Encoding Ancillary Benefits as Decrement-Triggered Cash Flows

From Benefit Rules to Cash Flows

A pension plan rule often describes a benefit in words: a refund is payable on withdrawal, a vested member receives a specified amount, or an eligible spouse receives a pension after the member dies. Valuation requires translating each rule into a precise contingent cash flow. That translation must identify what event triggers payment, who is eligible, how much is paid, and when it is paid.

The key modeling discipline is to keep the decrement event separate from the benefit condition. For example, death may trigger a survivor benefit, but a payment may occur only if an eligible spouse exists. Likewise, withdrawal may trigger either a vested benefit or a refund, depending on the member’s status at withdrawal. These conditions determine which cash flow applies; they do not change the definition of the decrement probability itself.

This tutorial treats hypothetical benefit rules as cash-flow specifications. It does not assume that every exit produces a payment, that every payment occurs at the exit date, or that eligibility is independent of the exit cause. The methods are aligned with the SOA Exam ALTAM (Advanced Long-Term Actuarial Mathematics) syllabus.

A General Cash-Flow Representation

Consider an active member at time 0. Let \(T\) be the time of exit, and let \(J\) identify the exit cause. For example, \(J\) might be withdrawal, death, or retirement. In a multiple-decrement model, the exit causes are mutually exclusive: at a given exit, exactly one cause is recorded. Let \(E\) denote an eligibility condition, such as having an eligible spouse at death, and let \(Z(T-)\) denote the member’s state immediately before exit.

A benefit rule can be represented as a collection of payments. For payment number \(r\), write its amount as \(B_r(J,E,Z(T-))\), its payment time as \(\tau_r\), and its discount factor as \(v^{\tau_r}\), where \(v=1/(1+i)\) for effective annual interest rate \(i\). The payment amount can be zero when its trigger or eligibility condition is not met.

This expression is a bookkeeping framework, not a substitute for specifying the model. The expectation must use the relevant probability of each decrement and the probability of any eligibility condition given that decrement. If benefits are paid only at the exit date, then \(\tau_r=T\). A survivor pension may instead have payment times after the member’s death.

In a discrete annual model, suppose \(q_{k,j}\) is the probability that a member active at the start of year \(k\) exits during that year from cause \(j\). These are multiple-decrement probabilities: the causes compete, and their probabilities are defined on the same active population. If \(N_k\) is the probability the member is still active at the start of year \(k\), then the probability of a cause-\(j\) exit in year \(k\) is \(N_k q_{k,j}\). If the exit occurs at year-end, a payment at exit is discounted by \(v^k\).

Why the Encoding Gives the Expected Present Value

The valuation formula follows by conditioning on the mutually exclusive events that determine the payment. First condition on the year and cause of exit. Then, within that event, condition on eligibility or member state if the benefit amount depends on either. Finally, discount each resulting payment at its actual payment time.

For a single payment at exit, this conditioning gives the useful form

The conditional expected benefit is essential when members exiting for the same cause can receive different amounts. If eligibility is already known for each member, the expectation is replaced by the appropriate status-specific amount. If a payment follows the exit, its own discount factor must be used inside the conditional expectation.

Encoding Eligibility and Payment Schedules

Eligibility can be represented either by explicit subgroups or by conditional probabilities. If, for example, a fraction of members dying in a period has an eligible spouse, use the probability of an eligible spouse conditional on death. Do not automatically multiply by the population-wide spouse proportion: the relevant proportion may differ among members who die.

The same principle applies to the payment schedule. A one-time refund at withdrawal is discounted to the withdrawal date. A survivor pension beginning after the member’s death is a set of payments at later dates. For an introductory calculation, the installments can be valued directly; in more involved applications, a suitable annuity value may summarize them. The benefit rule must specify which installments are payable and when.

Multiple Years and Member-State Exposure

For multiple years, each year’s decrement probabilities apply to members who are still active at that year’s start. If the probability of remaining active through one year is 0.85, then only 85% of the initial group is exposed to the next year’s decrement probabilities. This exposure factor must be applied before adding the next year’s expected payments.

A practical way to avoid omissions is to create one row for each exit cause and year, then record the probability of reaching that row, the conditional decrement probability, the eligibility rule, the amount, and each payment date. A benefit that is payable only in a specified member state should appear only in the row for that state and trigger. This makes mutually exclusive exits explicit and prevents counting two incompatible benefits for the same exit.

Common Mistakes and a Reliable Checklist

The benefit description, decrement model, and payment timing must agree. The most common errors are not difficult arithmetic; they are errors in the event being valued or the date at which money is paid.

• Using the wrong eligibility proportion. If the problem gives eligibility conditional on death or withdrawal, use that conditional proportion for the corresponding cause.
• Applying a population proportion to a cause-specific event without justification. A subgroup’s share of all members need not equal its share among members experiencing a particular decrement.
• Discounting every payment to the exit date. A pension installment paid after death is discounted to its own payment time.
• Counting incompatible benefits together. If a member receives either a vested amount or a refund, encode the alternatives as conditional amounts, not as two payments on the same withdrawal.
• Forgetting exposure in later years. Multiply each later year’s conditional decrement probability by the probability of remaining in the relevant state until that year.
• Treating multiple-decrement probabilities as independent single-decrement probabilities. The cause probabilities in a multiple-decrement model apply to the same exposed population and represent competing exits.

Before calculating an EPV, write down the trigger, eligibility condition, amount, and payment date for each benefit. Then identify the probability of reaching the relevant state and the probability of the specified decrement from that state. This separation is useful whenever a plan has several exit causes or conditional benefits.