Professional Documents
Culture Documents
Yufeng Guo
• Reference to Actuarial Mathematics and Models for Quantifying Risk. You can use
either textbook as a companion to go with Arch MLC.
• Low Cost: $99. For $99, you bring a quality study manual home.
• Convenience. After you buy Arch MLC PDF, you can print a hard copy. You can also
install the PDF in your computer.
0 INTRODUCTION 9
The origin of Arch Manual . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Praises of the Arch Manual originally written by Nathan and Robin . . . . . 10
About Yufeng Guo . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
How to use this manual . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3
actuary88.com Chapter 0
INTRODUCTION
In the late 90’s, Nathan Hardiman and Robin Cunningham worked full-time at the former
Jefferson Pilot Financial. Nathan and Robin, like other exam candidates, faced the daunt-
ing challenge of plowing through difficult textbooks and mastering the fundamentals of life
contingency theories and probability models to pass the Course 3 exam, the most difficult of
the four preliminary actuarial exams and the exam with the highest failure rate.
The difficulty of Course 3 was mainly due to its enormous scope. Candidates were required
to read encyclopedia-like textbooks such as Actuarial Mathematics and Probability Models,
gain sophisticated understanding of complex concepts such as multiple decrements, Markov
Chain, Brownian motion, and be ready to tackle tricky word problems on the exam.
Since Nathan and Robin both already had families and full-time jobs when they began study-
ing for exams, they created their own study framework and philosophy for quickly passing
Course 3. After passing the exam in one sitting using their unique study methods, Nathan
and Robin decided to jointly write a new study guide that would enable candidates to build a
core body of knowledge for Course 3 quickly. They wanted their manual to use straight talk
and down-to-earth examples to explain difficult fundamental concepts intuitively and simply.
Nathan and Robin published the first edition of their study manual for Course 3 in 2001.
They named their study manual the Arch-3. Since its publication, Arch has been a popular
study manual for Course 3 and Exam M.
Arch’s power lies in its simplicity. While textbooks talk fancy, Arch talks simple. While
textbooks rigorously prove theorems, Arch explains the intuition. While textbooks demand
attention to everything, Arch separates the critical from the trivial.
While Arch sells well, Nathan and Robin climbed corporate ladder higher and higher. With
9
actuary88.com Chapter 0
each day passing, they have less and less time to keep up with the SOA’s syllabus changes.
Finally, in fall, 2006, Nathan and Robin decided to withdraw from the Arch manual business
and passed on the copyright and ownership of Arch M manual to Yufeng Guo, who is the
author of Deeper Understanding manuals for Exam P, FM, and M.
Nathan and Robin’s contribution to actuarial education was not just the Arch manual but
more importantly the Arch’s effective teaching style. Before Arch was published, many
thought that learning difficult things such as Course 3 ought to be slow and painful. Arch’s
straight talk and down-to-earth examples showed the actuarial community that learning
difficult actuarial theories can indeed by fast-paced and enjoyable.
-I start a seminar on Friday, and I never would have been able to finish and
understand the material without your study guide.
-I want to personally thank both of you for the fantastic and brilliant work that
you did on ARCH. Seeing as it’s not my first time tackling this exam, I’ve had
the chance to use [several other manuals]; however this is by far superior to all
of those products. I have and will continue to recommend it to others in my
company.
-I would first like to say that I am very happy with your manual so far. I feel
that I am progressing through the syllabus much faster than I would have without
it, and the depth of understanding that I am on my own giving up due to my not
using the texts themselves is more than compensated for by the excellent coverage
of the important topics in your manual.
FAQ
I notice you have two study guides for MLC: Arch MLC and Deeper Understanding MLC.
What’s the difference? Which one should I buy?
Difference: Arch MLC focuses on thoroughly explaining the core concepts. Deeper Under-
standing MLC focuses on teaching conceptual insights and calculation shortcuts.
Which one to buy: If money is not an issue, consider buying both. If you want to buy only
one guide, choose the one that better fits your need. For example, if you already have a
study guide and want to learn calculation shortcuts, buy Deeper Understanding: MLC. If
your goal, on the other hand, is to master basic concepts, buy Arch MLC.
Each chapter of this manual contains plenty of examples with solutions. You are likely to
benefit a great deal if every time you get to an example, you cover up the solution and
attempt to work it. You will get many of them wrong, especially the first time you see them.
But the problem-solving experience will be extremely valuable!
On the exam, you will not be asked to explain anything. You will be asked to calculate
numerical answers. Therefore, much of our explanation of the material is done by way of
numerical examples and practice questions. Examples range from very simple ones (to make
sure you know the basic concepts), to thought provoking ones (to help you think about what
you’ve learned and really understand it), to exam questions from prior exams (to get you
ready for exam day).
I also suggest problems from the texts for you to work. Many of the problems in the
text are not transferable to the exam. Some, however, provide useful insight and prac-
tice. Solutions to these suggested problems are available on the Download Samples page at
www.archactuarial.com.
A formula summary for each chapter is included. These summaries are intended to serve as
a reference as you familiarize yourself with the syllabus material.
Finally, there’s a full length practice exam of new questions. This practice exam is designed
to be used in conjunction with the prior Course 3 and Exam M problems in the SOA website
at www.soa.org. Make sure you work all of these exams!
All materials contained herein are copyrighted by Yufeng Guo. This PDF study manual is for
individual use for the sole purpose of taking Exam MLC. Reselling this manual is prohibited.
Redistribution of this manual in any form is prohibited.
Please check www.archactuarial.com for errata and answers to suggested text exercises.
ACTUARIAL MATHEMATICS:
CHAPTER 3 SURVIVAL
DISTRIBUTIONS AND LIFE
TABLES
This text forms the heart and soul of the exam syllabus. The basic principles of life insurance
(and annuities) are explained throughout the book. You need to have a solid understanding
of this material in order to pass the exam. However, you do not need to understand the
majority of the underlying theory in this text. The key points that a student must learn
from this text are:
KEYPOINTS:
1. Notation – much of this notation is new. While it can be confusing at first, there is some
logic to it. It will help you to remember and understand the many symbols
if you regularly translate the notation into words as you read.
