AS - Actuarial Science Summary

Can you compile a summary of all the lectures

Lecture AS1: Introduction to Actuarial Science & Insurance Fundamentals

• Actuarial Science Defined:
– Actuarial science focuses on quantifying and managing risks using probability and
statistics.
– Actuaries advise insurers, financial institutions, and government entities on risk
assessment and decision-making.
• Core Concepts & Notation:
– Random variables (e.g., X, X1 , X2 , …) represent risks; insurers charge a nonrandom
​ ​

premium P .
– Wealth W (random) and utility u(w) capture decision‐makers’ preferences.
• Risk Transfer via Insurance:
– Insured pays premium P ; if loss X occurs, insurer pays X .
– Law of Large Numbers underpins pooling of independent risks, justifying why P ≈ E[X]
asymptotically.
• Expected Utility & Premium Bounds:
– Insured’s Reservation Price P + : solve

E[u(w0 − X)] = u(w0 − P + ) .

​ ​

– By Jensen’s inequality (risk aversion u′′ < 0), P + ≥ E[X].

– Insurer’s Minimum Premium P − : solve

U (W0 ) = E[ U (W0 + P − − X)],

​ ​

which similarly yields P −

≥ E[X].
−
– Any contract with P ≤ P ≤ P + increases expected utility of both parties.
• Approximate Formula for P + (small‐risk approximation):
If μ = E[X], σ 2 = Var(X), and A(w) = − u′′ (w)/u′ (w) is the absolute risk aversion, then

P+ ≈ μ + 1
2
​
σ 2 A(w0 − μ) .

• Arrow–Pratt Coefficient & HARA Utilities:

– Exponential utility u(w) = −e−αw gives constant absolute risk aversion (CARA): A(w) = α.
– Power utility u(w) = w γ (with 0 < γ < 1) gives decreasing absolute risk aversion (DARA).

1/10
Lecture AS2: Advanced Expected Utility & Reinsurance
• Recap of AS1: Insurance’s role, expected utility approach to premium determination, risk
aversion.
• Arrow–Pratt Absolute Risk Aversion:
′′
u (w)
– Definition: A(w) = − u′ (w) . ​

– Exponential: A(w)= α (constant); Power: A(w) = (1 − γ)/w (declining in w).

– HARA (Hyperbolic Absolute Risk Aversion) class: A(w) = γ + β/w .
• Refining Premium Approximations:
– Exponential premium for risk X :
1
πα (X) =
​

αlog(E[eαX ]),
​

1
which, for small α, expands to E[X] + 2
​
α Var(X) + O(α2 ).
• Introduction to Reinsurance:
– Definitions:
Cedent (primary insurer) cedes part of its risk to a reinsurer.
Total risk X splits into Xc
+ Xr , with Xr = I(X) the portion ceded.
​ ​ ​

Premium likewise splits into Pc + Pr . ​ ​

– Motivation:
1. Protect against catastrophic losses, smoothing volatility.
2. Allow insurer to underwrite larger limits than its balance sheet alone would permit.
– Proportional Contracts: cedent retains a fixed proportion a of every loss (I(X)
= aX ).
– Non‐proportional Contracts (Stop‐Loss): reinsurer covers (X − d)+ = max(0, X − d).
Stop‐loss net premium for retention d:
∞
π(d) = E[ (X − d)+ ] = ∫d [1 − FX (x)] dx, ​ ​

which is convex and strictly decreasing in d.

– Illustration: If X takes discrete values, π(d) = ∑x≥d (x − d) P (X = x). ​

Lecture AS3: Introduction to Non‐Life Insurance & Premium Principles

N
• Review of AS1–AS2 Basics: Total insurer risk can be modeled as ∑i Xi or ∑j=1 Yj . For multi‐ ​ ​ ​ ​

period solvency, premiums π(X) must exceed E[X].

