Report No. 26844-IN India Financing Rapid Onset Natural Disaster Losses in India: A Risk Management Approach August 2003 Document of the World Bank UNITS AND ABBREVIATIONS CurrencyUnit =Rupee (R) Rs 1.OO =US$.0204(astaken inthe study) US$l.OO =Rs 49.0 (astaken in the study) 1Crore= 10,000,000 1Lakh = 100,000 Abbreviations andAcronyms AAL Average annual loss ADB Asian DevelopmentBank AEP Aggregate exceedanceprobability AHDR Annualizedhazardsdamageratio ART Alternativerisk transfer CDF Cumulativedistributionfunction CDMP ComprehensiveDisasterManagement Program CEA California EarthquakeAuthority CRF CalamityRelief Fund DANA Damageassessmentandneeds analysis ELT Event losstable EP Exceedanceprobability FEMA US Federal EmergencyManagement Agency FHCF Florida Hurricane CatastropheFund FRPCJUA Florida ResidentialPropertyand CasualtyJoint UnderwritingAssociation GDP Grossdomesticproduct GEERP Gujarat EarthquakeEmergencyReconstructionProgram Go1 Governmentof India GSDMA Gujarat StateDisasterManagementAgency GSDP Gross statedomesticproduct HPC High Powered Committee IDA InternationalDevelopment Association IDB Inter-AmericanDevelopmentBank IDRC CanadianInternationalDevelopment ResearchCenter IRDA InsuranceRegulatoryand Development Authority IS0 Insurance ServiceOffice LOB Line of business MDR Mean damage ratio MMI ModifiedMercalliintensity MOA Ministry of Agriculture MOF Ministry of Finance NBER US National Bureau of EconomicResearch NCCF National CalamityContingencyFund NCCM National Center for CalamityManagement NCDC National ClimaticData Center NCDM National Center forDisasterManagement NDMD Natural DisasterManagement Division NFCR National Fund for CalamityRelief NGO Non-governmental organization OEP Occurrenceexceedanceprobability OPD Operationsand PolicyDepartment,World Bank PGA Peak ground acceleration PMF Probablemaximumflood PML Probablemaximumloss PMP Probablemaximumprecipitation R M S I Indian subsidiary of aglobalgeographic informationservices company(RMS) TCIP The Turkish CatastropheInsurancePool UNCTAD UnitedNationsConferenceon Trade andDevelopment UNDP UnitedNationsDevelopmentProgram US% USdollar equivalents USAID USAgency for InternationalDevelopment Vice President: MiekoNishimizu Country Director: Michael Carter Sector Director: Joseph Pemia Team Leader: Rodney Lester i FOREWORD Over the last 35 years o f the 20thCentury India suffered more then 150,000 fatalities as the result o f rapid onset natural disasters. Formally reported direct property and infrastructure losses from natural disasters over the same period amounted to US$30 billion, but actual losses will have been substantially higher. An increasing frequency and severity o f natural disasters poses a growing challenge to economic and social development and the country's fiscal balance. Thie resultant need to formally address the impact o f nature1hazards i s reinforced by the fact that the poor are almost invariably most affected by the occurence o f calamities. Inthe last decade the situation has been exacerbated by the fact that most Indianstates and the central government have been running deficits and resources for post disaster reconstruction in particular have become increasingly constrained. Giventhis context and a clearly expressed concern on the part o f the Indianauthorities as fiscal pressures grow, the World Bank undertook a detailed review o f India's catastrophe exposures, with indepth studies in four states. The purpose of the study was to examine the loss potentials from rapid onset natural disasters and to consider the opportunity to apply enhanced country and state level risk management techniques, with a particular emphasis on the financing o f post disaster reconstruction and the efficient allocation o fpublic funds. The report i s a product o f two years o f research to understand the natural catastrophe risks that India faces andthe way they are currently managed and financed. It is the first time that there has been an attempt to develop a comprehensive catastrophe risk management framework that brings together risk financing and mitigation techniques and contains an in-depth discussion o f the role o f institutional incentives innational disaster management. Signed: Marilou Uy, Director, Financial Sector, Operations and Policy Department .. 11 Acknowledgements This study was funded by the South Asia Region with support from the Financial Sector Vice Presidency o f the World Bank. Support for the study was also provided under the umbrella o f the Provention Consortium from the United Kingdom's Department for International Development (DFID). The team thanks DFIDfor their generous support. The authors are Rodney Lester and Eugene Gurenko o f the OPD Department o f the Financial Sector Vice Presidency. Vijay Kalavakonda, also o f OPD, provided operational support and contributed to the Appendixes and Chapter V. Other key members o f the original mission team in February 2002 were Alcira Kreimer and Margaret Arnold from the Disaster Management Facility and Priya Basu and Santosh Kumar from the Indian Country Office. Torben Andersen provided background data for Chapter Iand Mina Hamedani contributed her experience at Fannie Mae to Appendix V. Manisha Shah and Rick Zechter provided logistical support during the production o f this document. The World Bank team working on the Bhuj (Gujarat) earthquake reconstruction effort has also provided valuable input. The authors would also like to acknowledge the professional contribution by RMSI, an Indianrisk modeling consultancy, which was commissioned to carry out in-depth studies o f catastrophe loss potentials in four selected states. Inaddition, we would like to thank all those Indianprofessionals who providedinputs on the Indian disaster response system and on the development o f the Indian general (non-life) insurance sector. Thanks is also due to all those who contributed so willingly to this project through interviews and feedback, including representatives o f the Ministry of Finance, the members o f the High Powered Committee on Disaster Management, IRDA, various state disaster management bodies, federal departments handling key infrastructure, and representatives o f the insurance, reinsurance, risk management and finance sectors, and o f other development organizations. We should also thank the World Bank country office staff and consultants who proved to be so helpful during our two visits, and the three external (Paul Freeman, Andrew Dlugolecki, Professor Arya) and two internal (John Pollner, Christoph Pusch) reviewers who helped to shape this paper. Chris Hoban's positive skepticism was key to developing a more nuancedandrealistic view o f the nature o f the natural disasters fimdinggap. ... 111 TABLE CONTENTS OF ABBREVIATIONSAND ACRONYMS .................................................................................................................... i ACKNOWLEDGEMENTS ................................................................................................................................... EXECUTIVESUMMARY.................................................................................................................................... 111 1 INTRODUCTION ................................................................................................................................................ 5 I. NATURAL DISASTERS A THREATTOINDIA'SFISCALANDECONOMICDEVELOPMENT POSE AND JUSTIFY ...................................................................................... I1. A FORMAL RISKMANAGEMENT APPROACH 7 DISASTER MANAGEMENT 111. APPROACH............................................................................................................................ LossFUNDINGSTRATEGIESNEED BEINSTITUTED To AsPARTOFA FORMAL RISK 15 CATASTROPHE RISKMODELS POINT TO DIFFERENTSTATE LEVEL LOSS POTENTIALSAND R I S K MANAGEMENT IMPERATIVES ........................................................................................................................ 22 IV INDIA'S IsUNDEVELOPED INSURANCE MARKET PENETRATION AND~INTAIN ADEQUATE V. CAPACITY .................................................................................. ANDINTERVENTION I sREQUIRED To INCREA 30 FINDINGS, POLICY OPTIONS ANDRECOMMENDATIONS .................................................................... 40 ANNEXI:INDIA'S ......................................................................................................... ANNEX11: Loss SUMMARY TABLES............................................................................................................ DISASTER HISTORY 50 55 ANNEX111: INSURANCE CONSUMPTION BYSTATE....................................................................................... 57 ANNEX IV: CENTRALRELIEFFUNDSFLOWS STATES TO ............................................................................. 58 ANNEX V: ANDHRAPRADESHPOSTDISASTER EXPERIENCE ....................................................................... 60 ANNEXVI: BHUJEARTHQUAKE CAPABILITYEUNDING GAP....................................................................... 62 ANNEX VII: BRIEFOVERVIEW OFINDIANMORTGAGE MARKET ................................................................. 63 ANNEXVIII: U S CONSUMERS UNION PERSPECTIVEONNATURAL DISASTER INSURANCE .......................... 64 APPENDIXI:INTERNATIONAL .................................................... EXPERIENCE WITH CATASTROPHE FUNDS 66 APPENDIX11: INSURANCE MARKET ECONOMICS .......................................................................................... 74 APPENDIX 111: HAZARD AND VULNERABILITYMODELS CHAPTERIX - CALAMITY RELIEF................................ .............................................................................. 78 APPENDIXIV: ELEVENTH FINANCECOMMISSION: 90 APPENDIX v: LOAN .................. 94 BIBLIOGRAPHY............................................................................................................................................ HEDGING CATASTROPHERISKOFRESIDENTIALMORTGAGE PORTFOLIOS 101 GLOSSARY .................................................................................................................................................. 104 iv LISTOFTABLES TABLE1: REPORTED NATURAL CATASTROPHE IMPACTS ................................... INSOUTH ASIA. 1996-2000 5 TABLE2: DROUGHT INCIDENCE IN INDIA ....................................................................................................... 6 TABLE3: DISASTER HISTORYMAJOR BY HAZARDINDIA, 1996-2001....................................................... IN 8 TABLE5: NCCF DISBURSEMENTSALL STATES(RSMILLION)..................................................................... TABLE4: REPORTED NATURAL CATASTROPHE LOSSES, 1996-2001.............................................................. 8 11 TABLE6: FUNDINGGAPRECONSTRUCTIONAND REHABILITATION.............................................................. 17 TABLE7: MODELING SCOPE OF STATEAND PERILCOMBINATION............................................................... 23 TABLE8: AVERAGE ANNUAL LOSSSUMMARY .............................................................................................. 25 TABLE9: PR0BABLEMAXIMUM LOSS SUMMARY ( u s $ MILLION) .............................................................. 26 TABLE10: PURERISKPREMIUMSUMMARY -(I) ......................................................................................... 27 TABLE11: PURERISKPREMIUMSUMMARY -(11) ........................................................................................ 27 TABLE12: RETURNPERIOD LOSSESCOMBINED ASSETS FOR (US$ MILLION)-(I) ..................................... 28 TABLE13: RETURNPERIODLOSSESCOMBINEDASSETS FOR (us$MILLION)-........................................... (11) .................................... 28 TABLE15: CAPACITY UTILIZATION, (Rs CRORE)............................................................................... TABLE14: RANKINGS OFFOUR STATESINTERMSOFTHEIR RISKEXPOSURES 29 2002 33 TABLE16: CATASTROPHERISKEXPOSURESAS PERCENTAGEOFKEYECONOMIC FLOWMEASURESINFOUR SELECTEDSTATES ......................................................................................................................................... 40 TABLE17: LISTOFLARGEST CYCLONES INDIA.................................................................. IN (1891-2000) 50 TABLE18: LARGE EARTHQUAKES ININDIA ................................................................................................. 52 TABLE19: SUMMARY OFMAJOR FLOOD LOSSES ...................................................... 53 TABLE20: VALIDATIONOFM M I S WITH2001BHUJEARTHQUAKE............................................................. ININDIA(1953-2001) 53 TABLE21: VALIDATIONOFM M I S WITH 1993 LATUR EARTHQUAKE .......................................................... 54 TABLE22: VALIDATION OFWIND SPEEDSWITH 1977ANDHRAPRADESHCYCLONE.................................... 54 TABLE23: VALIDATION OFWIND SPEEDSWITH 1999ORISSACYCLONE...................................................... 54 TABLE24: AVERAGE (us$MILLION)-(I).......................................................... 55 TABLE25: AVERAGE"LiAL LOSSSUMMARY (us$ MILLION)-(11) ....................................................... ANNUAL LOSS SUMMARY A 55 TABLE26: PROBABLEMAXIMUM (us$ MILLION)............................................................ 56 TABLE27: INSURANCE PENETRATION%..................................................................................................... LOSS SUMMARY 57 TABLE28: ANNUAL MARGIN COMMISSIONS(Rs MILLION)........................................................................................................................ MONEY/Cw ALLOCATED THE STATESBY SUCCESSWEFINANCE TO TABLE29: RELEASESFROMNATIONAL FORCALAMITY RELIEF(RSMILLION).................................. 58 FUND 59 TABLE30: ANDHRA PRADESHPOSTDISASTER EXPERIENCE: 1996 CYCLONE ............................................. 60 TABLE31: ANDHRAPRADESHPOSTDISASTER EXPERIENCE :2001FLOODING TABLE32: SOURCESAND USESOFFUNDS (us$ MILLION) ............................................................................ 62 .......................................... 61 TABLEA 1.1: GOVERNMENT SPONSORED CATASTROPHE INSURANCE 67 PROGRAMS ...................................... TABLEA 1.2: CATASTROPHE PROGRAM DESIGN VARIABLES ...................................................................... 69 TABLEA 1.3: INSURANCE VEHICLES........................................................................................................... 71 TABLEA 1.4: RATES CHARGEDAND MITIGATION ................................................................... 72 TABLEA 2.1: INSURANCE......................................................................... INCENTIVES MARKET DEVELOPMENT PATHS 77 TABLEA 3.1: EXPOSURE VALUESUMMARY (US$ MILLION)...................................................................... 85 TABLEA 5.2: PRE-EVENT .AND TABLEA 5.1: ASSUMPTIONS CALCULATIONS FORABc BANK............................................................ TABLEA 3.2: HOUSING REPLACEMENT (Rs)...................................................................................... 87 COST 96 VS POST-EVENT LTV......................................................................................... 97 V LISTOFCHARTS CHART 1: REPORTEDCATASTROPHELOSSES ININDIA. 1965-2001 ............................................................... 9 CHART2: BHUJ(GUJARAT) EARTHQUAKE FUNDING/~APABILITY GAP.END2002.................................... 13 CHART 3: REINSURANCEPRICINGVOLATILITY ........................................................................................... 21 CHART4: GENERALINSURANCE CONSUMPTION.LOWINCOMECOUNTRIES .............................................. 30 CHART5: PENETRATIONv s.PERCAPITA INCOME. 1998-1999 ................................................................... 31 CHART6: POVERTYIMPACTON INSURANCEPENETRATION ......................................................................... 31 CHART7: NON-LIFE INSURANCE ........................................................ PENETRATIONININDIA. 1994-2000 32 CHART 8: INSURANCE PENETRATIONAND CREDITDISBURSEMENT. CHARTA 2.1: NON LIFEINSURANCEELASTICITYOFPREMIUMPERCAPITA VS.GDPPERCAPITA.............38 1998- 1999......................................... CHARTA 3.1: EXPOSURE CHARTA 3.2: GENERALBUILDINGVULNERABILITYCURVES..................................................................... VALUESUMMARY (US$) .................................................................................... 86 77 88 LISTOFFIGURES FIGURE1: COUNTRY/STATE RISKMANAGEMENT ........................................................................... MODEL 16 FIGURE2: THEPROBABILISTICRISKMODEL ............................................................................................... 24 FIGUREA 1.1: FRENCHNATCATSYSTEM.................................................................................................... 70 FIGUREA 1.2: JAPANESEEARTHQUAKE REINSURANCE PROGRAM .............................................................. 71 FIGUREA 2.1: THELIMITSOFINSURANCEMARKETS.................................................................................. 74 FIGUREA 3.1: MODELED SOURCES WITH MAXIMUMMAGNITUDES ............................................................ 79 FIGUREA 5.1: RISKCURVE......................................................................................................................... 