XXIV .
On the nature of the function expressive of the law of human mortality , and on a new mode of determining the value of Life Contingencies .
In a Letter to FRANCIS BAILEY , Esq. F. R. S. &amp;c. By BENJAMIN GOMPERTZ , Esq. F. R. S. Read June 16 , 1825 .
DEAR SIR , THE frequent opportunities I have had of receiving pleasure from your writings and conversation , have induced me to prefer offering to the Royal Society through your medium , this Paper on Life Contingencies , which forms part of a continuation of my original paper on the same subject , published among the valuable papers of the Society , as by passing through your hands it may receive the advantage of your judgment .
I am , Dear Sir , yours with esteem , gth June x825 .
BENJAMIN GOMPERTZ .
CHAPTER I. ARTICLE 1 .
IN continuation of Art .
2 .
of my paper on the valuation of life contingencies , published in the Philosophical Transactions of this learned Society , in which I observed the near agreement with a geometrical series for a short period of time , which must pervade the series which expresses the number of living at ages in arithmetical progression , proceeding by small intervals of time , whatever the law of mortality may be , provided the intervals be not greater than certain limits : I now call the reader 's attention to a lavw observable in the tables of mortality , for equal intervals of long periods ; and adopting the notation of my former paper , considering L to express the number of living at the age x , and using x for the characteristic of the common logarithm ; that is , denoting by x(L ) the common logarithm of the number of persons living at the age of x , whatever x may be , I observe that if x ( L ) x ( L + ) ( L+ -(L ) nl n +ml v m\ 71 + 2W/ m X ( L+ zm)-x L 3s ) , &amp;c. be all the same ; that is to say , if the differences of the logarithms of the living at the ages n , n m ; n+ m , n m ; n+ m , n+ m ; &amp;c. be constant , then will the numbers of living corresponding to those ages form a geometrical progression ; this being the fundamental principle of logarithms .
Art .
2 .
This law of geometrical progression pervades , in an approximate degree , large portions of different tables of mortality ; during which portions the number of persons living at a series of ages in arithmetical progression , will be nearly in geometrical progression ; thus , if we refer to the mortality of DEPARCIEUX , in Mr. BAILY'S life annuities , we shall have the logarithm of the living at the ages 15 , 25 , 35 , 45 , and 55 respectively , 2,9285 ; 2,88874 ; 2,84136 ; 2,79379 ; 2.72099 , for A ( L ) ; x ( L ) ; x ( L ) ; &amp;c. and we find ( L)X ( L)= , 04738 x ( L)-x ( L)= , 04757 , and con\25/ 35 35 5 sequently these being nearly equal ( and considering that for small portions of time the geometrical progression takes place very nearly ) we observe that in those tables the numbers of 514 iving in each yearly increase of age are from 25 to 45 nearly , ill geometrical progression .
If we refer to Mr. MILNE'S table of Carlisle , we shall find that according to that table of mortality , the number of living at each successive year , from 92 up to 99 , forms very nearly a geometrical progression , whose common ratio is 4 ; thus setting out with 75 for the number of living at 92 , and diminishing continually by .
, we have to the nearest integer 75 , 56 , 42 , 32 , 24 , 18 , 13 , lo , for the living at the respective ages 92 , 93 , 94 , 95 , 96 , 97 , 98 , 99 , which in no part differs from the table by -jIth part of the living at 92 .
Art .
3 .
The near approximation in old age , according to some tables of mortality , leads to an observation , that if the law of mortality were accurately such that after a certain age the number of living corresponding to ages increasing in arithmetical progression ; decreased in geometrical progression , it would follow that life annuities , for all ages beyond that period , were of equal value ; for if the ratio of the number of persons living from one year to the other be constantly the same , the chance of a person at any proposed age living to a given number of years would be the same , whatever that age might be ; and therefore the present worth of all the payments would be independent of the age , if the annuity were for the whole life ; but according to the mode of calculating tables from a limited number of persons at the commencement of the term , and only retaining integer numbers , a limit is necessarily placed to the tabular , or indicative possibility of life ; and the consequence may be , that the value of life annuities for old age , especially where they are deferred , should be deemed incorrect , though indeed for immediate annuities , where the probability of death is very great , the limit of the table would not be of so much consequence , for the present value of the first payment would be nearly the value of the annuity .
Such a law of mortality would indeed make it appear that there was no positive limit to a person 's age ; but it would be easy , even in the case of the hypothesis , to show that a very limited age might be assumed to which it would be extremely improbable that any one should have been known to attain .
For if the mortality were , from the age of 92 , such that i of the persons living at the commencement of each year were to die during that year , which I have observed is nearly the mortality given in the Carlisle tables between the ages 92 and 99 , * it would be above one million to one that out of three millions of persons , whom history might name to have reached the age of 92 , not one would have attained to the age of 192 , notwithstanding the value of life annuities of all ages above 92 would be of the same value .
And though the limit to the possible duration of life is a subject not likely ever to be determined , 'even should it exist , still it appears interesting to dwell on a consequence which would follow , should the mortality of old age be as above described .
For , it would follow that the non-appearance on the page of history of a single circumstance of a person having arrived at a certain limited age , would not be the least proof of a limit of the age of man ; and further , that neither profane history nor modern experience could contradict the possibility of the great age of the patriarchs of the scripture .