• Non‐Life Insurance Characteristics:
– Insurance period typically one year; insurer indemnifies all losses arising in that year.
– Examples: motor, property, liability, accident & health.
N
– Claim amount per insured: S = ∑k=1 Yk (compound distribution), with N = claim count, Yk
​ ​ ​

2/10
= claim size.
• Ruin Theory (Brief):
– Insurer’s surplus at time t: initial capital U0 plus collected premiums minus cumulative paid ​

claims.
– To avoid ruin (surplus < 0 w.p.1), require premium > E[annual loss].
• Premium Principles (multi‐period context):
– Premium = net premium + safety loading A(X).
– Common Loadings:
1. Expected‐Value Principle: π(X) = (1 + θ) E[X].
2. Variance Principle: π(X) = E[X] + α Var(X).
3. Standard‐Deviation Principle: π(X) = E[X] + α Var(X). ​

4. Exponential Principle: π(X) = α1 log(E[eαX ]). ​

– Desired Properties of π : Nonnegative loading (π(X) ≥ E[X]), no “rip‐off” (π(X) ≤ sup X ),

cash invariance (π(X + c) = π(X) + c), additivity on independent risks (for some principles),
iterativity (premium‐of‐premium property).
– Unique principle satisfying all preferred properties: the exponential principle.
• Introduction to Bonus‐Malus (experience‐based rating):
– Used primarily in automobile insurance: policyholder moves up/down a scale (“step”) based
on prior claims.
– Scale Example: 14 levels; each level corresponds to a % loading/discount on base premium
(e.g., Level 1 = 120%, Level 14 = 30%).
– Transition from state i to j depends on number of claims Nt in year t. If no claims, ​

policyholder moves down (bonus); if claims occur, moves up (malus).

– Markov Chain Model:
States: St ∈ {1, … , n}.
​

Transition matrix P = (pij ), where ​

∞
pij = P (St+1 = j St = i) = ∑k=0 P (Nt = k St = i) 1{step(i,k)=j} ,
​ ​ ​ ​ ​ ​ ​ ​ ​

computed from the claim count probabilities P (Nt = k ∣ St = i) = pk (i). ​ ​ ​

– Premium in Year t: If ℓt = (ℓ1 (t), … , ℓn (t)) is the distribution over steps at time t, and b =
​ ​ ​

(b1 , … , bn ) are the level‐specific multipliers of base premium, then annual expected %
​ ​

loading is
bt = ℓt b,
​ ​ bt = ℓ0 P t b.
ℓt+1 = ℓt P ,
​ ​ ​ ​

– Long‐Run (Steady State): Solve ℓ∞ = ℓ∞ P , ∑ ℓ∞,i = 1; then steady‐state loading b∞ =

​ ​ ​ ​

ℓ∞ b. ​

3/10
Lecture AS4: Bonus‐Malus Deep Dive & Efficiency
• Review of AS3 Bonus‐Malus Model: Markov chain (St ), transition matrix P , step‐dependent ​

claim counts pk (i). Expected premium in year t: bt​ ​ = ℓt b.​

• Premium Calculation Examples:

– Simple Two‐State Example: States 1&2 with b = (110%, 90%). If
1 1
2 2
P = ( ), ℓ0 = (1, 0),
​ ​

1 3
​ ​ ​

4 4
​ ​

then
ℓ1 = (1, 0) P = ( 12 , 12 ),
​ ​ ​ ℓ2 = ℓ1 P = ( 38 , 58 ),
​ ​ ​ ​

so
b0 = 1 ⋅ 110% + 0 ⋅ 90% = 110%, b1 = ( 12 , 12 ) (110%, 90%) = 100%, b2 =
​ ​ ​ ​ ​

( 38 , 58 ) (110%, 90%) = 97.5%.

​ ​

– Long‐Run: Solve ℓ∞ P = ℓ∞ , ℓ∞,1 + ℓ∞,2 = 1. Here ℓ∞ = ( 13 , 23 ), so steady‐state loading

​ ​ ​ ​ ​ ​ ​

b∞ = 13 110% + 23 90% = 96.67%.

​ ​ ​

• Loimaranta Efficiency (measures how well bonus‐malus differentiates risk levels):

– Let a driver’s steady‐state claim frequency be λ (Average number of claims per year). Under
a Poisson assumption N ∼ Poisson(λ), the transition probabilities pij (λ) can be calculated ​

explicitly for each step. Solve for ℓ∞ (λ) and then compute b(λ) = ℓ∞ (λ) b. ​ ​

– Compare the bonus‐malus premium b(λ) to the “ideal” actuarial loading c λ E[Y ]. A system
is “perfectly efficient” if b(λ) is proportional to λ.
– Loimaranta Efficiency Index:
b′ (λ) d(log b(λ))
e(λ) = λ b(λ) = d(log λ) .
​ ​

