on the optimal design of participating life insurance ... · common pitfalls, the journal of risk...

52
UNIVERSITÀ DEGLI STUDI DI TRIESTE On the optimal design of participating life insurance contracts Anna Rita Bacinello, Chiara Corsato and Pietro Millossovich Department of Economics, Business, Mathematics and Statistics University of Trieste (Italy) European Actuarial Journal Conference Leuven September 10, 2018

Upload: others

Post on 11-Aug-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

UNIVERSITÀDEGLI STUDI DI TRIESTE

On the optimal designof participating life insurance contracts

Anna Rita Bacinello, Chiara Corsato and Pietro Millossovich

Department of Economics, Business, Mathematics and StatisticsUniversity of Trieste (Italy)

European Actuarial Journal ConferenceLeuven

September 10, 2018

Page 2: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Outline

Motivations

● Optimal insurance, an overview

A constrained maximisation problem

● Setting of the problem

● Mathematical analysis

Study of the fairness condition

An overall picture of the solutions

● Numerical results

State-of-the-art and future work

Page 3: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

What do we do?

Framework

Policyholders and shareholders form a life insurance company, which issuesparticipating contracts.

Aim

● To maximise analytically the policyholders’ preferences.

● To consider fair valuation and solvency constraints.

● To study the life insurance demand.

● To investigate the problem numerically.

Warning: the results are very preliminary!

Page 4: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

References - Starting pointNon life insurance, origins of the optimal insurance problem

P J. Mossin, Aspects of Rational Insurance Purchaising, Journal of Political Economy76, 533-568 (1968).

P K.J. Arrow, Essays in the Theory of Risk Bearing, Markham Publishing Co., 1971.

P A. Raviv, The Design of an Optimal Insurance Policy, American Economic Review,69, 84–96 (1979).

Life insurance, literature of reference

P E. Briys, F. de Varenne, Life Insurance in a Contingent Claim Framework: Pricingand Regulatory Implications, The Geneva Papers on Risk and Insurance Theory,19, 53–72 (1994).

P E. Briys, F. de Varenne, On the Risk of Insurance Liabilities: Debunking SomeCommon Pitfalls, The Journal of Risk and Insurance, 64(4), 673–694 (1997).

P H. Schmeiser, J. Wagner, A Proposal on How the Regulator Should Set MinimumInterest Rate Guarantees in Participating Life Insurance Contracts, The Journal ofRisk and Insurance, 82(3), 659–686 (2015).

P A. Chen, P. Hieber, Optimal Asset Allocation in Life Insurance: the Impact ofRegulation, Astin Bulletin 46(3), 605–626 (2016).

Page 5: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Contract featuresAt time t = 0,

Assets LiabilitiesW0 L0 = αW0

E0 = (1 − α)W0

W0 W0

α ∈ [0,1]: leverage ratio (policyholders’ percentage initial contribution)

At time t = T ,

Assets LiabilitiesWT LT = αW0G´¹¹¹¹¹¹¹¸¹¹¹¹¹¹¶

guaranteed

+ δα(WT −W0G)+´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶

bonus

− (αW0G −WT )+´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶

default

ET =WT − LTWT WT

G ∈ [0,+∞[: guaranteed amount per euro insuredδ ∈ [0,1]: participation rate

The model follows Briys, de Varenne (1994, 1997): considered default risk, nomortality risk.

Page 6: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Contract featuresAt time t = 0,

Assets LiabilitiesW0 L0 = αW0

E0 = (1 − α)W0

W0 W0

α ∈ [0,1]: leverage ratio (policyholders’ percentage initial contribution)

At time t = T ,

Assets LiabilitiesWT LT = αW0G´¹¹¹¹¹¹¹¸¹¹¹¹¹¹¶

guaranteed

+ δα(WT −W0G)+´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶

bonus

− (αW0G −WT )+´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶

default

ET =WT − LTWT WT

G ∈ [0,+∞[: guaranteed amount per euro insuredδ ∈ [0,1]: participation rate

The model follows Briys, de Varenne (1994, 1997): considered default risk, nomortality risk.

