14) cannot properly be compared with those set out in columns (15) and (17) of Schedule (B), as the period of observation, and the data involved, are not identical in the two cases. * See pp. 12, 13. 10 ScirEDULE ((x). OnSEETATION ExTENDTXG OVER Table showing methods of deducing the Numbers E.rposed to Bish, and the with exact Fractional Fx^iositres. Curtate Dura- tion t Surviving Existing Withdrawals Deaths Total Decrement Not Movement ^IxH 1 T[x]+t + w + ci) *= ^lx]+« (.vi-F) 0) ( 2 ) (3) (4) (r>) (li) (7) (S) 0 332 35 35 d-297 297 i6'6 1 128 85 40 3 128 0 297 21-9 2-5 2 93 7 3t) 2 45 -f- 48 345 191 0-4 3 124 75 40 1 116 -h 8 353 18-5 o‘5 4 IIG 80 42 2 124 - 8 345 i8'2 0-8 5 121 77 32 2 111 + 10 355 18-5 o‘S G 9G ()2 29 91 + 5 3G0 iS‘9 7 74 78 17 2 97 - 23 337 10-3 o '8 8 84 84 24 2 110 - 2G 311 i4'5 I 3 () 82 91 11 1 103 - 21 290 5 '5 o'3 10 74 74 15 2 91 - 17 273 7'9 I’l 11 G3 50 17 6 73 - 10 2G3 8-9 33 12 48 G5 10 1 7G - 28 235 56 o'S 13 13 70 12 1 83 - 70 165 56 O’l 14 18 G 1 9 1 71 - 53 112 S'l 0-6 15 19 47 1 1 49 - 30 82 o‘3 O'l IG 9 3() 4 40 - 31 51 2’6 17 13 11 3 14 - 1 50 23 18 14 14 14 0 50 19 9 IG 1 17 - 8 42 0-8 20 4 7 3 10 - 6 3G 1-8 21 9 12 1 13 - 4 32 o ‘6 22 10 10 10 0 32 23 15 9 9 + G 38 24 15 4 4 + 11 49 25 10 9 1 1 11 - 1 48 O'l 0-6 2G 7 10 1 1 12 - 5 43 0-8 o'S 27 12 1 13 - 13 30 0-9 28 14 1 15 - 15 15 0-9 29 9 1 10 - 10 5 0'2 30 5 5 - 5 1,G00 1,184 385 31 1,G00 0 4,941 202 '3 i *5'o 47 Four Years of Durvtiox. — Exact Dfrattox Method. — Schedule (G). Rates^ of Mortality and of JVitlidrawal^ in true years of duration; and — Central Aye at Entry (20). Mortality Withdrawal Curt.ite WixW + d[_x^+t) Exposed Rate Expn.sed Rate 1 ( III ti— tion + {w' + d\x]+t 9i\x]+t d ^ E + d\xw) {ivE)ix]+t + + d')[x\-^t '111 ^ fiE ) t (9) (10) (11) (12) (13) (14) (15) 16-6 313-(5 35-0 332-0 •1054 0 24-9 321-9 -00932 42-5 339-5 •1178 1 21-1 36(3-1 -00546 36-4 381-4 •0944 2 19-5 372-5 -00269 40-5 393-5 •1017 3 20-2 305-2 -00548 42-8 387-8 •1083 4 20-5 375-5 -00533 32-5 387-5 •0826 5 15-9 375-9 29-0 389-0 •0746 6 12-3 349-3 -00573 17 8 354-8 •0480 7 16*5 327-5 -00611 25-3 336-3 •0714 8 6-5 296-5 -00337 11-3 301-3 •03()5 9 9-9 282-9 -00707 16-1 289-1 •0519 10 • 14-9 277-9 -02159 20 3 283-3 •0600 11 6-6 241-6 -00414 10-5 245-5 •0407 12 6-6 171-6 •00583 12-1 177-1 •01)77 13 0-1 118-1 •00847 9-6 121-6 •0740 14 1-3 83-3 •01200 1-1 83-1 •0120 15 2 (3 .53-6 4-0 .55-0 •0727 16 2-3 .52-3 3-0 53-0 •05()6 17 .50-0 50-0 18 0-8 42-8 1-0 43-0 •0233 19 1-8 37-8 3-0 39 0 •0769 20 0-6 32-6 1-0 33-0 •0303 21 32-0 32-0 22 38-0 38-0 23 49-0 49-0 24 1-1 49-1 •02037 1-6 49-6 •0202 25 1-8 44-8 •02232 1-5 44-5 •0225 26 1-0 31-0 •03226 0-9 30-9 27 0-9 15-9 1-0 16-0 •0625 28 10 (3-0 •1(;()67 0-2 5-2 29 30 m^3 5,174-3 400-0 5341-0 4.S ^Iean Dukation ]\Ietiioi). In tins casCj the formulie for the numbers exposed to risk beeome res|)ectively = + . . . (24) («E)„+, = S„'(G) + («,+ 0,,„. . . (27) and the tabular results are set out in Sehedule (H). Schedule (H). — Obseryation Extending over Table showing methods of deducing the Numbers Exposed to EisJe^ and the mean or average Fractional Exposures. Cuit'itc Diirii- tioii Surviving Existing With- drawals Deaths Total Decrement Net Movement t e^J+t ^[xHt (^[x\+t -^tu + d) =-^tn+« {s^-F) (1) (2) (3) (4) (5) (6) (7) (8) 0 332 35 35 + 297 297 1 128 85 40 3 128 0 297 2 93 7 36 2 45 + 48 345 3 124 75 40 1 116 + 8 353 4 116 80 42 2 124 - 8 345 5 121 77 32 2 111 + 10 355 6 96 62 29 91 + 5 360 7 74 78 17 2 97 - 23 337 8 84 84 24 2 110 - 26 311 9 82 91 11 1 103 - 21 290 10 74 74 15 2 91 - 17 273 11 63 50 17 6 73 - 10 263 12 48 65 10 1 76 - 28 235 13 13 70 12 1 83 - 70 165 14 18 61 9 1 71 - 53 112 15 19 47 1 1 49 - 30 82 16 9 36 4 40 - 31 51 17 13 11 3 14 - 1 50 18 14 14 14 0 50 19 9 16 ’ i 17 - 8 42 20 4 7 3 10 - 6 36 21 9 12 1 13 - 4 32 22 10 10 10 0 32 23 15 9 9 + 6 38 24 15 4 4 + 11 49 25 10 9 1 1 11 - 1 48 26 7 10 1 1 12 — 5 43 27 12 1 13 - 13 30 28 14 1 15 — 15 15 29 9 • ... i 10 - 10 5 30 5 5 - 5 1,600 1,184 385 31 1,600 0 4,941 49 Here again the operations are simplified as compared with those shown in Schedule (C),* and the values of E and of (wE) are deduced by direct processes. This method is especially rapid in its operations, and is very suitable for the investigation of the experience of a body of lives, where the average fractional exposure at exit does not greatly differ from half a year. * See pp. 16, 17. Four Years of Duration. — Mean Duration Method. — Schedule (H). Rales of Mortality and of Withdrawal, in true years of duration; and with — Central Age at Entry (20). /■"‘M+i V 2 + Mortality Withdrawal Curtate Dura- tion Exi)osed Rate Exposed Kate Kiii+i =2o‘(e) •(I-) 7[x]+< d ~ E \ (W-E')[a;]+< = V((7) («'7)[x]+t w ^ foE) t (9) ( 10 ) (11) (12) ( 13 ) ( 14 ) ( 15 ) 17-5 314-5 35-0 332-0 •1054 0 23-0 320-0 -00938 41-5 338-5 •1182 1 20-0 365-0 -00548 37-0 382 0 •0942 2 210 374-0 •00267 dO-5 393-5 •1017 3 23-0 368-0 -00544 43-0 388-0 •1082 4 18-0 373-0 •00536 330 388-0 •0825 5 14-5 374*5 29-0 389 0 •0746 6 10-5 347*5 •00576 18-0 355-0 •0479 7 14-0 325-0 •00615 25-0 336-0 •0714 8 6-5 296-5 •00337 11-5 301-5 •0365 9 9-5 282 5 •00708 16-0 289-0 •0519 10 14-5 277-5 •02163 20-0 283-0 •0601 11 60 241-0 •00415 10-5 245-5 •0407 12 7-0 172-0 •00581 12-5 177-5 •0676 13 5-5 117-5 •00851 9-5 121-5 •0741 14 1-5 83-5 •01198 1-5 83-5 •0120 15 2-0 53-0 4-0 55-0 •0727 16 1'5 51-5 3-0 53-0 •0566 17 50-0 50-0 18 b-5 42-5 1-0 43-0 •0233 19 1-5 37-5 3-0 39-0 •0769 20 0-5 32-5 1-0 33-0 •0303 21 . .« 32-0 32-0 22 38-0 38-0 23 49-0 49-0 24 i-5 49-5 •02020 1-5 49-5 •0202 25 1*5 44 5 •02247 1-5 44-5 •0225 26 1-0 31-0 •03226 0-5 30-5 27 0-5 15*5 1-0 16-0 •0625 28 10 6-0 •16667 0-5 5-5 29 30 223-5 5,161-5 400 5 5,341-5 D 50 Nearest Duration Method. Ill this case the foriiiulai hecoiiie respectively and + * = V(G) 4- («<^)[.r]4( • • • (29) (u;£;)[,.]+i = S„^(G)-f («u')[.v]+i . . • (30) where G^+f = [s’ — (e‘ -t- w 4 d )] 1 w and d representing, as before, the numbers of withdi-avvals and deaths as tabulated by the Nearest Duration Method. The values are tabulated in Schedule (J). The cases surviving and existing arc respectively set out, in columns (2) and (3), at their true integral durations, the new entrants figuring as and in columns (4) and (5) the withdrawals and deaths are respectively tabulated, according to their half-years of duration at exit. The numbers existing, withdrawing, and dying, as set out in columns (3), (4) and (5), arc then summed (as indicated by the brackets) in column (6) ; their sum being deducted from the numbers surviving in column (2), and the difference entered in column (7). The values thus arrived at, continuously summed in column (8), give the values of (G) the numbers cx])osed to risk up to the dates of death or withdrawal ; from which the values of E and of (wE) are deduced, by the above formulae, in columns (9) and (12) respectively. In columns (10) and (13) the total number of deaths and of withdrawals are tabulated for convenience, these numbers being simply the sums of those set out, in adjacent half-years, in columns (5) and (4) respectively. Finally, in columns (11) and (14), the rates of mortality and of withdrawal are computed, for each year of duration. I have explained these operations in detail, because this method appears to me to be, upon the whole, that best suited for practically dealing with a large body of assured lives, in such a way as to deduce, by a very simple and rapid series of operations, the rates of mortality and of withdrawal, in their true years of duration. Comparing Schedule (J) with Schedule (E),* where the observation was limited by calendar years, it will be seen that the tabular operations are practically identical throughout, and as See pp. 26, 27. 51 2„*(G) = E[.r]+f in this case^ this latter function is necessarily com- puted (in both Schedules) by the ordinary processes followed. Comparison op Aggregate Numbers Exposed to Risk. The aggregate numbers exposed to risk of death and of withdrawal, as deduced by the three methods, are as follow: 2(E) '2,{wE) Exact Duration Method (Schedule G) . 5,174*3 5,341*0 Mean Dui’ation Method (Schedule H) . 