Whole-Life Insurance Retrospective Reserves

Introduction

A whole-life policy can remain in force for many years, with premiums received while the insured is alive and a death benefit paid when death occurs. A retrospective reserve at a selected duration measures the accumulated net cash flow generated by the original cohort of policies, allocated among the policies still in force. The central calculation is not just an accumulation: it must also account for the fact that fewer policies survive to the valuation date.

This tutorial uses an annual discrete model. Premiums are paid at the beginning of each policy year while the insured is alive, and a death benefit is paid at the end of the year of death. A valuation at duration $k$ is made immediately before the premium due at time $k$. This timing convention makes the reserve at issue, before the first premium, zero.

Definition: For a cohort of identical whole-life policies issued at age $x$, the retrospective reserve at duration $k$, immediately before the premium at time $k$, is the accumulated net cash flow of the original cohort through time $k$, divided by the probability that a policy from that cohort survives to time $k$. It is a cohort-level accumulated fund allocated per survivor; it is not the expected accumulated net cash flow of an individual survivor’s own past premiums and claims.

The Cohort Accumulation Formula

Suppose each policy has a level annual premium $P$ and a death benefit $B$. Let $i$ be the effective annual interest rate, and let ${}_j p_x$ be the probability that a life aged $x$ survives $j$ years. Write $q_{x+j}$ for the probability of death during the year following duration $j$, conditional on survival to that duration, so ${}_{j+1}p_x={}_jp_x(1-q_{x+j})$.

Among one policy issued at time 0, the expected number of premiums paid at time $j$ is proportional to ${}_jp_x$. Those premiums accumulate for $k-j$ years to time $k$. The probability of a death claim in the year from $j$ to $j+1$ is ${}_jp_xq_{x+j}$. Since that claim is paid at time $j+1$, it accumulates for $k-j-1$ years to time $k$.

{}_kV_x= \frac{ P\displaystyle\sum_{j=0}^{k-1}{}_jp_x(1+i)^{k-j} - B\displaystyle\sum_{j=0}^{k-1}{}_jp_xq_{x+j}(1+i)^{k-j-1} }{{}_kp_x}.

The numerator is the accumulated cohort premium income less accumulated death claims through duration $k$. Dividing by ${}_kp_x$ allocates that net amount among the fraction of the original cohort still in force. For $k=0$, both sums are empty and ${}_0V_x=0$. The formula requires ${}_kp_x>0$; if no policies survive to duration $k$, there is no surviving cohort to which a reserve can be allocated.

Timing: The displayed formula values the reserve immediately before the premium at time $k$. It includes premiums at times $0,\ldots,k-1$, and claims paid at times $1,\ldots,k$. Immediately after the premium at time $k$, the reserve is larger by $P$, under the same cohort allocation and with no other cash flow at that instant.

Why Survival Conditioning Appears

The division by ${}_kp_x$ follows directly from tracking a cohort. Imagine issuing $N$ identical policies. The expected number surviving to duration $k$ is $N{}_kp_x$. The expected accumulated premiums less claims generated by the original cohort are $N$ times the numerator in the formula. Dividing the accumulated cohort fund by the expected number of survivors gives the displayed reserve, since $N$ cancels.

This is different from conditioning on one individual surviving to time $k$ and then accumulating only that individual’s past cash flows. A policy that died before $k$ did not survive, but its past premiums and its death claim still contributed to the original cohort’s accumulated fund. The cohort calculation retains those cash flows before allocating the net fund among survivors.

Find cohort probabilities.
For each policy year, calculate the probability alive at the premium date and the probability of death during that year.

Move cash flows to the valuation time.
Accumulate each premium from its payment date and each claim from its payment date using the stated interest rate.

Net and allocate.
Subtract accumulated claims from accumulated premiums, then divide by the probability of survival to the valuation duration.

Worked Calculations

The following examples use hypothetical whole-life contracts. Mortality probabilities are specified for every year relevant to the calculation; a final probability of death equal to 1 is used in some examples simply to close the hypothetical lifetime model.

Worked Example: A Net Premium and a Duration-Two Reserve

A hypothetical whole-life policy has benefit $B=\$29{,}800$, paid at the end of the year of death. The annual effective interest rate is $i=0\%$. Conditional death probabilities by policy year are $q_x=0.10$, $q_{x+1}=0.20$, $q_{x+2}=0.50$, and $q_{x+3}=1$. Premiums are level and paid annually in advance while the insured is alive. Find the annual net premium and the reserve at duration 2 immediately before that year’s premium.