Page 7: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Contract featuresAt time t = 0,

Assets LiabilitiesW0 L0 = αW0

E0 = (1 − α)W0

W0 W0

α ∈ [0,1]: leverage ratio (policyholders’ percentage initial contribution)

At time t = T ,

Assets LiabilitiesWT LT = αW0G´¹¹¹¹¹¹¹¸¹¹¹¹¹¹¶

guaranteed

+ δα(WT −W0G)+´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶

bonus

− (αW0G −WT )+´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶

default

ET =WT − LTWT WT

G ∈ [0,+∞[: guaranteed amount per euro insuredδ ∈ [0,1]: participation rate

The model follows Briys, de Varenne (1994, 1997): considered default risk, nomortality risk.

Page 8: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Policyholders’ final payoff profile

LT =

⎧⎪⎪⎪⎪⎨⎪⎪⎪⎪⎩

WT if WT < αW0G ,

αW0G if αW0G ≤WT <W0G ,

αW0G + δα(WT −W0G) if W0G ≤WT .

Page 9: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

AssumptionsP historical measure.

Capital market conditions

● Deterministic risk-free interest rate r .

● Rate of assets return R normally distributed, with (under P)

µ: expected assets return per year,σ: volatility of assets return per year.

R ∼P N ((µ − σ2

2)T , σ2T) .

● µ > r , or, equivalently, risk premium λ = µ−rσ

> 0.

● There exists a unique risk-neutral measure Q equivalent to P⇒ Complete and arbitrage-free market.

R ∼Q N ((r − σ2

2)T , σ2T) .

● Black-Scholes framework.

Policyholders’ preferences

u ∶ R→ R utility function: u twice differentiable, with u′ > 0, u′′ < 0 in R (homogeneouscohort of risk-averse policyholders).

Page 10: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

AssumptionsP historical measure.

Capital market conditions

● Deterministic risk-free interest rate r .

● Rate of assets return R normally distributed, with (under P)

µ: expected assets return per year,σ: volatility of assets return per year.

R ∼P N ((µ − σ2

2)T , σ2T) .

● µ > r , or, equivalently, risk premium λ = µ−rσ

> 0.

● There exists a unique risk-neutral measure Q equivalent to P⇒ Complete and arbitrage-free market.

R ∼Q N ((r − σ2

2)T , σ2T) .

● Black-Scholes framework.

Policyholders’ preferences

u ∶ R→ R utility function: u twice differentiable, with u′ > 0, u′′ < 0 in R (homogeneouscohort of risk-averse policyholders).

Page 11: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

AssumptionsP historical measure.

Capital market conditions

● Deterministic risk-free interest rate r .

● Rate of assets return R normally distributed, with (under P)

µ: expected assets return per year,σ: volatility of assets return per year.

R ∼P N ((µ − σ2

2)T , σ2T) .

● µ > r , or, equivalently, risk premium λ = µ−rσ

> 0.

● There exists a unique risk-neutral measure Q equivalent to P⇒ Complete and arbitrage-free market.

R ∼Q N ((r − σ2

2)T , σ2T) .

● Black-Scholes framework.

Policyholders’ preferences

u ∶ R→ R utility function: u twice differentiable, with u′ > 0, u′′ < 0 in R (homogeneouscohort of risk-averse policyholders).

Page 12: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

AssumptionsP historical measure.

Capital market conditions

● Deterministic risk-free interest rate r .

● Rate of assets return R normally distributed, with (under P)

µ: expected assets return per year,σ: volatility of assets return per year.

R ∼P N ((µ − σ2

2)T , σ2T) .

● µ > r , or, equivalently, risk premium λ = µ−rσ

> 0.

