Annual Losses and One-Year Capital

Why the Timing of Annual Losses Matters

One-year capital is tied to a measurement horizon and a valuation date. If the horizon ends one year from now, a payment made earlier in the year is not on the same time basis as a payment made at year-end. To assess the year-end solvency position consistently, losses paid before the horizon are accumulated to the end of the year using the assumed interest rate.

This tutorial treats the one-year loss as a random amount measured at the end of the year. The loss distribution then provides a basis for calculating expected loss, a quantile of annual loss, and the resources needed to meet a stated solvency standard. These calculations are aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.

Definition: Let $C_j$ be a loss payment made at time $t_j$, where $0\leq t_j\leq 1$. At a deterministic effective annual interest rate $i$, the end-of-year value of that payment is $C_j(1+i)^{1-t_j}$. For a scenario with several payments, its annual loss $X$ is the sum of their end-of-year values. The annual loss distribution is the probability distribution of $X$ across all modeled scenarios.

This definition uses a specific loss basis: the payments included in $X$ are the complete cash flows being measured over the year. It assumes the same deterministic interest rate applies over the year and that payments can be accumulated at that rate. If the model uses a different valuation basis, or includes other cash flows, the loss definition must be adjusted before calculating capital.

For a loss distribution with cumulative distribution function $F_X(x)=P(X\leq x)$, define the $\alpha$-level VaR as the smallest loss amount whose cumulative probability is at least $\alpha$. Under the common unexpected-loss convention, economic capital is VaR less expected loss. Expected loss is assumed to be funded separately; total loss-paying resources at the VaR level are therefore expected loss plus this capital buffer.

Key relationship: For a selected confidence level $\alpha$, unexpected-loss capital is $\operatorname{VaR}_{\alpha}(X)-E[X]$. If expected loss is funded separately, total resources at the measurement date are $E[X]+\bigl(\operatorname{VaR}_{\alpha}(X)-E[X]\bigr)=\operatorname{VaR}_{\alpha}(X)$. The probability that those resources cover the loss is $P(X\leq\operatorname{VaR}_{\alpha}(X))$, which can exceed $\alpha$ when the distribution has a probability mass at the quantile.

Converting Payments to an Annual Loss

A one-year loss model may describe payments at different times, rather than give a single end-of-year amount directly. The required conversion follows from accumulation: a payment $C_j$ made at time $t_j$ earns interest for $1-t_j$ years before the measurement date. Adding the accumulated payments gives the scenario’s annual loss.

X=\sum_j C_j(1+i)^{1-t_j}

For example, a payment at $t=1$ is already valued at year-end and is unchanged. A payment at $t=0.5$ is accumulated for half a year, giving the factor $(1+i)^{0.5}$. The annual loss is a random variable because the payment amounts, payment times, or occurrence of the payments can differ by scenario.

Fix the measurement date and loss basis.
Specify which one-year cash flows count as losses and measure all scenario outcomes at the end of the year.

Accumulate each payment.
For every payment, apply $(1+i)^{1-t_j}$, using its actual timing $t_j$.

Construct the annual loss distribution.
Add the accumulated payments within each scenario and retain that scenario’s probability.

Calculate loss and capital measures.
Use the distribution to find $E[X]$, the selected VaR, and—if using the unexpected-loss convention—the difference between them.

Check solvency on the same date.
Compare end-of-year loss-paying resources with end-of-year losses and sum the probabilities of covered outcomes.

Worked Applications

Worked Example: Accumulating a Midyear Claim

A hypothetical portfolio has an effective annual interest rate of 4%. During the year, it has no loss with probability 0.85, a $100,000 payment at $t=0.5$ with probability 0.12, or a $300,000 payment at $t=1$ with probability 0.03. Treat these as complete loss cash flows, with no other payments. Find the 95% VaR, expected annual loss, and unexpected-loss capital at year-end.

Method. First convert each scenario to an end-of-year loss. The no-loss outcome remains $0. The midyear payment accumulates for half a year:

100{,}000(1.04)^{0.5}=100{,}000(1.0198039027\ldots) =101{,}980.39027\ldots

The year-end payment remains $300,000. Thus the end-of-year loss outcomes are $0, $101,980.39027…, and $300,000, with probabilities 0.85, 0.12, and 0.03. The cumulative probability is 0.85 at $0 and 0.97 at $101,980.39027…, so the 95% quantile is the midyear-payment outcome.

