Skip to content

Actuarial Mathematics for Modelling · Valuing cashflows contingent on multiple transition events

Multiple Decrement Tables and Joint Life Applications

Updated 11 October 2026 · Fact-checked

A multiple decrement table follows a group of lives that can leave for several reasons, such as death or withdrawal. You value a benefit by multiplying the probability of each exit in each year by its benefit and discount factor. For joint life and last survivor cases, you work from the status's survival probabilities.

Understand Multiple Decrement Tables and Joint Life Applications

A multiple decrement table follows one group of lives as they leave by different causes. For example, an employee may leave service by death, withdrawal, ill-health retirement or age retirement. The table gives l(x), the number still in the group at age x. It also gives d(x, j), the number who leave by cause j between ages x and x+1. The total leaving in the year is the sum of the d(x, j) over all causes.

The key point is that each cause competes with the others. The dependent rate q(x, j) = d(x, j) ÷ l(x) is the chance of leaving by cause j in the year, in the presence of all other causes. The independent (absolute) rate q'(x, j) is the chance of leaving by cause j if no other cause existed. Dependent rates are always no larger than the matching independent rates. The table gives you dependent rates directly. Independent rates are theoretical, and you need an assumption to link them to the table.

To value a cashflow, treat each exit as an event with a probability, a benefit and a timing. The expected present value is the sum, over years and causes, of the probability of exit by that cause in that year, times the benefit, times the discount factor. In continuous form the density of exit by cause j at age x+s is s p(x, τ) × μ(x+s, j), where τ means all causes together.

Joint life applications use the same idea with two lives. The joint life status (xy) lasts until the first death. The last survivor status lasts until the second death. You can model this as a multiple state model with four states: both alive, only x alive, only y alive, both dead. The transition intensities may depend on whether the other life is alive. If the two lives are independent, the formulas collapse to simple products and sums of single life values.

In the exam you are asked to find a probability, a present value or a premium. Define the states or decrements, write the cashflow in each, and weight by the right probability. Then check that your probabilities add up sensibly.

Key rules to remember

Total survival and decrement identity
l(x+1) = l(x) − Σ d(x, j); p(x, τ) = l(x+1) ÷ l(x); q(x, j) = d(x, j) ÷ l(x)
Dependent rates use l(x), the whole group at age x, as the denominator, not the number left by one cause.
Probability of surviving t years in the table
t p(x, τ) = l(x+t) ÷ l(x)
Use it to find the chance of still being in the group at a later age.
Probability of exit by cause j in year k+1
(k p(x, τ)) × q(x+k, j) = d(x+k, j) ÷ l(x)
Survive k years in the group, then leave by cause j in the next year.
Expected present value of a benefit paid at year end of exit
EPV = Σ over k and j of B(j, k+1) × v^(k+1) × d(x+k, j) ÷ l(x)
B is the benefit for cause j. Use v = 1 ÷ (1 + i), and adjust the power if payment is at the moment of exit.
Force of decrement link
q(x, j) = ∫ from 0 to 1 of s p(x, τ) × μ(x+s, j) ds; s p(x, τ) = exp(−∫ from 0 to s of μ(x+r, τ) dr); μ(τ) = Σ μ(j)
The total force is the sum of the forces for each cause. If the forces are constant over the year, q(x, j) = (μ(j) ÷ μ(τ)) × (1 − e^(−μ(τ))).
Independent and dependent rates
p(x, τ) = Π p'(x, j) if the causes are independent; p'(x, j) = exp(−∫ from 0 to 1 of μ(x+s, j) ds)
This holds only when the causes are independent. The table alone does not tell you whether they are.
Joint life survival (independent lives)
t p(xy) = t p(x) × t p(y)
Assumes the future lifetimes are independent. State this assumption.
Last survivor survival
t p(x̄ȳ) = t p(x) + t p(y) − t p(xy)
At least one life alive. For independent lives, also equals 1 − t q(x) × t q(y).
Assurances on two lives
A(x̄ȳ) = A(x) + A(y) − A(xy)
Holds for assurances paid at the same time point after the last or first death, and is valid without independence.
Annuities on two lives
ä(x̄ȳ) = ä(x) + ä(y) − ä(xy); reversionary annuity to y after x = ä(y) − ä(xy)
Annuity while exactly one life is alive = ä(x) + ä(y) − 2 × ä(xy).
Assurance and annuity link
A(xy) = 1 − d × ä(xy), where d = i ÷ (1 + i)
Works for whole life joint and last survivor statuses with the same timing basis.
Equivalence principle premium
P × ä(premium status) = EPV of benefits
Use the premium status for the premium term and the benefit status for the benefit.