● There exists a unique risk-neutral measure Q equivalent to P⇒ Complete and arbitrage-free market.

R ∼Q N ((r − σ2

2)T , σ2T) .

● Black-Scholes framework.

Policyholders’ preferences

u ∶ R→ R utility function: u twice differentiable, with u′ > 0, u′′ < 0 in R (homogeneouscohort of risk-averse policyholders).

Page 13: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

AssumptionsP historical measure.

Capital market conditions

● Deterministic risk-free interest rate r .

● Rate of assets return R normally distributed, with (under P)

µ: expected assets return per year,σ: volatility of assets return per year.

R ∼P N ((µ − σ2

2)T , σ2T) .

● µ > r , or, equivalently, risk premium λ = µ−rσ

> 0.

● There exists a unique risk-neutral measure Q equivalent to P⇒ Complete and arbitrage-free market.

R ∼Q N ((r − σ2

2)T , σ2T) .

● Black-Scholes framework.

Policyholders’ preferences

u ∶ R→ R utility function: u twice differentiable, with u′ > 0, u′′ < 0 in R (homogeneouscohort of risk-averse policyholders).

Page 14: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

AssumptionsP historical measure.

Capital market conditions

● Deterministic risk-free interest rate r .

● Rate of assets return R normally distributed, with (under P)

µ: expected assets return per year,σ: volatility of assets return per year.

R ∼P N ((µ − σ2

2)T , σ2T) .

● µ > r , or, equivalently, risk premium λ = µ−rσ

> 0.

● There exists a unique risk-neutral measure Q equivalent to P⇒ Complete and arbitrage-free market.

R ∼Q N ((r − σ2

2)T , σ2T) .

● Black-Scholes framework.

Policyholders’ preferences

u ∶ R→ R utility function: u twice differentiable, with u′ > 0, u′′ < 0 in R (homogeneouscohort of risk-averse policyholders).

Page 15: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

AssumptionsP historical measure.

Capital market conditions

● Deterministic risk-free interest rate r .

● Rate of assets return R normally distributed, with (under P)

µ: expected assets return per year,σ: volatility of assets return per year.

R ∼P N ((µ − σ2

2)T , σ2T) .

● µ > r , or, equivalently, risk premium λ = µ−rσ

> 0.

● There exists a unique risk-neutral measure Q equivalent to P⇒ Complete and arbitrage-free market.

R ∼Q N ((r − σ2

2)T , σ2T) .

● Black-Scholes framework.

Policyholders’ preferences

u ∶ R→ R utility function: u twice differentiable, with u′ > 0, u′′ < 0 in R (homogeneouscohort of risk-averse policyholders).

Page 16: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

AssumptionsP historical measure.

Capital market conditions

● Deterministic risk-free interest rate r .

● Rate of assets return R normally distributed, with (under P)

µ: expected assets return per year,σ: volatility of assets return per year.

R ∼P N ((µ − σ2

2)T , σ2T) .

● µ > r , or, equivalently, risk premium λ = µ−rσ

> 0.

● There exists a unique risk-neutral measure Q equivalent to P⇒ Complete and arbitrage-free market.

R ∼Q N ((r − σ2

2)T , σ2T) .

● Black-Scholes framework.

Policyholders’ preferences

u ∶ R→ R utility function: u twice differentiable, with u′ > 0, u′′ < 0 in R (homogeneouscohort of risk-averse policyholders).

Page 17: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Constraints imposed

Fairness of the contract

L0 = EQ[e−rTLT ]

or equivalently

α = αGe−rT + αδe−rTEQ[(eR −G)+] − e−rTEQ[(αG − eR)+]

or elseα = αGe−rT + αδC(1,G) − P(1, αG).

Solvency requirements

Maximum ruin probability ε ∈ [0,1] fixed:

P(WT < αW0G) ≤ ε

or equivalently

αG ≤ exp((µ − σ2

2)T +N−1(ε)σ

√T).

