collection of earthquake risk and insurance

South Africa Spotlight on EarthquakeJune 2010

Introduction Aon Benfield and the Aon Benfield Natural Hazard Centre Africa at the University of Pretoria have collaborated to look into one of the largest risks facing the people, government and insurance community in South Africa: earthquake. Regarded as the natural hazard most likely to trigger the country’s largest financial loss, this report combines science and insurance industry expertise to help prepare for a seismic event.

The new research into potential earthquake magnitudes and probabilities, represented in catastrophe models, will enable insurers to obtain a more accurate estimate of their exposure and in turn purchase appropriate reinsurance cover.

The report also helps to set the record straight around the recent hype on predictions of a Haiti sized scenario for Cape Town and Durban. While we address the question of large earthquake magnitudes in South Africa, this is put in the context of return periods associated with these events.

With South Africa in the spotlight for the 2010 FIFA World Cup, we use the Cape Town Stadium and Durban’s Moses Mabhida stadia as case studies to illustrate the low probability of high magnitude earthquakes.

Even though Johannesburg has the highest total exposure, we focus on Cape Town and Durban due to their experience of the largest seismic events recorded in South African history and where we would most likely expect the largest damage.

Figure 1 Location and magnitude of historical earthquake events and key football stadia

The Science Seismicity in South Africa The south-western Cape has one of the highest levels of seismicity in South Africa, which is characterised by its dual source of seismicity comprising mine related events and tectonic origin earthquakes.

According to new research from Professor Kijko, director of the Aon Benfield Natural Hazard Centre Africa, the Western Cape Province, with Cape Town as the capital city, can expect a maximum earthquake close to magnitude 7.0. By comparison, the largest mine related event in the Johannesburg gold mine area is estimated at magnitude 5.6.

Historically, the most severe earthquake of magnitude 6.3 occurred on 29 September 1969 in Ceres, 100 km northeast of Cape Town. The event resulted in 12 lost lives and numerous damaged buildings in the town of Tulbagh. On 4 September 1809, a seismic event estimated at magnitude 6.3, occurred at the Milnerton Fault, a mere 10km from Cape Town CBD and the location of the Cape Town Stadium.

The largest mine related event in the history of South Africa occurred on 5 March 2005 in the Klerksdorp (Stilfontein) gold mining district, 200km west of Johannesburg, which reached a magnitude of 5.3. Below ground, substantial damage was observed within the mines, while above ground, the structural damage to property was relatively low.

South Africa Spotlight on Earthquake Page 2

Mining related events add a low magnitude, high frequency facet to earthquake risk in Johannesburg. It is also believed that accumulation of water within old mine shafts of several kilometres deep can trigger small, and occasionally moderate sized, seismic events.

The tectonic origin and mining related events are considered to be largely uncorrelated. As mining activity around Johannesburg diminishes with the depletion of gold reserves, so has the risk of mining induced seismic events originating below the city.

Seismic risk in Cape Town Cape Town was exposed to an earthquake at the Milnerton Fault in 1809, so the Aon Benfield Natural Hazard Centre investigated the potential probable maximum loss (PML) that could be caused by a similar earthquake in the future.

Seismic events were selected from a catalogue of earthquakes in South Africa that occurred between 1620 and 2006 within a 300km radius from Milnerton. The results of the hazard analysis, in terms of the mean return periods, were calculated for range of expected magnitudes from 3.0 to 6.87 – as referenced in Figure 2.

The figure shows for example that an earthquake of magnitude 6.0 and larger can be expected to occur once in 300 years in the selected area surrounding Cape Town. It must be emphasised that the predicted upper limit earthquake magnitude for the area is 6.87.


AREA: Cape Town 100,000

3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8

10,000

Ret

urn

Per

iod

[YE

AR

S]

1,000

100

10

0

Magnitude

Figure 2 Magnitudes and return periods for earthquakes in the Cape Town area (200km radius)

Assuming the worst case scenario of a 6.87 magnitude earthquake at the Milnerton Fault, the predicted Mercalli Magnitude Intensity (MMI) in the Cape Town CBD and the Cape Town Stadium would be about IX or ‘ruinous’.

The chart below shows expected damages from a MMI IX earthquake and their uncertainties for three types of buildings, which represent approximately 70% of all South African urban structures (Davis and Kijko, 2003). As the financial sector is robust and closely involved in financing insurable commercial and residential property, building standards for these are enforced and regulated by local authorities.

Building type Expected damage Uncertainty Unreinforced masonry with load bearing wall, low rise 45 % 30% - 61% Reinforced concrete shear wall without moment resisting frame, medium rise 20 % 12% - 29%

Reinforced concrete shear wall without moment resisting frame, high rise 27% 16% - 37%


Seismic risk in Durban Durban, the largest port city in South Africa, is located next to the warm Indian Ocean and is not regarded as being exposed to high seismic risk. A magnitude 6.3 event occurred at St Lucia Estuary, about 220 km north of Durban, on New Year’s Eve 1932 but – unlike Cape Town – there is not an active known fault close to the city.

The return periods for earthquakes of magnitude 5.0 and 6.0 were assessed considering all seismicity within a 300 km radius of Durban. From this research, it is estimated that an earthquake of magnitude 5.0 could cause structural damage only if its epicentre is less than 45km from Durban. The return period for such an event is about 735 years.

Using the Moses Mabhida Stadium as a case study, for a magnitude 6.0 event and larger to occur and have a damaging impact, the epicentre must be closer than 90km. The return period for such an event is about 5,000 years. It is clear that earthquakes are not a significant risk to insured structures in Durban, although the contribution to expected losses must be taken into account.

The (re)insurance Catastrophe reinsurance South Africa contributes 0.7% to the global reinsurance market with an estimated ZAR6.6 billion short term (non-life) reinsurance premium. An estimated ZAR23 billion (USD3 billion) of catastrophe reinsurance is purchased, which covers all natural perils including flood, windstorm and hail. Local insurers tend to purchase reinsurance programs to recover at least a 1 in 250 year event, on the advice of the Financial Services Board regulator.

Exposure in South Africa is divided into 16 CRESTA zones. CRESTA 5, 6 and 7, which include the financial heartland surrounding Johannesburg and Pretoria, contribute 40% of the total exposure, while Cape Town and Durban equal 25% combined.

Catastrophe purchase is usually determined by exposure aggregates in CRESTA 5, 6 and 7, which are characterised by high exposure and low magnitude seismic activity from gold mines 4km deep. Traditionally insurers purchased catastrophe reinsurance at 2.5% of CRESTA 5, 6 and 7 exposure; although this number has decreased to below 2% after the introduction of new catastrophe earthquake models for the region which present a more accurate picture of the risk.

Impact on the insurance market Catastrophe reinsurance premium rates have reduced in recent years and business is in demand with international reinsurers as it provides good diversification from traditionally predominant US and European exposures. There has also been a lack of catastrophe events over the last decade resulting in further downward pressure on premium rates. This also means that South African insurers’ catastrophe retentions are low compared to other markets.

An earthquake would mainly affect the local South Africa insurance market in terms of claims and would create a perception of increased risk. This could lead to higher insurance rates locally but would not impact the global reinsurance market significantly.

Conclusion South Africa's large magnitude earthquakes have a low frequency.

The new academic research into seismic risk in South Africa enables Aon Benfield to improve its catastrophe models by updating provision for both mining related and tectonic origin seismic events. At the same time, the research can be used as a method of quantifying required catastrophe reinsurance cover and help insurers better manage their balance sheets.

Using scientific expertise from one of South Africa’s leading universities, this report helps both insurers and reinsurers operating in South Africa to appropriately price catastrophe risk for potentially the largest natural hazard in the region.

References Davies, N. & A. Kijko, 2003. Seismic risk assessment: with an application to the South African insurance industry, South African Actuarial Journal, 3, 1–28.

Contact Pieter Visser Professor Andrzej Kijko Aon Benfield, Johannesburg Aon Benfield Natural Hazard Centre Africa t: +27 11 944 7265 t: +27 11 420 3613 e: [email protected] e: [email protected]

Probabilistic Assessment of Earthquake

Insurance Rates for Turkey

M. S. YUCEMENDepartment of Civil Engineering and Earthquake Engineering Research Center, Middle EastTechnical University, 06531 Ankara, Turkey; E-mail: [email protected]

(Received 7 May 2004; accepted 3 November 2004)

Abstract. A probabilistic model is presented to obtain a realistic estimate of earthquake

insurance rates for reinforced concrete buildings in Turkey. The model integrates informationon seismic hazard and information on expected earthquake damage on engineering facilities ina systematic way, yielding to estimates of earthquake insurance premiums. In order to dem-

onstrate the application of the proposed probabilistic method, earthquake insurance rates arecomputed for reinforced concrete buildings constructed in five cities located in differentseismic zones of Turkey. The resulting rates are compared with the rates currently charged by

the insurance companies. The earthquake insurance rates are observed to be sensitive to theassumptions on seismic hazard and damage probability matrices and to increase significantlywith increasing violation of the code requirements.

Key words: seismic hazard, earthquake engineering, earthquake insurance, building damage,

damage probability matrix, risk premium, insurance premium

1. Introduction

Turkey is located in one of the seismically most active regions of the world.Geological and seismological data indicate that in almost all parts ofTurkey seismic hazard is significant: 95% of the population, 92% of thetotal area and 98% of the industry are under earthquake threat (Gencogluet al., 1996). To reduce the vulnerability of human settlements and indus-trial facilities, the engineered structures must be designed and constructedto resist the effects of earthquakes. However, due to poor control of con-struction in Turkey, a large degree of damage is expected if a majorearthquake hits a large city, as observed during the recent earthquakes.Accordingly, different measures have to be considered to alleviate the post-disaster consequences of earthquakes as well as enforcement of the earth-quake resistant provisions of the Code. The implementation of obligatoryearthquake insurance is among such measures.

In Turkey, the State used to have a legal obligation to fund the costs ofreconstructing buildings after an earthquake. This responsibility of the Statenaturally brought an unplanned burden on the national economy and on the

Natural Hazards (2005) 35: 291–313 � Springer 2005DOI 10.1007/s11069-004-6485-8

already limited central budget in the case of catastrophic seismic events. Thetwo recent major earthquakes in Turkey on 17 August and 12 November1999 both occurred in or near urban settlements and caused widespreaddestruction of the building stock. This activated the idea of a nationwideobligatory earthquake insurance enforcement, which was first put forward in1978 by the Ministry of Reconstruction and Settlement. At that time theEarthquake Engineering Research Institute, Middle East Technical Univer-sity (EERI/METU), through a project sponsored by the same Ministry,investigated the feasibility of an obligatory insurance model (Gurpinar et al.,1978; Gurpinar and Yucemen, 1980). The obligatory insurance model con-sidered by EERI/METU aimed to serve for a number of purposes, whichmay be listed as: control and therefore improvement of construction practicesin seismic regions, decrease of time loss due to the interruption of services,accumulation of funds for reconstruction and resettlement. The fact that thefirst purpose was included is due to the peculiar control mechanism inTurkey, which may to some extent be generalized to other developingcountries. As control of construction is implemented unsatisfactorily, it isexpected that the earthquake insurance may constitute a complementarycontrol mechanism, based on the conflict of economic interests on the part ofthe insurer and the insured.

After the two major earthquakes in 1999, the Government of Turkey hasdecided to enforce the earthquake insurance on the nationwide basis with thesole purpose of privatizing the potential risk by offering insurance via theTurkish Catastrophic Insurance Pool (TCIP) and then exporting the majorpart of this risk to the international reinsurance and capital markets(Bommer et al., 2002). Although the main aim was to reduce government’sfiscal exposure, it was also intended to encourage risk mitigation and saferconstruction practices. To achieve these purposes all registered residentialdwellings, the total number being about 13 million, are required to be in thecompulsory earthquake insurance coverage. Initially funded by the WorldBank, the TCIP program became effective as of March 2001. In spite of itscompulsory nature, TCIP policy count was about two million as of Sep-tember 2004, corresponding only to 15.3% of the total property owners. Thehighest insured percentage was observed in the Marmara region (24.80%),where the two major earthquakes in 1999 caused the highest damage(www.dask.gov.tr). Milli-Re (Turkish Reinsurance Company) is responsiblefor running the pool, which is protected by the World Bank and a whole hostof reinsurers. TCIP has a potential to become one of the largest earthquakeinsurance companies in the world, provided that the penetration rate is in-creased significantly.