2. Basic ideas – for example, chapter four introduces a variety of types of insurance. You
will want to make sure you have an understanding of these different products and their
benefit designs. Another key point is that there are many parallels. Again in Chapter 4,
the first part of the chapter considers products which pay a benefit immediately upon
death. The second part of the chapter considers the same products except that the
benefit is paid at the end of the year in which death occurs. It is helpful to realize that
you are really learning only one set of products, with a couple of benefit options, rather
than two sets of products. These parallels run throughout the text (e.g., continuous vs
curtate functions).
13
actuary88.com Chapter 1
3. Learn key formulas – there is no substitute for being able to recall the formula for,
say, a net level premium reserve for term insurance. If you can do this for most of the
formulas, you will be ready to answer questions quickly. This manual has tools to help
you learn these formulas, so don’t feel overwhelmed!
To the text!!!
Chapter 3 is all about notation, definitions, and a few basic ideas that are essential to
life contingencies. If you can make yourself comfortable with the symbols and methods of
Chapter 3, the rest of Actuarial Mathematics will be easier to absorb.
Definitions
X = newborn’s age at death
You can also think of X as “the future lifetime of a newborn”.
Define F (x) = Pr (X ≤ x), where x ≥ 0. Read as “the probability that death will occur
prior to (or at) age x”. In statistics, F (x) is the cumulative distribution function for the
future lifetime of a newborn. If y > x, it is always true that F (y) > F (x).
This makes sense. For a newborn, F (98), the probability of dying before age 98, is greater
than F (94), the probability of dying before age 94.
Define s(x) = 1 − F (X) = 1 − Pr(X ≤ x). The function s(x) is a survival function. Read
it as “the probability that death does not occur by age x” or “the probability of attaining
(surviving to) age x”.
F(x) s(x)
the pdf
y=f(x)
O x z
The figure shows the probability distribution function f (x) for death at age x. For any value
of x, F (x) is equal to the area under the curve y = f (x) and to the left of x. Similarly s(x)
is equal to the area under the curve and to the right of x.
By the way, you may have noticed that in our discussion, we dropped the subscript X in
FX (x) ... you can ignore it. I don’t know if the authors realize it but they are being a little
inconsistent in their treatment of F and s! If two different random variables, say X and Y ,
referred to the future lifetimes of two different newborns, then you would need to keep the
F and s straight for each kid. That’s all the subscript is indicating.
Newborns are great, but if our pension and insurance companies are going to make money
we need to be able to deal with people who are older than 0. So ...
= probability that person living at age x will die between ages x and z
= the probability that an x-year-old will die before turning z
[F (z) − F (x)]
=
[1 − F (x)]
[s(x) − s(z)]
=
[s(x)]
Why is this a conditional probability? Because it is the probability that a newborn will die
before age z given that the newborn survives to age x.
EXAMPLE:
1. Write two expressions (one with F only and one with s only) for the prob-
ability that a newborn dies between 17 and 40, assuming the newborn dies
between 10 and 40.
2. Interpret the following expression in English (or the language of your choice!).
S(20) − S(35)
1 − S(80)
SOLUTION:
1.
F (40) − F (17) s(17) − s(40)
or
F (40) − F (10) s(10) − s(40)
2. “The probability of death between ages 20 and 35, given that the newborn
will not attain age 80.” ♦
Now, let the symbol “(x)” represent a person age x and let T (x) be the future lifetime of a
person age x. (So T (25) is the future lifetime of (25), a twenty-five-year-old.)
Two basic probability functions exist regarding T (x):
q
t x tp x
x+t Age
x
In the figure, t qx is the probability that (x)’s death will occur in the age-interval (x, x+t), and
t px is the probability that (x)’s death will occur in the age interval (x + t, ω). (ω represents
the oldest possible age to which a person may survive.)
Useful notes:
• t p0 is just s(t).
Remember, these are the two basic functions. The formulas that follow are simply take-offs
on t px or t qx which you will learn with practice.
The symbol
t|u qx
represents the probability that (x) (that is, a person age x) survives at least t more years,
but dies before reaching age x + t + u. This is equal to each of the following expressions, each
of which you want to be able to put into words:
t+u qx − t qx
t px − t+u px
(As with qx and px , if u = 1, we drop it, leaving t| qx , the probability that (x) will survive t
years but not t + 1 years.)
Useful formulas:
x+t p0 s(x + t)
t px = =
x p0 s(x)
s(x + t)
t qx =1−
s(x)
s(x + t) − s(x + t + u)
t|u qx =
s(x)
s(x + t) s(x + t) − s(x + t + u)
= ∗
s(x) s(x + t)
= t px ∗ u qx+t
This last equation makes sense. It says “The probability of (x) dying between t and t + u
years from now (t|u qx ) is equal to the probability that (x) will first survive t years (t px ) and
then die within u years (u qx+t ).”
If you don’t remember anything else from the above, remember the following!
s(x + t)
t px =
s(x)
CONCEPT REVIEW:
1. Write the symbol for the probability that (52) lives to at least age 77.
2. Write the symbol for the probability that a person age 74 dies before age 91.
3. Write the symbol for probability that (33) dies before age 34.
4. Write the symbol for probability that a person age 43 lives to age 50, but
doesn’t survive to age 67.
5. Write 5|6 qx in terms of F and then in terms of p.
SOLUTIONS:
1. 25 p52 2. 17 q74 3. q33 4. 7|17 q43
= 5 px (1 − 6 px+5 ) or = 5 px − 11 px . ♦
To help memorize symbols, practice translating symbols into words and express words in
symbols. You can also make flash cards and quiz yourself.
3.2.3 Curtate-Future-Lifetime
• Option A reference: Actuarial Mathematics Chapter 3.2.3
Suppose a person born on Jan 1, 1900 died on Sept 30, 1990. How old was he at death?
The true age was about 90.75 years old. The curtate age was 90. To find the curtate age,
first find the true age. Next, throw away all the decimals and keep the integer. If there’s no
decimal, then the curtate age is equal to the continuous (true) age. For example, if T (x) = 90,
then K(x) = 90. (This book and others use ‘Curtate’ and ‘Discrete’ interchangeably.)
