ACTS 4301 FORMULA SUMMARY Lesson 1: Probability Review
1. Var(X)= E[X2]- E[X]2
2. V ar(aX + bY ) = a2V ar(X) + 2abCov(X,Y ) + b2V ar(Y )
¯ V ar(X) 3. V ar(X) = n
4. EX [X] = EY [EX [X|Y ]] Double expectation 5. V arX [X] = EY [V arX [X|Y ]] + V arY (EX [X|Y ]) Conditional variance Distribution, Density and Moments for Common Random Variables Continuous Distributions
Name f(x) F (x) E[X] Var(X)
− x − x 2 Exponential e θ /θ 1 − e θ θ θ 1 x a+b (b−a)2 Uniform b−a , x ∈ [a, b] θ , x ∈ [a, b] 2 12 x x α−1 − θ α R 2 Gamma x e /Γ(α)θ 0 f(t)dt αθ αθ
Discrete Distributions
Name f(x) E[X] Var(X)
Poisson e−λλx/x! λ λ m x m−x Binomial x p (1 − p) mp mp(1 − p) 2
Lesson 2: Survival Distributions: Probability Functions, Life Tables
1. Actuarial Probability Functions tpx - probability that (x) survives t years tqx - probability that (x) dies within t years t|uqx - probability that (x) survives t years and then dies in the next u years.
t+upx = tpx ·u px+t
t|uqx = tpx ·u qx+t = = tpx −t+u px =
= t+uqx −t qx 2. Life Table Functions
dx = lx − lx+1
ndx = lx − lx+n dx qx = 1 − px = lx lx+t tpx = lx lx+t − lx+t+u t|uqx = lx 3. Mathematical Probability Functions
tpx = ST (x)(t)
tqx = FT (x)(t)
t|uqx = P r (t < T (x) ≤ t + u) = FT (x)(t + u) − FT (x)(t) = F (x + t + u) − F (x + t) = P r (x + t < X ≤ x + t + u|X > x) = X X sX (x) Sx(t + u) Sx+u(t) = Sx(u) S0(x + t) Sx(t) = S0(x) F0(x + t) − F0(x) Fx(t) = 1 − F0(x) 3
Lesson 3: Survival Distributions: Force of Mortality
f0(x) d µx = = − ln S0(x) S0(x) dx fx(t) d µx+t = = − ln Sx(t) Sx(t) dx Z t Z t Z x+t Sx(t) =t px = exp − µx+s ds = exp − µx(s) ds = exp − µs ds 0 0 x dtpx/dt d lnt px µx(t) = − = − tpx dt fT (x)(t) = fx(t) =t px · µx(t) Z t+u t|uqx = spx · µx(s) ds t Z t tqx = spx · µx(s)ds 0 ∗ ∗ −kt If µx(t) = µx(t) + k for all t, then spx = spxe If µx(t) =µ ˆx(t) +µ ˜x(t) for all t, then spx = spˆxsp˜x ∗ ∗ k If µx(t) = kµx(t) for all t, then spx = (spx) 4
Lesson 4: Survival Distributions: Mortality Laws
Exponential distribution or Constant Force of Mortality
µx(t) = µ −µt tpx = e Uniform distribution or De Moivre’s law 1 µ (t) = , 0 ≤ t ≤ ω − x x ω − x − t ω − x − t p = , 0 ≤ t ≤ ω − x t x ω − x t q = , 0 ≤ t ≤ ω − x t x ω − x u q = , 0 ≤ t + u ≤ ω − x t|u x ω − x Beta distribution or Generalized De Moivre’s law α µ (t) = x ω − x − t ω − x − tα p = , 0 ≤ t ≤ ω − x t x ω − x Gompertz’s law: x µx = Bc , c > 1 Bcx(ct − 1) p = exp − t x ln c
Makeham’s law: x µx = A + Bc , c > 1 Bcx(ct − 1) p = exp −At − t x ln c
Weibull Distribution n µx = kx −kxn+1/(n+1) S0(x) = e 5
Lesson 5. Survival Distributions: Moments