And that if any argument can be adduced to prove the necessary termination of life , it does not appear likely that the materials for such can in strict logic be gathered from the relation of history , not even should we be enabled to prove ( which is extremely likely to be the state of nature ) that beyond a certain period the life of man is continually becoming worse .
Art .
4 .
It is possible that death may be the consequence of two generally co-existing causes ; the one , chance , without previous disposition to death or deterioration ; the other , a deterioration , or an increased inability to withstand destruction .
If , for instance , there be a number of diseases to which the young and old were equally liable , and likewise which should be equally destructive whether the patient be young or old , it is evident that the deaths among the young and old by such diseases would be exactly in proportion of the number of young to the old ; provided those numbers were sufficiently great for chance to have its play ; and the intensity of mortality might then be said to be constant ; and were there no other diseases but such as those , life of all ages would be of equal value , and the number of living and dying from a certain number living at a given earlier age , would decrease in geometrical progression , as the age increased by equal intervals of time ; but if mankind be continually gaining seeds of indisposition , or in other words , an increased liability to death ( which appears not to be an unlikely supposition with respect to a great part of life , though the contrary appears to take place at certain periods ) it would follow that the number of living out of a given number of persons at a given age , at equal successive increments of age , would decrease in a greater ratio than the geometrical progression , and then the chances against the knowledge of any one having arrived to certain defined terms of old age might increase in a much faster progression , notwithstanding there might still be no limit to the age of man .
Art .
5 .
If the average exhaustions of a man 's power to avoid death were such that at the end of equal infinitely small intervals of time , he lost equal portions of his remaining power to oppose destruction which he had at the commencement of those intervals , then at the age x his power to avoid death , or the intensity of his mortality might be denoted by a q , a and q being constant quantities ; and if L. be the number of living at the age x , we shall have a Lx x q.x for Lx the fluxion of the number of deaths= ( Li ) ; , a bq X-_L .
abq " =h y p. log .
of bxhy p. log .
of Lx , and putting the common logarithm of bx square of the hyperbolic logarithm of lo c , we have c.qx common logarithm of Ldd being a constant quantity , and therefore L or the number of persons living at the age of xr= d.giq ; g being put for the number whose common logarithm is c. The reader should be aware that I mean ?
gxq to represent g raised to the power q and not gq raised to the x power ; which latter I should have expressed by g l , and which would evidently be equal to gq .
I take this opportunity to make this observation , as algebraists are sometimes not sufficiently precise in their notation of exponentials .
This equation between the number of the living , and the age , becomes deserving of attention , not in consequence of its hypothetical deduction , which in fact is congruous with many natural effects , as for instance , the exhaustions of the receiver of an air pump by strokes repeated at equal intervals of time , but it is deserving of attention , because it appears corroborated during a long portion of life by experience ; as I derive the same equation from various published tables of mortality during a long period of man 's life , which experience therefore proves that the hypothesis approximates to the law of mortality during the same portion of life ; and in fact the hypothesis itself was derived from an analysis of the experience here alluded to .
Art .
6 .
But previously to the interpolating the law of mortality from tables of experience , I will premise that if , according to our notation , the number of living at the age x be denoted by L , , and x be the characteristic of a logarithm , or such that x ( L ) may denote the logarithm of that number , that if x ( La , )=x(L , + r)m , A(L , Jr r ) ( La + 2r ) mp , A ( La + zr ) ( L + 3r ) = mp ; and generally x ( La+._ ) A ( La + ) m. p%-i ; that by continual addition we shall have x ( La ) -x ( Lan )=m(l J+p +P+ P+. .
P)=r m. - ; and therefore if p_ = q , and E be put equal to the number whose common logarithm is M , we shall have x(La+)-=(L)-x ( ) x(1 --q)=x ( + ( )q ; L. La+n = -x i ; and this equation , if for a+n we write L -a qL x , will give LX =-q X ; and consequently if be put -a.x = , and i1 g , the equation will stand L =d .
I q , and -a x(g)-=-()xq = -a ; and I observe that when q is affirmative , and x ( 0 ) negative , that x ( g ) is negative .
The equation L= d. - ] may be written in general x ( L= x(d ) the positive number whose common logarithm is { x2 ( g ) +xX ( g ) } , the upper or under sign to be taken according as the logarithm of g is positive or negative , X2 standing for the characteristic of a second logarithm ; that is , the logarithm of a logarithm , x ( q ) =xX ( p )( ( ) --a.X ( q ) x )(g-a .
( q ) =(m)--(l--p ) -aA(q ) ; also x ( d ) x ( La my a ' I-p Art .
7 .
Applying this to the interpolation of the Northampton table , I observe that taking a= 15 and r= to from that table , I find A ( La)(La + r)= , o66 m , X ( La + )(A + zr ) =,745 , ( La+zr ) ( L+ 3r ) 915 , and x ( La + 3r ) ( La+ 4 ) =,128 ; now if these numbers were in geometrical progression , whose ratio is p , we should have respectively m =,0566 ; mp =,0o745 ; mp2 =-,0915 ; mp3a,1228 .
No value of p can be assumed which will make these equations accurately true ; but the numbers are such that p may be assumed , so that the equation shall be nearly true ; for resuming the first and last equations we have P3= -66 ; .
logarithm of p= --(logarithm of 1228 logarithm of 566 ) =,111 , ' , X X(q ) =,011213 and p= 1,2944 .