In the implementation of the obligatory insurance program, no reliableinvestigations were conducted for the assessment of earthquake insurancerates. As a matter of fact the rates are set by the Prime Ministry of the

M. S. YUCEMEN292

Turkish Republic, the Under-secretariat of Treasury and charged by theinsurance companies under the control and coordination of TCIP. Thus, thisstudy aims at presenting a simple model for the assessment of earthquakeinsurance rates by considering the damage statistics and the potential seismichazard. The proposed model is used for the assessment of the earthquakeinsurance premiums for reinforced concrete buildings constructed in differentseismic zones of Turkey. The earthquake insurance premiums are computedby considering only the risk of damage due to earthquake shock and lossesresulting from fire due to earthquake are omitted.

In most of the countries, earthquake insurance programs are implementedwith the purpose of accumulating funds to cover the post-disaster expendi-tures. The practice and regulations of earthquake insurance vary fromcountry to country. Below, a brief description for some countries, where theseismic hazard is significant is presented.

In Japan, earthquake insurance has been available since the 1950’s(Kamei, 1976). Earthquake insurance covers damage from volcano eruptionsand tsunamis also. The buildings are classified according to their earthquakeresistance capacities and in this way constructions with poor quality anddesign are discouraged. The earthquake insurance for homes is an endorse-ment to fire coverage, and reinsurance is provided by the government(Brillinger, 1993, p. 16; Scawthorn et al., 2003, chapter 32, pp. 24–25).

In California, a state-run earthquake insurance company, called theCalifornia Earthquake Authority (CEA), was formed in 1994 for providingcoverage for homeowners. The all-residential CEA program was formed inorder to overcome some of the difficulties encountered by the insurancecompanies after the 1994 Northridge earthquake (Scawthorn et al., 2003,chapter 32, pp. 23–24).

In New Zealand, the earthquake insurance is a part of fire insurance. TheEarthquake Commission (EQC), which was formed in 1993, handles theearthquake insurance nationwide. For homes there is a limit of 100,000 NZdollars of coverage at a cost of 0.5 per 1,000 (Scawthorn et al., 2003, chapter32, p. 26). The New Zealand reinsurance program is one of the largestcatastrophe coverages in the world (Steven, 1992).

Studies on the assessment of earthquake insurance premiums based onstatistical methods and utilizing earthquake engineering concepts are limitedin number. Gurpinar et al. (1978) and Gurpinar and Yucemen (1980) haveconsidered the problems related to obligatory earthquake insurance imple-mentation in Turkey and a pilot application was carried out for Denizli.Yucemen and Bulak (1997) have developed a statistical model for theassessment of earthquake insurance rates for the different seismic zones inTurkey. Smolka and Berz (1989, 1991) have developed a methodology forobtaining insurance premiums consistent with seismic risk and estimation ofpotential losses due to large earthquakes. Some of the other major studies

ASSESSMENT OF EARTHQUAKE INSURANCE RATES 293

conducted in the field of earthquake insurance are due to Straub (1973), VereJones (1973), Lockett (1980), Brillinger (1993), Kunreuther (1996), Walker(2000), Amendola et al. (2000) and Scawthorn et al. (2003).

2. Probabilistic Model for the Estimation of Earthquake Insurance Premiums

The assessment of earthquake insurance premiums requires two types ofstudies, namely: seismic hazard analysis (SHA) and estimation of potentialearthquake damage to structures. In the following the models used for thesetwo types of studies are presented.

2.1. SEISMIC HAZARD ANALYSIS (SHA)

In the probabilistic sense the seismic hazard can be defined as the probabilityof exceeding different levels of ground motion at a given place and within agiven period of time due to expected seismic activity in the region.

Many models have been developed for seismic hazard analysis. Most ofthe earlier models of seismic hazard assessment were based on the assump-tion that earthquake occurrences are independent events in space and time,and utilized the Poisson model which is also known as the classical SHAmodel (Cornell, 1968) or the extreme value statistics. Later studies consideredthe temporal or spatial dependence of earthquakes only. Some models con-sistent with the ‘‘elastic rebound theory’’ considered the temporal dependenceof earthquakes based on the processes with Markovian characteristics.Attempts were also made to model the spatial dependence of earthquakes. Inrecent studies, the occurrence of earthquakes is treated as a space-timeprocess and the spatial and temporal correlations are taken into consider-ation (Yucemen, 1993; Akkaya and Yucemen, 2002). A detailed review andcomparison of different stochastic models of earthquake occurrence is givenin Yucemen and Akkaya (1996) and Akkaya and Yucemen (2002).

The probabilistic formulation adopted in this study is based on the clas-sical SHA model. However, the classical model is improved by taking intoaccount the uncertainties associated with the attenuation equation, geo-graphical location of seismic sources and estimation of seismicity parameters.The SHA model used here involves the following stages:

(i) Determination of the probability distribution of earthquake magni-tudes: The probability distribution of earthquake magnitudes is obtainedfrom the linear magnitude-frequency relationship (Gutenberg and Richter,1956) having a lower bound m0 and an upper bound m1 for earthquakemagnitudes. The resulting exponential probability density function is asfollows (Yucemen, 1992):

M. S. YUCEMEN294

fMðmÞ ¼ kb expð�bðm�m0ÞÞ ð1ÞHere, k is the standardization coefficient which is needed to normalize thedistribution function to unity at m ¼ m1 and obtained from the followingformula:

k ¼ ½1� expð�bðm1 �m0ÞÞ��1 ð2ÞThe parameter b, which is the slope of the magnitude-frequency relationship,is an indicator of the severity of seismic activity and is related to the tectonicstructure of the region.

(ii) Determination of the probability distribution of earthquake occur-rences: Earthquake occurrences are assumed to be independent events inthe time domain and can be modeled as a Poisson process. Poisson modelis preferred in this study since it is found to be in agreement with theobserved seismic activity related to moderate or large magnitude earth-quakes which affect engineering structures seriously (Cornell andWinterstein, 1998; Ferraes, 2003). According to the Poisson model, theprobability of n earthquakes having intensity m P m0 occurring during[0, t] is:

PðN ¼ nÞ ¼ ½expð�mtÞðmtÞn�=n! ð3Þwhere, N ¼ number of earthquakes (m P m0) occurring in the time interval[0, t] and m ¼ mean number of earthquakes having intensity mPm0 per unittime (generally taken as 1 year).

(iii) Determination of an attenuation relationship: Attenuation relation-ships describe the decay of the ‘‘severity’’ of an earthquake of magnitude mwith epicentral (or hypocentral) distance r. Due to the scarcity of localstrong-motion data in Turkey, it was necessary to select an attenuationequation from a spectrum of such equations that appear in the literature. Inthis study the following widely used attenuation relationship given by Joynerand Boore (1981) is adopted.

Log PGA ¼ �1:02þ 0:249m� logr� 0:00255r

þ 0:26p for 5:0 � m � 7:7ð4Þ

Here, r ¼ (d2 + 7.32)½ and p ¼ 0 or 1, respectively, for 50 and 84% prob-ability that the prediction will not exceed the real value. PGA denotes peakground acceleration in terms of g and d is the closest distance in km to thesurface projection of the fault rupture. The coefficient of p in the aboveequation represents the standard error of prediction, denoted by racc. Forthis attenuation equation racc ¼ 0.26.

(iv) Determination of the error in the location of seismic sources: in thisstudy, the location of seismic sources is taken to be random and the location


of seismic source boundaries are assumed to exhibit a Gaussian distributionwith mean zero and standard deviation, denoted by rloc. This standarddeviation quantifies the expected error in the location of seismic sources. Thedetails of this model are given in Yucemen and Gulkan (1994).

It is to be noted that the main source of uncertainty in SHA is theattenuation equation. Other sources of uncertainties are related to the seis-micity parameters of the region (b, m1, m) and the geographical location ofseismic zones.

2.2. ESTIMATION OF POTENTIAL DAMAGE TO STRUCTURES

The other component of the probabilistic model involves the assessment ofthe seismic vulnerability of buildings. Damage is commonly described by aloss ratio that varies with the strength of shaking and type of structure(Whitman, 1973; Blong, 2003a; Askan and Yucemen, 2003). Due to theuncertainties involved, the damage that may occur during future earth-quakes has to be treated in a probabilistic manner. For this purposedamage probability matrices (DPM) are constructed from observationaland estimated data. A DPM expresses what will happen to buildings, de-signed according to some particular set of requirements, during earth-quakes of various intensities (Whitman, 1973, ATC-13, 1985). An elementof this matrix Pk (DS, I) gives the probability that a particular damagestate (DS) occurs when the structure of kth-type is subjected to an earth-quake of intensity, I. The identification of damage states is achieved in twosteps:

(i) The qualitative description of the degree of structural and non-struc-tural damage by words: In the damage evaluation forms used by the GeneralDirectorate of Disaster Works prior to 1994, five levels of damage states werespecified. These are: No damage (N), light damage (L), moderate damage(M), heavy damage (H), and collapse (C) states. This categorization ofdamage states is also used in this study.

(ii) The quantification of the damage described by words in terms of thedamage ratio (DR), which is defined as the ratio of the cost of repairing theearthquake damage to the replacement cost of the building. For mathe-matical simplicity it is convenient to use a single DR for each DS (Blong,2003b, p. 3; Gurpinar and Yucemen, 1980). This single DR is called thecentral damage ratio (CDR). Based on interviews with experts in charge ofdamage evaluation and based on similar studies, the damage ratios corre-sponding to the five damage states are estimated by Gurpinar et al. (1978)and are shown in Table I. In the present study only DPMs for conventionalreinforced concrete frame buildings are considered, and the correspondingmatrices are constructed from observational and estimated data available forTurkey.

M. S. YUCEMEN296

Using the post-event observational data on past earthquakes, Pk(DS, I)values can be calculated from:

PkðDS; IÞ ¼ NðDS; IÞ=NðIÞ ð5Þwhere, N(I) ¼ number of kth-type of buildings in the region subjected to anearthquake of intensity I and N(DS, I) ¼ number of buildings which are indamage state DS, among the N(I) buildings.

DPMs can be obtained from past earthquake data and by using subjectivejudgment of experts. Techniques based on theoretical analyses for developingDPMs are also available (Whitman, 1973; Yucemen and Askan, 2003). Inthis study DPMs are obtained by using both empirical results and subjectivejudgment of experts. The form of a DPM is illustrated in Table II.

2.3. DETERMINATION OF THE PURE RISK PREMIUM

Expected annual damage ratio (EADRk) is used as a measure of the mag-nitude of earthquake damage to a kth-type of structure that will be built in acertain seismic zone and is defined as:

EADRk ¼X

I

MDRkðIÞ � SHI ð6Þ

Table I. Damage ratios and CDRs corresponding to different damage states.

Damage state (DS) Damage ratio (DR) % Central damage ratio (CDR) %

None 0–1 0

Light 1–10 5

Moderate 10–50 30

Heavy 50–90 70

Collapse 90–100 100

Table II. Damage probability matrix.

Damage state

(DS)

Central damage ratio

(CDR) %

Modified Mercalli Intensity (MMI)

V VI VII VIII IX

None 0

Light 5 Damage State Probabilities

Moderate 30 P(DS, I)

Heavy 70

Collapse 100


where, MDRk(I) ¼ average damage ratio for the kth-type of structuressubjected to an earthquake of intensity I and SHI ¼ annual probability of anearthquake of intensity I occurring at the site.

The information contained in the damage probability matrix and in thedamage ratios can be combined by defining the MDRk(I) as follows:

MDRkðIÞ ¼X

DS

PkðDS; IÞ � CDRDS ð7Þ

where, CDRDS ¼ central damage ratio corresponding to the damage stateDS.

