Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Home Grading Scale

Transcription

1 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Home Grading Scale Client: January 10, 2011

2 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Copyright 2011 AIR Worldwide. All rights reserved. Trademarks AIR Worldwide is a registered trademark. Contact Information If you have any questions regarding this document, contact: AIR Worldwide 131 Dartmouth Street Boston, MA USA Tel: (617) Fax: (617)

3 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Table of Contents List of Figures... 6 List of Tables... 6 Executive Summary... 7 Reliances and Limitations... 8 Distribution and Use... 8 Questions and Comments... 8 Introduction... 9 Limitations and Assumptions of Modeled Results Brief History of Florida Insurance Premium Factors for Wind Loss Mitigation Description of the AIR Hurricane Model Rationale for Modeling Catastrophic Events Basic Model Architecture Hazard: Hurricane Simulation Hazard: Local Intensity Vulnerability: Damage Functions Financial: Aggregating Losses Interpreting Model Output Model Validation Description of the AIR Individual Risk Model How the IRM Operates on Building Features Individual Risk Model Validation How Exposure Data Is Introduced to the Model Definition and Description of Mitigation Features Comparable to Those Underlying Home Grading Scale Roof Geometry

4 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Roof Covering and Attachments Roof Deck and Attachments Roof Anchorage Window Protection Secondary Water Protection Relationship to Defined Features Unmatched Mitigation Feature Definitions Year Built Modeling Approach Development of Notional Data Set Model, Catalog, and Analysis Options Determination of Loss Relativities by Wind Zone and Terrain Description of Calculation Choice of Base Property for Relative Loss Costs by Wind Zone Group, Using FHCF Exposure Distributions Using Relative Loss Costs to Formulate Actuarially Sound Premium Credits for Wind Loss Mitigation Incorporating the Effect of Mitigation on Premium Components Consideration of Overlap and Gaps with Existing Rating Factors Consideration of Applicability of Wind Loss Mitigation to Property Other than the Main Structure 45 Correlation of HSRS Scores to AIR Loss Cost Relativities AIR Score Ranking Details Relationship of Modeled Relative Loss Costs to ISO BCEGS BCEGS Grades and Wind Loss Mitigation Premium Credits Conclusion Appendices Appendix 1: FHCF Distribution of Structure Mitigation Features

5 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Appendix 2: Loss Cost Relativity Tables Appendix 3:

6 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale List of Figures Figure 1: Basic Catastrophe Model Architecture Figure 2: Major Parameters of a Simulated Hurricane Figure 3: Windfield Cross Section Figure 4: Terrain Effects on Wind Speed Profiles Figure 5: Representative Residential Damage Function Figure 6: Damage Distribution Convolution Logic Figure 7: Sample Exceedance Probability Curve Figure 8: Evaluation of Building Performance in the AIR Individual Risk Model Figure 9: Examples of Applying Mitigation Adjustment to Basic Damage Function Figure 10: Modeled vs. Actual Reduction in Vulnerability for Sample Company Data Figure 11: Illustration of Gable Roof Figure 12: Illustration of Hip Roof Figure 13: Roof Anchorage with Hurricane Ties Figure 14: Secondary Water Protection Figure 15: Map of Modeled Locations Figure 16: FBC Wind Speed Zone Contours List of Tables Table 1: Mapping of to AIR Building Features Table 2: Count of Locations Modeled by Terrain and Wind Zone Table 3: Listing of Mitigation Features Included Table 4: Relative Loss Costs of Six Base Properties Table 5: Sample Premium Breakdown

