You've flagged the currency blanks and the BogotΓ‘ TBC to Ardent Re. James passes by your desk and asks how the Meridian SOV looks. You tell him about the missing currencies. He nods "What about the COPE? Have you checked the construction data?"
You scroll to the construction column. You see: Steel Frame, Wood Frame, RC Frame, Masonry, Concrete, Mixed, RCC, Mixed, Concrete Frame, Mixed Use. You count quickly 34 records say "Mixed," "Unknown," or are blank. Another 12 say "Concrete" or "RCC" without any further description. That's 46 of 150 locations where the model will be forced to apply a default.
James looks at the Istanbul record 22 storeys, USD 45 million TIV, construction: "Mixed." He raises an eyebrow. "Istanbul is basically sitting on a fault line. We need to know what that building actually is."
COPE is the standard framework the property insurance industry uses to describe what a building is and how it behaves under stress. The acronym stands for Construction, Occupancy, Protection, and Exposure. It predates modern cat modelling fire underwriters used exactly these four categories to assess fire risk before any quantitative models existed.
When cat modelling emerged after Hurricane Andrew in 1992, COPE was adopted as the primary input to vulnerability assessment. The fit was natural: construction determines how a building responds to wind, earthquake, and flood forces; occupancy determines what is inside and how the building is used; protection affects fire suppression capability; and exposure (TIV) determines the financial consequence of damage. Together, the four fields give a model the minimum information it needs to select a vulnerability curve and calculate an expected loss.
The structural system and primary materials how the building is built and from what. The single most important vulnerability driver for earthquake and wind. A wrong construction class can move modelled losses by a factor of 5β10.
What the building is used for. Determines contents density, operational profile, and specific hazard exposures. Also drives the model's assumptions about demand surge and business interruption loss.
Fire detection, suppression, and security systems. Less critical for pure nat cat runs but becomes material for fire-following-earthquake and for all-risk policies covering non-nat-cat perils alongside cat.
The insured financial value building replacement cost, contents, business interruption limit. This is TIV. It is the multiplier that converts a damage ratio into a dollar loss. Covered in full in Lesson 1.3.
Construction type is the most consequential single field in the entire SOV for catastrophe modelling. It is the primary driver of how a building responds to physical forces β whether those forces are lateral accelerations from an earthquake, wind pressures from a hurricane, hydrostatic loads from a flood, or radiant heat from a wildfire. Different structural systems respond to identical hazard intensities in dramatically different ways.
The KahramanmaraΕ 2023 doublet earthquake illustrates this with brutal clarity. Modern reinforced concrete shear wall buildings in the same city blocks survived with minimal damage while older RC frame structures and non-ductile concrete buildings collapsed catastrophically. Both were described as "reinforced concrete" in the exposure data. Both would receive the same initial construction class in a poorly prepared SOV. The vulnerability difference between them was enormous not because of the hazard, which was identical, but because "reinforced concrete" meant entirely different things in each case.
| Construction Class | Definition | Earthquake | Wind | Flood | Wildfire |
|---|---|---|---|---|---|
| Wood Frame | Timber studs and joists. Dominant in US residential, especially pre-1970s. | MODERATE | HIGH | HIGH | VERY HIGH |
| Unreinforced Masonry | Brick, stone, or block with no steel. Common pre-WWII globally and across much of Africa and South Asia. | VERY HIGH | HIGH | HIGH | MODERATE |
| Reinforced Masonry | Masonry with embedded steel reinforcing bars. Substantially more resilient than unreinforced. | MODERATE | MODERATE | HIGH | LOW-MOD |
| RC Frame Moment Resisting | RC columns and beams designed to resist lateral forces through ductile frame action. Post-code standard for earthquake-prone regions. | LOW | LOW | MODERATE | LOW |
| RC Frame Shear Wall | RC structure with reinforced concrete walls providing lateral stiffness. Most resilient against wind; excellent for moderate seismicity. | LOW | VERY LOW | MODERATE | LOW |
| RC Non-Ductile (Pre-Code) | RC built before modern seismic codes column ties too widely spaced, inadequate transverse reinforcement. Very common pre-1980 globally. | VERY HIGH | MODERATE | MODERATE | LOW |
| Steel Frame | Structural steel columns and beams. Highly ductile yields before fracturing. Standard in modern high-rise commercial construction. | LOW | LOW | MODERATE | LOW |
| Pre-Engineered Metal | Light-gauge steel frame with metal cladding. Common in industrial warehouses and retail sheds. | LOW-MOD | HIGH | MODERATE | LOW |
| Mixed / Unknown | Submitted as "Mixed," "Other," or blank. Model applies the most common regional default. Results unreliable. | DEFAULT | DEFAULT | DEFAULT | DEFAULT |
The earthquake vulnerability spread across these classes is the widest of any peril. At 0.3g peak ground acceleration a moderate-distance major earthquake the difference between modern RC shear wall and unreinforced masonry is a factor of roughly 10 in expected damage ratio. That is not a modelling nuance. At USD 45 million TIV, it is the difference between a USD 1.5 million loss and a USD 22 million loss from the same event.