Page 18: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Constraints imposed

Fairness of the contract

L0 = EQ[e−rTLT ]or equivalently

α = αGe−rT + αδe−rTEQ[(eR −G)+] − e−rTEQ[(αG − eR)+]

or elseα = αGe−rT + αδC(1,G) − P(1, αG).

Solvency requirements

Maximum ruin probability ε ∈ [0,1] fixed:

P(WT < αW0G) ≤ ε

or equivalently

αG ≤ exp((µ − σ2

2)T +N−1(ε)σ

√T).

Page 19: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Constraints imposed

Fairness of the contract

L0 = EQ[e−rTLT ]or equivalently

α = αGe−rT + αδe−rTEQ[(eR −G)+] − e−rTEQ[(αG − eR)+]

or elseα = αGe−rT + αδC(1,G) − P(1, αG).

Solvency requirements

Maximum ruin probability ε ∈ [0,1] fixed:

P(WT < αW0G) ≤ ε

or equivalently

αG ≤ exp((µ − σ2

2)T +N−1(ε)σ

√T).

Page 20: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Constraints imposed

Fairness of the contract

L0 = EQ[e−rTLT ]or equivalently

α = αGe−rT + αδe−rTEQ[(eR −G)+] − e−rTEQ[(αG − eR)+]

or elseα = αGe−rT + αδC(1,G) − P(1, αG).

Solvency requirements

Maximum ruin probability ε ∈ [0,1] fixed:

P(WT < αW0G) ≤ εor equivalently

αG ≤ exp((µ − σ2

2)T +N−1(ε)σ

√T).

Page 21: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

The problem setup

Constrained maximisation problem

Problemmax

(α,G)∈[0,1]×[0,+∞[

EP[u(e−rTLT − L0)]

Constraints

● Fairness conditionL0 = EQ[e−rTLT ]

● Ruin probability condition

P(WT < αW0G) ≤ ε

Remark: the problem is well-posed.

Page 22: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

The problem setup

Constrained maximisation problem

Problemmax

(α,G)∈[0,1]×[0,+∞[

EP[u(e−rTLT − L0)]

Constraints

● Fairness conditionL0 = EQ[e−rTLT ]

● Ruin probability condition

P(WT < αW0G) ≤ ε

Remark: the problem is well-posed.

Page 23: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

A possible interpretation of e−rTLT − L0

The objective function is given by

EP[u(e−rTLT − L0)].

Main characters: policyholders.

Initial investment (t = 0) Final payoff (t = T )Insurance company L0 = αW0 LT

Risk-free (r) (1 − α)W0 (1 − α)W0erT

W0 LT + (1 − α)W0erT

Policyholders’ final profit discounted at t = 0:

e−rT(LT + (1 − α)W0erT) −W0 = e−rTLT + (W0 −W0) − αW0

= e−rTLT − L0.

Page 24: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

A possible interpretation of e−rTLT − L0

The objective function is given by

EP[u(e−rTLT − L0)].

Main characters: policyholders.

Initial investment (t = 0) Final payoff (t = T )Insurance company L0 = αW0 LT

Risk-free (r) (1 − α)W0 (1 − α)W0erT

W0 LT + (1 − α)W0erT

Policyholders’ final profit discounted at t = 0:

e−rT(LT + (1 − α)W0erT) −W0 = e−rTLT + (W0 −W0) − αW0

= e−rTLT − L0.

Page 25: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

A possible interpretation of e−rTLT − L0

The objective function is given by

EP[u(e−rTLT − L0)].

Main characters: policyholders.

Initial investment (t = 0) Final payoff (t = T )Insurance company L0 = αW0 LT

Risk-free (r) (1 − α)W0 (1 − α)W0erT

W0 LT + (1 − α)W0erT

Policyholders’ final profit discounted at t = 0:

e−rT(LT + (1 − α)W0erT) −W0 = e−rTLT + (W0 −W0) − αW0

= e−rTLT − L0.