And to examine how near this is to the thing required , continually to the logarithm of , o566 namely 2,75282 , adding,11213 which is the logarithm of p , we have respectively for the 5eo logarithms of mp , of mp2 , of mp3 the values 2,8649 , 2,9771 , 1,0892 ; the numbers corresponding .
to which are,07327 ; ,09486 ; ,1228 ; and consequently m , mnp , mp2 , and mp3 respectively equal to , o566 ; ,07327 ; , o9486 , and,1228 which do not differ much from the proposed series , o566 ; ,07327 ; ,09486 , and,1228 ; and according to our form for interpolation , taking m = , o566 and p -= 1,2944 ; we have = I-p 56.= - , lg 22 ; and x(L ) agreeably to the Northampton tables , being = 3,734 we have x ( d)= 3,7342 +,1922 3,9264 , d= 8441 , X ( q ) , that is to say , the logarithm of the logarithm of q =x ) -a x ( q ) = 1,28375,16819 i , i56 , x ( g ) =,130949= 1,8695 , the negative sign being taken because X ( g ) ( ) x qa = q , andg =,744 .
And therefore x being taken between the limits , we are to examine the degree of proximity of the equation LL 8441 x 7404j 02 or x ( L ) , that is , the logarithm of the number of living at the age x =3,9264 --number whose logarithm is ( T,11556 +x x.011213 ) , as the logarithm of g is negative .
The table constructed according to this formula , which I shall lay before the reader , will enable him to judge of the proximity it has to the Northampton table ; but previously thereto shall show that the same formula , with different constants , will serve for the interpolations of other tables .
Art .
8 .
To this end let it be required to interpolate DEPARCIEUX'S tables , in Mr. BAILY'S life annuities , between the ages 15 and 55 .
, t ; 21 The logarithms of the living at the age of 15 are 2,92840 differences =,0o3966 = X(L15)(L25 ) 25 2,8887+,04738 = x(LS5)(L35 ) 35 2,84136,04757 = ( L35)-(L45 ) 45 2,79379,07280 = 45 ) ( L 5 ) 55 2,72099 Here the three first differences , instead of being nearly in geometrical progression are nearly equal to each other , showing from a remark above , that the living , according to these tables , are nearly in geometrical progression ; and the reader might probably infer that this table will not admit of being expressed by a formula similar to that by which the Northampton table has been expressed between the same limits , but putting , X ( L. s'15 ) * ?
***= 2,92840 ' on the .
supposi(L15 ) = tion of the possia(L 5)= 1w5 ) n * interpolations .
Thus because A ( L1o)(L o ) , o2561 , the first term of the differences , and x ( Lo ) ( L6 ) =,0817 , the fifth term of the differences : take the common ratio 8= t5 , and m = , o56 ; .- .
( m ) =,4o824 .
These will give A ( p),126 ; p 1,3365 a= o , r= o1 , A ( q)= , o26 , x ( ) =_ = , 36 ; .
A ( g ) negative ; AAg= 2,40824 -A( , s36S)-,126 = 2,75526 ; and A(d)= ( LLo ) + 336 = 3,8863i , and accordingly , to interpolate the Carlisle table of mortality for the ages between lo and 6o , we have for any age x , X(L)z=3,8863 i-number whose logarithm is ( 2,8812 + , ol02:6 ) .
Here we have formed a theorem for a larger portion of time than we had previously done .
If by the second method the theorem should be required from the data of a larger portion of life , we must take r accordingly larger ; thus if a be taken 1o , r= 12 , then the interpolation would be formed from an extent of life from to to 58 years ; and referring to Mr. MILNE'S tables , our second method would give x ( L ) = 3,89063 the number whose logarithm is ( 2,784336 +,0120948 x ) ; this differs a little from the other , which ought to be expected .
If the portion between 60 and too years of Mr. MILNE'S Carlisle table be required to be interpolated by our second method , we shall find p= 1,86466 ; A ( m)= T,3o82 ; m =,20329 , &amp;c. and we shall have A ( L ) 3,79657the number whose logarithm is ( g,74767 +,02706 x ) .
This last theorem will give the numbers corresponding to the living at 60 , 80 , and ioo , the same as in the table ; but for the ages 70 and 90 , they will differ by about one year : the result for the age of 70 agreeing nearly with the living corresponding to the age 71 ; and the result for the age go , agreeing nearly with the living at the age 89 of the Carlisle tables .
Art .
11 .
Lemma .
If according to a certain table of mortality , out of a , persons of the age of lo , there will arrive b , c , d , &amp;c. to the age 20 , o , 40o , &amp;c. ; and if according to the tables of mortality , gathered from the experience of a particular society , the decrements of life between the intervals lo and 20 , 20 and 30 , o3 and 40 , &amp;c. is to the decrements in the aforesaid table between the same ages , proportioned to the number of living at the commencement of those intervals respectively , as 1 to n , i to n ' , i to n " , &amp;c. it is required to construct a table of mortality of that society , or such as will give the above data .
Solution .
According to the first table , the decrements of life from io to 20 , 20 to 30 , 30 to 40 , &amp;c. respectively , will be found by multiplying the number of living at the commencement of each period by ab c &amp;c. , and " ' &amp; C. , and therefore , in the Society proposed , the corresponding decrements will be found by multiplying the number of living at those ages by a b- , nl ; C &amp;c. ; and the number of persons who will arrive at the ages 20 , 30 , 40 , &amp;c. will be the numbers respectively living at the ages io , 20 , 30 , &amp; c , *. .
mortality between the ages of 2o and 60 , which appears to me to be the experience of the ' Equitable Society , is x ( L ) 3,813 the number whose log .
is ( -.60o51 +,0149875 x ) .