Previously, we defined T (x) to be the future lifetime of (x). This is a continuous function.
Now we define
EXAMPLE:
100−x
If s(x) = 100 for every x, what is the probability that K = 19 for (18)?
SOLUTION:
s(37) − s(38)
Pr(K(18) = 19) = 19| q18 =
s(18)
63 − 62 1
= = . ♦
82 82
The force of mortality can be thought of as the probability of death at a particular instant
given survival up to that time. This is an instantaneous measure, rather than an interval
measure. There is good bit of theory in this section, but the most important items are the
following formulas and the table of relationships.
It is very important to know the relationships and requirements given in Table 3.2.1. These
will probably be tested on the exam. Below is a summary of the useful information in this
table. Each row shows 4 ways to express the function in the left column.
Rx
e−
R∞ µ(t) dt
s(x) 1 − F (x) s(x) x f (u) du 0
Rx
f (x) F 0 (x) −s0 (x) f (x) µ(x) e− 0
µ(t) dt
SOLUTION:
1. Rx
s(x) = e− 0
µ dt
= e−µx .
2.
s(x + t)
t px = = e−µt . ♦
s(x)
Life Table is widely used actuarial practice. Even today, Life Tables are often loaded into
systems for calculating reserves, premium rates, and the surrender cash value of an insurance
policy. Learning Life Tables will not only help you pass Exam MLC, it also helps you when
you become an actuary.
Definitions:
Relationships:
lx = l0 ∗ s(x)
lx − lx+1
qx =
lx
lx − lx+n
n qx =
lx
lx+n
n px =
lx
d
n x = lx − lx+n
1. Find s(42).
2. Find 40 d2 .
3. Find 38 q3 .
4. Find 2| q40 .
SOLUTION:
1. s(42) = 92,595.7
100,000 = 0.925957.
2. 40 d2 = l2 − l42 = 5230.6
3. 38 q3 = 1 − 38 p3 = 1 − = 1 − 92,872.6
l41
l3 97,706.6 = 0.04947.
4. 2| q40 = 2 p40 · q42 = 92,595.7
93,131.6 (0.0032) = 0.003182. ♦
The complete-expectation-of-life is the expected value of T (x) (or E[T (x)] for fans of
◦
Statistics) and is denoted ex . If you remember how to find the expected value of a continuous
random variable, you can figure out that
Z ∞
◦
ex = E[T (x)] = t px dt
0
Z ∞
◦2
Var[T (x)] = 2 t · t px dt− ex (3.5.4)
0
The book shows how to figure both of these formulas out with integration by parts in Sec-
tion 3.5.1. I suggest that you memorize these two expressions.
The median future lifetime of (x) is denoted m(x) and simply represents the number m such
that m px = m qx . In other words, it is the number of years that (x) is equally likely to survive
or not survive. It can be found by solving any of the following:
1
Pr[T (x) > m(x)] =
2
or
s[x + m(x)] 1
=
s(x) 2
or
1
m px = .
2
∞
X
Var[K(x)] = (2k − 1) · k px − e2x
1
SOLUTION:
Z ∞ Z ∞ ∞
−1 −µt 1
◦
e0 = t p0 dt = e−µt dt = e =
0 0 µ 0 µ
Z ∞ Z ∞ ∞
−1 −µt 1
◦
e50 = t p50 dt = e−µt dt = e =
0 0 µ 0 µ
If the force of mortality is constant, your future expected lifetime is the same
whether you are 0 (a newborn) or 50. ♦
SOLUTION:
50 50 50
X X 50 − t 1 X
e0 = t p0 = = 50 − t
1 1
50 50 1
1 (50)(51)
= 50 − = 24.5
50 2
5 5 5
X X s(45 + t) X 5−t
e45 = t p45 = =
1 1
s(45) 1
5
4+3+2+1+0
= = 2. ♦
5
The expression mx is the central death rate over the interval x to x + 1. Make sure not to
confuse mx with m(x), the median future lifetime!
(lx − lx+1 )
mx =
Lx
Lx and mx can be extended to time periods longer than a year:
Z n
n Lx = lx+t dt
0
lx − lx+n
n mx =
n Lx
The remaining of Section 3.5.1 has obscure symbols Tx and α(x). They rarely show up in
the exam. Don’t spend too much time on them.
Let Tx be the total number of years lived beyond age x by the survivorship group with l0
initial members (i.e. the lx people still alive at age x). Be careful with notation. This is not
T (x), the future lifetime of (x).
Z ∞
Tx = lx+t dt (3.5.16)
0
The final symbol is α(x). It’s the expected death time given x dies next year.
R1 R1 R 1 lx+t R1
tf (t)dt tt px µ(x+t)dt t· l µ(x+t)dt t·lx+t µ(x+t)dt
α (x) = E [T |T < 1] = R 1
0
= R1
0
= R 1 lx+t
0 x
= R0 1
f (t)dt t px µ(x+t)dt lx
µ(x+t)dt lx+t µ(x+t)dt
0 0 0 0
R1 R1
t·cdt tdt
= R01 = 01 tdt = 12 .
R
If UDD, then f (t) = qx = c is a constant. Then α(x) = R 1
0
cdt dt
0 0
(A) L5
(B) m5
(C) T5
SOLUTION:
(A)
Z 1 Z 1
L5 = l5+t dt = t p5 l5 dt.
0 0
Since
l5 = l0 e−µ·5 = 1000e−0.5 = 606.5,
we have Z 1 h i1
L5 = 606.5 e−0.1t dt = 606.5 −10e−0.1t
0 0
h i
= 606.5 10(1 − e−0.1 ) = 577.16.
(B)
l5 − l6 606.5 − 548.8
m5 = = = 0.10
L5 577.16
This approximates the rate at which people were dying between the 5th and
6th years.
(C)
Z ∞ Z ∞
−0.1(5+t)
T5 = 1000e dt = 606.5 e−0.1t dt = 6065
0 0
Relationship:
Tx ◦
= ex
lx
◦
This relationship makes sense. It says that the average number of years lived, ex , by the
members of lx is equal to the total number of years lived by this group divided by lx .
