EDUCATIONAL NOTE EARTHQUAKE EXPOSURE COMMITTEE ON PROPERTY AND CASUALTY INSURANCE FINANCIAL REPORTING

Transcription

1 EDUCATIONAL NOTE Educational notes do not constitute standards of practice. They are intended to assist actuaries in applying standards of practice in specific matters. Responsibility for the manner of application of standards in specific circumstances remains that of the practitioner. EARTHQUAKE EXPOSURE COMMITTEE ON PROPERTY AND CASUALTY INSURANCE FINANCIAL REPORTING JANUARY Canadian Institute of Actuaries Ce document est disponible en français Canadian Institute of Actuaries Institut Canadien des Actuaires

2 Canadian Institute of Actuaries Institut Canadien des Actuaires MEMORANDUM TO: All Members of the Canadian Institute of Actuaries Interested in Property and Casualty Matters FROM: Christopher Townsend, Chairperson Committee on Property and Casualty Insurance Financial Reporting DATE: January 26, 1998 SUBJECT: Earthquake Eposure Educational Note In their 1997 Instructions for Actuarial Reports on Property and Casualty Business, the Office of the Superintendent of Financial Institutions (OSFI) has included a new Section 4.5, Earthquake Eposure. The section reads as follows: While it is not epected that the author of the report will perform a quantitative review of the company s management and monitoring program with respect to earthquake eposure, it is nevertheless important that the author provides a qualitative summary description of the company s program including, but not limited to, the following: data quality the type of modelling approach used (if any) basic assumptions methodology In this regard, it is epected that the author of the report will discuss earthquake matters with the company s management, the personnel (if any) having epertise in this area, and the eternal auditors of the company. The author of the report should comment on the reasonableness of the company s procedures. If the author of the report believes that unusual problems are not known to eist, that belief should be clearly epressed. The attached educational note has been prepared by a subcommittee of the Committee on Property and Casualty Financial Reporting. It is intended to assist actuaries in complying with OSFI s request. The actuary should also refer to OSFI Guideline B-9, Earthquake Eposure Sound Practices, together with any annees or appendices. The actuary will note that some key components of Guideline B-9 are not fully effective until December 31, Accordingly, it would be appropriate to focus on the company s plans or progress towards complying with Guideline B-9 in the 1997 actuarial report. CT Secretariat: 360 Albert Street #820, Ottawa, Ontario, Canada, K1R 7X7 (613) FAX: (613)

3 Subcommittee of the Committee on Property and Casualty Insurance Financial Reporting France Fortin Brett A. MacKinnon Patrick Maillou (Chairperson)

4 EARTHQUAKE EXPOSURE INTRODUCTION Scientists have developed the theory of Plate Tectonics to eplain the origin of earthquakes. Essentially, plates of solid rock that include continents float on top of the earth s molten core. Almost all of the world s earthquakes occur along lines (or a boundary) joining one crustal plate with another. There are three types of boundary plates where earthquakes occur: neutral, diverging, and converging. A neutral boundary plate is where one plate slides along another. This is the type of movement that occurs along California s San Andreas Fault, and was responsible for the Northridge quake. Diverging is a situation where plates are moving away from each other. Finally, converging, also known as subduction, is when one plate pushes under another. The plate being pushed down is the subducting plate and the region in which this activity occurs is called the subducted zone (British Columbia scenario). Another category of earthquakes which is of concern in Canada is caused by the comple interaction of plates pushing off each other, which can result in the occurrence of an intraplate earthquake. This is the case with earthquake activity in the Québec Charlevoi region. The impacts of an earthquake are often described using scales commonly known as Richter and Modified Mercalli (MM). Modern seismographic instruments permit the measurement of the period and ground amplitude of the energy wave from an earthquake. By making allowance for the distance of the earthquake from the recording station and its depth, this recorded amplitude can be translated into the magnitude of an earthquake. The magnitude (M) so recorded is measured on the logarithmic Richter Scale, with each one point (1.0) increase representing a 10-fold increase in the amount of ground shaking and a 32-fold increase in the amount of energy that is released. The effect of an earthquake, as opposed to its magnitude, is measured with the help of the Modified Mercalli Scale (MM). It shows the intensity of a quake (i.e., what can be felt and observed). The earthquake insurance peril consists of many components, of which two are considered most significant: shake, which is damage caused by the actual physical shaking, and fire following, or fire damage caused from a fire initiated by the shaking. Within each of the Canadian Common Law jurisdictions, shake cover is provided as additional cover by endorsement, whereas the fire following is deemed by statute (the Insurance Act) to be included within the definition of fire. In the Province of Québec, shake is treated the same; however, with respect to the fire following, the civil code allows insurers to eclude fire following altogether. If the coverage is requested by the insured, then the insurers are able to charge a separate premium and use a different deductible. In actual business practice, P.Q. insurers are also providing fire following cover. Canada has a number of zones of significant seismic activity. The main ones, because of population and value concentrations, are Vancouver, B.C. (including Victoria), Montréal, P.Q. (including Ottawa Valley) and Québec City, PQ (see attached map). Insurance companies that write business in these areas, therefore, need to assess the etent of financial loss that could result from an earthquake occurrence. In order to determine the amount of earthquake risk being assumed, the insurer must eamine the frequency and severity of possible events. Net, the eposed values need to be determined, and, based on the insurance policy conditions in force, the likely loss impact on the insured portfolio is then estimated. The amount so determined is the probable maimum loss (PML) amount. Improvements in technology and research data have greatly improved the ability to quantify this figure. Through the use of computer models, a multitude of calculations are possible in order to simulate the effects of an earthquake event on an insurance portfolio. Despite improvements, our knowledge of earthquakes is still very limited (i.e., we cannot predict when they will happen and PML models can only produce estimates of actual loss impact). Also, often the quality of the data being input in models and the lack of seismic activity in Canada make scenarios educated guesses.

