LXII .
A Letter to the Right Honourable George Earl of Macclesfield , P. R. S. concerning the Value of an Annuity for Life , and the Probability of Survivorship .
By Mr. James Dodson .
My Lord , YOUR Lordship 's approbation of my essay , concerning the method of forming the series for making logarithms , and that for rectifying the circle , without the assistance of fluxions , has emboldened me to lay before you the following investigations , concerning the value of an annuity for life ( secured by land ) , and the probability of survivorship , between two persons of given ages ; neither of which have ( as I believe ) been before attempted to be found , otherwise than by fluxions ; at least , no such attempt has appeared in public .
I am sensible that I ought to apologize for reciting the principles , upon which calculations of this kind are commonly founded , because I know that your Lordship is perfectly well acquainted with them ; but I plead in excuse thereof , that I found a great difficulty would attend the rendring my argument intelligible and conclusive , if they were omitted .
I am , with the greatest deference , my Lord , Your Lordship 's most obliged , and most humble servant , Bell-Dock , Wapping , Oct. 18 , 1753 .
J. Dodson .
Read Jan. 17 , The writers upon the subject of annuities on lives have very justly distinguished them into two kinds : In the first , the annuitant is entitled to receive a payment , if he be alive on the day on which it becomes due ; but if he dies on the preceding day , or sooner , his heirs have no claim to any part of the payment , so to have became due ; but in the second , if the annuitant dies at any intermediate time , between the days of payment , his heirs are to receive a part of the annuity , proportional to the time elapsed , between the preceding day of payment and the annuitant 's decease .
This latter kind of annuities have been distinguished from the former , by the words , secured by a grant of lands ; because , where lands are leased for lives , the conditions are generally such as are above described .
The values of the first kind of annuities have been investigated upon principles purely arithmetical ; but , in order to perform the latter , fluxions have been used ( as I humbly conceive ) without any necessity : But as the investigation of the former may be usefully made a part of the latter , I shall first recite the method of performing that , and then proceed to attempt the other , upon the same principles .
If , with the sagacious mathematician Mr. Abraham De Moivre , we suppose the decrements of life to be equal ( viz. that out of a number of persons , alive at a given age , equal to the number of years that a person of that age hath a possibility of living , there will die one in each year , till they are all extinct ) ; then , out of a number of chances equal to that number of persons , which may , for instance , be 36 , all but one are favourable , in the first year , to any individual ; and , consequently , it is 35 to 1 , that he receives one payment of the annuity , by living till it becomes due ; that is , the probability of his receiving it , is 35/ 36 , and that of his not receiving it~ 1/ 36 .
Again ; since , by supposition , there dies but one person in the first year , and one in the second ; there are but two chances , in the 36 , against his receiving the second payment , by living till it becomes due ; and , consequently , 34/ 36 will be the probability of his receiving that also ; the probability of his dying in that year being 1/ 36 , as before .
In like manner it may be proved , that the probability of his receiving the third , fourth , fifth , &amp;c. payment , will be 33/ 36 , 32/ 36 , 31/ 36 , &amp;c. and therefore , if the annual payments were each 1 l. and if the interest of money was not to be considered , we might conceive these several probabilities , as the present worths of the several payments , and the sum of them would be the value of an annuity of the first kind .
But since the interest of money necessarily enters the process , and since the payments become due at the end of the first , second , third , &amp;c. year ; therefore the first of these payments must be discounted for one year , the second for two years , the third for three years , &amp;c. and the sum of their present worths will be the value of an annuity of the first kind , to continue during the life of a person , who may possibly live 36 years ; and this sum may be found by an easy and well-known process ( from the common tables of compound interest and annuities ) , which need not be inserted here .
The annuity secured by land must necessarily be of greater value than the above ; because , altho ' the annuitant dies before the payment becomes due , yet his heirs are to receive a part thereof ; the annuitant , therefore , in this case , hath not only the probability 35/ 36 of receiving the first payment , but he hath also an expectation upon part of the probability 1/ 36 , which , in the first case , was wholly against him .