99 vi Executive Summary Background Natural catastrophes pose a serious and growing threat to India's development. Twenty-two o f India's 31 states are regarded as particularly prone to natural disasters: 55% o f its land i s vulnerable to earthquake, 8% i s vulnerable to cyclone and 5% i s vulnerable to flood. MunichRe. has ranked India's four megacities as amongst the 50 most vulnerable mega cities in the world. On average, direct natural disasters losses amount to up to 2% o f India's GDP and up to 12% of central government revenues. * Despite being centred in a relatively underdeveloped area, the Gujarat earthquake i s estimated to have caused a US$491-655 million loss o f output, a US$2.2 billion negative impact over three years on the state's fiscal deficit2 and led to a national 2% tax surcharge. Total losses reported due to natural catastrophes have been growing. Reported direct losses from natural catastrophes more than quadrupled during the 15-year period 1981-1995 ($13.4 billion) compared to the losses registered during the previous 15 years ($2.9 billion). This alarming trend continues; the total losses o f US$13.8 billion reported in the most recent six-year period (1996-2001) have already exceeded total losses incurredover the previous 15-year period. Responsibility for disaster funding in the aftermath o f a natural catastrophe has been shared by the state and central govemments. While the affected state manages the relief work and reconstruction efforts, the central govemment provides financial support. Originally, the central government financed catastrophe relief efforts through margin money allocated to the states through the successive Finance Committees. However, the general experience under this system was that actual calamity expenditures consistently outpaced underlying budget expectations. Under the Ninth Finance Commission, the government revised the system and created a Calamity ReliefFund(CRF) from which states can draw upon under emergencies. The Eleventh Finance Commission limited the use o f CRF funds to items which provide immediate relief to the affected population. This Commission also proposed an enhanced role for the insurance markets. One limitation o f the current formal disaster relief funding mechanisms involves the funding of the restoration o f infrastructure. While states are required to maintain and restore infrastructure from planned capital budgets, these budgets have become increasingly constrained with a growing share o f state budgets going to recurrent expenditures and debt service on burgeoning public sector deficits. Faced with dwindling capital budgets some states have resorted to diverted development loans to find infrastructure repairs. These often involve intense renegotiation and ongoing rigorous procurement rules, although there have been efforts to expedite the process in India. As a result, in the absence o f adequate and timely funding for capital repairs the expected future lives o f some o f the infrastructure assets are reduced post disaster, while future capital projects necessary to support a growing economy are not undertaken. ' Direct losses are stock losses (mainly infrastructureand housing). Indirect losses are flow items such as state revenues and GDP. Fiscal effects are sometimes called secondary losses. World Bank/ ADB Assessment Report 1 The growing problem o f funding natural catastrophe losses has been recognized by the Finance Commissions; every Finance Commission since the Second has devoted a full chapter to calamity relief hnding. Despiite these efforts, India continues to suffer from underdeveloped state level risk management capacity and underutilizationo fprivate insurancemechanisms. The Study Inlight of India's vulnerability to growing losses due to natural disasters and escalating fiscal pressures at the central and state levels, the World Bank undertook a detailed review o f India's catastrophe exposures. The goal o f this project was to examine loss potentials from rapid onset natural disasters and to consider the opportunity to apply enhanced country and state level risk management techniques, with a particular emphasis on the financing o f post disaster reconstruction and the efficient allocation o f public funds. The role o f insurance markets has also been examined given their major contribution to the effective transfer o f private sector catastrophe risk inother countries, but relatively insignificant role inIndia to date. The country risk management approach developed by the World Bank is based partly on corporate risk management principles, but accounts for key economic and social metrics such as government fiscal profiles and the living conditions o f the poor. The first step under this methodology i s to assess the potential losses from natural hazards on a probabilistic basis, and detailed studies were carried out in four states. The next step involves a formal and structured approach to understanding the funding o f natural calamity losses and identifying the "natural disasters funding gap," which i s the difference between the expected fiscal cost o f an event and available ex post sources o f government revenue. The World Bank team recognizes that enhancing implementation capacity and reducing asset vulnerability by employing mitigation techniques (such as improving housing construction standards) are also integral to reducing direct losses from natural catastrophes. The risk management framework ideally includes ex ante capacity building, risk reduction andmethods to transfer or finance residual risk. Inparticular, inthe course o fthe studyit became clear that even iffunds are accessiblepost disaster, they may not be availed because of a lack ofcapacity and capability. The fundinggap concept has beenmodified appropriately. The main body o f the report i s divided into five chapters. The first chapter explores the fiscal impact o f historical natural disaster losses and the hnding methods used by the satte and central governments to date. It demonstrates that the fiscal and economic pressures caused by these calamities are significant andjustify a formal risk management approach. Chapter I1introduces the formal risk management framework used in the report: the CountryBtate Risk Management Model. It also discusses various ex ante risk management and ex post coping strategies the various governments could adopt. Chapter I11develops the risk management framework with an in-depthanalysis of the natural catastrophe loss potentials infour states. Chapter IV is a detailed review o f India's insurance market and examines various demand and supply drivers. It concludes that given current conditions, government intervention would be requiredin order to develop an effective natural catastrophe insurance market inIndia. The final chapter presents the team's findings, highlight policy options and make recommendations based on our findings. The five appendices highlight topics and present information related to the main document that will supplementthe reader's knowledge. 2 Conclusionsand Recommendations The study concludes that India still adopts a primarily reactive, or coping, approach to dealing with natural disasters. Though considerable progress has been made via mitigation anddisaster preparednessto reduce both financial and human losses, at the center and in some states, India's current approach to funding natural catastrophe losses remains fragmented. It lacks a comprehensive catastrophe risk management framework to quantify, analyze and manage potential losses. The current program, particularly at the national level, lacks institutional incentives and underplays the role o f risk financing through ex ante mechanisms such as catastrophe reinsurance and contingent credit facilities. The development o f ex ante funding programs i s particularly critical because these programs typically serve as a primary source o f immediate liquidity that would reduce human suffering, economic loss, and fiscal pressures in the aftermath of a natural disaster, and kick-start economic recovery. Ex ante funding approaches can also foster mitigation andprovide incentives for institutional capacity building. Based on the study, the World Bank team has identified several policy options and recommendations for the Government o f India. Policy Options: Mitigation and risk financing are the two pillars o f effective catastrophe risk management at the country and state level, and GoI's mitigation efforts could be augmented by a formal approach to risk financing. A risk financing strategy would consist o f three parts: formal risk assessmentsat the state and the central levels; identification o f fimding gaps; and finally, development o f state and national risk management plans aimed at closing the identified fundinggaps over time. Create fiscal incentives for states to pursue active risk management strategies, including buildinginstitutional capacity at the statelevel. The existing institutional framework for catastrophe risk management could be further developed in two ways. First, a Risk Financing Facility could be created to provide additional financial assistance to those states which are adopting and implementing an agreed risk management approach. Second, the use of contingent credit facilities could be explored for catastrophe risk financing and in support o f risk management incentives at the state level. Sucha a contingent credit facility would become available to meet claims o f the states inthe aftermath o fnatural disasters, provided an acceptable state risk managementprogram i s inplace. Introduce incentives and perhaps mandated requirements to increase the utilization of catastrophe insurance mechanisms by the private sector, including better off households. This could be done by requiringthat replacement cost catastrophe insurance i s purchased when mortgage financing is granted; tying catastrophe insurance to land tax or land registration systems; or making it clear, if necessary through regulation, that households inthe upper and middle income brackets are not eligible for government reconstruction funding.3 The World Bank team does not suggest that Go1should stop fmancing housing reconstruction inthe aftermath o f a disaster. We recommend that the government should adopt a clear policy o f not helping those who can afford to help themselves bybuying insurance (or through self-insurance). According to various surveys this group accounts for between7% and 10% o f the population. The poorer segments o fpopulation with substandard housing should continue to be entitledto government post disaster assistance. Recommendations While the options outlined above will require consideration within the larger Indian fiscal and sectoral policy framework, the scope for further reform in the insurance sector to add capacity and increase the penetration o f disaster insurance is relatively clear. For this reason we have characterized the relevant policy steps as recommendations. 0 The insurance sector should be further liberalized by removing current restrictions on, andcross subsidies from, the householdand smallbusiness insurance markets. 0 Claims handlingprocedures in the event of natural disasters should b e streamlined and formalized. 0 More explicit rules should be introduced regarding insurers' minimum premium retentions and maximum risk retentions, and exposure accumulation data should b e gathered and reported to IRDA. Progressand Challenges The paper is a product o f a two year study to understand the natural catastrophe risks that India faces. It is the first time that there has been an attempt to develop a comprehensive catastrophe risk management framework for India. Similarly the models developed for the four states are pioneering efforts. The main challenge that had to b e faced in developing models to precisely assess the risk for India was the availability of accurate data. The models could have been built to a greater degree of detail to assess vulnerability more accurately, for example, by accounting for differences in house layouts and number o f stories in houses, but relevant and reliable data was not available. Nonetheless, the team believes that both the models and the framework provide a firm basis for understanding India's exposure to natural catastrophes, the resultant hding gaps it faces, and for developing appropriate incentives to encourage active risk management. 4 Introduction The Commonwealth Disasters Index, despite being developed to support a case for better off small states to access development funds, includes India in the 5 countries most vulnerable to natural disasters. Perhaps the most telling measure o f India's exposure and vulnerability i s the human death toll (defined as killed and missingpeople). Within the last five years of the 20th century alone, various natural catastrophes claimed more than 45,000 victims across South Asia with the majority o fthese fatalities occurringinIndia (Table 1). Table 1: ReportedNatural CatastropheImpactsin SouthAsia, 1996-2000. Sources: Swiss Re, Natural catastrophes and man-made disasters 1996-2000; CRED, International disaster database, UniversitC Catholique de Louvain, Belgium; World Factbook. India i s also estimated to have suffered direct losses in excess o f $9 billion over the five years from 1996 to 2000, reflecting loss estimates on approximately 20% o f the reported catastrophe events during the period.4 These have disproportionately affected the poor5, although this i s largely unrecorded in the monetary loss data. In addition to killing people and destroyin property and infrastructure, natural disasters can have lasting economic and social effectsl , F including areallocationo fincome both geographically andbetweensocial groups. State level taxation and private insurance mechanisms (see below) are relatively underdeveloped in India and inpractice the major responsibility for ex post funding of relief and recovery has rested, directly and indirectly, with budget transfers from the central government. EveryFinance Commission since the Second has devoted a full chapter to calamity relief funding. In the last decade the situation has been exacerbatedby the fact that most states and the central government have been running deficits on their revenue accounts because o f burgeoning current Direct losses refers to losses o f economic capital or stock, but inpractice published insurance losses include any insured loss o fprofits. Indirect losses refers to flow items such as GDP. See Litan (1999) for a full discussion o f natural disaster loss metrics. 'See for example Bhatt (1999). Anderson (1995) has pointed out that indirect economic losses tend to be larger relative to direct material costs in poor countries than inrich countries. Litan (1999) points to evidence that indirect losses constitute a larger fraction oftotal losses for large disasters. 5 expenditures, and resources for post disaster reconstruction inparticular have been increasingly constrained and dependent on donor funding.7 Given this background and a clearly expressed concern on the part of the Indian authorities as fiscal pressures grow, the World Bank undertook a detailed review o f India's catastrophe exposures, with in depth studies in four states. The purpose o f the mission was to examine the loss potentials from rapid onset natural disasters and to consider the opportunity to apply enhancedcountry and state level risk management techniques, with a particular emphasis on the financing o fpost disaster reconstruction andthe efficient allocation o fpublic funds. The role o f insurance markets in has been examined given their contribution to the effective transfer o fprivate sector catastropherisk inother jurisdictions, but relatively insignificant role in India. In 1999, which is one o f the worst years on record for natural hazard related insurance losses, South Asian countries did not rate among the top 20 in terms of insurance losses; however, they did account for five o f the 20 worst events interms o f lives lost. Despite having close to a fifth o f the world population, South Asia only accounts for about 0.3% o f global non- life insurance premiums. The region, to all intents and purposes, has not been a serious participant inthe global markets for disaster loss risk transfer. This exercise focuses on rapid onset disasters. It is true that droughts affect more people than other natural disasters (Table 2) andtheir cumulative indirect economic effects can be substantial over time. However, direct losses tend to be substantially smaller for droughts than for rapid onset disasters. Because slow onset disasters such as drought have different characteristics from and are more difficult to quantify than rapid onset events, they would require a separate study usinga different risk management paradigm thanthe one applied inthis study. This dichotomy o f natural hazard risk was discussed by the Seventh Finance Commission but i s not currently recognized in India's expenditure planning and revenue sharing processes. Given weather insurance and other rapidly developing technologies in this area, some investment in investigating and perhaps even pilot testing ex ante funding o f slow onset disasters may now be justified. Number of Disaster Events Period People Affected -millions Total Drought ~~ Total Drought 121 8 1965 - 1980 662 500 181 5 1980 1995 - 849 502 75 4 1996 - 2001 283 90 1Source: CRED, International disaster database, UniversitC Catholique de Louvain, Belgium. I I I I I 'McCarten (2003). 6 I. NaturalDisastersPoseaThreatToIndia'sFiscalandEconomic Development andJustify a FormalRiskManagementApproach Disaster Exposure/ History - India has a significant exposure to natural hazards; 55% o f India's land i s vulnerable to earthquake, 8% i s vulnerable to cyclone and 5% i s vulnerable to flood.' andpotential impact o fnatural disaster^.^ Additionally, there is growing evidence that calamities Demographic and economic trends in the past three decades have magnified the actual can contribute to environmental degradation leading to a vicious cycle o f increasing disaster impacts." India has a long coastline, which i s exposed to tropical cyclones, especially along its eastern coastline. Around 85 cyclones from the Bay o f Bengal and Arabian Sea have affected the country over the past 35 years: inNovember 1996 over 7 millionpeople were displaced when a major cyclone hit Andhra Pradesh. These cyclones are frequently accompanied by tidal waves. Low-lyinglands, typical o f the Eastern shore o f India, permit storm surges o f even a few meters to intrude far into the hinterland, causing widespread flooding and seawater incursion. Flooding is a common phenomenon in India and i s exacerbated due to the silting up o f rivers, reduced soil absorption, lack o f urban planning, and deforestation. Floods are caused due to heavy rainfall during the three to four month long monsoon season. Large floods occurred in 1997 and 1998. Heavy monsoon rains flooded South West India in 1997 and affected Assam, Bihar, and Andhra Pradeshin 1998. Recent flooding events have been aggravated by increased urbanization and unplanned growth. For example, in Mumbai, where migration has increased the population significantly, large segments o f the population live inunauthorized slums close to drainage systems. Because o f these settlements, the width o f the "nallas" (man-made canals for sewage water and refuse) are reduced and the accumulation o f solid waste causes inner city floods. As noted above, about halfof India is exposed to earthquakes. The vulnerable areas are mostly inHimalayan and sub-Himalayan regions, and inAndaman andNicobar Islands (Vinod, 1999). The most recent earthquake occurred on January 26, 2001, and mainly affected the state of Gujarat. The Gujarat earthquake, which measured 6.9 on the Richter scale", i s considered one o fworst single disasters of the decade, causing severe destruction to buildings and other property inBhuj, in the Kutch district, and several urban cities including Ahmedabad. This earthquake affected 182 talukas covering 7,904 villages in 16 districts o f Gujarat: 13,800 people were reported killed and more than 167,000 injured. Nearly one million residences were destroyed completely or partially. More than 360 natural disasters have beenrecorded over the past 35 years and the frequency has been increasing (see Annex I). number o f reported events increased by around 50% during The the 15-year period 1981-1995 (181 events or 15 per year) compared to the previous period 1965- Dheri, in Sahni et al. (2001) For a disaster to occur human lives and property need to be exposed and the frequency of disasters should not be confused with the frequency o f natural events. lo Seejoint UNEP/OCHA EnvironmentalUnitEnvironmental Emergency Notification (ENRA) for a taxonomy. India Meteorological Department figure. Other sources gave higher values. 