This formula will give At the ages .
0l 20 3 4o ' 50 60 70 8o No. of living .
5703,2 5403,5 5007 4496 3862 31I6 * I500 Differs from the 28 7 CHAPTER II .
ARTICLE 1 .
The near proximity to the geometrical progression of the series expressing the number of persons living at equal small successive intervals of time during short periods , out of a given number of persons living at the commencement of those intervals , affords a very convenient mode of calculating values connected with life contingencies , for short limited periods ; by offering a manner of forming general tables , applicable ( by means of small auxiliary tables of the particular mortalities ) to calculations for any particular mortality ; and by easy repetition , to calculate the values for any length of period for any table of mortality we please .
If , for instance , it were required to find the value of an annuity of an unit for p years , on three lives of the age b , c , d , the rate of interest being such that the present value of an unit to be received at the expiration of one year , be equal to r , them the Lb+I Lc Ld value of thefirst payment would be Lxxx r ; bcd and of the pth payment the present value would be L+ x Lc+p Ld+p pwhether p be y+p Xd+ Xr ; butifLb =L x(L whet be i , , 3 , &amp;c. which will be the case when Lb , Lb+ , Lb , &amp;c. form a geometrical progression , and similarly , if L+ =Lc ) , and also , Ld + LdX( L ) the preLc d 530 sent value of the pt payment will be L : d , r ) ; hence , if L : b , c , d r be put = a , the value of the annuity will be , , dL -iFa b,.t a3 + p~a-a i p-a a+a2+a3+a..aP4 =jaaArt .
2 .
Consequently , let a general table be formed of the logarithm of -a-for every value of the log .
of aP ; and also let a particular table be formed for every value of the log .
of m~2 according to the particular table of mortality to be adopted ; from the last table take the log .
of pCL +P Lbc Ld+p -- ; and also from a table constructed for the purpose , d take the log .
of rl , add these four logs .
together , and the sum will be the log .
of P , which being sought for in the general table , will give the log .
of ( ) which will be the log .
of the annuity sought for the term p , on supposition of the geometrical progression being sufficiently near .
Here I remark , that were it not for more general questions than the above , it would be preferable to have general tables formed for the values of -- , instead of the log .
of such a._I values ; but from the consideration that for most purposes a table of the logs .
of will be found most convenient , I have had them calculated in preference .
Art .
3 .
The shorter the periods are , the nearer does the series of the number of persons living at the equal intervals of successive ages approximate to the geometrical progression ; and consequently this mode , by the assumption of sufficiently short periods , and frequent repetitions , will answer for any degree of accuracy the given table of mortality will admit of , but then the labour will be increased in proportion .
Art .
4 .
There are different modes of obviating , in a great measure , this inconvenience , by assuming an accommodated ratio for the given age , instead of the real ratio , from amongst which I shall only for the present select a few .
The first is as follows : find for every value of a , the log .
of 1 , that L Lx LLLL is , the log .
of L ?
+ L ... .
; seek x Lx xx this value in the general table , which will give the corresponding value of the log .
of aP ; and construct a table of such values for every value of x , and adopt these values for log .
of a P , instead of the abovenamed values of the log .
of L ?
+P for the determination of the values of the limited periods : the preference of this to the first proposed method consists inm this ; that if the series rx ( L6+1 + Lb + Lb+ ... L^+p= +C9+C'+ &amp; cC ... , , the series b+l , b+ ' , &amp;c. being nearly L Lb in geometrical progression , and Lb+ .
=e^ --.z , + Lb b &amp;c. E1 , s , &amp;c. will be small , and Of -2+ e ... .
Ep=O , and therefore , ifthe series L - , , &amp;c. and LL C. L , ,L Lc dL Ld formed accurately geometrical progressions , and the value of LXLl C+I , d+l .
r=m , the value of the annuity for the term , Cd would be accurately equal to mC+m+g m3 s ... .
+ mPg +m 1+ m g+ m3 ... ... .+ mPEp , but because in 532 general L-'. .
Ld+ and r differ very little from unity , m Lc d will not differ much from unity ; and therefore if p be not great , m , m m3 , &amp;c. will not differ much from unity ; and consequently , as 1 , E , E , &amp;c. are small , m E+ mi %+ m3 e3 ... .
mPep will not differ much from E1+ g2+ E-+ ... . .
p,4 ; but this has been shown to be o ; consequently me l m2 + m3 3 ... .
+ mPE differs very little from o , or in other words is very small ; and consequently , the value of the annuity differs very little from mg+ m2 2 m3 3 , ... .
mP P ; and the same method of demonstration would apply with any one of the other ages , the remaining ages being supposed to possess the property of the accurate geometrical progression ; notwithstanding this , however , as none of them probably will contain that property , but in an approximate degree , a variation in the above approximations may be produced of a small quantity of the second order ; that is , if the order of the product of two small quantities ; but , as in this approximation , I was only aiming at retaining the quantities of the first order , I do not consider this as affecting the result as far as the approximation is intended to reach : thus far with regard to the first accommodated ratios .
Art .
5 .