Now it may be esteemed an equal chance ( supposing him to die in the first year ) whether that decease happens before the expiration of half that year , or after it ; and if it happens before , he is to receive less than half the annual payment ; but if after it , more .
The annuitant may , therefore , be supposed to have an equal chance , if he fails of receiving the whole first payment , yet of receiving half thereof ; and , consequently , half of the probability , 1/ 36 , which was before totally against him , will ( in this case ) be favourable to him ; and his expectation of receiving either the whole , or at least half of the first payment , will be 35/ 36 + 1/ 72 .
In like manner , since the probability of his dying , in the second year , is , also , 1/ 36 ; we may ( by arguing in the same manner ) prove , that one half thereof will ( in this case ) become favourable to him ; and , consequently , that ~+ 1/ 72 will be the probability of his receiving the whole , or at least half , of the second payment .
It appears , therefore , that , for every year which he hath the possibility of living , he will ( in this case ) have the probability ( 1/ 72 or ) half of 1/ 36 in his favour , more than he had in the former case ; and therefore , if the present worths of the constant sum 1/ 72 l. ( supposed posed to be due at the end of one , two , three , &amp;c. years ) be found , and added to the value of the annuity , according to the former case , the sum will be the value of an annuity , secured by land , to continue during the life of a person who may possibly live 36 years .
Now since the sums , whereby the former annuity is to be increased , consist of the present worths of that fraction of a pound sterling ( whose numerator is unity , and denominator twice the number of years that the annuitant can possibly live ) supposed to be due at the end of each of one , two , three , &amp;c. years ; it will follow , that their amount , or the difference between the values of the two annuities , will be equal to the quotient , found by dividing the value of an annuity of 1 l. certain , for as many years as the annuitant can possibly live , by twice that number of years : And , therefore , if to the value of an annuity for life , of the first kind , we add the quotient so found ; the sum will be the value of an annuity , of the second kind , for the same life .
When I had thus investigated the value of this annuity , I compared the result with that Mr. De Moivre has deduced from fluxions , which is published in No 473 .
of the Philosophical Transactions ; and found , that they agree to more than a sufficient exactness , for computations of this nature : I have , therefore , annexed that comparison hereto .
The above-mentioned author , and others , have also calculated the probabilities of survivorship , by a fluxional process ; which probabilities will flow from what is above premised , by a very easy calculus ; in the doing which , the above yearly probabilities of the anQ qq2 nuitant nuitant 's receiving the whole first , second , third , &amp;c. payment , or at least the halves of them ( being , in fact , the probabilities of his living the whole of those years , or at least the halves of them ) , are called the expectations of life , in those years .
Mr. De Moivre has proved , on another occasion , that , if there be an expectation , to take place upon the happening of two independent events , the probability of the compound event , will be the product of the two single probabilities : Now , if there be an expectation , depending on one man 's surviving another , it must necessarily be compounded of the probabilities of the one 's living , and of the other 's dying .
Now let the person , upon whose surviving the expectation depends , be of such an age , that he may possibly live 36 years ; and let the person , who is to be survived , be of that age , that he may possibly live 43 years : Then since the survivor 's expectation of life will , in the first , second , third , &amp;c. year , be 35/ 36 + 1/ 72 , 34/ 36 + 1/ 72 , 33/ 36 + 1/ 72 , &amp;c. and since the probability of the other 's dying , in any one year , will be 1/ 43 ; therefore , if the former be severally multiplied by the latter , the products arising , viz. 35/ 36 + 1/ 72 ~ 1/ 43 , 34/ 36 + 1/ 72 ~ 1/ 43 , 33/ 36 + 1/ 72 ~ 1/ 43 , &amp;c. will exhibit the probabilities of the survivorship 's taking place , in the first , second , third , &amp;c. years .