7 1980 (121 events or 8 per year). This trend has continued inrecent years with 75 events reported inIndiaduringtheperiod 1996-2001(Table 3). Table 3: Disaster History by Major Hazardin India, 1996-2001 1 Hazard 1 No. of No. of People Reported No. of loss Percent Average reported reported affected losses reports reported loss per events 1 deaths (thousands) ($million) 1submitted I I report 1 Source: CRED, International disaster database, Universitt Catholique de Louvain, Belgium. Economic and Fiscal Impacts - Reported direct losses on public and private economic infrastructure in India have amounted to approximately $30 billion over the past 35 years (nominal values at then applying exchange rates). Since less than 25% of the registered loss events actually provide any loss estimates, the official numbers substantially understate the true economic impact o f direct losses. A crude grossing up for reporting frequency indicates that direct natural disasters losses equate to up to 2% o f India's GDP and up to 12% of federal government revenues (Table 4). Table 4: Reported NaturalCatastrophe Losses, 1996-2000 South Asia Reportec Percentage ~ sities - country incidents assessed losses [$ mill.] revenue: ~ pct. GDP nt. revenues India 73 19.2% $9,176 $407,850 $75,500 2.25% 12.15% Pakista 22 0.0% $52,280 $9,150 Afghanista 20 0.0% $3,895 bangl lades 48 8.3% $2,879 $37,650 $4,360 7.65% 66.03% ,sri 9 0.0% $11,625 $2,185 'Bhuta 0 0.0% $430 $165 NeDal 15 26.7% $52 $6,250 $690 0.84% 7.58% 187 7.7% $12,107 $519,980 $92,050 3.58% 13.15% Estimates based on factor income data, current foreign exchange rates, and extrapolation of comparative country figures. Estimates basedon comparative data on central government and state government operations. Sources: CRED, Intemational disaster database, Universite Catholique de Louvain, Belgium; International Monetary Fund, Recent Economic Developments - Country Report Series; World Factbook. Furthermore, the reported monetary losses seem to be increasing (Chart 1). Reported direct losses from natural catastrophes more than quadrupled during the 15-year period 1981-1995 ($13.4 billion) compared to the losses registered duringthe previous 15 years ($2.9 billion). This 8 alarming trend i s continuing with total losses o f $13.8 billion reported during the period from 1996-2001 (Table 3). Hence, the losses reported during the most recent six-year period have, in nominal dollars, already exceededtotal losses incurred over the previous 15-year period. The economic impact o f natural disasters extends beyond the directly measurable losses on economic infrastructure. There are often significant secondary effects and indirect losses associated with natural disasters. For example, the destruction o f productive assets and public infrastructure inhibits economic activity, while the increased demand for public expendituresfor relief and recovery disrupts fiscal planningand prejudices public and private capital investment. Numerous studies carried out in the last decade confirm the negative short term economic and social impacts o f natural disasters.12 A World Bank/ ADB assessment report estimated that the Bhujearthquakecaused a $491-655 millionloss ofoutput andhada $2.2 billionnegative impact over three years on Gujarat's fiscal position. The medium to longer term impact o f natural disasters has been examined in a number o f studies and the results are ambiguous, although it appears clear that both the timely availability o f hnding post disaster and institutional capabilities affect the extent and sustainability o frecovery. Chart 1: ReportedCatastropheLosses in India, 1965-2001 NominalUS$ Million at then applyingexchange rates Source: CRED, International Disaster Database, UniversitC Catholique de Louvain, Belgium. Discussions on natural calamities have been part o f the fiscal scene in India since the Second Financial Commission first focused on the problem. Every subsequent report has dedicated a complete chapter to the topic. More recently, the Tenth Planning Commission devoted a whole chapter to the development implications o f natural disasters, with a particular emphasis on disaster management, prevention and mitigation. Inlate 2001, a HighPowered Committee on Disaster Management submitted a report which recommended that 10% o f Plan Funds13at the national, state and district levels be earmarked and apportioned for prevention, reduction, preparedness andmitigation ofdisasters. The evolution o f disasters fimdinginIndia largely reflects the five year fiscal planning cycle and has been shaped by the federal structure o f the country. The Constitution does not directly specify which level o f government i s responsible for managing disasters. By convention, this 12See for example Benson (1997). l3India follows a plan approach to economic management and plan funds are those relating to items appearing inthe Plan. Non-plan funds largely cover ongoing expenditures. 9 responsibility has been taken up by individual states while the federal government provides financial support. The history o f formal post disaster funding inIndia can be captured along four vectors: 1. What is funded. This has generally been divided into three distinct categories, namely gratuitous relief (including emergency water, food and shelter, drainage works and seed), relief work on plan projects as a better alternative to gratuitous relief, and repairs and reconstruction o f government assets. 2. The funding role of the central government, which also has three categories. Normal transfers to the states under the tax sharing arrangements (this allowed for a margin for calamity relief from the Second to the NinthFinance Commissions, when the Calamity Relief Fund (CRF) was established), advances against the current Plan, and supplementary transfers from the federal level inthe event o f catastrophic losses. 3. The nature of federal supplements, which can be straight grant, loan or advance. In practice a large portion o f the loans and advances become de facto grants. The national CalamityReliefFundwas established whenthis was recognized. 4. The nature of the event. The Sixth, Seventh and Eighth Finance Commissions recognized that droughts have different characteristics from other natural calamities and are best respondedto with heavy investment inrelief works. The approaches adopted from time to time have reflected the ongoing tension between the central government's concern, on one hand, about fiscal discipline and efficient use o f funds and, on the other hand, the reality o f the growing intensity o f natural disasters and deteriorating state fiscal positions. While there have been many ad hoc adjustments over the last five decades, particularly with regard to the nature o f central supplementary transfers, the funding arrangements can essentially be divided into three main periods: Second to Sixth Finance Commissions - During this period an explicit margin for relief, usually including relief works, was built into state non-Plan budget planning. Excess requirements over the margin were partly or wholly met by the central government through combinations o f grant and lending. Repairs and restoration tended to be handled through Plan supplements, advances and development loans. There was an ongoing debate as to whether relief works came under Plan or non-Plan heads. As costs grew, various controls were built in including the involvement o f central inspection teams and the introduction o f ex post expenditure ceilings. Ceilings were removed in 1972/73 but expenditures grew rapidly andthe SixthFinance Commission called for states to live within their Plan allocations and for disaster funding to become anintegralpart o fthe planningprocess. Seventh and Eighth FinanceCommissions - The dictates o f the Sixth Commission proved to be impossible to sustain inpractice and central advances against repairs andreconstruction crept back in(in 1975/76 over 90% o f advance Plan assistance was for reconstruction and replacement o f roads, buildings, flood control, irrigation works and other public assets). The Seventh Commission reiterated that Plan assistance should only be available for the creation o f new assets and recommended that repairs and reconstruction should become part o f the non-Plan margin allowance, with 75% o f the excess over the margin being met through grants by the central government. However, the commission also recommended that drought associated relief expenditures which require new investmentshould be transferred to the relevant state's Plan. Up to 5% o f additional Plan funds were available for this purpose inany year, as an advance against fkture Plan allocations. Inthe event o f extreme disasters the center would contribute via grants and loans. 10 Ninth to Eleventh Finance Commissions - The Ninth Finance Commission proposed the cancellation o f the marginal fimding approach with heavy interventionfrom the central level and instead introduced the Calamity Relief Fund, with the central government contributing 75% in the form o f non-Plan grants. Any balance in the Fund could be carried forward to future Plan periods, and in the event o f heavy calamity expenditures up to 25% o f the following year's central allocation could be drawn upon. The NinthCommission also began to canvass the idea o f an "Expert Group" to monitor the actions o f the states, but restricted this to relief work. The Tenth Finance Commission introduced the National Calamity Relief Fund(NCRF), managed by a National Calamity Relief Committee (NCRC), to cover calamities o f rare severity. However, the states allegedly then projected "any calamity as one o f rare severity," resultingin an upward trend inrelief requests (Table 5). Central control processes, including visiting teams, had to be reintroduced. Inpractice it was sometimes found that hnds disbursedhad not been employed even after considerable periods. The Eleventh Finance Commission modified the Ninth by restricting capital expenditures from the CRF to items that provide immediate relief to the affected population and are o f short duration. Reconstruction and repair were reallocated to Plan fimds "on priority" and the distinction between drought and other calamities was removed. The EleventhFinance Commissionalso proposed a role for the insurance markets andrecommended the creation o f a National Center for Calamity Management (NCCM) to provide advice to the central government on the ex post financing o f calamity recovery efforts. In many ways it anticipated the recommendations o f this report (see Chapter V) and has been included as Appendix IV. The Twelfth Financial Commission, which has to report byApril 2004, has beenaskedto look, once again, at this topic. Table 5: NCCFdisbursements allstates (RsMillion) Year 1997-98 1998-99 1999-00 2000-01 2001-02 2002-03 Total 2,774 4,971 12,910 9242 13,687 16,000 Keyelements o fthe current approach are as follows: 0 The Calamity Relief Fund meets immediate relief needs for the victims o f cyclone, drought, earthquake, fire, flood and hailstorm. Under this arrangement a Calamity Relief Fund (CRF) is constituted in each state to receive funds, 75% o f which come from the central level inthe form o f non-plan grants. Individual state funding levels are based on relatively short term averages, adjusted for inflation and mitigation efforts.l4 Central government transfers are subject to receipt o f evidence from the states that the funding o f the CRF is being appropriately managed. States may also draw on up to 25% o f central funds due in the following year, subject to subsequent adjustment. The state CRFs are administered by committees consisting o f officials connected with relief work or who ~~ l4 This is consistent with the Arrow -Lind(1970) expected cost formulation for nations, but as Mechler (2002) has demonstrated, the underlying assumptions break down for highlyvulnerable developing countries such as India, as shown by the fact that extreme events have had to be discounted inthe past inamving at CRF allocations. The Arrow - Lindwork built on work by Hirshleifer (1966) and made it clear that insecond best situations public investment could replace private investment. 11 have expertise inthe natural calamity field. A state committee is responsible for ensuring that money drawn from the CRF i s applied only to expenditure items approved by the Minister of Home Affairs. CRF funds are to be invested in prescribed assets to ensure their availability when needed. CRF funds may be applied to existing capital works, but only if this i s required for the provision o f immediate relief, such as restoration o f drinking water and shelter. Other capital restoration has to be funded, from state plan funds, if necessary by reallocation, and can include donor contributions. Unused CRF funds maybe carried forward to the next fiscal planningperiod. 0 Following a severe disaster, the central National Calamity Contingency Fund (NCCF) meets relief expenditures in excess o f a state's CRF fund, subject to oversight by the National Center for Calamity Management (NCCM), constituted by the Ministry o f Home Affairs. The N C C M monitorsthe occurrence and impact o f the hazards mentioned above under the CRF. Funds are released to states after a decision by a High Level Committee on Calamity Relief. Assistance provided to the states by the central govemment from the NCCF i s financed by an immediate levy o f a special surcharge on federal taxes for a limited period. Overall, there i s very limited scope to fund the restoration o f infrastructure from the formal disaster relief funding mechanisms currently in place. Instead, as noted earlier, states are required to maintain infrastructure from Plan funds. Plan funds have become increasingly constrained because an increasing level o f public sector borrowing i s required to fill gaps in the aggregate national revenue account.15 There i s also some uncertainty over the capacity o f some states to effectively employ capital funds, especially when implementation needs to be expedited.16 Post Bhuj (Gujarat, 1999) earthquake reconstruction and rehabilitation expenditures provides a contemporary snapshot o f the reality o f reconstruction fundinginIndia (Chart 2). An analysis demonstrates: 1. The dependency o f the states, even relativelywealthy states such as Gujurat, on extemal funding, from and through the central govemment when massive reconstruction and repair i s required. 2. Substantial resources were allocated to housing reconstruction (rather than infrastructure or drought relief works). InGujarat, house owners well above the poverty line became eligible for government supported reconstruction. 3. The very slow and in some cases non-existent disbursement o f funds under current ex post funding arrangements. This almost certainly reflects institutional weaknesses as well as funding issues, however the relative contributions are not clear. At the very least it is possible to say that there i s ajoint fundinglcapability gap. ''McCarten (2003). l6Anand (1999),Eleventh Finance CommissionReport, section 9.27 12 Chart 2: Bhuj (Gujarat)EarthquakeFunding/CapabilityGap End2002 - Gujarat Earthquake 4000 3500 -2*E3000 2500 2000 * 2 1500 1000 500 0 1 2 3 4 Expenditure and Funding Source: MOF (see Annex VI for detailed analysis). Furtherevidence o fthe limitations ofthe current approach canbefound by examiningtwo recent disasters in Andra Pradesh (AP), the November 1996 cyclone (affecting the East and West Godavari Districts o f AP with 120 knot winds) and the October 2001 floods (affecting four districts). Memoranda summarizing total damages to major public and private sectors for both events were prepared by the ReliefDepartment. A request was also made for relief fundingfrom the Government o f India. Data summarizing AP's post disaster experience in 1996 and 2001 can be found inAnnex V; this includes data on damage estimates andrelief requests by sector. The analysis compares damage estimates by sector to the capital budgets for the year o fthe event, although it was not possible to compare line by line budgets with damage estimates due to reporting differences in 1996 and 2001. The table also allocates funds received from the Go1 to different sectors. Major conclusions drawn from the data include: 0 Private housing accounted for very large losses in both events, while horticulture suffered enormously in the 1996 cyclone. Both sectors are viewed by AP finance officials as primarily private sector activities 0 The natural disaster funding gap between damages within AP and funds received from the central government i s very large. The percent o f damage to the public sector not funded by the center was 98% for the 2002 floods and 96% for the 1996 cyclone. Afier allowing that the original damage estimates are probably inflated, it i s still clear that the state still bore most o f the losses. As these events do not represent the full range o f natural calamity severity, it is possible that future events o f a higher severity would have smaller funding gaps because o f larger assistance from the Go1and international sources. However the absolute financial burdenon AI? would inall probabilitybe higher insuch a case. 13 0 Damage to the public sector as a percent o f AP's capital budget i s estimated to be 16% for the floods and 80% for the much more severe cyclone. Each event also affected the public sectors differently, with roads beingmore susceptible to flooding, while electric power was more affected by cyclonic winds. 0 No damage data is available for the private commercial sector. The government has indicated that businesses are expected to buy insurance or otherwise take full responsibility for their losses from natural disasters. After the 1996 cyclone, private commercial sector firms applied for government assistancebut were deniedany assistancebythe state. Thus, the state o fAP relied inpractice largely on its own resources and other sources o f funding rather than on the center for fimding reconstruction o f damaged assets. Discussions with state finance officials indicate that as a matter o f practice, the state does not borrow any funds from the center or from banks specifically to fimdreconstruction. As there i s no other likely source o f fimds, reconstruction is funded by reallocating current budgets. Losses occumng early in the fiscal year before budgets are fully committed are likely to be funded earlier than events occurring later in the year. Should current budgets prove inflexible, budgets from future years will be reprioritized to fimd reconstruction at a future time. Such a process results in delayed restoration o f important assets, usually only to a level o fminimumfunctionality. This i s likely to lead to heightened maintenance and substandard reconstruction, with attendant future costs because such reconstruction may be more vulnerable to future natural disasters. Government road officials estimated that it took over two years to replace lost roads from the 1996 cyclone and some roads andbridges may have been abandoned. Insummary, the expected future lives o f some o f the assets will be reduced, while future infrastructure projects necessary to support a growing economy will not be constructed. Substandard capital investment over the long term will retardeconomic growth inthe state. The current funding approach clearly involves a reactive response to each event. Minimal proactive effort i s applied to reducing the future financial and human costs through mitigation, land planning, improved building codes and construction practices, and ex ante fimding programs that provide immediate funds for reconstruction. Without adoption of modem risk management programs, the current lack o f proactive risk management practices will exacerbate future financial losses. This effect will be greater with increases in the population and supporting infrastructure. Application o f ex ante intellectual and financial capital i s the recommended approach for assisting Indian states inmanaging their catastrophic risk. Earthquakes, cyclones and floods will continue to affect India, buttheir human and financial costs can be reduced. 