5,164*5 5,341*5 Nearest Duration Method (Schedule J) 5,148* 5,337* These results are practically identical, the greatest difference not exceeding 1 in 200, or one-half per-cent. The rates of mortality q and of withdrawal [wq) also give closely identical results, in each year of duration, as deduced by the three methods. Assumed Distribution of Emergents over Years of Duration current at Exit. In Schedule (K) I have tabulated, by the Nearest Duration Method, the cases included in the period of observation as limited by years of duration, upon the basis of an assumed distribution of the withdrawals over the years of duration current at exit; and, as before, I have assumed that the withdrawals, taken one with another, are exposed for nine-tenths of the year of exit. The methods followed, and the tabular arrangement, are identical with those set out in Schedule (J), the only difference in the data being in column (4), where the withdrawals are set out in half- years of duration, according to their assumed distribution, and so that [bw) = 9(aw), throughout. If we now compare Schedule (J) with Schedule (K), we shall see the results, upon the numbers exposed to risk, and upon the rates of death and withdrawal, of a variation in the distribution of the withdrawals over their year of exit. It will be seen that the aggregate numbers exposed to risk arc as follow : — n 2 52 ScnEDULE (J). — Observation Extending over Table showing methods of deducing the Numbers Nxgosed to Risk^ and the Fractional Exposures being taken to the Dura- tion Surviving Existing Witlidrawals Deaths Total Decrement Net Movement {bd\x]+t-i '( fol = V((i) t (««f’)[xl+< ) 0^)[xl+« ) “T \\ T- U j = "'[x]+« = 1 1 Survivors Existing Methods Employed 1 1 Cases as tabulated, (A) Period OF Obsekvation LIMITED BY Calendar Years. (1) Exact Duration Method — (B) Duration (t) as tabulated Cases as tabulated . curtate* curtate* «tx]+« Aggregate Fractional Exposures ^'lx]+< (2) .Mean Duration Method — (C) Duration (/) as tabulated curtate* curtate * Cases as tabulated . ^[xJ+< (3) Nearest Duration Method— (l))(a) (a) For Mortality only — Duration {t) as tabulated Cases as tabulated . (ft) For Withdrawal only — .. (I>)(ft) nearest Integral SM+< = nearest integral Cfx]+« = (^^)[x]+«-l + («^)[x]+I Duration (ft) as tabulated Cases as tabulated . (e) For Mortality and With- dratoal — (K),(F) nearest integral S[a;]+I = (^-y)[x]+I-l + nearest integral eLx]+« = + («^)[x]+< Reference to beginning or| end of Year of Dui*ation ) Duration (ft) as tabulated Cases as tabulated . ( B) Period of Observation Limited BY Years of Duration. ...^ («^)[x]+« nearest integral f'[x]+< = (*«)[x]+<-l + («5)[x]+< {ae)ix]+t (^^)[x]+< nearest integral e[x]+I = (^^)[x]+«-l + (««)[x]+< (1) Ex.act Duration Method — (G) j Duration (ft) as tabulated Cases as tabulated . true integral ^\x]+« true integral e’[x]+I Aggregate Fractional Exposures (2) Mean Duration Method — (H) Duration (ft) as tabulated Cases as tabulated . true integral «\x]+< true integral (3) Ne.arrst Duration Method — (.T),(K) ! Reference to beginning or ) end of Year of Duration ) Duration (ft) as tabulated Cases as tabulated . ■■■ ( true integral s\x]+« true integral e\x]+< * Integral durations of {t + 1) years being considered as of Symbols common to Numbers exposed to Risk — Of Death ........ Era:]+i Of Withdrawal Of Death or Withdrawal ..... '^[x]+t New Entrants at age [x] ..... G1 employed in Part I. Withdrawals Deaths Total Decrement Net Movement at Age at Entry [x], and Duration (<). curtate* curtate * W(x]+« f[x^+t 9[x-]+t — {e + IV -f- d)[x]+t = i.^-f)ixW w'w+t ^'\xW /'[x]+< g'ixW = (d + w' -f- d')ix+t} = {d -f)\.x]+t curtate * curtate * f[x]+t 9[x-\+t = {e + ^v + (^)[x-]+« = (*-/)[x]+t nearest integral current year %]-(-« -1 AxW g’M+i = (e + w + (?) = (s-f)[x]+« current year nearest integral f'[x]+< s'[x-]+t {bd)ix]+t - 1 + {fid){x^+t = (e + fc + d) = (s-fO[x]-i-< («/)[x]+< 1 {ag\x]+t (¥)[x]+t \ id9\x-\+t nearest integral nearest integral ... f[x]+i g'M+t (M[x]+t-i + (aw)[x]+t + (ac?)[xH-t = (e -f- w -i- d)[^]+< == (s - f)[x]d-t cui-tate * 1 curtate* «'[x]+« i d[x]+t -^[x]+< ^[x]+i \ = (el d- + d\x]+t =('^l-^)[x]4-« '^'[xW d\x-\+t F'[x-\+t <^'[x]+< = {w' + (?')[x]+« =-(w'-K?')[x]+< curtate * i 1 curtate * ^M+< Pix]+t <^[x]+« = (el + w + c?)[x]+« = (-^^--^)[xl+< {ad\x]+t («/)[x]+< (^??)[x]+< {bd\:c]+t W\x]+t (%)[xJ-i-« nearest integral nearest integral W[x]+« = = ^\x]+« Grx]+« (iw)[a:]+«-i + («w)M+i (S(^)[x]+<-i + {ad){x]^t = (ei + w+d)[:cHt = (si-F)[x]+i ‘‘curtate” duration {t), and “fractional exposure” (I'O). all above Methods. Rates — Of Mortality Of Witlulrawal . Central Death-rate Central Withdrawal Rate (II). Application of Methods to the computation of the Rates experienced, and the Special Benefits GRANTED, BY ClEKKS’ AsSOCIATIOxNS. General Characteristics of Clerks’ Associations, and NATURE OF SpECIAL BENEFITS GrANTED. It will be convenient, at the outset, to set out as concisely as ])ossible the leading ohjccts and general characteristics of these Clerks’ Associations, which are established in several of the cities and larger towns of England and Scotland. The nieinbership is restricted to bond fide clerks employed within a definite local area; and the main object of the Associations is to render assistance to the members when out of employment, or when jiayment of their salaries is tcmjiorarily suspended by reason of sickness. The following is a condensed summary of the main provisions of the rules of one of these Associations established in an English city. These particulars are given by way of illustration merely, and while the main principles will not vary, the actual provisions as to age, amount, incidence of allowances, &c., may materially differ in individual Associations. Members are admitted between the ages of 18 and 45. The rates of monthly subscriptions vary from 25-. to 2s. 9^/. per month, under the ^^Low” Scale: but members can, at their option, subscribe at an increase of 50 per-cent upon the above rates (on the Middle” Scale), or at double the minimum rates (on the ^Mligh” Scale). The benefits secured under the three scales are strictly proportionate to the increased subscriptions. The benefits granted under the Low Scale of Subscriptions arc as follow: — (1) A weekly allowance during non-employment, or during sickness involving temporary suspension of pay, granted to members of six months’ standing and upwards, of 245. weekly during the first four weeks; 155. weekly during the following nine weeks; and 85. weekly during the remaining 13 weeks, at the end of which the allowance ceases, but is subject to renewal, for similar amounts and durations, after the member shall have held a permanent situation for at least six months. This restrictive term is reduced to three months in respect of members who have been three years upon the 63 books without rcceiviug benefit: and is altogether abrogated in the case of members who have been five years upon the books without receiving benefit. There are also limitations as to the maximum aggregate . amounts which may be received during the whole of membership by way of allowance, which need not be here specified in detail. (2) An increasing death-benefit on the following scale : — To members of between six months’ and five years’ standing, £7. IO 5 .; to members of between five and ten years’ standing, j011. 5s.; to members of 10 years’ standing and upwards, <£15. (3) Allowances to members of not less than five years’ standing, who shall be permanently unfitted for employ- ment; and to members of not less than 15 years’ standing, who shall, from old age or permanent ill- health, have no prospect of obtaining remunerative employment; also to necessitous widows and orphans of members. These allowances may be either by way of benevolent grants, or annuities, and they are at the entire discretion of the executive as to grant, amount, and duration, and are usually made from a Special Annuity Fund” or Reserve Fund”, created and augmented by voluntary contributions, and appro- priations of surplus moneys. They are not, therefore, a charge upon the members’ monthly contributions. (4) Medical attendance and medicine during sickness, the charges under this head being provided by the allocation of a fixed portion of the members’ subscriptions, averaging about 45. per member per annum. (5) An “Employment Bureau”, or classified register of candidates and situations, by which the members are assisted in obtaining situations when out of employment; this being equally a benefit to the individual member, and to the Association, which is thus liberated from a charge upon its funds in respect of the allowance during non-employment. (6) A restricted allowance of death-benefit upon the Minimum Scale, or of death-benefit and medical attendance only, to members who, by removal outside the local district, or by entry into business as employers (and thus ceasing to be “clerks”) have forfeited the right to the allowance (luring non-ein])loymcnt; rcclucecl subscriptions of 155. and of 215. per annum respectively being paid by such members. The management expenses of the Association are usually defrayed (1) by an allocation of a fixed proportion (say 35 per- cent) of the members^ subscriptions; (2) by incidental reeeipts from entry fees, donations, and miscellaneous sources of income. Form of Card adopted, and Method of Recording the Experience. The following is the form of card, which was, on the whole, found to be that most appropriate for recording the data: — FRONT. Number at Entry ^ (J, Subscription £/ ' Date of Entry J . J J . S Si. Date of Exit j . Mode of Exit Duration (Entry) JZ , Si Duration (Exit) S . Remarks 65 A few explanatory notes are needed to make clear the mode of filling up the several particulars upon the cards : — (1) Number . — This was filled in with the consecutive number of membership^ corresponding to the policy number in an insurance office. (2) Scale . — This was entered as H ” (High), M (Middle), or 1 j ” (Low). (3) Age at Entry . — This was, in all cases, the office age at entry, as charged, and stated in the office books. As, however, the scale of subscriptions was constant over groups of entry ages (18-30, 30-35, 35-40, and 40-45) it is probable that there was no great exactitude in stating the precise age at entry ; nor does it appear BACK. Allowances During Sickness OR Non-Employment Duration Date Weeks Days Amount i886 6 k 6 6!i§ § ' ^ \ j 888 i jii ^o'c) 8- n "^JL. i8(jj \ i i 1 1 8 j 1 i 1 / ! ! 