Expected annual loss is the probability-weighted mean:

E[X]=0.85(0)+0.12(101{,}980.39027\ldots)+0.03(300{,}000) =21{,}237.64683\ldots

Therefore, to the nearest cent, the 95% VaR is $101,980.39 and expected loss is $21,237.65. Unexpected-loss capital is $101{,}980.39027\ldots-21{,}237.64683\ldots=80{,}742.74344\ldots$, or $80,742.74 to the nearest cent. The calculation checks: expected loss plus the unrounded capital buffer equals the VaR. Resources equal to the VaR cover the first two outcomes, so their solvency probability is $0.85+0.12=0.97$, which exceeds 0.95 because the quantile outcome has positive probability.

Common mistake avoided: Using $100,000 as the midyear loss at year-end ignores six months of accumulation. The loss must be valued at the specified horizon before computing either the quantile or the mean.

Worked Example: Aggregating Payments Within Scenarios

A hypothetical risk has an effective annual interest rate of 4%. Its complete one-year payment scenarios are: no payment with probability 0.70; $60,000 at $t=0.5$ with probability 0.20; $100,000 at $t=0.5$ and $50,000 at $t=1$ with probability 0.08; or $400,000 at $t=1$ with probability 0.02. Calculate the 95% VaR, expected annual loss, and unexpected-loss capital.

Method. Accumulate payments within each scenario, then order the resulting annual losses. The half-year accumulation factor is $(1.04)^{0.5}=1.0198039027\ldots$. The scenario losses are:

No payment: $X=0$.
One midyear payment: $X=60{,}000(1.0198039027\ldots)=61{,}188.23416\ldots$.
Two payments: $X=100{,}000(1.0198039027\ldots)+50{,}000=151{,}980.39027\ldots$.
One year-end payment: $X=400{,}000$.

The cumulative probabilities at these ordered losses are 0.70, 0.90, 0.98, and 1.00. The first loss amount with cumulative probability at least 0.95 is $151,980.39027…, so the 95% VaR is $151,980.39.

The expected annual loss is:

\begin{aligned} E[X]&=0.70(0)+0.20(61{,}188.23416\ldots)\\ &\quad+0.08(151{,}980.39027\ldots)+0.02(400{,}000)\\ &=32{,}396.07805\ldots \end{aligned}

To the nearest cent, expected loss is $32,396.08. Unexpected-loss capital is $151{,}980.39027\ldots-32{,}396.07805\ldots=119{,}584.31222\ldots$, or $119,584.31. The check is that expected loss plus the unrounded buffer returns $151,980.39027…; resources at that level cover the first three scenarios, with probability $0.70+0.20+0.08=0.98$.

Common mistake avoided: The third scenario contains two payments, not two alternative losses. Both accumulated amounts belong in that scenario’s annual loss before the scenario is assigned its probability.

Worked Example: From Opening Funds to Year-End Solvency

A hypothetical insurer models complete year-end losses of $10,000, $40,000, and $100,000, with probabilities 0.50, 0.40, and 0.10. The effective annual interest rate is 5%, and there are no other cash flows. Calculate expected loss, 90% unexpected-loss capital, and total year-end resources needed for 90% solvency. Then assess the solvency probability if $38,000 of loss-paying resources are available at the start of the year and earn 5%.

Method. The losses are already stated at year-end, so no payment-timing adjustment is needed. Expected loss is:

$$ E[X]=0.50(10{,}000)+0.40(40{,}000)+0.10(100{,}000) =5{,}000+16{,}000+10{,}000=31{,}000 $$

The cumulative probabilities at $10,000, $40,000, and $100,000 are 0.50, 0.90, and 1.00. Hence $\operatorname{VaR}_{0.90}(X)=\$40{,}000$. Unexpected-loss capital is $40{,}000-31{,}000=\$9{,}000$; with expected loss funded separately, total required resources at year-end are $31{,}000+9{,}000=\$40{,}000$. These resources cover losses of $10,000 and $40,000, giving solvency probability $0.50+0.40=0.90$.

The available opening resources accumulate to $38{,}000(1.05)=\$39{,}900$ at year-end. They cover the $10,000 outcome but not the $40,000 outcome, so the resulting solvency probability is 0.50. To have $40,000 at year-end, the required opening amount is $40{,}000/1.05=\$38{,}095.238\ldots$, or $38,095.24 to the nearest cent.

