Continuous-Time Retrospective Reserve Accumulation

Why continuous-time accumulation needs survival weighting

A retrospective reserve measures the accumulated value of past cash flows, viewed at a valuation time. In a continuous-time contract, premiums may arrive as a rate while the policy is in force, and a death benefit may be paid at the instant of death. The calculation must therefore account for both the timing of each cash flow and the chance that a policy is still in force when that cash flow occurs.

The reserve is stated per policy surviving to the valuation time. To obtain it, first calculate accumulated net cash flows for the original group of policies, allowing for the fact that not all policies survive to that time. Then divide by the proportion that does survive. This conditioning step is central to the formula.

Definition: Consider a policy issued at age $x$, with no initial reserve. Let $V_t^R$ denote its retrospective reserve at time $t$, conditional on survival to time $t$. Premiums are paid continuously at rate $p(s)$ per year while the policy is in force. A death at time $s$ produces an immediate benefit of $b(s)$. The mortality force at duration $s$ is $\mu_{x+s}$, and the force of interest is $\delta(r)$ at time $r$. With no expenses or other cash flows, $$ V_t^R= \frac{\displaystyle\int_0^t {}_sp_x \left[p(s)-b(s)\mu_{x+s}\right] \exp\!\left(\int_s^t \delta(r)\,dr\right) ds} {{}_tp_x}. $$ Here ${}_sp_x$ is the probability of surviving from issue to time $s$, and ${}_tp_x$ is the probability of surviving to time $t$. If the force of interest is constant, the accumulation factor is $e^{\delta(t-s)}$.

In this expression, $p(s)$ is a premium rate, not a single premium. Likewise, $b(s)\mu_{x+s}$ is the expected death-benefit outflow rate conditional on the policy having survived to time $s$: the benefit is paid only if death occurs, and the probability of death over a short interval of length $ds$, conditional on survival to its start, is approximately $\mu_{x+s}ds$.

Why the integral formula holds

Imagine a large hypothetical group of identical policies, all issued at the same time. The group makes the cohort interpretation precise: the cash flows are first accumulated for the original group and then divided by the number of policies still in force at time $t$. This is the continuous-time counterpart of a retrospective cash-flow balance.

Count premium income at time $s$.
For each original policy, the probability of being alive at $s$ is ${}_sp_x$. Thus the expected premium income rate per original policy is ${}_sp_xp(s)$.

Count death claims at time $s$.
The probability density of death at $s$ is ${}_sp_x\mu_{x+s}$. The expected benefit outflow rate per original policy is therefore ${}_sp_x\mu_{x+s}b(s)$.

Accumulate each net cash flow to $t$.
A dollar received at $s$ accumulates by the factor $\exp(\int_s^t\delta(r)\,dr)$. Integrating the weighted premium income less claims over $0\leq s\leq t$ gives accumulated net cash flow per original policy.

Condition on survival to $t$.
The proportion of original policies still in force is ${}_tp_x$. Dividing the accumulated cohort amount by this proportion gives the amount per policy surviving to $t$, which is $V_t^R$.

The argument uses expected cohort cash flows, or equivalently the limiting proportions in a large group. It does not say that every individual policy has received the same past cash flows: policies that died have made claims and no longer contribute premiums. If interest or mortality is discontinuous at isolated times, the integral formulation remains valid when those points do not create separate lump-sum cash flows.

Level premium and benefit under constant forces

A useful special case has a constant mortality force $\mu$, a constant force of interest $\delta$, a level continuous premium rate $P$, and a level death benefit $B$ paid immediately on death. In this model, ${}_sp_x=e^{-\mu s}$, and the death density at duration $s$ is $\mu e^{-\mu s}$. Substitution into the general formula gives

V_t^R =\frac{1}{{}_tp_x}\int_0^t e^{-\mu s}(P-B\mu)e^{\delta(t-s)}\,ds =\frac{P-B\mu}{\mu+\delta} \left(e^{(\mu+\delta)t}-1\right),

when $\mu+\delta\ne 0$. The middle expression shows the two distinct ingredients of the expected cash-flow rate: premium income $P$ and expected claim outgo $B\mu$. For positive $\mu+\delta$, the last equality follows by integrating $e^{-(\mu+\delta)s}$ from 0 to $t$ and dividing by $e^{-\mu t}$. If $\mu+\delta=0$, the integral form should be used directly; the corresponding limit is $(P-B\mu)t$.