Complete Future Lifetime Z ∞ ˚ex = ttpxµx+t dt 0 Z ∞ ˚ex = tpx dt 0 1 ˚e = for constant force of mortality x µ ω − x ˚e = for uniform (de Moivre) distribution x 2 ω − x ˚e = for generalized uniform (de Moivre) distribution x α + 1 Z ∞ h 2i E (T (x)) = 2 ttpx dt 0 1 V ar (T (x)) = for constant force of mortality µ2 (ω − x)2 V ar (T (x)) = for uniform (de Moivre) distribution 12 (ω − x)2 V ar (T (x)) = for generalized uniform (de Moivre) distribution (α + 1)2(α + 2) n-year Temporary Complete Future Lifetime
Z n ˚ex:n = ttpxµx+t dt + nnpx 0 Z n ˚ex:n = tpx dt 0 ˚ex:n =n px(n) +n qx(n/2) for uniform (de Moivre) distribution
˚ex:1 = px + 0.5qx for uniform (de Moivre) distribution 1 − e−µn ˚e = for constant force of mortality x:n µ Z n h 2i E (T (x) ∧ n) = 2 ttpx dt 0
Curtate Future Lifetime ∞ X ex = kk|qx k=1 ∞ X ex = kpx k=1 ex =˚ex − 0.5 for uniform (de Moivre) distribution ∞ h 2i X E (K(x)) = (2k − 1)kpx k=1 1 V ar (K(x)) = V ar (T (x)) − for uniform (de Moivre) distribution 12 6 n-year Temporary Curtate Future Lifetime n−1 X ex:n = kk|qx + nnpx k=1 n X ex:n = kpx k=1 ex:n =˚ex:n − 0.5nqx for uniform (de Moivre) distribution n h 2i X E (K(x) ∧ n) = (2k − 1)kpx k=1 7
Lesson 6: Survival Distributions: Percentiles, Recursions, and Life Table Concepts
Recursive formulas
˚ex =˚ex:n +n px˚ex+n
˚ex:n =˚ex:m +m px˚ex+m:n−m, m < n
ex = ex:n +n pxex+n = ex:n−1 +n px(1 + ex+n)
ex = px + pxex+1 = px(1 + ex+1)
ex:n = ex:m +m pxex+m:n−m =
= ex:m−1 +m px(1 + ex+m:n−m) m < n ex:n = px + pxex+1:n−1 = px 1 + ex+1:n−1
Life Table Concepts
Z ∞ Tx = lx+tdt, total future lifetime of a cohort of lx individuals 0 Z n nLx = lx+tdt, total future lifetime of a cohort of lx individuals over the next n years 0 Z ∞ Yx = Tx+tdt 0 Tx ˚ex = lx nLx ˚ex:n = lx Central death rate and related concepts
ndx nmx = nLx qx mx = for uniform (de Moivre) distribution 1 − 0.5qx nmx = µx for constant force of mortality L − l a(x) = x x+1 the fraction of the year lived by those dying during the year dx 1 a(x) = for uniform (de Moivre) distribution 2 8
Lesson 7: Survival Distributions: Fractional Ages
Function Uniform Distribution of Deaths Constant Force of Mortality Hyperbolic Assumption
s lx+s lx − sdx lxpx lx+1/(px + sqx) s sqx sqx 1 − px sqx/(1 − (1 − s)qx) s spx 1 − sqx px px/(1 − (1 − s)qx) s sqx+t sqx/(1 − tqx), 0 ≤ s + t ≤ 1 1 − px sqx/(1 − (1 − s − t)qx)
µx+s qx/(1 − sqx) − ln px qx/(1 − (1 − s)qx) s 2 spxµx+s qx −px(ln px) pxqx/(1 − (1 − s)qx) 2 mx qx/(1 − 0.5qx) − ln px qx/(px ln px)
Lx lx − 0.5dx −dx/ ln px −lx+1 ln px/qx
˚ex ex + 0.5
˚ex:n ex:n + 0.5nqx
˚ex:1 px + 0.5qx 9