14 11. Disaster Loss FundingStrategies NeedTo BeInstitutedAs Part Of A FormalRiskManagementApproach In the larger industrial countries natural disaster recovery is typically funded through a combination o fprivate insurance arrangementsand an efficient public revenue system relyingon wide and deep taxation catchments. Inthe case o f developing countries, which have relatively low tax ratios and ongoing fiscal pressures, funding sources for post disaster reconstruction tend to be more varied, with a strong emphasis on assistance from international donors.17 The most common sources o f such funding are multilaterally sourced infrastructure loans and relief aid from donor agencies." Some countries have explicitly factored these sources into their fiscal planningby ensuring that they would have co-funding immediately available in the event o f a disaster, and taking steps to maketheir international public relations efforts effective. l9 2o As Mechler (2002) has pointed out, contrary to the standard Arrow and Lind(1970) formulation, "a number o f developing countries with high natural hazard exposure and a limited ability to cope with disaster impacts need to be risk averse to natural risk." To this end, the World Bank has been developing a country risk management model which i s partly based on corporate risk management principles21but also factors inkey economic andsocial metrics such as government fiscal profiles, the living conditions o f the poor and investments in risk mitigation. The methodological framework (described in Figure 1) implicitly assumes a growth oriented development model appropriately modified by risk management and distributional objectives.22 The first step under this methodology involves assessingpotential losses from natural hazards on a probabilistic basis (see Chapter 111). While loss control planning i s implied to be a distinct activity by the model, price discovery signals indicated through the risk funding and transfer marketsoften act as apositive influence indirectingthe mitigationeffort. Once the assessment o f potential losses i s complete, the second step in this methodology i s to determine how an array o f risk reduction techniques (mitigation) can be used to reduce the identified loss exposures. Reducing the loss from hture catastrophic events should be an essential part o f any risk management program. The most beneficial mitigation programs are those that are implementedbefore or at the time o f new construction, when the incremental cost o f adding disaster-resistant design features to withstand wind, water or shake forces i s usually a small percentage o fthe total capital cost. " McCarten(2003)pointsoutthatcapitalexpendituresbythestateshavedeclinedfrom31%oftheirrevenue aggregates in 1980/81 to 17% in 1996197. Investment inpower, irrigation, roads andurban infrastructure has stagnatedand operations and maintenance expenditures have declined. '* Phase 1o fthe Gujarat Earthquake EmergencyReconstructionProgram, 2001, was financed byre-allocating the proceeds o f one loan and eleven credit agreements already approved or already active. l9 See the case o f Bolivia inFreemanand Martin (May 2002) 2o Inthis regard a number o f commentators have compared the coordinated and very effective performance o f the Central American States after hurricane Mitch and the relative lack o f coordination o f the Caribbeanstates after hurricane George. India does not make requests for post disaster aid as a matter o fpolicy, but accepts assistance offered suo moto. 21Doherty (2000). 22 J.M. Albara Bertrand (1994) and others have argued that the growth model increases vulnerability and should be modified, but this debate is outside the scope o f this paper, which seeks better solutions within the existing paradigm. 15 Land use planning can also provide substantial risk reduction benefits, by banning or freezing construction in areas prone to windwave/ erosion/landslide/liquefaction/earth quake faulting or ground settlement. Other mitigation projects, involving retrofitting buildings originally constructed with little attention to their disaster resistance performance will be more expensive than for new buildings, pointig to the importance o f adequate enforcement o f construction codes through on-site inspections at the construction stage. Even for poorly constructed buildings such measures as providing roof tie downs for anti-cyclone design improvements, i s often a worthwhile improvement. Other measures for flood control via levees and drainage culverts can also be very cost effective. In general, mitigation i s more cost effective against events which occur with higher frequency, since the benefit o f the mitigation will be higher the more times the event occurs. Conversely, mitigating for events with very low frequencies will probably not be economical, and it i s inthose situations where risk transfer via insurance i s likely to prove more cost effective. The third stage o f the described decision model i s to provide guidance as to the most effective fundingandrisk transfer mechanisms, allowing for longer term economic and social imperatives. Inthis regardFreemanand Martinin examining optimal natural disaster funding arrangements for four Latin American and Caribbean countries have built on a framework outlined by UNCTAD in a 1995 s i ~ d y . The methodology they have developed takes a formal and ~ ~ structured approach to the funding question, and in particular identifies what i s called the "natural disasters resource gap," which i s the difference between the expected fiscal cost o f an event and known ex post sources o f government funding that could be tapped. As there is inevitably a positive and usually non-linear correlation between the severity o f an event and its rarity, the resource gap itself is a non linear function o f event frequency. This non-linearity is further complicated by the fact that aid and other resources are themselves likely to vary according to the nature and size o f a catastrophe. Figure 1: Country/State Risk Management Model 23UNCTAD (1995). 16 The following table (Table 6) illustrates the presentation o fthese results ina format that would support policy decisions. Table 6: FundingGapReconstruction and Rehabilitation In practice it is considered unrealistic for cognitive, financial and political reasons to expect govemments with inherently limitedresourcesto provide for extremely rare events intheir fiscal planning. However, the mission team believe good economic management would cater for events with a probability o f 1%or more o f occurring. The main types o f ex post funding, with the possible exception o f direct cash aid, have potential costs as well as benefits (see Box 1). Despite this, experience to date indicates that govemments, especially those running fiscal deficits, will usually employ ex post funding before resorting to ex ante funding arrangements. The main ex post sources o f funds are redirected budget, direct aid, tax increases, diverted loans (usually involving the development banks), and increased borrowings including from the central bank.24 Direct aid i s assumedto be an important source o f funds in some of the models now being developed. However, experience to date indicates that only a small proportion o f this i s usually inthe form o f cash and available for reconstructi~n.~~ In addition, intemational aid as a source has been relatively static.26 Diversion of already approved development loans i s also seen to be attractive source largely because this can usually be effected with fewer bureaucratic roadblocks and conditionalities than are involved in producing new loan arrangements. Taxation i s probably the next most popular ex post funding source, although this is normally partly offset by exemptions and deferrals given to those affectedby the event. 24This is frowned uponbythe international financial institutions and is ultra vires insome countries. 25Freeman(2002) estimates that globally, approximately 9% o f direct stock losses are covered by direct aid on average. 26OECD DAC wwwl.oecd.org/dachtm/aidglancehome.htm. 17 Box 1: Tradeoffs inEx Post Funding Sources of ex post fundinginvo tradeoffs and often have economiccosts. Development1 reconstruction potentially pro reconstruction, particularly that of hou purpose. Increasedborrowings may a increasing the opportunity t, andthis can potential umablymeanthat what wer ificant and sustained, c Evendirect aid couldhave The only practical way of future based on probabilistic 1 portfolio chosen(given the Even if funds are available, public sector delivery mechanisms can pose challenges. This i s not just a developing country problem, however the situation is exacerbated in a poorer economic environment, and there i s a need to develop transparent and objective aid allocationtriggers and guidelines. For example, after the Latur earthquake o f 1993, illiterate widows, unaware o f the laws, lost out on reconstruction or home replacement aid27, and, as a general observation, existing channels o f distribution o f govemment assistance to affected people have been costly and inefficient. It is estimated that in India only one-fourth o f govemment assistance expenditures reach the intendedbeneficiary.28 Inaddition, and as noted inthe previous chapter, much o f the available funding i s often not employed, reflecting a lack o f planning and limited relevant humanand other resources and inadequate institutional capacity. The four generic ex ante fundingmethodologies are risk transfer (usually specialized catastrophe insurance and reinsurance), the establishment o f insurance reserve funds (backed by hedging instruments such as reinsurance), inter-temporal smoothing (finite reinsurance) and the arrangement o f contingent debt f a ~ i l i t i e s . ~It~can be easily demonstrated that fairly priced insurance usually dominates the other three alternatives in the case o f infrequent catastrophic events. The main reasons for this are related to timing and human nature; insurance reserve funds without an insurance base or other financial support will typically take decades to reach sufficiency and inthe interim are subject to political bids andthe possibility o f investment losses. Contingent debt can often be arranged; however, simulations demonstrate that if commercial interest rates are applied, contingent debt i s a superior approach to insurance only if a disaster happens relatively quickly after the facility i s established or if the price o f insurance becomes exce~sive.~' Inaddition, experience to date indicates that some countries are able to borrow on unchanged terms after a natural disaster, making the facility and arrangement fees a deadweight cost. A combination o f insurance, reserve funds and contingent debt can be optimal when 27World Disaster Report (2001), page 21. 28Per discussion with Joint Secretary (Expenditure), Ministry o fFinance, Government o f India, New Delhi. 29 Stand alone reserve funds are so clearly inferior to the other alternatives for low frequency events that they have been ignored. They may have a role for less severe more frequent events, along with finite reinsurance. 30Freeman (2002). 18 reinsurance prices peak or ifthe contingent debt approximates grant money (e.g. IDA facilities). Inter-temporalsmoothing by its very nature i s best suitedto more frequent less severe events. Thus insurance can dominate on the basis o ftiming benefits alone. It may also lead to marginal economic benefits when opportunity costs, behavioral affects and variability are taken into account. For example, if insurance changes economic behavior by enabling economic agents to pursue riskier but more productive activities it can have a permanent positive impact on the economy, independent o f the consumption smoothing benefits it creates. Giles and Bigger (2000) in an economic study for Turkey state that "under the assumption that individuals withhold 30% o f their savings and firms withhold 15% o f their capital for the purposes o f self- insurance, we find that an economy operating without insurance operates at 8% below potential GDP." 31 Insurance can act conceptually as a signaling advice ifpricing accurately reflects risk levels and i s passed on to the consumer, encouraging cost effective mitigation and appropriate avoidance behavior, such as not buildingon flood plains. Equilibrium modeling also raises questions about the returns obtained from the altemative allocations o f post disaster funding.32 As mentioned earlier, governments normally allocate priority to the replacement o f essential infrastructure and to basic consumption maintenance for the poor (particularly inrural areas). The replacement o f keyproductive public infrastructure i s hardlycontroversial, and it has been amply demonstrated that the poor suffer disproportionately from natural disasters.33As Lava11has pointed out, the use o f economic criteria and cost-benefit equations for attempting tojustify risk mitigation and reductionmay reap rewards for the modem sector economy, but this i s not necessarily the case for the poor and traditional sectors that make up the majority o fthe victims of disaster.34 The attainment o fmore secure livingconditions for the poor and a substantial reduction in their vulnerability i s more a case o f ethics and social justice than o f economic rationale andefficiency. A more questionable area where political pressure often leads to misallocation o f scarce government resources is reconstruction o f private housing in the aftermath o f natural disasters for those who normally would not be viewed as poor.35 This use o f funds i s o f marginal economic benefit. In fact the direct allocation o f government support to the small business sector would probably have a much greater impact on restoring economic activity, although this i s an equally questionable use o f public funds if insurance i s available. In practice, formal government support to the small business sector (including cash crop fanners) after a disaster i s normally restricted to the relaxationo f credit terms from banks and other lenders. The only way governments can avoid political pressures on the housing front from the better off sections o f society is to ensure that a fairly priced catastrophe insurance product i s available and that those who can afford it are either requiredto purchase it, or at least have a good incentive to do so. This requires that they are in tum sufficiently convinced that the government will not assist them inreplacing or repairing their property inthe event o f a future disaster (see Appendix 11). 31This employed a two stage approach using standard loss modeling instage one to provide inputs into a static equilibriummodel. 32 Lahiri et al. (2001) have also raised this issue based on pragmatic observation. 33 See for example Annual Disaster Report, Red Cross and RedCrescent (2001). 34A. Lava11(1999). 35 See R. Gibert (2001) for a full discussion o f issues relating to the post disaster funding o f housing reconstruction. 19 A complication inIndiais the prevalence of informal housing, reflecting the large sections ofthe Indian economy that operate outside the formal economy. Thus any mandated scheme may appear to be inequitable given that many o f those inthe informal sector, such as small business people, have greater financial resources than those in the formal sector, a high proportion o f whom are low paid government employees. However, government reconstruction support for informal sector housing is likely to support the rapid revival o f the SME sector after a disaster and can possibly be justified on these grounds alone. Thus any practical approach in the foreseeable future i s likely to be built around protecting the balance sheets o fthe housing finance intermediaries (see Chapter V, Annex VI1 and Appendix V) and possibly mandating insurance for registered housing, subject to income level. A further possible legitimate use ofpublic resourcesinvolves the enhancement ofmitigation and disaster recovery capacity. Unfortunately, the measurement and cognitive challenges to justifying mitigation expenditure are even greater than for insurance.36 Discussions with local officials point to two key criteria for investingin mitigation in India. The first i s that it clearly saves lives and the second i s that sufficient ongoing funds will be available to sustain the effort. Even in wealthy countries with well developed insurance markets the loss potential can be so large that the insurance markets are unable to provide sufficient capacity at acceptableprices. In some o f these cases, special state mandated catastrophe insurance arrangements have been made, usually in the form o f a private/public partnership riding on the base o f the private insurance system. Industrial countries and states with such arrangements include France, California, Florida, New Zealand, Norway and more recently Taiwan (see Appendix I). Inaddition, someof the wealthier industrial countries provide subsidies through specialized government catastrophe insurers, such as the National Flood Insurance Program inthe U.S. A number o f countries with less developed private insurance systems have implementedor are considering implementing a variation on this modality, which involves the establishment o f a mandatory specialist catastrophe insurance mutual, but with the private sector having a distribution role. The Turkish earthquake pool is probably the best known o f these more recent efforts and has become the basic model for other transition andpost transition jurisdictions. The main purpose o f catastrophe pools is to act as efficient intermediariesbetweenthe ultimate consumer and reinsurance markets. Inaddition, becausereinsurance capacity and pricing can be highlyvolatile (Chart 3 shows rate variation for the US.prior to September 2001) the poolsneed to accumulate sufficient h d s to be able to smooth the domestic cost o f risk transfer by varying the level o f local risk retention. Further financial support and smoothing capacity can be arrangedinthe form o fcontingent debt; international financial institutions have recently shown a willingness to provide such facilities. 36See P. Kleindorfer and H.Kunreuther (2000). 20 Chart3: Reinsurance PricingVolatility (1984 = 1.OO) 2.25 2.00 1.75 1.50 1.2s 1.oo 0.2s 01984 1986"1988 1990 1992"1994 1996 1998 2001 ' ~ ' ~ " " " ' " Source: Congressional Budget Office based on data from Paragon Reinsurance Risk Management Services. Inthe case of India, the situation is complicated by the centerhtate flows of information and hnds described earlier and inAnnex IV. Inpractice there is no guarantee that the funds made available are appropriate either in terms o f quantum or allocation and a more rigorous and objective approach, based on the methodology described above, is desirable. Given that reconstruction hnds at the margin are often sourced de facto if not de jure (loans and advances have often been forgiven) from the central government, there i s substantial scope for the center to develop incentives for the states to adopt rigorous ex ante risk hnding and mitigation approaches to natural disasters in an effort to control hrther fiscal blow outs and optimize resource allocation. Options and recommendations as to how incentives could be incorporated into an integrated funding, mitigation and advisory institutionalframework appear inChapter V. 21 111. CatastropheRiskModels Point To Different StateLevelLoss Potentials and RiskManagement Imperatives3' The first step inthe risk management model described inthe previous chapter involves a detailed assessment o f loss potential in the jurisdictions o f interest. Given the costs o f such studies, an initial broad filter i s applied, often based on historical patterns, to determine which specific exposures should be examined. Duringthe last 110 years, the coasts of Orissa, Andhra Pradesh, Gujarat and Maharashtra were hit by 102, 73, 21 and 6 cyclones, respectively. Extremely violent winds and heavy rains associated with tropical cyclones led to major floods, storm tides (combination o f storm surge and astronomical tides) and coastal inundation. Inthe case of the May 1990 cyclone inAndhra Pradesh, the total loss o f public and private properties was estimated as $480 million, while the estimated economic loss due to the 1999 Orissa super-cyclone was $2.5 billion. (Table 17 in Annex 1presents a list o f 34 largest cyclones that made landfall inIndia over the last century.) The history o f devastating seismic events has been no less frequent. A large part o f continental India i s prone to shallow earthquakes o f magnitudes (M) o f 5.0 or more on the Richter scale. Giant earthquakes of M> 7.5 have occurred inKutch, the Andaman islands and the Himalayas. The largest earthquake in India, o f 8.7 magnitude, took place in the Shillong Plateau in 1897. The extremely highintensityo f this quake and the 1950 quake o f M 8.6 in Sadiya region ledto serious consequences. Rivers changed their course, ground elevations were permanentlyaltered and huge rocks were thrown highup inthe air. The most recent massive earthquake, measuring 6.9 on the Richter scale, struck Bhuj in Gujarat on January 26, 2001, and i s conservatively estimated by the World Bank to have causedproperty losses (public and private) o f $2.1 billion, entailing a reconstruction cost o f $2.4 billion. Table 18 in Annex 1 has been com lied from various sources andpresents a list o f some o fthe most damaging earthquakes inIndia.83 Although perhaps not as catastrophic as cyclones and earthquakes interms o f loss o f life, floods are India's most frequent peril and cause large economic damage. Occurring almost annually in peninsular India, floods are caused by inland rainfall, rivers in flood plains and storm surges along the coast. Average annual rainfall in India i s approximately 115 mm; almost 80% o f the rain falls duringthe south-west monsoon which lasts from June to September. Tropical cyclone storms which occur during the pre and post monsoon periods during the months o f May, October, November and December also bring heavy rainfall in short durations o f 1-2 days. Although high, the average rainfall numbers are somewhat misleading due to a considerable variation in seasonal occurrence and spatial distribution o f the rainfall. An analysis o f national damage figures since 1953 shows that on average every year, floods affect about 7.5 million hectares and cause losses o f over US$200 million (in real terms) in India. The loss includes damagesto an average o f 1.17 millionhouses, amounting to a direct loss o fUS$28 million, and a loss to public utilities o f US$78 million. In 1988, the losses amounted to nearly a billion dollars 37All factualfindings presentedinthis chapter are the result o fa detailedriskmanagement study conductedby RMSI, an Indiansubsidiary of RMS, an international risk modeling consultancy, retained for this task. 38Major sources: I) of the expert group onNatural Disaster Prevention, Preparedness & Mitigation having Report bearing on Housing and Related Infrastructure-Part I, 11) GSI, Seismotectonic Atlas of India & its Environs, 1998; 2000, Calcutta. 