Method. First calculate the whole-life net premium by equating the expected present value of premiums and benefits. Then apply the cohort accumulation formula through duration 2.

The survival probabilities at durations 0 through 3 are $1$, $0.90$, $0.90(0.80)=0.72$, and $0.72(0.50)=0.36$. The death probabilities by year are $0.10$, $0.90(0.20)=0.18$, $0.72(0.50)=0.36$, and $0.36$. They sum to $1$, as required for this mortality model.

At zero interest, the whole-life insurance present value per unit benefit is $A_x=1$, because death occurs with total probability 1 and every payment has discount factor 1. The whole-life annuity-due value is $\ddot{a}_x=1+0.90+0.72+0.36=2.98$. By the net premium equivalence principle,

P=\frac{B A_x}{\ddot{a}_x} =\frac{29{,}800(1)}{2.98} =\$10{,}000.

By duration 2, the time-0 premium has accumulated for two years and the time-1 premium for one year; with $i=0\%$, the accumulation factors are both 1. The expected accumulated premium income per original policy is $10{,}000(1+0.90)=\$19{,}000$. Claims paid in years 1 and 2 have probabilities $0.10$ and $0.18$, so expected accumulated claims are $29{,}800(0.10+0.18)=\$8{,}344$. Thus,

{}_2V_x= \frac{19{,}000-8{,}344}{0.72} =\frac{10{,}656}{0.72} =\$14{,}800.00.

Check. At duration 2, the expected present value of the remaining benefit is $\$29{,}800$, because death is certain eventually and interest is zero. Expected future premiums are $10{,}000(1+0.50)=\$15{,}000$. Their difference is $\$14{,}800$, agreeing with the retrospective calculation. The check avoids treating the reserve as only the premiums paid by a surviving individual.

Worked Example: Interest and Claims Paid Before Duration Three

A hypothetical whole-life contract has annual premium $P=\$3{,}000$, death benefit $B=\$10{,}000$, and effective annual interest rate $i=10\%$. The conditional death probabilities are $0.10$, $0.20$, $0.50$, and $1$ in policy years 1 through 4. Find the reserve at duration 3 immediately before the premium due then.

Method. Calculate survival and death probabilities year by year, accumulate each cash flow to time 3, and divide the cohort net amount by the duration-3 survival probability.

The survival probabilities at durations 0, 1, 2, and 3 are $1$, $0.90$, $0.90(0.80)=0.72$, and $0.72(0.50)=0.36$. Therefore ${}_3p_x=0.36$. Death probabilities in years 1, 2, and 3 are $0.10$, $0.90(0.20)=0.18$, and $0.72(0.50)=0.36$.

The time-0, time-1, and time-2 premiums accumulate to time 3 with factors $1.1^3=1.331$, $1.1^2=1.21$, and $1.1$. Their expected accumulated value per original policy is

3{,}000\left(1.331+0.90(1.21)+0.72(1.1)\right) =3{,}000(1.331+1.089+0.792) =\$9{,}636.

The year-1 claim is paid at time 1 and accumulates for two years; the year-2 claim accumulates for one year; the year-3 claim is paid at time 3 and does not accumulate further. Expected accumulated claims are

10{,}000\left(0.10(1.21)+0.18(1.1)+0.36\right) =10{,}000(0.121+0.198+0.36) =\$6{,}790.

The accumulated cohort net amount is $\$9{,}636-\$6{,}790=\$2{,}846$. Allocating it among the $0.36$ expected fraction of policies still in force gives

{}_3V_x=\frac{2{,}846}{0.36}=\$7{,}905.56.

Check. The result is positive because accumulated premium income exceeds accumulated claims for the original cohort. The year-3 claim has accumulation factor 1, not 1.1: it is paid exactly at the valuation time.

Worked Example: A Reserve with a Different Valuation Duration

A hypothetical whole-life contract has $P=\$15{,}000$, $B=\$50{,}000$, and $i=5\%$. The conditional death probabilities in policy years 1 through 4 are $0.20$, $0.25$, $0.40$, and $1$. Find the reserve at duration 2 immediately before the premium at time 2.

Method. Use the cohort formula at $k=2$, carefully matching each accumulation factor to the cash-flow date.

Survival to duration 1 is $0.80$, and survival to duration 2 is $0.80(0.75)=0.60$. The probability of a year-1 death is $0.20$; the probability of a year-2 death is $0.80(0.25)=0.20$.

