Conditioning Cash-Flow Values on Future States

Why Condition on a Future State?

A state-contingent contract may pay at several future dates, with each payment depending on the state occupied at that date. One direct approach is to calculate the probability of each payment event and value each payment separately. Another is to choose an intermediate time, calculate the probability of being in each state at that time, and then calculate the expected value of the remaining cash flows conditional on each state. The second approach can make a long calculation easier to organize and check.

The key is to keep two different quantities separate: the probability of reaching a state and the conditional value of future cash flows given that state. The conditional value is not itself a probability, and it must be measured at a clearly specified time so that discounting is applied correctly.

Definition: Let $X_k$ be the state occupied at time $k$, and let $C_j$ be the cash flow paid at time $j$. At annual effective interest rate $i$, write $v=1/(1+i)$. For cash flows strictly after time $k$, define the value at time $k$, conditional on state $s$, by $$ V_k(s)=E\left[\left.\sum_{j=k+1}^{n}v^{j-k}C_j\right|X_k=s\right]. $$ This is a conditional expected present value at time $k$, not at time 0. If the state at time $k$ has probabilities $p_k(s)=P(X_k=s)$, the expected value at time 0 of those later cash flows is $$ v^k\sum_s p_k(s)V_k(s). $$ Any cash flows paid at or before time $k$ must be valued separately at their own payment dates.

In the examples, the state at the conditioning time is sufficient to determine the relevant future transition probabilities and payment rules. This is the usual state-based Markov modelling assumption. If a payment rule depends on information about the past that is not captured by the current state, the state description must be expanded before this conditioning method can be applied using state alone.

The Conditioning Identity

The calculation follows from the law of total expectation. At time $k$, the events $X_k=s$, over all possible states $s$, are mutually exclusive and collectively exhaustive. Therefore, for the future value $Z_k=\sum_{j=k+1}^{n}v^{j-k}C_j$,

E[Z_k]=\sum_s E[Z_k\mid X_k=s]P(X_k=s) =\sum_s V_k(s)p_k(s).

The expression on the right is an expectation at time $k$. Multiplying by $v^k$ converts it to time 0. This gives a useful calculation sequence: determine the state probabilities at the conditioning time, determine the conditional values from that time forward, weight those values by the state probabilities, and discount the resulting time-$k$ value to time 0.

Choose the conditioning time.
State which future state will be used to divide the possible outcomes.

Find the state probabilities.
Use the transition information to calculate $P(X_k=s)$ for every relevant state.

Value future cash flows conditionally.
For each state $s$, start at time $k$, use the transitions from $s$, and discount each future payment back to time $k$.

Average and discount.
Weight each conditional value by $P(X_k=s)$, then discount the time-$k$ result to time 0.

Worked Applications

Worked Example: One Later Payment and a One-Year Conditional Value

A hypothetical contract has states $A$, $B$, and $X$. At time 1, the state probabilities are $P(X_1=A)=0.50$, $P(X_1=B)=0.40$, and $P(X_1=X)=0.10$. State $X$ is absorbing. At time 2, the contract pays \$300 if the state is $B$, and pays nothing otherwise. Given state $A$ at time 1, the probability of $B$ at time 2 is 0.20; given $B$ at time 1, that probability is 0.50. The annual effective interest rate is $i=4\%$. Find the time-0 expected present value.

Method. Condition on the state at time 1. First calculate the value at time 1 conditional on each state, then average those values using the time-1 state probabilities and discount one year.

Since $v=1/1.04$, the conditional time-1 value in state $A$ is the expected time-2 payment discounted one year:

V_1(A)=\frac{0.20(300)}{1.04} =\frac{60}{1.04} =57.6923.

For state $B$, the corresponding calculation is:

V_1(B)=\frac{0.50(300)}{1.04} =\frac{150}{1.04} =144.2308.

State $X$ is absorbing and has no payment, so $V_1(X)=0$. The expected value at time 1 is therefore

$$ 0.50(57.6923)+0.40(144.2308)+0.10(0) =28.8462+57.6923 =86.5385. $$

Discounting this value one year to time 0 gives

\frac{86.5385}{1.04}=83.2101.