Worked Example: Level Continuous Premium and Immediate Death Benefit

A hypothetical whole-life contract has a continuous premium rate of $P=\$100$ per year and pays $B=\$3{,}000$ immediately on death. Assume a constant mortality force $\mu=0.02$ per year and a constant force of interest $\delta=0.03$ per year. Find the retrospective reserve per surviving policy at duration $t=2$ years.

Method. Use the constant-force formula. The expected claim outgo rate per policy in force is $B\mu=3{,}000(0.02)=\$60$ per year. Thus the net cash-flow rate used in the accumulated balance is $P-B\mu=100-60=\$40$ per year. Also, $\mu+\delta=0.02+0.03=0.05$.

Substituting these values,

V_2^R =\frac{40}{0.05}\left(e^{0.05(2)}-1\right) =800\left(e^{0.1}-1\right).

Using $e^{0.1}=1.105170$, the reserve is $800(1.105170-1)=800(0.105170)=\$84.136$, or \$84.14 to the nearest cent. The positive result is consistent with the assumed premium rate being greater than the expected claim outgo rate $B\mu$. The common mistake avoided here is subtracting the benefit amount $B$ directly from the premium rate; claims occur at the mortality-weighted rate $B\mu$, not at rate $B$.

Changing premium rates over time

The general integral is useful when the premium rate changes by duration. Under constant $\mu$ and $\delta$, suppose $p(s)$ is piecewise constant and the benefit remains $B$. Then each interval contributes its own premium rate less $B\mu$, weighted for survival and accumulation. It is often clearest to integrate separately over the intervals rather than replace the varying premium by a single average.

Worked Example: Two Continuous Premium Rates

A hypothetical two-year contract pays an immediate death benefit of $B=\$2{,}500$. The premium is paid continuously at rate $\$60$ per year during the first year and $\$100$ per year during the second year. Assume constant mortality force $\mu=0.02$, constant force of interest $\delta=0.03$, and no other benefits or expenses. Find the reserve at duration 2, conditional on survival.

Method. The expected claim outgo rate while in force is $B\mu=2{,}500(0.02)=\$50$ per year. The net rate is therefore $60-50=\$10$ in the first year and $100-50=\$50$ in the second. In the general integral, the accumulation and survival weights combine as $e^{0.1}e^{-0.05s}$ when $t=2$.

V_2^R =e^{0.1}\left[ 10\int_0^1 e^{-0.05s}\,ds +50\int_1^2 e^{-0.05s}\,ds \right].

The first integral is $(1-e^{-0.05})/0.05=(1-0.951229)/0.05=0.975412$. The second is $(e^{-0.05}-e^{-0.10})/0.05=(0.951229-0.904837)/0.05=0.927840$. Therefore,

V_2^R =1.105170\left[10(0.975412)+50(0.927840)\right] =1.105170(9.75412+46.39200) \approx \$62.05.

As a check, multiplying the accumulation factor into each interval contribution gives $10(1.077982)+50(1.025422)=\$62.0509$, which also rounds to \$62.05. The method preserves the timing of the rate change; using one premium rate over the full two years would generally give a different result.

Piecewise mortality and survival to the valuation time

A constant mortality force is not required. When mortality changes by duration, the survival probability and claim density must reflect the force applicable at each time. For a force that is constant on successive intervals, survival across an interval is exponential in that interval's force, and survival to a later time is the product of the interval survival probabilities. The denominator ${}_tp_x$ must use the mortality over the entire period from issue to $t$.