5 MODELS A number of computer models have been developed by various sources. Historically, there are two concepts of seismic hazard analysis: deterministic, and the more common probabilistic analysis. Deterministic models assume a specific earthquake scenario including a given event return period, calculate the level of ground motion at the site of interest and then estimate a PML. Probabilistic models, on the other hand, consider a number of possible earthquake scenarios based on potential earthquake sources that could affect the site, model their magnitude, frequency characteristics, and calculate the ground motion at the site, including variability in the attenuation. Probabilistic models produce an estimate of the possible loss severity distribution showing a loss return period associated with each loss amount. There are numerous companies specializing in this field, and some reinsurers have also developed proprietary models. Also available as an alternative is the IBC proposed Benchmark, which simply applies percentage factors to the total insured values within certain zones. This method of determining a PML estimate was designed to have a conservative bias. It is, nonetheless, an alternative for those companies that have too little eposure to justify the epense of a computer model. For those companies using computer models, the IBC Benchmark can also be used as a second indication to complement model estimates. COMMON MODEL PARAMETERS To be able to assess the risk from earthquakes, the four elements that must be considered are: Hazard past seismic activity recording, potential frequency and strengths of earthquakes in the area of consideration. Vulnerability type and quality of the insured values occupancy, building age, height, construction type and design, and soil conditions. These are the physical characteristics that can influence the susceptibility to loss. Policy Conditions common policy restrictions coinsurance, deductible and loss limit. Policy conditions will impact on the amount of eposure borne by the insurer. Portfolio Sums insured both on a gross and net (of reinsurance) aggregate basis for building, contents, business interruption, and additional living epenses. Aggregates must be segregated for shake and fire following coverages, as well as for personal and commercial business. Ideally, each of these categories must be accumulated by location of the risk at a minimum, this should be done by CRESTA Zone (see attached Earthquake Accumulation Assessment Zones). ISSUES/COMMON PROBLEMS 1. In the past, the collection of eposure data has proven to be a major problem, and this still continues. Some companies have only recently begun to accumulate this information, which has necessitated changes in procedures and systems. This has been further complicated by the fact that the integrity or quality of the data has been found to be lacking. 2. When calculating net PMLs, it is important to remember that reinsurance covers may have occurrence limits. These occurrence limits have to be taken into account when estimating net PMLs. It is not sufficient to simply use net insured values as input to a model. 3. With multiple locations insured under one policy, often the correct values are not coded or distributed properly among the regions in which the different risks might be located. Also, with multiple locations, the loss limit should be applied separately to individual locations.