Lesson 8: Survival Distributions: Select Mortality When mortality depends on the initial age as well as duration, it is known as select mortality, since the person is selected at that age. Suppose qx is a non-select or aggregate mortality and q[x]+t, t = 0, ··· , n − 1 is select mortality with selection period n. Then for all t ≥ n, q[x]+t = qx+t. 10
Lesson 9: Insurance: Payable at Moment of Death - Moments - Part 1
Z ∞ −δt A¯x = e tpxµx(t) dt 0 Actuarial notation for standard types of insurance
Name Present value random variable Symbol for actuarial present value
T Whole life insurance v A¯x
vT T ≤ n Term life insurance A¯1 0 T > n x:n
0 T ≤ n Deferred life insurance |A¯ vT T > n n x
0 T ≤ n T ¯1 ¯ Deferred term insurance v n < T ≤ n + m n|Ax:m =n |mAx 0 T > n
0 T ≤ n Pure endowment A 1 or E vn T > n x:n n x
vT T ≤ n Endowment insurance A¯ vn T > n x:n 11
Lesson 10: Insurance: Payable at Moment of Death - Moments - Part 2
Actuarial present value under constant force and uniform (de Moivre) mortality for insurances payable at the moment of death
Type of insurance APV under constant force APV under uniform (de Moivre)
a¯ µ ω−x Whole life µ+δ ω−x
µ −n(µ+δ) a¯n n-year term µ+δ 1 − e ω−x
e−δna¯ µ −n(µ+δ) ω−(x+n) n-year deferred life µ+δ e ω−x
−n(µ+δ) e−δn(ω−(x+n)) n-year pure endowment e ω−x
j-th moment Multiply δ by j in each of the above formulae
Gamma Integrands Z ∞ n −δt n! t e dt = n+1 0 δ Z u u −δt 1 −δu a¯u − uv te dt = 2 1 − (1 + δu)e = 0 δ δ Variance
If Z3 = Z1 + Z2 and Z1,Z2 are mutually exclusive, then
V ar(Z3) = V ar(Z1) + V ar(Z2) − 2E[Z1]E[Z2] 12
Lesson 11: Insurance: Annual and m-thly: Moments
∞ ∞ X k+1 X k+1 Ax = k|qxv = kpxqx+kv 0 0 Actuarial present value under constant force and uniform (de Moivre) mortality for insurances payable at the end of the year of death
Type of insurance APV under constant force APV under uniform (de Moivre)
a q ω−x Whole life q+i ω−x
q n an n-year term q+i (1 − (vp) ) ω−x
vna q n ω−(x+n) n-year deferred life q+i (vp) ω−x
n vn(ω−(x+n)) n-year pure endowment (vp) ω−x 13
Lesson 12: Insurance: Probabilities and Percentiles
Summary of Probability and Percentile Concepts • To calculate P r(Z ≤ z) for continuous Z, draw a graph of Z as a function of T . Identify parts of the graph that are below the horizontal line Z = z, and the corresponding t’s. Then calculate the probability of T being in the range of those t’s. Note that ∗ ∗ ∗ ∗ t ∗ ln z ∗ ln z ln z P r(Z < z ) = P r(v < z ) = P r t > − = ∗ p , where t = − = − δ t x δ ln(1 + i) The following table shows the relationship between the type of insurance coverage and corre- sponding probability:
Type of insurance P r(Z < z∗)
Whole life t∗ px
∗ n npx z ≤ v n-year term ∗ n t∗ px z > v
∗ n q + ∗ p z < v n-year deferred life n x t x 1 z∗ ≥ vn
0 z∗ < vn n-year endowment ∗ n t∗ px z ≥ v
∗ m+n nqx +n+m px z < v m+n ∗ n n-year deferred m-year term nqx +t∗ px v ≤ z < v 1 z∗ ≥ vn
• In the case of constant force of mortality µ and interest δ, P r(Z ≤ z) = zµ/δ • For discrete Z, identify T and then identify K + 1 corresponding to those T . In other words, ∗ ∗ ∗ ∗ P r(Z > z ) = P r (T < t ) =k qx, where k = bt c - the greatest integer smaller than t ∗ ∗ ∗ ∗ P r(Z < z ) = P r (T > t ) =k+1 px, where k = bt c - the greatest integer smaller or equal to t • To calculate percentiles of continuous Z, draw a graph of Z as a function of T . Identify where the lower parts of the graph are, and how they vary as a function of T. For example, for whole life, higher T lead to lower Z. For n-year deferred whole life, both T < n and higher T lead to lower Z. Write an equation for the probability Z less than z in terms of mortality probabilities expressed in terms of t. Set it equal to the desired percentile, and solve for t or for ekt for any k. Then solve for z (which is often vt) 14
Lesson 13: Insurance: Recursive Formulas, Varying Insurances Recursive Formulas
Ax = vqx + vpxAx+1 2 2 Ax = vqx + v pxqx+1 + v2pxAx+2
Ax:n = vqx + vpxAx+1:n−1 1 1 Ax:n = vqx + vpxAx+1:n−1
n|Ax = vpxn−1|Ax+1 1 1 Ax:n = vpxAx+1:n−1 Increasing and Decreasing Insurance Z ∞ n −δt n t e dt = n+1 0 δ Z u u −δt 1 −δu a¯n δ − uv te dt = 2 1 − (1 + δu) e = 0 δ δ µ I¯A¯ = for constant force x (µ + δ)2 2µ E[Z2] = for Z a continuously increasing continuous insurance, constant force. (µ + 2δ)3 ¯ ¯1 ¯ ¯1 ¯1 IA x:n + DA x:n = nAx:n ¯1 ¯1 ¯1 IA x:n + DA x:n = (n + 1)Ax:n 1 1 1 (IA)x:n + (DA)x:n = (n + 1)Ax:n Recursive Formulas for Increasing and Decreasing Insurance 1 1 1 (IA)x:n = Ax:n + vpx (IA)x+1:n−1 1 1 1 (IA)x:n = Ax:n + vpx (IAA)x+1:n−1 1 1 1 (DA)x:n = nAx:n + vpx (DA)x+1:n−1 1 1 1 (DA)x:n = Ax:n + (DA)x:n−1 15
Lesson 14: Relationships between Insurance Payable at Moment of Death and Payable at End of Year Summary of formulas relating insurances payable at moment of death to insurances payable at the end of the year of death assuming uniform distribution of deaths
i A¯ = A x δ x i A¯1 = A1 x:n δ x:n i |A¯ = |A n x δ n x i IA¯1 = (IA)1 x:n δ x:n i ID¯1 = (ID)1 x:n δ x:n i A¯ = A1 + A 1 x:n δ x:n x:n i A(m) = A x i(m) x 2i + i2 2A¯ = 2A x 2δ x 1 1 I¯A¯1 = IA¯1 − A¯1 − x:n x:n x:n d δ Summary of formulas relating insurances payable at moment of death to insurances payable at the end of the year of death using claims acceleration approach
