Appendix A.

Commutation Functions

A.l Introduction

In this appendix we give an introduction to the use of commutation functions.

These functions were invented in the 18th century and achieved great popu-
larity, which can be ascribed to two reasons:

Reason 1
Tables of commutation functions simplify the calculation of numerical values
for many actuarial functions.

Reason 2
Expected values such as net single premiums may be derived within a deter-
ministic model closely related to commutation functions.

Both reasons have lost their significance, the first with the advent of powerful
computers, the second with the growing acceptance of models based on prob-
ability theory, which allows a more complete understanding of the essentials
of insurance. It may therefore be taken for granted that the days of glory for
the commutation functions now belong to the past.

A.2 The Deterministic Model

Imagine a cohort of lives, all of the same age, observed over time, and denote
by lx the number still living at age x. Thus dx = lx - lx+1 is the number of
deaths between the ages of x and x + l.
Probabilities lind expected values may now be derived from simple pro-
portions and averages. So is, for instance,
(A.2.1)
120 Appendix A. Commutation Functions

the proportion of persons alive at age x + t, relative to the number of persons

alive at age x, and the prob ability that a life aged x will die within a year is

(A.2.2)

In Chapter 2 we introduced the expected curtate future lifetime of a life

aged x. Replacing kP x by IX+k/lx in (2.4.3), we obtain

(A.2.3)

The numerator in this expression is the total nu mb er of complete future years

to be "lived" by the Ix lives (x), so that ex is the average number of completed
years left.

A.3 Life Annuities

We first consider a life annuity-due with annual payments of 1 unit, as intro-

duced in Section 4.2, the net single premium of which annuity was denoted
by äx ' Replacing kP x in (4.2.5) by IX+k/1x, we obtain

(A.3.1)

or
(A.3.2)
This result is often referred to as the equivalence principIe, and its interpre-
tation within the deterministic model is evident: if each of the Ix persons
living at age x were to buy an annuity of the given type, the sum of net single
premiums (the left hand side of (A.3.2)) would equal the present value of the
benefits (the right hand side of (A.3.2)).
Multiplying both numerator and denominator in (A.3.1) by vX, we find

(A.3.3)

With the abbreviations

(A.3.4)

we then obtain the simple formula

.. Nx
ax = D x ' (A.3.5)

Thus the manual calculation of äx is extremely easy if tables of the com-

mutation functions D x and N x are available. The function D x is called the
"discounted number 01 survivors".
A.4. Life Insurance 121

Similarly one may obtain formulas for the net single premium of a tempo-
rary life annuity,
(A.3.6)
immediate life annuities,
Nx+l
ax = ----ri; 1 (A.3. 7)
and general annuities with an nu al payments: formula (4.4.2) may naturally
be translated to
E(Y) = roDx + r1Dx+l + r2 D x+2 + ...
(A.3.8)
Dx
For the special case rk = k + 1 we obtain the formula
(Iä)x = ~: ; (A.3.9)
here the commutation function Sx is defined by
Sx D x + 2D x+1 + 3D x+2 + .. .
N x + Nx+l + N X+2 + ... . (A.3.1O)

A.4 Life Insurance

In addition to (A.3.4) and (A.3.10) we now define the commutation functions

Cx vX+1d x ,
Mx C x + Cx+l + CX+2 + ... 1
Rx C x + 2Cx+1 + 3Cx+2 + .. .
Mx + Mx+l + M X +2 + ... . (A.4.l)
Replacing kPxqx+k in equation (3.2.3) by dx+k/l x1 we obtain
vd x + v 2dx+1 + v 3 dx+2 + ...
lx
C x + C x+1 + C X +2 + ...
Dx
(A.4.2)

Similarly one obtains

vd x + 2v 2dx+1 + 3v 3 dx+2 + ...
Ix
C x + 2Cx+l + 3Cx+2 + ...
Dx
(A.4.3)
122 Appendix A. Commutation Functions

Obviously these formulae may be derived within the deterministic model by

means of the equivalence principle. In order to determine A x one would start
with
(A.4.4)
by imagining that lx persons buy a whole life insurance of I unit each, payable
at the end of the year of death, in return for a net single premium.
Corresponding formulae for term and endowment insurances are
Mx - Mx+n
Dx
Mx - Mx+n + Dx+n
Dx
Cx + 2Cx+! + 3Cx+2 + ... + nCx+n - 1
Dx
Mx + Mx+! + M X +2 + ... + Mx+n - 1 - nMx+n
Dx
Rx - Rx+n - nMx+n
(A.4.5)
Dx
which speak for themselves.
The commutation functions defined in (A.4.I) can be expressed in terms
of the commutation functions defined in Seetion 3. Prom dx = lx -lx+! follows

(A.4.6)

Summation yields the identities

(A.4.7)

and
R x = N x - dSx . (A.4.8)
Dividing both equations by,D x , we retrieve the identities

1- d äx ,

äx - d(Iä)x, (A.4.9)

see equations (4.2.8) and (4.5.2).