7 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Executive Summary This report was commissioned by the (OIR) to satisfy statutory requirements for the development of an updated method for insurers to assign premium adjustment factors for wind loss mitigation by February 1, The statute states that the method must be consistent with the existing Home Structure Rating System (HSRS), a separate consumer friendly 0 to 100 scale measuring the wind resistance of individual homes, developed by OIR in consultation with the engineering firm Applied Research Associates () and adopted in March In September 2010, AIR Worldwide (AIR), a subsidiary of Verisk Analytics, Inc., proposed to prepare this report in response to an OIR Request for Proposal (RFP) and was awarded a contract to do so by OIR in December AIR has produced relative loss costs, the core risk measure that actuaries must convert to premium differentials in rate filings, for a large number of possible types and locations of single family residential structures based on: Geographic location reflecting severity of hurricane risk by Florida Building Code Wind Speed zones and Terrain Exposure zones; Year of Construction reflecting the adoption of the first comprehensive statewide Florida Building Code in early 2002; Roof system reflecting detail regarding roof coverings, geometry, deck type, resistance to leaks, and anchorage to walls; Window protection reflecting various types of shutters and impact resistant glass. Relative loss costs measure the annual average risk of insurable losses by hurricanes compared to a base structure. AIR used statewide data on the prevalence of these features, collected annually by the Florida Hurricane Catastrophe Fund, to relate hurricane risk to a base defined as the most common type of structure in each area. AIR has also demonstrated that its relative loss costs are directly correlated to the HSRS as required by statute. Statistical ranking and regression techniques applied to each type of structure in all areas of Florida indicate that AIR s measure of hurricane risk is quite consistent with the 0 to 100 HSRS scale. Finally, AIR has offered examples of how to use the results in this report to derive actuarially sound premium differentials for wind loss mitigation, with appropriate consideration of all actuarial components of the insurance premium formula, as well as applicability to miscellaneous covered property and potential overlap with existing rating factors such the ISO Building Code Effectiveness Grading System. Actuaries and regulators should be able to use the results presented here to develop insurer specific or broadly applicable premium adjustment tables that are appropriate for insurance product manuals and require minimal adjustments to information systems and actuarial work papers supporting rates and rating factors. 7

8 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Reliances and Limitations This report relies on public technical documents supporting the current HSRS. AIR has made its best effort to interpret these work products accurately but cannot be responsible for the quality or accuracy of these documents. This report also relies on the AIR hurricane model, including its limitations on the use of model results as discussed in the Limitations and Assumptions of Modeled Results section. No party other than the OIR may rely on this report for its insurance rates, rating factors, and regulatory filings, and AIR accepts no liability for such reliance. Distribution and Use This report is solely for the use of the OIR. The final report may be distributed in accordance with Florida s public records laws. AIR requests that when distributed, the report be provided in its entirety, including all notes and exhibits, as individual sections or results read without context may be misleading. Questions and Comments Questions and comments regarding the report by outside parties should be directed to OIR, which will contact AIR at its discretion. The lead consultants associated with this report are available to answer questions about it. 8

9 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Introduction On September 15, 2010 the (OIR) issued a Request for Proposals (RFP) for Actuarial, modeling and consulting services to relate Insurance Discounts to the Home Grading Scale. More specifically, the RFP pertains to insurance premium discounts for windstorm loss mitigation, commonly known in Florida as mitigation credits. Discount factors considering the wind resistance of various construction vintages and features in various regions of Florida are already used by all insurers in compliance with Florida Statute (1)(a), and most insurers use a specific system of discounts developed by OIR and the Florida Department of Community Affairs (DCA) in consultation with Applied Research Associates () which has been referenced by regulation (Rule 69O , revised Sept. 2007). Independently and also in response to a statutory mandate (F.S ), a Uniform Home Grading Scale (UHGS) was developed, also by OIR in consultation with, and adopted in March 2007 as the Home Structure Rating System (HSRS) for Florida. This scale was designed to assign any given home a linear (0 to 100) wind resistance rating. Such a rating was seen as more practical for consumer education and real estate transaction purposes. The RFP underlying this report was issued in response to yet another statutory mandate (F.S (1)(b)), which requires a proposed method for insurers to establish...rate differentials for hurricane mitigation measures which directly correlate to the numerical rating assigned...[by] the Home Grading Scale. In short, the task for the vendor is to develop a system of wind loss mitigation factors that can readily be converted to premium differentials for insurance rating, yet are consistent with the adopted linear Home Structure Rating System. AIR Worldwide (AIR) responded to the RFP and was awarded the contract in early December of AIR has performed the work in accordance with the contract terms and RFP and presents the results in this report. AIR leveraged our existing expertise on the subject, developed over more than twenty years of experience in simulating hurricanes and their financial impacts on insurable structures and related assets. Meteorologists, statisticians, civil and structural engineers, construction managers, actuaries, software developers, and other professionals work together to create a catastrophe modeling architecture in which embedded scientific expertise within each component is used to simulate thousands of years of storm activity and its impact on any set of properties real or hypothetical. The model considers the hurricane parameters, develops a wind footprint for each simulated event, and applies each event s wind speed and duration to every property described by geographic location and building characteristics entered into the software. Engineering calculations based on those wind loads then determine the damage ratio to each location, and actuarial calculations based on the damage ratio and the stated replacement value and insurance coverage of the property are used to translate the raw damage estimates into insurable losses. 9