After calculating EADRk, the pure risk premium (PRPk) is computedbased on the insured value of the building (INSV) under consideration fromthe following relationship:

PRPk ¼ EADRk � INSV ð8Þ

2.4. DETERMINATION OF THE TOTAL EARTHQUAKE INSURANCE PREMIUM

The total earthquake insurance premium (TPk) that will be charged by aninsurance company for the kth-type of structure is found by increasing thePRPk by some margin as follows:

TPk ¼ ðPRPkÞ=ð1� LFÞ ð9Þwhere, LF ¼ load factor which covers the hidden uncertainties, administra-tion, business and taxation expenses and a reasonable profit allowance forthe insurance firm. Here, LF is set equal to 0.4 (Gurpinar and Yucemen,1980) and with this value of LF, the total insurance premium that will becharged by the insurance companies becomes:

TPk ¼ 1:67� PRPk ð10ÞA flowchart showing the algorithm for the computation of earthquake

insurance premiums is given in Figure 1.

3. Application: Assessment of Earthquake Insurance Rates for Different

Seismic Zones in Turkey

The implementation of the proposed probabilistic method is illustrated bycomputing the earthquake insurance rates for reinforced concrete buildingslocated in different seismic zones of Turkey. Cities are selected from thedifferent zones according to the new seismic zonation map of Turkey(Gencoglu et al., 1996). These cities and the corresponding seismic zones,(shown in parentheses) are as follows: Erzincan (Zone I), Denizli (Zone I),Istanbul/south (Zone I), Istanbul/north (Zone II), Ankara (Zone III) andKonya (Zone IV). In computing the earthquake insurance rates, complianceand non-compliance with the Code (1975) is also considered.

M. S. YUCEMEN298

3.1. SEISMIC HAZARD ANALYSIS

In order to use the proposed model it is first necessary to carry out a SHA forthese cities. For this purpose the comprehensive study carried out by Gulkanet al. (1993) for the assessment of seismic hazard in Turkey is utilized. Thelocations of the seismic sources are shown in Figure 2. These seismic sourcesare all modeled as area sources. The past seismic activity data withmagnitude ‡4.5 are distributed to these sources according to their epicenterallocations and closeness to the sources. Earthquakes whose epicenters cannot

EARTHQUAKE INTENSITIES, I

EXPECTED ANNUAL DAMAGE RATIO

EADR k=∑ MDR k (I) x SH (I)I

SEISMIC HAZARD, SH (I)

PURE RISK PREMIUM PRP k = EADRk x INSV

MEAN DAMAGE RATIOS

MDR k (I) = ∑ Pk(DS, I) x CDR (DS)DS

DAMAGE PROBABILITY MATRIX[ P k (DS, I) ] L

AR

TN

EC

EG

AM

AD

IT

AR

OS

,R

DC

(S

D)

TOTAL INSURANCE PREMIUM

TPk =PRP k

1-LF

MA

DS

ET

AT

SE

GA

(S

D)

Figure 1. Algorithm for the computation of earthquake insurance premiums.


be associated with anyone of these seismic sources are considered as ‘‘float-ing’’ earthquakes and assigned to ‘‘background’’ seismicity.

The values given in Table III are taken as the ‘‘best’’ estimates of theseismicity parameters and are used as the input to the computer program.The attenuation relationship given by Equation (4) is utilized as the atten-uation model for PGA. The standard deviation racc is assigned a value of 0.3,which is believed to be a good estimate of the uncertainty involved in theattenuation model adopted herein. The location uncertainty is assumed to beisotropic, and the corresponding standard deviation is taken as rloc ¼ 20 km.With these input values the seismic hazard is computed for these five citiesusing the SEIS-HAZARD software package. A detailed description of thissoftware is presented in Ozgur and Yucemen (1997).

Although intensity is not a very reliable and objective measure of theseverity of ground shaking, it is used in this study mainly because earthquakedamage to buildings is much better correlated with the modified Mercalliintensity (MMI). Accordingly, the seismic hazard values computed in termsof PGA are converted to MMI scale and for this conversion the empiricalrelationships given by Trifunac and Brady (1975) and Wald et al. (1999) areutilized. The resulting conversion curve is smoothed in the highermagnitude-intensity levels in order to achieve a better correlation. Based onthe resulting seismic hazard values expressed in terms of MMI, the annualprobabilities of observing different intensity levels are computed for each oneof these cities. MMI scale provides 12 discrete levels of intensity withincreasing severity. Consistent with the existing seismic activity, only levels

Figure 2. Seismic source zones for Turkey (Gulkan et al., 1993).

M. S. YUCEMEN300

V–IX are considered. The reader is referred to Gulkan et al. (1993) for fur-ther details of the SHA and the results concerning the variation of seismichazard for each city.

3.2. DAMAGE PROBABILITY MATRICES

In order to compute EADR according to Equation (6), it is necessary toobtain the DPMs that are applicable for the seismic zones that these cities arelocated in. For this purpose the previous studies on the assessment of DPMsfor reinforced concrete buildings located in different seismic zones of Turkeyare examined and revised. The development of DPMs for Turkey was firstconsidered by Gurpinar et al. (1978) and Gurpinar and Yucemen (1980). Inthese studies, based on the available damage evaluation records from the1976 Denizli and 1971 Bingol earthquakes, it was only possible to estimatethe damage probabilities in the MMI ¼ VI and MMI ¼ VIII columns of theDPM. Since the empirical data were inadequate for establishing a DPMcompletely, it was decided to estimate damage state probabilities by makinguse of the subjective judgment of experts. For this purpose a questionnairewas prepared and sent to thirty engineers experienced in the field of

Table III. Values of the seismicity parameters for different seismic sources (Gulkan et al.,1993).

Source no. m1 m0 b m

1a 7.4 4.5 1.84 3.899

1b 7.2 4.6 1.60 1.164

1c 7.9 4.5 1.64 1.873

2 6.4 4.5 1.29 0.379

3 7.6 4.5 1.68 0.621

4 6.4 4.5 2.29 1.880

5 5.2 4.5 2.08 0.280

6a 7.0 4.5 1.62 0.663

6b 7.2 4.5 2.62 2.750

6c 7.4 4.5 1.92 4.254

7 7.7 4.5 2.25 8.567

8 6.0 4.5 1.99 0.100

9 6.3 4.7 0.73 4.000

10 6.3 4.7 0.73 0.010

11 6.3 5.0 0.73 0.030

12 6.4 4.5 3.47 0.020

13 6.5 4.8 1.39 0.070

14 6.3 4.7 1.02 0.070

15 5.8 4.5 1.85 0.162


earthquake engineering. Only ten engineers responded and the responses ofthese ten engineers are averaged to obtain the subjective DPMs. The limitedempirical data were combined with the DPMs obtained by the subjectivemethod and DPMs for reinforced concrete buildings were proposed for thedifferent seismic zones of Turkey. These DPMs give two sets of subjectivedamage probabilities for reinforced concrete frame buildings constructed inthe different seismic zones of Turkey. The first set corresponds to buildingsthat are designed and constructed in conformance with the specificationsdesignated in the Code (1975), and in the second set it was assumed that theearthquake resistant design provisions are violated. In the following sectionsthese two conditions will be referred to as ‘‘According to the Code’’ (AC) and‘‘Not According to the Code’’ (NAC), where ‘‘Code’’ refers to ‘‘Specifica-tions for Structures to be Built in Disaster Areas’’ which was prepared andput into regulation in 1975. Here, because of space limitation only the DPMfor seismic Zone I, where seismic hazard is the highest, is presented(Table IV).

Later, Yucemen and Bulak (1997, 2000) have obtained empirical DPMsby using the post-earthquake damage assessment reports compiled by theGeneral Directorate of Disaster Affairs for the 1971 Bingol, 1976 Denizli,1983 and 1992 Erzincan and 1986 Malatya earthquakes. In a series of studiesconducted by Yucemen (2002) and Yucemen and Askan (2003), these DPMsare revised and updated in view of the additional information (Sucuoglu andTokyay, 1992) assessed on these earthquakes. Also the damage assessmentreports prepared by various institutions (Wasti and Sucuoglu, 1999;Elnashai, 2000; Ozmen and Bagci, 2000; www.seru.metu.edu.tr/archives/da-tabases) concerning the recent earthquakes, namely: 1995 Dinar, 1999Kocaeli (for the city of Adapazari) and 1999 Duzce are also utilized, espe-cially for complementing the empirical DPM at higher intensity levels. Thedamage state probabilities are computed by using Equation (5). The resulting

Table IV. Subjective damage probability matrix for seismic Zone I (Gurpinar et al., 1978).

Damage

State (DS)

CDR

(%)

MMI = V MMI = VI MMI = VII MMI = VIII MMI = IX

AC NAC AC NAC AC NAC AC NAC AC NAC

None 0 1.00 0.95 0.95 0.70 0.70 0.50 0.50 0.20 0.30 0.05

Light 5 0 0.05 0.05 0.15 0.20 0.20 0.20 0.20 0.30 0.20

Moderate 30 0 0 0 0.10 0.10 0.15 0.20 0.40 0.20 0.40

Heavy 70 0 0 0 0.05 0 0.10 0.10 0.10 0.20 0.20

Collapse 100 0 0 0 0 0 0.05 0 0.10 0 0.15

MDR (%) 0 0.25 0.25 7.25 4.0 17.5 14.0 30.0 21.5 42.0

(AC: According to the Code; NAC: Not according to the Code).

M. S. YUCEMEN302

empirical DPM is shown in Table V. In the sixth row of this table the numberof buildings for which damage assessments were made is given.

The information on the date of construction of buildings was generallymissing in the damage assessment reports compiled by the General Direc-torate of Disaster Affairs. Accordingly, it was not possible to classify thebuildings as constructed before or after the 1975 Code, except for the 1976Denizli and 1971 Bingol earthquakes. The building stock considered in thesetwo earthquakes is assumed to be constructed not in accordance with therequirements of the Code (1975), since they were constructed before the 1975Code became effective. For the buildings involved in the other three earth-quakes, i.e. 1983, 1992 Erzincan and 1986 Malatya, it was difficult to decide.However, the general opinion of experts in charge of damage evaluation wasthat: most of the buildings did not comply with the requirements of the 1975Code. A similar situation was valid for the 1999 Kocaeli and 1999 Duzceearthquake reports. However, for the 1995 Dinar earthquake, where only 39buildings were examined, it was possible to decide for each building thedegree of compliance with the Code. Accordingly, the empirical values givenin Table V are assumed to be valid for reinforced concrete structures that areconstructed not in accordance with the Code, except for the case of Dinar,where AC and NAC conditions are differentiated (Yucemen, 2002; Yucemenand Askan, 2003).

The values given in Table V are used to obtain the empirical damage stateprobabilities valid for different seismic zones by relating the cities with theseismic zones. Table VI gives the damage state statistics for Zone I withMMI ¼ VI, VIII and IX and for Zone II with MMI ¼ VII and VIII. Forcases where more than one earthquake damage data are available for thesame zone, weighted average damage state probabilities, based on thenumber of buildings, are computed. The empirical damage state probabilitiesgiven in this table correspond to the NAC case, as explained above.

3.3. BEST ESTIMATE DAMAGE PROBABILITY MATRICES

The DPMs will show differences from zone to zone. Therefore for each zonea DPM is needed. Besides, whether a building has been constructedaccording to the requirements of the code or not should be taken into con-sideration. In selecting these DPMs it is desirable to utilize all of the relevantinformation in a systematic way. In this respect, we note the following pointsconcerning the information presented in the previous section.

(i) It seems that the most reliable method for constructing DPMs is theempirical method, which is based on the observed damage statistics, providedthat personal biases in damage evaluation are controlled. However, due tolack of data it was only possible to quantify the empirical damage stateprobabilities for Zones I and II as given in Table VI.


Table

V.Empiricaldamagestate

probabilitiesforreinforced

concretebuildings.

Damage

State

(DS)

CDR

(%)19.8.1976Denizli

MMI=

VI

18.11.1983

Erzincan

MMI=

VI

6.6.1986

Malatya

MMI=

VII

22.05.1971

Bingol

MMI=

VIII

13.3.1992

Erzincan

MMI=

VIII

1.10.1995

Dinar

MMI=

VIII

17.08.1999

Kocaeli

MMI=

IX

12.11.1999

Duzce

MMI=

IX

AC

NAC

None

00.49

0.74

0.45

0.12

0.31

0.23

0.24

0.04

0.17

Light

50.37

0.23

0.39

0.29

0.48

0.31

0.24

0.34

0.16

Moderate

30

0.13

0.03

0.12

0.31

0.09

0.38

0.41

0.27

0.28

Heavy

70

0.01

0.00

0.03

0.18

0.07

0.04

0.05

0.175

0.19

Collapse

100

00

00.10

0.05

0.04

0.06

0.175

0.20

Number

of

Buildings

378

112

89

46

415

39

13240

5420

MDR

(%)

6.45

2.05

7.65

32.9

15

19.75

23

39.6

42.5

(AC:Accordingto

theCode;

NAC:Notaccordingto

theCode).

M. S. YUCEMEN304

Table

VI.

Empiricaldamagestate


concretebuildingsclassified

accordingto

theseismic

zones.

DamageState

(DS)

CDR

(%)

ZoneI

MMI=

VI(N

AC)

ZoneII

MMI=

VII

(NAC)

ZoneII

MMI=

VIII(N

AC)

ZoneI

MMI=

VIII(N

AC)

ZoneI

MMI=

IX(N

AC)

None

00.54

0.45

0.04

0.30

0.08

Light

50.34

0.39

0.43

0.45

0.29

Moderate

30

0.11

0.125

0.26

0.13

0.27

Heavy

70

0.01

0.035

0.135

0.07

0.18

Collapse

100

00

0.135

0.05

0.18

MDR

(%)

5.7

8.15

32.9

16.1

40.15

(NAC:Notaccordingto

theCode).


(ii) The empirical damage state probabilities are supplemented by theinformation available from other sources. For this purpose the subjectivelyassessed DPMs given by Gurpinar et al. (1978) are utilized. Although thisstudy was conducted long time ago, the experts answering the questionnairewere quite experienced and their responses were examined carefully, cross-checked and was rated very reliable at that time. Therefore, it is an importantsource of information which can not be ignored, especially if we consider thescarcity of the empirical damage data. However, this limitation of the DPMsbased on subjective judgment of experts is taken into consideration by givinga smaller weight in computing the best estimate DPMs, as explained in thefollowing paragraph (item iii).

(iii) The combination of the empirical DPM (Table VI) with the subjectiveDPMs is achieved by computing a set of weighted average DPMs. A sub-jective weight of 0.75 is assigned to empirical values whenever they areavailable and a weight of 0.25 is given to the subjective DPMs that arereflecting expert opinion.

(iv) A building that is not constructed according to the requirements of thecode, is expected to experience the same degree of damage irrespective of thezone, when subjected to a given earthquake intensity. Therefore, the prob-abilities listed under the NAC columns of the DPMs for a given intensitylevel should be the same in all of the zones.

The resulting DPMs are called the ‘‘best estimate’’ DPMs and again dueto space limitation only the DPM for seismic Zone I is given (Table VII).Modified Mercalli intensities of X–XII are not shown in the DPMs given inTables IV and VII. This is due to the fact the seismic hazard correspondingto these high intensity levels are very small, and consequently their contri-bution to risk is negligible compared to that of smaller intensities.In Tables IV and VII the term MDR denotes the ‘‘mean damage ratio’’ andis to be computed from Equation (7). The variation of MDR with MMI for

Table VII. ‘‘Best estimate’’ damage probability matrix proposed for seismic Zone I.

Damage

State (DS)

CDR (%)MMI = V MMI = VI MMI = VII MMI = VIII MMI = IX


None 0 1.00 0.95 0.95 0.58 0.70 0.46 0.50 0.28 0.30 0.07

Light 5 0 0.05 0.05 0.29 0.20 0.34 0.20 0.39 0.30 0.27

Moderate 30 0 0 0 0.11 0.10 0.14 0.20 0.20 0.20 0.30

Heavy 70 0 0 0 0.02 0 0.05 0.10 0.07 0.20 0.19

Collapse 100 0 0 0 0 0 0.01 0 0.06 0 0.17

MDR (%) 0 0.25 0.25 6.2 4 10.4 14 18.9 21.5 40.7


M. S. YUCEMEN306

different seismic zones and depending on the degree of conformance with theCode (1975) is shown in Figure 3. As observed in this figure, buildings notconforming with the requirements of the Code (1975) yield to significantlyhigher mean damage ratios, especially at lower intensity levels. As a matter offact, the curve for the NAC case forms an upper bound envelope for theMDRs applicable to the different seismic zones.

3.4. COMPUTATION OF THE EARTHQUAKE INSURANCE RATES

In computing the earthquake insurance rates for the different cities the bestestimate seismicity parameters and the best estimate attenuation equation(with uncertainty measure, racc ¼ 0.3) and the best estimate DPMs are used.Based on the seismic hazard results obtained for these cities, the EADRscorresponding to reinforced concrete buildings that are constructed inaccordance and not in accordance with the requirements of the Code (1975)are computed and given in Table VIII. These EADR’s can be interpreted asthe pure risk premiums (PRP) to be charged for every 1,000 Turkish Lira(TL) of insured property. The corresponding total premium rates which areobtained by multiplying the PRPs by the load factor of 1.67 (Equations (9)and (10)) are also shown in Table VIII. Since the southern part of Istanbulfalls into seismic Zone I and the northern part into Zone II, these regions aretreated separately and for Istanbul two different rates are given. For seismicZone I, where the seismic hazard is highest, three cities, Erzincan, Istanbul/south and Denizli are considered. The earthquake insurance rates computedfor these three cities are quite close to each other; therefore the average rates

0

5

10

15

20

25

30

35

40

45

V VI VII VIII IX

MMI

)%(

RD

M

Zone I (AC)

Zone II (AC)

Zone III(AC)Zone IV(AC)All Zones(NAC)

Figure 3. Variation of mean damage ratio with MMI for different seismic zones and

degree of compliance with the Code (AC: According to the Code; NAC: Not accordingto the Code).


obtained based on the values assessed for these three cities can be used forseismic Zone I. These average rates are presented in Table VIII.

In order to come up with a single value for the total earthquake insurancepremium rate, it is assumed that the reinforced concrete buildings in thesecities are constructed on the average with 50% compliance with the coderequirements. This leads to a rate, which is equal to the average of the ratescomputed for the cases of construction in accordance and not in accordancewith the code. The resulting rates are called as the ‘‘best estimate’’ totalearthquake insurance premium rates for these cities and are also shown inTable VIII.

Finally, the earthquake insurance premium rates obtained in this studyunder different assumptions are compared with the current practice of theinsurance companies in Turkey which implement the tariff specified byTCIP. These premium rates cover only insurance against earthquakedamage and are shown in Table VIII. It is to be noted that on the averagethe best estimate total insurance premium rate is about 2.7 times morethan the rate that the insurance companies currently are charging againstearthquake risk, making the purchase of earthquake insurance quitefeasible. This difference results from the fact that, in this study theearthquake insurance rates are computed considering a single property,whereas the insurance firms have a portfolio of n policy holders, where nis generally a very large number. The ratemaking of insurance companiesis based on the law of large numbers (central limit theorem) which implies

Table VIII. Pure risk premium (PRP) and total earthquake insurance premium (TP) rates fordifferent cities.

City (seismic zone) Pure risk

premium rate

PRP (1/1000)

Total earthquake insurance

premium rate TP (1/1000)

AC NAC AC NAC Best

estimate

Charged by

the insurance

firms

Erzincan (Zone I) 1.41 6.37 2.35 10.62 6.49 2.20

Istanbul/South (Zone I) 1.25 6.01 2.08 10.02 6.05 2.20

Denizli (Zone I) 1.07 6.08 1.78 10.13 5.96 2.20

Average for Zone I 1.24 6.15 2.07 10.25 6.16 2.20

Istanbul/North (Zone II) 1.05 3.31 1.75 5.52 3.64 1.55

Ankara (Zone III) 0.92 1.82 1.53 3.03 2.28 0.83

Konya (Zone IV) 0.73 1.22 1.22 2.03 1.63 0.55


M. S. YUCEMEN308

that as n increases the uncertainty on the expected loss becomes less,because the standard deviation of the expected (mean) loss decreases in-versely proportional to the square root of n. Since it is a more predictablerisk and dispersed to a large number of households, it becomes a moremanageable risk from the point of view of the insurance companies. Thisis the main reason for the difference between the rates estimated in thisstudy based on a single household and the rates currently charged by theinsurance companies in Turkey.

Within the context of the probabilistic approach it is also important tocomment on the different types of uncertainties that contribute to theinsurance rates and their variation with the degree of code compliance.The uncertainties can be categorized into two, namely: the aleatory(probabilistic) uncertainty that is inherent in earthquake occurrence anddamage potential, and the epistemic (knowledge-based) uncertainty that isassociated with the degree of code compliance. The aleatory uncertaintycannot be reduced by acquiring additional information (McGuire, 2004),whereas the epistemic uncertainty over the code compliance or violationcould be reduced by implementing a systematic program of buildinginspection and rehabilitation.

4. Conclusions

In this paper, a probabilistic model is presented for the calculation of theearthquake insurance rates and its application is illustrated by computing theearthquake insurance rates for five cities located in different seismic zones ofTurkey. The following main conclusions can be stated based on this study:

1. The computation of earthquake insurance rates requires information onfuture earthquake hazard and expected seismic vulnerability of engi-neering structures.

2. In this study all empirical data are utilized in the preparation ofDPMs. However, since observed data were not sufficient for estab-lishing DPMs for all of the seismic zones completely, other sources ofinformation, including subjective judgment of experts, are utilized. Itis to be emphasized that the determination of the appropriate DPMsis crucial as far as the validity of the resulting insurance rates areconcerned.

3. For seismic Zone I, three cities, namely: Erzincan, Denizli and thesouthern part of Istanbul are considered. No significant difference isobserved among these three cities with respect to the PRP and TP rates.Therefore the average value obtained from these three cities can beapplied for seismic Zone I.


4. For reinforced concrete buildings constructed according to the Code(1975), the pure and total premium rates decrease gradually, but consis-tently as it is moved from Zone I to Zone IV (Table VIII). Actually TPrates forZones I and II are quite close to each other (2.07& versus 1.75&).Similarly Zones III and IV have almost the same TP rates (1.53& versus1.22&). This trend suggests that for reinforced concrete buildings con-structed according to the requirements of the Code (1975), earthquakeinsurance premiumsmaybe implementedby treatingZones I and II as onegroup and Zones III and IV as another group. The TP rates computed forbuildings satisfying the code requirements in Zones I and II are observedto be slightly below and above, respectively, what has been currentlycharged by the insurance firms. On the other hand, for Zones III and IVthis difference is rather high; 1.85 and 2.22 times more, respectively.

5. For the case where the requirements of the Code (1975) are violated,total premium rates are again observed to decrease consistently as it ismoved from Zone I to Zone IV, but this time the differences are quitesignificant (Table VIII). In this case grouping is not possible. Besides theTP rates are much higher (3.6–4.7 times more) than the rates currentlycharged by the insurance companies.

6. As the final outcome of this study the best estimate total premium ratesare computed for each zone based on the assumption that reinforcedconcrete buildings are constructed on the average with 50% compliancewith the code requirements. The resulting best estimate TP rates areabout 2.7 times more than the rates currently charged for insuranceagainst earthquakes (Table VIII).

7. Much higher (up to five times more) insurance premium rates that resultfrom the violation of the code requirements strongly suggest thatcompliance with the code should be an important factor in deciding onthe earthquake insurance rates. In other words, significantly differentrates should be charged for buildings depending on the degree ofcompliance with the code. It is also believed that enforcement of such acriterion, will not only encourage the implementation of the coderequirements with respect to earthquake resistant design provisions, butit will also create a control mechanism.

8. It is to be emphasized that the earthquake insurance rates presentedherein are based on the SHA carried out and DPMs assessed within thescope of this study. Both of these inputs are obtained based on the datacurrently available and are subject to revision as new data becomeavailable. It is also to be noted that the rates are valid for reinforcedconcrete buildings and 1975 Code is considered, throughout the study.

9. The earthquake insurance rates are observed to be sensitive to theassumptions on SHA and DPMs. Therefore more consideration shouldbe given to the assessment of proper input parameters. Future studies

M. S. YUCEMEN310

should also concentrate on the collection of earthquake damage data,which is essential for obtaining a realistic estimate of the damage stateprobabilities.

Acknowledgements

Thanks are due to Nazan (Yilmaz) Ozturk, research assistant and AykutDeniz, graduate student at the Department of Civil Engineering, Middle EastTechnical University, for carrying out the numerical computations related tothe assessment of seismic hazard.

References

Akkaya, A. D. and Yucemen, M. S.: 2002, Stochastic modeling of earthquake occurrences andestimation of seismic hazard: A random field approach, Probab. Eng. Mech. 17, 1–13.

Amendola, A., Ermoliev, Y., Ermolieva, T. Y., Gitis, V., Koff, G., and Linnerooth-Bayer, J.:

2000, A systems approach to modeling catastrophic risk and insurability, Nat. Hazards 21,381–393.

Applied Technology Council: 1985, Earthquake Damage Evaluation Data for California,

Report ATC-13, Redwood City, California.Askan, A. and Yucemen, M. S.: 2003, Stochastic methods for the estimation of potential

seismic damage. In: Proceedings of the Ninth International Conference on Application of

Statistics and Probability in Civil Engineering, Berkeley, California, Vol. I, pp. 497–502.Blong, R.: 2003a, A review of damage intensity scales, Nat. Hazards 29, 57–76.Blong, R.: 2003b, A new damage index, Nat. Hazards 30, 1–23.Bommer, J., Spence, R., Erdik, M., Tabuchi, S., Aydinoglu, N., Booth, E., del Re, D., and

Peterken, O.: 2002, Development of an earthquake loss model for Turkish catastropheinsurance, J. Seismol. 6, 431–446.

Brillinger, D. R.: 1993, Earthquake Risk and Insurance, Environmentrics 4, 1–21.

Code: 1975, Specifications for Structures to be Built in Disaster Areas, Earthquake ResearchInstitute, Turkish Government Ministry of Reconstruction and Resettlement, Ankara.

Cornell, C. A.: 1968, Engineering seismic risk analysis, Bull. Seismol. Soc. Am. 58(5), 1583–

1606.Cornell, C. A. and Winterstein, S. R.: 1998, Temporal and magnitude dependence in earth-

quake recurrence models, Bull. Seismol. Soc. Am. 78, 1522–1537.

Elnashai, A. S.: 2000, Analysis of the damage potential of Kocaeli (Turkey) earthquake of 17August 1999, Eng. Struct. 22(7), 746–754.

Ferraes, S. G.: 2003, The conditional probability of earthquake occurrence and the next largeearthquake in Tokyo, Japan, J. Seismol. 7, 145–153.

Gencoglu, S., Ozmen, B., and Guler, H.: 1996, Settlement Locations and Earthquake, TurkishEarthquake Foundation, Ankara (in Turkish).

Gulkan, P., Yucemen, M. S., Basoz, N., Kocyigit, A., and Doyuran, V.: 1993, Seismic Zoning

Map for Turkey Based on the Most Recent Data, METU/EERC; Report No. 93–01,Ankara (in Turkish).

Gurpinar, A. and Yucemen, M. S.: 1980, An obligatory earthquake insurance model for

Turkey, In: Proceedings of the International Conference on Engineering for Protection fromNatural Disasters, Asian Institute of Technology, Bangkok, Thailand, pp. 895–906.


Gurpinar, A., Abali, M., Yucemen, M. S., and Yesilcay, Y.: 1978, Feasibility of ObligatoryEarthquake Insurance in Turkey, METU/ EERI Report No.78–05, Ankara (in Turkish).

Gutenberg, B. and Richter, C.: 1956, Earthquake magnitude, intensity, energy and accelera-tion, Bull. Seismol. Soc. Am. 46, 105–143.

Joyner, W. B. and Boore, D. M.: 1981, Peak horizontal acceleration and velocity from strong

motion records including records from the 1979 Imperial Valley, California earthquake,Bull. Seismol. Soc. Am. 71(6), 2011–2038.

Kamei, K.: 1976, Japanese report on outline of earthquake insurance in Japan, In: UNESCOConference on Assessment and Mitigation of Earthquake Risk, Paris.

Kunreuther, H.: 1996, Mitigating disaster losses through insurance, J. Risk Uncertainty, 12,171–187.

Lockett, J. E.: 1980, Catastrophes and catastrophe insurances, J. Inst. Actuaries Student Soc.

24, 1980, 91–134.Lomnitz, C.: 1974, Global Tectonics and Earthquake Risk, Elsevier, Amsterdam.McGuire, R. K.: 2004, Seismic Hazard and Risk Analysis, EERI, MNO-10, California.

Ozgur, C. and Yucemen, M. S.: 1997, Assessment of Seismic Hazard for Cyprus, Chamber ofCivil Engineers, Report No.: UCCTEA/CCE-97–01, Nicosia.

Ozmen, B. and Bagc�, G.: 2000, 12 November 1999 Duzce Earthquake Report, Ministry of

Public Works and Resettlement, General Directorate of Disaster Affairs, EarthquakeResearch Department, Ankara (in Turkish).

Scawthorn, C., Kunreuther, H., and Roth Jr, R.: 2003, Insurance and financial risk transfer,In: W.-F Chen. and C. Scawthorn (eds.), Earthquake Engineering Handbook, CRC Press,

Chapter 32.Smolka, A. and Berz, G.: 1989, The Mexico earthquake of September 19, 1985 – an

analysis of the insured loss and implications for risk assessment, Earthquake Spectra, 5,

223–248.Smolka, A. and Berz, G.: 1991, Seismic zoning in earthquake insurance: past and future, In:

Proceedings of the Fourth International Conference on Seismic Zonation, Stanford, Cali-

fornia, pp. 245–252.Steven, L.: 1992, The earthquake and war damage commission – a look forward (and a look

back), Bull. N. Z. Nat. Soc. Earthquake Eng., 25, 52–55.Straub, E.: 1973, Earthquakes and statistics from the insurance point of view, Bull. Inter.

Statist. Inst. 45(3), 462–466.Sucuoglu, H. and Tokyay, M.: 1992, Engineering Report of 13 March 1992 Erzincan

Earthquake, Turkish Society of Civil Engineers, Ankara (in Turkish).

Trifunac, M. D. and Brady, A. G.: 1975, On the correlation of seismic intensity scale withpeaks of recorded ground motions, Bull. Seismol. Soc. Am. 45, 139–162.

Vere Jones, D.: 1973, The statistical estimation of earthquake risk, N. Z. Stat. 8, 7–16.

Wald, D. J., Quitoriano, V., Heaton, T. H., and Kanamori, H.:1999, Relationships betweenpeak ground acceleration, peak ground velocity and modified Mercalli intensity in Cali-fornia, Earthquake Spectra 15(3), 557–564.

Walker, G. R.: 2000, Earthquake engineering and insurance: past, present and future, Aon ReAustralia (available on the Internet at: http://www.aon.com.au/reinsurance/knowledge/quake_engineering.asp).

Wasti, S. T. and Sucuoglu, H.: 1999, Rehabilitation of Moderately Damaged Reinforced

Concrete Buildings after the 1 October 1995 Dinar Earthquake, METU/EERC, ReportNo. 99–01, Ankara (in Turkish).

Whitman, R. V.: 1973, Damage Probability Matrices for Prototype Buildings, Dept. of Civil

Engineering, MIT, R73–57, Cambridge.

M. S. YUCEMEN312

Yucemen, M. S.: 1992, Seismic hazard maps for Jordan and vicinity, Nat. Hazards 6(3), 201–226.

Yucemen, M. S.: 1993, A random field model for the estimation of seismic hazard, Struct.Safety 12, 187–203.

Yucemen, M. S.: 2002, Prediction of potential seismic damage to reinforced concrete buildings

based on damage probability matrices, In: Proceedings of the 6th International Conferenceon Concrete Technology for Developing Countries, Amman, Vol. 3, pp. 951–960.

Yucemen, M. S. and Akkaya (Dener), A.: 1996, A comparative study of stochastic models forseismic hazard estimation, In: M. I. El-Sabh, S. Venkatesh, H. Denis and T. S. Murty (ed),

Land-Based and Marine Hazards, Kluwer Academic Publishers, The Netherlands, Vol. 7,pp. 5–24.

Yucemen, M. S. and Askan, A.: 2003, Estimation of earthquake damage probabilities for

reinforced concrete buildings, In: S. T. Wasti and G. Ozcebe (eds), Seismic Assessment andRehabilitation of Existing Buildings, NATO Science Series IV: Earth and EnvironmentalSciences, Kluwer Academic Publishers, The Netherlands, Vol. 29, pp. 149–164.

Yucemen, M. S. and Bulak, S.: 1997, Assessment of earthquake insurance premiums based onstatistical methods, In: Proceedings of the Fourth National Earthquake Engineering Con-ference, Ankara, pp. 699–707 (in Turkish).

Yucemen, M. S. and Bulak, S.: 2000, Estimation of earthquake damage on structures: damageprobability matrices, Bull. Chamb. Civil Eng. 3, 15–22 (in Turkish).

Yucemen, M. S. and Gulkan, P.: 1994, Seismic hazard analysis with randomly located sources,Nat. Hazards 9, 215–233.


1Earthquake Risk Insurance

KNOWLEDGE NOTE 6-2CLUSTER 6: The economics of disaster risk, risk management, and risk financing

Earthquake Risk Insurance

2 KNOWLEDGE NOTE 6-2

Prepared by Olivier Mahul and Emily White, World Bank


Earthquake Risk Insurance

The March 2011 earthquake that hit East Japan was the fourth-largest ever recorded. It was not only a human tragedy but an economic shock with losses estimated in excess of ¥16,900 billion, making it the costliest disaster in history. Despite this, the Japanese insurance industry is expected to emerge without significant financial impairment, thanks to a well-developed residential earthquake risk insurance dual program (with private nonlife insurers and cooperative mutual insurers) based on conservative control of insurers’ liabilities (through insurance policy structures and reinsurance). Meanwhile, more than half of Japanese homeowners are still unin-sured, creating a significant fiscal burden for the government.

FINDINGS

Residential earthquake insurance: A dual program with carefully controlled liabilities

Residential earthquake insurance coverage in Japan relies on two major actors: nonlife private insurers and cooperative mutual insurers. Despite major differences in their finan-cial management of earthquake risk, these two insurance systems demonstrated their efficiency in claims settlements and their financial viability after the Great East Japan Earth-quake (GEJE). Table 1 compares the residential earthquake insurance scheme offered by the private nonlife insurance companies with the scheme offered by the largest cooperative mutual insurer, the National Mutual Insurance Federation of Agricultural Cooperatives (also known as JA Kyosai*). While the perils covered, assets covered, and extent of coverage are similar across the two programs, earthquake coverage is offered on a voluntary basis with risk-based premium rates by private insurers, and on an automatic basis with flat rates by cooperative mutual insurers.

Both programs are based on conservative control of insurers’ liabilities. In both programs, the claims payments are not intended to provide complete coverage: the maximum

KNOWLEDGE NOTE 6-2CLUSTER 6: The economics of disaster risk, risk management, and risk financing

* Also known as Zenkyoren.


coverage is limited at 50 percent of the fire insurance amount (subject to upper limits). Like-wise, both programs rely on sophisticated reinsurance strategies. The reinsurance protec-tion of the private insurance scheme relies on a catastrophe insurance pooling mechanism, the Japanese Earthquake Reinsurance Co. (JER), backed by the government of Japan. In contrast, reinsurance protection for cooperative mutual insurers is provided by the inter-national reinsurance and capital markets, with no government intervention. In both cases, the use of reinsurance serves to limit the liability of the private or cooperative risk carriers.