8 m E G6 that any cvidcaice of tlic true age was required or furnished. Upon the whole, and after enquiry, it seemed ])robable that the offiee age at entry was that at nearest birthday, which would be identical with Dr. Sprague’s conmiencirg age.” {J.I.A., xxxi, 208.) (4) Subscription. — This was tilled in from the office books, the amount entered on the card being the year’s subscription In practice all subscriptions were paid at intervals of calendar months, excepting the relatively few cases of restricted benefit, where the reduced subscription of 15s. or 21s. was paid annually. (5) Date of Entry. — This was entered up with the month and year in which the first subscription was ])aid. The subscriptions being payable in all cases on the first day of a calendar month, the date of entry always coincides wdth the commencement of the month recorded upon the cards. (G) Date of Exit. — This was recorded, in cases of with- drawal, as the month and year in which the subscrij)tion ceased ; and in cases of death, as the month and year in which death actually took ])lace. (7) Mode of Exit. — This was either by withdrawal, including («) non-payment of subseription (b) removal outside the local district (c) entering upon business as an employer — all of wffiieh were marked L ” — or by death, which was marked D.” (8) Duration (Entry). \ These particulars were not entered (9) Duration (Exit) . ) upon the cards from the office books, but were computed and recorded later on, according to the methods specified in the first part of this paper.* It may be added that, where it is desired to deduce and tabulate the experience according to ages attained, these headings would be replaced by ‘^Age at Entry on Observation” and ^^Age at Exit” respectively. The additional lines afford opportunity for similar entries at subsequent investigations. (10) Remarks. — Here particulars were entered of any special incidents of the case, including changes of scale during membership ; revival after lapse ; transfer to restricted scale of benefit, by removal or entry upon business; transfers to Annuity or Benevolent Funds; &c. * See pp. 5, 6. G7 (11) Allowances during Sickness or Non-Ernployment . — -The entries under this head being sometimes numerous^ and increasing with the duration of membership^ it was found preferable to preserve the back of the cards for these particulars. The first column duration (representing the duration of actual membership at the date when allowance commenced) was not supplied from the office books^ but computed and recorded later. The d'dte” was that of the month and year when benefit commenced; the ^“^weeks^^ the number of weeks (and days) during which the allowance was continuously paid; and the amount” the aggregate sum received in respect of each such series of continuous payments. Methods adopted in deducing Rates of Mortality, With- drawal, AND Allowance during Non-Employment. The method actually selected for investigating the particular experience here under review was that of Mean Durations, as this method offered great facilities in the actual computations, and undoubtedly gave very accurate results in this particular case. The processes of deducing the numbers exposed to risk, and the rates of mortality and withdrawal, in central ages at entry and individual years of duration, have been fully explained in the first part of this paper, the tabulation being in the form set out in Schedule (C), on pages 16 and 17. The rate of mortality was experimentally computed in this form ; but it was seen from the outset that the aggregate number of deaths (128) was quite too few to give trustworthy, or indeed, workable, results ; and it was ultimately decided to treat the rate of mortality as a function of the age only, irrespective of duration of membership. The age attained was assumed to be equal to the sum of the office age at entry (a:) and the curtate duration (/); and all cases having an (assumed) age of (oc-\-t) were brought together so that the numbers exposed to risk (E^+f) and the deaths (^/x-+«) could be compared, and the rate of mortality computed. The rate thus deduced was equal to the central death-rate [mx+t), and the values of ((pc+t) were deduced by the formula of relation _ 2mx+t 2 + mxj^t It will be seen that, as tlie cases had been grouped in central ages at entry, and years of duration, the effect of this was that the values of lUx+t eonsequently those of qx+^ were deduced from the numbers exposed to risk, and the deaths at five grouped ages (x + / — 2), (.r + /— 1), (x + / + l), and [x-\-t-\-2). This was in effect a first graduation by summation in fives. The results thus deduced were combined with the experience of the previous seven years 1880-7, and a graduated rate was ultimately adopted, which represented approximately the experience of the 12 years 1880-1892. The rate of mortality thus adopted, as a basis for money- values, is set out in Table 1.* The aetual experience did not extend beyond age 60, and at this age a junetion was conveniently effected with a table of mortality re])resenting the experience of males in city districts employed in similar occupations to those of the members whose experience was under investigation. As regards the rate uf ivithdrawal, the experience was much more extensive, there having been 1,222 cases of withdrawal during the period of observation. The actual rate of withdraAval was deduced for each central age at entry and each year of duration by the formula the values of (w(j)[-c]+t being then computed by the formula of relation 2 + + ^ The resulting values are set out in Table II for each central age at entry and each year of duration. It will be seen that the rate of withdrawal (at all entry ages) averages about 16 per-cent in the first year of duration, about 9 })er-cent in the fifth year, 5 ])er-cent in the tenth year, 4 per-cent in the fifteenth year, and thence diminishes steadily to practical extinction at about the thirtieth year. It is also to be noted that the rates are somewhat materially lower at central age at entry 20 (especially in the earlier years of duration) than the rates at higher entry ages in corresponding years of duration. As there was a considerable preponderance of cases at the earlier entry ages and durations, this diminution in the rate of withdrawal was somewhat note- worthy, and eould hardly be disregarded in the computation of money-values. * For Tables I to VIII, see pages 78-84. 69 A rate of withdrawal was ultimately adopted^ whlch^ while having regard to the results set out in Table was well within that actually experienced^ and thus left a fair margin for contin- gencies in this respect. As regards the rate of rion-enq^loyment , the method adopted was as follows: The dates of the several allowances^ the number of weeks and days during which they were continuously paid^ and the aggregate amounts paid in respect of each continuous allow- ance, having been recorded on the backs of the cards, as already explained, the duration of membership as at the date of each such allowance was computed to one decimal place, and entered in the column headed Duration.’^ The cards were then sorted (1) according to central ages at entry; (2) according to year of duration when entering on benefit; and the cases counted and tabulated, so as to show, for each central age at entry and each successive year of duration, the number of cases of benefit, the total number of weeks during which benefit was actually paid, and the aggregate amount paid in each case. The allowance was, as has been said, a variable one, diminish- ing after 4 weeks’, and again after 13 weeks’, continuous pay, and ceasing after 26 weeks of total pay. In order to give effect to this, in a form convenient for computation, a further column was added, giving the number of weeks over which the allowance would have extended, if the amount of the weekly pay had been, throughout, the maximum amount which was in fact payable during the first four weeks only. This was readily arrived at, by dividing the aggregate amount received, by the maximum weekly allowance, which, of course, varied according to the scale of membership and subscription (High, Middle, or Low); and could be computed in each individual case, or in each group of cases where the cential entry age, the year of duration, and the scale of membership, were all identical. The number of weeks thus deduced, which I have termed equivalent weeks”, numbered 2,776 in the aggregate, as compared with 4,206 weeks during which the reducible allowance Avas actually paid; and the effect of this Avas to reduce the tabulated number of weeks in the average ratio of about 3 to 2. •By the adoption of this device, a good deal of labour Avas saved in the folloAving stages. It proceeds, as will have been seen, upon the reasonable assumption that the value of an allow- ance extending, for example, over fifteen weeks, of £1. 