For the above expectations contain the probabilities of the survivor 's living the whole , or ( at least ) the halves of the first , second , third , &amp;c. years ; and the constant factor , 1/ 43 , is the probability of the other 's dying , some time within one of those years ; which may , therefore , be either in the first , or in the second , half year , of either of them ; thro ' both of which periods , the probability of the survivor 's living is above exhibited ; and , consequently , the probabilities of the survivorship 's taking place , in those times , will be severally represented by the above products .
Now the sum of 36 terms of this series of products , will , upon computation , appear to be 18/ 43 ; which is , therefore , the probability of the survivorship required .
But if , instead of requiring , as above , the probability of the elder person 's surviving the younger , it were required to find the probability of the younger person 's surviving the elder ; then , since it is almost a certainty , that both of them will not die in the same moment of time ; we may , by denoting that certainty by unity ( agreeable to another principle established in the doctrine of chances ) determine the probability required to be ( 1 18/ 43 or ) 25/ 43 : The algebraic investigation of each of these cases is annexed , to which I beg leave to refer , for a farther illustration ; and shall only observe , farther , that the above two results will , upon comparison with those given , upon a fluxional process by Mr. De Moivre , in his treatise of annuities on lives , appear to coincide with them exactly .
The probability of any order of survivorship , that can happen among three persons , and , consequently , that of one person 's surviving two others , may , likewise , be investigated upon similar principles , without the assistance of fluxions ; but as this problem admits of six cases , and the algebraic process is of a length , too great for the designed limits of this essay , I beg leave to postpone it .
A comparison of the result of the beforegoing investigation of the value of an annuity for life ( secured by land ) , with the value thereof , given by Mr. De Moivre , in p. 67 .
of No 473 .
of the Philosophical Transactions .
If the annual payment , or rent , be supposed to be 1 l. ; if the rate of interest ( that is , the amount of 1l .
and its interest for one year ) be denoted by r ; the number of years which the annuitant may possibly live ( called by Mr. De Moivre the complement of life ) by n ; and the value of an annuity of 1 l. per annum , certain for n years , by P : Then , by Prob. I. of Mr. De Moivre 's treatise of annuities on lives , the value of an annuity ( not secured by land ) for the life of a person , whose complement is n , will be 1 r/ n P/ r 1 ; to which if P/ 2 n be added , the sum 1 r/ n P/ r 1+ P/ 2 n will , according to the result of the above investigation , be the value of an annuity ( secured by land ) for the life of the same person : And if the hyperbolic logarithm of r be denoted by ~ ; then , by the above-quoted Philosophical Transaction , 1/ r 1 P/ ~ n will be the value of the same annuity , according to Mr. De Moivre : Let us proceed , therefore , to examine , whether there these two expressions , viz. 1 r/ n P/ r 1+ P/ 2 n and 1/ r 1 P/ ~ n are equal , or nearly so .
It is well known , that the hyperbolic logarithm of any number , r , is equal to the infinite series , 2/ 1 ~ r-1/ r+1 + 2/ 3 ~ r-1/ r+1 ~3+ 2/ 5 ~ r-1/ r+1 ~5 &amp;c. and , in this case ( since the interest of money has not , for many years , exceeded 5 l. per cent. and is continually decreasing ) , r will be always expounded by some of the following numbers , 1,05 .
1,04 .
1,03 .
&amp;c. or others intermediate to them ; in the greatest of which ( viz. 1,05 ) r1 will be expounded by ( ,05 or ) 1/ 20 , and r+1 by ( 2,05 or ) 41/ 20 ; whence , 2/ 1 ~ r-1/ r+1 , the first term of the series will be 2/ 41 , and 2/ 3 ~ r-1/ r+1 ~3 will be ( 2/ 3 ~ 1/ 41 3 = ) 2/ 206763 , a fraction too small to affect a calculation of this kind ; and therefore , 2/ 1 ~ r-1/ r+1 , the first term of the series , will be nearly equal to ~ , and may be wrote for it .