Meridian's facilities register captures "Structure Type" with five options: Steel, Concrete, Brick, Wood, Mixed. Their facilities manager selected the closest match for each property. The result: a 1998 RC moment frame warehouse in Osaka appears as "Concrete." A 1963 unreinforced brick masonry apartment block in Munich appears as "Brick". A 2015 steel frame warehouse with precast concrete panels appears as "Mixed." None of these three descriptions is useful to a cat model. The Munich apartment block is the gap between "Brick" and the range of actual vulnerabilities it represents in earthquake is enormous. But all three require investigation before the run proceeds.
Occupancy describes the primary use of the building and determines the nature of what is at risk inside it. It affects vulnerability in at least three ways simultaneously: it drives the type and density of contents at risk, it shapes the model's assumptions about business interruption loss, and for some perils it directly modifies how hazard exposure is assessed.
The most common occupancy error is not submitting a wrong class β it is submitting a class that is too broad to be useful. "Residential" covers a detached timber-frame bungalow and a 42-storey reinforced concrete tower block. For wind modelling, the roof and structural profile are completely different. For earthquake, the resonance period is an order of magnitude apart. "Retail" covers a corner newsagent and a 50,000 mΒ² shopping mall with USD 80 million in anchor tenant stock. Occupancy sub-class matters, and pushing for it is always worth the effort.
| Occupancy Class | Key Risk Characteristics | Common Data Issues |
|---|---|---|
| Residential β Single Family | Lower contents density relative to structure; no BI element; wind and flood dominant. Roof type and first-floor elevation are the critical secondary modifiers. | Often not distinguished from mobile home or row house β meaningfully different wind vulnerability profiles. |
| Residential β Multi-Family | Higher occupancy density amplifies casualty risk in seismic events. Height matters for earthquake resonance; lower floors are flood-vulnerable. | "Multi-family" spans a 2-unit conversion and a 32-storey tower. Number of storeys is the essential companion field. |
| Office | Moderate contents density; daytime occupancy pattern; BI often significant β professional services firms lose revenue per employee per day. | Generally reliable. Main issue: ambiguity between owned and leased (affects TIV scope) and conflation with mixed-use. |
| Retail | High contents density in some sub-types; high public footfall; flood sensitivity high (ground-floor display areas). Tenant vs. landlord insurance split is complex. | "Retail" without GFA or sub-type is almost meaningless β a jewellery store and a supermarket have completely different contents profiles. |
| Hotel | High BI exposure β room revenues cease immediately on total loss. Complex contents (FF&E). High-rise hotels have specific earthquake and wind profiles. Post-event hotels often become emergency housing, extending the BI period. | Room count or star rating sometimes provided instead of GFA. BI limit must reflect daily revenue Γ realistic rebuild timeline β often severely underinsured. |
| Industrial / Warehouse | Contents can vastly exceed building value. Supply chain disruption from a single warehouse loss can trigger contingent BI claims far beyond the property itself. | "Warehouse" spans a cold storage facility and a large fulfilment centre. Sub-class (light, medium, heavy) matters significantly for contents assumption. |
| Mixed Use | Retail ground floor, residential or office above. The flood profile of the ground floor, the wind profile of upper floors, and the contents of both must be treated separately. | Very common in European and Asian markets. Rarely decomposed in the SOV. Model must apply a blended occupancy assumption unless the split is provided. |
For a pure nat cat model run covering earthquake and wind only, missing protection data is a low-priority gap. A sprinkler system cannot stop earthquake shaking or suppress a wildfire advancing from an adjacent hillside. For those perils, the Protection field has no direct effect on modelled loss.