0.5 A¯x = (1 + i) Ax ¯1 0.5 1 Ax:n = (1 + i) Ax:n 0.5 n|A¯x = (1 + i) n|Ax ¯ 0.5 1 1 Ax:n = (1 + i) Ax:n + Ax:n (m) (m−1)/2m Ax = (1 + i) Ax 2 2 A¯x = (1 + i) Ax 16
Lesson 15: Annuities: Continuous, Expectation Actuarial notation for standard types of annuity
Name Payment per Present value Symbol for actuarial annum at time t random variable present value
Whole life 1 t ≤ T a¯T a¯x annuity
1 t ≤ min(T, n) a¯T T ≤ n Temporary life a¯x:n 0 t > min(T, n) a¯n T > n annuity
0 t ≤ n or t > T 0 T ≤ n Deferred life annuity n|a¯x 1 n < t ≤ T a¯T − a¯n T > n
0 t ≤ n or t > T 0 T ≤ n Deferred temporary 1 n < t ≤ n + m and t ≤ T a¯T − a¯n n < T ≤ n + m n|a¯x:m 0 T > n + m a¯n+m − a¯n T > n + m life annuity
1 t ≤ max(T, n) a¯n T ≤ n Certain-and-life a¯x:n 0 t > max(T, n) a¯T T > n
Relationships between insurances and annuities 1 − A¯ a¯ = x x δ A¯x = 1 − δa¯x 1 − A¯ a¯ = x:n x:n δ A¯x:n = 1 − δa¯x:n A¯ − A¯ |a¯ = x:n x n x δ General formulas for expected value
Z ∞ a¯x = antpxµx+t dt 0 Z ∞ t a¯x = v tpx dt 0 Z n t a¯x:n = v tpx dt 0 Z ∞ t n|a¯x = v tpx dt n
Formulas under constant force of mortality 17
1 a¯ = x µ + δ 1 − e−(µ+δ)n a¯ = x:n µ + δ e−(µ+δ)n |a¯ = n x µ + δ
Relationships between annuities
a¯x =a ¯x:n +n Exa¯x+n
a¯x:n =a ¯n +n |a¯x 18
Lesson 16: Annuities: Discrete, Expectation Relationships between insurances and annuities 1 − A a¨ = x x d Ax = 1 − da¨x 1 − A a¨ = x:n x:n d Ax:n = 1 − da¨x:n 1 1 1 Ax:n = va¨x:n − ax:n (1 + i)Ax + iax = 1 Relationships between annuities
a¨x:n =a ¨n +n |a¨x a¨x:n =a ¨x −n Exa¨x+n
n|a¨x =n Exa¨x+n
a¨x =a ¨x:n +n |a¨x
ax =a ¨x − 1
a¨x:n = ax:n + 1 −n Ex = ax:n−1 + 1 Other annuity equations n−1 X a¨x:n = a¨k k−1pxqx+k−1 +a ¨nn−1px k=1 n−1 n X k X k a¨x:n = v kpx, ax:n = v kpx k=0 k=1 1 + i a¨ = , if q is constant x q + i x Accumulated value a¨x:n s¨x:n = nEx sx:n =s ¨x+1:n−1 + 1 1 sx:n = sx+1:n−1 + n−1Ex+1 1 sx:n =s ¨x:n + 1 − nEx mthly annuities ∞ k (m) X 1 m a¨x = v k px m m k=0 19
Lesson 17: Annuities: Variance General formulas for second moments Z ∞ ¯ 2 2 E[Yx ] = a¯t tpxµx+t dt 0 ∞ ¨ 2 X 2 E[Yx ] = a¨k k−1|qx k=1 n n−1 ¨ 2 X 2 2 X 2 2 E[Yx:n] = a¨k k−1|qx + npxa¨n = a¨k k−1|qx + n−1pxa¨n k=1 k=1 Special formulas for variance of whole life annuities and temporary life annuities 2A¯ − (A¯ )2 2(¯a −2 a¯ ) V ar(Y¯ ) = x x = x x − (¯a )2 x δ2 δ x 2A¯ − (A¯ )2 2(¯a −2 a¯ ) V ar(Y¯ ) = x:n x:n = x:n x:n − (¯a )2 x:n δ2 δ x:n 2A − (A )2 2(¨a −2 a¨ ) V ar(Y¨ ) = V ar(Y ) = x x = x x +2 a¨ − (¨a )2 x x d2 d2 x x 2A − (A )2 V ar(Y¨ ) = V ar(Y ) = x:n x:n x:n x:n−1 d2 20
