A Timed Workflow for Integrated Profit and Reserve Problems

Why a Fixed Workflow Matters

An integrated profit-and-reserve problem can involve several moving parts: a policy changes state, cash flows occur at different points in the year, and reserves depend on what remains to be paid or received. A sound calculation can still produce a wrong result if, for example, a year-end claim is treated as a beginning-of-year payment or a reserve conditional on disability is applied to the active population.

The practical safeguard is to establish the timeline before calculating. Then handle the state probabilities, cash flows, reserves, and profits in a fixed order. The workflow in this tutorial is aligned with the SOA Exam ALTAM (Advanced Long-Term Actuarial Mathematics) syllabus.

Definition: A profit vector records profit by duration and, when needed, by policy state. The profit signature is the sequence of expected aggregate profits by duration, normally expressed per policy at issue. The net present value (NPV) of that signature is the sum of its profits discounted to the chosen valuation time.

A Timed Workflow

For a discrete annual model, use the following sequence. The times shown are a study aid, not a substitute for allocating time according to the question’s marks and complexity.

Fix the clock.
Mark the valuation time, the beginning and end of each year, and the time at which every premium, expense, benefit, and reserve is measured.

Write the state model.
List the states and transition probabilities for one year. Check that every row sums to one, including any absorbing state such as death.

Project probabilities and cash flows.
Use the transition model to find the probability in each state at each duration. Apply each payment only to the transitions or states specified.

Calculate reserves by state.
State whether a reserve is just before or just after the cash flows at that time. Use the stated valuation interest and cash-flow assumptions.

Build and check the profit signature.
Calculate each period’s profit using the reserve movement and cash flows. Discount profits to the valuation time and reconcile with direct cash flows.

A useful time allocation is about one-fifth of the available time for the timeline and assumptions, one-fifth for probabilities, two-fifths for cash flows and reserves, and the balance for the profit signature, NPV, and checks. Do not start by plugging figures into a profit formula before deciding what “at time $t$” means.

State Probabilities and Benefit Timing

In a discrete-time Markov model, the probability of the next state depends on the current state, not on the earlier path, conditional on that current state. Let $p_{jk}$ be the probability of moving from state $j$ at the start of a year to state $k$ at the end. Each row of the transition matrix must sum to one. A benefit specified on a transition is paid when that transition occurs; a benefit specified for a state at year-end is paid to policies that occupy that state at year-end.

Worked Example: Projecting Disabled Lives and Transition Benefits

A hypothetical contract has active, disabled, and dead states, denoted $A$, $D$, and $†$. At each year-end, the transition probabilities are $A\to A:0.7$, $A\to D:0.2$, $A\to †:0.1$, $D\to A:0.1$, $D\to D:0.8$, and $D\to †:0.1$. Death is absorbing. The contract pays $30 on a transition to death and $8 to anyone in state $D$ at year-end. A policy is active at issue. Find the state probabilities after two years and the expected benefits at each year-end.

Method. First project the state distribution, then apply the payment rule to each transition or ending state. The first-year probabilities are $P(A)=0.7$, $P(D)=0.2$, and $P(†)=0.1$.

At the second year-end, the probability of being active is $0.7(0.7)+0.2(0.1)=0.51$. The probability of being disabled is $0.7(0.2)+0.2(0.8)=0.30$. The remaining probability is death: $0.1+0.7(0.1)+0.2(0.1)=0.19$. The probabilities sum to $0.51+0.30+0.19=1.00$.

In year one, expected death benefits are $0.1(30)=\$3.00$, and the year-end disability benefit is $0.2(8)=\$1.60$. Total expected benefits are $\$4.60$.

For year two, conditional on starting active, expected benefits are $0.1(30)+0.2(8)=\$4.60$. Conditional on starting disabled, they are $0.1(30)+0.8(8)=\$9.40$. The year-two starting probabilities are 0.7 active and 0.2 disabled. Thus expected year-two benefits are $0.7(4.60)+0.2(9.40)=\$3.22+\$1.88=\$5.10$. The dead state contributes no further benefits.

The check is that the year-two calculation conditions on the state at the start of year two, while the disability payment is triggered by the state at the end of that year. A common mistake avoided here is paying the disability benefit only on $A\to D$ transitions and omitting $D\to D$.

Prospective Reserves by State

A reserve is not fully specified until its timing and conditioning state are clear. In the following examples, $V_t^j$ is the prospective reserve at time $t$, conditional on state $j$, just before the cash flows at the beginning of the next policy year. It is the present value of future benefits and expenses minus future premiums, using the stated valuation assumptions. A negative net reserve is possible when future premiums exceed expected future outgo.