45 . during the first four weeks, of 155. during the following nine weeks, 70 and of 8s. for the remaining two weeks^ — or c€l2. 7s. in all — cannot differ materially from tlie value of an allowance at the constant rate of £1. 4s. weekly extending over 10-3 weeks, and also amounting, in the aggregate to £12. 7s. Upon the basis of the number of “equivalent weeks^’ thus deduced, the average number of weeks’ allowance per member was deduced (1) as a function of the year of duration only; (2) as a function of the age attained only; the formulae being respectively where Vt and represent the aggregate number of “equivalent weeks” arising in the (/ + l)th year of duration, and in the year following age (,r-|- /), respectively ; and E.r^^ re})resent the numbers exi)osed to risk in the (^-f l)th year of duration, and in the year following age (.r + Z) res|)eetively ; and Ut and Vx+t rei)resent the resulting rates of non-employment. In deducing these rates, as at ages attained, a special correction was made, in respect of the non-j)ayment of benefit during the first six months of membershi]). In Table III, I have set out the rate of non-em])loymcnt computed uj)on these two bases. As regards the rate experienced in years of duration, it might be anticipated that there would be some indications of a selection against the Association in the early years of duration, arising from members entering who were in expectation of benefitting by the allowances in this respect at an early date. If the r^tes,as tabulated in years of duration, are closely examined, as set out in column (4) of Table III, regard being given to the fact that the rate during the first year should practically be doubled for purposes of comjiarison, it will be seen that there are indications, in the early years of duration, of the effects of some such causes: but these indications cannot be said to be strongly marked. As regards the tabulation of the rates according to ages attained, it will be seen from column (2) that the rate tends to diminish from age 20 to about age 31, after which it tends on the whole to increase with the age, up to age 64, the highest age under observation. 71 It was, upon the whole, thought to be preferable to deduee the money values upon the basis of the rate of non-employrnent as computed at ages attained. As this function is one of some interest, I have appended in Table IV the rate of non-employment as experienced in the several periods 1862-79, column (2); 1880-87, column (3); and 1888-92, column (4), at ages attained; and also, in column (5), the rates adopted for the computation of money values, after consideration of the whole experience thus available. The rates above age 64 not being obtainable, an assumed rate of allowance was tentatively adopted ; and, having regard to the restrictions under the rules as to the duration of allowance, and the conditions under which the allowance could be resumed, as well as those limiting the aggregate amount receivable by a member, a rate increasing up to a maximum of two weeks^ allowance per annum per member, was deemed to be fully sufficient to meet the case. Methods of Computing Values of Benefits, and Valuation Tables and Results. Upon the bases specified above, as to the rates of mortality, of withdrawal, and of allowance during non-employment, there were computed, for each central age at entry, and for each successive year of duration, the value (1) of annuities jiayable throughout membership (2) of the varying assurance at death and (3) of the benefit payable during non-employment; interest being taken throughout at a rate of 3 per-cent per annum. (1) The values of the annuity ])ayable throughout member- ship, that is, until cessation by death or withdrawal, were computed upon the basis of the approximate formula (See Appendix E) ^i\x]+t=q{x]+t+ [_q x (w.’<7)][.i-]+^ where q' represents the probability of either death or withdrawal during the year of duration; whence we have // T f/ V u-]+t—^ — q [x]^t and log = ^ log ~^''[x] + t whence the values of and of a'\x]-\ t can be deduced by the ordinary ])rocesscs. (2) The values of the assurance, payable at death, with allowance for mortality and withdrawal during life, were based upon the approximate formula (See xVppendix E) where q represents the probability of death, allowing for withdrawals: and we have ^ [.1-] + ^ X H + i XV = d + ^ W + t or 1|>S '/'w+!+ log DVi+(+ log «= log C'(j,n.i whence the values of of and of the varying assurance, can readily be deduced in the usual way. (3) The values of the allowance during non-employment were obtained by the formula IT' U']4< — 7^// ^ lx] + t where represents the value of a constant benefit of one per week during non-employment, with allowance for mortality and withdrawals, and + « + + + < + + « + D'%] + ( + U-W[a;] + i+l+ . . . The tabular values, computed as above, represent the values of the several benefits to members entering at office age [. 2 ], who were in existence (at the date of valuation) at the precise durations 0, 1, 2, 3, . . . t years. By taking the means of these values throughout, factors were obtained a})propriate for the valuation of the benefits or contributions of members whose curtate durations were, at the date of valuation 0, 1, 2, 3, ... / years, and whose fractional exposures were, on the average, half-a-year. The cards representing the cases ^^existing^^ at the date of the valuation having been already sorted according to central ages at entry and curtate durations, the mean valuation factors, arrived at as above, were then applied to the valuation of the cases in each year of duration, regard being also given to the scale under which the members had severally subscribed. By this means the liability in respect of the death-benefit, and of the allowance during non-employment, was ascertained; and the values of the members^ annual subscriptions were similarly computed by the employment of mean annuity-values. In computing the values of the subscriptions, allowance was made for the non-payment of the monthly subscriptions by 73 members whilst in reeeipt of benefit. This was arrived at by increasing the amount of the maximum weekly allowance during non-employment_, by the amount of the average weekly subscription of the members (in each of the three scales) as a whole, increased in the proportion of actual to equivalents^ weeks of benefit. The value of the estimated net premiums was arrived at by deducting from the value of the ofiice subscriptions payable (1) the percentage available under the rules for management expenses, (2) the value of the constant deduction in respect of medical attendance, &c. Finally, by deducting the aggregate value of the estimated net premium from the aggregate values of the death benefit and the allowance during non-employment, and carefully eliminating negative values, the amount of the estimated liability was arrived at. As TO THE EFFECT OF AN ESTIMATED ALLOWANCE FOR WITH- DRAWALS UPON THE Net Premiums and Valuation Factors computed, and upon the Reserves or Net Liability. The effect of making an allowance for secessions or withdrawals, in the valuation of Benefit Societies, is a subject that does not appear to have received much attention in our published trans- actions; although it must, as I imagine, come frequently under the attention of the actuary. It is quite impossible for me, within the limits of the present paper, to discuss this interesting question at any length, or at all adequately; but I have thought that it would be of interest to append some comparative results, as to the effect of making an allowance for withdrawals, in the case of the particular benefits here dealt with (1) on the amounts of the net premium and valuation factors, (2) on the amount of the reserves, or net liability. I have assumed, for the sake of simplicity, that the benefit at death is a constant amount of J01O; that the allowance during non-employment is at the constant rate of per week, without I’eduction, but limited as to incidence and duration according to the conditions s})ecified on pages 164 and 165; and I have adopted the rate of mortality as specified in Table I; a graduated rate of withdrawal based upon the rates shown in Table II, and ceasing at the expiration of 30 years^ duration; and a rate of non-employ- ment as shown in Table IV, column (5) ; with interest throughout at 3 per-cent. 74 I have preferred, in these illustrative examples, to assume a rate of withdrawal whieh, upon the whole, is fully equal to that aetually experienced; it will, however, be understood that in the j)ractical valuation of such an Association, the rate of withdrawal assumed as likely to operate in the future should be materially below that actually obtaining in the immediate past. I have computed, and give in Table V, the valuation factors A'[a-]+o and 'vith allowance for withdrawals; also the factors A[a:]+<, and a[x]+t, without allowance for with- drawals; taking values of [,r] = 20, 30, and 40; and of / = 0, 1, 3, 5, 10, 15, 20, 25, and 30. 