Protection becomes material in two specific situations. The first is fire-following-earthquake β a component that major earthquake models include based on the historical record of post-earthquake urban fires, most famously in the 1906 San Francisco earthquake and again documented in the TΕhoku 2011 case study. Buildings with working automatic suppression systems suppress early-stage post-earthquake fires at far higher rates than unsprinklered buildings. The second situation is all-risk policies that cover fire, explosion, and malicious damage alongside nat cat perils. For these policies, sprinkler status, alarm type, and distance to the nearest fire station all contribute to the modelled loss distribution.
The practical rule: check your run configuration before deciding how urgently to fill protection gaps. If your model run covers nat cat perils only, protection data is genuinely not material. If it covers all-risk or includes a fire-following-earthquake module, protection gaps for high-value locations in seismic zones are worth resolving before the run proceeds.
Beyond the four COPE fields, every major cat model accepts additional parameters that further refine the vulnerability estimate. These are called secondary modifiers. Providing them can shift modelled losses by 15β50% relative to the COPE-only baseline β making them worth obtaining for high-value locations and portfolios where loss estimate accuracy is critical.
Determines which building code was in force at construction. Pre-1960 masonry and post-2000 RC frame are both just "Masonry" and "RC" without year built. The most important secondary modifier for earthquake.
Up to Β±40% earthquake AALDetermines the natural period of vibration β low-rise buildings respond to high-frequency ground motion; high-rise to long-period motion. Essential for earthquake curve selection. Also drives flood depth ratio and wind height exposure.
Up to Β±25% earthquake AALHip roofs perform dramatically better than gable roofs under wind loading. The single most important secondary modifier for hurricane and cyclone modelling.
Up to Β±35% wind AALFlood damage is a threshold phenomenon β no interior damage until water reaches the first floor. A building elevated 1m above grade requires a significantly deeper flood event to sustain interior loss.
Up to Β±50% flood AALBasements sustain near-complete contents loss in any inundation event. Their presence substantially increases flood loss for the contents component of TIV.
Up to Β±15% flood AALFor hail β the dominant loss driver in US SCS events as noted in the Severe Convective Storms 2023β2025 case study β roof cover is the most important input. Impact-resistant shingles reduce hail losses by 70β90% vs. standard asphalt.
Up to Β±70% hail AALUsed for TIV plausibility checks β TIV Γ· GFA produces a cost per mΒ² that can be benchmarked against market rates. A building where this figure is implausibly low signals a possible book-value TIV entry.
Primarily a QA fieldAverage shear-wave velocity in the top 30m of soil. Soft soils amplify earthquake ground motion; hard rock attenuates it. Most models derive this from global databases at the geocoded location, but site investigation data is always more accurate.
Up to Β±100% earthquake AAL in soft soil zonesYou rarely have time to obtain every secondary modifier for every location β so prioritise by the combination of hazard severity and financial exposure at each specific location.
When COPE fields are missing, the model does not refuse to run β it applies a pre-set default. Almost all defaults are calibrated to be conservative: they overstate rather than understate losses. This is deliberate β it is better for an insurer to price cautiously than to understate risk. The consequence for the analyst is that a portfolio with significant missing COPE data will produce higher modelled losses than the same portfolio with complete data.
This has a counterintuitive implication: improving data quality usually reduces modelled losses, not increases them. A high AAL from a poorly specified portfolio should never be taken as confirmation of high physical risk β it may simply reflect conservative defaults stacking up across dozens of records.
| Field Left Blank | Typical Default Applied | Loss Impact | Why |
|---|---|---|---|
| Construction class | Most common vulnerable class for region | +15 to +40% AAL | Default is conservative β almost always more vulnerable than the actual building |
| Year built | Pre-1980 in most models | +10 to +30% EQ AAL | Pre-modern-code assumption applied to a 2015 building dramatically overstates earthquake vulnerability |
| Roof shape | Gable (more vulnerable) for residential | +15 to +35% wind AAL | Gable default is conservative β overstates loss if actual building has hip roof |
| First floor height | At-grade (0m) | +10 to +50% flood AAL | At-grade maximises flood vulnerability β any elevation reduces expected loss |
| Occupancy class | Residential or light commercial | Β±10 to +30% AAL | Wrong occupancy drives wrong contents assumption and wrong BI profile simultaneously |
// Halcyon Syndicate 2247 Β· Ardent Re Brokers Β· 150 Locations Β· 13 Countries
You sit back down after James's question and run a COPE completeness check across all 150 records. Here is what you find β and what you decide to do about each gap.