A.5 Net Annual Premiums and Premium Reserves

Consider a whole life insurance with I unit payable at the end of the year of
death, and payable by net annual premiums. Using (A.3.5) and (A.4.2) we
find
(A.5.1)
A.5. Net Annual Premiums and Premium Reserves 123

Of course, the deterministic approach, i.e. the condition

(A.5.2)

leads to the same result.

The net premium reserve at the end of year k then becomes

- A P" _ Mx+k - PxNx+k

k Vx - x+k - xax+k - D (A.5.3)
x+k

This result may also be obtained by the deterministic condition

k Vxl x+k+ Pxlx+k + vPxlx+k+1 + v 2P xlx+k+2 + ... (A.5.4)

= vdx+k + v 2d x+k+1 + v 3 d x+k+2 + ....
Here one imagines that each person alive at time k is allotted the amount k Vx ;
the condition (A.5.4) states that the sum of the net premium reserve and the
present value of future premiums must equal the present value of all future
benefit payments.
The interested reader should be able to apply this technique to other, more
general situations.
Appendix B. Simple Interest

In practice, the accumulation factor for a time interval of length h is occa-

sionally approximated by

(1 + i)h :::::: 1 + hi 0 (Bol)

This approximation is obtained by neglecting all but the linear terms in the
Taylor expansion of the left hand side above; alternatively the right hand side
may be obtained by linear interpolation between h = 0 and h = 1. Similarly
an approximation for the discount factor for an interval of length h is

v h = (1 - d)h :::::: 1 - hd 0 (Bo2)

The approximations (Bol) and (Bo2) have little practical importance since the
advent of pocket calculatorso
Interest on transactions with a savings account is sometimes calculated
according to the following rule: If an amount of r is deposited (drawn) at
time u (0 < u < 1), it is valued at time 0 as

rv U :::::: r(l - ud) 0 (Bo3)

At the end of the year (time 1) the amount is valued as

r(l + i)l-u r(l + i)v r(l

U
:::::: + i)(l - ud)
r{l + (1 - u)i} 0 (B.4)

This technique amounts to accumulation from time u to time 1 according to

(Bol) or discounting from u to 0 according to (Bo2)o With a suitably chosen
variable force of interest the rule is exact; this variable force of interest is
determined by equating the accumulation factors:

1 + (1 - u)i = exp (11 O(t)dt) (Bo5)

Differentiating the logarithms gives the expression

i d
O(u)=l+(l_u)i I-ud (Bo6)
126 Appendix B. Simple Interest

for 0 < u < 1. The force of interest thus increases from 15(0) = d to 15(1) = i
during the year.
The technique sketched above is based on the assumption that the accu-
mulation factor for the time interval from u to 1 is a linear function of u; this
assumption is analogous to Assumption C of Section 2.6, concerning mortality
for fractional durations. The similarity between (B.6) and (2.6.10) is evident.
References

A thorough introduction into the theory of compound interest is given in Butcher-

Nesbitt [3]. The textbook by Bowers-Gerber-Hickman-Jones-Nesbitt [2] is the nat-
ural reference for Chapters 2-10; it contains numerous examples and exercises.
The "classical method" of Chapter 11 is documented in Batten [1] and updated
in Hoem [7]. In this respect the reader mayaIso orient himself with the text of
Elandt-Johnson [6].
An extensive bibliography is given by Wolthuis-van Hoek [14].
The classical texts in life insurance mathematics are those by Zwinggi [15], Saxer
[12] and Jordan [9]; the newer book by Wolff [13] is of impressive completeness. The
monographs by Isenbart-Münzner [8] and Neill [10] are written in the traditional
style; however, the former book may appeal to non-mathematicians. The three
volumes by Reichel [11] have an unconventional approach and will appeal to the
mathematically minded reader. The books by De Vylder [4] and De Vylder-Jaumain
[5] give a very elegant presentation of the subject.