1 have also computed the value of 7r^[^], the net premium, with allowance for withdrawals, required to provide .€10 at death, and €1 weekly during non-employment; and of 7r[.r] the net premium, without allowance for withdrawals, com])uted to ])rovide the same benefits. These valuation factors and net ])remiums are, as might have been expected, materially reduced where the element of withdrawal is introduced, and especially in the early years of duration. In Table VI are given what may be termed the true net ])remium reserve values at the ages at entry, and after the several durations, above mentioned. Here the reserve values in columns (2), (3), and (4) are computed on the basis of a net premium and valuation factors allowing for withdrawals; while in columns (5), (()), and (7) the reserves are upon the basis of a net premium and of valuation factors without allowance for withdrawals. These reserve values will be built up -respectively by the accumulations of the net premiums as com])uted, assuming, of course, that the rates of mortality, of withdrawal, and of non-employment do not differ from those assumed. It will be seen that in these examples the reserve values are, in the early years of duration, less when the element of withdrawal is introduced; but that after between 10 and 15 years^ duration, the introduction of the withdrawals increases the reserve values throughout. It does not of course follow, that these relations would always hold good, and the comparative results would probably be modified, according to the rate of withdrawal assumed in successive years of duration, and the progression of that rate. It is, however, clear, where (as in the present examples) the rate of withdrawal is assumed to operate over a fixed term of years only, that the reserve values after the exinration of that term must throughout be greater where the element of withdrawal enters into the net premium : for the other valuation factors (the effect of withdrawal having ceased to operate) are now identical^ and the lower net premium will necessarily produee a greater reserve value. In the above examples it is assumed that the Actuary has a free hand in the selection and computation of his valuation factors, and that the office scale of subscriptions will, after a reasonable allowance for expenses and other charges, be sufficient to provide the larger net premium required where no allowance is made for withdrawals. As, however, the expenses of manage- ment and other charges will usually absorb a stated proportion of the office subscriptions, the more usual case will be that where the available net premium applicable to benefits is represented by the fixed proportion of the subscriptions remaining. Let us first take the case where this available benefit premium is found to be just sufficient to provide the risks without any allowance for withdrawals. I have computed in Table VII the comparative reserve values upon this basis, with and without allowanee for withdrawals; but always upon the assumption that the net premium valued is that available as above, and which will provide for the benefits, assuming that there are no with- drawals. Taking the columns (2), (3), and (4), and comparing the results with those set out in columns (5), (6), (7), it will be seen that the element of withdrawal introduces (upon this basis) negative values in the early years of duration, and that the reserve values are throughout diminished, where allowance is made for withdrawals. After 30 years’ duration, however, when the effeet of withdrawals ceases, the reserve values will be identical under both assumptions; as the net premium employed in this case is throughout the same in the two cases. It may be added that the reserve values set out in columns (5), (6), and (7) will be throughout built up (without allowance for withdrawals) precisely by the net premiums assumed; and that the reserve values set out in columns (2), (3), and (4) will consequently be materially less than those that would have been built up (with allowance for withdrawals) by the assumed net premiums; the difference being represented by the anticipated value of the future profit arising from withdrawals. Taking now the case where the available net premium, after providing for expenses and charges, is only just suffieient to ])r()vide the benefits granted, Avith full allowance for withdrawals, we have the results set out in Table VIII. Here the element 70 of withdrawal materially reduces the reserve values in columns (2), (3), and (4), which are througdiout less than those in columns (5), (0)^ and (7)j until the effects of withdrawals ceases^ when they are identical with those deduced without allowance for withdrawals. The values in columns (2), (3), and (4) are here precisely built up (with allowance for withdrawals) by the net premiums assumed; but these premiums would be quite insufficient to build up the greater reserve values (without allowance for withdrawals) set out in columns (5), (6), and (7); the difference representing the present value of the loss which would arise if there should in the future be no withdrawals. In the case of the Association whose experience has been under review in the earlier portion of this paper, it was found that the available office subscriptions, after allowing for expenses and medical charges, was somewhat more than sufficient to provide the benefits granted, upon the assumption that the future rate of withdrawal would be reasonably below that actually experienced. It will probably be found, in many cases, that there is a very narrow margin in this resj)ect; and that, therefore, the premium available for benefits would not be sufficient to provide reserve values computed without allowance for withdrawals. The rate of withdrawal likely to obtain in the future history of the Association thus becomes a most important element in the case; as, if from any cause the rate experienced falls materially below that assumed in the valuation, the net premiums available will no longer be sufficient to provide the necessary reserves. It therefore behoves the Actuary to exercise great caution in his assumptions as to the rate of withdrawal. If, on the one hand, he altogether ignore this element, or adopt a rate materially below that indicated by the experience, he will ])robably bring out an immediate and large (but at the same time somewhat illusory) deficiency in the funds; while if, on the other hand, he assume a rate of withdraAval practically identical with that actually experienced in the past, he runs the risk that the rate obtaining in the future may be materially below his estimates, and that a grave deficiency may thus arise at future valuations. In practice, a careful judgment must be exercised, as to the sufficiency, upon reasonable assumptions as to the rate of with- drawal, of the available net premium to provide for the risks; and the basis upon which to proceed in the computations of the reserves, must be largely determined by a careful analysis of the 77 circumstances and experience of the particular Association^ and its general financial position. As an illustration of the important financial considerations involved in these questions^ I have roughly computed the aggregate amount of reserves which would have been required, upon the different assumptions made in Tables VI, VII, and VIII, in respect of the 2,881 members actually existing, as at 31 December 1892, in the Association whose experience I have dealt with in this paper. I have taken, throughout, supposititious benefits of J01O at death, and of £l weekly during non-employment. The reserve values were computed, at quinquennial ages and durations, by means of the values given in Tables VI, VII, an d VIII, so as to give the aggregate reserves (1) with allowance for withdrawals, (a) in both valuation factors and net premium, {b) in valuation factors only; (2) without allowance for with- drawals, {a) in both valuation factors and net premium, {b) in valuation factors only. The application of these four cases to the model office^’ necessarily involves the assumption that the available benefit premiums are, throughout, sufficient to cover the net premiums computed without allowance for withdrawals. The resulting Aggregate Reserve Values are as follow: — Aggregate Reserve