Now by writing 2/ 1 ~ r-1/ r+1 for ~ , the value of the expression , 1/ r-1 P/ ~ n , will become 1/ r 1r+ 1 r+1~ P/ r-1 ~2 n ; and it remains , only , to prove the equality of the expressions , 1 ~/ n P/ r 1+ P/ 2 n and 1/ r 1r+1~ P/ r-1 ~2 n , which may be done as follows : Put 1 r/ n P/ r 1+~ P/ 2 n= N. Then 1 r/ n P+ r-1/ 2 nP=r1~ N ; But , because r/ n =2 r/ 2 n , And 2 r/ 2 n r-1/ 2 n=r+ 1/ 2 n ; Therefore 1 r+1/ 2 nP= r-1 ~ N , Whence 1/ r-1 r+1 ~ P/ r-1 ~ 2n = N : And the equality of the two expressions is evident : From whence it appears , that the value of an annuity for life ( secured by land ) might have been found , exactly enough , without the assistance of fluxions .
The investigation of the probabilities of survivorship between two persons of given ages .
Let the complement of the younger life be denoted by n , and that of the elder by p : Then , First , If the probability of the elder person 's surviving the younger be required , let the symbol p be used instead of the number 36 , and it will ( from the argument above used ) appear , that the expectations of the survivor 's life , for the first , second , third , &amp;c. years , will be represented by ( p-1/ p + 1/ 2 p , p-2/ p + 1/ 2 p , p-3/ p + 1/ 2 p , &amp;c. or ) 2 p-1/ 2 p , 2 p-3/ 2 p , 2 p-5/ 2 p , &amp;c. which expectations , being severally multiplied into 1/ n , the yearly probability of the other 's dying , will give 2 p-1/ 2 p n , 2 p-3/ 2 p n , 2 p-5/ 2 p n , &amp;c. for the probability of the survivorship 's taking place , in the first , second , third , &amp;c. years .
Now since , by the hypothesis , the survivor cannot possibly out-live p years ; therefore , only , p terms of the above series are to be used ; which series , being an arithmetical progression , whose greatest term is 2p 1/ 2 p n , least term 1/ 2 p n , and number of terms p ; the sum thereof will , by a well-known rule , be ( 2 p 1/ 2 pn+ 1/ 2 pn~ p/ 2 =2 p/ 2 np~ p/ 2 = ) p/ 2 n ; which expression is exactly the same as Mr. De Moivre gives , in the 44th page of the 3d edition of his treatise of annuities on lives ; which by the demonstration thereof , given in fol , 116 .
of the same work , appears to have been obtained by a fluxional process .
Secondly , If the probability of the younger person 's surviving the elder be required ; then , if 2n-1/ 2 n , 2n-3/ 2 n , 2n-5/ 2 n , &amp;c. the yearly expectations of the continuance of the life of the younger , be severally~ multiplied by 1/ p , the yearly probability of the elder 's dying ; the products , 2n-1/ 2 n p , 2n-3/ 2 n p , 2n-5/ 2 n p , &amp;c. will give the probabilities of the survivorship 's taking place , in the first , second , third , &amp;c. years .
And here , as before , the survivorship must necessarily take place , before , or at , the end of p years , the longest time that the elder can possibly live , by supposition ; and therefore , the sum of p terms , of the series above given , will be the answer .
Now this series is , also , an arithmetical progression , whose greatest term is 2n-1/ 2 n p ; least , 2n-2p+1/ 2 n p ; and number of terms , p ; whence ( 2n-1/ 2 np+ 2n-2p+1/ 2 np ~ p/ 2 = 4n-2p/ 2 np~ p/ 2 = 4n-2p/ 2 n~ 1/ 2 = 2n-p/ 2 n = ) 1 p/ 2n will be the probability required ; to which , if p/ 2n , the probability of the elder 's surviving the younger , be added ; the sum will be unity , as was above observed .