10 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Limitations and Assumptions of Modeled Results Although AIR s simulation methodology is a state of the art technique for estimating potential catastrophe losses, the method has certain limitations. It is based on mathematical and statistical models that represent real world systems. As with all models, these representations are not exact. The simulated events generated by the AIR model do not represent catastrophes that have occurred, but rather events that could occur. The AIR model relies on various assumptions, some of which are subject to uncertainty. Accordingly, the loss estimates generated by the model are themselves subject to uncertainty. AIR refines and updates its model assumptions from time to time in light of new meteorological and other information as and when such information becomes available. Such refinements and updates may materially alter the loss estimates generated by the AIR model. Note that extreme occurrence losses are possible, though they have a very low probability of occurrence. The largest simulated event losses do not represent worst possible scenarios. The loss estimates and their associated probabilities are estimates of the magnitude of losses that may occur in the event of such hazards; they are not factual and do not predict future events. Actual loss experience can differ materially. The estimates are intended to function as one of several tools for use in analyzing estimated expected and potential losses from certain hazards. The assumptions that AIR used in creating them may not constitute the exclusive set of reasonable assumptions and methodologies. The use of alternative assumptions and methodologies could yield materially different results. 10

11 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Brief History of Florida Insurance Premium Factors for Wind Loss Mitigation To better understand the need for this report and the relationship between the UHGS/HSRS and mitigation credits used in insurance rating manuals, it is appropriate to provide some background on the history of mitigation credits in Florida. Though enforcement of a building code (from one of four minimum standard models approved by the state) went into law in 1974, and was later revised to be a single statewide code with state oversight 1, insurance mandates relating to loss mitigation for existing structures did not take effect until after Hurricane Andrew. In 1993, the Florida Legislature passed F.S and required insurers to offer shutter discounts or deductible reductions for fixtures designed to reduce hurricane losses. The regulatory rules associated with this law became known as the shutter discount rule since hurricane shutters were the most prevalent loss preventive technique. The Department of Insurance (now OIR) issued Rule of the Florida Administrative Code (FAC) in 1997, which required premium discounts for shutters or other wind mitigation devices or fixtures that were at least equal to discounts issued by any statewide rating organization. The requirement was functionally equivalent to requiring discounts to be at least equal in magnitude to the Insurance Services Office (ISO) discounts. F.S was amended in 2000 to provide that rate filings for residential property insurance must include actuarially reasonable (rather than appropriate ) discounts, credits, or other rate differentials, or appropriate reductions in deductibles, for properties on which fixtures or construction techniques demonstrated to reduce the amount of loss in a windstorm have been installed or implemented. 2 The amendment stated that fixtures or construction techniques shall meet the requirements of the Florida Building Code and shall include fixtures or construction techniques which enhance roof strength, roof covering performance, roof to wall strength, wall to floor tofoundation strength, opening protection, and window, door, and skylight strength. 3 In 2002, a later modification also instructed all insurance companies to make a rate filing reflecting these measures by February 28, Applied Research Associates (), contracting with the Florida Department of Community Affairs (DCA), completed two studies in 2002 to quantify hurricane loss reduction for wind mitigation 1 August% pdf /0627/Sections/ html /0627/Sections/ html 11

12 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale construction features. Development of Loss Relativities for Wind Resistive Features of Residential Structures focused on single family homes and Development of Loss Relativities for Wind Resistive Features for Residential Buildings with Five or More Units addressed condominium and renter occupancies in buildings with five or more units. The Florida OIR issued Informational Memorandum OIR M on January 23, 2003, which required property insurers in Florida to use the mitigation credit tables created by the OIR from the loss relativity tables in the 2002 studies. The Memorandum was an Implementation of Revision to Section (1), F.S., and suggested sets of credits for new and existing construction that were determined and tempered by 50%. For example, a loss relativity of.60, indicating that a certain reinforced structure is expected to incur 40% lower losses than a given base structure, was converted to an OIR credit of 20%. In the words of the Memorandum, This tempering was applied in view of the large rate changes which might otherwise be induced, the approximations needed to produce practical results (such as the specifications of the houses used for modeling and the number of rating factors used), and the potential for differences in results using different hurricane models. 4 In 2005, the Florida Legislature required insurance companies to notify homeowners of the availability of and the range of each premium discount, credit, other rate differential, or reduction in deductibles, and combinations of discounts, credits, rate differentials, or reduction in deductibles, for properties on which fixtures or construction techniques demonstrated to reduce the amount of loss in a windstorm can be or have been installed or implemented. 5 In the 2006 regular session, the Legislature modified F.S (1)(a) to require OIR to reevaluate the mitigation credits in use. In response, in October 2006, Rule 69O F.A.C. was amended to require insurers to make new rate filings that doubled the credits to 100% of the indicated value, or provide alternatives justified by a detailed alternate study with all assumptions available to the Office for review, by March 1, Informational Memorandum OIR 07 03M was issued February 27, 2007, stating windstorm mitigation discount filing shall not include any modification of the rating factors or base rates for any purpose, including the offset of revenue impact on current business. 6 Also in 2006, Chapter , Section 39, Laws of Florida, required the OIR to develop a program that provided an objective rating system reflective of a property s relative ability to withstand wind loads. In response to this, presented a Home Structure Rating System (HSRS) to the OIR in March The HSRS was created to provide a measure of the ability of a structure to withstand severe tropical windstorm or hurricane forces. Based on loss cost relativities, the HSRS produced by the was designed for consistent interpretation throughout the state of Florida and each of the eight wind zones m.pdf /0627/Sections/ html M.pdf 12