If b(λ) ∝ λ, then e(λ) ≡ 1. In practice, e(λ) < 1, and efficiency is assessed by how close
e(λ) stays to 1.
– Under Poisson claim counts and a given steps‐vs‐claims rule, one can derive closed‐form
expressions for ℓ∞ (λ) and hence for b(λ), then evaluate e(λ). ​

Lecture AS5: Life Insurance Basics, Annuities & Pension Systems

• Life vs. Non‐Life Insurance:
– Non‐life (property/casualty) covers events within a fixed policy year.
– Life insurance deals with uncertain human lifetimes; payoffs often occur at death or
annuitization.
– Historical note: “losrenten” (government bonds paying fixed 1 gulden/year at price 25, circa
1670) vs. “lijfrenten” (life annuities paying 1 gulden/year as long as annuitant lives, sold at

4/10
price 14 guilders); de Witt argued the fair price should exceed 16 guilders once survival
probabilities were accounted for.
• Life Annuity Types & Notation:
– Survival probabilities: t px= P (age x survives t years).
​ ​

– Flat yield curve: r constant, so discount factor for t years is (1 + r)−t .

(T ) ∞
– Deferred Annuity: ax = ∑t=T t px (1 + r)−t .​ ​ ​ ​

(1) ∞
– Immediate Annuity (payment begins at age x + 1): ax = ax = ∑t=1 t px (1 + r)−t .
​ ​ ​ ​ ​

∞
– If r = 0, ax = ∑t=1 t px , which equals the curtate expectation of future lifetime (denoted ex
​ ​ ​ ​ ​

).
– Pricing Examples (flat 3%):
* Age 25, immediate annuity: a25 ≈ 26.285. ​

(35)
* Age 25, 35‐year deferred annuity: a25 ≈ 5.424. ​

• Pensions & Pension Funds:

– Definition: Pension is an income stream after retirement (typically until death). Payments
arise from accumulated contributions plus investment returns.
– Three Pillars:
1. First Pillar (State Pension, AOW in NL): Pay‐As‐You‐Go (PAYG), funded by current payroll
taxes; everyone with 50 years of residency is entitled at retirement age. Avoids old‐age
poverty.
2. Second Pillar (Occupational Pension): Employer‐sponsored plans:
- Defined Benefit (DB): Retirement benefit is (approximately) predetermined, e.g., a fraction
f of final or career‐average salary; contributions must be set so that accumulated assets can
support those promised benefits.
- Defined Contribution (DC): Contributions are fixed; ultimate benefit depends on
investment performance. Shifts investment/longevity risk to participant.
3. Third Pillar (Private Pension): Individual, voluntary, often tax‐favored retirement savings
(e.g., IRAs).
– State Pension Parameters:
AOW Retirement Age Formula (from 2026 onward):
2
V = 3 ​ (L − 20.64) − (R − 67),
if V ≥ 0.25, R ← R + 0.25 year.
Here L = projected remaining life expectancy at age 65; R = prior year’s retirement age;
increment by 3 months if V ≥ 0.25.
– E.g., projected L2026 = 20.82 → increment zero; subsequent years update.
​

• Domestic Pension Statistics (March 2024, DNB):

– Total pension assets end 2024: €1.913 trillion.
– Dutch pension assets by sector: civil servants > €500 billion; health > €250 billion.

5/10
– Demographic trends (CBS projections): remaining life expectancy at 65 increasing from
about 20.96 in 2030 to beyond 21.5 by 2035.
– Dependency ratios rising, pressuring first‐pillar costs (AOW ~5% GDP).
• Stylized DB Pension Example:
– 40‐year career (age 25–64), salary y . Aim for replacement rate f = 0.8 of y from ages 65–
84. Uniform accrual → per‐year accrual factor f /w = 0.8/40 = 0.02.
(65−x) (T )
– At age x, contribution per unit salary = 0.02 ax ​, where ax is value of T -year deferred
​