Values. Basis of Reserves Basis of Net Preiiiiums Aggregate Reserves Ratios per-cent Tal)lc With Allowance for Withdrawals With £8,856 100-0 VI (a) Without „ „ Without 8,901 100-5 VI (b) With Without 5,327 60-2 VII (a) Without ,, ,, With 1 14,156 159-9 VIII (5) 78 Table I. — Clerks’ Association. Graduated Itate of MorfaJHy^ as employed in computation of money values^ based upon actual experience^ 1880-1892. Ago (.r + 0 Rate of Mortality 9x+t ' (1) (-0 20 •0010 21 •0047 22 •0048 23 •0049 21 •0052 25 •0054 20 •0057 27 •0059 28 •0002 29 •0005 30 •0008 31 •0071 32 •0073 33 •0070 31 •0079 35 •0082 30 •0085 37 •0088 38 •0092 39 •0090 40 •0101 41 •0100 42 •0112 43 •0119 44 •0120 45 •0133 40 •0142 47 •0150 48 •0159 49 •0107 50 •0170 51 •0180 52 •0197 53 •0211 54 •0230 55 •0255 50 •0285 57 •0321 58 •0300 59 •0402 79 T.vble II. — Clerks’ Association. Rates of Withdraival, as actually experienced (1888—92) scheduled according to Central Ages at Entry and Years of Duration. Year of Dura- tion t Central Age . \T Entry All Entry A^^es Y(‘ar of Dura- tion t 20 25 30 35 40 : 45 Rate of VVitlulrawal 100(«y(7[a;]+9 (1) (2) Ki) (4) (5) («) (7) (B) 0^) 0 11-4 19'7 17*2 18-4 17-8 22-1 16-7 0 1 10-9 13-1 10-2 18*4 23-4 120 1 2 9-8 11-9 13*1 14*5 9-0 10-9 2 3 9-2 14*9 13-9 5-9 14-5 18*’9 12*2 3 4 111 9-6 7-2 1*3 10*5 9*4 4 5 9-2 9-2 9*7 9*8 9-8 9-7 5 6 7-2 8-6 6*8 9-2 61 10-5 8*0 6 7 5-2 6*2 3-7 7-3 7-1 5-5 7 8 6-4 6-3 4*7 4*7 12*1 8-7 61 8 9 4-1 3-7 4-3 3*2 4*3 3-8 9 10 5*3 5*1 4-7 3*6 4-8 10 11 6-0 5-6 5-3 6*3 5*8 11 12 3-9 7-9 2-6 7*3 5-2 12 13 5*9 2-2 6*4 3*0 4*3 13 14 6-8 3-9 4*0 3*9 200 5*2 14 15 1-8 5-6 2*8 5*2 3-7 15 16 6-2 3*5 7*2 4-7 16 17 4-6 4*1 2-5 17 18 4-7 2-5 4-8 3-7 18 19 1-9 3-4 13*3 2-8 19 20 6-4 4*4 4 6 20 21 2-3 4*6 3-8 21 22 22 23 20 10 23 24 24 25 1-7 2*6 7*4 2*2 25 26 2-0 5*0 3*5 30 26 27 4-3 0-9 27 28 4*4 1-4 28 29 e’V 2-3 29 80 Taiu.e hi. — Clerks’ Assoc fat ion. Bates of Non-Employjnenf^ as actually experienced ( 1888 - 92 ), scheduled ( 1 ) according to ages attained ( 2 ) according to years of duration. A-e attaineil x-\-t Rate of Non- Eiii])loynieiit «/x+< 1 Year of 1 Duration t Rate of Non- Einployinent «< (1) (-0 ' (3) 0) 20 •106 0 •087 21 •262 1 1 •245 22 •]71 2 •122 23 •236 3 •204 21 •283 4 •311 25 •139 5 •159 20 •219 6 •167 27 •175 7 •248 28 •143 8 •165 29 •151 9 •100 30 •216 10 •185 31 •127 11 •222 32 •167 12 •217 33 •141 13 •113 31 •165 14 •152 35 •151 15 •139 36 •278 16 •189 37 •217 17 •304 38 •195 18 •349 39 •283 19 •177 40 •125 20 •012 41 •109 21 •290 42 •143 22 •061 43 •196 23 44 •171 24 •013 45 •139 25 •206 46 •341 26 •158 47 •509 27 •508 48 •360 28 •383 49 •117 29 •864 50 •269 51 •066 52 •709 53 54 •255 55 •141 i! 56 •660 i 57 i ^535 i It ... 58 ! ! j 59 !!! i il i: 60 •680 61 62 ! 63 •300 64 2-787 i 81 Table IV. — Clerks’ Association. Graduated Bates of Non-Employment^ based upon the Experience of successive periods 1862-1879, 1880-87, and 1888-92; also the rate adopted for computation of 3Ioney Values. Age attained (0 Period of Experience Rate einj)loyed in computation of Money Values Age attained (0 1862-1879 1880-1887 1888-1892 Rate of Non-Employment ( ux + t ) (1) (2) (3) (4) (5) (6) 20 •266 •259 •214 •252 20 21 •268 •256 •213 •250 21 22 •268 •258 •212 •249 22 23 •263 •260 •211 •249 23 24 •250 •264 •209 •249 24 25 •243 •269 •199 •248 25 26 •232 •270 •192 •246 26 27 •226 •269 •182 •242 27 28 •228 •266 •172 •237 28 29 •232 •259 •166 •231 29 30 •238 •248 •165 •226 30 31 •247 •241 •160 •219 31 32 •256 •231 •163 •214 32 33 •260 •224 •169 •211 33 34 •260 •216 •177 •210 34 35 •256 •217 •189 •213 35 36 •261 •223 •203 •220 36 37 •265 •239 •204 •229 37 38 •269 •253 •200 •238 38 39 •270 •277 •194 •244 39 40 •281 •288 •179 •249 40 41 •291 •292 •166 •250 41 42 •308 •289 •168 •252 42 43 •341 •284 •188 •256 43 44 •385 •280 •215 •261 44 45 •421 •277 •244 •268 45 46 •435 •275 •277 •276 46 47 •446 •274 •290 •284 47 48 •433 •269 •297 •293 48 49 •448 •252 •283 •303 49 50 •408 •249 •276 •316 50 51 •274 •259 •335 51 52 •344 •277 •358 52 53 •440 •280 •384 53 54 •585 •297 •416 54 55 •719 •298 •454 55 56 •820 •322 •499 56 57 •841 •305 •538 57 58 •845 •292 •577 58 59 •774 •271 •616 59 60 •733 •303 •655 60 61 •295 •710 61 62 ... •352 •780 62 63 •496 •870 63 64 •682 •980 64 F 82 Table V. — Clerks’ AssociATioif. Values of Assurances at Death, of Allowance during Non- Dmployment, of Annuities, and of Net Premiums, based upon the Rate of JSIortalitg as shotun in Table I; a Graduated Pate oj Withdrawal (ivhere assumed'), based upon Table 11 ; and the Pate of Non- Employment as shown in Table IV column (5), with Interest at 3 per-cent. The Net Premiums are computed {icith and without Withdrawals) to provide Penefits of £^1.0 at Death, and of £\ tveeklg during Non-Kmplogment. Duration With Allowanck fou W lTHDUAWALS Without Allowance fou Withdrawal'* Duration Value of Assuranee of 1 at Death Value of Allowance of 1 Weekly (luring Noii-Eiii- ployiueiit Value of Anmiity of 1 ixjr annum Val ue of Assurance of 1 at Death Value of Allowance of 1 Weekly during Non-Eni- i iiloyiuent 1 Value of Annuity of 1 per annum (0 A'|x]+< ] — ■‘156 Age at Entry [ar]=30 7q3*>j = -585 0 •117 2-30 7-61 '413 7-60 19-86 0 1 •144 2'85 8-72 •422 7-65 19-57 1 3 •192 3-62 10-27 •439 7-78 18-99 3 5 '238 4'33 11-36 '457 7-95 18-39 5 10 •332 5-63 12-21 •505 8-38 ]6 77 10 15 •437 7-03 12-44 •557 8-87 15-01 15 20 •538 8-38 11-85 •611 9-50 13'16 20 25 •636 9-65 10-69 •669 10-13 11-19 25 30 •723 10-72 9-39 •723 10-72 9-39 30 ’*■'[40] = ‘606 Age at Entry [a-] =40 7r[4o] = -794 0 •154 2-62 6-86 •505 8-38 16-77 0 1 •189 3-26 7-78 •515 8-46 16-42 1 3 •250 4-17 8-97 '535 8-65 15-73 3 5 •307 5-00 9-73 '557 8-87 15-01 5 10 •420 660 9-92 •611 9-50 13-16 10 15 •545 8-26 9-48 '669 10-13 11-19 15 20 •651 9-62 8-55 '723 10-72 9-39 20 25 25 30 30 83 Table VI.— Clerks’ Association. Comparison of Heserve Values for Benefts of £10 at death, and of £1 weekly during Non-Employment ; computed upon the basis of the factors given in Table V, respectively with and without allowance for Withdrawals. Duration ( a ) With Allowance FOR Withdrawals ( b ) Without Allowance FOR Withdrawals Duration Age at Entry Ag-e at Entry (/) 20 30 i 40 20 30 40 (0 (1) .3) (4) CO (7) («) 1 •24 •31 •43 •29 •42 •57 1 3 •56 •86 1-23 •69 1-06 1-51 3 5 •95 1-53 2-17 1-11 1-76 2-52 5 10 2-12 3-38 4-79 2-28 3-62 5-16 10 15 3-85 5-73 7-96 3-77 5-66 7-91 15 20 5-83 8-36 10-95 5-45 7-91 10-49 20 25 8-01 11-13 7-29 10-27 25 30 10-38 13-67 9 34 12-46 30 ’Tlxl (as valued) •397 •456 •606 •476 •585 •794 nx ] (as valued) Table VII. — Clerks’ Association. Comparison of Reserve Values for Benefits of £10 at death, and of £1 weekly during Non-Employment ; computed upon the basis of the factors given in Table V; the net premiums employed being in all cases those computed without allowance for Withdrawals. Duration (o) With Allowance FOR Withdrawals (b) Without Allowance FOR Withdrawals Duration Ag-e at Entry Age at Entry (0 20 30 40 20 1 30 40 (/) (1) C) CO (4) (5) (6) (T) (S) 1 (- -57 ) (- -81 ) (-1-03 ) •29 •42 •57 1 3 (- -36 ) {- --I'S ) (- -45 ) •69 1-06 1-51 3 5 (- -06 ) •05 •34 1-11 1-76 2-52 5 10 1-01 1-81 2-92 2-28 3-62 5-16 10 15 2-66 4-15 6-18 S-77 5-66 7-94 1 5 20 4-65 6-80 9-34 5-45 7-91 10-49 20 25 6-88 9-75 7-29 10-27 25 30 9-31 12-40 9-34 12-46 30 "^Ix] (as valued) •476 •585 •794 •476 •585 I •794 '^[X] (as valued) F 2 84 Table VIII. — CLEitKs’ Association. Comparison of JRcserve Values for Benefts of £10 at death and of £1 weekly during Non- Employment ; computed upon the basis of the factors given in Table V ; the net premiums employed being in all cases those computed with allowance for Withdrawals. , Duration 1 (0 ( a ) With Allowance FOR Withdrawals Q >) Without Alia)Wance FOR AVithdrawals Duration (0 Age at Entry Age at Entry 20 1 30 40 20 1 30 40 (0 (2) C ) (4) (5) 63) U) (8) 1 1 •24 •31 •43 2-05 2-94 3-66 1 *56 '86 1-23 2-41 3-51 4-47 3 ! 5 '95 1-53 2'17 2-78 4-13 5'34 5 i 10 2-12 3-38 4'79 3-84 5-78 7-63 10 1 15 3-85 5'73 7-96 5'22 7'59 10-04 15 1 20 5-83 8-36 10-95 6-77 9-61 12-26 20 ' 25 8-01 11'13 8-48 11-72 25 j 30 10-38 , 13-67 10-38 13-67 30 (as valued) •397 : •456 •606 •397 •456 •606 ’Tlx] (as valued) APPENDIX (A). Period of Observation Limited by Calendar Years. Exact Duration FoRMULiE. (Schedule B.) We have^ for the number exposed to risk at age [oc], [X] = -I- (^ — ^0 M — (^ + + d) + [e -Cw' -\-d') = ^[x] + (^ -/) [.rl - («' -/ ') [xj = ^[x]+i7[.r]"“.