Worked Example: A Piecewise Mortality Force

A hypothetical policy pays premiums continuously at $\$100$ per year while in force and pays $\$2{,}000$ immediately on death. There is no interest, so $\delta=0$. The mortality force is $0.01$ per year during the first year and $0.03$ per year during the second year. Calculate the reserve at $t=2$, conditional on survival to that time.

Method. With no interest, accumulated cash flows are not changed by time. In the first year, the net expected cash-flow rate per policy in force is $100-2{,}000(0.01)=\$80$. In the second year it is $100-2{,}000(0.03)=\$40$. Survival to the start of the second year is $e^{-0.01}$, and survival to time 2 is $e^{-0.01}e^{-0.03}=e^{-0.04}$.

The accumulated net cash flow per original policy is the sum of the two survival-weighted interval integrals:

80\int_0^1 e^{-0.01s}\,ds +40\int_1^2 e^{-0.01}e^{-0.03(s-1)}\,ds =80\frac{1-e^{-0.01}}{0.01} +40\frac{e^{-0.01}-e^{-0.04}}{0.03}.

The first term is $80(0.995017)=\$79.6013$. The second is $40(0.975347)=\$39.0139$. Their sum is $\$118.6152$ per original policy. Divide by the probability of survival to time 2:

V_2^R=\frac{118.6152}{e^{-0.04}} =\frac{118.6152}{0.960789} \approx \boxed{\$123.46}.

This is the reserve per surviving policy, not the accumulated amount per original policy. The result is checked by multiplying $\$123.46$ by the survival proportion $0.960789$, which returns approximately $\$118.62$ per original policy.

Common conditioning and cash-flow errors

The most important exam trap is to stop after calculating accumulated cash flow per original policy. That amount is not yet the retrospective reserve per policy surviving at time $t$. In the preceding example, omitting the denominator gives $\$118.62$, whereas the required reserve is $118.6152/0.960789=\$123.46$. Since fewer policies survive than were issued, dividing by the survival proportion increases this positive accumulated amount.

A second error is to treat the death benefit as if it were paid continuously at rate $B$. The benefit amount is paid only on death. Over a short interval, the expected claim outgo rate per policy in force is $B\mu_{x+s}$, so the mortality force must be included. A third error is to use the survival probability to time $t$ for every cash flow. Cash flows at time $s$ are weighted by survival to $s$, then accumulated to $t$; only after integrating is the result divided by survival to $t$.

Key takeaway: For a continuous-time retrospective reserve, weight premium income by the probability of being alive when it is paid, weight death claims by the death density, accumulate each cash flow to the valuation time, and divide the cohort balance by ${}_tp_x$. Keep the premium rate, death-benefit amount, and mortality force distinct.

Check Your Understanding

For each problem, calculate a reserve per policy surviving to the stated duration. Assume no expenses or cash flows other than those specified.

A contract has constant $\mu=0.04$, $\delta=0.01$, continuous premium rate $P=\$90$, and immediate death benefit $B=\$1{,}500$. Find $V_{1.5}^R$. (Answer: $\$46.73$)
A two-year contract has constant $\mu=0.02$, $\delta=0.03$, premium rate $P=\$75$ continuously, and immediate death benefit $B=\$2{,}000$. Find $V_1^R$. (Answer: $\$35.89$)
There is no interest. A policy pays a continuous premium rate of $\$70$ and a $\$1{,}000$ immediate death benefit. The mortality force is $0.02$ in year 1 and $0.04$ in year 2. Find the reserve at time 2. (Answer: $\$83.17$)
A two-year policy has constant $\mu=0.01$, $\delta=0.02$, and immediate death benefit $B=\$4{,}000$. Its continuous premium rate is $\$30$ in year 1 and $\$80$ in year 2. Find $V_2^R$. (Answer: $\$30.15$)

What You'll Learn