For a one-year step with beginning-of-year net outgo $C_j$ (expenses less premiums), end-of-year transition benefit $B_{jk}$, annual effective interest rate $i$, and next-time reserve $V_{t+1}^k$, the recursion is:

V_t^j=C_j+\frac{1}{1+i}\sum_k p_{jk}\left(B_{jk}+V_{t+1}^k\right).

Here $B_{jk}$ is the payment associated with the transition from $j$ to $k$; if a payment depends on the ending state, it is included for every transition ending in that state. At the terminal duration, set the reserve to zero if no future benefits or expenses remain. The formula does not apply unchanged if cash-flow timing differs, such as a premium paid at year-end; that timing must be represented explicitly.

Worked Example: Calculating Active and Disabled Reserves

Use a hypothetical two-year contract with states active $A$, disabled $D$, and dead. For each year, an active policy pays a premium of $6 and incurs an expense of $2 at the beginning of the year; a disabled policy pays no premium and incurs an expense of $1. At year-end, $20 is paid on death and $4 is paid to each policy ending the year disabled. From $A$, the transition probabilities are $0.8$ to $A$, $0.1$ to $D$, and $0.1$ to death. From $D$, they are $0.2$ to $A$, $0.7$ to $D$, and $0.1$ to death. The annual effective interest rate is $i=0.05$. Calculate the reserve after year one, before the second year’s beginning cash flows.

Method. Calculate the expected net outgo separately for each state. The terminal reserve after year two is zero. For an active policy, second-year beginning net outgo is expense less premium, $2-6=-4$. Expected year-end benefits are $0.1(20)+0.1(4)=\$2.40$. Therefore:

V_1^A=-4+\frac{2.40}{1.05} =-4+\frac{16}{7} =-\frac{12}{7} =-\$1.71.

For a disabled policy, beginning net outgo is $1. Expected benefits are $0.1(20)+0.7(4)=\$4.80$; the transition to active has no year-end benefit. Therefore:

V_1^D=1+\frac{4.80}{1.05} =1+\frac{32}{7} =\frac{39}{7} =\$5.57.

The displayed dollar amounts are rounded to the nearest cent only at the final step; the exact fractions are used in later calculations. The active reserve is negative because the future $6 premium exceeds the active-state expected benefits and expense on a present-value basis. This is consistent with the stated net-reserve definition, not a sign error.

From Reserves to Profit Signature

For a discrete annual profit test, take $V_{t-1}^j$ as the reserve at the start of a year before its beginning cash flows. Let $N_j$ be beginning-of-year net cash flow, premiums less expenses, and let $C$ be the end-of-year benefit. If the policy survives in state $k$, the reserve carried forward is $V_t^k$. The period-end profit conditional on starting state $j$ is:

\Pi_t^j=(V_{t-1}^j+N_j)(1+i)-C-V_t^k.

For multiple possible next states, expected profit conditional on starting state $j$ is $(V_{t-1}^j+N_j)(1+i)-\sum_k p_{jk}(B_{jk}+V_t^k)$. The profit signature per policy at issue is found by weighting those conditional profits by the probabilities of being in each starting state at that duration. Be explicit about whether an initial reserve is given or calculated: setting it to zero is not the same as using a prospective reserve at issue.

Worked Example: Constructing a Profit Signature and Checking NPV

Continue the two-year contract and assumptions of the reserve example. All policies are active at issue, the initial reserve is zero, and the actual transition, benefit, expense, and interest assumptions equal the valuation assumptions. Calculate expected profits at the ends of years one and two, then calculate and reconcile the NPV at issue. Report dollar amounts to the nearest cent.

Year one. The beginning net cash flow is premium less expense, $6-2=\$4$. Expected year-one benefits are $\$2.40$, and the year-one ending-state probabilities are 0.8 active, 0.1 disabled, and 0.1 dead. The expected reserve carried forward is:

E[V_1]=0.8\left(-\frac{12}{7}\right)+0.1\left(\frac{39}{7}\right)+0.1(0) =-\frac{57}{70}.

Thus the expected profit at the end of year one is:

\Pi_1=4(1.05)-2.40-\left(-\frac{57}{70}\right) =\frac{183}{70} =\$2.61.

Year two. At the start of year two, the probabilities of being active and disabled are 0.8 and 0.1. Expected beginning net cash flow is $0.8(4)+0.1(-1)=\$3.10$. The expected reserve at that time is $-57/70$, so the expected amount accumulated to year-end is $\left(3.10-57/70\right)(1.05)=\$2.40$. Expected year-two benefits are $0.8(2.40)+0.1(4.80)=\$2.40$. There is no terminal reserve, so $\Pi_2=2.40-2.40=\$0.00$. The profit signature per policy at issue is therefore $(\$2.61,\ \$0.00)$.