6 4. Many policies contain a guaranteed replacement cost agreement, which can greatly inflate the final settlement. For a single loss, this eposure is manageable; however, this is of particular concern in the event of a catastrophe having a multitude of losses. 5. Co-insurance Do the underwriters check to confirm that the risk is insured to its proper value? The concern is that companies may not collect premium for a risk they may be required to pay, and that the eposures (and losses) are significantly higher. 6. Recent changes in the underlying data and assumptions for some models have produced significant fluctuation in results, particularly when compared to the previous estimates. Also, there are significant differences in estimates between models. 7. Automobile Physical Damage OSFI is suggesting that there be a PML estimate for damage to automobiles arising from an earthquake. This would be done by taking a percentage factor times the total number of vehicles insured, times an estimated average insured value per vehicle. At this time, most models do not incorporate auto physical damage into the estimation, since previous occurrences have shown this not to be a significant factor in the overall eposure. For eample, in the 1994 Northridge quake, the loss from automobile damage was approimately 0.5% of the total insured loss. 8. Some insurers have been specifically ecluding fire following within the insurance contract, unless the insured agrees to pay a premium to obtain this as additional coverage. Governments have refused to address this ambiguity within the Insurance Act, leaving the courts to interpret the reasonableness of this approach, only after the event has occurred. There are differing legal interpretations as to the outcome of such a scenario. QUESTIONS Hazard 1. Does the company have a significant earthquake eposure in B.C. or Québec (including Ottawa Valley)? 2. Does the company have any underwriting guidelines and restrictions for this eposure? 3. Does the company use a model to determine its earthquake PML? 4. Was this model developed internally, or is one by an outside consultant used? Is this a deterministic or a probabilistic model? 5. Does the model contain data detailing seismicity frequency and strengths of earthquakes in the area of consideration? Vulnerability 6. Does the model contain factors which take into account the vulnerability of the portfolio? Is this done individually by each building or on the portfolio as a whole? Policy Conditions 7. What are the company s guidelines with respect to policy conditions (coinsurance, earthquake deductibles and loss limits)? Does the model factor in these variables?

7 8. Were the following coverages considered in the PML estimates? Shake Fire Following Yes No Yes No Commercial buildings Commercial contents Business interruption Sprinkler leakage Dwelling (building) Personal property contents Additional living epenses Outbuildings Auto physical damage 9. Were the following factors affecting ultimate loss considered? Shake / Fire Following Yes No Post-event inflation Co-insurance Loss adjustment epenses Debris removal Portfolio 10. How does the company track accumulations of earthquake eposures (e.g., postal code, civic address, CRESTA Zones, etc.) 11. Can you etract and identify the limits of Limits of Loss policies, and the relationship to the sum insured, both on a pro rata and ecess of loss basis? 12. How does the company ensure data quality and integrity?

8 TERMS AND DEFINITIONS Deterministic Model assumes a specific earthquake scenario, and then calculates the level of ground motion at the site of interest. Probabilistic Model considers potential earthquake sources that could affect the site, modelling their magnitude, frequency characteristics, and calculating the ground motion at the site, including variability in the attenuation. PML Probable Maimum Loss is the threshold dollar value of losses beyond which losses caused by a major earthquake are unlikely. Gross PML is the PML after policyholder deductibles but before catastrophic and other reinsurance protection. Net PML is the PML after deductibles and catastrophic and other reinsurance protection; sometimes also epressed as EML (estimated maimum loss). CRESTA Zone CRESTA stands for Catastrophe Risk Evaluation and Standardizing Target Accumulation, an organization founded in 1977, mainly to improve accumulation control systems, and, to a lesser degree, assist in catastrophe loss management. There are, altogether, 20 accumulation assessment zones created by CRESTA. The high-risk zones for earthquake are Vancouver/Victoria (1, 2, 3, 4), Greater Montréal (5, 6, 7, 8) and Québec City (9, 10). Event Return Period corresponds to the reciprocal of the probability of an earthquake event occurring. For eample, an event return period of 250 years equates to a 0.4% (1/250) probability of a particular earthquake event occurring in a given year. Lower magnitude events (based on the Richter scale) occur more often than higher magnitude events, and generally are epected to produce lower damage figures. A higher event return period translates into a lower probability of an earthquake occurrence, but a higher potential dollar value of loss. Loss Return Period is the reciprocal of the annual probability of a certain loss level being eceeded for any possible earthquake event. Vulnerability Risk Characteristics: 1. Occupancy single residence dwelling, light industrial, high rise office, etc. 2. Building Age buildings retrofitted or renovated for earthquake should be recorded as having been built during the year that this work was completed. 3. Building Heights number of stories. 4. Building Construction use Applied Technology Construction classifications (i.e., masonry, steel, wood). 5. Unusual Design large unsupported spans (e.g., a domed stadium), or other unusual configurations. 6. Soil Conditions derived from Geological Survey Division of Natural Resources Canada. For eample, during an earthquake, loose or sandy soil becomes unstable or more fluid (known as liquefaction). A structure built on top of this type of soil will often collapse due to the unstable foundation. 7. Geocoding postal code or CRESTA Zones to establish a location. Seismic Definitions: 1. Magnitude is a measure of the earthquake s energy, independent of the place of observation. 2. Epicenter is the point on the earth s surface directly above the focus. 3. Focus is usually an instrumentally located point below the earth s surface where the faulting first occurs. 4. Attenuation is the decrease in intensity depending on the distance from the focus. 5. Intensity is a measure of all earthquake effects at a location or site. RRC Reinsurance Research Council

9 This map is based on information taken from the National Atlas of Canada Natural Hazards posters Her Majesty the Queen in Right of Canada with permission of Natural Resources Canada. 6

10 7