1. Batten, R.W.: Mortality Table Construction. Prentice-Hall, Englewood Cliffs,

New Jersey, 1978
2. Bowers, N.L., Gerber, H.U., Hickman, J.C., Jones, D.A., Nesbitt, C.J.: Ac-
tuarial Mathematics. Society of Actuaries, Itasca, Illinois, 1986
3. Butcher, M.V., Nesbitt, C.J.: Mathematics of Compound Interest. Ulrich's
Books, Ann Arbor, Michigan, 1971
4. De Vylder, Fl.: Theorie generale des operations d'assurances individuelles de
capitalisation. Office des Assureurs de Belgique, Bruxelles, 1973
5. De Vylder, FI.,Jaumain, C.: Expose moderne de la theorie mathematique des
operations viageres. Office des Assureurs de Belgique, Bruxelles, 1976
6. Elandt-Johnson, R.C., Johnson, N.L.: Survival Models and Data Analysis.
John Wiley & Sons, New York London Sydney, 1980
7. Hoem, J.M.: A Flaw in Actuarial Exposed-to-Risk Theory. Scandinavian
Actuarial Journal 1984(3), 187-194
8. Isenbart, F., Münzner, H.: Lebensversicherungsmathematik für Praxis und
Studium. Gabler, Wiesbaden, 1977
9. Jordan, C.W.: Life Contingencies. Second edition. Society of Actuaries,
Chicago, Ill., 1967. Also available in braille
128 References

10. Neill, A.: Life Contingencies. William Heinemann, London, 1977

11. ReicheL G.: Mathematische Grundlagen der Lebensversicherung. Volumes
3, 5 and 9 of the series Angewandte Versicherungsmathematik der DGVM.
Verlag Versicherungswirtschaft E.V., Karlruhe, 1975, 1976, 1978
12. Saxer, W.: Versicherungsmathematik. Volumes 1 and 2. Springer, Berlin
Göttingen Heidelberg, 1955, 1958
13. Wolff, K.-H.: Versicherungsmathematik. Springer, Wien New York, 1970
14. Wolthuis, H., van Hoek, I.: Stochastic Models for Life Contingencies. Insur-
ance: Mathematics & Economics, 5, 1986(3), 217-254.
15. Zwinggi, E.: Versicherungsmathematik. Second edition. Birkhäuser, Basel
Stuttgart, 1958
Index

Accumulation factor 2 De Moivre 17, 50, 60

Acquisition expenses 103 Debt 11
Adequate premium 104 Decrement
Administration expenses 103 cause 75
Aggregate life table 20 force 76
Annuity certain 9 prob ability 75
Arrears 4 Deferred insurance 25
Assumption Deferred life annuities 54
a 21, 26, 28, 30, 38-40, 46, 65, Deterministic model 119
77 Difference operator 89
b 21, 28, 111 Differential equation 4, 32, 43,
c 22, 110, 126 71-73,80
Asymmetrie annuities 90 Disability insurance 75
Asymmetrie insurances 91 Discount 4
Discount factor 2
Balducci 22
Discounted number of survivors 120
Bayesian approach 116, 117
Dispersion 94
Bernoulli 24
Due 6,35
Beta distribution 118
Duration of premium payment 49
Claim frequency 98
Collection expenses 103 Effective interest rate 1
Collective risk model 97 Endowment 24, 54, 59, 84, 104, 122
Commutation Functions 119 Equivalence principle 49, 120
Compound interest 1 Excess of loss 100
Compound Poisson distribution 96 Expected
Confidence interval 113 curtate lifetime 19
Continuous remaining lifetime 16
model 71,80 Expense loadings 103
payments 3, 40 Expense-loaded premium reserve 105
Conversion of insurance 68 Exponential
Conversion period 1 growth 8
Convolution 94 mortality 18
Credibility theory 116 Exponentially
Cross-sectionallife table 118 growing annuity level 41
Current life table 118 increasing sum insured 28
Curtate future lifetime 18, 76 Exposure 110
130 Index