22 andinthe 1978 floods, 3.5 millionhouses were damaged. A summaryo f flood losses between 1953-2001 is given inTable 19, Annex 1. Selection of states and perils - Given the above described exposures o f the country to natural disasters, the goal o f the study was to analyze and quantify the impact o f historical and probable fiiture natural catastrophes on four States that suffered extensively from natural disasters in the recent past. A s a result, Andhra Pradesh, Gujarat, Orissa and Maharashtra were selected as case studies. The study's key major objectives were to: 0 Create a reasonably comprehensive exposure database for residential buildings and public infrastructure. 0 Assess the nature o f the hazards affecting the region, measurethe exposures and vulnerability o f districts/ blocks in the region to catastrophic shocks, and construct hazard maps based on the severity and frequency o f hazards involved.39 0 Develop an "actuarially sound" flexible loss model that can be used for catastrophe risk management at the state level. The selected state and perils combinations are listed inTable 7 Table 7: Modeling Scope of State and Peril Combination Andhra Pradesh Gujarat Maharashtra Ollssa Cyclone X X X Earthquake X X Flood X X X I Note: "X" meansincludedinthe modelingscope. I Furthermore, due to the limited availability o f data, the scope o f the modeling with regard to potential losses was limited to: public infrastructure (consisting o f educational, medical building, roads and bridges) and housing (residential dwellings). Government buildings, utilities, minor irrigation systems and commerciahdustrial property are not included in the study and this translates into lower damage estimates thanwould be expected inpractice. Methodology - To arrive at probabilistic loss estimates, stochastic events from the characteristics o f historical events were generated using simulation techniques. The simulations were carried out on occurrence parameters o f the peril and the probability o f occurrence o f all events likely to cause damage to assets. The occurrence parameters incase o f an earthquake are location, magnitude and depth, and in case o f a cyclone are central pressure, forward velocity and direction o f landfall. The generated set o f stochastic events was then used in four modules o f the probabilistic risk model, as shown in Figure 2. These modules are explained in brief below. 39 For the purposeofthe study each state was divided.intoseveralblocks of a fixed size. 23 Figure2: The ProbabilisticRiskModel HazardModule I T ExposureModule I Vulnerability Module + Loss Analysis Model Hazard module: Once the parameters o f each event in a stochastic set are defined, this module can analyze the intensity at a location once an event in the stochastic set has occurred. In earthquakes, the intensity o f ground shaking i s represented as MMI, and in the case o f cyclones the unit i s Peak Gust to measure wind speed. This module models the attenuatioddegradation o f the event from its location to the site under consideration and evaluates the propensity o f local site conditions to either amplify or reduce the impact. The potential intensities o f the three selected hazards: cyclone, earthquake, and flood, were assessed in separate hazard modules, which are discussed inAppendix 111. Exposure model: The exposure values o f "assets at risk" at block level for the four states were estimated either from available secondary data sources or were derived from the distribution o f population at the district or state level. Based on this data, the module then computes the value for all types o f exposures as a product o f multiplication o f the area o f total building inventory andthe averagereplacement cost perunit o finventory. Vulnerability module: The model quantifies the damage caused to each asset class by the intensity o f a given event at a site. The development o f asset classification i s based on a combination o f construction material, construction type (for example a wall and roof combination), buildingusage, number o f stories and age. Estimation o f damage i s measured in terms o f a mean damage ratio (MDR). The MDR i s defined as the ratio o f the repair cost to the replacement cost o fthe structure. The curve that relates the MDRto the peak gust or intensity o f ground shaking at the site i s called a vulnerability function. The study has developed vulnerability functions for differentasset classes andperils. Loss analysis module: To calculate losses, the damage ratio derivedinthe vulnerability module i s translated into dollar loss by multiplyingthe damage ratio by the value at risk. This i s done for each asset class at each location. Losses are then aggregated at block, district, or state level as required. Loss estimates The quantification o f risk for the four selected states i s the key objective o f this - risk assessment. The study yielded estimates o f average annual loss (AAL) with standard deviation and o f probable maximum loss (PML). Further outputs include loss exceeding probability cuwes (OEP/AEP) and the pure risk premium (PRP). It is worth mentioning however that due to the stochastic nature o f risk modeling undertaken for the purposes o f this 24 research and significant data limitations, all estimates o f risk exposures produced by this work are likely to suffer to a greater or lesser degree from statistical uncertainty, a factor to be considered inmaking policy decisions. Average annual loss: Average annual loss (AAL) i s the expected loss per year when averaged over a very long period. Computationally, AAL i s the summation o f products o f event losses andeventprobabilities of occurrence for all events inthe event loss table (ELT). The eventsare an exhaustive list affecting the location/ region under consideration generated by stochastic modeling. Inprobabilistic terms AAL i s a mathematical expectation and broadly represents the Arrow Lindannual cost that would be budgeted for ina large andwell diversified economy. The A A L s expected from future events are presentedexhaustively in technical annexes o f the main study by district, by peril and by asset class for each o f the four states. Table 8 presents AAL summaries for the four states; tables 24 and25 inAnnex I1present AAL summaries along with standard deviation (SD). Andhra Pradesh suffers the highest onging losses followed by Gujarat, Orissa and Maharashtra. Maharashtra suffers far lower losses despite having the highest exposed value when compared to other three states. Following the trends inthe exposed values, housingaccounts for most o fthe losses inall four states. Table 8: Average annual loss summary State All perils (US$ Million) Andhra Pradesh 82.9 Gujarat 64.9 Maharashtra 2.8 Orissa 43.2 Probable maximum loss: The concept o f probable maximum loss (PML) i s commonly used by insurance professionals as a measure o f loss severity. Typically expressed as a percentage o f value, P M L i s not ordinarily the "maximum possible loss," which is the worst possible scenario and which would, in many cases, be 100percent of the property replacement value. Although actual losses can often exceed the PML estimates, they provide useful statistical approximations o f underlyingrisk exposures. Stochastic catastrophe risk models, including the one used inthis study, are now available inthe marketplace to define and compute the PML. For the purposes o f this study the PML i s defined as the largest likely loss to housing and infrastructure in a given state from all perils corresponding to an event with a 150-year retum period. Under this definition, the annual probability o f losses from any single catastrophic event exceeding the givenPML estimate would be equal 0.66 percent. Although various definitions o f PML are available for earthquake risk, there i s little information on hurricane. A.M. Best, a leading insurance rating agency, considers a hurricane PML corresponding to 100-year retumperiod (and an earthquake PML o f 250-year retum period) in recommends a 500-year wind speed for the ultimate load design ofbuildingsand structure^.^^ its capital adequacy e~aluation.~' The American Society o f Civil Engineers standard 40 Dunleavy (1998) (http:l/www.casact.orglconeduclspecsemi98catast/dunleavey.ppt). 41ASCE (1998). 25 Inthe case of flood risk, the PML evaluation involves a 3-step process: first, estimate probable maximum precipitation (PMP); second, compute probable maximum flood (PMF); and third, determine PML corresponding to PMF. The PMF is the flood that may be expected from the most severe combination o f critical meteorological and hydrologic conditions that are reasonably possible in the region.42 The PMF is calculated from the PMP. The methodology adopted for flood modeling in the present study takes historical flood discharges at a particular gauging station as the input andthe startingpoint. Underthis approach it is not possible to estimate PMP andthen compute PMF. The PML corresponding to a 150-year returnperiod i s given inTable 26 by state, by asset class and by peril, and is summarized below in Table 9. Again, similarly to the AAL estimates, housingaccounts for most o fthe loss. Table 9: ProbableMaximumLoss Summary(US$ Million) State Peril Combinedassets Andhra Pradesh All Perils Gujarat All Perils 1,009 Maharashtra Orissa All Perils 479 Pure risk premium: In insurance literature, pure risk premium i s defined as the portion o f insurance rate or premium intended to pay for insured loss under the insurance policy, for the cost o f repairing or rebuildingthe damagedproperty. It does not include adjusting for expenses, underwritingcosts, profit, other contingencies, and inflation, which insurers add to the purerisk premiumto obtain a final rate. Risk models are often used to quantify pure risk premiums for insuredperils. To normalize, risk modelers consider pure risk premium as AAL per thousand dollars of exposed value. For modelers, the major advantage o f pure risk premiumover AAL is that it can be compared across perils, coverages, or geographic areas and usefbl conclusions can be drawn for validating the models. PRP eliminates the effect o f differences in exposed values between comparables and thus simplifies comparisons. Tables 10 and 11contain summaries o f PRPbystate, peril and asset class inunitsperthousand. 42U.S. Army Corps of EngineersManual(1997). 26 Table 10: PureRiskPremiumSummary -(i) Combinedassets-per mille State All perils Cyclone Earthquake Flood AP 1.91 1.41 0.50 GJ 1.37 0.55 0.36 0.46 MR 0.04 0.04 OR 2.64 2.11 0.53 Table 11: PureRiskPremiumSummary-(ii) While in the case of absolute AAL estimates, Orissa did not figure prominently due to a relatively low asset base at risk compared to the larger states, once a ratio o f AAL to value of assets at risk is computed, the state ends up with the highest PRP o f 3.96 per mille among all four states. Andhra Pradesh comes as a close second with its pure risk premium of 3.83 per mille. Again, housing exposures to natural disasters in all four states account for most of the risk. Loss exceedance curves: Aggregate exceeding probability (AEP) and occurrence exceeding probability (OEP) curves are the other two powerful statistical tools for quantifying the severity of losses. Exceedingprobability curves are cumulative distributions showing the probability that losses from a single catastrophic event will exceed a certain monetary threshold. What these losses represent is key to understanding the difference between the AEP and the OEP curves. The AEP curve deals with aggregate annual dollar losses (vs. dollar losses per event in case o f the OEP curve). It shows the probability that aggregate losses per year (Le., the sum o f all losses from all annual events) would exceed a certain threshold. The OEP curve deals with losses from individual events occurring in a given year. It shows the annual probability o f losses from at least one occurrence exceeding a certain monetary value. This distinction between the two AEP and OEP curves is crucial. Since OEP is the cumulative distribution for the largest occurrence in a year, it can be used to analyze occurrence based situations. For example, one can calculate the probability o f activating and exhausting occurrence based contracts such as a policy or reinsurance quota share treaty using the OEP curves. In addition, the OEP curve can provide statistical information on single event covers. Loss EP curves (AEP and OEP) and loss return period tables by peril and by asset class are provided in Annexes of the main study for each o f four states. Tables 12 and 13 present return period losses derived from the AEP curves for all public and private assets combined. For instance, by reading the AEP loss exceedance data for Andhra Pradesh, one can infer that there is a 1percent chance that losses from all natural perils in a given year will exceed US$ 811million. 27 As larger catastrophic events occur rather infrequently, the probability o f events causing losses in excess o f US$ 1.43 billion is only 0.1 percent, which roughly corresponds to a 1,000 year event. Table 12: ReturnPeriodLosses for CombinedAssets (US$ Million)-(i) Source: R M S Delhi Table 13: ReturnPeriodLosses for CombinedAssets (US$ Million)-(ii) 0.002 500 1,997 873 1,742 287 132 0.001 1000 2,436 1,183 2,271 308 230 0.0002 5000 3,126 2,013 2,946 1,244 0.0001 10000 3,283 2,125 3,102 1,553 Despite the limitations mentioned earlier this study presents the first comprehensive effort to quantify the aggregate catastrophic risk exposures in four Indian states. The results o f risk modeling displayed in this chapter confirm that three out of four selected states have large exposures to natural disasters which warrant active risk management. Table 14 below attempts to summarize the findings of the study by ranking the four states by their risk exposures for each of the above described riskmeasures. 28 Table 14: RankingsofFour States inTermsofTheir RiskExposures While in terms o f pure risk premium, Orissa and Andhra Pradesh are most vulnerable to natural disasters, when such relative measures o f risk as PML and PRP are considered, Orissa due to its highly concentrated risk exposures to severe although rare earthquakes and coastal cyclones clearly appears to be in the worst position o f the selected states. Gujarat is the second worst in terms o f PML, followed by Andhra Pradesh. Maharashtra's exposures are found to be rather moderate by any measure. 29 I V India's InsuranceMarketI s UndevelopedAnd InterventionI s Required To IncreasePenetrationAnd MaintainAdequate Capacity Inmost industrialcountries, between 30% and 60% ofall direct and some indirect catastrophe losses are typically funded through private insurance and reinsurance markets. Typically insurers cover private sector property (including housing) and lost profits, although in some cases the public sector also buys insurance (see Annex VI11for an industrial country consumer's perspective on catastrophe insurance). This i s sometimes supplemented by state mandated catastrophe pools, supported by contingent public hnding when the potential loss i s large relative to the premium pool that can be generated inthejurisdiction concerned (see Appendix I). Inmost cases such catastrophepools are closely integrated with the domestic insurance market, which typically has a penetration in excess o f 90% o f households. A major driving force for the establishment o f such pools has been the need to protect the balance sheets o f mortgage providers (Jaffee and Russell (1997)). General insurance consumption in India i s low (Chart 4), even when compared to a trend line based on international norms, although it is not out o f alignment with a number o f other Asian countries inits peer group (as measuredby GDP per capita). This i s despite havinghad an active insurance sector for well over a century. Countries below the trend line have historically been subject to either strong central government control or have had restricted foreign entry into the insurance and/or reinsurance sectors, while those above the line have had active and open insurancemarkets. Chart4: GeneralInsuranceConsumption Low IncomeCountries - 0 400 800 800 l.000 1,210 GDPl Capita $US While India under-performs against its overall peer group as measured by GDP per capita, an examination o f insurance penetration by state shows that an income effect i s at work within the country (Chart 5): 43 43Underlyingdata appears inAnnex 11. 30 Chart 5: Penetrationvs. Per CapitaIncome, 1998-1999 0 1 0 , 0 0 0 2 0 , 0 0 0 3 0 , 0 0 0 4 0 , 0 0 0 P e r C a D i t a 1 n c o m e Source: Preparedfrom data inAnnex I1 While it could be argued that the shiftingdown o f the Indianinsurance consumption curve arises fiom income distribution and inparticular the highlevel o fpoverty inthe country, are-charting o f developing marketsaccording to poverty level shows no apparent causality and indicates that some country specific factors are at work (Chart 6). Chart6: Povertyimpacton InsurancePenetration 0 10 20 30 40 YObelow poverty line Source: CIRE study To explore the issues specific to Indiait is usehl to consider supplyand demand issues separately. Supply Issues- The relatively low level of insurance sector development inIndia has to some extent been attributed to the fact that the non-life insurance industry in India, consisting o f 107 31 domestic and international insurers, was nationalized in 1972, which eventually led to a loss o f service standards and entrepreneurial drive.44 Upon nationalization the industry was consolidated into the four large regional government owned insurers (based in Mumbai, Delhi, Chennai and Calcutta), with GIC as the holding company and national supplier o f supplementary capacity through proportional reinsurance. The negative developmental implications o f this oligopoly were ultimately recognized, and in 1994 the Malhotra Committee recommended that private sector competition be reintroduced. After some resistance the relevant legislation was passedin 1999 and 2000. The need for reform became manifest after the Commission reported; ihe non-life sector showed no growth in penetration (Chart 7), even under the threat o f competition: Chart 7: Non-LifeInsurancePenetrationinIndia,1994-2000 45 1 9 9 4 1 9 9 5 1 9 9 6 1 9 9 7 1 9 9 8 1 9 9 9 2 0 0 0 Sources: Swiss Re., SIGMA. The reforms have included the setting up o f a modem and well resourced supervisor, the Insurance Regulatory and Development Authority (IRDA), under the control o f a senior government official. The IRDA has since carefully and successfully guided the re-opening o f the sector to market competition. Key elements o f the liberalization strategy have included the following requirements: that a substantive Indian enterprise hold at least 74% o f the equity in a new insurer; that minimumcapital be set at a relatively highlevel by international standards (Rs 100 Crores or approximately US$21 million); only a limited number o f highly reputable international players would initially be allowed to enter the market; the existing pricing tariff regime would be maintained for a period; and ensuring that the actuarial profession has a key role to play and should be developed accordingly. Inaddition, steps were taken to expand and energize distribution. As the market matures, some o f these controls will be gradually relaxed. This has already begun with the recent splitting up o f the GIC group and significant policy moves towards removal o fthe tariff system are expected inthe next two years. Initially, four privately owned general insurers were granted entry and this number has recently expanded to eight, all with foreign partners, thus adding Rs 800 crores (US$ 168 million) o f capital or approximately US$400 million o fpremium underwritingcapacity to the market.46 The first full year o f business under the market model was 2002/3, and after ninemonths o f business the new players had booked written premiums o f Rs 965 crores (US$ 203 million) or 9.3 % o f 44 IRDAAnnual Report, 2000-2001, page 2. 45 Penetration i s premium as a proportion o f GDP and is a broadmeasure o f consumptionpreference. 46 Minimumcapital for anew direct nonlife insurer is Rs 100crores. 32 the market. This in part reflects the transfer o f the industrial and fire accounts o f the Indian partners in the joint ventures to their associated insurers. Overall, there remains capacity for further growth o fthe privateplayers. However, establishment expenses and market growth may, at some point, limit their scope to acquire more market share unless they bring in additional capital. The four government owned insurers are writing very conservative premium volumes given their available capital and surplus (Table 15), and have substantial scope to grow in real terms. Direct capacity is further increased by the fact that all insurers are requiredto cede 20% o f their business to GIC (subject to limits for fire, engineeringand energy risks), which i s transforming itself into a professional reinsurer under the liberalized industry structure initiated in 2002. The capital resources o f GIC's stand alone operation were approximately Rs 4,000 crores (US$ 842.10 million) at the end o fthe 2001 financial year. The IRDAhas also suggestedthat insurers establish catastrophe reserves varying from 0.5% to 5% o f relevant net premiums, which will bringIndia into closer alignment with other catastrophe prone countries. Thus ina direct writing sense aggregate premium capacity i s unlikely to be an issue in the foreseeable hture, even if India achieves its potentialor greater according to intemationalnorms (see Appendix 11). Table 15: Capacity Utilization, 2002 (Rs Crore) Source: IRDA Annual~eport(2001-2002). Despite the clear underutilization o f capacity relative to normal risks (Table 15), there i s inadequate capacity to cover peak industrial and infrastructure risks and possibly some catastrophic loss aggregating events such as a windstorm in Mumbai or an earthquake inDelhi. The largest general insurer retains Rs 10 crores (US$2 million) o f catastrophe exposure and buys excess o f loss coverage up to a limit o f Rs 260 crores (US$57 million). There i s an umbrella cover o f Rs 100 crores (US$22 million) on top o f this. This reinsurance schedule is probably more than adequate given the current take up rate o f disaster insurance in India. The average P M L ratio applied o f 30% i s very high for a well diversified portfolio, and reinsurers no doubt reflect the lack o f accurate accumulation data in their pricing, pointing to the probability that India i s currently overpaying for reinsurance, and has some latent capacity which would be released ifbetterdatawere available. The need to access external reinsurance markets is normal for the great majority o f countries in the world, particularly as penetration o f general insurance grows. According to some altemative estimates the Bhuj earthquake generated insurable direct losses o f up to Rs 21,000 crores (US$4.4 billion), although actual insuredlosses amounted to only Rs 750 crores (US$16 million) 33 because o f low levels o f insurance penetrati~n.~~addition, there are individual peak risks In (known as mega risks in India), such as the Reliance petrochemical plants in Jamnagar, which have total sums insuredthat are currently a multiple o f Indian aggregate non-life premiums and have an earthquake P M L (75% o f total exposed value, compared to the more t ical Indian figure o f 30%), which at least approximates the size of the nationalpremium pool!`GIC rightly limits itselfto a retained exposure of Rs 50 crores (US$11million) for industrial fire and Rs 90 crores (US$20 million) for advance loss o f profits (ALOP) and buys Rs 850 crores (US$l85 million) o f excess loss cover, with Rs 100 crores (US$22 million) on top. Earthquake i s an optional add on coverage, as opposed to RSMD49which is covered by default. Regardless o f packaging the take up o f additional coverages over and above basic fire and engineering wordings has been disappointingly low. After the Bhuj event the President o f the Bengal Chamber o f Commerce was quoted as pointing out that despite reasonable pricing "owners o f large buildings in Gujarat excluded both these policies and only an estimated five percent o f households with fire policy paid extra for earthquake."50 The nexus between industry structure and capacity i s subtle for general insurance and India's strategy to date o f attracting enough serious players to ensure adequate competition while avoiding fragmentation appears to be appropriate. Studies carried out to date appear to demonstrate that industry fragmentation actually lowers domestic risk bearing capacity while foreign entry increases it. 51 Despite this, and as personal lines and small business property insurance increase in popularity, India will need to ensure that it has a sophisticated and competitive capacity to deal with international risk transfer markets. This points to allowing a small number of additional qualified reinsurance intermediaries into the local market to add technology and to provide a healthy level o f competition. A number o f countries, including Turkey and China, have withdrawn mandatory cessions to their national reinsurers on a phased basis, andthis may become appropriate inIndia at some time inthe future. Where the private sector cannot cover a risk the Indian government has shown a willingness to step in. After the events o f September 2001, reinsurance cover largely ceasedto be available for terrorism cover in India and this coverage was delinked from the basic fire wording. An initial alternative considered involved a surcharge on premiums which would build a reserve, with the government providing initial reinsurance, to be repaid over time by the insurers inthe event o f a claim. This was to be backed by a formal insurer catastrophe reserving system, with tax incentives. In the event, the industry asked that the levy be treated as premium with central poolingto be administered by GIC.52 Evenifreinsurance is available, highprices are another potential stumblingblock to increasing catastrophereinsurance penetration. One way o f lowering reinsurance pricing i s to provide good information about catastrophe event risk. In this regard India i s currently not well served, despite having a leading edge intellectual and technical capacity to do the necessary applied 47GIC showed an estimate ofRs 361 crores (US$78 million) as its loss fromBhuj inits 29" annualreport (2000- 2001). By contrastHurricaneAndrew inFlorida in 1992resultedinaninsured loss ofUS$17.0billionagainst an estimatedtotal loss ofUS$30.0 billion, muchof which was retained by direct insurers. 48PMListhe levelofriskupto which the insuredwishes to seek coverage giventhe relevantlossexceedance curve. 49Riot, strike andmalicious damage cover. This i s arequiredcoverage ifterrorism cover is provided. ''Business 50 Line,April 23,2001. Outreville (2000). 52IRDA2ndReport2001-2002, page40. 34 research and a huge but fragmented database. Munich Re recently53 presented a paper on catastrophe potential in India and this appears to have acted as a catalyst for a more active debate. In particular, it appears that there i s a need to do more physical modeling o f the earthquake process on the subcontinent, to review earth uake zones and to allow for soil type, particularly in the large cities on the Ganges flood plainP4 Inthis regard, the recent decision to move to a single country wide solidarity based earthquake premium loading o f lop per RslOOO for smaller property risks potentially limits the scope to encourage mitigation and to ensure a contribution to the Indian disaster premium pool from those in less earthquake prone areas. When a similar pure solidarity approach was attempted in Turkey it became very clear very quickly that people in general are aware o f relative levels o f risk and that those in areas o f low seismicity had a limited willingness to subsidize their countrymen in high risk zones. The 15 risk zone premium structure finally adopted by TCIJ? now provides for relative risk levels reflecting aproperty's locationandconstruction quality. Other supply constraints in India, and poor service in particular, will presumably reduce in importance as competition forces better performance; however it is likely to be on the claims handlingside that the big four will ultimately rise or fall. Anecdotal evidence, backedby GIC's own published productivity and establishment data, point to very slow and potentially bureaucratic claims handlingprocesses, which favor the insurer inthe event o f dispute.55 A key problem inthis regard appears to be a hierarchical approval process and an unwillingness on the part of front line officials to exercise discretion, even when the relevant authority has been delegated.56 The insurance supervisor i s now taking vigorous action to redress the balance between insurers and claimants, including the establishment o f various recourse mechanisms for consumers. Possibly the most important supply innovation will be the opening up of new distribution systems and the creation o f a more professional and well remunerated sales force. Commissions for household and small business coverage have historically been lo%, although these were increased to 15% as an incentive to agents to market to the rural sector and certain other under- serviced communities. Other distribution channels being explored include brokers (a substantial number of licenses have already been issued, including one JV involving an international broker), bancassurance and direct selling using the internet and other electronic media.57 These are highly desirable reforms as the sales and distribution system has been handicapped by a lack o f incentives and hence entreprene~rship.~' However it seems unlikely that any o f these mechanisms will efficiently or effectively reach the less affluent sections of the population, particularly the poor, although the government has mandated that the new players should underwrite certain minimumproportions o fbusiness inthe rural and "social" sectors as follows: 53 Institute o f Administration Natural Disasters Conference, Delhi, February 2002. 54 K.Mishra (2002). " CIREestimatesthatapproximately45%ofclaimsoutstandingatanyonetimehavebeenpendingforatleastone year, and 23% have beenpending for at least three years. '`The Comptroller and Auditor General o f India was reported as questioning the flexibility that insurers showed in settling clams after the Bhuj earthquake, but evidently accepted that special circumstances deserve special ''''Banksfour approaches. may enter the market as corporate agents, strategic investors or promoters. The state owned insurers announced inMarch 2003 that they have reduced their traditional marketing force by 32% through a voluntary retirementscheme. 35 Rural Sector (of gross written premium): First Year 2% SecondYear 3% Thirdyear 5% Social Sector (lives insured): First Year 5,000 Second year 7,500 Thirdyear 10,000 Fourthyear 15,000 Fifthyear 20,000 Performance against these targets has been mixed, with the private sector players generally performing adequately, although two insurers have received notices relevant to the social sector. The definition o f rural sector was recently modified to include townships where 25% of the male population is engaged in farming (rather than 75%) and one o f the new insurers appears to be making some inroads into this market. The performance of the large state insurers has been "tepid" and the IRDA has had discussions with the relevant management." Despite these mixed results, the delivery o f risk management products to the rural and social sectors is a legitimate objective and mechanisms need to be found to effect this result. The k e y issue appears to be the creation o f special intermediaries between the formal sector and the target group, which can deal with moral hazard, design andprice appropriate products and provide relevant education on risk management at the level o f the household and individual economic activity. This is an area undergoing a rapid evolution around the developing world and a number o f the most advanced experiments are being carried out in India.60 Early results indicate that insurance can b e sold to the poor as part of other service offerings, but that as a stand alone product it suffers from lack o f understanding and its inherently intangible nature. If a working model can b e found there may be arguments for reducing the minimum capital and other relevant requirements for mutual insurers, as has been the case under EUlaw, subject to minimummembership and other relevant prudential andgovemance requirements. Another mechanism that has been developed inIndia is the state level insurance hnds, which are regulated by IRDA and come under the provisions of the federal insurance law. Their retentions are small and most risk is transferred to one or more o f the big four non- life insurers. In Gujurat, for example, the Government Insurance Fund insures government infrastructure and provides coverage for death and disability under group personal accident policies for the socially disadvantaged.61 The five groups covered are small and marginal farmers, police personnel, people below the poverty line, landless laborers and college students. Premiums are paid by the relevant government department, with the Commissioner of Rural Development being responsible for those below the poverty line. The final supply issue is pricing, which is subject to a tariff for all major classes (75% o f all business) and has historically been distorted by a heavy cross subsidy from property classes to motor insurance, and to motor third party liability insurance in particular. There have been j 9IRDA2ndAnnual Report, page 6. 6o See for example the Self Employed Women's Association insurance model. 61 Inpracticeonly state andmunicipal buses, power generatingassets andassets ofsome state enterprises are covered inGujarat and the situation is likely to be similar inother states 36 numerous attempts to date bythe rate setting body (TAC) to revert to actuarially sound pricing or at least a less egregious transfer, but without success because o f the lobbying efforts o f the haulage industry. Inthe interim, fire rates have been reduced three times based on experience, but continue to produce relatively good results for the underwriters. Recently, the insurance supervisor, as part o f the market liberalization package and recognizing increased competition, has announced that tariffs, startingwith the commercial motor sector, will be phased out. Market based pricing should lead to lower and more risk sensitive property rates over time, ensure that databases become aligned with the key rating factors, and increase the scope to add catastrophe coverages. DemandIssues - There is a price at which insurance ceases to be attractive, even if it is hlly understood and is seen as an intrinsically attractive service by the risk averse consumer (see Appendix 11). The price established by the insurer includes the estimated expected loss, expense loadings and allowances for the cost o f the capital backing the solvency o f the insurer. This equation tends to work best when the pure risk component is large relative to the fixed costs o f the insurance enterprise, including its distribution system, and there i s a degree o f confidence about the risk pricing model being used. This in turn implies good and credible data, limited scope for moral hazard or fraudulent behavior and a relatively large customer base.62 The equation tends to break down for sections o f society that are difficult or expensive to reach, do not understand or trust formal sector financial institutions and have incentives to influence the probability or quantumo f a claim. Thus it seems likely that for parts o f Indiansociety the formal sector insurers cannot directly provide an insurance product at an acceptable price, at least without subsidy. However there are large sections o f Indian society for which a fair premium would, by international standards, be acceptable in term o f their income levels, and which have already shown a propensity to purchase life insurance. It is estimated that personal lines insurance (including compulsory motor) i s between 4% and 6% o f total premium income, and the Bhuj earthquake pointed to less than 2% o f domestic residences being insured in what i s a relatively industrialized state. Possibly the most important factor underlying this low penetration i s a lack of knowledge and understanding o f the insurance mechanism, which in turn tends to be a function o f education and awareness. CIRE reports that "except for policies which are purchased due to govemment mandate (e.g. motor third party liability insurance, public liability insurance etc.) or insistence by lending institutions such as banks or housing societies, customers are largely unaware o f the existence o f non-life insurance policies." The correlation with credit generation inparticular i s quite noticeable (Chart 8). 62 Credible data is largely a function o f the number o f claims generated by a rating class. Typically a minimumo f 1,000 claims are requiredbefore a premiumrate canbe determinedwith a degree o f confidence, for a personal lines rating category where the individual claims distribution is not highly skewed. 37 Chart 8: InsurancePenetrationand CreditDisbursement, 1998- 1999 1 2 1 .-E0 - 0 8 0 6 C n 0 4 0.2 I 0 0 200 400 600 800 1000 1200 ScheduledCommercialBanks'Credit Disbursements1998-1999 Source: CIRE Inthis context, the importance ofeffective public education about the financial consequencesof catastrophe risk exposures and the insurance products available to address them should not be underestimated. International experience demonstrates convincingly that effective public awareness campaigns conducted through mass media and education channels can dramatically improve the public perception o f government sponsored insurance programs and thus contribute to increasing insurancepenetration. An excellent exampleo fhow an effective public information campaign can help to change public attitudes and boost insurance is the Turkish Catastrophe Insurance Pool (TCIP). Only three years after its launch, and despite the initial rather hostile attitude o f the population to any government mandated insurance programs, the TCIP has managed to buildnationwide acceptance o f its insurance coverage. It has now become the most trusted name in the Turkish insurance market due to an active and creative ongoing public relations and mass education campaign, which, according to the TCIP's estimates, has enabled it to attract an additional 350,000 homeowners annually. Another major issue for disaster related coverages is that precedent has been established for the provision o f public fimds for the reconstruction o f the housing sector, creating a potential moral hazard and a lack o fpropensity to seek alternative insurance arrangements. Inthe case o f Bhuj, housing reconstruction accounted for close to a half o f estimated post disaster public sector reconstruction While some o f these outlays may have financed post disaster relief and reconstructions needs for the poor and thus may be justified, it appears that those who could afford earthquake insurance also received hnds which could have beenbetter usedelsewhere. The ease o f access to insurance, and hence its cost to potential consumers, may also be an issue. Anecdotal informationpoints to a highly competitive market developing inthe cities64and there may be opportunities for specialist rural community based insurers to emerge in India as has happened in a number o f industrial countries with large agricultural sectors. These latter 63Lahiri et al. (2001), Table 9. 64There have been a number o f tariff breaches inthe last year, necessitating a 400% increase inthe relevant fine. 38 organizations are often closely associated with farmers cooperatives and the main agricultural inputsuppliers, includingcredit providers. Other factors that have been quoted as accounting for lack o f demand include a fear o f disclosing assets (partly a tax issue), a perceived lack o f secure or negotiable property titles and an inefficient tort system (see earlier comments on claims management). National catastrophe insurance programs require massive enrollment to achieve a balanced and well diversified portfolio o f risk and affordable pricing for insureds, even at the most hazardous locations. That can be achieved either (i) by makingcatastrophe insurance coverage compulsory for all registered homeowners (perhaps, with some minor exceptions), or at least for those borrowing; or (ii) voluntarily, through active public education and mass marketing campaigns. While each country requires a unique solution, a key consideration i s the tradeoff between achieving wide participation through compulsion and the creation o f a public impression o f catastrophe insurance premiumsbeing a tax with consequent adverse effect on households' risk management behavior. Inaddition, the level o f solidarity build into the rating structure requires a tradeoffbetween simplicity, social equity andthe encouragement o fmitigation efforts. 39 V. Findings,Policy Options And Recommendations Findings The combination of an increasingincidence of naturaldisasters and the current approach to funding and applying post disaster relief and rehabilitation(which is being effected in the context of chronic revenuedeficits)detractsfrom India's development program. More positively, significant progress has been made insome states over the last three years (andinthe last 12 months at the national level) in building institutional capacity for disaster management. However this is fragmented and there appears to be no overriding and comprehensive catastrophe risk management framework in existence, although the central government has clearly evidenced a desire to move inthis direction (see Tenth Planning CommissionReport and Report o f the High Powered Committee on Disaster Management). Inparticular, the current national approach to disaster management at the central and state levels suffers from a lack o f institutional incentives and underplays the role o f risk financing, including ex ante mechanisms such as catastrophe reinsurance and contingent credit facilities. As a result, the potentialfunding gap between damages sustained by the states and funds available from all sources to finance theminthe aftermath ofnatural disastershasbeenincreasing (Table 16). Table 16: Catastropherisk exposures as percentage of key economicflow measuresin four selectedstates 150Year PML65 State US$ MM GSDP(%) Tax Revenue (%) FiscalDeficit(%) AP 921 3.3% 28.7% 61.5% GJ 1.009 4.4% 43.7% 32.8% MR 59 0.1% 1.1% 2.7% OR 479 6.5% 41.9% 19.9% The infrastructureof India is in danger of beingsignificantly degraded as fiscal/ capability constraints limit capital expenditure options. The need for rapid emergency repairs post disaster affects the quality o f work, and in one case (Gujarat after the recent earthquake) scarce public funds have been diverted to rebuild housing for self sufficient sections of the population. As financial assistance from the NCCF and CRF accounts for a small fraction o f expected losses, reconstruction o f destroyed or damaged infrastructure is h d e dby redirectingcurrent budgets to the extent possible. Should current budgetsprove inflexible, budgets from fbture years are used to fund the reconstruction at a future time. Such a process results in delayed and inadequate restoration o f important assets, and consequently, reduced hctionality and operating lives. Other consequenceso f current practices are heightenedmaintenance with attendant future costs, as well as increased vulnerability o f the affected assets to future natural disasters. In addition plannedcapital projects necessaryto support a growing economy will not be undertakenor will be deferred, prejudicing future economic growth. 65P M L = Probable Maximum Loss for a 1-in-150 Year event ( similar to the magnitude to the recent Gujaratj Earthquake). 40 These findings are supported by a detailed case study o f two recent catastrophic events inAP, a sources and uses o f fbnds analysis following the Bhuj (Gujarat) earthquake, as well as by more general fiscal statistics from other states. The mission team believes that most states are not financially prepared to deal with the consequences o f severe catastrophic events and that a numbero fstateswould findit difficult to use finds even iftheywere made available. Natural disasters, partly through the destructionor damage of life line infrastructuresuch as roads and clean water supply, increase the gap between the poor and other sections of society. This effect has already beenwell documented elsewhere, but a comprehensive approach to dealing with it has yet to be developed.66 The current fbnding approach has unpredictable impacts on the poor, since the poor have few accumulated financial assets to rely upon following a catastrophe: even a slight decline in government assistance arising from reallocations o f government budgets (for example reconstruction o f houses for the non-poor) can leave the poor fkther behind. As many are dependent on the agriculturehorticulture sector, delays in rebuildinghestoration o f rural infrastructure (such as roads, water supply and electricity) immediately affect those with no or limitedother income and minimal consumption cushions. Ongoingand effectivemitigationis not encouragedby the current fundingmethods,except when donor fbnds are involved and the relevant donor makes this a conditionality. Mitigation has several meaningsand there i s a need to concentrate on those forrns o f mitigation which have the best demonstrable impact (building standards sufficient to at least save lives, early warning systems, etc.) andhave credible ongoing hnding sources that will ensure their sustainability. In light of the overall importance of reducing the country's risk exposures to natural disasters, serious thought should be given to the use o f fiscal and institutional incentives to promote active riskreductionefforts at the locallevel. The lack of institutional capacity at the local level to absorb donor funds following large natural disasters frequently results inthe slower than expected utilization o f extemal aid, as well as leakages and misuse o f fbnding. These factors impair speedy economic recovery and reconstruction efforts. The problem i s frequently exacerbated by the rigid and rather bureaucratic procurement and disbursement guidelines attached to the receipt o f development and reconstruction aid. These guidelines require the creation o f specialized project implementationagencies at the local level, along with specially trained staff that may be inshort supplyat the time ofa disaster. General insurance consumption in India is low, even when controlling for the level of economic development. There are numerically large sections of Indiansociety for which a fair premium would, by intemational standards, be acceptable in term o f their income levels, and which have already shown a propensity to purchase life insurance.67 However, it has been estimated that personal lines insurance (excluding compulsory motor) account for only 4% to 6% o f total premium income, and the Bhuj earthquake pointed to less than 2% o f domestic residences being insured in what i s a relatively industrialized state. Despite a clear underutilization o f local insurance capacity relative to normal risks, there is inadequate capacity See for example Bhatt, Natural Disasters as National Shocks to Poor and Development, World Bank, 1999 and ADB JFPR:IND 36029. 67 Itis estimated that approximately 7% (both rural and urban) o f the Indian population fall into the "non- poor" categories (Deshpande (2003)). 41 to cover peak industrial and infrastructure risks, and possibly some catastrophic loss aggregating events such as a severe earthquake affecting Delhi, without resorting to intemationalreinsurance markets. On the "positive" side, India i s probably overpaying for reinsurance because o f a lack o f detailed risk accumulation data, and has some latent risk absorption capacity which would be releasedifbetter information were available. In summary, the current funding approach to severenatural disasters inIndia involves a largely reactive response to each event. Some proactive efforts are now being made to reduce the future financial and human losses through mitigation (including land planning, improved building codes and construction practices) and disaster preparedness, but there has been limitedscope for the design o f ex ante funding programs which provide immediate funds for reconstruction. This is becoming an increasingly important policy issue as the adverse effects o fnatural disasters will almost certainly only become larger with increased population concentrations and concomitantly increasing concentrations o f social and productive capital. Having examined existing institutional arrangements for mitigation and risk financing at the central and state level and takinginto account the potential finding gaps revealed bythe in-depth risk assessments of four selected states, the mission team has developed a number of policy options and some specific recommendations. Policy Options Develop a Risk Financing Strategy as an Integral Part of National Disaster Management: Mitigation and risk financing are the two pillars o f effective catastrophe risk management at the country and state level. InIndia government actions inrisk reduction andprevention (commonly referred to as mitigation), should be augmented by a formal approach to risk financing. Such a risk financing strategy would consist o f three parts. First, formal risk assessments at the state and the central levels; second, identificationo f fundinggaps; andthird, development of state and national risk management plans aimed at closing the identified funding gaps over time. Such risk management plans are likely to consist of a combination o f mitigation and risk financing initiatives, which inter alia, could include vulnerability reduction programs, catastrophe insurance and access to a federally maintained contingent credit facility. Introduce Fiscal Incentives for Active Risk Management at the State Level. The mission team believes that the size o f identified funding gaps can partially be explained by a lack o f institutional incentives for better risk management at the state Currently, states rely on six main sources to fund relief and rehabilitation work in the aftermath of natural disasters: (i) fundingfrom the state Calamity ReliefFunds(CW) to provide immediatereliefto the victims of natural disasters and urgently restore life-line infrastructure; (ii)the National Calamity Contingency Fund (NCCF), which provides financing for expenditures by state governments in excess o f balances available intheir CRFs following particularly severe events; (iii) annual state capital budgets; (iv) reallocation o f Plan funds, which can be usedfor reconstruction o f damaged infrastructure; (v) contingency funds, including the Prime Minister's ReliefFund; and finally (vi) intemationaland domestic donor finds, uponthe occurrence o f calamities o f great magnitude. The level o fpost-disaster funding to the states is based largely on the size of disaster losses and the state economy. The current system does not provide incentive for those states that may have taken proactive steps inrisk reducing measures. 42 For the purposes o f our further analysis, we have grouped these funding sources into two funding categories: (a) Ex ante funding sources, under which the amount o f financing available for relief and rehabilitation i s pre-arranged and possibly allocated prior to the occurrence o f disasters. This presently consists o fthe CRFs and current capital budgets. (b) Ex post funding sources, which provide funds in the aftermath o f natural disasters. These sources include reallocation o f planned funds from future capital budgets, central post-disaster assistance and donor funds. Overall, it appears that the States have little fiscal flexibility to pursue ex ante risk management initiatives that are not funded from external sources.69 To reduce the fundingkapability gap and the vulnerability o f infrastructure, the existing national system o f post disaster financing could be redesigned to provide strong fiscal incentives for the states to adopt more proactive approaches to risk management. Such a "carrot" based approach to disaster risk management at the local level by the GO1 would be consistent with best international practice available today in developed countries. Two cases in point are the U.S. Federal Emergency Management Agency (FEMA) and the FrenchNatural Catastrophe Program (NatCat). Both programs learned early in their experience that affordable insurance, supplemented by federal grants for disaster management, can provide a strong incentive for disaster prone communities to join national risk reduction programs and adopt these programs' mitigation standards. Inthe case of FEMA, no federal grants or loans are allowed for capital improvements in the flood-prone areas o f non-participating communities. Inaddition, the Flood Disaster Protection Act o f 1973 requires that anyone who applies for a mortgage from federally connected lenders - which means most lenders inthe U S - or who seeks federal disaster assistance or federal loans, must buy flood insurance ifthe property is in a high risk, flood hazard area. By making flood insurance and consequently mortgage financing unavailable to homeowners in communities not participating in risk reduction programs, the law created strong local pressures as well as powerhl incentives for local politicians to join and actively implement the FEMA risk reduction programs. The proposed reforms in risk financing should also translate into improved awareness o f catastrophe risk by homeowners and enterprises, raising the level o f insurance coverage in the country. Policy actions at the state and central level could include: a. Inthe case o f ex post sources o f risk financing, having the Go1 reward states pursing active catastrophe risk management with additional fiscal resources for rehabilitation o f destroyed state-owned assets. It would be desirable to make the quantity o f such additional financial assistance known in advance. For instance, the Government may consider offering amultipleo f aid typically expected from the NCCF to the states that are advanced in risk-management. The financial resources for these extra-budgetary allocations could come from donors and IFIs, including the World Bank, or through reallocation o f the GoI's planned financing for natural calamities. Ideally, over time, more government funding for natural calamity related work would be allocated through this channel. 69W. McCarten (2003). 43 b. Introductionbythe states of a special tax onproperty or a surcharge onpublicly provided services, the proceeds o f which would be earmarked for rehabilitation o f destroyed or damagedpublic infrastructure and would accumulate inthe local CRFs. c. Making infrastructure investments financed by IFIs contingent upon states submitting comprehensive risk management plans for the proposed investment. This would require government policy action that would not only safeguard the contemplated public investmentsbut also promote broader active riskmanagement approaches, including loss reductionand capability enhancement measuresby the states. Modify the existing institutional arrangements for disaster management at the center. While the existing institutional framework for catastrophe risk management i s well developed and comprehensive, the following changes in the system would further facilitate active mitigation, build the capacity to effectively employ funds at the state level, and augment the existing ex post risk financing approach: 70 a. The creation o f a designated Risk Management TechnicalAssistance Unit (RMTAU) could be considered. The R M T A U would have two primary functions: (i) serve as a to technical resource for the RFF (see below), and (ii) provide Technical Assistance and Capacity Building support to the states preparing and implementing risk management initiatives. The RMTAU would operate as an independent unit hosted by the Central Relief Commission. It would be staffed with insurance and risk management professionals andwould have an arms-length relationship with the RFF. b. To promote better mitigation practices, the NCCF may also consider instituting a specially designated grant facility for mitigation initiatives o f those states committed to reducing their funding gap. The RMTAU could house such a facility, funded by international donors or the GoI. c. Adoption o f Risk Management Plans (RMPs) by the states, with technical assistance from the RMTAU, would be formalized through an official document guiding all disbursements o f disaster relief from the RFF. The RMPs for individual states are likely to include: (i) assessment o f risk exposures and identification o f the funding gap; (ii) targeted risk reduction measures to reduce the vulnerability o f life-line infrastructure assets, including enforcement o f building codes, improved land use practices, and structural re-enforcement o f exposed assets; (iii) identification o f risk exposures, such as privately-ownedhousing, which can be covered by private insurance; and(iv) acquisition o f catastrophe insurance for peak risks for key public infrastructure, particularly when fundedbythe development lenders. The facilitation of any risk financing initiative would require the creation of a new Risk Financing Facility (RFF) to provide additional financial assistance to those states that are adopting and implementing a risk management approach. The RFF would provide additional resources, sourced from donors, IFIs and the Go1 for rehabilitation and repair o f infrastructure. Disbursements from this facility would be made contingent upon (i)the occurrence o f catastrophe events and (ii) achievement o f risk management performance targets that would be agreed upon between the state and the facility, and certified by RMTAU. 