Page 26: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Pareto-optimality

The maximisation problem studied is related to a Pareto-optimalityproblem, characterised by the following criteria:

Policyholders’

Insurer’s

U(α,G) = EP[u(e−rTLT − L0)]Π(α,G) = P(WT < αW0G)

Theorem

For any given maximal ruin probability ε, the solutions (α,G) of thecorresponding maximisation problem are also pareto-optima.

Page 27: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Pareto-optimality

The maximisation problem studied is related to a Pareto-optimalityproblem, characterised by the following criteria:

Policyholders’

Insurer’s

U(α,G) = EP[u(e−rTLT − L0)]Π(α,G) = P(WT < αW0G)

Theorem

For any given maximal ruin probability ε, the solutions (α,G) of thecorresponding maximisation problem are also pareto-optima.

Page 28: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Study of the fairness condition

Page 29: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Study of the fairness condition

Page 30: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Set of admissible solutions

Page 31: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Set of admissible solutions

Page 32: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

The extremal case δ = 1

Theorem (δ = 1)

The solutions of the constrained maximisation problem are given by

● (α̂,0), for a unique α̂ ∈ ]0,1[, if EP[u′(W0(eR−rT − 1))(eR−rT − 1)] ≥ 0.

● (1,G), for all G ≤ exp((µ − σ2

2)T +N−1(ε)σ√T), otherwise.

Page 33: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

The extremal case δ = 1

Theorem (δ = 1)

The solutions of the constrained maximisation problem are given by

● (α̂,0), for a unique α̂ ∈ ]0,1[, if EP[u′(W0(eR−rT − 1))(eR−rT − 1)] ≥ 0.

● (1,G), for all G ≤ exp((µ − σ2

2)T +N−1(ε)σ√T), otherwise.

Page 34: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

The case δ ∈ [0,1[ - Preliminary results

Theorem (δ ∈ [0,1[)The only solution of the constrained maximisation problem is given by(α̃,Gδ(α̃)), for some α̃ ∈ ]0,1[.

Page 35: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Numerical results - 1

0.0 0.2 0.4 0.6 0.8 1.0

−15

−10

−5

05

δ=0.80

α

cert

aint

y eq

uiva

lent

0.0 0.2 0.4 0.6 0.8 1.0

−15

−10

−5

05

δ=0.95

α

cert

aint

y eq

uiva

lent

Certainty equivalent related to the policyholders’ expected utility, with δ = 0.80 (left),δ = 0.95 (right): T = 20, W0 = 100, r = 0.03, λ = 0.1, σ = 0.20 (µ = 0.05), ε = 0.05,

u(x) = (x+W0)1−ρ1−ρ , ρ = 1.5.

The ruin probability condition at the optimal solution may be binding or not!

Page 36: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Numerical results - 1

0.0 0.2 0.4 0.6 0.8 1.0

−15

−10

−5

05

δ=0.80

α

cert

aint

y eq

uiva

lent

0.0 0.2 0.4 0.6 0.8 1.0

−15

−10

−5

05

δ=0.95

α

cert

aint

y eq

uiva

lent

Certainty equivalent related to the policyholders’ expected utility, with δ = 0.80 (left),δ = 0.95 (right): T = 20, W0 = 100, r = 0.03, λ = 0.1, σ = 0.20 (µ = 0.05), ε = 0.05,

u(x) = (x+W0)1−ρ1−ρ , ρ = 1.5.

The ruin probability condition at the optimal solution may be binding or not!

Page 37: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Comparative statics (δ = 0.80)

●●

0.0 0.1 0.2 0.3 0.4 0.5 0.6

0.0

0.2

0.4

0.6

0.8

risk premium comparative statics

λ

optim

al α

0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.400.

00.

20.

40.

60.

8

volatility comparative statics

σ

optim

al α

Dependence of the optimal leverage ratio αon the risk premium λ, with σ = 0.20.