(40)
life annuity. E.g., with r = 2%: p25 = 0.02 a25 ≈ 0.151.
​ ​

– If population per age is uniform, a uniform contribution p is set so that

∑64 (65−x)
x=25 0.02 ax
p = ,
​ ​

40
​

independent of individual age. Plotted contribution vs. age shows increasing px for older
​

cohorts.
– Assets & Technical Reserves: Contributions accumulate in an asset pool; technical reserve
for each cohort = promised pension PV. The funding ratio = total assets / total technical
reserves. If <1, future contributions or benefits must adjust.
– Investment Strategies:
1. Replication: buy zero‐coupon bonds for every future pension payment; requires bonds to
exist with maturities up to, say, 59 years.
2. Rebalancing: when only 40‐year zero‐coupon bonds exist, invest each cohort’s
contribution in the bond with maturity 65 − x. At age 65, lump‐sum payoff partially funds
age‐65 pensions and rest is reinvested in bonds maturing at 66–84.
(1)
3. Interest Rate Risk: If interest rates drop from 2% to, say, 1%, then a65 rises from 13.34 to
​

14.58, so available funds to rebalance are insufficient → inability to meet promised benefits
(funding ratio falls <1).
4. Inflation & Longevity Risk:
– If funding ratio >1, funds may be used to index benefits. Without indexation, real benefits
decline.
– Unanticipated improvements in life expectancy (e.g., OECD trends) increase liabilities;
pension plan must hold extra assets or raise contributions.
• Defined Contribution (DC) Illustration:
– Instead of promising a fixed pension f y , contributions are invested, and accumulated
account at retirement determines annuity. Participants bear all investment and longevity
risk. Shifts risk from plan sponsor to individual.
• Pension Reform Discussion (Bas Werker, ISL):
– Emphasizes that individuals dislike uncertainty; guaranteeing benefits over long horizons is
impossible without fully flexible contributions.

6/10
– Proposed shift from DB to DC: stop promising fixed pensions; use individual accounts
where accumulated funds drive benefits.

Lecture AS6: Advanced Pension Topics—DB Dynamics, Investment Risk & DC

• Review of DB Stylized Example:
– Recap of uniform contribution p over ages 25–64; contribution per y per cohort is p y .
– Cohort x in DB → promised benefit 0.8 y from ages 65–84; technical reserve at time 65 =
(1)
0.8 a65 y . ​

– Asset accumulation and intergenerational flows illustrated via cohort diagrams:

contributions, past contributions, entitlements, reserves.
• Funding Ratios & Historical Shocks:
Assets
– Funding ratio FR = Technical Reserves
​ . If FR<1, benefits must be cut or contributions
increased.
– DNB data: post‐2007–09 financial crisis, equities plunged and interest rates fell → DRs<1;
during COVID‐19, similar impacts.
• Investment Strategies (Continued):

1. Replication vs. Rebalancing revisited with formulas:

Replication requires zero‐coupon bonds for all maturities; track each cohort’s
(65−x)
required ax .
​

Rebalancing invests at each age x in a bond of maturity 65 − x; at age 65 pivot to

bonds maturing 66–84.

2. Interest‐Rate Risk:
(1)
– If pre‐rebalancing rate r0 = 2% yields a65 = 13.34, but post‐shock rate r1 = 1%
​ ​ ​

(1) (1)
yields a65 ​ = 14.58 (+9%), funds [0.8(1 + a65 )] at the new rates cannot purchase
​

required bonds to cover 66–84 benefits → underfunding (FR <1).

– Consequence: younger cohorts only partially affected (built‐up part), older ones
unaffected (already retired).
– Overall FR: at 2% initially FR =1; at 1% FR =0.95; at 0% FR =0.89.
• Inflation & Longevity Risks:
– Inflation Risk: Euro Area HICP inflation (CBS): if FR>1, indexation possible; else, real
benefits drop.
– Longevity Risk: Life expectancy frontier (Oeppen & Vaupel, 2002) rising at ~0.25
years/year → plan liabilities increase unexpectedly.

7/10
 – Investment Policy: To outpace inflation/longevity, plans allocate to equities, bonds;
failure to do so leaves benefits under‐indexed.
• DC Illustration (Revisited):
– Under DC, each cohort’s contributions become individual account balances. At
retirement, participant buys a variable annuity or systematic drawdown.
– Stylized diagrams show contributions, invested assets, and benefit flows for DC vs. DB
cohorts.
• Pension Reform (Bas Werker):
– Traditional DB impossible to guarantee long‐term benefit under uncertain
markets/longevity.
– Shift to DC: no promises, contributions invested; final account drives benefit. Removes
promise‐risk misalignment.

