Why a Catastrophe Model Uses Separate Modules
A catastrophe loss estimate is not produced by one calculation on a single portfolio total. A model combines information about insured assets, a damaging event, the assets’ response to that event, and the applicable financial terms. Keeping those steps separate makes it possible to see what each input contributes and where an error has entered.
The four modules are commonly described as exposure, hazard, vulnerability, and financial. Software and model providers may use different labels or combine steps internally, but the conceptual sequence is the same: identify what is exposed, estimate the event’s effect at each location, translate that effect into physical damage, and apply coverage terms to determine insured loss.
These stages answer different questions. The exposure module does not say how severe an event is. The hazard module does not decide how much a building is damaged. The vulnerability module does not determine what a policy covers. And the financial module does not repair missing or inaccurate descriptions of the insured assets.
The Four Modules and Their Outputs
| Module | Question it addresses | Typical output passed forward |
|---|---|---|
| Exposure | Which insured assets are being modeled, and what relevant characteristics and values describe them? | Identified locations or assets, values, attributes, and coverage references |
| Hazard | What event conditions reach each exposed location? | Location-level hazard measures for the event |
| Vulnerability | What physical damage is associated with those conditions and asset characteristics? | A damage amount or damage ratio for each asset |
| Financial | What portion of that damage is paid under the specified insurance terms? | Location, policy, or portfolio insured losses |
The exposure module may use information such as an asset’s value, use, construction characteristics, and applicable coverage reference. The exact fields depend on the peril and model. For this tutorial, assume the required exposure information has already been supplied and accepted; assessing data quality and representing exposure fields are separate tasks.
The hazard module evaluates a specified event against the exposed locations. Its output might be a wind speed, flood depth, ground-motion measure, or another peril-specific intensity. A location outside the event’s affected area may receive a negligible or zero modeled intensity. The hazard output is not itself a dollar loss.
The vulnerability module relates hazard conditions to expected physical damage for the asset type and characteristics represented. A convenient summary is a damage ratio: the fraction of an asset’s value lost as physical damage. For example, a damage ratio of 0.10 corresponds to physical damage equal to 10% of the value used as the model’s exposure basis. It is not automatically the insurer’s payment.
Finally, the financial module applies the relevant coverage structure. Depending on the model and contract, this may include deductibles, limits, participation, or other terms. The order and level at which terms apply matter. A location-level payment, a policy-level loss, and a portfolio total are not interchangeable quantities.
From Damage Ratio to Ground-Up Damage
For a simple deterministic calculation, let \(V_j\) be the modeled value of asset \(j\), and let \(r_j\) be its damage ratio for the event. Assume the value and ratio use consistent coverage and valuation bases, and that the ratio directly represents a fraction of that value. Then ground-up physical damage is:
Here, “ground-up” means physical damage before deductibles, limits, or other financial terms. The equation is a simplified representation: actual models may use more detailed vulnerability calculations, and the result depends on the modeled peril and asset assumptions.
The reason for the multiplication is the definition of a ratio. If \(r_j\) is the fraction of the asset value damaged, then \(r_j=D_j/V_j\). Multiplying both sides by \(V_j\) gives \(D_j=V_jr_j\). If the damage ratio is uncertain and \(V_j\) is fixed, the same relationship implies \(E[D_j\mid V_j]=V_jE[r_j\mid V_j]\). That statement concerns expected physical damage; it does not generally justify applying policy terms to expected damage to obtain expected insured loss.
Worked Example: Tracing One Warehouse Through All Four Modules
A hypothetical insurer models one warehouse for a specified wind event. The exposure record gives a modeled value of $2,000,000 and identifies a location-specific policy. The hazard module assigns an event wind intensity of 120 miles per hour at the location. Under the supplied asset characteristics, the vulnerability output is a deterministic damage ratio of 0.15. The policy has a $25,000 deductible and a $250,000 location limit. Assume the deductible applies to ground-up damage and the limit caps the payment after the deductible.
Method. Follow the module sequence, first calculating physical damage and then applying the stated location-level financial terms. This example assumes the supplied hazard intensity and vulnerability ratio are already valid model outputs; it does not estimate either one from raw event data.
Exposure and hazard. The exposure module supplies the $2,000,000 value and the location’s policy reference. The hazard module supplies 120 miles per hour for this event at that location. Neither value is itself a loss estimate.
Vulnerability. The physical damage is the modeled value multiplied by the damage ratio:
The event therefore produces $300,000 of modeled ground-up physical damage at the warehouse.
Financial. Apply the $25,000 deductible first:
The resulting amount exceeds the $250,000 location limit, so the insurer’s modeled payment is $250,000. The ground-up damage not covered by that payment is $300,000 − $250,000 = $50,000. This consists of the $25,000 deductible and $25,000 above the payment limit. The result checks: $250,000 paid plus $50,000 not paid equals $300,000 of ground-up damage.
Common mistake avoided: Calling $300,000 the insured loss skips the financial module. It is the modeled physical damage; under the stated terms, the insurer’s payment is $250,000.
Aggregating Location Results for an Event
For a single realized event, location-level ground-up damages can be added to obtain total modeled physical damage, provided they refer to the same event and compatible loss definitions. Insured payments can also be added when the applicable contract terms have first been applied at the correct level. A portfolio total should therefore be understood as the endpoint of calculations at the relevant asset, location, or policy level—not as a substitute for those calculations.
For example, if two locations have different deductibles, applying one deductible to their combined damage will generally not match applying each deductible separately. The exposure and financial records must preserve the relationships needed to apply terms as written. This is one reason the model carries identifiers and coverage references through its stages.
Worked Example: Two Locations With Different Hazard and Coverage
A hypothetical event affects two insured buildings. The accepted exposure records give Building A a value of $1,200,000 and Building B a value of $2,500,000. The hazard module assigns intensities of 80 and 110 miles per hour, respectively. The vulnerability outputs for these building characteristics and intensities are damage ratios of 0.05 for A and 0.20 for B. A has a $10,000 deductible and a $100,000 location limit; B has a $20,000 deductible and a $400,000 location limit. Assume each deductible and limit applies separately to its building, with the deductible applied before the limit.
Method. Calculate physical damage separately for each location, apply that location’s financial terms, and then sum the event results. This preserves the different hazard, vulnerability, and policy inputs.
Building A. Its ground-up damage is:
After the deductible, the amount eligible for payment is $60,000 − $10,000 = $50,000. That is below A’s $100,000 limit, so the payment for A is $50,000.
Building B. Its ground-up damage is:
After the deductible, the amount eligible for payment is $500,000 − $20,000 = $480,000. The $400,000 limit caps B’s payment at $400,000.
Event totals. Total ground-up damage is $60,000 + $500,000 = $560,000. Total insured payment is $50,000 + $400,000 = $450,000. The difference is $560,000 − $450,000 = $110,000, which checks against the amounts not paid at each location: $10,000 for A and $100,000 for B.
Common mistake avoided: Using one portfolio-wide damage ratio or applying one building’s policy terms to both buildings would discard location-specific model inputs. The correct aggregation adds the two results only after each location’s damage and coverage have been calculated.
Why the Financial Module Must Follow Vulnerability
The financial module acts on a loss amount, not on the hazard intensity or the exposure value by itself. A deductible is a term on covered loss, and a limit caps a payment under a specified contract basis. Applying either term directly to an exposure value, before estimating damage, changes the question being answered.
The sequence also matters when terms apply separately to multiple locations or policies. Consider two buildings with equal ground-up damage of $30,000 each, each insured with a $10,000 deductible and a $25,000 location limit. Under these assumptions, each location pays $30,000 − $10,000 = $20,000, which is below its limit. Total payment is $20,000 + $20,000 = $40,000.
If the damages are aggregated first and one $10,000 deductible is incorrectly applied once, the calculation gives $60,000 − $10,000 = $50,000, before considering limits. That is not the contract calculation: it treats two location deductibles as one. The correct result remains $40,000 because terms apply separately at each location. Aggregating before applying terms is valid only when the contract specifies terms at that aggregate level.