Dependence of the optimal leverage ratio αon the volatility σ, with λ = 0.1.

Page 38: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Comparative statics - 2

2 3 4 5 6 7 8

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

(relative) risk aversion comparative statics

ρ

optim

al α

Dependence of the optimal leverage ratio α on the relative risk aversion ρ, with λ = 0.1,

σ = 0.20, δ = 0.80.

Page 39: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

To sum up

● We consider a model for the creation of a life insurance company issuingparticipating contracts.

● We maximise the policyholders’ preferences.

● We impose fair valuation and solvency constraints.

● Analytical results: if λ > 0 and G is upper unbounded, then the optimal solutioninduces the policyholders to insure (α > 0).

The properties of the optimal contracts depend on the choice of δ (δ = 1, δ < 1)and, if δ = 1, on a condition involving the policyholders’ risk aversion.

● Numerical results: according to the parameters of the contract chosen, the ruinprobability condition evaluated at the optimal solution may be binding or not.

This contradicts the assumption made in Schmeiser, Wagner (2015).

Page 40: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

To sum up

● We consider a model for the creation of a life insurance company issuingparticipating contracts.

● We maximise the policyholders’ preferences.

● We impose fair valuation and solvency constraints.

● Analytical results: if λ > 0 and G is upper unbounded, then the optimal solutioninduces the policyholders to insure (α > 0).

The properties of the optimal contracts depend on the choice of δ (δ = 1, δ < 1)and, if δ = 1, on a condition involving the policyholders’ risk aversion.

● Numerical results: according to the parameters of the contract chosen, the ruinprobability condition evaluated at the optimal solution may be binding or not.

This contradicts the assumption made in Schmeiser, Wagner (2015).

Page 41: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

To sum up

● We consider a model for the creation of a life insurance company issuingparticipating contracts.

● We maximise the policyholders’ preferences.

● We impose fair valuation and solvency constraints.

● Analytical results: if λ > 0 and G is upper unbounded, then the optimal solutioninduces the policyholders to insure (α > 0).

The properties of the optimal contracts depend on the choice of δ (δ = 1, δ < 1)and, if δ = 1, on a condition involving the policyholders’ risk aversion.

● Numerical results: according to the parameters of the contract chosen, the ruinprobability condition evaluated at the optimal solution may be binding or not.

This contradicts the assumption made in Schmeiser, Wagner (2015).

Page 42: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

To sum up

● We consider a model for the creation of a life insurance company issuingparticipating contracts.

● We maximise the policyholders’ preferences.

● We impose fair valuation and solvency constraints.

● Analytical results: if λ > 0 and G is upper unbounded, then the optimal solutioninduces the policyholders to insure (α > 0).

The properties of the optimal contracts depend on the choice of δ (δ = 1, δ < 1)and, if δ = 1, on a condition involving the policyholders’ risk aversion.

● Numerical results: according to the parameters of the contract chosen, the ruinprobability condition evaluated at the optimal solution may be binding or not.

This contradicts the assumption made in Schmeiser, Wagner (2015).

Page 43: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

To sum up

● We consider a model for the creation of a life insurance company issuingparticipating contracts.

● We maximise the policyholders’ preferences.

● We impose fair valuation and solvency constraints.

● Analytical results: if λ > 0 and G is upper unbounded, then the optimal solutioninduces the policyholders to insure (α > 0).

The properties of the optimal contracts depend on the choice of δ (δ = 1, δ < 1)and, if δ = 1, on a condition involving the policyholders’ risk aversion.

● Numerical results: according to the parameters of the contract chosen, the ruinprobability condition evaluated at the optimal solution may be binding or not.

This contradicts the assumption made in Schmeiser, Wagner (2015).

Page 44: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

State-of-the-art and possible developments

● The maximisation problem has been completely solved analyticallywhen δ = 1.