Fixed costs 104 Makeham 18, 85

Flexi ble life 68 Maximum likelihood 112, 117
Force Mortality
of decrement 76 gain 69
of interest 3 ratio 113
of mortality 16, 84, 112
Fractional Nesbitt 87, 89
age 46 Net amount at risk 61, 79
death probabilities 21 Net annual premium 52, 122
durations 64 Nominal interest rate 2
payment 37 Normal approximation 93
premiums 54
Orphans' insurance 91
Future lifetime 15, 75, 83
Paid-up insurance 68
Gamma distribution 113, 116, 118
Panjer 98
General type 27, 55, 77
Pension fund 70, 96
Generating function 90, 96
Perpetuity 6
Generation life table 118
Poisson distribution 90, 112, 115
Gompertz 18, 84, 92
Premium 49
Gross premium 105
difference formula 64
Hattendorff's theorem 66, 79 paid m times a year 54
paying period 104-106
Immediate 6, 37 rate 71
Inclusion-exclusion formula 86 refund 56
Individual claim amount 97 reserve 59, 78, 105, 122
Interest Present value 2, 4, 23, 35
in advance 4 Principal 11
rate 1 Profit sharing 70
Internal rate of return 13 Prospective debt 11
Inventar 105 Pure endowment 24, 53
Investment
gain 69 Reinsurance 51, 100
yield 13 Retention 100
Retrospective debt 11
Jensen's inequality 43 Reversionaryannuity 91
Joint-life status 83 Risk aversion 51
Risk premium 62, 71, 79
Laplace transform 44 Rounding 94
Last-survivor status 85
Lexis diagram 109 Safety loading 50
Life Savings premium 62, 71, 79
annuity 35 Schuette 87, 89
contingencies 1 Security 14
insurance 23 Select life table 20
table 20 Selection 20
Longitudinallife table 118 Shift operator 89
Loss in a policy year 65, 79 Simple interest 125
Index 131

Standard deereasing Term insuranee 23, 52

annuity 10 Thiele's differential equation 71, 80
insuranee 29 Tontines 72
Standard inereasing Total
annuity 10, 121 claim amount 93
insuranee 29, 121 loss 49, 66, 79
Status 83
Stoehastic interest 56 Ultimate life table 20
Stop-Ioss reinsuranee 101 Universallife 68
Sum insured 23 Utility function 50
Survival
Variable life annuity 39
probability 15
risk 63 Waring 90
risk premium 63 Weibull 18
Symmetrie sum 86 Whole life insuranee 23, 52, 63
Widows' insuranee 91
Teehnieal gain 69, 80
Temporary annuity 36 Zillmer 106
E. Straub, Zurich, Switzerland

Non-Life Insurance
Mathematics
1988. VII, 136 pp. 12 figs. Hardcover DM 84,- ISBN 3-540-18787-1

Contents: Problems. - Tools. - Premiums. - Reinsurance. - Retentions.-

Statistics. - Reserves. - Solutions. - References. - Subject Index.

The book gives a comprehensive overview of modem non-life actuarial

science. It starts with a verbal description (i. e. without using mathematical
formulae) ofthe main actuarial problems to be solved in non-life practice.
Then in an extensive second chapter all the mathematical tools needed to
solve these problems are dealt with - now in mathematical notation.
The rest of the book is devoted to the exact formulation of various problems
and their possible solutions.
Being a good mixture of practical problems and their actuarial solutions, the
book addresses above all two types of readers: firstly students (of mathe-
matics, probability and statistics, informatics, economics) having some
mathematical knowledge, and secondly insurance practitioners who
remember mathematics only [rom some distance. Prerequisites are basic
calculus and probability theory.

Jointly published by
Springer-Verlag Berlin Heidelberg New York
London Paris Tokyo Hong Kong Barcelona and
Association of Swiss Actuaries Zurich
J. GrandelI, Stockholm

Aspects of Risk Theory

1990. Approx. 175 pp. 5 figs. (Applied Probability, Vol. 6) Hardcover
ISBN 3-540-97368-0

Risk theory, which deals with stochastic models of an insurance business, is

a classical application of probability theory. The fundamental problem in risk
theory is to investigate the ruin possibility of the risk business. Traditionally
the occurrence of the claims is described by a Poisson process and the cost
of the claims by a sequence of random variables.
This book is a treatise of risk theory with emphasis on models where the
occurrence ofthe claims is described by more general point processes than
the Poisson process, such as renewal processes, Cox processes and general
stationary point processes. In the Cox case the possibility of risk fluctuation
is explicitly taken into account. The presentation is based on modem proba-
bilistic methods rather than on analytic methods. The theory is accompanied
with discussions on practica1 evaluation of ruin probabilities and statistical
estimation. Many numerical illustrations of the results are given.

H.Bühlmann

Mathematical
Methods
in Risk Theory
1970. XII, 210 pp. 39 figs. (Grundlehren der
mathematischen Wissenschaften, Band 172)
Hardcover DM 114,- ISBN 3-540-05117-1