^^[x] (1) also for the number exposed at age + l E[x ]+1 = E[a;] + 9 {x\+ 9 [_x]+\ — = E£^j+^[a;]+i — (2) and^ generally = 9[x\+t — ^9\^']+t-'^ ' • • (^) If now M e insert in formula (2) tbe value of as given in formula (l),we have: E[a;] + 1 = n^x-\ + y[x] +^[X] + 1 —9\x-^ ~ ^9\x-\ — ^[X] -k 9 W +yM+i —9\x]+i = 7Z[x] + Sj(^) — /[x]+l and, generally 'Ei^^+t = ri^.r^-\-^[A9)-9'm+t • • • • (4) "4iere So (^) = (^[x]+^M+i+ • • . • +^w+i) 85 Mean Duration Formulae. (Schedule C.) Here , s , e , w d *=3’"=2’“'=3>'^=3’ also. , , e^-w + d f f=e+w' + d= 2 =2 and ^ I 3 3' Inserting* these mean values in formulse (1) (3) and (4) respectively, we have : — + (5) ... ( 6 ) EM+« = «M + 2„'(y)--^ (7) Nearest Duration Formula^^ as applied (in Schedule D) FOR deducing either THE RaTE OF MORTALITY OR THAT OF M^ITHDRAWAL. {a) Rate of Mortality. Let s, e and w_, represent the numbers surviving*^ existing and withdrawing, as tabulated by this method; so that we have, for initial values: Ste] = {as) e^.] = {ae) Wf^j = {aw)^cc] and for subsequent values, + ^ 1 "b (^^)[a?] + if] W[a’] + ^ — [ (6 ?u)[aO + ^ _ 1 + (« W) + 1\ the symbols a and b, representing throughout the reference to the beginning and the end respectively of the year of duration. Also let f*[x]+^=(e + w)[a;]+^ + 5)[a;] — [(6c) + {bw) + {bd)\x] + (<5f5)[a7] + i — [(ffe) + {aw) + («C?)][a.*]+i — E[a.]+ (S — f)[a:] + i = E[a^] + g[a'] + l and generally E[ar ]4 ^ = E[x]+^-i + g[a-]+^ .... (10) a formula for deducing the numbers exposed to risk by a continuous method. 88 Inserting now in formula (9) the value of Eui from formula (8), we have: + i = n^x] + g[.r] + g[a:] + l and generally = + (11) a formula for deducing the numbers exposed to risk by a process of summation, as applied in Schedule (E) . If now we analyze this formula, we have: E[a;] + <=W[*r]+ (g[a;] + g[.r]+l + • • • • + gM + ^) = ^[a;] + [(^^)[a’]+ (^^)[a’]+(«5)U’]+l+ .... H- («^)[a’] + J — [(«e)[.r]4- (^^)[-r] + l + • • • • + («^)[a,-] + ^] — + .... + — [(fff/)[a.]+ («^)[.r] + l + .... + (^^)[.r] + ^] = //[.rj + -o^Cs‘) — (/>'«) + ^ — = n[x] + '^o^{g) — {fj^)[x]+t (12) Comparing this now with formula (4) of Exact Durations, E[x] + ^= ?i[.r] + 2o’^(^) [x] + t .... (4) we see that the two formulje differ only in their last terms, and that the difference between the numbers exposed to risk, as deduced by the Nearest and by the Exact Duration Methods, solely arises from the error introduced in the year of duration at entry (for survivors), or at exit (for cases emerging), by the assumption that (bg)[a:]+t} the balance of nearest integral durations in the year, is a])proximately equal to g'[x]+t (see pages 29, 30), the balance of true fractional exposures in the year. APPENDIX (B). Period of Observation Limited by Years of Duration. Exact Duration Formula. (Schedule G). Here become e\:c]+< (the numbers sur- viving and existing at precise duration t), and, as the new entrants, during the period (72[^]) now fall into rank as ^^survivors^^ at precise age [.r], (6-^[a;]), and, as there are now no cases existing at precise age [,z], we have e\,;] = 0. 89 Representing by G and G' respectively^ the modified values of g and g under these conditions^ we have — d)[x]} G^x]+t=s^{x]+t— d\x]+t) also, as 5^[a?]+^ = 0, c%]+^ = 0, we have G\x^+t— — + d')ix]+t for all values of t. Then formula (1) becomes Et,i=G[,]-GV]=G'. + K + ^')u-] . . • . (13) Similarly, formula (3) becomes l^[a-]+^=E[^]+#-i + LG\x]+t-\ = ^[x]U-\'^ ^[x]+t + d\x]^-t-\ • (Id) also formula (4) becomes E[a'] + ^ = ^0^(G) — G\x] + t = ^,\G)+{w'-Vd'\x^^t (15) These formulae give the numbers exposed to risk (E), from which the usual functions (E) and [wE) can be deduced. It is however, in this case equally convenient to deduce these latter functions by a direct process. For we have E = E + {d'-d) and E = {wE) -{-[w' — id ) , and inserting these values successively in formulae (13), (14) and (15) we have for the numbers exposed to the risk of death. E[.r] — G[j-] + [w' + d)[x-\ (16) E[a;]+^=E[a;]+^_i+ G[a;]+^ + A(^/;' + . . (17) and = + d\x]+t (18) Similarly, for the numbers exposed to the risk of withdrawal, we have {wE)^x\— G^x]-^ d')]^x] (19) [wE)[x]+t— ['^E)^x\ + t-\+ G^[.r] + e + A( 2 C + fl(')[j.]^.^_l (20) {wE)[x\Jrt=-'^Q{G) 4- {'^ + d')^x]->rt (^il) ^Iean Duration Formulae. (Schedule H.) Here w'='^, d'=~, and formulae (16), (17) and (18) become respectively 90 + ( 22 ) E[a:] + < = E[a:]4.^_l + A^— + + . . (23) E[.]+.=2o^(G) + g+^).]+. (24) Similarly, for the numbers exposed to the risk of withdrawal, formuhc (19), (20) and (21) become respectively {W E)[a:] = G[a;] + g + (25) = + G^[a-] + ^ + ^g + ^^[.r] + ^_i ( 26 ) {w E ) [a-] + ^ — ^ 0 ^ ( G) H- ^ lu + [.I-] (27) Nkauest Duration Formulae. (Schedules J, K.) Reverting to, and analyzing formula (11) we have l^V]+^=”[j-] + So^(g) = nix] + — o^[s — (e + w + d) ] which becomes, when the observation is limited by years of duration (since s = 5', e = c’, and 7i[^.] = 5’[a-]), E[a -] + 1 = [ 5 * - (e> + w + d) ] , which we may call So^(G) ; therefore E,.h,-V(G) (28) but, as E[a^]+^ = E[.^]+^— (««?)M+i; = {wE)ix]^t — {aw\x]^t we obtain the following direct formulae for the numbers exposed to the risk of death and of withdrawal respectively E[a:]+^ = "o^(G) + («<^)m+^ (29) and (tcE)[.r]+i 5 = So^(G) + («^c)[a.]+i 5 (30) Here it may be shown that 2o^(G) = So^(G)+[(^.^c) + (^^)][,H^, and the above formula3 thus become E[a?]+i{ = 2!o^(G) + [(/^w) +<^][a:] + ^ (31) and {wE)ix]+t = ^Q[G) + \w+{bd)']ixut .... (32) which may be compared with the Exact Duration formulae (18) and (21), the quantities w' and d' being replaced, in formulae (31) and (32), by [bw) and (bd), respectively. 91 Appendix (C). The values of and of as deduced from that of E[^j+^ by the several formulae (see pages 5, 15, and 18) E[a:]+^ = E[a’]+^ + {^d {^wE\x]+t—^[x-\+t'\‘ (j^ ^ )[•!’]+ ^ Eia.i+i=K,^,+«+ {wE)^xnt=^m+t-{- ~~2~ E[a7]+if — Ej-ipj+^d- {ad^\x]+t (^wE\x^+t — E|-j.]+^d- (^aw)[x-^+t according to the Exact, Mean, and Nearest Duration Methods respectively, do not appear, in the particular case here investigated, to represent quite correctly the true relations. As the period of observation terminates at the close of a calendar year, the cases “existing” are necessarily under observation for a portion only of the year of duration then current, and some of the cases of death (and of withdrawal) during the last calendar year, would, if treated as “existing”, have in like manner completed only a portion of the year of duration current at exit. In strictness, therefore, such cases should contribute to the number exposed to risk, not the full year of duration current at exit, but only that portion of the year which actually fell within the period of observation. The effect of the assumptions made, and the nature and extent of the correction required, will be seen if we suppose that 1,000 members complete their year of duration in the last calendar year of the period of observation, on 80 June; and that of these 1,000 cases none withdraw, but five die during the following six months. The number exposed to risk in respect of the 995 survivors would clearly be 995 = -^=497‘5; and to this would be added, by the above formulie, a full year’s exposure in respect of each of the five deaths, making the total number exposed to risk 502‘5. The annual rate of mortality deduced in respect of these cases only would thus be 502-5 = •00995; but it would seem to be clear that the true annual rate of mortality = •01, and that the number exposed to risk should, in strictness, be 497-5 + 2-5 = 500; that is, that the death cases should be considered as exposed to risk for six months each only. The cases affected are those of death (or withdrawal) occurring in the last calendar year of the period of observation, where the anniversary of entry in that year precedes the date of death (or with- drawal). Assuming that this would include about one-half of the number of cases emerging in the year ; that the number of emergents in each year of a quinquennial period do not greatly vary ; and that, upon the average, two-thirds of the year of duration current at exit fiills, in these cases, within the period of observation ; the amount by which the exposures of the cases of death (or of withdrawal) would be overstated, may be roughly estimated at about one-thirtieth part (or per-cent) of the total number of deaths (or withdrawals) occurring in the quinquennium. 92 I find that, of the 481 withdrawals included in the illustrative experience employed in this paper, 34 occurred in the calendar year 1892, subsequently to their anniversaries of entry in that year ; and that, of these 34 cases, the actual portion of the year of duration current at exit which fell within the period of observation, was, in the aggregate 25*5 years. Of the 48 deaths, seven occurred in the calendar year 1892, subsequently to their anniversaries of entry in that year ; in respect of which the actual portion of the current year of duration which fell within the period was 5 7 years. The number exposed to the risk of withdrawals was thus overstated by (34 — 25 5 = ) 8 5 years ; and the number exposed to the risk of death, by (7 — 5'7 = ) 1*3 years. The aggregate years of risk being 6,660 and 6,458 respectively, the amount of the error was (in this particular case) about 13 per 10,000 in the numbers exposed to the risk of withdrawal, and about two per 10,000 in the numbers exposed to the risk of death. The correction is, therefore, extremely small, and may, in practice, be disregarded. ApeENOix (D) (see i)age 45). In the case of an experience where the value of tends to increase steadily with the age, the probability will be for the death cases to congregate towards the end of the year of duration, if we assume that the progression of the decrements in half-t/ears follows generally the same law as their annual progression. In the Table, for example, the value of cl increases between ages 14 and 22, and between ages 25 and 74, after which it steadily diminishes until the end of the table ; and with such an experience, it would appear to be more probable that death will happen (on the whole) in the later half of the year of duration. Under the Table the same tendency is observable, but in a less strongly marked form ; the value of d increasing between ages 10 to 16, 19 to 28, 39 to 45, 49 to 76, after which it steadily diminishes. Appendix (E). The formuliB given on pp. 72, 73, for the values of ^'and of are closel}^ approximate, the true formulae being, upon the assumption of a uniform distribution of deaths throughout the year, ^ 2 q+{wq)~q.