Moreover , on the supposition that L. , L+ , L *..L , and also Ld L+ Ld Ldp are c+z C+v ' d+ ' d+ z d+p L Ld+ series in geometrical progression , and that r.x Lcdn= n.q. Since the annuity for p years on the three lives is Lb t+ L. +2 Lb+v my it fb ob+p equal to L. m -.it follows +B LL ] T^1 LLb+ L Lb+ P that if n+ +.n3*. .
T n= L L^ ' LL Lbbbb g.n + g8 n+ 3 .
n 3 ... .
+ gPnP that if n be very nearly equal Lb+I Lb2 Lb+ Pp hic to m , --.n.q J+ -L..q +J &amp; c4 * .
n*q which bbb will be the value of the annuity on the three lives , will be nearly =n.q + g.n2.q2 + &amp;c. ... .
tV If q were equal to unity , or , which is the same thing , m--n , the equality would be accurate ; but it may not be so when m differs from i ; but the nearer n is to m , at least when the difference does not exceed certain limited small quantities , the nearer will be the coincidence .
It appears therefore , that if instead of taking the accommodated ratio for gp so that L , X ( Lb+x + Lb+ L++3 Lb+p ) +g *+ g riods of ten years ; that is , for x ( ) , corresponding to a given value of A ( a'o ) .
N ? .
, for seven years , or for x ( 1 , w _ _ ) corresponding to ( a ) , and the sd for five years or for xa - ) , corresponding to X ( a ) ; calculated ( whether p =1 lo , 7 or )for every value of ( a ) , answering to , oo ; 3 , oi ; 3,0o , &amp;c. ... .
o. The first column containing the aforesaid value of A. ( aP ) , corresponding to which , in an horizontal line , is placed the log .
of _- , , and between each successive value is placed the difference , retaining a decimal figure more ; at the head of the other columns for the proportional parts of the differences , are placed a column showing the number of ciphers to be prefixed to the differences entered in the column following , which are headed 983 s5 nearest under i , 2 , s , 4 and 5 , and opposite the number ; suppose,1 6 of table log .
of a o , stands 2757 , 0520 , oS26 , ,26 Il 376 the upper , with the addition of the two ciphers , give the proportional parts for , ool , oo , oog , ,004 , oo5 : and the under , with the two ciphers , shows the proportional parts for,009 , oo8 , 007 , 00 6 ; and the reason of choosing this arrangement , is the advantage which it offers of proof of correctness ; thus the sum of the higher an lower numbers of each of the above row with the two ciphers = 002752 , which is double,01376 , and equal to the whole difference between the successive terms .
Let it be required to find the logarithm of ( I1 corresponding to log .
of a ' ?
= 1.7954 .
In the General Table I , Opposite to I.79 we have..,88868 For , oo5 we have proportional part .
256 For,0oo4 .
ditto. .
20 The sum.,89144 is the answer .
S35 , If log .
of aP is less than 3- , oo , then it will be necessary to calculate x ( , aL ' by common methods , as the tables do not a.W0 go lower .
And generally it will be then sufficient , omitting a P , only to calculate the value of-x ( da ) ; but from this , if more accuracy be required , subtract the number whose common logarithm is ( T,6378 +X ( r)P ) .
If x ( ZiT_- ) be given , and x ( a ) be required , proceed thus , A ( alT ) being =,89144 for example .
In Table I , the next value of to X( -is,88868 to which x ( a o1 ) corresponding is i,79 Difference , oo256 belonging to 1.79 gives. .
00 Difference , oooo. .
ditto ... , ooo ifx t^ I -a,8944 then we have x ( aI ' ) 1,7954 If X ( aP ) is less than 3 , proceed thus : put the given value ofA('_ aX q , and we have the common logarithm of a =--p xx(+ q- ' ) a small correction if great accuracy be required ; which correction is nearly equal to px the number whose common log .
is{.6378-4x-p+-i .
( I+ ' ) } These methods and tables only apply immediately to xk F( _ , j when a is a proper fraction ; but if a be greater than unity , put it equal to b ' then will b be a proper fraction ; pP -P -Pb P+1 butl ' -I-bb x - &gt; ax F- ) ; conseZITic i-a " l~ I~a Ib----quently ) =p.(pa)+ -P )I have likewise had Table IV .
calculated , which is a general table , for the common log , of ( _ ) , corresponding to a given value of Xa , commencing with ( a ) = 1.7 ; 1,701 ; 1,702 , &amp;c. with the differences between them .
I have not , in this table , had the proportional parts inserted , though it would be attended with advantage , as the table is not meant to be of general use ; but only given to be applied for rough purposes , or where accuracy is not particularly required for calculating at once the value of a life annuity for the whole term of life , or the whole remaining terms of life , after a given term , by considering the present value of each successive payment to form the successive terms of a geometrical progression whose first term and common ratio are each equal to a. And as A ( --1 will represent the log , of the sum of the said geometrical progression , it will likewise express approximatively the logarithm of the value required .
For many purposes , a table of , answering to given values of a , would be preferable , but not for general purposes .
Art .
7 .
I have already , in Art .
4 and 5 , Chap. II , introduced the term accommodated ratios , or chances , and endeavoured to explain the methods to be adopted to reap the advantage of the ideas there expressed .
Table V , for Carlisle , Deparcieux , and Northampton , are the logarithms of tenth terms of the accommodated ratios , or the logarithms of the accommodated chances for living ten years , calculated according to a mode laid down in Art .
5 , Chap. II ; that is , it expresses for every age , or value of b , the logarithm of C ' , when L(1 , o05L .