Discount the profit signature at $i=0.05$, so $v=1/1.05=20/21$. The NPV is:

\operatorname{NPV}=\frac{183}{70}\frac{20}{21} =\frac{122}{49} =\$2.49.

Reconciliation. Directly discount the net cash flows: the issue-time net cash flow is $4; at time one, expected net cash flow is $3.10-2.40=\$0.70$; and at time two, expected benefits are $2.40. Hence:

4+\frac{0.70}{1.05}-\frac{2.40}{(1.05)^2} =4+\frac{2}{3}-\frac{320}{147} =\frac{122}{49} =\$2.49.

The two methods agree. The reserve transfers profit between durations but, with consistent assumptions and a zero terminal reserve, does not change the NPV of the underlying timed cash flows. The key mistake avoided is discounting a cash flow from time two as though it occurred at time one.

Deriving the NPV Reconciliation

The reconciliation follows from the reserve movement, rather than being an unrelated check. Let $N_t$ be the net cash flow at the beginning of period $t$, $C_t$ the end-of-period benefit, and $V_t$ the reserve at the end of the period. In expectation over states, the profit at period end can be written:

\Pi_t=(V_{t-1}+N_t)(1+i)-C_t-V_t.

Multiply by $v^t$, where $v=1/(1+i)$, and sum from $t=1$ to $n$. The reserve terms cancel between adjacent periods: the $V_t$ released at the end of one period is the same reserve carried into the next, with its accumulated interest. This gives:

\sum_{t=1}^{n}v^t\Pi_t =V_0+\sum_{t=1}^{n}v^{t-1}N_t -\sum_{t=1}^{n}v^t C_t-v^nV_n.

This identity assumes that each profit and reserve uses the same cash-flow timing and interest basis. If the reserve basis differs from the experience basis, or if a reserve strengthening is introduced, the reserve movement need not cancel against the same expected cash flows; the difference is part of the profit result. Always check the stated bases before interpreting the reconciliation.

Common Timing Trap and Final Checks

Suppose a hypothetical profit vector has a loss of $5 at the end of year one and a profit of $22 at the end of year two, with $i=0.04$. Discounting both amounts by one year gives the incorrect calculation:

\frac{22}{1.04}-\frac{5}{1.04} =\$16.35 \quad\text{(incorrect: both cash flows are treated as year-one amounts).}

The correct NPV places each amount at its own time:

-\frac{5}{1.04}+\frac{22}{(1.04)^2} =-4.807692\ldots+20.340237\ldots =\$15.53.

The incorrect expression’s arithmetic is $16.35 to the nearest cent, but the error is its timing, not its arithmetic. Before submitting an integrated calculation, check that transition probabilities sum to one, payments match their triggers, reserves are conditional on the correct state and time, and each profit is discounted from its actual duration.

Key takeaway: Fix the timeline first; then project states, calculate cash flows and state reserves, form the profit signature, and reconcile its discounted value to direct cash flows. A consistent sequence makes timing and conditioning errors easier to detect.

Check Your Understanding

Work through each problem by stating the relevant timing and state conditioning before calculating.

A hypothetical model starts active. Its transitions are $A\to A:0.7$, $A\to D:0.2$, $A\to †:0.1$, $D\to A:0.1$, $D\to D:0.8$, and $D\to †:0.1$, with death absorbing. Find the probabilities of active, disabled, and dead after two years. (Answer: 0.51 active, 0.30 disabled, 0.19 dead.)
At $i=0.04$, a one-year reserve conditional on a healthy state has a beginning expense of $2 and premium of $5. At year-end, death probability is 0.05 with benefit $30, disability probability is 0.15 with benefit $8, and all other outcomes have no benefit; the terminal reserve is zero. Calculate the reserve before the beginning cash flow. (Answer: $-\$0.40$, to the nearest cent.)
A policy has a reserve of $10 at the start of a year, a beginning net cash inflow of $3, end-year expected benefit of $8, and end-year reserve of $5. At $i=0.05$, calculate the expected end-year profit. (Answer: $(10+3)(1.05)-8-5=\$0.65$.)
A profit signature has a loss of $4 at the end of year one and a profit of $12 at the end of year two. At $i=0.05$, calculate its NPV at issue. (Answer: $-4/1.05+12/(1.05)^2=\$7.98$, to the nearest cent.)
A model has expected end-year profit $\$2.00$ in year one and $\$3.00$ in year two, with $i=0.05$. A candidate discounts both by one year. Give the correct NPV and identify the timing error. (Answer: $2/1.05+3/(1.05)^2=\$4.63$, to the nearest cent; the year-two profit must be discounted for two years.)

What You'll Learn