{wq) _ q.jwq) _ q.(wq) 4 4 These formulae are deduced directly from those given by Dr. Sprague (J LA. j xxi, 416-7) for computing the elements of the mortality table in the case of a double decrement, which, expressed in the notation here employed, become (icqy d=l. 1 - q.(icq) 1 + m-\- (wm) 4 2 93 4 "^2 q-\-{ioq)-q{ivq) 1- q.{wq) 4 1 - = Z. 1 + 2 ~ 2 The alternative expressions given on the right-hand side are con- venient for deducing the mortality elements in terms of the central rates of death and of withdrawal; and, in the case of an experience such as that here investigated, it is, on the whole, more convenient to compute the values of and of {wm^^x^+t direct from that of and thence to deduce the values of \x\^u and tv^x^^t- The values of and of {wE)^x]+tt as well as those of q[x\+u and of {wq)[x^+f, can then be dispensed with. It will be convenient to call m m-\- {lum) (wm)f.^,J+^= (ivm') 1 + 711 {wm) 1 + 2 1 - 711 4- {tvill) and 1 — [m[ar]+^+(wm)ta, 3 +,f]: then we have, 14- 711 -{■ {WTll) and ^[x]+^[l (ce[a:*]+^4' (wm)[-.r]+j()] and by these formulje the values of d, w, and Z, can be readily computed. The appended Table IX shows, for age at entry (20), the central rates of mortality and of withdrawal, the logarithms of the factors m, (wm), and {1 — [m+ (wm)]}, respectively, and the numbers living, dying, and withdrawing, as deduced in each successive year of duration by the above formulsB. The rate of withdrawal here employed is a graduated rate, based upon that experienced in years of duration, as set out in column (8) of Table (II). In Table X the Commutation columns are deduced, at a rate of interest of 3 per-cent per annum, from the mortality elements given in Table IX; also the values of assurances and annuities, and of the benefit during non-employment; with allowance, throughout, for mortality and withdrawal. 94 Table IX. — Mortality Table. Table showing the Central Rates of Mortality and of Withdrawal^ and the Mortality Table (^with decrements by Death and Withdrawal) as deduced therefrom. — Central Age at Entry [a7] = 20. Dura- tion Central Rate of I.OOAUITH.M.S OF FACTORS FOR Deducing Mortality Table Mortality Table .Mortality With- drawal Numbers Living Deatlis With- drawals Deaths and With- drawals (/) ^n\x]+i log log ! log iTl — (m-fwiii) [*]+< hx]+t {d -f «’) W+t (1) (-9 (3) (i) (3) (6) (7) (8) ( 9 ) (10) 0 •00461 •16220 7-62891 9-17526 9-92739 100,000 426 14,970 15,396 1 •00471 •13330 •64404 •09585 •93997 84,604 373 10,549 10,922 2 •00481 • 1 1080 •65775 •02014 •94974 73,682 335 7,717 8,052 3 •00491 •09424 •67007 8-95323 •95691 65,630 307 5,892 6,199 4 •00521 •08333 •69803 •90199 •96153 59,431 297 4,741 5,038 5 •00541 •07792 •71547 •87392 •96379 54,393 282 4,070 4,352 (3 •00572 •06718 •74185 •81169 •96832 50,041 276 3,244 3,520 7 •00592 •05656 •75896 •73915 •97286 46,521 267 2,551 2,818 8 •00622 •05128 •78148 •69764 •97502 43,703 264 2,179 2,443 9 •00652 •01604 •80298 •65187 •97716 41,260 262 1,852 2,114 10 •00682 •04604 •82245 •65181 •97704 39,146 260 1,756 2,016 11 •00713 •04604 •84170 •65175 •97691 37,130 258 1,66 4 1,922 12 •00733 •04604 •85366 •65170 •97681 35,208 251 1,580 1,831 13 •00763 •04082 •87214 •60018 •97896 33,377 219 1,329 1,578 14 •00793 •04082 •88881 •60041 •97882 31,799 246 1,268 1,514 15 •00823 •03562 •90598 •54227 •98095 30,285 244 1,056 1,300 16 •00854 •03046 •92307 •47534 •98306 28,985 243 865 1,108 17 •00884 •02532 •93909 •39610 •98516 27,877 242 695 937 18 •00924 •02532 •95823 •39602 •98499 26,940 245 670 915 19 •00965 •02020 •97809 •29891 •98703 26,025 248 518 766 20 •01015 •02020 •99993 •29881 •98682 25,259 253 502 755 21 •01066 •02020 8-02111 •29870 •98600 24,504 257 487 744 22 •01126 •02020 •04176 •29857 •98033 23,760 263 474 737 23 •01197 •01511 •07225 •17342 •98824 23,023 272 343 615 24 •01268 •01511 •09712 •17326 •98792 22,408 280 334 614 25 •01339 •01511 •12064 •17312 •98763 21,794 2S8 324 612 26 •01430 •01511 •14900 •17292 •98723 21,182 298 316 614 27 •01511 •01005 •17383 7-99674 •98907 20,568 307 204 511 28 •01603 •01005 •19930 •99654 •98867 20,057 317 200 517 29 •01684 •00501 •22162 •69512 •99051 19,540 325 97 422 30 •01776 •24560 •99229 19,118 337 31 •01877 •26940 •99185 18,781 349 32 •01990 •29455 •99136 18,432 363 33 •02132 •32119 •99075 18,069 381 34 •02327 •36177 •98989 17,688 407 35 •02583 •40655 •98879 17,281 440 36 •02891 •45482 •98744 16,841 480 37 •03262 •50645 •98583 16,361 525 38 •03728 •56346 •98381 15,836 580 39 •04102 •60418 •98218 15,256 613 95 Table X. — Values of Benefits. Table showing the Commutation Columns, allowing for Mortality and Withdrawal, and the values of Annuities, Assurances and Allowance during Non-Employment, as deduced therefrom. — Central Age at Entry [_x] = 20. — Interest at ^ per-cent per annum. Dura- tion Commutation Columns, Allowing for Mortality and Withdrawal Values of Annuity and Assurance Rate of Non-Em- 1 ployment Commutation Columns, Allowing FOR Mortality AND Withdrawal Values OF Allow- ance (0 W[x]+t N "[ a : i +« C[_x\+t u'\x]+t V[a-]+i 2(1)"^) ( i ) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) 0 55,368* 488,939* 228*7 5,721*0 8*831 *104 •252 6,3530 146,845* 2*652 1 45,479* 443,460* 194*5 5,492*3 9 751 •121 •250 10,492*0 140,442* 3*088 2 38,454* 405,006* 169*8 5,297*8 10*532 *138 •249 8,927*7 129,949* 3*379 3 33,254* 371,752* 151*0 5,128*0 11*170 *155 •249 7,780*0 121,022* 3*640 4 29,236* 342,516* 141*6 4,977*0 11*716 *170 •249 6,874*2 113,242* 1 3*873 5 25,978* 316,538* 131*0 4,835*4 12*185 •187 •248 6,098*6 106,368- 4*095 6 23,204* 293,334* 124*3 4,704*4 12*642 *203 •246 5,430*2 100,269* 4*321 7 20,944* 272,390* 116*7 4,580*1 13*006 •219 •242 4,845*6 94,839* 4*528 8 19,102* 253,288* 112*1 4,163*4 13*260 *234 •237 4,338*5 89,993* 4*711 9 17,509* 235,779* 108*0 4,351*3 13*467 *249 *231 3,885*2 85,655* 4-892 10 16,128* 219,651* 104*0 4,243*3 13*620 *263 •226 3,500*7 81,770* 5*070 11 14,852* 204,799* 100*1 4,139*3 13*790 •279 *219 3,123*6 78,269* 5*270 12 13,673* 191,126* 94*8 4,039*2 13*979 *295 *214 2,809*6 75,145* 5*496 13 12,584* 178,542* 91*0 3,944*4 14*188 *313 *211 2,555*7 72,336* 5*748 14 11,640* 166,902* 87*5 3,853*4 14*339 *331 •210 2,352*4 69,780* 5*995 15 10,763* 156,139* 84*1 3,765*9 14*507 *350 •213 2,211*6 67,428* 6*265 16 10,001* 146,138* 81*3 3,681*8 14*613 *368 •220 2,127*3 65,216* 6-521 17 9,338*1 136,800* 78*8 3,600*5 14*650 *386 •229 2,072-4 63,089* 6*756 18 8,761*7 128,038* 77*3 3,521*7 14*614 *402 •238 2,020*6 61,016* 6*964 19 8,217*5 119,821* 75*8 3,444*4 14*581 *419 •244 1,947-2 58,996* 7*179 20 7,743*4 112,078* 75*2 3,368*6 14*474 *435 •249 1,872*0 57,048* 7*369 21 7,293*1 104,785* 74*3 3,293*4 14*367 *452 •250 1,769*9 55,176* 7*566 22 6,865*6 97,919* 73*9 3,219*1 14*262 *469 •252 1,666*3 53,406* 7-743 23 6,459*1 91,460* 74*1 3,145*2 14*160 •487 •256 1,608*0 51,740* 8*010 24 6,103*4 85,356* 74*1 3,071*1 13*985 *503 •261 1,548*6 50,132* 8-214 25 5,763*0 79,593* 73*9 2,997*0 13*811 •520 •268 1,501-0 48,584* 8*430 26 5,438*1 74,155* 74*4 2,923*1 13*635 •537 •276 1,457*9 47,083* 8*658 27 5,126*7 69,029* 74*3 2,848*7 13*465 •556 •284 1,417*2 45,625* 8*899 28 4,853*7 64,175* 74*6 2,774*4 13*222 *572 •293 1,384-5 44,207- 9*108 29 4,591*0 59,584* 74*2 2,699*8 12*979 •588 •303 1,356-2 42,823- 9*328 30 4,360*9 55,223* 74*5 2,625*6 12*663 *602 •316 1,346-2 41,467- 9*509 31 4,159*4 51,064* 75*1 2,551*1 12*277 *613 •335 1,360*6 40,121- 9*646 32 3,963*2 47,100* 75*8 2,476*0 11*884 •625 •358 1,384*6 38,760- 9*780 33 3,772*0 43,328* 77*3 2,400*2 11*487 *636 •384 1,412*5 37,375- 9*909 34 3,584*9 39,743* 80*1 2,322*9 11*086 *648 •416 1,453*0 35,963- 10*032 35 3,400*4 36,343* 84*2 2,242*8 10*688 *660 •454 1,502*2 34,510- 10*150 36 3,217*3 33,126* 89*0 2,158*6 10*296 •671 •499 1,559-8 33,008- 10*260 37 3,034*5 30,091* 94*6 2,069*6 9*916 *682 i -538 1,583*3 31,441- 10*363 38 2,851*5 27,240* 101*3 1,975*0 9*553 *693 •577 1,592*2 29,865- 10*473 39 2,667*2 24,573* ' 104*1 1,873*7 9*213 *702 *616 1,587*0 28,272* 10*600