Lesson 18: Annuities: Probabilities and Percentiles
• To calculate a probability for an annuity, calculate the t for whicha ¯t has the desired property. Then calculate the probability t is in that range. • To calculate a percentile of an annuity, calculate the percentile of T , then calculatea ¯T . • Some adjustments may be needed for discrete annuities or non-whole-life annuities • If forces of mortality and interest are constant, then the probability that the present value of payments on a continuous whole life annuity will be greater than its actuarial present value is µ µ/δ P r(¯a > a¯ ) = T (x) x µ + δ 21
Lesson 19: Annuities: Varying Annuities, Recursive Calculations Increasing/Decreasing Annuities 1 I¯a¯ = , if µ is constant x (µ + δ)2 ¯ ¯ Ia¯ x:n + Da¯ x:n = na¯x:n
(Ia)x:n + (Da)x:n = (n + 1)ax:n Recursive Formulas
a¨x = vpxa¨x+1 + 1
ax = vpxax+1 + vpx
a¯x = vpxa¯x+1 +a ¯x:1
a¨x:n = vpxa¨x+1:n−1 + 1
ax:n = vpxax+1:n−1 + vpx
a¯x:n = vpxa¯x+1:n−1 +a ¯x:1
n|a¨x = vpxn−1a¨x+1
n|ax = vpxn−1ax+1
n|a¯x = vpxn−1a¯x+1 a¨ = 1 + vq a¨ + vp a¨ x:n x n−1 x x+1:n−1 a = v + vq a + vp a x:n x n−1 x x+1:n−1 a¯ =a ¯ + vq a¯ + vp a¯ x:n 1 x n−1 x x+1:n−1 22
Lesson 20: Annuities: m-thly Payments In general:
1 a(m) =a ¨(m) − x x m !m !−m i(m) d(m) 1 + i = 1 + = 1 − m m
(m) 1 i = m (1 + i) m − 1
(m) − 1 d = m 1 − (1 + i) m
For small interest rates:
m − 1 a¨(m) ≈ a¨ − x x 2m m − 1 a(m) ≈ a + x x 2m
Under the uniform distribution of death (UDD) assumption:
m − 1 a¨(m) =a ¨ − x x 2m m − 1 a(m) = a + x x 2m (m) a¨x = α(m)¨ax − β(m) (m) a¨x:n = α(m)¨ax:n − β(m)(1 −n Ex) (m) n|a¨x = α(m)n|a¨x − β(m)nEx a¯x = α(∞)¨ax − β(∞) 1 1 a(m) =a ¨(m) − + E x:n x:n m m n x 1 − A(m) a¨(m) = x x d(m)
Similar conversion formulae for converting the modal insurances to annuities hold for other types of insurances and annuities (only whole life version is shown).
id α(m) = i(m)d(m) i − i(m) β(m) = i(m)d(m) i(∞) = d(∞) = ln(1 + i) = δ 23
(m) Woolhouse’s formula for approximatinga ¨x : m − 1 m2 − 1 a¨(m) ≈ a¨ − − (µ + δ) x x 2m 12m2 x 1 1 a¯ ≈ a¨ − − (µ + δ) x x 2 12 x m − 1 m2 − 1 a¨(m) ≈ a¨ − (1 − E ) − (µ + δ − E (µ + δ)) x:n x:n 2m n x 12m2 x n x x+n m − 1 m2 − 1 |a¨(m) ≈ |a¨ − E − E (µ + δ) n x n x 2m n x 12m2 n x x+n 1 1 ˚e = e + − µ x x 2 12 x When the exact value of µx is not available, use the following approximation: 1 µ ≈ − (ln p + ln p ) x 2 x−1 x 24
Lesson 21: Premiums: Fully Continuous Expectation The equivalence principle: The actuarial present value of the benefit premiums is equal to the actuarial present value of the benefits. For instance: