ACTUARIAL MATHEMATICS

Math 161 — Survival Models: Complete Exam Study Notes

Coverage: Ch. 2 Survival Models | Up to Curtate Future Lifetime

1. FUNDAMENTAL CONCEPTS & DEFINITIONS

1.1 The Life (x) and Future Lifetime

In actuarial mathematics, we study the future lifetime of a living individual. We use the following
notation:

Notation Meaning
(x) A life (person) currently aged x, where x ≥ 0
(0) A newborn life (aged 0)
(21) A life aged 21, e.g., a college student
T_x or T(x) The future lifetime of (x) — a continuous random variable
x + T_x The age-at-death of (x), also a random variable

KEY DEFINITION — Future Lifetime T_x

T_x is the time (in years) from age x until death. It is a continuous, non-negative random variable.
If a person aged x dies at age y, then T_x = y - x. Example: If (40) dies at age 55.3, then T_40 =
55.3 - 40 = 15.3 years.

1.2 The Distribution Function F_x(t) — Lifetime Distribution

The cumulative distribution function (CDF) of T_x is:
F_x(t) = P(T_x ≤ t) ← probability that (x) dies within t years
This is called the lifetime distribution from age x. It gives the probability that (x) does NOT survive
beyond age x + t.

Properties of F_x(t):
• F_x(0) = 0 (you are alive at time 0)
• lim F_x(t) = 1 as t → ∞ (everyone eventually dies)
• F_x(t) is non-decreasing in t
1.3 The Survival Function S_x(t)
Because insurance problems are often more interested in survival than death, we define the survival
function:
S_x(t) = 1 - F_x(t) = P(T_x > t) ← probability that (x) survives at least t more years

Properties of S_x(t):
• S_x(0) = 1 (alive at the start)
• S_x(t) is non-increasing in t (survival probability decreases over time)
• lim S_x(t) = 0 as t → ∞
• S_x(t) is differentiable for all t ≥ 0 (Assumption 1 from class)
• lim t · S_x(t) = 0 as t → ∞ (Assumption 2)
• lim t² · S_x(t) = 0 as t → ∞ (Assumption 3 — needed for variance formula)

1.4 Connecting S_x(t) to the Newborn's Survival Function S_0

Let S_0(x) = survival function of a newborn to age x. Being 'a life aged x' means T_0 > x. Using
conditional probability:
F_x(t) = P(T_x ≤ t) = P(T_0 ≤ x+t | T_0 > x) = [F_0(x+t) - F_0(x)] /
S_0(x)

Therefore, the survival function from age x is:

S_x(t) = 1 - F_x(t) = S_0(x+t) / S_0(x) ← FUNDAMENTAL FORMULA

IMPORTANT RESULT — Chain Rule for Survival

S_0(x + t) = S_0(x) · S_x(t) This says: the probability of surviving from birth to age x+t equals
(probability of surviving from birth to x) times (probability of surviving t more years given you are
already aged x).

1.5 The Multiplicative Property of Survival

For any ages and time periods:
S_x(t + u) = S_x(t) · S_{x+t}(u) ← survival for t+u years = survive t years × survive u
more years

Base case (set x = 0):

S_0(x + t) = S_0(x) · S_x(t)

Rearranging gives:
S_{x+t}(u) = S_x(t + u) / S_x(t)
Interpretation: If we know the survival function from any starting age x, we can derive survival
probabilities from any future age x + t.

2. THE PDF OF T_x

2.1 Probability Density Function

Since F_x(t) is the CDF of T_x, the probability density function (pdf) is its derivative:
f_x(t) = d/dt F_x(t) = -d/dt S_x(t) ← rate of death at time t

Key relationship between pdf, survival function, and force of mortality:

f_x(t) = S_x(t) · mu_{x+t} ← density = survival × force of mortality at age x+t

Properties of f_x(t):
• f_x(t) ≥ 0
• ∫₀^∞ f_x(t) dt = 1 (total probability equals 1)
• At t = 0: f_x(0) = mu_x (since S_x(0) = 1)

3. FORCE OF MORTALITY μ_x

3.1 Definition
The force of mortality at age x is defined as:
mu_x = lim_{dx→0+} (1/dx) · P(T_x ≤ dx) ← instantaneous death rate at age x

Equivalently, in terms of the survival function:

mu_x = -(1/S_x(t)) · dS_x(t)/dt evaluated at t=0

Or more usefully, using S_0:

mu_x = -(1/S_0(x)) · dS_0(x)/dx = -d/dx [ln S_0(x)] ← Equation (2.9)

WHAT IS μ_x ? (Is it a probability? A rate? Something else?)

μ_x is the instantaneous rate of death at age x, given survival to age x. It is measured per year (if
time is in years). It is not a probability itself, but μ_x · dx ≈ P(T_x ≤ dx) for very small dx. Think of
it as: the speed at which people are dying at age x.
3.2 Force of Mortality at Future Age x + t
If x is fixed and t is variable, the force of mortality at age x + t is:
mu_{x+t} = -(1/S_x(t)) · dS_x(t)/dt = f_x(t) / S_x(t) ← Equation (2.10)

Key derivative result (used often!):

d/dt [_t p_x] = -mu_{x+t} · _t p_x

3.3 Expressing S_x(t) Using Force of Mortality

Starting from mu_x = -d/dx [ln S_0(x)] and integrating from x to x+t:
S_x(t) = exp( -∫_x^{x+t} mu_r dr ) = exp( -∫_0^t mu_{x+s} ds ) ←
FUNDAMENTAL

This is the most important formula connecting force of mortality to survival. Survival = exponential of
minus accumulated hazard.

FULL EQUIVALENCE (Key Insight)

The force of mortality function {mu_x, x ≥ 0} fully describes the lifetime distribution, just as S_0
does. If you know mu_x for all x, you can recover all survival probabilities and vice versa.

3.4 Visual Intuition for Force of Mortality

Imagine a timeline from age x to age x + t:
• mu_r is the instantaneous death rate at each age r along the way
• ∫_x^{x+t} mu_r dr is the total accumulated hazard from age x to x+t
• S_x(t) = exp( - accumulated hazard ) = the survival probability
• Higher force of mortality → faster accumulated hazard → lower survival

4. STANDARD MORTALITY MODELS

4.1 De Moivre's Law (Uniform Distribution)

Given: F_0(x) = 1 - (1 - x/120)^{1/6} for 0 ≤ x ≤ 120 [Example 2.1 from class]

Then: S_0(x) = 1 - F_0(x) = (1 - x/120)^{1/6}

Deriving force of mortality for this example:

mu_x = -(1/S_0(x)) · dS_0(x)/dx
 dS_0(x)/dx = (1/6)(1 - x/120)^{-5/6} · (-1/120) = -1/(720) ·
(1-x/120)^{-5/6}
mu_x = 1/(720-6x) = 1 / [6(120-x)]

4.2 Gompertz Law of Mortality

Gompertz (1825) proposed that the force of mortality increases exponentially with age:
mu_x = B · c^x , x > 0 , 0 < B < 1 , c > 1

The corresponding survival function is derived by integrating:

S_x(t) = exp( -B·c^x · (c^t - 1) / ln c ) ← derived from S_x(t) = exp(-∫ mu_{x+s}
ds)

The CDF is:

F_x(t) = 1 - exp( -B·c^x · (c^t - 1) / ln c )

The pdf is:

f_x(t) = exp( -B·c^x(c^t-1)/ln c ) · B·c^{x+t} , t ≥ 0

Quick checks:
• S_x(0) = exp(0) = 1 ✓ (alive at t=0)
• S_x(t) is monotone decreasing in t ✓
• c^x increases with age x → exponent becomes more negative → S_{x+Δ}(t) < S_x(t) for Δ > 0 ✓

GOMPERTZ STRUCTURE
The key quantity in S_x(t) is c^x — the 'risk multiplier' at age x. Older starting ages mean
exponentially higher mortality, which is a realistic feature. This is why Gompertz is widely used in
life tables and insurance pricing.

Makeham extension: mu_x = A + B·c^x (adds a constant background mortality A to Gompertz)

5. ACTUARIAL NOTATION (Standard Exam Notation)

5.1 Survival and Mortality Rates

These are the most-used notations in actuarial science. Learn them cold.

Notation Formula Interpretation

_t p_x P(T_x > t) = S_x(t) Prob. (x) survives at least t years. Called
 t-year survival probability.
_t q_x P(T_x ≤ t) = F_x(t) = 1 - _t p_x Prob. (x) dies within t years. t-year
mortality probability.
p_x _1 p_x = S_x(1) 1-year survival probability (t=1 omitted by
convention)
q_x _1 q_x = F_x(1) 1-year mortality probability (t=1 omitted)
_{u|t} q_x P(u < T_x ≤ u+t) = S_x(u) - u-year deferred t-year mortality: dies
S_x(u+t) between u and u+t years from now

5.2 Key Identities — Must Know!

These identities will appear in problems. Know them by heart:

Identity Meaning
_t p_x + _t q_x = 1 Survival and mortality are complementary
_{t+u} p_x = _t p_x · _u Chain rule: survive t years, then u more years
p_{x+t}
_{u|t} q_x = _u p_x - Deferred mortality = two survival probs subtracted
_{u+t} p_x
_t p_x = exp(-∫_0^t Survival as exponential of accumulated hazard
mu_{x+s} ds)
d/dt [_t p_x] = -mu_{x+t} Derivative of survival = force of mortality × survival
· _t p_x
f_x(t) = mu_{x+t} · _t pdf = force of mortality × survival probability
p_x
_t q_x = ∫_0^t mu_{x+s} · Mortality = integral of (death rate × survival) over time
_s p_x ds

5.3 Relationship Between Notation and Formulas

Since _t p_x = S_x(t), all the S_x formulas translate directly:
_t p_x = S_x(t) = S_0(x+t) / S_0(x)
mu_x = -(1 / _x p_0) · d(_x p_0)/dx
mu_{x+t} = -(1 / _t p_x) · d(_t p_x)/dt

6. USEFUL APPROXIMATION
6.1 Midpoint Approximation for q_x
When q_x is small (i.e., _s p_x ≈ 1 for 0 ≤ s ≤ 1), the midpoint rule gives:
q_x = ∫_0^1 mu_{x+s} · _s p_x ds ≈ mu_{x + 1/2}

This approximation says: the 1-year mortality rate is approximately equal to the force of mortality at the
midpoint of the year. Useful for checking and estimation.

7. MEAN AND VARIANCE OF T_x

7.1 Complete Expectation of Life ê_x

The mean future lifetime of (x) is called the complete expectation of life:
e°_x = E(T_x) = ∫_0^∞ t · f_x(t) dt

Using integration by parts (IBP) with u = t, dv = f_x(t) dt, together with the boundary conditions (t · _t
p_x → 0 as t → ∞), we get the elegant formula:
e°_x = ∫_0^∞ _t p_x dt ← MAIN FORMULA for mean lifetime

This says: the expected future lifetime of (x) equals the total area under the survival curve _t p_x.

ALTERNATIVE DERIVATION (Indicator Function Method)

For any non-negative random variable T ≥ 0, we have the identity: T = ∫_0^∞ 1_{T > t} dt Taking
expectations: E(T_x) = ∫_0^∞ P(T_x > t) dt = ∫_0^∞ _t p_x dt ✓ This is an elegant measure-
theoretic result that your instructor showed in class.

7.2 Second Moment E(T_x²)

Using the same indicator function trick, since T² = 2∫_0^T t dt, taking expectations:
E(T_x²) = 2 ∫_0^∞ t · _t p_x dt

7.3 Variance of T_x

Using the standard formula Var(X) = E(X²) - [E(X)]²:
Var(T_x) = 2 ∫_0^∞ t · _t p_x dt - (e°_x)² ← MAIN FORMULA for variance

EXAM TRAP — Don't Forget!

When computing Var(T_x), you need BOTH parts: 1. E(T_x²) = 2∫_0^∞ t · _t p_x dt 2.
[E(T_x)]² = (e°_x)² Var(T_x) = E(T_x²) - (e°_x)²
 8. CURTATE FUTURE LIFETIME K_x

8.1 Definition
The curtate future lifetime K_x is the floor (integer part) of T_x:
K_x = floor(T_x) = ⌊T_x⌋ ← a discrete random variable

K_x represents the number of complete years lived by (x) after age x before death.

Example (T_x value) K_x value

T_x = 15.3 yrs K_x = ⌊15.3⌋ = 15
T_x = 0.8 yrs K_x = ⌊0.8⌋ = 0
T_x = 50 yrs K_x = ⌊50⌋ = 50

8.2 Distribution of K_x

K_x takes non-negative integer values k = 0, 1, 2, 3, ...

The probability mass function (PMF) of K_x is:

P(K_x = k) = P(k ≤ T_x < k+1) = _k p_x - _{k+1} p_x = _k p_x ·
q_{x+k}

This says: (x) lives exactly k complete years, meaning (x) survives k years and then dies in the (k+1)-th
year.

Survival and CDF for K_x:

P(K_x ≥ k) = _k p_x ← (x) survives at least k years
P(K_x < k) = _k q_x = 1 - _k p_x ← (x) dies before completing k years

8.3 Mean of K_x — Curtate Expectation of Life

The mean of K_x (notation: e_x, without the circle) is the curtate expectation of life:
e_x = E(K_x) = sum_{k=0}^∞ P(K_x ≥ k+1) = sum_{k=1}^∞ _k p_x ← curtate
expectation

Note: the circle notation (e°_x) is for continuous T_x; no circle (e_x) is for curtate K_x.
 Notation Formula / Meaning
e°_x = ∫_0^∞ _t p_x dt Complete (continuous) expectation of life for T_x
e_x = Σ_{k=1}^∞ _k p_x Curtate (discrete) expectation of life for K_x

8.4 Variance of K_x

The variance of the curtate lifetime is:
Var(K_x) = 2 sum_{k=1}^∞ k · _k p_x - e_x · (e_x + 1) ← or use E(K_x²) -
[e_x]²

Alternatively, compute E(K_x²) = Σ k² · P(K_x = k) = Σ (2k-1) · _k p_x, then subtract e_x².

8.5 Relationship Between e°_x and e_x

Since T_x = K_x + fractional year remainder S_x (where S_x ∈ [0,1)):
e°_x = e_x + E(S_x)

Under the Uniform Distribution of Deaths (UDD) assumption, E(S_x) = 1/2, giving:
e°_x ≈ e_x + 1/2 ← under UDD assumption

9. WORKED EXAMPLES

9.1 Example — Survival, CDF, Probability Calculations

Given: F_0(x) = 1 - (1 - x/120)^{1/6} for 0 ≤ x ≤ 120 [Class Example 2.1]

So: S_0(x) = (1 - x/120)^{1/6}

① P(newborn survives beyond age 30):

S_0(30) = 1 - F_0(30) = (1 - 30/120)^{1/6} = (3/4)^{1/6} ≈ 0.9532
Answer: about 95.32% chance of surviving past age 30.

② P((30) dies before age 50) → use F_{30}(20):

F_{30}(20) = [F_0(50) - F_0(30)] / [1 - F_0(30)]
= [(1-(50/120)^{1/6}) - (1-(30/120)^{1/6})] / (30/120)^{1/6}
= [(0.75)^{1/6} - (7/12)^{1/6}] / (0.75)^{1/6} ≈ 0.0410
Answer: about 4.10% chance.

③ P((40) survives beyond 65) → use S_{40}(25):

S_{40}(25) = S_0(65) / S_0(40) = (1-65/120)^{1/6} / (1-40/120)^{1/6}
≈ 0.9395
Answer: about 93.95% chance.

9.2 Example — Gompertz Survival Function Derivation

Given mu_x = B·c^x. Find S_x(t).

Step 1: Use S_x(t) = exp( -∫_x^{x+t} B·c^r dr )

Step 2: ∫ B·c^r dr = (B/ln c) c^r + constant
Step 3: Evaluate from x to x+t:
S_x(t) = exp( -(B/ln c) · c^x · (c^t - 1) )

9.3 Example — q_x and mu_{x+1/2} Calculations

Using F_0(x) = 1-(1-x/120)^{1/6} from Example 2.1:

For p_x (1-year survival), we have:

p_x = _1 p_x = S_x(1) = S_0(x+1)/S_0(x) = (1 - (x+1)/120)^{1/6} / (1
- x/120)^{1/6}
p_x = (1 - 1/(120-x))^{1/6}

At x = 20: p_20 = (1 - 1/100)^{1/6} = 0.99^{1/6} ≈ 0.99833

So q_20 = 1 - 0.99^{1/6} ≈ 0.00167 (about 0.167%)