What can be said more on the case δ < 1 (and λ > 0)?

● The model has been constructed considering the investment risk, only.

What if the mortality risk is also taken into account?

● The risk-free rate r is assumed to be deterministic.

What if it is stochastic (Briys, de Varenne, 1994)?

● The rate of assets return R is modelled by a normal distribution.

What if R has a different type of distribution?

Page 45: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

State-of-the-art and possible developments

● The maximisation problem has been completely solved analyticallywhen δ = 1.

What can be said more on the case δ < 1 (and λ > 0)?

● The model has been constructed considering the investment risk, only.

What if the mortality risk is also taken into account?

● The risk-free rate r is assumed to be deterministic.

What if it is stochastic (Briys, de Varenne, 1994)?

● The rate of assets return R is modelled by a normal distribution.

What if R has a different type of distribution?

Page 46: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

State-of-the-art and possible developments

● The maximisation problem has been completely solved analyticallywhen δ = 1.

What can be said more on the case δ < 1 (and λ > 0)?

● The model has been constructed considering the investment risk, only.

What if the mortality risk is also taken into account?

● The risk-free rate r is assumed to be deterministic.

What if it is stochastic (Briys, de Varenne, 1994)?

● The rate of assets return R is modelled by a normal distribution.

What if R has a different type of distribution?

Page 47: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

State-of-the-art and possible developments

● The maximisation problem has been completely solved analyticallywhen δ = 1.

What can be said more on the case δ < 1 (and λ > 0)?

● The model has been constructed considering the investment risk, only.

What if the mortality risk is also taken into account?

● The risk-free rate r is assumed to be deterministic.

What if it is stochastic (Briys, de Varenne, 1994)?

● The rate of assets return R is modelled by a normal distribution.

What if R has a different type of distribution?

Page 48: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

State-of-the-art and possible developments

● The maximisation problem has been completely solved analyticallywhen δ = 1.

What can be said more on the case δ < 1 (and λ > 0)?

● The model has been constructed considering the investment risk, only.

What if the mortality risk is also taken into account?

● The risk-free rate r is assumed to be deterministic.

What if it is stochastic (Briys, de Varenne, 1994)?

● The rate of assets return R is modelled by a normal distribution.

What if R has a different type of distribution?

Page 49: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

State-of-the-art and possible developments

● The maximisation problem has been completely solved analyticallywhen δ = 1.

What can be said more on the case δ < 1 (and λ > 0)?

● The model has been constructed considering the investment risk, only.

What if the mortality risk is also taken into account?

● The risk-free rate r is assumed to be deterministic.

What if it is stochastic (Briys, de Varenne, 1994)?

● The rate of assets return R is modelled by a normal distribution.

What if R has a different type of distribution?

Page 50: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

State-of-the-art and possible developments

● The maximisation problem has been completely solved analyticallywhen δ = 1.

What can be said more on the case δ < 1 (and λ > 0)?

● The model has been constructed considering the investment risk, only.

What if the mortality risk is also taken into account?

● The risk-free rate r is assumed to be deterministic.

What if it is stochastic (Briys, de Varenne, 1994)?

● The rate of assets return R is modelled by a normal distribution.

What if R has a different type of distribution?

Page 51: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

State-of-the-art and possible developments

● The maximisation problem has been completely solved analyticallywhen δ = 1.

What can be said more on the case δ < 1 (and λ > 0)?

● The model has been constructed considering the investment risk, only.

What if the mortality risk is also taken into account?

● The risk-free rate r is assumed to be deterministic.

What if it is stochastic (Briys, de Varenne, 1994)?

● The rate of assets return R is modelled by a normal distribution.

What if R has a different type of distribution?

Page 52: On the optimal design of participating life insurance ... · Common Pitfalls, The Journal of Risk and Insurance, 64(4), 673{694 (1997). P H. Schmeiser, J. Wagner, A Proposal on How

Thank you for your attention!