Worked Example: Locating an Error in a Model Output
A hypothetical portfolio has two locations. Each has an exposure value of $600,000. For the same event, the hazard and vulnerability modules produce a damage ratio of 0.05 at each location. Each location has its own $5,000 deductible and $40,000 limit. A preliminary report shows $50,000 of insured event loss, calculated by adding the physical damages and subtracting one $10,000 deductible. Determine the correct modeled payment and identify the stage where the error occurred.
Method. Reconstruct the location-level path through the modules and compare it with the report’s calculation. Assume the stated ratios are deterministic and the terms apply separately at each location.
Exposure and vulnerability. For each location, ground-up damage is:
Thus, total ground-up damage is $30,000 + $30,000 = $60,000.
Financial module. At each location, the deductible leaves $30,000 − $5,000 = $25,000 eligible for payment. This amount is below the $40,000 limit, so each location pays $25,000. The correct total insured payment is $25,000 + $25,000 = $50,000.
The preliminary report’s final number happens to match the correct total, but its method is wrong: it aggregated damage and applied one $10,000 deductible, even though the contract has two separate $5,000 location deductibles. The equality here is accidental, because the sum of the two deductibles is also $10,000 and neither location limit binds. A changed deductible or a binding limit could make the incorrect method produce a different answer.
Check. The correct total not paid is $5,000 + $5,000 = $10,000. Adding the $50,000 payment and $10,000 not paid gives $60,000, equal to total ground-up damage. This reconciliation confirms the amounts, while the contract terms—not the matching preliminary total—determine the valid method.
Model Outputs, Assumptions, and Interpretation
A model result is conditional on its event description, exposure data, hazard estimates, vulnerability assumptions, and financial terms. A difference between a modeled loss and an observed loss can arise at any of these stages. A useful review therefore asks which module produced each input and whether the output passed to the next stage has the intended meaning and units.
- Ground-up damage describes modeled physical damage before the financial contract is applied.
- Insured loss reflects the specified coverage terms and may be smaller than ground-up damage.
- Event-level loss aggregates the relevant location or policy results for one defined event.
- Portfolio loss estimates depend on the selected event set and modeling assumptions as well as the module calculations.
A single event trace helps explain the architecture, but it does not establish how frequently such events occur or the full distribution of annual losses. Those questions require additional modeling inputs and analysis. Similarly, an event-level modeled payment should not automatically be treated as a payment under a separate trigger or contract; the financial basis and covered terms must match the quantity being reported.
Check Your Understanding
For each problem, assume all stated module outputs are valid and deterministic, and use the specified financial terms.
- A location has a modeled value of $800,000. The vulnerability output for a specified event is a damage ratio of 0.125. Find the ground-up damage before applying coverage terms. (Answer: $100,000)
- Two locations have values of $1,000,000 and $2,000,000. For one event, their hazard and vulnerability outputs imply damage ratios of 0.04 and 0.10. Calculate each location’s ground-up damage and the total. (Answer: $40,000 and $200,000; total $240,000)
- A location has $90,000 of ground-up damage, a $15,000 deductible, and a $60,000 limit that caps payment after the deductible. Calculate the insured payment and the damage not paid. (Answer: $60,000 payment; $30,000 not paid)
- Two locations each have $45,000 of ground-up damage. Each has a separate $5,000 deductible and a $40,000 location limit, applied after the deductible. Calculate the total insured payment and explain why applying only one deductible to the aggregated damage is inappropriate. (Answer: $80,000 total payment; the two location-level payments are $40,000 each, so the terms must be applied separately)
- A report gives $300,000 of physical damage across a portfolio and states that the financial module output is also $300,000. No policy terms or coverage assumptions are supplied. What can be concluded, and what additional information is needed to identify insured loss? (Answer: $300,000 is the stated ground-up damage, but insured loss cannot be determined without the applicable coverage terms and their level of application)