At x = 110: p_110 = (1 - 1/10)^{1/6} = 0.9^{1/6} ≈ 0.98259

So q_110 ≈ 0.01741 (about 1.741%)

Force of mortality at midpoint: mu_x = 1/[6(120-x)]

mu_{20.5} = 1/[6(99.5)] ≈ 0.001676 → ≈ q_20 ✓ (midpoint approximation works!)
mu_{110.5} = 1/[6(9.5)] ≈ 0.01754 → ≈ q_110 ✓

9.4 Example — Complete Expectation of Life

To find e°_x, compute ∫_0^∞ _t p_x dt. For De Moivre's law U(0, ω):
_t p_x = (ω - x - t)/(ω - x) for 0 ≤ t ≤ ω - x
e°_x = ∫_0^{ω-x} (ω-x-t)/(ω-x) dt = (ω-x)/2
For De Moivre U(0,120): e°_x = (120-x)/2. E.g., e°_40 = 40 years.

10. INTERPRETATION GUIDE

Your instructor includes interpretation questions. Here are the key phrases:

Expression How to interpret it

_t p_x = 0.94 A life currently aged x has a 94% probability of surviving at
least t more years.
_t q_x = 0.06 A life currently aged x has a 6% probability of dying within
the next t years.
mu_x = 0.02 At exact age x, the instantaneous death rate is 0.02 per year
(≈ 2% per year).
mu_x · dx ≈ p For a very small interval dx, p ≈ probability of dying between
ages x and x+dx given alive at x.
e°_x = 35 A life aged x is expected to live 35 more years on average.
e_x = 34 A life aged x is expected to live 34 more complete (whole)
years on average.
K_x = 14 The person died during their 15th year after age x (survived
exactly 14 complete years).
S_x(t) = exp(-∫) Survival probability decays as a function of accumulated
mortality hazard.

EXAM STRATEGY — Flow of Topics

Your instructor typically follows this flow in exam questions: 1. DEFINITION: 'What is T_x?' /
'What does _t p_x represent?' 2. FORMULA: 'Write the formula for S_x(t) in terms of S_0.' 3.
APPLICATION: 'Given F_0(x) = ..., calculate P((40) survives beyond 65).' 4. ANALYSIS: 'Derive
mu_x for the given model.' 5. INTERPRETATION: 'Interpret e°_x in the context of this model.'

11. FORMULA QUICK REFERENCE CARD

Formula What it is
T_x Future lifetime of (x) — continuous r.v.
K_x = ⌊T_x⌋ Curtate future lifetime of (x) — discrete r.v.
F_x(t) = P(T_x ≤ t) CDF = lifetime distribution (mortality side)
S_x(t) = 1 - F_x(t) Survival function
S_x(t) = S_0(x+t)/S_0(x) Survival from age x in terms of newborn's S_0
S_x(t+u) = S_x(t)·S_{x+t} Multiplicative property
(u)
_t p_x = S_x(t) Actuarial notation for survival
_t q_x = F_x(t) Actuarial notation for mortality
_{u|t} q_x = _u p_x - Deferred mortality
_{u+t} p_x
mu_x = -d/dx ln S_0(x) Force of mortality from S_0
mu_{x+t} = f_x(t)/S_x(t) Force of mortality at age x+t
S_x(t) = exp(-∫_0^t Survival from force of mortality
mu_{x+s} ds)
f_x(t) = mu_{x+t} · _t pdf of T_x
p_x
d/dt[_t p_x] = -mu_{x+t} Derivative of survival
· _t p_x
e°_x = ∫_0^∞ _t p_x dt Mean of T_x
Var(T_x) = 2∫_0^∞ t·_t Variance of T_x
p_x dt - (e°_x)²
e_x = Σ_{k=1}^∞ _k p_x Mean of K_x (curtate)
P(K_x=k) = _k p_x - PMF of curtate lifetime
_{k+1} p_x
q_x ≈ mu_{x+1/2} Midpoint approximation (when q_x small)
Gompertz: mu_x = Bc^x c>1, 0<B<1
Gompertz S_x(t) = exp(- Gompertz survival function
Bc^x(c^t-1)/lnc)

Good luck on your exam! Math 161 — Actuarial Mathematics | Survival Models Ch. 2