Rounded to the nearest cent, the time-0 expected present value is \$83.21. As a check, the probability of state $B$ at time 2 is $0.50(0.20)+0.40(0.50)+0.10(0)=0.30$. Direct valuation gives $300(0.30)/(1.04)^2=\$83.21$, agreeing with the conditional calculation.

Common mistake avoided: The payment is at time 2, so it must be discounted for two years to time 0. Using only one year of discounting would value it at time 1, not time 0.

Worked Example: Including a Payment at the Conditioning Time

A hypothetical contract has states $A$, $B$, and absorbing state $X$. At time 1, their probabilities are 0.60, 0.30, and 0.10, respectively. A \$80 payment is made at time 1 if the state is $B$. From time 1 to time 2, the probability of $B$ is 0.30 given $A$, 0.60 given $B$, and 0 given $X$. A further \$250 payment is made at time 2 if the state is $B$. At $i=5\%$, find the time-0 expected present value.

Method. The definition of $V_1(s)$ above covers payments strictly after time 1. The time-1 payment must therefore be valued separately, while the time-2 payment is valued conditionally from time 1.

The expected time-1 payment is $0.30(80)=\$24$, with time-0 value

\frac{24}{1.05}=22.8571.

For the time-2 payment, the time-1 conditional values are $250(0.30)/1.05=\$71.4286$ in state $A$, $250(0.60)/1.05=\$142.8571$ in state $B$, and zero in $X$. Weighting by the time-1 probabilities gives a conditional-value average of

$$ 0.60(71.4286)+0.30(142.8571)+0.10(0) =42.8572+42.8571 =85.7143. $$

This is the expected time-1 value of the time-2 payment. Its time-0 value is $85.7143/1.05=\$81.6327$. Adding the two time-0 values gives

$$ 22.8571+81.6327=104.4898. $$

The time-0 expected present value is \$104.49, rounded to the nearest cent. Check by direct probabilities: $P(X_1=B)=0.30$, while $P(X_2=B)=0.60(0.30)+0.30(0.60)+0.10(0)=0.36$. Thus the value is $80(0.30)/1.05+250(0.36)/(1.05)^2=\$104.49$.

Common mistake avoided: A conditional value starting at time 1 does not automatically include a payment already made at time 1. Separating the time-1 payment prevents either omitting it or discounting it as if it were paid later.

Worked Example: Conditioning Before Two Future Payment Dates

At time 1, a hypothetical contract has probabilities 0.50 in state $A$, 0.40 in state $B$, and 0.10 in absorbing state $X$. The transitions from time 1 to time 2 are: from $A$, probabilities 0.50 to $A$, 0.30 to $B$, and 0.20 to $X$; from $B$, probabilities 0.25 to $A$, 0.50 to $B$, and 0.25 to $X$. At time 2, the contract pays \$120 if in $B$. From time 2 to time 3, the probability of $B$ is 0.20 given $A$, 0.40 given $B$, and 0 given $X$. At time 3, it pays \$200 if in $B$. There are no other payments, and $i=5\%$. Find the time-0 value.

Method. Condition on the time-2 state, calculate the value of the time-2 and time-3 payments from each state, then weight by the probability of each time-2 state.

First find the time-2 probabilities. The probability of $A$ is $0.50(0.50)+0.40(0.25)=0.25+0.10=0.35$. The probability of $B$ is $0.50(0.30)+0.40(0.50)=0.15+0.20=0.35$. The remaining probability is $0.50(0.20)+0.40(0.25)+0.10=0.10+0.10+0.10=0.30$ in $X$. These probabilities sum to 1.

At time 2, the conditional value of future payments in state $A$ is the expected time-3 payment discounted one year:

V_2(A)=\frac{0.20(200)}{1.05}=38.0952.

In state $B$, the \$120 payment is made immediately at time 2, and the expected time-3 payment is $0.40(200)=\$80$:

V_2(B)=120+\frac{80}{1.05} =120+76.1905 =196.1905.

There are no payments from absorbing state $X$, so $V_2(X)=0$. The expected value at time 2 is

$$ 0.35(38.0952)+0.35(196.1905)+0.30(0) =13.3333+68.6667 =82.0000. $$

Discount two years to time 0:

\frac{82}{(1.05)^2} =\frac{82}{1.1025} =74.3764.