, o5 .
L,0 Lb + &amp;c. ... 1 , o5 PL jL ?
b-\i 2bo-+I2 b+z '+ isequal tol , oo.l4 1,05_~ 2+ &amp;c. ... 1 , o05 .
10 I and to show , by example , how these are calculated , let it be required to find the logarithm of the accommodated chance for living 537 ten years , for the age o2 , calculated according to the Carlisle table upon the consideration of interest at 5 per cent. Accord1,05l1 ing to the Carlisle tables , I find X ' ; that is , the logarithm of the annuity of one pound on a life of 20 , for ten years , at 5 per cent =,87176 , and putting a= -105 - , by hypothesis we shall have xi a ; that is the logarithm of ( a +a+ a3..Y a 1),87176 ; that is , xQ ) =,87176 ; hence proceeding , as shown above , to find from General Table I. ( a ' ?
) Having given..,87176 .
- ) ( '--l ?
We have next less=,86842 corresponding to .
*.75,00334 difference,00302 proportional p art.,006 30 .
ditto .
, ooo6 2 .
0shall be equal to the value of the whole life annuity at I 05"x -11 five per cent. at such age , namely ' ; consequently x 1 , o05_ o1,05X1 i , o5~1 1051 / 1 051 ( 1+ i , )=T ; . .
A:=A( )+A(1,0)-0^ : * This table is constructed for Carlisle , Deparcieux , and Northampton , and is to be used in conjunction with Table IV .
, where only a rough value of the contingency is required ; and though this table applies as the other tables of accommodated chances , to different rates of interest , still it would be of advantage more particularly here for the greater approximation to have similar tables constructed from the formula x ( )= )+x ( rl ) X( ) for different values of r. Art .
9 .
In calculating the value of life annuities for long periods , by means of adding together the values of portions of those periods , the portions of the distant periods contain factors of the real chance of living to these periods , and likewise of the discounted value of the money of which the payment is not immediate ; thus if t be greater than io , I_ s_\ 1_ aal b , c. a , b , cL ^a , b , c 0abco a , b ? ? ? ? ?
l _ 10 t L. ri x+L a , b , cxr it " bb G ( -|)aIo , +io , b+O , .
It will be therefore convenient to have a table of the logarithm of the real chance of living 10 , 0o , go years , &amp;c. and also for other terms ; and some of these are given by Tables VII .
, VIII .
, IX .
Time will not allow me , for the present , to offer more than a very few examples of the method to be employed in calculating by these tables , which are as follow : Example i. Required , according to the Carlisle table , the value of a life annuity , for ten years , on the joint lives 30 and o4 , at 3 per cent interest .
In Table VIII .
for Carlisle , log .
of accommodated chance for Io years , at the age 30 ... 9552 Ditto 40 ... = 1.9383 Ditto x 1,03 ... -1.8716 Sum. .
1.7651 = ( a'0 ) In Table I , 1.76 corresponds to ... .
8734 In proportional parts , oo5 corresponds to.253 Ditto .
oooi corresponds to .
5 Consequently i,7651 corresponds to..87604 which is the log .
of the required value : the number corresponding to this is 7,5169 , for the value of the annuity , according to the Carlisle mortality , at 3 per cent. on the joint lives 30 and 4o ; and by calculation from Mr. MILNE 's tables , I find the value should be 7,5168 ; the difference of the two is evidently insignificant .
In this way I calculated the log .
of the value of the life annuity , at the Carlisle mortality , at 3 per cent. for 1o years , for the joint lives o and o , 1o and o , 20 and 30 , 30 and 40 , 40 and 5o , 50 and 60 , to be,76580 ; ,90247 ; ,89139 ; ,87604 ; ,86295 ; ,8167 ; and the annuity , or the numbers corresponding to the said logarithms , 5,8318 ; 7,9874 ; 7,7874 ; 7,5169 ; 7,2937 ; 6,4665 ; and , according to calculation from Mr. MILNE'S tables , I get 5,8595 ; 7,992 7 ; 7,7906 ; 7,5168 ; 7,2916 ; 6,4679 .
The difference between the two sets is insignificant , except 540 1 05 perhaps in the values of 1i ( _o ; that is , the value of the annuity on the joint life of a child just born , with one of the age of io , at 3 per cent. Had we divided the period in portions , the value might have been obtained as near as we pleased ; or we should likewise have obtained greater accuracy , had we assumed an accommodated chance deduced at a more appropriate interest than 5 per cent. See Art .
, Chap. II .
Example 2 .
Let it be required to find the value of a life annuity at 3 per cent. for io years , at the Carlisle mortality , for the five lives of the age 20 , 30 , 40 , 45 and 5o .
In Table VIII .
log .
of accom .
chance for xo years at age 20 X.9685 Ditto ... 30 = I.955 Ditto ... 40 79383 Ditto ... 45 = 1.9367 Ditto ... 50 = i.99z 10 X 1,05 =.8716 x ( a o1 ) =.5995 This sought in Table I. ; thus , 1,59 giving,79035,009 427,0005 23 gives 79485 the N ?
to which log .
is 6,2352 1,03"1 I , 40 , 45 for the value of lo 404 , 0 .
1,03"1 Example 3 .
Let it be required to find the value of bb+i ?
Carlisle mortality , when b-= o , that is , for the whole joint lives of lo and 0o .
By dividing the whole in portions of ten 1 03 '1 years the operation will stand thus forlo years , the operation will stand thus fo.b ?