13 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale After the 2004 and 2005 hurricane seasons, the Legislature instructed the OIR to conduct another mitigation study; in 2008, completed OIR s study, called 2008 Florida Residential Wind Loss Mitigation Study. This study evaluated windstorm loss relativities for construction features, involved analysis of damage and loss data from the seasons, and included new engineering load and test data analyses to update the mitigation discount relativities. 7 The Legislature in 2009 amended F.S to require the Florida Commission on Hurricane Loss Projection Methodology (FCHLPM) to undertake a study of the implementation of mitigation credits under F.S during late FCHLPM delivered its Windstorm Mitigation Discounts Report to the legislature on Feb 1, This report summarized public testimonies, as well as delivered sets of recommendations on wind mitigation credits. In March of 2010, Risk Management Solutions, Inc. (RMS) was commissioned by the Florida Department of Financial Services to assess the impact of wind mitigation credits on the Florida insurance market in order to improve the management and mitigation of hurricane risk in Florida. RMS proposed that wind mitigation credits are not operating as intended and had a number of recommendations for insurance companies and the legislature pdf 13

14 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Description of the AIR Hurricane Model Rationale for Modeling Catastrophic Events Natural catastrophes such as hurricanes have tremendous impacts on society in many ways, from the immediate loss of life and property to far reaching and long lasting impacts on the economy and insurance system. Fortunately, large loss events are relatively infrequent. Unfortunately, they are also severe in impact and unpredictable. It is the combination of these three characteristics that make the estimation of damage and insurance losses from future catastrophes so difficult. The scarcity of historical loss data makes standard economic and actuarial techniques of loss estimation inappropriate for catastrophe losses. Further, the relevance of the loss data that does exist is limited because of the constantly changing landscape of properties. Property values change, sometimes rapidly, along with the costs of repair and replacement. Building materials and designs change, and new structures may be more or less vulnerable to wind loads than were the old ones. New properties continue to be built in areas of high hazard. Finally, it is difficult to understand how losses emerge and should be financed over long periods of time, because storms can strike in any year. AIR developed catastrophe modeling technologies as alternatives to the actuarial and rule of thumb approaches that had previously been relied upon for estimation of potential catastrophe losses. The technical expertise of meteorologists, other physical scientists, engineers, statisticians, actuaries, and computer technology specialists is augmented by the years of experience that AIR has accumulated in this field and integrated into a system for modeling the impact of potential events. The result is the delivery of reliable and credible loss estimates needed to make informed risk management decisions. Basic Model Architecture AIR developed the first commercial software application for catastrophe modeling based on sophisticated stochastic simulation procedures and powerful computer models of how natural catastrophes behave and act upon the man made environment. The years since have seen the models undergo a continual process of review, refinement, enhancement, and validation. The ongoing research ensures that the models incorporate the latest advances in the scientific, engineering, mathematical and other fields that are pertinent to their development. Figure 1 illustrates the component parts of the AIR state of the art catastrophe models. It is important to recognize that each component, or module, represents both the analytical work of the research scientists and engineers who are responsible for its design and the complex computer programs that run the simulations. 14