How to solve Multiple Decrement Tables and Joint Life Applications questions

Use this method for any question on multiple decrements or joint life cashflows. Do not start calculating before you have set up the structure.

  1. 1Identify the lives and the statuses. Write down the decrements or the states, for example death, withdrawal, joint life, last survivor, both alive.
  2. 2List each cashflow and say exactly when it is paid and on which event. Mark premiums separately from benefits.
  3. 3Choose the probability you need. For tables, use d(x+k, j) ÷ l(x) for exit by cause j in year k+1. For two lives, use joint or last survivor survival probabilities.
  4. 4State the assumptions you rely on, such as independence of lives, independence of causes, uniform distribution of decrements, or constant forces.
  5. 5Apply the discount factor for each payment time. Match the factor to the timing, either year end or moment of exit.
  6. 6Sum the terms to get the expected present value. Use the equivalence principle if you need a premium.
  7. 7Check the result. Probabilities should not exceed 1, and the last survivor value should be no smaller than the joint life value.
  8. 8Write the final answer with units, and give a one-line comment if the question asks for interpretation.

Quickest way: Table columns and status identities

When to use it: Use this when you are given a table or single life values and have limited time.

  1. For a table, add three columns next to it: probability of exit by the cause you need (d ÷ l at the starting age), discount factor, and benefit. Multiply across and add down.
  2. Always divide by l at the valuation age, not by l at the age of exit.
  3. For two lives, convert at once to single life and joint values. Use the identities A(x̄ȳ) = A(x) + A(y) − A(xy) and ä(x̄ȳ) = ä(x) + ä(y) − ä(xy).
  4. For a reversionary annuity, use ä(y) − ä(xy) directly, with no probabilities.
  5. Only build a full multiple state model with Kolmogorov equations if the question gives intensities that depend on the other life.

Common mistakes in Multiple Decrement Tables and Joint Life Applications

  • Using l(x+k) as the denominator when finding the chance of exit in a later year from age x.

    The table gives d ÷ l at each age, and students read it as the probability from the start.

    Fix: Divide by l at the starting age. The chance of exit by cause j in year k+1 is d(x+k, j) ÷ l(x). Or multiply k p(x, τ) by q(x+k, j).

  • Treating the dependent rate q(x, j) as if it were the independent rate q'(x, j).

    Both are labelled as the rate for cause j, and a single decrement table uses q' only.

    Fix: Check which one the question gives. Dependent rates come from the multiple decrement table. Independent rates apply when the cause acts alone. Convert only with a stated assumption.

  • Multiplying survival probabilities of two lives without stating independence.

    The product formula t p(xy) = t p(x) × t p(y) is so common that it becomes automatic.

    Fix: Write the assumption. If the question gives a joint model with dependence, such as a four-state model with different intensities after one death, do not use the product.

  • Writing the last survivor value as the sum of the single life values, forgetting to subtract the joint value.

    Students remember a plus sign but not the correction for double counting the period when both are alive.

    Fix: Use ä(x̄ȳ) = ä(x) + ä(y) − ä(xy). Then check that the result is above both single life values.

  • Valuing benefits with the wrong timing of payment, such as year end instead of moment of exit.

    Tables give yearly decrements, and the discount power is not adjusted.

    Fix: Read the payment basis. For year-end payment, use v^(k+1). For payment at exit with uniform distribution, apply the adjustment the question's basis requires, such as a half-year factor.

  • Adding probabilities of the same event under different causes to get a rate of exit by one cause.

    Students mix up total decrement with one cause.

    Fix: Sum d(x, j) over all causes to get total exits. Use a single cause only when the benefit depends on that cause.

Worked examples

Example 1

A multiple decrement table for employees aged 60 shows: l(60) = 1,000; deaths d(60, death) = 20 and withdrawals d(60, withdrawal) = 80. At age 61: deaths 27 and withdrawals 63. A death benefit of ₹1,00,000 is paid at the end of the year of death. No benefit is paid on withdrawal. Interest is 5% a year. Find (a) the probability that an employee aged 60 is still in service at age 62, and (b) the expected present value of the death benefit for an employee aged 60 over the two years.