+ I. b=io b=z5o b4 b50 b6 b=7 o b-8o Log.of accom .
ratio.9768 i.9685 1.9552 I.9383 I.9292 1.8318 I.6689..3134 from Tab. Vl .
for o years = 1.9685 1.9552 1-9383 1.9292 1.83I8 I.6689 I*3134 2.66955 Carlisle .
x(I , o03"o)= 1.87161.8716 18716:.87I6 1.8716 1.8716 1.8716 1.8716 sum. .
1_.81691.7953 ?
.7651,.7391 1.6326 1.3723 2.8539 3.8545 No ' corresponding tosuminTableJ.g } 90247 -89139.876o04.86295.81067'.69156.48781.19146 Log. of rats f.97438 o.94120 .
89520r .
83292 7t23 1 570I6I 6886 Log.ofratiosrOoyears.968 I.gz92o82 ; I.85854I.77684 I 59577 I-19448 2.36767 X 1 , o3_-0 .
=i.87i63 1.74325.6I4881.4865 I.358r41.Z2977 1.0139 The log .
of the present.70 481 123157 I 90694 396702 Z4822 829 worth of each portion 3 23157 994 397 And the present worth of each , or the numbers corresponding to the last logarithms are arranged below .
For first Io years 7.9886 As the method by which the logarithms of the present worth of nd ditto 5.0607 the different portions are found , may not be seen by every reader , I 22nd aO~ 5 06o7 will explain the operation in the third portion ; that is , when the 3d d ? ?
3.0291 logarithm of the portion first found is anticipated for 20 years .
4th do I1 7044 Resume ... 87604 5th d ?
.8071 Table VII .
log .
of real chance for age }1.94 20 6th do.2492 io living o2 years. .
7th d ?
.0303 Ditto 20 years living. .
1 92082 8th do.0007 ( I , o03-^ ?
). .
74325 sum 18.8701.48131 which differs but insignificantly from Mr. MILNE'S table , which gives 18.873 .
In a similar way , I find the value of the joint lives for ages 20 and 30 , at 3 per cent. and Carlisle mortality to be 16.745 ; which , according to Mr. MILNE'S table , should be 16.749 ; which appears to be an insignificant difference .
Example 4 .
To find , when particular accuracy is not required , according to the formula for the whole of life , 1,03 the approximate value of 5a4-+ 10 at the Carlisle mortality , when a= -o , 20 , 30 , &amp;c. call the logarithm of accommodated ratios for an unlimited time at the age a , R standing for the accommodated ratio in Table VI .
at the age a. 542 a 10 0 30 40 50 6o 70 80 90 Ra 1 ?
.99529 1.99455 1.99265 1.98991 I.98546 I.97514 1.95755 i:9246I '.866 6i R.i , ,s.9945 19965 1.98991 '1.98546 1.975t4 ?
1.95755 7.92461 ?
.8666o 7.8 2 82 X i,3-91:.98716 1.98716 I.98716 I.98716 I.98716 1.98716 &lt; I.987I6 I.987I6 i.98716 7-.97700 1.97436 1.972 96972z.96253 1.94776 cI.91985 I.86932 1.77837 Y.66658 Log. which I26 f 275.13083 1*03886.88674.68817.45337.17571 corresponds to 1 o00631.o6.0o641 667.00502.02 93..oo6o67.693 i.45.6 o.21i606 1.14145 1.04527.89341.69319.45460.17664i Fi_ !
I.166 6 !
44 X. 0 l7q82-X 85.47.94 i. Numbers. .
18.37 16.446 13.850 11.099 7.824 4.9339 2.8485 1.501g Instead of .
18.873 1:6.749 14.449 11.954 8.729 5.565 i 3.229 1.589 To find the value corresponding to T.66658 , not in the table , find the number corresponding complement of the log .
T.6658 , which number is 2,159 ; subtract i , and find the complement of the log .
which is =.9359165 , whose number is,8628 .
Mr. MILNE 's table gives.979 .
But as it is not always the same rate of interest which gives the best accommodated ratios , in order to try when , for instance , the interest of money is s per cent. what rate of interest should be used in determining the ratios , use the following table:* Interest .
1.o8 X ( 1.o8 X I.03 ) =.g979 0.o7 x ( I.071 X x.o3 ) I9.g83 i.o6 AX ( io61 X I1.03 ) = 1.987 neary ; I.o5 x &lt; ( z.o5 X 1.03)= Y.991 I.04 x ( I.04 X I1.03 ) = 7.996 I observe that I have not given any table of the logarithm of the accommodated ratios for an unlimited term , except that calculated with 5 per cent. as a radix ; but by the assistance of a table of life annuities , for single life at different rates per cent. , this will enable us , independent of certain exceptions , to derive the quantity for the same rates per cent. for any radix at the per cent. contained in the second table ; thus to find R Carlisle mortality , radix 8 per cent. I look to 50 the Carlisle table of single lives at 8 per cent. , and I find the value of the annuity on the life of 50-o 8.987 , I search the age to which this will correspond at 5 per cent. and I find sufficiently nearly 59,82 for the age corresponding , to which from my table ( with the radix at 5 per cent. ) for the log .
of ratios I find 1.97536 ; to this I add log .
of l.o8 ; that is , 1.05 , o0123 , and we get T.98759 , the same as given on the other side .