We can determine the average number of years lived between x and x + n by the lx survivors
at age x as:
Z n
n Lx
= t px dt
lx 0
n Lx
= n-year temporary complete life expectancy of (x)
lx
◦
= ex:n (p.71)
These are basically ways to avoid working integrals. They are based on the Trapezoid Rule
for integration – maybe you remember the trapezoid rule from calculus.
Backward:
u(x) = c(x) + d(x) ∗ u(x + 1)
Forward:
u(x) − c(x)
u(x + 1) =
d(x)
Note that the Forward Method is simply an algebraic recombination of the Backward Method.
Note also that this Forward formula is different from the book – work out the formulas
yourself to convince yourself of their equivalence. Then, learn whichever form you find more
straightforward.
◦ ◦
The text shows how to use these formulas to compute ex and ex starting with eω and eω and
working backward. For ex , using the recursion once will produce eω−1 , the second iteration
will produce eω−2 , etc. until you get all the way back to e0 , when you will have produced a
list of ex for every x between 0 and ω.
c(x) = px
d(x) = px
◦
For ex ,
◦
u(x) =ex
Z 1
c(x) = s px ds
0
d(x) = px
◦
Starting Value =eω = u(ω) = 0
The random variable T is a continuous measure of remaining lifetime. The life table has been
developed as an approximation of T , using a curtate variable K. As we’ve discussed, K is
only defined at integers. So, we need some way to measure between two integer ages. Three
popular methods were developed.
For all of the methods that follow, let x be an integer and let 0 ≤ t ≤ 1. Suppose that we
know the value of s(x) for the two integers x and x + 1 and we want to approximate s at
values between x and x+1. In other words, we want to approximate s(x+t) where 0 ≤ t ≤ 1.
This method is also known as “Uniform Distribution of Deaths”, or UDD. Under UDD,
s(x + t) and t px are both straight lines between t = 0 to t = 1. This method assumes that
the deaths occurring between ages x and x + 1 are evenly spread out between the two ages.
As you might imagine, this is usually not quite correct, but is a pretty good approximation.
(Please note: the linearity of s(x + t) and t px is only assumed to hold up to t = 1!)
One key formula for UDD you might want to memorize is:
f (t) = qx
To see why, please note that the number of deaths from time zero to time t is a fraction of
the total deaths in a year
Here for convenience we interpret s(x + t) as the number of people alive at age x + t. For
example, if s(x + 0.5) = 0.9, we say that for each unit of people at age x, we have 0.9 unit
of people at age x + 0.5, with one unit being one billion, one million, or any other positive
constant.
You can also come up with f (t) = qx by intuitive thinking. Under UDD, death occurs at
a constant speed. If 12 people died in one year, then one person died each month. So f (t)
must be a constant. Then:
Z 1 Z 1
qx = f (t)dt = f (t) dt = f (t)
0 0
Method 3: Harmonic Interpolation: This method is more commonly called the “Balducci
assumption” or the “Hyperbolic assumption.”
The above formula says that if you are x + t years old (where 0 ≤ t ≤ 1), then your chance
of dying in the remainder of the year is a fraction of your chance of dying in the whole year.
You can derive all the other formulas using 1−t qx+t = (1 − t)qx . Later I’ll show you how to
derive other formulas in Balducci assumptions.
Uniform Constant
Function Distribution Force Hyperbolic
tqx
t qx tqx 1 − ptx 1−(1−t)qx
qx qx
µ(x + t) 1−tqx − log px 1−(1−t)qx
(1−t)qx
1−t qx+t 1−tqx 1 − px1−t (1 − t)qx
yqx yqx
y qx+t 1−tqx 1 − pyx 1−(1−y−t)qx
px
t px 1 − tqx ptx 1−(1−t)qx
qx px
t px µ(x + t) qx −ptx log px [1−(1−t)qx ]2
Table 3.6.1
Table 3.6.1 summarizes UDD, constant force of mortality, and the Balducci assumption.
Don’t try to memorize the whole table. Learn basic formulas and derive the rest on the spot.
µx (t) = µ
Rt
− µdt
t px =e 0 = e−µt
px = e−µ
µ = − ln px
t
t px =e −µt
= (e−µ ) = (px )t
t+y px (px )t
y px+t = = (t+y)
= (px )y
t px (px )
s(x + t) = s(x)e−µt
You can derive all the other formulas from this starting point. For example, let’s derive
the formula for t px .
T (x) 0 t 1
Age x x+t x+1
px
Number of people alive s(x) = 1 s(x + t) = 1−(1−t)qx
s(x + 1) = px
px
This is how to derive s(x + t) = 1−(1−t)qx .
First, we set the starting population at s(x) = 1 for convenience. You can set it to any
positive constant and get the same answer. After setting s(x) = 1, we’ll have s(x + 1) = px .
This is because px = s(x+1)
s(x) .
Next, let’s find s(x + t) using the formula 1−t qx+t = (1 − t)qx .
s(x + 1) px
1−t qx+t =1− =1− = (1 − t)qx
s(x + t) s(x + t)
px
s(x + t) =
1 − (1 − t)qx
s(x+t) px
However, t px = s(x) = s(x + t). This gives us: t px = 1−(1−t)qx .
d d px qx px
f (t) = − t px = − =
dt dt 1 − (1 − t)qx [1 − (1 − t)qx ]2
Derive y qx+t :
s(x + t + y)
y qx+t =1−
s(x + t)
px
Use the formula: s(x + t) = 1−(1−t)qx
Replace t with t + y, assuming 0 ≤ t + y ≤ 1:
px
s(x + t + y) =
1 − (1 − t − y)qx
s(x + t + y) 1 − (1 − t)qx px
y qx+t =1− =1− =
s(x + t) 1 − (1 − t − y)qx 1 − (1 − t − y)qx
1 1−t t
Finally, let’s derive s(x+t) = s(x) + s(x+1)
px
s(x + t) =
1 − (1 − t)qx
1−t t t
+ =1−t+
s(x) s(x + 1) px
1 1−t t
⇒ = +
s(x + t) s(x) s(x + 1)
Now you see that you really don’t need to memorize Table 3.6.1. Just memorize the following:
◦ 1
ex = ex +
2
1
Var(T ) = Var(K) +
12
EXAMPLE:
You are given that qx = 0.1. Find (A) 0.5 qx , (B) 0.5 qx+0.5 under each of
SOLUTION:
(A) • UDD:
0.5 qx = (0.5)(0.1) = 0.05
• CF:
0.5 qx = 1 − (0.9)0.5 = 0.0513
• Balducci:
(0.5)(0.1)
0.5 qx = = 0.0526.
