State Benefits With Several Eligible States

Introduction

A contract may pay a fixed benefit if a policyholder is disabled, requires long-term care, or is in either condition at a specified date. The valuation question is not simply how likely each condition is: it is also whether the eligible conditions are mutually exclusive states or overlapping events. That distinction determines whether probabilities can be added directly or whether overlap must be allowed for.

This tutorial values benefits payable at a specified time when the policyholder is in any of several eligible states. It also considers a different structure: benefit triggers defined by conditions that can hold simultaneously. The benefit wording matters. A payment made once if at least one trigger applies is not valued in the same way as separate payments for each applicable trigger.

Adding Probabilities Across Eligible States

In a multiple-state model, states such as healthy, disabled, requiring long-term care, and dead are defined as distinct categories. At a fixed time, the policyholder occupies one of them. Therefore the events \(J_t=j\) for different states \(j\) are mutually exclusive: the policyholder cannot be in two of these states at once. If the benefit is the same amount in each eligible state, the probability of payment is the sum of the probabilities of those states.

If the amount depends on the eligible state, write \(b_j\) for the payment in state \(j\). The expected present value at a fixed payment time is then

This calculation uses the state probabilities at the payment time. It does not require the paths taken to reach those states to be mutually exclusive across different times. For benefits at several payment dates, expected discounted payments can be added date by date, with each date's payment rule valued using the relevant state probabilities.

Using a Transition Probability Matrix

When state probabilities at the payment date are not supplied directly, they can be obtained from a transition probability matrix. In a time-homogeneous discrete-time Markov model, each matrix entry \(p_{ij}\) is the probability of moving from state \(i\) to state \(j\) in one period. With row probability vectors, the distribution after \(n\) periods is the initial row vector multiplied by the \(n\)th power of the transition matrix. This assumes the same transition probabilities apply in each period and that the current state contains the information needed for the model's transitions.

Once the distribution at the benefit date is found, the valuation is the same: add the probabilities of eligible states, multiply by the appropriate benefit amount, and discount to the valuation date. A probability for the payment union should be calculated from the final state distribution, not from a sum of transition probabilities that may describe different routes or periods.

Why the Probability Sum Works

The valuation formula follows from writing the payment as a random variable. Let \(I_j\) equal 1 if the policyholder is in state \(j\) at time \(t\), and 0 otherwise. For a common benefit amount \(B\), the payment at time \(t\) is \(B\sum_{j\in\mathcal{E}}I_j\). Since exactly one state is occupied, this sum equals 1 when the policyholder is in an eligible state and 0 otherwise. Taking expectations and applying linearity of expectation gives the result:

For state-specific amounts \(b_j\), the payment is \(v^t\sum_{j\in\mathcal{E}}b_jI_j\); taking its expectation gives \(v^t\sum_{j\in\mathcal{E}}b_jP(J_t=j)\). This argument depends on the payment definition: it treats each eligible state as one alternative state and applies the corresponding amount once.

Overlapping Triggers Are Not Mutually Exclusive States

A contract may define eligibility using characteristics rather than mutually exclusive state labels. For example, a person may both need help with activities of daily living and have a qualifying cognitive impairment. These conditions overlap. If a single benefit is paid once when either condition is satisfied, adding the two individual probabilities counts people meeting both conditions twice.

For two overlapping events \(A\) and \(C\), the probability that at least one occurs is

If the contract instead pays a separate amount for each satisfied trigger, the payment amount in the overlap case may be the sum of both amounts. In that case, value each trigger payment according to its own wording; do not replace the structure with a once-only union benefit. One way to avoid ambiguity is to expand the model into distinct joint categories—neither condition, A only, C only, and both—and state explicitly what is paid in each category.

Exam Checks and Common Traps

For a fixed-time benefit, first translate the contract wording into a payment rule. Then identify whether the qualifying categories are mutually exclusive states or overlapping events. Finally, calculate the payment probability or expected amount and apply the discount factor for the actual payment time.

• Do not confuse state probabilities with transition probabilities. A one-period transition probability is not automatically a probability of being in that state after several periods.
• Do not double-count overlapping events. For a once-only benefit triggered by A or C, use the union probability, not the sum of marginal probabilities unless \(P(A\cap C)=0\).
• Read the payment rule carefully. A benefit paid once if any trigger applies differs from separate amounts paid for each trigger that applies.
• Discount from the benefit date. A benefit at time \(t\) is discounted by \(v^t\), not by the factor for one period unless \(t=1\).

These calculations are aligned with the SOA Exam ALTAM (Advanced Long-Term Actuarial Mathematics) syllabus. The central discipline is to represent the actual contract rule faithfully: state sets are disjoint by construction, while benefit triggers may overlap unless the wording or model makes them mutually exclusive.