Overall Synthesis of AS1–AS6

1. Foundations of Risk & Insurance (AS1–AS2):
– Actuarial science formalizes risk analysis via random variables and expected utility.
– Insurance transfers risk X from insured (pays premium P ) to insurer; both parties
benefit if P lies between their reservation bounds [P − , P + ].
– Risk aversion (concave utility) ensures P + , P − ≥ E[X]. Approximate formula uses
Arrow–Pratt coefficient. Exponential utility yields the “exponential premium”
1
α
​ log E[eαX ].
– Reinsurance (proportional, stop‐loss) mitigates insurer’s tail risk; stop‐loss premium
π(d) = E[(X − d)+ ] is fundamental.
2. Non‐Life Insurance & Rating (AS3–AS4):
– Non‐life insurers pool many independent risks; to avoid multi‐period ruin, premiums
must exceed expected losses by a loading. Common loading principles: expected‐value,
variance, standard‐deviation, exponential. Exponential principle uniquely satisfies
preferred axioms.
– In motor insurance, Bonus‐Malus systems provide experience‐based rating:
policyholders move along a discount/loading scale based on claim history.
Mathematically a finite‐state Markov chain; premiums evolve via ℓt P b.
​

– Steady‐state distribution ℓ∞ solves ℓ∞ P

​ ​ = ℓ∞ . Long‐run loading b∞ = ℓ∞ b.
​ ​ ​

– Loimaranta Efficiency measures how well b(λ) tracks true claim frequency λ. Defined
as λ b′ (λ)/b(λ); a perfectly discriminating system has index 1.

8/10
 3. Life Insurance & Pensions (AS5–AS6):
(T )
– Life annuities priced via survival probabilities t px and discount rates rt : ax
​ ​ ​ ​ =
∑∞ ​ ​
−t
t=T t px (1 + r) . Special case r = 0: ax = ex (expected future life years).
​ ​ ​

– Pension funds operate on DB or DC paradigms:

• DB: Prescribe promised benefit (e.g., fraction f of salary) for each cohort; contributions
calibrated via life‐annuities; technical reserves track present value of future liabilities.
Contributions typically uniform across ages (actuarially smoothed).
• Funding Ratio: Assets/technical reserves; shocks in interest rates, equities, mortality, or
inflation can push FR<1, forcing benefit cuts or higher contributions.
• Investment Strategies: Replication (perfect match via zero‐coupon bonds) vs.
Rebalancing (practical approach with limited maturities) illustrate interest‐rate risk.
• Inflation & Longevity Risks: New risks require investments in equities or inflation‐
linked bonds. Longevity improvements push liabilities upward.
• DC: Contributions go into individual accounts; final account value drives annuity
purchase or drawdown. Implements “no‐promise” philosophy, exposing individuals to
market/longevity risk.
– Pension System Structure in NL: Three pillars—AOW (state PAYG), occupational
(funded DB/DC), and private (individual, tax‐favored). Rising life expectancy and shifting
demographics have led to retirement age formula linking life expectancy to statutory
retirement age (increase by 3 months when 23 (L − 20.64) − (R − 67)
​ ≥ 0.25).
– Pension Reform: Recognizing that long‐term DB guarantees conflict with market
uncertainty, many systems shift toward DC or mixed designs, passing
investment/longevity risk to participants.

Key Takeaways Across All Lectures

Risk aversion and expected utility underpin why insurance exists and how premiums
must exceed expected losses.

Reinsurance (especially stop‐loss) lets insurers manage tail exposures.

Premium principles (ES, variance, exponential) formalize loading for solvency.

Bonus‐Malus systems in non‐life insurance use Markov‐chain models to adjust

premiums based on claim experience; Loimaranta efficiency gauges how well such
systems differentiate risk.

9/10
 Life annuity valuation requires mortality (survival probabilities) and discount factors;
annuity values drive DB pension contributions and reserves.

Pension funds’ challenges: funding ratios sensitive to interest rates, equity returns,
inflation, and mortality; investment strategy critical to meet indexed benefits.

Shift from DB toward DC reflects the difficulty of guaranteeing long‐term benefits in

uncertain financial/mortality environments.

10/10