Penetration under the private nonlife insurance program is estimated at about 25 percent of Japanese households, with just under 13 million residential earthquake insurance policies in force: an estimated 48 percent of all fire insurance policies in force include earthquake coverage. Cooperative mutual insurance programs cover about 14 percent of Japanese households, so that total penetration is estimated at 39 percent.† JA Kyosai holds a very large share of the cooperative mutual insurer market, with 5.4 million households holding building endowment policies covering residential earthquake risk (11 percent of total Japa-nese households). The cooperative mutual insurer Zenrosai has an additional 1.7 million

TABLE 1: The dual residential earthquake insurance system in Japan

Private non-life insurersCooperative mutual insurer

JA Kyosai

Perils covered Earthquake, volcanic eruption, tsunami

Earthquake, volcanic eruption, tsunami

Assets covered Residential dwelling and content

Residential dwelling and content

Extent of coverage 30–50 percent of fire insurance amount with limits

Up to 50 percent of fire insurance amount with limits

Coverage purchase Optional endorsement to residential fire insurance policy

Automatically included in building endowment policy

Premium rate Risk-based rates (by risk zone and type of construction)

Flat rates (wooden/nonwooden)

Reinsurance Japan Earthquake Reinsurance Co. (JER) and Japanese government

International reinsurance and capital markets

Loss adjustment 3-step system Proportional system

Penetration of earthquake coverage (percent households)

25% 11%

† The number of households is estimated at about 51 million (Government of Japan, Statistics Bureau). Policy-in-force data from the Japanese Non-Life Insurance Rating Organization (2010), JA Kyosai Business Operations (2011), and Zenrosai Annual Report (2010). Cooperative mutual insurer figures extrapolated based on 85 percent estimate of the JA Kyosai market share.


Liability of government

Liability of JER

Liability of insurance companies

¥115billion

¥871billion

¥5,500billion,220-yearreturnperiod ¥4,397.55

billion

¥378billion

¥115.7billion

¥115.75billion

¥305.7billion

¥115 billion

¥72.3 billion

FIGURE 1: Japanese earthquake reinsurance program (as of May 2011)

Source: JER 2011a.

natural disaster policies covering residential earthquake risk, accounting for a further 3 percent of total Japanese households.

Private nonlife insurance companies and the Japanese earthquake reinsurance company

Earthquake insurance offered by private nonlife insurance companies is available as an optional endorsement to fire insurance policies. Earthquake coverage is available at policy limits of 30 percent to 50 percent of the fire insurance limit, with maximum limits of ¥50 million per dwelling and ¥10 million for personal property.

A three-step claims settlement allows for rapid damage assessment and claims settle-ment. Payouts are not proportional to damage, but based on a three-step system: total loss, half loss, and partial loss—which allow for 100 percent, 50 percent, and 5 percent of the earthquake insurance policy limit, respectively.

The premium rates are risk based, and vary according to the prefecture where the dwelling is located (divided into eight risk zones) and type of construction (wooden or nonwooden). For an insured amount of ¥10 million, the annual premium varies between ¥5,000 for a nonwooden structure in Nagazaki Prefecture, and ¥31,300 for a wooden structure in Tokyo. Discount rates of up to 30 percent apply when the building is earthquake resistant, according to the Japanese Housing Performance Designation Standards, including a 10 percent discount for buildings constructed after 1981. The premium rates, calculated by the Non-Life Insurance Rating Organization, consist of the pure premium rate and a loading


factor. It should be noted that the rates do not include any loading for profit since the program is not for profit. Despite this rating and because of Japan’s considerable earth-quake exposure, rates are still considered high.

The 1966 Earthquake Insurance Law (enacted after the Niigata earthquake of 1964) estab-lished the JER, to whom private nonlife insurers were obliged to offer earthquake insur-ance and cede 100 percent of the earthquake premium and liabilities. The JER thus acts as the sole earthquake reinsurer for the private insurance market. The JER can be seen as an earthquake reinsurance pool, retaining a portion of the liability and ceding the rest back to private insurers (based on their market share) and to the Japanese government through reinsurance treaties. The reinsurance program is designed such that the liability of private insurers and the JER itself does not exceed the accumulated reserves from earthquake insurance premiums. Figure 1 describes the Japanese earthquake reinsurance program as revised in May 2011 after the GEJE. The total claims-paying capacity of the program is currently ¥5,500 billion, which is estimated to correspond to the scenario of the 1923 Great Kanto earthquake with a return period of 220 years.‡ Should insured earthquake losses exceed this amount, claims would be prorated.

The role of the Japanese government is central to the program. The maximum liability of the government of Japan, JER, and private insurers is 87 percent, 10 percent, and 3 percent, respectively. It should be noted that under the previous reinsurance program (before May 2011), the government’s liability was only 78 percent, and the rest was shared equally between the JER and private insurers. The revision of the reinsurance program, leading to an increase of the government’s liability share, is the direct consequence of a depletion of the earthquake reserves of both the JER and private insurers after the GEJE.

Japanese accounting standards allow the insurers to build up preevent catastrophe reserves (by accumulating the earthquake insurance premiums received, less expenses and any underwriting gains and investment income) over time with separate resources to pay claims, the size of which is based on the probable maximum loss of the insurer’s portfolio. Likewise, the government of Japan has set up a special account to accumulate its reserves. Table 2 shows the amount of reserves at end of fiscal years 2010—that is, before the GEJE. The GEJE wiped out about half of the program’s earthquake reserves.

‡ The total claims-paying capacity of the program will increase to ¥6.2 billion in 2012 (Ministry of Finance 2012).

TABLE 2: Reserves under the earthquake insurance program

Source: JER 2011a.

¥ billion End of fiscal year 2010

Government 1,343

JER 424

Private insurers 489

Total 2,256


It is noteworthy that the total reserves supporting the Japanese Earthquake Reinsur-ance Program, even before the GEJE, represent only a fraction of the liability of all stake-holders. The size of this potential gap is largely due to the government’s reserve-to-liability ratio under the program, which appears low. In case of a major earthquake exceeding the reserves available, it would be critical to immediately mobilize additional resources to ensure the financial solvability of the program.

Cooperative mutual insurers

Residential earthquake insurance is also available through cooperative mutual insurers. These insurers conduct insurance operations on behalf of Japan’s cooperative societies. The largest of these cooperatives is JA Kyosai, which holds an estimated 85 percent market share of all the homeowners insurance written through cooperative mutual insurers. Like any cooperative, JA Kyosai operates on a nonprofit basis. Its insurance products are different from those of private insurers. Cooperative mutual insurers offer building endow-ment policies: these policies offer more comprehensive coverage than the policies avail-able through the private insurers and can therefore be seen as a savings mechanism that provides funding for home repairs, whether caused by natural disasters or other adverse events. The five-year (or longer) term policy automatically covers residential dwellings and personal property from damage caused by fire, flood, earthquakes, and other natural disas-ters. If the policy expires and the policyholder has not claimed a total loss, he or she is entitled to a partial refund of the premium. At the start of 2011, JA Kyosai’s client base comprised more than 11 million building endowment policies.

Earthquake insurance is automatically included in the building endowment policies offered by JA Kyosai. The policy limit is 50 percent of the fire insurance limit, up to ¥250 million. The average fire insurance amount is ¥30 million, hence the average earthquake insurance limit is ¥15 million.

Under the building endowment policy available through JA Kyosai, the claims settlement process in case of an earthquake is proportional: a loss assessor estimates the damage percentage of the house, and this rate is applied to the earthquake policy limit.

The premium rate is flat, that is, the same wherever the dwelling is located. It only differs according to whether the building is a wooden or nonwooden structure.

Cooperative mutual insurers are not subject to the Earthquake Insurance Law and do not participate in the JER. They work outside the nonlife insurance regulatory framework and are instead accountable to their respective ministries; for example, JA Kyosai reports to the Ministry of Agriculture, Forestry, and Fisheries. In contrast to private nonlife insurers, cooperative mutual insurers cede a significant portion of their liabilities to the interna-tional reinsurance market. JA Kyosai is known to have one of the largest reinsurance programs in the world, with reinsurance capacity in excess of ¥75 billion. Its large and well-diversified asset base also allows it to retain a significant portion of its liability. In addition to traditional reinsurance, JA Kyosai has issued catastrophe (Cat) bonds to better spread its risk (see box 1).


Industrial and commercial earthquake insurance

Traditionally, industrial and commercial earthquake insurance has been issued as a reduced indemnity policy, which provides limited coverage on a proportional basis. The extent of the coverage depends on the location of the asset, for which the country has been divided into 12 risk zones. The indemnity limit varies from 15 percent in Tokyo up to 100 percent in Niigata. Following the enactment of the Insurance Business Law in 1996, which largely deregulated the insurance market in Japan, insurance policies on a first-loss basis were also offered, which generated a significant increase in the sum insured (the maximum amount that could be paid out). Loss of revenue and business interruptions caused by earthquakes have not traditionally been marketed and have low penetration rates.

Other classes include earthquake fire expense insurance. This is a limited amount for fire following an earthquake, which is provided automatically with some insurance policies, such as the storekeepers’ comprehensive policy. The coverage is limited to 5 percent of the fire sum insured, up to certain fixed limits. Other insurance policies that generally include earthquake coverage are cargo insurance, motor insurance, and engineering insurance.

BOX 1: Innovative catastrophe risk financing: Capital markets protect Japanese farmers against earthquake

In 2008, Munich Re, a reinsurance company based in Germany, issued JA Kyosai’s second catastrophe (Cat) bond, a $300 million issue, through the special-purpose vehicle, Muteki Ltd.

Cat bonds are index-linked securities that secure financial resources on the capital markets, to be disbursed in case of the occurrence of a predefined natural disaster. Cat bonds generally cover the highest level of risk and are mainly issued for specific perils with an annual probability of occurrence of 2 percent or less (that is, a return period of 50 years or more). Unlike traditional reinsurance, Cat bonds are fully collateralized and offer multiyear coverage (usually 3 to 5 years).

The three-year Muteki Cat bond provided fully collateralized protection for Japanese earthquake exposure indirectly to JA Kyosai/Zenkyoren, through a reinsurance agree-ment with Munich Re, which served as counterparty on the transaction. Like other Cat bonds in Japan, the Muteki Cat bond was parametric, triggered by the location and magnitude of an earthquake rather than the actual losses. Following the GEJE disaster, the Muteki Cat Bond became the first Cat Bond to pay out on the occurrence of an earthquake event. The instrument released the full coverage limit of $300 million in response to the event.

In February 2012 Guy Carpenter and Company announced the placement of a $300 million Cat bond, through the SPV Kibou Ltd, which would ultimately benefit JA Kyosai. It provided protection on a parametric basis, using earthquake data gathered from various recording stations from the Kyoshin-Net network of seismographs.


Economic and insured losses

The GEJE caused major direct economic losses, with current estimates of ¥16,900 billion (KN 6-1). Private (residential, commercial, and industrial) buildings represented 62 percent, and public infrastructure represented 13 percent of the (direct) economic losses (see annex 1). Insured losses were estimated at ¥2,750 billion, or 16 percent of total economic losses. Residential assets represented 78 percent of insured losses. Fifty-six percent of the resi-dential insured losses were covered by private insurers and the JER, and 44 percent were covered by cooperative mutual insurers (see annex 1).

Despite significant differences, both private and mutual residential earthquake insur-ance programs had adequate capacity to meet their claims obligations, thanks to efficient management of exposure to losses through a combination of policy limits and reinsurance protection. The earthquake insurance program managed by the private nonlife insurance companies faced an estimated total loss of ¥1,200 billion, with 42 percent retained by private insurers, 13 percent retained by the JER, and 45 percent retained by the govern-ment. This event, however, severely depleted the earthquake reserves of both the private insurers and JER, leading to an increase in government liability in the revised reinsurance program of 2012. Earthquake losses incurred by JA Kyosai were estimated at ¥830 billion, 90 percent of which were residential losses. It is estimated that about 58 percent of those losses were reinsured.

The three-step earthquake claims settlement system implemented by the private insur-ance companies allowed claims to be settled rapidly. Satellite images were also used to identify total losses on buildings, which helped further speed up claims settlements. In the aftermath of the disaster, the General Insurance Association of Japan designated specific total loss zones, based on satellite imagery (KN 5-2). Any total loss claims filed within these areas did not require additional confirmation of incurred losses, thereby speeding up the payout process. Out of ¥1,200 billion generated by the 741,000 claim payments made after the GEJE, 60 percent were paid within two months and 90 percent within five months.

Comparative analysis of the GEJE with other recent earthquakes

It is interesting to compare the economic and fiscal impact of the GEJE with the impact of other recent earthquakes: the 2010 earthquake in Chile and the 2011 earthquakes in Canter-bury, New Zealand. All three earthquakes were very large in magnitude and caused severe economic losses in their countries. Table 3 summarizes this comparative analysis. While the GEJE caused the largest economic losses in absolute terms, losses as a percentage of gross domestic products (GDP) are lower than those in Chile and New Zealand given the size of the Japanese economy. The government’s portion of direct losses (that is, addi-tional expenditures), expressed as a percentage of total government expenditures, were estimated at 8 percent for the GEJE and 11 percent for the Canterbury earthquake in New Zealand. Finally, the fraction of the insured losses covered by international reinsurance was estimated at 95 percent in Chile, 29 percent in New Zealand (where the Earthquake Commission EQC retained a large fraction of the losses), and 23 percent in Japan. This last figure hides a large difference between the JER, which relies on public reinsurance and cooperative mutual insurers, such JA Kyosai, that purchase most of their reinsurance capacity abroad.


LESSONS

Some key lessons can be drawn from the review of Japan’s earthquake insurance programs in the light of the GEJE:

• No one-size-fits-all. The dual earthquake insurance programs in Japan illustrate that there is no one-size-fits-all catastrophe insurance program. Two very different schemes can coexist successfully within a country significantly exposed to earth-quakes, offering earthquake coverage to about four households out of ten in Japan.

• Resilience is critical for earthquake insurance programs. Both programs managed to fulfill their obligations after the GEJE without difficulties, because of the sound management of policy limits and conservative reinsurance coverage. The apparent resilience of the current setup does not mean, however, that there is no room for these schemes to improve without compromising sustainability. For example, the earthquake insurance limit offered by JA Kyosai started at 10 percent and has increased progressively to 50 percent currently.

TABLE 2: Comparative analysis of the Tohoku (GEJE), Canterbury, and Maule earthquakes

Source: Swiss Re 2011; Ain Benfield 2011; Ministry of Finance Japan 2012; New Zealand Treasury 2011; RMS 2011.

Note: Direct economic losses are defined as damage to physical assets (including infrastructure).

Tohoku, JapanCanterbury,

New Zealand Maule, Chile

Year 2011 2011 2010

Magnitude 9.0 6.3 8.8

Estimate direct economic losses ($ billion)

225 15 20

Estimated direct economic losses (% GDP)

4 9 9

Estimated direct losses borne by government (as % of government expenditures)

8 11 n/a

Estimated insured losses (% of direct economic losses)

16 80 40

Estimated insured losses covered by international rein-surance

23 73 95


• Rapid claims settlement can be achieved, even after a major disaster. The three-step claims adjustment system implemented by the private insurers allows for rapid damage assessment and claims settlement. It also takes into account that, immediately after a major disaster, large numbers of loss assessors have to be deployed at the same time. The simplicity of the three-step system allows this to happen.

• Insurance penetration in Japan is high, but there is still considerable room for expansion. About 40 percent of Japanese households have earthquake insurance coverage, leaving 60 percent of households without coverage. International experi-ence shows that it is very difficult, if not impossible, to increase the penetration rate beyond a certain level on a voluntary basis. Compulsory earthquake insurance could therefore be considered.

The GEJE also highlighted certain challenges of earthquake insurance programs run by private insurance companies:

• The JER claims-paying capacity is limited in the aggregate. The aggregate limit is currently set at ¥5,500 billion (to be increased to ¥6,200 billion in 2012), which would be sufficient for a major earthquake such as the Great Kanto earthquake in 1923. But this does not take into account the occurrence of consecutive major earthquakes, which could jeopardize the solvency of the program.

• The government’s liability under the JER exceeds its ex-ante financing arrange-ments. The government’s maximum liability is adjusted based on the balance of earthquake reserves of the private insurers and the JER and the maximum defined liability under the program. The government currently holds 87 percent of the total liability of the program. Its current special account would not be sufficient to cover this level of liability and would require an immediate budget appropriation or real-location in case of a major disaster.

• Limited policy coverage may not meet the needs of the insured. The program is designed to provide partial coverage (up to 50 percent of the fire insurance coverage limit) to “stabilize the livelihood of the earthquake victims” (article 1 of the 1966 Earthquake Insurance Law). There seems to be a growing demand for higher coverage, but such an increase in coverage should be carefully evaluated to maintain the financial sustainability of the system.

• The claims settlement process introduces significant basis risk and could be revised. Although the three-step claims adjustment process allows for rapid settle-ment of claims, there is a big gap between payouts for partial loss (5 percent) and half loss (50 percent). This increases the risk that payments will not match the needs of the insured party following the occurrence of damage (basis risk). A fourth intermediate step could be introduced to reduce this risk.

• Catastrophe risk modeling for Japan is sophisticated, but could be improved. State-of-the-art catastrophe risk models have been developed for Japan, but need to be further refined as secondary loss perils such as tsunamis (which caused about 30 percent of the total losses from the GEJE) and liquefaction are not included


as standard in all models. These models could also be used to further assess the catastrophe risk exposure of public buildings and infrastructures.

RECOMMENDATIONS FOR DEVELOPING COUNTRIES

Developing viable and affordable catastrophe risk insurance programs

Japanese earthquake insurance programs demonstrated considerable resilience after the GEJE. From this experience, recommendations can be made to disaster-prone developing countries willing to promote catastrophe risk insurance to help them promote viable and affordable programs and clearly define the role of the government in public-private partner-ships (PPPs).

Structure policies to allow for sustainable and affordable programs. Catastrophe risk insurance policies should be designed to enable insurance companies control their liabili-ties and offer affordable coverage. The policy structure can be revised over time to better respond to the needs of the policyholders, while also ensuring the system’s resilience to major disasters. The partial coverage produced by both Japanese earthquake insurance programs and the simplified loss adjustment process of the private insurer system help to keep costs down.

Price insurance premiums based on the underlying risks. Insurance premiums should reflect the underlying risks with respect to the various risk zones and types of construction. Risk-based insurance premiums make policyholders aware of the underlying cost of risk, thereby providing financial incentives to engage in disaster risk mitigation. Even in cases where the full cost of cover is not passed onto the policyholder, it is still possible to signal the underlying cost of risk by making subsidies transparent.

Provide incentives to invest in disaster risk mitigation. Additional financial incentives, such as discounts on premium rates or lower deductibles, can be offered to the policy-holders who invest in risk reduction.

Consider mechanisms for enforcing insurance purchase. Voluntary catastrophe risk insurance does not typically generate high penetration rates, even in highly developed insurance markets. Some type of compulsory mechanism, such as an automatic catas-trophe guarantee in fire insurance policies, may be necessary to ensure that a large propor-tion of the population is insured against natural disasters.

Promote multiple-catastrophe risk insurance delivery channels. Catastrophe risk insurance should leverage existing nonlife insurance delivery channels, such as private insurers or mutual insurers. The Japanese system demonstrates that different segments of the population may be best served by different delivery channels, even for very similar products. Multiple distribution channels for catastrophe risk insurance should therefore be explored.


Develop detailed catastrophe risk models. Detailed catastrophe risk models and data-bases are essential for detailed risk assessment, premium rate calculation, and efficient management of catastrophe risk insurance liabilities. In addition to a strong hazard model, such assessments also require detailed exposure databases of at-risk assets (buildings and infrastructure) and detailed vulnerability functions to translate hazard values into dollar losses. These models are typically developed by private risk modeling firms and licensed to the insurance industry. But for some less-developed insurance markets, governments and donors have funded or partially funded the development of such models as public goods to support market development.

Develop catastrophe risk insurance market infrastructure. Catastrophe risk insurance markets require major investments in basic infrastructure, such as catastrophe risk models, exposure databases, product design and pricing, and the like. Governments can play a major role in developing this kind of infrastructure to help the private insurance industry can offer cost-effective and affordable insurance solutions.

Promote enabling legal and regulatory environments. Unlike traditional lines of insur-ance business such as automobile insurance, catastrophe risk insurance can generate large correlated losses for insurers. The legal and regulatory framework should enforce adequate pricing, reserving, and reinsurance buying to ensure that insurers will meet their claims in full in the event of a disaster.

Promote PPPs for catastrophe insurance programs. Governments can play an important role in building an affordable and sustainable earthquake insurance program. As the private insurance sector brings its technical expertise and financial capacity to the table, govern-ments can support the development of public goods and risk-market infrastructure to foster sustainable market-based insurance solutions.

Governments can play a role as the financier of last resort. Governments may want to act as financiers of last resort when private reinsurance capacity is unavailable or too expensive to allow domestic insurers to offer cost-effective insurance solutions. Govern-ments should not compete with the private reinsurance market but rather complement it. When needed, governments should make financial capacity available to domestic insurers through public reinsurance or (contingent) credit.


Insurance schemes in agriculture and fishing helped farmers and fishermen stabilize their businesses by compensating them for losses and damages caused by the GEJE. Insurance paid for some level of damage sustained by almost all fishing boats. In Japan, these schemes began as cooperative activities by local farmers and fishermen. They were subsequently turned into voluntary mutual aid programs established by the government, which subsidizes the premiums paid by farmers and fishermen, covers part of the administrative costs, and reinsures the insurance associations.

Fishery insurance

The earthquake and tsunami damaged some 25,000 fishing vessels, at a cost of ¥170 billion. Ninety percent of the vessels in Iwate, Miyagi, and Fukushima prefectures were damaged, which had an enormous effect on the fishing industry since these vessels were used for aquaculture as well as fishing. Before the tsunami, the three prefectures accounted for 10 percent of the total catch in Japan (excluding aquacul-ture). Aquaculture industries were also severely damaged, particularly in the Iwate and Miyagi prefectures, where production of oysters and wakame seaweed is wide-spread. Damage to aquaculture amounted to ¥131 billion: 57 billion for production and 74 billion for facilities.

Policies in force for agricultural, fishing boat, and fisheries insurance in 2009

Number of households

underwritten (thousands)

Area underwritten (thousands of

hectares)Value covered (¥

million) Penetration

Farm products

Paddy rice 1,752 1,479 1,223,157 91% (area)

Field rice 0.4 0.2 46 5% (area)

Wheat and barley 49 252 83,277 95% (area)

Fruit trees

Harvest mutual relief 76 45 (number of boxes)

107,200 26% (number)

Tree mutual relief 4 1 7,000 2% (number)

Livestock 89 6.665 (Number of livestock)

724,585 42% (number)

Field crops 82 259 140,400 62 %

Fishing boats 192 (boats) n.a. 1,028,517 >100% (number of boats)

Fisheries 61 n.a. 394,155 52% (households)

n.a. = Not applicable.

BOX 2: Agriculture and fishery insurance


BOX 2 ,CONTINUED

The fisheries insurance system in Japan is well organized, providing essential insur-ance services at a reasonable cost to all fishermen including small-scale producers. The fishing vessel insurance system, which was established in 1952 under the Fishing Vessel Damage Compensation Law, aims at stabilizing fishing businesses by covering the loss of and damages to their fishing vessels. The system includes the following insurances:

• Fishing vessel insurance covering basic damage caused by accidents and disas-ters, and including special insurance for damage caused by war and seizure.

• Protection and indemnity insurance covering compensation for the crew and damages incurred during navigation.

• Owner-operator insurance covering the death of owner-operators.

• Cargo insurance covering the loss of catches or cargo.

• Pleasure boat insurance covering compensation, rescue costs, and damages.

• Transshipped catches insurance.

• Crew salary insurance covering crew salaries if vessels are seized.

The fisheries mutual insurance scheme, which was established in 1964 under the Fish-eries Disaster Compensation Law, aims at stabilizing small- and medium-size fishing and aquaculture operations by covering losses from poor catches caused by natural disasters. The system insures fish harvests, aquaculture, special aquaculture, and fishing gear.

The government subsidizes one-third to one-half of the premium. While fishing vessel insurance enjoyed a surplus of ¥16.5 billion in 2010, the Fisheries Mutual Insurance Scheme suffered a deficit of ¥28.9 billion.

Fishing vessel insurance system

Fisheries mutual insurance scheme Total

Government 72.7 (78%) 21.3 (77%) 94.0 (78%)

Reserve of government special account 11.0 (12%) — 11.0 (9%)

Associations at national level 1.4 (2%) 3.0 (11%) 4.4 (4%)

Associations 7.8 (8%) 3.2 (12%) 11.0 (9%)

Total 92.9 (100%) 27.5 (100%) 120.4 (100%)


BOX 2 ,CONTINUED

The Ministry of Agriculture, Fishery, and Forests estimates that total claims would amount to ¥120.4 billion, of which the central government will cover ¥94 billion, or 78 percent for the GEJE. As of March 13, 2012, ¥63.4 billion in claims have been paid out: ¥47.5 billion under the fishing vessel insurance system, and ¥15.9 billion under the fisheries mutual insurance scheme. Sixty percent of vessels were insured under the vessel insurance scheme, of which some 80 percent of boats were over 20 tones. Some 80 percent of the insured vessels were more than 15 years old. Since the schemes cover the residual value of the vessels, the claims paid out may not cover the replacement costs.

Agriculture insurance

Damage to agricultural production and facilities from the GEJE event amounted to ¥63 billion. Rice is an important crop in Japan, but because the GEJE happened before the rice-growing season, insurance almost did not cover rice production losses. Since compensation related to the accident at the Fukushima Nuclear Power Plant has not yet been decided, the total payout on agricultural insurance is uncertain. In Miyagi Prefecture, the agricultural insurance scheme has covered damages to greenhouses in the amount of ¥1billion.

The Farm Losses Compensation Law introduced the agricultural insurance scheme in 1947 to help farmers stabilize their businesses by covering damages caused by natural disasters; the scheme offers insurance coverage for almost all major agricultural prod-ucts. It was started by local farmers as a cooperative initiative to set up a reserve fund


BOX 2 ,CONTINUED

to pay for insurance premiums, which evolved into agricultural mutual relief associa-tions. The insurance scheme includes: rice, wheat, and barley insurance (mandatory for paddy fields of more than 20 hectares); livestock insurance; fruit and fruit tree insurance; field crop and horticultural insurance; greenhouse insurance; and houses and properties. The government subsidizes half of farmers’ premiums.

Prepared by Mikio Ishiwatari, World Bank.

REFERENCES

Benfield, Aon. 2011. “Earthquake Insurance Business in Japan.” December 2011.

General Insurance Association of Japan. 2011. Annual Report 2010−2011.

JA Kyosai. 2011. Annual Report 2010, Business Operations.

Japan Credit Rating Agency Ltd. 2011. “JCR Affirmed AAp/Stable on Japan Earthquake Reinsurance.” December 28.

JER (Japan Earthquake Reinsurance Co., Ltd). 2011a. Annual Report 2011.

———. 2011b. “Response to the Great East Japan Earthquake by the General Insurance Industry.” Presentation, World Forum, Jamaica, October 25−26, 2011.

McAllister, S., and E. Cohen. 2011. “Japanese Casualty Insurers Show Resilience.” www.contingencies.org.

Muir-Wood, R. 2011. “Designing Optimal Risk Mitigation and Risk Transfer Mechanisms to Improve the Management of Earthquake Risk in Chile.” OECD Working Papers on Finance, Insurance and Private Pensions No. 12, Organisation for Economic Co-opera-tion and Development, Paris.

Non-Life Insurance Rating Organization of Japan. 2011. www.nliro.or.jp.

SCOR Global P&C. 2011. Technical Newsletters, December and October 2011.

Swiss Re. 2012. “Lessons from Major Earthquakes.” Economic Research and Consulting, January 2012.

Zenrosai. 2011. Annual Report 2010.


Annex 1. Economic and insured losses of the Great East Japan Earthquake (GEJE)

GEJE: Economic losses by sector, as percent of total loss (¥16,900 billion)

GEJE: Insured losses by sector, as percent of total insured losses(¥2,750 billion)

GEJE: Insured residential losses by scheme, as percent of total insuredresidential losses (¥2,137 billion)

Mutualinsurers43.8%

Privateinsurers56.2%

Commercial/industrial

22%

Residential78%

62%

7%

8%

11%

13%

Private buildings

Public infrastructure

Agriculture

Lifeline infrastructure

Other buildings


Annex 2. Estimated GEJE insured residential losses, by earthquake insurance program

GEJE: JER earthquake insurance claims (¥1,200 billion)

GEJE: JA Kyosai earthquake insurance claims (¥830 billion)

Government45.2%

Insurers42.0%

JER12.8%

Retention42%

Reinsurance58%

Probabilistic Assessment of Earthquake

Insurance Rates for Turkey

M. S. YUCEMENDepartment of Civil Engineering and Earthquake Engineering Research Center, Middle EastTechnical University, 06531 Ankara, Turkey; E-mail: [email protected]

(Received 7 May 2004; accepted 3 November 2004)

Abstract. A probabilistic model is presented to obtain a realistic estimate of earthquake

insurance rates for reinforced concrete buildings in Turkey. The model integrates informationon seismic hazard and information on expected earthquake damage on engineering facilities ina systematic way, yielding to estimates of earthquake insurance premiums. In order to dem-

onstrate the application of the proposed probabilistic method, earthquake insurance rates arecomputed for reinforced concrete buildings constructed in five cities located in differentseismic zones of Turkey. The resulting rates are compared with the rates currently charged by

the insurance companies. The earthquake insurance rates are observed to be sensitive to theassumptions on seismic hazard and damage probability matrices and to increase significantlywith increasing violation of the code requirements.

Key words: seismic hazard, earthquake engineering, earthquake insurance, building damage,

damage probability matrix, risk premium, insurance premium

1. Introduction

Turkey is located in one of the seismically most active regions of the world.Geological and seismological data indicate that in almost all parts ofTurkey seismic hazard is significant: 95% of the population, 92% of thetotal area and 98% of the industry are under earthquake threat (Gencogluet al., 1996). To reduce the vulnerability of human settlements and indus-trial facilities, the engineered structures must be designed and constructedto resist the effects of earthquakes. However, due to poor control of con-struction in Turkey, a large degree of damage is expected if a majorearthquake hits a large city, as observed during the recent earthquakes.Accordingly, different measures have to be considered to alleviate the post-disaster consequences of earthquakes as well as enforcement of the earth-quake resistant provisions of the Code. The implementation of obligatoryearthquake insurance is among such measures.

In Turkey, the State used to have a legal obligation to fund the costs ofreconstructing buildings after an earthquake. This responsibility of the Statenaturally brought an unplanned burden on the national economy and on the

Natural Hazards (2005) 35: 291–313 � Springer 2005DOI 10.1007/s11069-004-6485-8

already limited central budget in the case of catastrophic seismic events. Thetwo recent major earthquakes in Turkey on 17 August and 12 November1999 both occurred in or near urban settlements and caused widespreaddestruction of the building stock. This activated the idea of a nationwideobligatory earthquake insurance enforcement, which was first put forward in1978 by the Ministry of Reconstruction and Settlement. At that time theEarthquake Engineering Research Institute, Middle East Technical Univer-sity (EERI/METU), through a project sponsored by the same Ministry,investigated the feasibility of an obligatory insurance model (Gurpinar et al.,1978; Gurpinar and Yucemen, 1980). The obligatory insurance model con-sidered by EERI/METU aimed to serve for a number of purposes, whichmay be listed as: control and therefore improvement of construction practicesin seismic regions, decrease of time loss due to the interruption of services,accumulation of funds for reconstruction and resettlement. The fact that thefirst purpose was included is due to the peculiar control mechanism inTurkey, which may to some extent be generalized to other developingcountries. As control of construction is implemented unsatisfactorily, it isexpected that the earthquake insurance may constitute a complementarycontrol mechanism, based on the conflict of economic interests on the part ofthe insurer and the insured.

After the two major earthquakes in 1999, the Government of Turkey hasdecided to enforce the earthquake insurance on the nationwide basis with thesole purpose of privatizing the potential risk by offering insurance via theTurkish Catastrophic Insurance Pool (TCIP) and then exporting the majorpart of this risk to the international reinsurance and capital markets(Bommer et al., 2002). Although the main aim was to reduce government’sfiscal exposure, it was also intended to encourage risk mitigation and saferconstruction practices. To achieve these purposes all registered residentialdwellings, the total number being about 13 million, are required to be in thecompulsory earthquake insurance coverage. Initially funded by the WorldBank, the TCIP program became effective as of March 2001. In spite of itscompulsory nature, TCIP policy count was about two million as of Sep-tember 2004, corresponding only to 15.3% of the total property owners. Thehighest insured percentage was observed in the Marmara region (24.80%),where the two major earthquakes in 1999 caused the highest damage(www.dask.gov.tr). Milli-Re (Turkish Reinsurance Company) is responsiblefor running the pool, which is protected by the World Bank and a whole hostof reinsurers. TCIP has a potential to become one of the largest earthquakeinsurance companies in the world, provided that the penetration rate is in-creased significantly.

In the implementation of the obligatory insurance program, no reliableinvestigations were conducted for the assessment of earthquake insurancerates. As a matter of fact the rates are set by the Prime Ministry of the

M. S. YUCEMEN292

Turkish Republic, the Under-secretariat of Treasury and charged by theinsurance companies under the control and coordination of TCIP. Thus, thisstudy aims at presenting a simple model for the assessment of earthquakeinsurance rates by considering the damage statistics and the potential seismichazard. The proposed model is used for the assessment of the earthquakeinsurance premiums for reinforced concrete buildings constructed in differentseismic zones of Turkey. The earthquake insurance premiums are computedby considering only the risk of damage due to earthquake shock and lossesresulting from fire due to earthquake are omitted.

In most of the countries, earthquake insurance programs are implementedwith the purpose of accumulating funds to cover the post-disaster expendi-tures. The practice and regulations of earthquake insurance vary fromcountry to country. Below, a brief description for some countries, where theseismic hazard is significant is presented.

In Japan, earthquake insurance has been available since the 1950’s(Kamei, 1976). Earthquake insurance covers damage from volcano eruptionsand tsunamis also. The buildings are classified according to their earthquakeresistance capacities and in this way constructions with poor quality anddesign are discouraged. The earthquake insurance for homes is an endorse-ment to fire coverage, and reinsurance is provided by the government(Brillinger, 1993, p. 16; Scawthorn et al., 2003, chapter 32, pp. 24–25).

In California, a state-run earthquake insurance company, called theCalifornia Earthquake Authority (CEA), was formed in 1994 for providingcoverage for homeowners. The all-residential CEA program was formed inorder to overcome some of the difficulties encountered by the insurancecompanies after the 1994 Northridge earthquake (Scawthorn et al., 2003,chapter 32, pp. 23–24).

In New Zealand, the earthquake insurance is a part of fire insurance. TheEarthquake Commission (EQC), which was formed in 1993, handles theearthquake insurance nationwide. For homes there is a limit of 100,000 NZdollars of coverage at a cost of 0.5 per 1,000 (Scawthorn et al., 2003, chapter32, p. 26). The New Zealand reinsurance program is one of the largestcatastrophe coverages in the world (Steven, 1992).

Studies on the assessment of earthquake insurance premiums based onstatistical methods and utilizing earthquake engineering concepts are limitedin number. Gurpinar et al. (1978) and Gurpinar and Yucemen (1980) haveconsidered the problems related to obligatory earthquake insurance imple-mentation in Turkey and a pilot application was carried out for Denizli.Yucemen and Bulak (1997) have developed a statistical model for theassessment of earthquake insurance rates for the different seismic zones inTurkey. Smolka and Berz (1989, 1991) have developed a methodology forobtaining insurance premiums consistent with seismic risk and estimation ofpotential losses due to large earthquakes. Some of the other major studies