1 − (0.5)(0.1)
(B) • UDD:
(0.5)(0.1)
0.5 qx+0.5 = = 0.0526
1 − (0.5)(0.1)
• CF:
0.5 qx+0.5 = 1 − (0.9)0.5 = 0.0513
• Balducci:
0.5 qx+0.5 = (0.5)(0.1) = 0.05. ♦
Although computers have rendered Analytical Laws of Mortality less imperative to our pro-
fession, they are still important for understanding mortality and particularly for passing the
exam. The book describes four basic analytical laws/formulas. Of these 4 laws, De Moivre’s
Law is frequently tested in the exam. The other 3 laws are rarely tested in the exam (I would
skip these 3 laws).
De Moivre’s Law:
x
µ(x) = (ω − x)−1 and s(x) = 1 − , where 0 ≤ x < ω
ω
Gompertz’ Law:
µ(x) = Bcx and s(x) = exp[−m(cx − 1)]
B
where B > 0, c > 1, m = , x≥0
log c
Makeham’s Law:
Weibull’s Law:
µ(x) = kxn and s(x) = exp(−uxn+1 )
k
where k > 0, n > 0, u = , x≥0
(n + 1)
Notes:
• In Makeham’s law, A is the “accident hazard” while Bcx is the “hazard of aging”.
We will be seeing more of these laws later in the book. To be ready for the exam, you should
become intimately familiar with De Moivre’s Law. DeMoivre’s Law says that at age x, you
are equally likely to die in any year between x and ω.
Here are some key life-functions for DeMoivre’s Law in the form of an example. If you don’t
read the proofs, still make an effort to understand the formulas and what they mean.
EXAMPLE:
Under DeMoivre’s law, show that
1.
ω−x−t
t px =
ω−x
2.
t
t qx =
ω−x
3.
◦ ω
e0 =
2
4.
ω−1
e0 =
2
5.
◦ ω−x
ex =
2
6.
ω−x−1
ex =
2
SOLUTION:
1.
s(x + t) ω−x−t
t px = =
s(x) ω−x
2.
t
t qx = 1 − t px =
ω−x
3. " #
(ω − t)2
Z ω
◦ ω−t ω
e0 = dt = − =
0 ω 2ω 2
4.
ω ω ω
X X ω−k 1X ω−1
e0 = k p0 = =ω− k=
1 1
ω ω 1 2
5. Z ω−x
◦ ω−x−t ω−x
ex = dt =
0 ω−x 2
6. Since mortality is uniform over all years under DeMoivre’s law, it is uniform
over each individual year, so UDD applies. Therefore
◦ 1 ω−x−1
ex =ex − =
2 2
Suppose you are trying to issue life insurance policies and two 45 year-old women apply for
policies. You want to make sure you charge appropriate premiums for each one to cover the
cost of insuring them over time. One of the women is simply picked from the population
at large. The second women was picked from a group of women who recently passed a
comprehensive physical exam with flying colors – significantly healthier than the general
population.
Is it equitable to charge the same premium to the two women? No – because you have
additional information about the second woman that would cause you to expect her to have
better “mortality experience” than the general population. Thus, to her premiums, you might
apply a “select” mortality table that reflects better the mortality experience of very healthy
45 year olds. However, after 15 years, research might show that being very healthy at 45
does not indicate much of anything about health at age 60. So, you might want to go back
to using standard mortality rates at age 60 regardless of status at age 45. After all, 15 years
is plenty of time to take up smoking, eat lots of fried foods, etc.
This simple scenario illustrates the idea behind select and ultimate tables. For some period
of time, you expect mortality to be different than that for the general population – the “select
period”. However, at some point, you’re just not sure of this special status anymore, so those
folks fall back into the pack at some point – the “ultimate” table.
Consider table 3.8.1. The symbol [x] signifies an x-yr old with “select” status. Note that
for the first two years (columns 1,2), select mortality applies with q[x] and q[x]+1 . However,
at duration 3 (column 3), it’s back to standard mortality, qx+2 . This table assumes the
“selection effect” wears off in just 2 years.
Why would this hold? The expression qx represents a pick from the general population.
The expression q[x] indicates special knowledge about the situation – for example, recently
passing a physical exam (This formula assumes we are trying to select out healthy people,
of course). The expression q[x−1]+1 indicates special knowledge about the situation as before
– for example, recently passing a physical exam, but this time the applicant has had a year
for health to deteriorate since she was examined at age x − 1 (one year ago) rather than at
age x.
SOLUTION:
3 q32deals only with the ultimate table so I am only interested in the values of
lx + 2 in Column (6).
l35 9,888
3 q32 = 1 − 3 p32 = 1 − =1− = 0.001313
l32 9,901
For 3 q[32] , we need l[32] and l[32]+3 which is just l35 since the select period is only
2 years. So
l35 9,888
3 q[32] = 1 − =1− = 0.001111
l[32] 9,899
So the answer is 0.202. ♦
• The probability that [32] will die in the next 3 years is lower if [32] is taken from a
select group. People you are sure are healthy should be less likely to die than someone
drawn from the general population.
• To follow the people alive from the 9898.75 ‘selected’ at age 32, you first follow the
numbers to the right until you hit the ultimate column and then proceed down the
ultimate column. This is useful! You can quickly evaluate that
9882
5 p[31] =
9903
by counting off 5 years – 2 to the right and then 3 down.
Conclusion
Chap. 3 introduces a lot of new concepts and notation. Make sure you understand the
notation in Table 3.9.1 – this is the foundation for the rest of the text.
Chapter 3 Suggested Problems: 1(do first row last), 5, 6, 7, 9, 12, 18abc, 20, 28,
30, 36, 39 There are lots for this chapter, some chapters in this book will have very few.
(Solutions are available at archactuarial.com on the Download Samples page.)
h R i
t
−
0
µ(x+s)ds s(x + t)
t px =e t px = t qx = 1 − t px
s(x)
Pr(K(x) = k) = k px − k+1 px
= k px ∗ qx+k
= k| qx
Life Tables:
lx = l0 ∗ s(x) n dx = lx − lx+n
lx − lx+1 lx+1
qx = px =
lx lx
lx − lx+n lx+n
n qx = n px =
lx lx
◦ 1 1
ex = Var[T ] =
µ µ2
∞
X ∞
X
ex = E[K(x)] = k px Var[K(x)] = (2k − 1) · k px − e2x
1 1
◦ 1 1
ex = ex + Var(T ) = Var(K) +
2 12
Make sure not to confuse mx with m(x), the median future lifetime!
Z 1 Z n
Lx = lx+t dt n Lx = lx+t dt
0 0
Z n
Tx ◦ n Lx ◦
= ex = t px dt =ex:n
lx lx 0
a(x) is the average number of years lived between age x and age x + 1 by those of the
survivorship group who die between age x and age x + 1. With the assumption of uniform
distribution of deaths over the interval (x, x + 1),
a(x) = 1/2.
Without this assumption,
R1 1 R
t · lx+t µ(x + t)dt
0 t · t px µ(x + t)dt
a(x) = R 1 = 0R 1
0 lx+t µ(x + t)dt 0 t px µ(x + t)dt
By the way, a(x) is a minor concept in MLC. I’ll be surprised if SOA tests this obscure
concept. I would skip it.
x
De Moivre’s Law: µ(x) = (ω − x)−1 and s(x) = 1 − , where 0 ≤ x < ω
ω
B
where B > 0, c > 1, m = , x≥0
log c
1 c
µx (t) = ω−x−t µx (t) = ω−x−t
Be careful on the exam - Modified DeMoivre problems are often disguised in a question that
starts something like
You are given
x 2
• s(x) = 1 − 80
• ···
In cases like this, you have to recognize that the question is just a modified DeMoivre written
in a different algebraic form.
Note that in this table, the function listed at left are given in terms of the functions across
the top row!
Rx
e−
R∞ µ(t) dt
s(x) 1 − F (x) s(x) x f (u) du 0
Rx
f (x) F 0 (x) −s0 (x) f (x) µ(x) e− 0
µ(t) dt
qx qx
µ(x + t) 1−tqx − log px 1−(1−t)qx
(1−t)qx
1−t qx+t 1−tqx 1 − px1−t (1 − t)qx
yqx yqx
y qx+t 1−tqx 1 − pyx 1−(1−y−t)qx
px
t px 1 − tqx ptx 1−(1−t)qx
qx px
t px µ(x + t) qx −ptx log px [1−(1−t)qx ]2
1. Given:
◦
(i) e0 = 25
(ii) lx = ω − x, 0 ≤ x ≤ ω
(iii) T (x) is the future lifetime random variable.
Solution:
ω2
Z ω
t ω
◦
e0 = 1− dt = ω − = = 25 ⇒ ω = 50
0 ω 2ω 2
402
Z 40
t
◦
e10 = 1− dt = 40 − = 20
0 40 (2)(40)
" #40
t2 t3
Z 40
t
Var [T (x)] = 2 t 1− dt − (20)2 = 2 − − (20)2 = 133
0 40 2 3 · 40 0
Key: C
Calculate the probability that a person age 80.5 will die within two years.
(A) 0.0782 (B) 0.0785 (C) 0.0790 (D) 0.0796 (E) 0.0800
Solution:
q81
0.0408 = µ(81.5) = ⇒ q81 = 0.0400
1 − (1/2)q81
Similarly,
q80 = 0.0200 and q82 = 0.0600
h i
2 q80.5 = 1/2 q80.5 + 1/2 p80.5 q81 + p81 · 1/2 q82
0.01 0.98
= + [0.04 + 0.96(0.03)] = 0.0782
0.99 0.99
Key: A
3. Mortality for Audra, age 25, follows De Moivre’s law with ω = 100. If she takes up hot
air ballooning for the coming year, her assumed mortality will be adjusted so that for
the coming year only, she will have a constant force of mortality of 0.1.
Calculate the decrease in the 11-year temporary complete life expectancy for Audra if
she takes up hot air ballooning.
(A) 0.10 (B) 0.35 (C) 0.60 (D) 0.80 (E) 1.00
Solution: STANDARD:
t2
Z 11
◦ t
e25:11 = (1 − )dt = t − |11 = 10.1933
0 75 2 × 75 0
MODIFIED: R1
p25 = e− 0
0.1ds
= e−0.1 = 0.90484
Z 1 Z 10
◦ t
e25:11 = t p25 dt + p25 (1 −
)dt
0 0 74
Z 1 Z 10
t
= e−0.1t dt + e−0.1 (1 − )dt
0 0 74
!
1−e −0.1 t 2
−0.1
= +e t− |10
0.1 2 × 74 0
= 0.95163 + 0.90484(9.32432) = 9.3886
◦ ◦ ◦
Here we use a recursive formula: ex:n = ex:1 + 1 px × ex+1:n−1
The difference is 0.8047. Key: D
4. You are given the following extract from a select-and-ultimate mortality table with a
2-year select period:
(A) 0.0102 (B) 0.0103 (C) 0.0104 (D) 0.0105 (E) 0.0106
Solution:
l[60]+0.6 = (0.6)(79,954) + (0.4)(80,625) = 80,222.4
80,222.4 − 79,396.5
0.9 q[60]+0.6 = = 0.0103
80,222.4
Key: B
5. Given:
Calculate F .
(A) -0.20 (B) -0.09 (C) 0.00 (D) 0.09 (E) 0.20
Solution: R 0.4
− (F +e2x )dx
0.4 p0 = 0.5 = e 0
h i0.4
e2x e0.8 −1
−0.4F − 2 −0.4F − 2
=e 0 =e
⇒ 0.5 = e−0.4F −0.6128 ⇒ ln(0.5) = −0.4F − 0.6128
⇒ −0.6931 = −0.4F − 0.6128 ⇒ F = 0.20
Key: E
6. An actuary is modeling the mortality of a group of 1000 people, each age 95, for the
next three years.
The actuary starts by calculating the expected number of survivors at each integral age
by
l95+k = 1000 k p95 , k = 1, 2, 3
The actuary subsequently calculates the expected number of survivors at the middle of
each year using the assumption that deaths are uniformly distributed over each year of
age.
Age Survivors
95 1000
95.5 800
96 600
96.5 480
97 −−
97.5 288
98 −−
The actuary decides to change his assumption for mortality at fractional ages to the
constant force assumption. He retains his original assumption for each k p95 .
(A) 270 (B) 273 (C) 276 (D) 279 (E) 282
l96 + l97
Solution: From UDD, l96.5 = .
2
600 + l97
480 = −→ l97 = 360
2
Likewise, from l97 = 360 and l97.5 = 288, we get l98 = 216.
l98 216
For constant force, e−µ = = = 0.6
l97 360
1
0.5 px = e−0.5µ = (0.6) 2 = 0.7746
l97.5 = (0.7746)l97 = (0.7746)(360) = 278.86
Key: D
7. For a 4-year college, you are given the following probabilities for dropout from all causes:
• q0 = 0.15
• q1 = 0.10
• q2 = 0.05
• q3 = 0.01
(A) 1.25 (B) 1.3 (C) 1.35 (D) 1.4 (E) 1.45
Solution: Z 1.5
◦
e1:1.5 = t p1 dt
0
Z 1 Z 0.5
= t p1 dt + 1 p1 x p2 dx
0 0
Z 1 Z 0.5
= (1 − 0.1t)dt + 0.9 (1 − 0.05x)dx
0 0
" #1 " #0.5
0.1t2 0.05x2
= t− + 0.9 x −
2 0
2 0
= 0.95 + 0.444 = 1.394 Key: D
8. For a given life age 30, it is estimated that an impact of a medical breakthrough will
◦
be an increase of 4 years in e30 , the complete expectation of life.
Prior to the medical breakthrough, s(x) followed de Moivres law with ω = 100 as the
limiting age.
Assuming de Moivres law still applies after the medical breakthrough, calculate the
new limiting age.
◦ ω − 30
Solution: For deMoivre’s law, e30 = .
2
◦ 100−30
Prior to medical breakthrough ω = 100 ⇒e30 = 2 = 35.
◦0 ◦
After medical breakthrough e30 =e30 +4 = 39.
◦0 ω 0 − 30
⇒e30 = 39 = ⇒ ω 0 = 108. Key E
2
(i)
x q[x] q[x]+1 q[x]+2 qx+3 x + 3
60 0.09 0.11 0.13 0.15 63
61 0.10 0.12 0.14 0.16 64
62 0.11 0.13 0.15 0.17 65
63 0.12 0.14 0.16 0.18 66
64 0.13 0.15 0.17 0.19 67
(ii) White was a newly selected life on 01/01/2000.
(iii) White’s age on 01/01/2001 is 61.
(iv) P is the probability on 01/01/2001 that White will be alive on 01/01/2006.
Calculate P .
(A) 0 ≤ P < 0.43 (B) 0.43 ≤ P < 0.45 (C) 0.45 ≤ P < 0.47
(D) 0.47 ≤ P < 0.49 (E) 0.49 ≤ P ≤ 1.00
Solution:
5 p[60]+1 = 1 − q[60]+1 1 − q[60]+2 (1 − q63 ) (1 − q64 ) (1 − q65 )
= (0.89)(0.87)(0.85)(0.84)(0.83) = 0.4589
Key: C
◦
Calculate e25:25 .
(A) 14.0 (B) 14.4 (C) 14.8 (D) 15.2 (D) 15.6
Solution: Z 15 Z 10
◦
e25:25 = t p25 dt + 15 p25 t p40 dt
0 0
Z 15 R 15 Z 10
−0.04t − 0.04 ds
= e dt + e 0 e−0.05t dt
0 0
1 1
= 1 − e−0.60 + e−0.60 1 − e−0.50
0.04 0.05
= 11.2797 + 4.3187 = 15.60
Key: E
Calculate 10 p40 .
(A) 0.81 (B) 0.85 (C) 0.88 (D) 0.92 (E) 0.96
Z ∞ Z 50 Z ∞
1= fT (t) dt = kf1 (t) dt + 1.2 f2 (t) dt
0 0 50
= kF1 (50) + 1.2 (F2 (∞) − F2 (50))
= k (1 − 50 p0 ) + 1.2(1 − 0.5)
= k(1 − 0.8951) + 0.6
1 − 0.6
⇒k= = 3.813
1 − 08951
For x ≤ 50, Z x
FT (x) = 3.813f1 (t) dt = 3.813F1 (x)
0
(A) 0.45 (B) 0.53 (C) 0.58 (D) 0.64 (E) 0.73
SOLUTION:
We can use the exponential formulation for p0 (x)
R1 (1)
R1 (τ )
R1
0 0 0.2kt2 dt
0.04 = qx(1) = 1−px(1) = 1−e− 0
µx (t) dt
= 1−e− 0
0.2µx (t) dt
= 1−e− 0 = 1−e−0.2k/3
⇒ e−0.2k/3 = 0.96
µ(1) (τ )
x (t) = 0.2µx (t) ⇒ µ(2) (τ )
x (t) = 0.8µx (t)
Z 2 Z 2
(2) (τ )
2 qx = t px · µ(2)
x (t)dt = (τ )
t px · (0.8)µ(τ ) (τ )
x (t)dt = 0.82 qx
0 0
(τ )
To get 2 qx , we use
R2 (τ )
R2
− kt2 dt
(τ )
2 qx = 1 − 2 p(τ )
x =1−e 0
µx (t) dt
= 1 − e− 0 = 1 − e−8k/3
40
= 1 − e−0.2k/3 = 1 − (0.96)40
⇒ 2 qx(τ ) = 0.8046
⇒ 2 qx(2) = (0.8)(0.8046) = 0.644
(A) 332
(B) 352
(C) 372
(D) 392
(E) 412
SOLUTION:
◦
e30:40 = 040 t p30 dt
R
= 040 ω−30−t
R
ω−30 dt
h i
t2
= t− 2(ω−30) |40
0
800
= 40 − ω−30
= 27.692
⇒ ω = 95
65−t
t p30 = 65
Now, realize (after getting ω = 95) that T (30) is uniformly on (0, 65), its variance is
just the variance of a continuous uniform random variable:
2
V ar = (65−0)
12 = 352.08
Key: B
14. For a life table with a one-year select period, you are given:
◦
x l[x] d[x] lx+1 e[x]
(i) 80 1000 90 − 8.5
81 920 90 − −
(ii) Deaths are uniformly distributed over each year of age.
◦
Calculate e[81] .
(A) 8.0
(B) 8.1
(C) 8.2
(D) 8.3
(E) 8.4
SOLUTION:
Complete the table:
l81 = l[80] − d[80] = 910
l82 = l[81] − d[81] = 830
910
p[80] = 1000 = 0.91
830
p[81] = 920 = 0.902
p81 = 830
910 = 0.912
◦ ◦ ◦
Use the recursive formula: ex:n = ex:1 + 1 px × ex+1:n−1
◦ ◦ ◦
We have: e[80] = e80:1 + 1 p[80] × e81
◦
Notice under UDD, t px is a straight line. Since ex:1 is the area of the function t px
◦
bounded by t = 0 and t = 1, under UDD we have ex:1 = 0.5(0 px + 1 px ) = 0.5(1 + 1 px )
◦ R1
You can also derive the above equation using the standard formula: ex:1 = 0 t px dt =
R1 R1
0 (1 − t qx )dt = 0 (1 − tqx )dt
Since qx is a constant, you can verify that:
R1
0 (1 − tqx )dt = 1 − 0.5qx = 0.5(1 + 1 px )
◦ ◦ ◦
Hence: ex:n = 0.5(1 + 1 px ) + 1 px × ex+1:n−1 = 0.5(1 qx + 1 px + 1 px ) + 1 px × ex+1:n−1 =
◦
0.51 qx + 1 px (1 + ex+1:n−1 ).
Set n = ∞, we have: ėx = 21 qx + px (1 + ėx )
(A) 0.38
(B) 0.39
(C) 0.41
(D) 0.43
(E) 0.44
SOLUTION:
4 p50 = e−(0.05)(4) = 0.8187
−(0.05)(10) = 0.6065
10 p50 = e
−(0.04)(8) = 0.7261
8 p60 = e
18 p50 = (10 p50 )(8 p60 ) = 0.6065 × 0.7261 = 0.4404
16. For a population which contains equal numbers of males and females at birth:
(A) 0.076
(B) 0.081
(C) 0.086
(D) 0.091
(E) 0.096
SOLUTION:
s(60) = P [T (0) > 60] (i.e. the probability that a newborn survives to age 60)
Using the double expectation:
s(60) = P [T (0) > 60] = P (M aleN ewborn)∗P [T (M aleN ewborn) > 60]+P (F emalenewborn)∗
P [T (F emaleN ewborn) > 60] = 0.5e−(0.1)(60) + 0.5e−(0.08)(60) = 0.005354
Similarly,
s(61) = 0.5e−(0.1)(61) + 0.5e−(0.08)(61) = 0.00492
0.00492
q60 = 1 − 0.005354 = 0.081
Key: B
(A) 5.25 (B) 6.08 (C) 8.52 (D) 26.08 (E) 30.00
(A) 4.5 (B) 7.2 (C) 28.7 (D) 35.9 (E) 44.8
◦
3. Calculate ex
(A) 0.5 (B) 2.0 (C) 10.0 (D) 20.0 (E) 40.0
4. Calculate Var[T ].
(A) 0.25 (B) 0.50 (C) 1.00 (D) 2.00 (E) 4.00
• q60 = 0.3
• q61 = 0.4
• f is the probability that (60) will die between ages 60.5 and 61.5 under the uniform
distribution of deaths assumption.
• g is the probability that (60) will die between ages 60.5 and 61.5 under the Balducci
assumption.
Calculate 10,000(g − f ).
(A) 0 (B) 85 (C) 94 (D) 178 (E) 213
Solutions to Chapter 3
1. Key: A We need to find t such that t p20 = 0.5
Rt Rt
(60−s)−0.5 ds
⇒ 0.5 = e− 0
µ(20+s) ds
= e− 0
0.5 t √ √
= e2[(60−s) ]0 = e2[ 60−t− 60]
h√ √ i
ln 0.5 = 2 60 − t − 60
√
⇒ 60 − t = 7.40 ⇒ t = 5.25
2. Key: D Z ∞
Tx+1 = lx+1+t dt
0
6.4
lx+1 = =8 ⇒ lx+1+t = 8(0.8)t
0.8
Z ∞
8
Tx+1 = 8 (0.8)t dt = (0.8)t |∞
0
0 ln 0.8
−8
= = 35.9
ln 0.8
1
4. Key: A (CFM) Var[T ] = µ2
= 0.25
5. Key: C
0.5 = t px = e−µ·t = e−2t
ln 2
⇒t=
2
6. Key: E
lx − lx+1 lx (1 − e−µ )
mx = = R1
Lx 0 lx · t px dt
lx (1 − e−2 ) 1 − e−2
= R1 == 1 −2
=2
lx 0 e−µt dt 2 (1 − e )
(0.5)(0.3)
UDD: [1 − (0.5)(0.3)] 1−(0.5)(0.3) + (0.7) · (0.5)(0.4) = 0.2900 = f
10,000(g − f ) = 85