Show the solution
  1. Find l(61): 1,000 − 20 − 80 = 900.
  2. Find l(62): 900 − 27 − 63 = 810.
  3. (a) The probability of being in service at 62 is l(62) ÷ l(60) = 810 ÷ 1,000 = 0.81.
  4. (b) The chance of death in year 1 is 20 ÷ 1,000 = 0.02. The chance of death in year 2 is 27 ÷ 1,000 = 0.027, using l(60) as the denominator.
  5. Discount factors: v = 1 ÷ 1.05 = 0.952381; v² = 1 ÷ 1.1025 = 0.907029.
  6. Year 1 term: 0.02 × 0.952381 = 0.0190476. Year 2 term: 0.027 × 0.907029 = 0.0244898.
  7. Sum: 0.0190476 + 0.0244898 = 0.0435374.
  8. Multiply by ₹1,00,000: 0.0435374 × 1,00,000 = ₹4,353.74.

Answer: (a) 0.81. (b) The expected present value of the death benefit is about ₹4,354 (₹4,353.74).

Example 2

Two independent lives, (x) and (y), are valued at the same interest rate. You are given ä(x) = 12, ä(y) = 14 and ä(xy) = 10. An annuity of ₹1,00,000 a year is paid annually in advance while at least one of them is alive. Find (a) its expected present value, (b) the expected present value of a reversionary annuity of ₹1,00,000 a year paid to (y) after the death of (x), starting from the first payment date after that death, and (c) the value of ₹1 a year payable only while exactly one life is alive.

Show the solution
  1. (a) Last survivor annuity factor: ä(x̄ȳ) = ä(x) + ä(y) − ä(xy) = 12 + 14 − 10 = 16.
  2. Multiply by ₹1,00,000: 16 × 1,00,000 = ₹16,00,000.
  3. (b) The reversionary annuity is paid while y is alive and x is dead. Its factor is ä(y) − ä(xy) = 14 − 10 = 4.
  4. Multiply by ₹1,00,000: 4 × 1,00,000 = ₹4,00,000.
  5. (c) Exactly one alive: ä(x) + ä(y) − 2 × ä(xy) = 12 + 14 − 20 = 6.
  6. Check: the last survivor factor 16 equals the joint factor 10 plus the exactly-one factor 6. This is consistent.

Answer: (a) ₹16,00,000. (b) ₹4,00,000. (c) The factor is 6 for ₹1 a year, which is ₹6,00,000 for ₹1,00,000 a year.

Exam tips

  • Write the denominator before you write the numerator. Examiners award a mark for showing l at the starting age, and that is where most errors come from.
  • State assumptions in one line each: independence of lives, independence of causes, UDD or constant forces. These carry marks in written answers.
  • For joint life questions, reduce to single life and joint values before doing any arithmetic. It saves time and avoids errors.
  • In a Paper B style R or Excel question, build the decrement table with columns for l, each d, and the exit probabilities. Check that l(x+1) equals l(x) minus total d before you value anything.
  • In multiple choice questions, sanity-check the result. A last survivor value must be at least the joint life value, and an exit probability cannot exceed 1.

Practice questions from Valuing cashflows contingent on multiple transition events

Multiple Decrement Tables and Joint Life Applications: frequently asked questions

What is the difference between dependent and independent decrements?

Dependent decrement rates are the chances of leaving by a cause when all the causes act together, and the table shows these. Independent rates are the chances of leaving by that cause if it acted alone. Independent rates are at least as large as the dependent rates, and you need a stated assumption to convert between them.

How do I value a benefit paid on death of two lives?

Decide whether the benefit is paid on the first death or the second. For the first death, use the joint life status (xy). For the second death, use the last survivor status. Then use the matching assurance value, with A(x̄ȳ) = A(x) + A(y) − A(xy) for the last survivor.

When do I need a multiple state model for two lives?

You need one when the intensity of death for one life changes after the other life dies, or when the question gives intensities by state. If the lives are independent, single life and joint life values are enough and are quicker.

Why does the denominator in a multiple decrement table stay l at the starting age?

You are finding the chance that a life now aged x leaves by a cause at a later age. That is a probability from the starting point, so the count at the start is the base. Using l at the later age would give only the chance of leaving in that year, given still in the group.