Common mistake avoided: Calling $40,000 the unexpected-loss capital would confuse the VaR-level total resources with the buffer above expected loss. Under the stated convention, the buffer is $9,000; the $31,000 expected loss is funded separately. The solvency check compares total year-end resources with the full loss distribution.

Why the Relationships Hold

The accumulation formula follows directly from the definition of an effective annual rate. A dollar invested for one year grows to $1+i$ dollars. For a fractional period of $s$ years under a constant effective annual rate, the accumulation factor is $(1+i)^s$. A payment at time $t_j$ has $1-t_j$ years to accumulate to the end of the year, so its year-end value is $C_j(1+i)^{1-t_j}$. Since scenario cash flows add, their year-end values add as well.

Expected annual loss is calculated by weighting each scenario loss by its probability. In the discrete case, if $X=x_k$ with probability $p_k$, where the probabilities sum to one, then $E[X]=\sum_k p_kx_k$. The VaR is found from the cumulative probabilities, not by averaging losses. Subtracting expected loss from VaR gives the amount above the mean required to reach that quantile; adding the expected loss back recovers the total resources at the quantile.

For a discrete distribution, solvency at resources $R$ means $X\leq R$. Thus its probability is the sum of the probabilities of all outcomes no larger than $R$. At $R=\operatorname{VaR}_{\alpha}(X)$, this probability is at least $\alpha$, but need not equal $\alpha$. This is why checking the actual cumulative probability is important.

Common Traps and Practical Checks

Mixing valuation dates. Do not combine a midyear payment and a year-end payment as if both were measured at year-end. Accumulate the earlier payment first.
Applying the rate for the wrong period. A payment at $t=0.5$ is accumulated for 0.5 years, not a full year; its factor is $(1+i)^{0.5}$.
Confusing VaR with unexpected-loss capital. VaR is the total loss level at the selected quantile. Under the convention used here, capital is VaR less expected loss.
Assuming discrete solvency probability equals the confidence level. A probability mass at the VaR can make the actual probability of coverage larger than the target.
Comparing resources and losses at different dates. Accumulate opening funds to year-end, or convert the required year-end amount back to the opening date. Do not compare a beginning-of-year balance directly with an end-of-year loss.

A reliable one-year calculation states the measurement date, loss basis, interest assumption, and confidence level; builds the loss distribution on one consistent date; and verifies the solvency probability from the cumulative distribution. This keeps the interpretation of the capital amount separate from the mechanics used to calculate it.

Check Your Understanding

For each hypothetical problem, use the stated loss basis and measurement date. Round dollar amounts to the nearest cent only at the final step.

At an effective annual rate of 21%, a portfolio has no payment with probability 0.85, a $50,000 payment at $t=0.5$ with probability 0.10, or a $120,000 payment at $t=1$ with probability 0.05. Find the 90% VaR, expected annual loss, and unexpected-loss capital. (Answer: VaR $55,000.00; expected loss $11,500.00; capital $43,500.00.)
At an effective annual rate of 21%, a hypothetical risk has no payment with probability 0.60, a $20,000 midyear payment with probability 0.25, both a $40,000 midyear payment and a $10,000 year-end payment with probability 0.10, or a $100,000 year-end payment with probability 0.05. Find the 90% VaR, expected annual loss, unexpected-loss capital, and solvency probability at resources equal to VaR. (Answer: VaR $54,000.00; expected loss $15,900.00; capital $38,100.00; solvency probability 0.9500.)
A year-end loss is $15,000 with probability 0.40, $40,000 with probability 0.50, and $100,000 with probability 0.10. Calculate expected loss, 90% VaR, unexpected-loss capital, and the coverage probability at total resources equal to VaR. (Answer: expected loss $36,000.00; VaR $40,000.00; capital $4,000.00; coverage probability 0.9000.)
At an effective annual rate of 10%, year-end losses are $0 with probability 0.75, $25,000 with probability 0.15, and $80,000 with probability 0.10. Find the expected loss and 95% VaR. If opening loss-paying resources are $30,000, find year-end resources and the resulting solvency probability. (Answer: expected loss $11,750.00; VaR $80,000.00; year-end resources $33,000.00; solvency probability 0.9000.)
At an effective annual rate of 0%, year-end losses are $5,000 with probability 0.60, $20,000 with probability 0.30, and $90,000 with probability 0.10. Find expected loss, 95% VaR, unexpected-loss capital, and solvency probability if total resources are $50,000. (Answer: expected loss $18,000.00; VaR $90,000.00; capital $72,000.00; solvency probability 0.9000.)

What You'll Learn