| COPE Field | % Complete | Primary Issues | Priority |
|---|---|---|---|
| Construction | 77% | 34 records as "Mixed," "Concrete," "RCC," or blank β concentrated in Turkey, India, Nigeria, Colombia | HIGH β earthquake-exposed locations |
| Occupancy | 94% | 9 records as "Commercial" without sub-type; 3 mixed-use records with no residential/commercial split | MEDIUM β affects contents assumption |
| Protection | 41% | 59% blank β but this run covers nat cat perils only, so protection data is not used in the primary model | LOW for this run configuration |
| Year Built | 78% | 33 records missing β concentrated in older UK residential, Nigerian locations, and some South American properties | HIGH β code era critical for seismic vulnerability |
| Number of Storeys | 89% | 17 records missing β mostly single-storey industrial where the default (1 storey) is probably correct anyway | LOW β default likely appropriate for blanks |
| Roof Shape | 28% | 72% missing β matters significantly for Florida, Texas, Australia, and Philippines locations | HIGH β wind peril locations without roof shape |
| First Floor Height | 62% | 57 records missing β concentrated in Thailand, Germany, and UK flood-exposed locations | HIGH β flood-exposed locations |
You summarise these five records in a data request to Ardent Re, prioritised by financial impact. You also inform James that the initial model run will proceed with conservative defaults for the gaps β and that a revised run will follow once the priority data comes back. This two-run approach is standard practice for large accounts with incomplete COPE data.
"Concrete" is a material, not a structural system. Unreinforced concrete masonry and modern RC shear wall are both "concrete" β with earthquake vulnerability factors 7β10Γ apart. Always investigate for seismic-zone locations.
A building constructed in Chile in 1985 sits differently on the code timeline than a 1985 building in California. Year built is only meaningful when interpreted against the specific jurisdiction's building code revision history.
True for pure earthquake or wind runs. Not true for all-risk policies that include fire alongside nat cat β or for runs that include a fire-following-earthquake module. Always check the run configuration first.
"Residential" and "Retail" are not complete occupancy descriptions when the portfolio spans single-family homes to high-rise towers and corner shops to anchor-tenant malls. Sub-class moves both the vulnerability curve and the contents assumption.
Spending time on roof cover material for a German earthquake-exposed location while first floor height is missing for a Thai flood-exposed location is a prioritisation failure. Always match the modifier to the dominant peril at each specific location.
When revised COPE data reduces modelled losses, analysts sometimes worry they have made an error. In almost all cases, this is the correct and expected direction β conservative defaults inflate losses, and accurate data reduces them.
Construction, Occupancy, Protection, Exposure β the four standard physical and operational fields that drive vulnerability curve selection in every cat model.
A mathematical relationship describing expected damage ratio as a function of hazard intensity. Selected by the model based on construction class, occupancy, and secondary modifiers.
Brick, stone, or block construction with no embedded steel reinforcement. Among the most earthquake-vulnerable building types globally β responsible for the majority of earthquake fatalities in developing-world events.
Reinforced concrete built to pre-modern seismic codes β lacking the transverse reinforcement for ductile seismic response. Highly vulnerable in earthquake; common in pre-1980 construction globally.
Any exposure field beyond basic COPE that refines vulnerability curve selection β year built, storeys, roof shape, roof cover, basement, first floor height, and soil type.
The post-event increase in construction costs driven by simultaneous demand from many damaged properties. Cat models apply a demand surge multiplier; occupancy class affects its expected magnitude.
The building code in force at the time of construction. Buildings designed to pre-modern seismic codes are more vulnerable than post-modern equivalents even when described by the same construction class label.
Average shear-wave velocity in the top 30 metres of soil β the standard measure of site conditions for earthquake. Low Vs30 (soft soil) amplifies ground motion; high Vs30 (hard rock) attenuates it.
Five questions Β· 4 of 5 correct (80%) to pass