IAI Actuarial Core Principles · Actuarial Mathematics for Modelling

Valuing Cashflows Contingent on Multiple Transition Events

This chapter values cashflows that depend on a person moving between states, such as healthy, sick, dead or withdrawn. You model the states as a Markov process, use transition intensities to get probabilities, then take the expected present value of each payment. Thiele's equations give policy values step by step.

What this chapter covers

This chapter extends the two-state alive-dead model to models with several states. A person can be healthy, sick, disabled, withdrawn or dead. Each move between states is a transition. Payments depend on the state you are in, or on the move you make. Examples are income protection, critical illness, and pensions with ill-health retirement.

You start with the Markov property: the future depends only on the present state, not on how you got there. Then you describe the model with transition intensities (forces of transition) μ_ij(x). From these you build transition probabilities using the Kolmogorov forward equations. With the probabilities, you value cashflows as expected present values. You then switch to the differential view with Thiele's equations, which track the policy value over time.

In CM1 this chapter sits inside the Pricing and reserving and Decrement and multiple life models areas. It uses the theory of interest and equation of value from earlier chapters. It also feeds directly into premium calculation, reserving and profit testing. Expect it in the written Paper A, and in Paper B where you may solve the equations numerically in R or Excel.

Multi-state models and Thiele's equations are core tools for the actuarial work CM1 tests: pricing and reserving. Questions here tend to be long, multi-part and mark-rich, so one well-prepared topic can earn a large share of a written question. The same ideas also appear in Paper B, where you must code the method and show working. If you master the logic once, you can handle many product types, because every question follows the same pattern: define states, write intensities, get probabilities or differential equations, then value.

Valuing cashflows contingent on multiple transition events: topics in the order to study them

1Multiple State Models and Markov ProcessesEverything else needs you to define states, transitions and the Markov property clearly, so start here.
2Transition Intensities and Kolmogorov Forward EquationsIntensities are the inputs of the model, and the forward equations turn them into transition probabilities.
3Valuing Cashflows Using Multiple State ModelsOnce you have probabilities, you can write expected present values for benefits and premiums and set up premium equations.
4Thiele's Differential Equations for Policy ValuesThis re-expresses the reserve as a differential equation. It needs the earlier valuation ideas and the intensities.
5Multiple Decrement Tables and Joint Life ApplicationsThese are special cases of the multi-state framework, so they are easiest to learn last, once the general method is clear.

How to prepare Valuing cashflows contingent on multiple transition events

Treat this chapter as one method applied to different models. Learn the method first, then practise it on varied products.

Learn the definitions: states, transitions, Markov property, transition intensity. Be able to state each in a sentence.
Draw a state diagram for every question before doing any algebra. Label each arrow with its intensity and mark which states pay benefits or premiums.
Practise writing the Kolmogorov forward equation for any given model, for example d/dt p_ij(t) = Σ over k≠j of [p_ik(t) μ_kj(t+x) − p_ij(t) μ_jk(t+x)]. Check each term is inflow minus outflow.
Value cashflows by writing the expected present value as an integral or sum of payment × discount factor × probability of being in the right state. State your assumptions, such as constant intensities or a fixed interest rate.
Write Thiele's equation for each state: the rate of change of reserve equals interest on the reserve, plus premium rate, minus benefit outgo, minus the expected cost of transitions, which is the intensity times the change in reserve plus any lump sum. Then solve numerically with Euler's method, working backwards from the terminal condition.
Do full past-paper questions under time. Then repeat the numerical parts in R or Excel for Paper B, showing the formula, the working and the result.
Finish with decrement tables and joint life problems, and check that you can express them as multi-state models.

Common mistakes in Valuing cashflows contingent on multiple transition events

Mixing up transition intensities with transition probabilities.
Fix: Remember that an intensity is a rate per unit time, and a probability is a number between 0 and 1 over a period. Convert with the Kolmogorov equations or integration, and check units.
Missing a term in the Kolmogorov or Thiele equation.
Fix: Go through every arrow in the diagram. For each state, list all arrows in and out, and check each one appears exactly once.
Using the wrong sign or wrong reserve in the transition term of Thiele's equation.
Fix: Read it as the extra cost to the insurer of moving from i to j: the lump sum paid plus the new reserve needed, minus the reserve already held.
Running Euler's method forwards instead of backwards, or ignoring the terminal condition.
Fix: Write the terminal reserve first, for example zero at the end of a term with no maturity benefit. Then step back using V(t−h) ≈ V(t) − h × dV/dt.
Assuming independence or constant intensities without saying so.
Fix: State each assumption in one line. Examiners give marks for stating assumptions and may penalise results that rely on unstated ones.
Showing only a final number in computer-based answers.
Fix: Write the formula in standard notation, show the key working or code, and state the result with a short interpretation.

Last-day revision: Valuing cashflows contingent on multiple transition events

Markov property: the future depends only on the current state, not on the past path.
μ_ij(x) is the transition intensity from state i to state j at age x, where i ≠ j.
Over a small time h, P(move i to j) ≈ h × μ_ij(x), with an error that is o(h).
Forward equation: the rate of change of p_ij is inflow to j minus outflow from j.
Boundary condition for the forward equation: p_ii(0) = 1 and p_ij(0) = 0 for i ≠ j.
Expected present value = Σ or ∫ of payment × discount factor × probability of the right state or transition.
Thiele: dV_i/dt = δ V_i + P_i − B_i − Σ over j≠i of μ_ij (S_ij + V_j − V_i).
Solve Thiele backwards from the known terminal reserve, not forwards from time zero.
Euler's method with a small step gives an approximate reserve; state the step size you use.
In a multiple decrement model, the total force of decrement is the sum of the individual forces.
For joint life problems, state independence assumptions clearly before using products of probabilities.
Always draw the diagram and state your assumptions before writing any equation.

Valuing cashflows contingent on multiple transition events practice questions

Valuing cashflows contingent on multiple transition events: frequently asked questions

What is a multiple state model in CM1?

It is a model with several states, such as healthy, sick and dead, and defined transitions between them. Each transition has an intensity. You use the model to find probabilities of being in each state and to value benefits and premiums that depend on state.

Do I need to memorise Thiele's equation?

You should know its structure well enough to write it from a state diagram. Learn it as interest on the reserve plus premium, minus benefits, minus the expected cost of transitions. Understanding the meaning of each term is safer than rote learning.

Will this chapter appear in Paper B?

It can. Paper B is computer-based, so you may need to solve Thiele's equations or probability equations numerically in R or Excel. Practise showing the method, the formula and the result, not only the output.

How is a multiple decrement table related to multi-state models?

A multiple decrement model is a multi-state model with one starting state and several exit states, and no moves back. Once you understand the general framework, decrement tables and many joint life problems become special cases.