The time-0 value is \$74.38, rounded to the nearest cent. A direct check gives $P(X_2=B)=0.35$ and $P(X_3=B)=0.35(0.20)+0.35(0.40)+0.30(0)=0.21$. Hence $120(0.35)/(1.05)^2+200(0.21)/(1.05)^3=\$74.38$.

Common mistake avoided: The \$120 payment is at time 2, so it belongs in $V_2(B)$ without an extra one-year discount. The time-3 payment is discounted one year within $V_2$, and the full time-2 value is then discounted two years to time 0.

Checks, Assumptions, and Exam Use

Conditioning is especially useful when many possible histories lead to the same state at a chosen time. Instead of retaining every path separately, the calculation groups paths by their state at that time. If the future payment rules and transition probabilities depend only on the current state, all paths in the same state can be summarized by one conditional value.

Before using the identity, check that the state probabilities cover all possible states and sum to 1, and that the transition probabilities needed for later payments are specified. A conditional probability for one destination state is not enough to determine all future values unless the omitted transitions are irrelevant to every later payment. Also check that the time basis is consistent: a value calculated at time $k$ is discounted by $v^k$ to reach time 0, while payments after $k$ are first discounted back to time $k$.

This conditioning method is aligned with the SOA Exam ALTAM (Advanced Long-Term Actuarial Mathematics) syllabus. It is a general organization principle for state-contingent valuation: it does not change the contract’s payments or probabilities; it rearranges the expectation into a weighted sum of conditional values.

Key takeaway: For future cash flows after time $k$, calculate a value $V_k(s)$ conditional on each state $s$, average using the probabilities $P(X_k=s)$, and then discount the result from time $k$ to time 0. Keep cash flows at time $k$ separate unless the definition of the conditional value explicitly includes them.

Check Your Understanding

For each problem, identify the relevant state probabilities and keep the payment dates explicit.

At time 1, a hypothetical contract has probabilities 0.70 in state $U$ and 0.30 in state $V$. At time 2 it pays \$500 if in $V$. The probabilities of $V$ at time 2 are 0.10 given $U$ and 0.60 given $V$. There are no other states or payments, and $i=0\%$. Find the time-0 expected value. (Answer: \$125.00)
At time 1, the probabilities of states $A,B,X$ are 0.40, 0.50, and 0.10. State $X$ is absorbing. At time 2, a \$240 payment is made in $B$; the probability of $B$ at time 2 is 0.25 given $A$ and 0.40 given $B$. The annual effective rate is 5%. Find the time-0 expected present value. (Answer: \$65.31)
At time 1, the probabilities of $A,B,X$ are 0.50, 0.30, and 0.20. A \$30 payment is made immediately in $A$. At time 2, a \$200 payment is made in $B$. The probability of $B$ at time 2 is 0.20 given $A$, 0.50 given $B$, and 0 given $X$. State $X$ is absorbing and $i=0\%$. Find the time-0 expected value. (Answer: \$65.00)
At time 1, probabilities of $A$ and $B$ are 0.60 and 0.40. From time 1 to time 2, the probabilities of $A,B,X$ are $0.40,0.40,0.20$ given $A$, and $0.20,0.50,0.30$ given $B$. At time 2, a \$100 payment is made in $B$. From time 2 to time 3, the probability of $B$ is 0.10 given $A$, 0.60 given $B$, and 0 given $X$, which is absorbing. A \$300 payment is made in $B$ at time 3. At $i=4\%$, find the time-0 expected present value. (Answer: \$119.62)
Suppose the time-2 probabilities of $A,B,X$ are 0.35, 0.45, and 0.20. State $X$ is absorbing. From time 2 to time 3, the probability of $B$ is 0.20 given $A$, 0.50 given $B$, and 0 given $X$. A \$160 payment is made in $B$ at time 3. Write the time-2 expected value using conditional values, then give the time-0 value if $i=0\%$. (Answer: $V_2(A)=\$32,\ V_2(B)=\$80,\ V_2(X)=\$0;\ E[V_2(X_2)]=\$47.20$, so the time-0 value is \$47.20)

What You'll Learn