The time-0 premium accumulates two years, and the time-1 premium is paid by the $0.80$ of policies alive then and accumulates one year:

15{,}000\left(1.05^2+0.80(1.05)\right) =15{,}000(1.1025+0.84) =\$29{,}137.50.

The year-1 claims accumulate one year to time 2; year-2 claims are paid at time 2. Therefore,

50{,}000\left(0.20(1.05)+0.20\right) =50{,}000(0.21+0.20) =\$20{,}500.

The cohort net amount is $\$29{,}137.50-\$20{,}500=\$8{,}637.50$. Dividing by the survival probability gives

{}_2V_x=\frac{8{,}637.50}{0.60}=\$14{,}395.83.

Check. Premiums and claims paid at time 2 have no accumulation beyond that date. Applying an additional factor of $1.05$ to the year-2 claims would value them at time 3 instead.

Recursion and Common Traps

The accumulated cohort formula also gives a useful one-year check. Suppose ${}_kV_x$ is measured immediately before the premium at time $k$. At time $k$, the surviving cohort receives premium $P$, and the combined amount earns interest through the year. At time $k+1$, the fraction $q_{x+k}$ dies and receives benefit $B$; the remaining fraction is $p_{x+k}=1-q_{x+k}$. Thus, when $p_{x+k}>0$,

{}_{k+1}V_x = \frac{\left({}_kV_x+P\right)(1+i)-q_{x+k}B}{p_{x+k}}.

The recursion is a check on timing and survival allocation: add the premium before interest, subtract claims paid at the end of the year, then divide by the fraction that survives. For example, in the second worked example, the duration-1 reserve is $(3{,}000(1.1)-0.10(10{,}000))/0.90=\$2{,}555.56$. The recursion then gives duration 2 as $((2{,}555.56+3{,}000)(1.1)-0.20(10{,}000))/0.80=\$5{,}138.89$; applying it once more gives duration 3 as $((5{,}138.89+3{,}000)(1.1)-0.50(10{,}000))/0.50=\$7{,}905.56$, matching the direct calculation up to rounding.

Common mistake. A tempting but incorrect approach is to condition on survival to duration $k$, count only the premiums that the eventual survivor personally paid, and omit claims on policies that died earlier. That is not the cohort retrospective reserve. For example, in the first worked example, that mistaken calculation would count premiums of $\$10{,}000$ at time 0 and $\$10{,}000$ at time 1, with no earlier claim for the survivor, and report $\$20{,}000$. The correct cohort calculation includes expected claims of $\$8{,}344$ from the original group and allocates the net $\$10{,}656$ across the $0.72$ surviving fraction, giving $\$14{,}800$.

A second common error is shifting a cash flow by one year. Premiums are paid at the start of a year; claims are paid at its end. A claim in the final year before valuation has no further accumulation, while the premium at the start of that year accumulates for a full year. Keep the valuation instant explicit, and do not include the premium due at time $k$ in a reserve measured immediately before it.

Key takeaway: For a whole-life retrospective reserve, accumulate the original cohort’s premiums and death claims to the valuation date, subtract claims from premiums, and divide by the fraction surviving to that date. State the premium and claim timing before calculating.

Check Your Understanding

For each hypothetical contract, use the annual discrete model and value the reserve immediately before the premium due at the stated duration. Round money to the nearest cent.

A whole-life policy has $P=\$6{,}000$, $B=\$20{,}000$, and $i=0\%$. The first two annual conditional death probabilities are $0.20$ and $0.25$. Find ${}_2V_x$. (Answer: $\$4{,}666.67$)
A whole-life policy has $P=\$4{,}000$, $B=\$15{,}000$, and $i=10\%$. The first two conditional death probabilities are $0.10$ and $0.20$. Find ${}_2V_x$. (Answer: $\$6{,}180.56$)
A whole-life policy has $P=\$5{,}000$, $B=\$12{,}000$, and $i=4\%$. The first-year conditional death probability is $0.25$. Find ${}_1V_x$. (Answer: $\$2{,}933.33$)
A cohort begins with 1,000 policies. There are 800 survivors at time 1 and 600 at time 2; 200 policies die in each of the first two years. Each policy has $P=\$1{,}000$, $B=\$4{,}000$, and $i=5\%$. Find the duration-2 reserve per survivor. (Answer: $\$504.17$)
A whole-life policy has $B=\$12{,}000$, $i=0\%$, and conditional death probabilities $0.50$ in year 1 and $1$ in year 2. Find the level net annual premium by equivalence, then find ${}_1V_x$. (Answer: Net premium $\$8{,}000$; ${}_1V_x=\$4{,}000.00$)

What You'll Learn