From Catastrophe Losses to a Capital Plan
A catastrophe model can produce event scenarios, annual loss distributions, and estimates of how reinsurance changes losses. Capital planning turns those outputs into a forward-looking decision: how much financial resource should be available, and will the projected capital position be sufficient? A modeled loss is not itself a capital requirement. The result depends on the chosen risk measure, the horizon, the treatment of reinsurance and other management actions, and the capital buffer set by governance.
A sound plan keeps three quantities distinct: a scenario loss under a specified stress, a probability-based measure of the loss distribution, and the funds available to absorb losses. Scenario analysis helps test resilience to a specified event or change in assumptions. A tail measure such as TVaR summarizes a part of the modeled loss distribution. Available capital is the resource against which a selected requirement is assessed. These quantities inform one another, but they are not interchangeable.
The model basis must be explicit. For example, the annual aggregate loss distribution should reflect the exposure set, modeled perils, loss definition, event aggregation, and reinsurance terms being assessed. If a calculation uses a discrete distribution, the probabilities must sum to one. If losses are net of reinsurance, recoveries should be calculated under the stated contract terms before applying the selected risk measure. A risk measure calculated for one horizon or basis should not be compared directly with funds or a requirement defined on another.
Calculating Tail Capital From a Discrete Model
For a discrete loss model, the VaR is the smallest loss value whose cumulative probability reaches or exceeds \(q\). TVaR is not generally the same as that VaR: it averages losses in the worst \(1-q\) fraction of outcomes. If a probability mass straddles the boundary, only the portion needed to fill that tail fraction is included.
Let \(v=\operatorname{VaR}_q(X)\), let \(r=P(X>v)\), and let \(\alpha=1-q\). The discrete-tail calculation is
The term \(\alpha-r\) is the remaining tail probability needed after including all outcomes strictly larger than \(v\). This formula follows by taking exactly the worst \(\alpha\) probability of outcomes: include every outcome above \(v\), then fill the remaining tail probability with outcomes at \(v\). The numerator is the loss-weighted probability in that tail, and division by \(\alpha\) gives its average. This boundary adjustment matters whenever probability is concentrated at a loss value.
Worked Example: Converting a Catastrophe Distribution Into a Capital Gap
A hypothetical insurer uses this annual aggregate catastrophe loss model, in millions of dollars, for its current exposure. Assume the outcomes are mutually exclusive and exhaustive, and the stated probabilities are the complete model for this calculation. Calculate the 99% VaR, 99% TVaR, expected loss, economic capital, and the gap to a stated capital target if eligible funds are $155 million.
| Annual loss, $ millions | Probability |
|---|---|
| 0 | 0.90 |
| 40 | 0.06 |
| 100 | 0.03 |
| 180 | 0.01 |
Method and model check. This is a discrete annual loss model. Its probabilities sum to \(0.90+0.06+0.03+0.01=1.00\). The cumulative probabilities at losses \(0,40,100,180\) are \(0.90,0.96,0.99,1.00\).
VaR. The first cumulative probability reaching 0.99 occurs at a loss of $100 million. Thus \(\operatorname{VaR}_{0.99}(X)=100\) million.
TVaR. The worst 1% of outcomes consists exactly of the $180 million outcome, which has probability 0.01. Therefore,
Expected loss. Weight each outcome by its probability:
Economic capital and gap. Under the stated convention, economic capital is \(180-7.2=172.8\) million. If the plan targets 99% TVaR as the required funds, the gap is \(180-155=25\) million. The economic-capital amount and the target-funds gap answer different questions: the former subtracts expected loss; the latter compares the chosen gross requirement directly with available funds.
Check. The 99% VaR is no greater than the 99% TVaR, as expected here, and the expected loss is below both tail measures. The reported gap is positive, so funds fall short of the stated target by $25 million.
Common mistake avoided. Treating the 99% VaR of $100 million as the average worst 1% loss would understate this model's tail average by $80 million. VaR locates a quantile; TVaR averages the specified tail.
Reinsurance, Net Loss, and the Requirement Basis
A catastrophe capital plan should show whether its requirement is gross or net of reinsurance. For an occurrence excess-of-loss layer attaching at \(a\) with limit \(L\), the recovery on a single covered event loss \(x\) is \(\min(\max(x-a,0),L)\), subject to the actual contract wording and any applicable conditions. The insurer's net event loss is gross loss less recovery. This event-level formula does not by itself capture reinstatement premiums, aggregate limits, multiple events, exclusions, or counterparty default; those require explicit treatment if material to the model.
Reinsurance can lower modeled net tail loss, but it can also have a premium cost and may not respond to every scenario. Therefore the plan should calculate net losses under the contract, identify the associated premium or other cost on the appropriate financial basis, and test whether the intended protection is available in the planning horizon.
Worked Example: Reinsurance Changes the Tail Requirement
Use the annual loss distribution in the preceding example as a hypothetical single-event model for this calculation. A hypothetical occurrence treaty pays losses in excess of $20 million, up to a $50 million limit, per event. Ignore reinstatements and counterparty default. Calculate the net loss distribution, its 99% TVaR and expected loss, and compare it with the gross results. The hypothetical treaty premium is $3 million.
Method. For each gross loss \(x\), calculate ceded recovery \(C(x)=\min(\max(x-20,0),50)\), then net loss \(N(x)=x-C(x)\). Apply the risk measure to the resulting net distribution, not to the gross loss values.
| Gross loss | Recovery calculation | Recovery | Net loss | Probability |
|---|---|---|---|---|
| 0 | \(\min(0,50)\) | 0 | 0 | 0.90 |
| 40 | \(\min(40-20,50)\) | 20 | 20 | 0.06 |
| 100 | \(\min(100-20,50)\) | 50 | 50 | 0.03 |
| 180 | \(\min(180-20,50)\) | 50 | 130 | 0.01 |
The net cumulative probabilities are 0.90 at $0 million, 0.96 at $20 million, 0.99 at $50 million, and 1.00 at $130 million. Hence net 99% VaR is $50 million. The worst 1% has a net loss of $130 million, so net 99% TVaR is $130 million.
Expected net loss.
Net economic capital is \(130-4=126\) million. Compared with the gross 99% TVaR of $180 million, the modeled reduction in this tail measure is \(180-130=50\) million. The $3 million premium should be reflected in the financial projection of funds according to the plan's timing and accounting basis; it is not a catastrophe loss recovery and should not be subtracted from TVaR as though it changed the modeled loss outcomes.
Check. At the largest gross loss, the recovery is capped at $50 million, leaving \(180-50=130\) million net. The net tail result therefore respects the layer limit. A calculation that deducted $80 million from the $180 million outcome would have ignored that limit.
Forward-Looking Capital Planning
A capital plan projects both the requirement and the resources available at future dates. Exposure growth, portfolio changes, inflation assumptions, model revisions, reinsurance renewal terms, mitigation, and planned distributions can alter one or both sides. The projection should state its horizon and assumptions, then compare resources with the requirement on the same date and basis. Scenario analysis is especially useful for showing how the plan behaves under a specified adverse event or assumption change; it does not assign that stress a probability unless a probability model supports one.
Specify the horizon, exposure and peril scope, gross or net loss basis, risk measure and confidence level, and any separate buffer.
Apply stated exposure changes, modeled mitigation effects, and reinsurance terms in the correct sequence. Keep scenario stresses distinct from probability-based results.
Reflect retained earnings, capital contributions, distributions, premiums, and action costs using consistent timing and units.
Compare projected funds with the requirement plus any stated buffer, quantify any shortfall or surplus, and identify a feasible response and its timing.
Worked Example: Projecting Capital After Growth and Mitigation
A hypothetical insurer starts with $175 million of eligible funds. Over the next planning year, it expects $8 million of retained earnings and plans a $15 million distribution. A portfolio growth factor of 1.25 is applied to each positive gross loss in the discrete model from Worked Example 2. A hypothetical engineering program then reduces each positive grown loss by $10 million, with no loss reduced below zero. The existing $20 million excess-of-loss attachment and $50 million limit remain in force. The program costs $4 million, paid within the planning year. The insurer requires projected net 99% TVaR plus a $10 million buffer. Calculate the target, projected funds, and any additional capital needed.
Method. First grow the gross outcomes, then apply the stated mitigation reduction, then calculate treaty recoveries and net outcomes. This order represents the problem's assumption that mitigation changes the loss before reinsurance responds.
| Original gross loss | After growth | After mitigation | Treaty recovery | Projected net loss |
|---|---|---|---|---|
| 0 | 0 | 0 | 0 | 0 |
| 40 | \(40(1.25)=50\) | \(50-10=40\) | \(\min(40-20,50)=20\) | 20 |
| 100 | \(100(1.25)=125\) | \(125-10=115\) | \(\min(115-20,50)=50\) | 65 |
| 180 | \(180(1.25)=225\) | \(225-10=215\) | \(\min(215-20,50)=50\) | 165 |
The probabilities remain \(0.90,0.06,0.03,0.01\), because the assumed growth and mitigation change loss amounts but not the modeled outcome probabilities. The projected net 99% VaR is $65 million because cumulative probability reaches 0.99 at that outcome. The worst 1% outcome is $165 million, so projected net 99% TVaR is $165 million. The target including the separately stated buffer is \(165+10=175\) million.
Projected eligible funds after earnings, distribution, and program cost are
The shortfall is \(175-164=11\) million. If a hypothetical capital contribution of $12 million is available in time, projected funds become \(164+12=176\) million, leaving a $1 million surplus over the target.
Check. The largest mitigated gross loss is $215 million; the treaty pays its maximum $50 million, leaving $165 million net. The capital contribution exceeds the calculated shortfall by exactly $1 million. The $4 million mitigation cost is included in projected funds, rather than silently treated as free.
Common mistake avoided. Applying the treaty to the original $180 million loss and then scaling the net result would not match the stated contract and action assumptions. The specified sequence is growth, mitigation, treaty recovery, then net risk measurement.
Interpreting the Plan and Avoiding Traps
A capital plan is decision support, not a claim that a modeled quantile will be the realized loss. The model represents specified exposures, assumptions, and contract responses. A stress scenario may reveal a vulnerability even when it is outside the modeled tail metric, while a probability-based measure summarizes only the distribution supplied to it. Decision-makers should review the loss basis, assumptions, uncertainty, timing, and execution risks behind any proposed response.
- Do not confuse VaR with a tail average. VaR is a quantile; TVaR averages the worst specified portion of outcomes.
- Do not compare unlike bases. Gross losses, net losses, economic capital, required funds, and available capital are different quantities. State which one is being compared.
- Do not treat a scenario as a probability statement. A stress loss is conditional on a described scenario unless a model assigns probabilities to it.
- Do not omit costs or timing. Premiums, mitigation expense, planned distributions, and capital actions affect projected funds even when they do not directly change modeled losses.
- Do not ignore contract limits. Reinsurance recoveries depend on attachment, limit, and wording; they are not a blanket percentage reduction in gross loss.
The practical result of the analysis is a traceable plan: a defined catastrophe loss basis, a stated risk measure and buffer, a projection of available funds, and a quantified response if the projected position misses its target. This approach is aligned with the CAS Exam 9 (Risk Management for Actuaries) syllabus.
Check Your Understanding
For each hypothetical problem, state the method and keep the loss basis and planning assumptions explicit.
- An annual loss model has outcomes of $0 million with probability 0.92, $30 million with probability 0.05, and $90 million with probability 0.03. Find the 97% VaR and 97% TVaR. (Answer: VaR = $30 million; TVaR = $90 million.)
- A hypothetical event loss is $75 million. A treaty pays $15 million excess of $25 million, subject to a $40 million limit. Calculate the recovery and net loss. (Answer: Recovery = $40 million; net loss = $35 million.)
- A model has losses of $0, $20, and $100 million with probabilities 0.90, 0.08, and 0.02. Calculate expected loss and 98% TVaR. (Answer: Expected loss = $3.60 million; 98% TVaR = $100 million.)
- Projected net TVaR is $142 million and the plan adds a $12 million buffer. Funds before planned actions are $160 million; retained earnings are $5 million, a distribution is $18 million, and a mitigation program costs $3 million. Find the target, projected funds, and shortfall. (Answer: Target = $154 million; projected funds = $144 million; shortfall = $10 million.)
- A hypothetical model's 99% TVaR is $210 million gross. A modeled reinsurance program reduces net 99% TVaR to $165 million and costs $7 million in premium during the plan year. If funds before premium are $170 million and the target is net TVaR plus a $10 million buffer, calculate the target and funds after premium, then identify the gap. (Answer: Target = $175 million; funds after premium = $163 million; shortfall = $12 million.)