This method is accurately consistent with the definition of accommodated ratios for unlimited periods ; and if this description of accommodated ratios at a certain rate per cent. be given for one table , for which at the same rate per cent. we have the value of single lives , we may find the same description of accommodated ratios for any other table of mortality for which , at the same rate per cent. we have a table of the value of single lives : thus , suppose the logarithm of this description of accommodated ratios be given for the Carlisle table at five per cent. , and the same be required for -1 1,05 the Northampton for the age 60 , at the same rate ; 6_ o Northampton = 8,392 , this being sought in the Carlisle 545 How the value of particular assurances may be determined from the value of annuities , is shown in my Paper in the Philosophical Transactions for the year 1820 , many of the cases of which are solved by methods essentially the same as those which have been long adopted ; but when such assurances are but for terms , which are not of great extension , very near approximations may be had by using a geometrical progression , without confining the arithmetical operations to the same route , since the chance of extinction of the joint lives of the present age a , b , c , &amp;c. taking place between the period commencing with the time n+t-1 , and finishing with the time n+ t , from the present , is = ( L+ t : a , b , c , &amp;c. L+ t : a , b , c , &amp;c. )+ , b , c , &amp;c. ; it follows that if r be the present value of unity , to be received certain in the time i , and L+ t:a , b , &amp;c. = L : a , b , x , t , whatever tr a , b , c , &amp;c. may be , that , b , C &amp; or the assurance of unity to be received at the first of the equal periods i , from the commencement of the time n-i to the expiration of the time m , which shall happen after the extinction of the joint lives , L is equal to n:b &amp; c Xr x ( 1--r ) +rx ( r x--y ) + a , , C , &amp; C. r^t2 3 ) .
t. .
m-nI x -"i ir-f ) } -=(l- , r)X L-:a , b , c , &amp; c n+I n+zt n+ r3 m " -n-I = ( rr-L r..r .
a , b , c , &amp;c. -r n-I:a , A c , &amp;c. r , .
r ... . .
L rx n. i a , b c &amp; c r+:rr4r a. b , 6 , c , &amp;c. .
If the assurance be not deferred , n will be equal to i , and r we shall have , according to the hypothesis , l4b , c , &amp;c. = rr ( i-r ) .
rx la , b , &amp; c , ; and also = a. If the 579 4F taken equal to , we shall have from the equation L+t_ a , b c , &amp; C. LI a , , C 7r , X c , &amp; bc .
, and this would be the real n- ] ' a , b , c , &amp;c. ' _-i : a , b , c , &amp;c. value which should be taken for , r , if the geometrical progression coincided perfectly with the fact ; and it would be indifferent whether we made it equal to n+t : a , c , or i-L + t:a , b , c , &amp;c. L.n : a , ib , c , &amp;c. .n. .
- " ; bc , a as the two would be the same ; but this not L:a , b , c , &amp;c. being the case , there will be a preference ; and generally , if not always , Xr should be taken an intermediate value between the two ; and when the term is not very long , it will answer a good purpose to take it about the middle between them , inclining generally , though perhaps not always , rather nearer the last than the first , as the first terms are generally of more consequence than the last .
If the said assurance be not deferred , and instead of being paid for immediately , be to be paid for by equal periodic payments , at an unite of time from each other , up to the time m-1 inclusive , and the first payment be to be made immediately , then will the present value of such periodic payment be !
: , and consequently each payment , from what is shown above , is equal to b a , c , K - .
C= ( i--r ) .
r. From whence we may draw an inference worthy of remark , namely ; when an assurance of joint lives is meant to commence immediately , and to continue for a term of t years , which is not large , and to be paid for by t annual payments , that those payments will not differ much with the increase of the time t , provided , as I have said , that t be not large , and the ages 580 be not at the extremes of life , a consequence which follows from the near agreement to a geometrical progression which takes place in the number of living at each small equal increment of time ; that is to say , from the near coincidence LLo an , 6 , c , &amp;c. with +t : a , b , c , &amp;c. of::a , bc , C , &amp;c. ith nt:a , b , ^ ' &amp; C , or the small variation of ir b1 : a , b , c , &amp;c. -ti:a , c for the different values of t : and also , that when the number of years for which an assurance continues be not very long , and the ages be not at the extremes of life , the annual premiums will not differ widely from the premiums to be paid for an assurance of one year of a life older than the proposed life by about half the term : thus , according to the Northampton table , at three per cent. to assure oo 1.at the Age. .
I5 20 30 4.0 50 6o 64 For 7 years , the annual 1 premium by the com ?
I..2..Ix I. .
9 .
5..14.1 2. .
4. .
1 3. .
o. .
8 4 ... ... i 54..1o moan modes of calculation And the premium for one year assuranceforanage..3 .
3 i. .
98 i..i5. .
o 2. .
6 3. .
47 85 5 .
63 years older. .
JI the difference of which is very small.-As another example , let Age. .
or .
o yer , tenul FPI 10 20 30 40 5o 6o or Io years , the annual ; premium will be , by com ?
o..92.9. .
I. , ... 8 2. .
5 .
8 3. .
3. .
4 4.'12 .
6 mon modes of calculation remiumforoneyearas o.17 .
i..1..7 lI16 ... .
6 .
8 3. .
5. .
4..a15 surance , age 5 years older Here , except at the age lo , the excess is rather more in the approximation than in the first set of examples ; but it should be recollected , that we took the exact middle , instead of inclining to the early age .
I.---T I-