15 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Figure 1: Basic Catastrophe Model Architecture Hazard: Hurricane Simulation The event generation module determines the frequency, overall intensity, and other characteristics of potential hurricanes by geographical area. This requires a thorough analysis of the characteristics of historical events. Event generation begins by collecting the available scientific data pertaining to these parameters from many different sources. The data are cleaned and verified. When data from different sources conflict, a detailed analysis and the use of expert judgment is required before they are suitable for modeling purposes. After rigorous data analysis, AIR researchers develop probability distributions for each of the parameters, testing them for goodness of fit and robustness. Figure 2 shows the major parameters of a simulated hurricane. The selection and subsequent refinement of these distributions are based not only on the expert application of statistical techniques, but also on well established scientific principles and an understanding of how hurricanes behave. Figure 2: Major Parameters of a Simulated Hurricane 15

16 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale These probability distributions are then used to produce a large catalog of simulated storms. By sampling from these distributions, the model generates simulated years of activity. A simulated year represents a hypothetical year of catastrophe experience, but one that realistically could happen in the current year. AIR allows for the possibility of multiple events occurring within a single year. That is, each simulated year may have zero, one, or multiple events, just as might be observed in an actual year. Many thousands of these years are generated to produce a complete and stable range of potential annual experience of hurricane activity, and to ensure full coverage of extreme (or tail ) events, as well as full spatial coverage. Hazard: Local Intensity Once the model generates the characteristics of a simulated storm, it propagates the event across the affected area. For each location within the affected area, local intensity a function of a variety of local conditions is estimated. Researchers base local intensity formulae on empirical observation as well as on theoretical relationships between the variables. The wind profile of a hurricane over space, the area affected by the footprint, and time, the duration of damaging winds at each location, is developed using the following simulated parameters for each storm: 1. Frequency of occurrence: number of storms affecting each area of the coastline in a season; 2. Landfall location: number of landfalls affecting each coastal segment of 50 nautical miles; 3. Heading at landfall: compass direction of movement of the storm center at landfall; 4. Minimum central barometric pressure: intensity of storm at landfall; 5. Radius of maximum winds: distance from center of storm to the eye wall; 6. Forward speed: velocity of the storm center at landfall. 7. Gradient wind reduction factor: the relative scaling of the wind speed at 10 meter height to that measured at flight level. These meteorological characteristics are used to simulate the storm s movement along its track. Calculations of local intensity take into account the effects of the asymmetric nature of the hurricane windfield, storm filling over land, surface friction, and relative wind speed profiles. The generation of local wind fields is a complex procedure requiring the use of many variables. First, the maximum over water wind speed is calculated. Adjustments are then made for asymmetry effects, filling, surface friction, and relative wind speeds as a function of distance from the eye. In the Northern Hemisphere, hurricane winds rotate in a counter clockwise direction. The combined effects of hurricane winds and forward motion will produce higher wind speeds on the right hand 16

17 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale side of the storm. The model accounts for the dynamic interaction of translational and rotational speeds, and the inflow angle. The effect is visualized in Figure 3. Storm Path Eye Left Side Right Side Stronger Winds Weaker Winds R max Storm Center Figure 3: Windfield Cross Section As the storm moves inland, its intensity begins to dissipate. Central pressure rises and the eye of the hurricane begins to fill as it moves away from its energy source, warm ocean water. The model s filling equations are functions of the geographic location (particularly distance from coastline), and time elapsed since landfall. Rates of fill also vary by region, as is consistent with historical observation. Differences in surface terrain also affect wind speeds. Wind velocity profiles typically show higher wind speeds at higher elevations, as depicted in Figure 4 below. Winds travel more slowly at ground level because of surface friction. In the AIR model, the initial step in calculating the friction coefficient for each point of interest is to obtain an estimate of the surface roughness at the site. At ground level, horizontal drag forces induced by the surface roughness are exerted on the wind flow, causing retardation of the wind near the ground. The surface roughness is estimated based on high resolution digital USGS land use/land cover data. The land use/land cover categories vary from urban or built up land, to agricultural land, to forest land or wetlands, to water. Each terrain type has a different roughness value that will lead to different frictional effects on wind speeds. In general, the rougher the terrain the larger the frictional effect on wind speeds. The magnitudes of the friction coefficients in the AIR method are consistent with accepted scientific literature and empirical hurricane wind speed data. 17

18 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Figure 4: Terrain Effects on Wind Speed Profiles Vulnerability: Damage Functions AIR scientists and engineers have developed mathematical relationships called damage functions, which describe the interaction between buildings structural and nonstructural components as well as their contents and the local intensity to which they are exposed. Damage functions have also been developed for estimating time element losses. These functions relate the mean (average) damage level, as well as the variability of damage, to the wind speed profile at each location. Because different structural types will experience different degrees of damage, the functions vary according to construction type and occupancy. Damage functions also vary with year built to reflect changes in building codes, construction practices, structural aging, and upgrading over time. As illustrated in Figure 5, the AIR model estimates a complete distribution around the mean level of damage allowing for the possibility of zero damage and 100% or total loss damage for each local intensity and each structural type. Figure 5: Representative Residential Damage Function 18

19 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Losses are calculated by applying the appropriate damage function to the replacement value of the insured property. This fact has an important implication, which is that modeled losses are only as good as the replacement values which underpin them. If replacement values are understated by 50%, the modeled losses will be as well. This is why using exposure data of the highest possible quality and completeness is critical to obtaining the most reliable model results. In the case of building damageability, the damage ratio is the dollar loss to the building divided by the corresponding replacement value of the building. The contents damage ratio is the dollar loss to the contents divided by the replacement value of the contents. In the case of time element, the damage ratio represents per diem expenses or business interruption losses associated with the expected number of days that the building is uninhabitable (residential) or unusable (commercial). Time element damageability is a function of the mean building damage, as well as the time it takes to repair or reconstruct the damaged building. The functional relationship between building damage and loss of use is established using detailed published building construction and restoration data and on engineering judgment. Estimated time element losses have been validated using actual time element loss data from client companies. The AIR tropical cyclone damage estimation module develops a complete time profile of wind speeds for each location affected by the storm, thus capturing the effects of wind duration on structures as well as the effect of peak wind speed. Design loads are routinely exceeded in tropical cyclones of even moderate intensity. With no reserve strength, a fastener or connector that has been pulled out or pulled through as a result of uplift load can compromise the integrity of the building envelope. Wind damage manifests at the weak links in a structural system. As each connector is overwhelmed, loads are transferred to the next point of vulnerability. The longer the duration of high winds, the longer this process continues and the greater the resulting damage. Calculating damage only when winds are at their maximum and applying a single damage ratio to the full replacement value would completely disregard the cumulative effects of prolonged winds. The vulnerability relationships in AIR s model incorporate the results of well documented engineering studies, tests, and structural calculations. They also reflect the relative effectiveness and enforcement of local building codes. AIR engineers refine and validate these functions through the use of postdisaster field survey data and through exhaustive analysis of detailed claims data as it becomes available after events. Financial: Aggregating Losses Raw damage estimates must be collected for each coverage at each location within each policy, and proper policy terms applied, to yield the insurable losses for the exposure at hand. The final component of the catastrophe model, the financial module, aggregates damage and converts to losses as viewed from whatever perspective is relevant first dollar or ground up, or paid by an insurance company after deductibles and policy limits gross or net of reinsurance. Policy conditions may be 19

20 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale rather complex, including deductibles by coverage, site specific or blanket deductibles, coverage limits and sublimits, loss triggers, coinsurance, attachment points and limits for single or multiple location policies, and risk specific reinsurance terms. However, in this analysis, the main perspectives of interest are ground up losses and insured losses gross of a standardized industry wide deductible. Aggregation of losses while preserving the uncertainty in the simulated damage distribution is not as easy as just adding losses by coverage/location/policy; a statistical technique called numerical convolution is applied to each simulated storm loss for each policy/location to get an accurate representation of the total loss, including the variability of the total loss around its average value. The technique is illustrated in Figure 6. f ( η) f ( η) ( η) a b f c h( x) Figure 6: Damage Distribution Convolution Logic The analytical solution obtained is the most theoretically sound and yields the most precise results. Interpreting Model Output The results of simulating the entire catalog against the exposure at hand are condensed into outputs of use to risk managers. The fundamental result is the complete probability distribution of losses, also known as exceedance probability (EP) curves. EP curves can be assembled on two bases: occurrence (largest event in each simulated year) or annual aggregate (sum of all losses in each simulated year). They can also be viewed from the ground up, gross, or net loss perspective. The probabilities can also be expressed as return periods. Sometimes this is more intuitive. That is, the loss associated with a return period of 20 years is likely to be exceeded in only 5 percent of years, or on average, in one year out of twenty. Figure 7 illustrates a sample (not applicable to this analysis) EP curve. Each potential loss amount (on the horizontal axis) can be paired with the probability that occurrence or annual losses meet or exceed that amount (on the vertical axis). 20

21 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Figure 7: Sample Exceedance Probability Curve Output may be customized to any desired degree of geographical resolution down to location level, as well as by line of business, construction class, coverage, and other factors. The model also provides summary reports of exposures, comparisons of exposures and losses by geographical area, and detailed information on potential large losses caused by extreme tail events. Model Validation AIR s rigorous validation process is not just limited to the final model results. Throughout the model development process, every component is carefully verified against data from historical events. Of course, the goal of catastrophe models is not simply to replicate the historical record; the model should reflect the full range of potential future catastrophe experience, including the most extreme events events that may not have occurred historically. Therefore it is critical that the model be vetted and validated by the domain experts both internal and external for each model component to ensure reasonability. 21

22 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Description of the AIR Individual Risk Model Within the vulnerability module of AIR s hurricane model is a specific subsidiary model for individual risk analysis. This is the part of the modeling architecture that operates on what catastrophe modelers call secondary modifiers, or detailed construction and wind loss mitigation features those beyond the overall construction type, year built, and height introduced to the model in a property data record. It may be visualized as a smaller box within the Engineering module in basic model architecture shown in Figure 1. The Individual Risk Model (IRM) was developed using a knowledge based expert system and a structured approach. Based on structural engineering expertise and building damage observations made in the aftermath of historical hurricanes, several building features have been identified as having a significant impact on the losses. Options or categories for each feature are identified based on construction practice. Algorithms for modifying the vulnerability functions 8 are developed based on engineering principles and building performance observations. The modification captures the changes to building vulnerability that result when certain features are present and when information on such features is known. The function varies with the wind intensity to reflect the relative effectiveness of a building feature when subjected to different wind speeds. The first step in the development of the individual risk model is to identify building and environmental characteristics that impact the performance of a building in high winds. These features are selected based on academic research and controlled experiments, as well as knowledge of building performance in high winds as obtained in the course of hurricane damage surveys. The AIR model includes four groupings of building and environmental features that influence damageability. These are: Nonstructural features (e.g. wall siding); Structural features (e.g. roof and wall systems); General building features (e.g. building condition, occupancy); Environmental features (e.g. tree exposure, surrounding terrain). Note that the IRM supports any combination of multiple building features and produces an integrated modification function to the vulnerability function. In other words, the IRM does not operate in a simple additive or multiplicative fashion every combination is different. 8 The IRM modifies damage functions for building coverage (structure). Since damage to contents and time element coverages is dependent upon the building damage, changes in the building damage also modifies the damage to contents and time element coverages. 22

23 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Figure 8 shows a schematic of the IRM. The model considers each building to be an assemblage of many building components. Each building component can have many subcomponents. For example, a roof system is made up of subcomponents for roof covering, roof decking, and roof attached structures. Each building component and its subcomponents are associated with certain building features that can affect the overall vulnerability. Figure 8: Evaluation of Building Performance in the AIR Individual Risk Model Embedded in the IRM are two primary metrics rates and weights for evaluating the impact of a building feature (e.g., roof covering type) or building environmental feature (e.g., tree exposure) on overall building performance. The rate is a weighted value assigned to the various options for building or environmental features. The rate for any given option of a particular feature reflects the relative prevalence of use among the available options, and is independent of other features. That is, the value is designed such that the most commonly used option is assigned a value close to 1.0. The implication is that a building with this option is expected to perform very similarly to the average, or typical building represented by the base damage functions. An option that is considered to be more vulnerable (i.e. less wind resistive) than the more commonly chosen options will be assigned a rate greater than 1.0. That is, a building with this option will be more vulnerable than the average building to damage from wind. Similarly, the rate assigned for an option 23

24 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale that is considered to be less vulnerable (i.e. more wind resistive) than the most prevalent one will be less than 1.0. Such a building will be less vulnerable than the average building. The value for a given option is a constant. If no information is available on the option, the default value is 1.00, which means that the base damage function is used without modification. The second metric, the weight, is a value of one of two types. The first weight type is used to develop simple weighted averages that are used to evaluate the loss contribution of several features that together constitute a system, such as roof. They are wind speed dependent; the contribution of each feature varies with wind speed. For example, a roof may consist of three features: roof covering, roof deck and roof attachment. The loss contribution to the roof system from these three features is expected to be different at different wind speeds. At low wind speeds, the roof covering drives the damage since it is at relatively low wind speeds that damage to roof covering occurs. As wind speeds increase, the roof deck becomes vulnerable. In this case, roof deck failure will result in loss of roof covering regardless of the type (or option) of roof covering present. Therefore, as wind speed increases, the weight for roof deck increases. In contrast, at higher wind speeds, the weight for roof covering decreases because it is already lost. The sum of the weights for a system should add up to 1.0. The second type of weight metric is used to combine the effects of features whose interaction is complex and not necessarily additive. These are introduced to evaluate features that modify the performance of the system. Consider the roof system as an example; the age, pitch, and geometry of the roof all modify the performance of the system as a whole. Hence the weight should be used as a multiplier. Weights are dependent on wind speed and construction class, and are appropriately selected to reflect the importance of a feature at certain levels of a building s damage state. The general nature of the IRM framework allows estimation of the loss relativities for a wide variety of building types prevalent in hurricane prone areas. To estimate loss relativities for a building built to a specific building code, appropriate features and the options for those features can be selected in the analysis to estimate the building damageability. The Individual Risk Model has been developed and reviewed by wind engineering experts, and has been accepted by the Florida Commission on Hurricane Loss Projection Methodology since its inception in How the IRM Operates on Building Features Figure 9 illustrates the application of modifications for storm shutters to the basic damage functions for a given AIR construction type code in this case, wood frame. Modifications for non engineered shutters (e.g. removable plywood covers) and engineered shutters (e.g. metal roll up type) are stored in the IRM. A property known to have no storm shutters will be modeled as slightly more vulnerable than vulnerability assumed for the basic damage function, raising its relative losses and showing a debit for lack of mitigation, but a property with non engineered shutters will be modeled as somewhat 24

25 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale less vulnerable and show a mitigation credit in the relative losses, and of course a property with engineered shutters will show even less vulnerability and a higher mitigation credit. It should be noted that there is a varying marginal impact of input of incremental secondary features into the model, as the current AIR model assumes different default secondary features depending upon the location and year built of a structure. Figure 9: Examples of Applying Mitigation Adjustment to Basic Damage Function Individual Risk Model Validation To validate the IRM, AIR s engineers conduct detailed damage survey investigations, analyze claim reports and utilize engineering judgments. Damage surveys as well as engineering analyses indicate that building features such as opening protections can significantly mitigate building damage. Due to insurance incentives in some states for homeowners for the use of mitigation features, many insurance companies have started collecting detailed building feature data in those areas. Over time, it is expected that claims data will include more detailed information regarding these features. For example, AIR has received claims data for Florida from the 2004 and 2005 hurricane seasons that contains detailed building features information. Figure 10 shows the comparison between modeled 25

26 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale and actual percentage reduction in vulnerability due to mitigation features, using sample company data. It is clear from the figure that all of the key mitigation features provide significant reductions in vulnerability. Though there can be variation in actual reductions seen in any given data set, modeled results are similar to those observed in the data. Figure 10: Modeled vs. Actual Reduction in Vulnerability for Sample Company Data How Exposure Data Is Introduced to the Model Each structure subject to the analysis is input into the model in the form of a data record. The AIR data format includes fields for the following major attributes of a property: Identifying information (policy and location number, line of business); Geographic location (street address, city and ZIP Code which the model converts to a latitude longitude pair, or geocode ); Replacement value (by location and coverage); Insured values and policy terms (limits and deductibles by coverage, how the deductibles are applied; any sublimits, layers, or per risk reinsurance by location and policy); Primary property attributes (construction type, occupancy type, year built, height in number of stories); Secondary modifiers to property attributes (mitigation features, detail of surroundings). User input of secondary features will override the default assumptions for a given year and location of a property in the model. By varying the construction types, heights, and mitigation features in otherwise equal property data records, we are able to determine the effect of the IRM on loss costs and thus the indicated insurance benefits associated with more or less reinforced structures in various locations in Florida. 26

27 Florida Insurance Discounts for Wind Loss Mitigation and Relationship to the Florida Uniform Grading Scale Definition and Description of Mitigation Features Comparable to Those Underlying Home Grading Scale The following sections outline the set of construction features determined most relevant to the scope of this study. Roof Geometry A roof s geometry largely determines the magnitude of aerodynamic loads experienced by the roof when exposed to hurricane force winds. The geometry affects the intensity of wind pressures and the resulting uplift resistance. The roof geometry categories modeled in this study are as follows: Gable without end bracing a gable roof slopes in two directions so that the end formed by the intersection of slopes is a vertical triangle (see Figure 11). A gable roof without end bracing has no additional bracing securing the overhanging ends to the main structure. Figure 11: Illustration of Gable Roof Hip this roof slopes in four directions such that the end formed by the intersection of slopes is a sloped triangle. A hip roof, shown in Figure 12, generally resists wind loads better than unbraced or braced gable roofs. Figure 12: Illustration of Hip Roof 27