'OThis ideally wouldbepart of a largerrisk managementstrategy. 44 Preliminary analysis suggests that the National Calamity Contingency Fund (NCCF) could be usedas a hosting and managing organization for the RFF. However, to host the RFF, the NCCF would require sufficient loss assessment and claims settlement capabilities to ensure an expedient, fair andtransparent compensationprocess. Explore the use of contingent credit facilities for the purposes of catastrophe risk financing, and in support of risk management incentives at the state level. To finance reconstruction o f public infrastructure and housing, the Go1has been utilizing World Bank and ADB emergency and reconstruction loans made in the aftermath o f natural disasters. Some o f the funding for reconstruction has also come from reallocations in the portfolios o f developmental lenders. Despite the fact that most o f these lending operations contained mitigation components, they have major inherent limitations. First, as evidenced by large funding gaps that exist at the state level, the GoI's reliance on ad hoc post disaster reconstruction loans from the development banks has done little to improve systemic risk management inthe disaster prone states. Second, despite the considerably shortened time frames requiredfor the preparation o f emergency reconstruction loans when compared to the World Bank's other lending operations, emergency loans can be relatively slow to disburse (compared to immediately disbursing ex ante mechanisms) due to the World Bank's project procurement rules (although simplified) and other safe-guard policies.71 As a result, these lending instruments are not appropriate for meeting the Government's immediate and often significant liquidity needs in the aftermath o f natural disasters which, if unsatisfied, canhave far reaching negative social and economic implications. For instance, a contingent credit facility similar to that supporting the Turkish Catastrophe Insurance Pool (Box 2) could be extended to the NCCF in support o f the RFF. Such a facility would then become available to meet claims o f the states in the aftermath o f natural disasters, provided an acceptable state risk management program i s in place. Compliance with the terms set out in the risk management plan would be viewed as a major disbursement criteria. A matching contribution from the central government budget would be expected under such an arrangement. Such a contingent credit line would enable the RFFto operate as an effective fully pre-funded provider o f liquidity to the disaster stricken states. If disbursed, the facility backing the RFF could be then replenished without any major costs. The above suggested funding approach for natural disasters would enable the Bank to switch to a proactive mode o f lending for natural disasters by replacing multiple ex post future emergency lending operations with a single line o f credit, and provide the Go1with immediate liquidity to meet reconstruction needs inthe aftermath o fnaturaldisasters. While unconstrained funds can be more expeditiously reallocated to changing project needs following a major catastrophic event compared to earmarked funds, the advantages o f fungibility should be balanced by the increased importance o f budgetary discipline. Inthe immediate post- loss environment, information i s often scarce and the capability o f the government to respond is stretched. Demands for shelter, food, water and health services for affected populations are immediate, as are those for the restoration o f power and other critical services necessary for the resumption o f economic activities. Conflicting demands as well as alternative visions for the future makeit difficult to pursue value maximizing budgetinginthe disordered and emotionally " Note,thisisbaseduponglobalexperienceandmayormaynotbedirectlyrelevanttoIndia.Insomecountriesit has been observed that though the World Bank's funding may be forthcomingpost disaster, there i s a significant delay inspending by the client. 45 charged post-disaster condition^.^^ However, the problem o f post-loss liquidity inherent to the proposed model o f hnding natural disasters i s not insurmountable. The key i s to have a clear and well-prepared risk financing plan that can be used as the main framework for a post-emergency disaster fundingbudget. Such a plan can be worked out inadvance inconsultations with disaster prone communities, local and state governments, international donors and development lenders. 72Clarke and Doherty, "Development Enhancing RiskManagement," Working paper, August 2003, 46 Box 2: Turkish ContingentCredit Facility Turkisheconomyis over 50percent ofthe n 47 Introduce incentives and perhaps mandated requirements to increase the utilization of catastrophe insurancemechanisms by the private sector, includingbetter off ho~seholds.7~ These incentiveshequirements could be part o f the requirements mentioned above for additional funding from the center and are likely to vary between states according to local realities. Approaches that couldbe considered include: Requiring that replacement cost catastrophe insurance is purchased when mortgage financing i s granted (see Appendix V). This in some cases could be effected through a relatively small addition to the interest rate and could even be accompanied by a slight offset reduction inthe underlyinginterest rate reflecting the reduction incredit risk. Makingit clear, ifnecessarythrough regulation, that households inthe upper or middle income brackets are not eligible for government reconstruction funding (although they would continue to be eligible for relief). Tying catastrophe insurance into the landtax or landregistration systems. Sales o f catastrophe risk insurance policies to households and small businesses could be counted as partially contributing to the quotas specified by IRDA for the rural and social sectors, even ifthe risk concernedi s inan urbanarea andthe policyholder does not fall into the social category. Alternatively, specific requirements for catastrophe insurance penetration couldbe introduced. Increase catastrophe reinsurance capacity in India by pooling all domestic catastrophe business written by insurers. This would produce a more balanced portfolio and conceptually should increase local retention capacity. A precedent already exists in India with terrorism insurance and such arrangements exist in a number o f other developing markets (most recently Indonesia). However a precondition for this to work would be the upgrading and auditing o f underwriting standards within the established insurance sector and the accurate and complete collection o f accumulation data (see recommendations). A more formal catastrophe reserving system, basedbroadly on systems developed incountries suchas Mexico and Canada, could also be institutedto increasecapacity, andpotentially be supportedbyshort term tax incentives (inthe long runtaxes inthis context areonly atimingissue). Suchfacilities also lendthemselvesto contingent debtback up. AppendixIprovides acomplete descriptiono finternational practice inthis arena. A final innovation that could be tested is to allow a very limited and select number of microfinance institutions (MFIs) to distribute catastrophe insurance products, with lower minimum capital requirements than those currently imposed on the formal sector insurance intermediaries. This would be subject to very strict criteria regarding management skills, minimum size of established membership, target markets and reinsurance arrangements. While micro catastrophe insurance i s unlikely to be an attractive single purchase for most clients o f MFIs, the technology exists for it be added to credit and other products, possibly at the village and selfhelp level. Recommendations While the options outlined above will require consideration within the larger Indian fiscal and sectoral policy framework, the scope for further reform in the insurance sector to add capacity 73There is some question about the constitutional validity of any law that would require all households to purchase catastrophe insurance and some thought would be required as to how this constraint could be accommodated. 48 and increase the penetration o f disaster insurance i s relatively clear. For this reason we have characterized the relevant policy steps as recommendations. The insurancesector shouldbe further liberalizedby removingcurrentrestrictionson, and cross subsidies from, the householdand small businessinsurancemarkets. Inparticular, fire premium rates for households and small businesses should be completely liberalized over a relatively short time. While a standard policy wording should be maintained for market conduct purposes, insurance companies could be allowed to vary this wording through a derogation statement approved by the regulator and attached to the policy document. This will encourage contract innovation and introduce effective price competition. Prior to the complete liberalization of rates a modem claims experience database should be established, categorized according to relevant rating factors, and technically advanced rating methodologies should be introducedto the industry. The authorities could then keep overall control in the medium term byintroducing a file and write system.74Advisory catastropheprimarypremiumrates, based on technically sound assessment o f the relevant hazards, long term reinsurance pricing, and vulnerabilities and uncertainties could also be made available to the insurance industry as a socially desirable public good. Claims handling procedures in the event of natural disasters should be streamlined and formalized. In this regard the facilitating actions of the national insurers following the Bhuj earthquake should be encouraged rather than questioned, subject to adjusters and claims officers demonstrating adequate levels o fprofessionalism andpreparation. More explicit rules should be introducedas to insurers' minimumpremiumretentionsand maximum risk retentions. Inparticular, IRDA should begin to require all insurers to gather detailed aggregate catastrophe accumulation data and to monitor insurers' relevant exposures on at least an annual basis. 74Undera file andwrite system the insurer submitsrates to the supervisor but canbeginto use them ifthere is no response after a definedperiod, typically 60 days. 49 Annex I:India's DisasterHistory Table 17: Listof Largest Cyclones inIndia (1891-2000) Pradesh Godavari andKrishna districts (15N-17N). 6 Andhra 1946, November 6-11 Southeast ofNellore Deaths: 750 people and 30,000 cattle head. Damage Pradesh to property androads alsoreported. 7 Andhra 1949, October 21-30 Crossednear 10kmsoutheast Deaths: 800 people and 30,000 cattle head. Houses Pradesh ofNarasapuramnear destroyed: 0.25 million. Crops destroyed:over 1 Masulipatnam million acres. Heavy damages to hutsibuildingslplantations. 8 Andhra 1955, October 6-14 CrossedKalingapatnam Deaths: 500people and 100000cattle head. Heavy 66 Pradesh loss o fproperty. Communication completely T A n d h r a 1969, November 4-9 Crossedbetween Deaths: 900people. Kolletikota Islandhamlet of 174 Pradesh Masulipatnamand Kakinada Krishna district was completely submergedunder 8- 10 ft of water. Property damageof Rs. 200 crores. 10 Andhra 1977, November 14-20 Crossednorth of Chirala 60 Deaths: 10,000 people, 5,74204 cattle head/ other 259 Pradesh kmeast ofOngole animals, Populationaffected: 71 lakhs. Cropped areaaffected at acres: 36 lakhs. Houses damaged/destroyed: 10,10,336. Damageto public utilities: Rs. 11crores. 11 Andhra 1979, May 5-13 Crossednear Ongolebetween Populationaffected: 40 lakhs. Deaths:700 people, 202 Pradesh Nellore and Kavali 300,000 cattle heads. Loss ofproperty: Rs.170 crores. Crops destroyed:over 0.7 lakh acres. 12 Andhra 1984,November 9-14 Crossedsouth Andhra Pradesh Deaths:541 people. 84,000 peoplehomeless. 112 Pradesh coastjust northo f Sri Harikota Extensive damage to several installations at Rocket Launchingand Tracking Station at Sri Harikota. 91 metersMeteorological Tower broken. 13 Andhra 1987, October 14-19 Crossednorthof Ongole 17 deaths, substantialdamage. 67 Pradesh 14 Andhra 1987, October 31- CrossedNellore Deaths:50 people and 25,800 cattle head. 8400 72 Pradesh November 3 housesdamaged. Roadsand communication disninted Source: Various sources. 50 Table 17: Listof LargestCyclonesin India(1891-2000) (cont'd) Wind speed No. State Date Location Damage at landfall (kmph) 15 Andhra 1989, November 3-9 Crossednear Kavali Destruction inNellore and Kavali. Deaths69 213 Pradesh people. 55.5 la!& families homeless. 16 Andhra 1990,4-10 May Crossed 40 kmsouthwest of Deaths:967 people, 3.6 million livestock. 14,000 164 Pradesh Machilipatnam housesdamaged. Loss ofproperty Rs.2289.6 crores. 17 Andhra 1996, June 12-16 Crossednear Vishakhapatnam Deaths:68 people. Damageswere mainly due to 65 Pradesh breachof tanks andreservoirs, not due to wind and surges. Property/infrastructure loss: Estimated to be Rs.82 crores. 18 Andhra 1996, November4-7 Crossed Andhra Pradeshcoast Heavy damages caused to infrastructure, roads, 119 Pradesh nearly SO kms southof buildings, etc. 7 million families were affected. Kakinadainthe east Godavari Deaths: About 1,057 people. 925 people (mostly fishermen) missing. 1.74 la!& hectaresof crops .- .- 19 IGujarat 11964, June 9-13 ICrossednear Naliya IDeaths: 27 people. Extensive damage. 1161 20 IGujarat 11975, October 19-24 ICrossedSaurashtracoastnear IDeaths: 85 people. Severalthousandhouses . . 1185 Porbandar damaged. 21 Gujarat 1976, May 29-June 5 CrossedSaurashtracoastnear Deaths: 87 people, 4500 cattle head. Extensive 157 Gopinathpoint (between damage. Mahuvaand Bhavnagar) 22 Gujarat 1982, November 4-9 CrossednearVeraval Deaths: 542 people, 1,50,332 cattle head. Extensive 149 damagehouses andbuildings. 23 Gujarat 1996, June 17-20 Crossed southGujarat coast Deaths: 47 people. 30,000 housesdestroyed. 109 lcloseto Diu 24 IGujarat 11998, June 4-10 ICrossedGujarat coastnear IDeaths: 1,250 people, 11,700 animals. Total damage1165 Porbandar causedby the cyclone inGujarat alonewas estimatedto be Rs.1334 crores. The cyclone caused considerable damage in Rajasthanas well. The KandlaPort Area was the most severelyaffected area within the Kutchh District. About 2.57 lakhs houses were damaged. 25 Orissa 1909,October 24-27 Near Gopalpur Extensive damage. 26 Orissa 1959, September 27- Crossednorth of Balasorein Low lyingredons roundCalcutta heavily flooded 139 October 2 the night of30th September for two days- 27 Orissa 1971, October 26-30 Crossednear Paradip Deaths: about 10,000 people. 185 28 Orissa 1972, September 7-14 Crossednear Barua I I I Storm surge o fheight varying from 1-3 mabove 195 astronomical tide affected the coast from Chandbali tn Rania .__. - 29 Orissa 1972, September 20-25 Crossednear Gopalpur Inundation inPun district. 185 30 Orissa 1973, October 6-12 CrossedChandbali Deaths: 100people. 83 31 Onssa 1982, May 31-June 5 Crossednear Paradip Deaths: 245 people. Very heavy damagewas caused 134 all along from Paradip to Balasore 32 Orissa 1995,November 7-10 Crossednear Gopalpur Deaths: 96 people. 28,4253 hectareso f crops 104 33 Orissa 1999, October 15-19 Crossednear Berhampur Deaths: 205 people. 331000housesdamaged. 182 158,000 cropped area damaged. 5,181 villages were affected. 34 Orissa 1999, October 25-31 CrossedOrissacoast close to Deaths: 9,893 people, 444,531 livestock. The super 259 Paradipbetween Ersamaand cyclone affected 15 millionpeople andmore than 2 Balikuda(southwest of million householdsin the state. Paradip) 51 Table 18: LargeEarthquakes inIndia Source: Various sources. 52 Table 19: Summary of Major FloodLosses in India (1953-2001) Area Damage to Damage to State / Country Item affected Houses uti.ities Public Total Damage (M-Ha) (million $1 (million$) (million $) Andhra Pradesh Maharashtra I Maximum I 17.50 1 272.04 I 659.65 I 1215.96 Source: CWC. Morvi 7.0 6.4 0.91 Viramgam 6.0 6.1 1.01 Jamnagar 7.0 6.0 0.86 Rajkot 6.0 5.7 0.95 Gandhinagar 6.0 5.6 0.94 Ahmadabad City 6.0 5.5 0.91 53 Table 21: Validation of MMIswith 1993 Latur Earthquake Latur 6.0 5.0 0.83 Source: R M S Delhi. Table 22: Validation of wind speeds with 1977 Andhra Pradesh cyclone Ratio ~ Station Observed peakgust (mph) Modeledpeakgust (mph) Modeled/Observed Ongole 93.64 90.60 0.97 Mawlipamam 110.18 108.72 0.99 Gannavaram 120.78 116.54 0.96 Madras 29.58 31.09 1.05 Table 23: Validation of wind speeds with 1999 Orissa cyclone Source: R M S Delhi. 54 Annex 11: Loss Summary Tables The following tables are referred to inChapter111. Table 24: Average annualloss summary (US$ Million) Source: R M S Delhi. Table 25: Average AnnualLoss Summary (US$ Million) Source: R M S Delhi 55 Table26: ProbableMaximumLoss Summary(US$ Million) - infrastructure AP 1 AllPerils I 9211 7391 205 ! Cyclone 911 733 203 2.1% 2.6% 1.4% Flood 191 142 49 0.4% 0.5% 0.3% GJ All Perils 1.009 888 128 2.1% 3.1% 0.7% Cyclone 517 461 61 1.1% 1.6% 0.3% Earthquake 733 669 76 1.5% 2.3% 0.4% Flood 223 155 71 0.5% 0.5% 0.4% MR Earthquake 59 49 9 0.1% 0.1% 0.0% OR All Perils 479 288 177 3.2% 4.1% 2.2% Cyclone 477 290 177 3.2% 4.1% 2.2% Flood 130 63 67 0.9% 0.9% 0.8% 56 Annex 111: InsuranceConsumptionBy State Table27: InsurancePenetration% Assam 223.87 I 8.700 I 0.86 0.4 Source: CIRE (Indian Institute of Management), Report Commissioned by World Bank, 2002. 57 Annex IV: Central Relief Funds Flows To States Table 28: AnnualMarginMoney/CWAllocatedto the States by Successive Finance n Commissions(RSMillion) SL No. State 1 1 1 1 1 ~ Sixth ~Seventh Eighth Ninth' Tenth' Eleventh 1974-79 1979-84 1984-89 1990-95 1995-2000' 2000-2005* * Indicatesthe the Calamity ReliefFund. Indicates annual average ofthe five year devolution 1995-2000and 2000-2005. Source: DisasterManagementFacility with Consultants, June 2002. 58 Table 29: Releasesfrom National Fundfor Calamity Relief (Rs Million) States 1995-9 1996-9 1997-98 1998-9 1999-0 Tota Source: Disaster Management Facility with Consultants, June 2002. 59 Annex V: Andhra PradeshPostDisasterExperience Table 30: Andhra Pradesh Post Disaster Experience: 1996 cyclone 1996 Cyclone Capital Budget Damage as a Damage Damage % 19964997 %Capital Budget Relief Request %Damage Cn USD-Millions (lesshort&housing) CIS(') CIS griculture 397 1117 34 6% 12 3277.1% 50 13% anchayath Raj 150 42.3 146% 130 87% igation & CAD 100 28.2 9.7% 611 164% 80 80% lunicipal Administration 120 33.8 11.7% 100 83% nimalHusbandry 45 12.7 4.4% 30 67% isheries 40 11.3 3.9% 40 100% ledical and Health oads and Buildings 35 9.9 3.4% 196 17.9% 35 100% . P. TFANSCO-Electricity 102 28.9 10.0% 102 100% ther 37 3.6% 33 mediate Relief& Public Health 150 otal 6126 1,715.3 2143 350, otal lessHorticultureand housing 1,026 278.59 100.0% 1276 80.4% Distribution ex-gratia private )-1996-97 Budget is assumedto be awage of 1995/96and 96/97dueto energy budgetRuctuationsinthose two years IS.=crores Public Sector FundingGap Ratio 96% CIS. 10,000,000 rps 281,690 USDlcrs rps = ,02816 USD (35 rps= 1 USD) source: Governmentof AP 60 Table 31: Andhra Pradesh Post Disaster Experience :2001 Flooding 2001 Flooding Capital Budget Damage as a Damage Damage % 2001-2002 %Capital Budget Relief Request %Damage Crs USD-Millions less hortand housing Crs CIS 4griculture 68 14 5 13 6% 18 370.0% 20 29% 'anchayath Raj 60 12 8 11.9% 25 42% rrigation& CAD 72 15.4 14.4% 1170 6.2% 50 69% vlunicipalAdministration 41 8.6 8.1% 15 37% 4nimal Husbandry 3 0.6 0 6% 0% :ishenes 20 4.3 4 0% 5 25% dedicaland Health 54 11.5 10.8% 15 28% qoads and Buildings 160 341 31.8% 759 21.1% 50 31% 4. P. TRANSCO-Electricity 25 5.2 4.9% 22 1108% 25 103% rota1 925 196.7 278 30% rotal- lessHort and Housing 503 107.0 100.0% 3091 16.3% Reliefgrant-from Delhi Distribution public private PublicSector FundingGap Ratio 98% :IS.= crores I CIS. 10,000,000rps 212,766 USDlcrs I rps = ,0212USD (47 rps= 1 USD) Source: Government of Ap 61 Annex VI: BhujEarthquakeCapability/FundingGap Table 32: Sources and uses of funds (US$ million) PlannedExpenditurefor 2000 and Sourcesof Funds 2001 Item Amount Source I Tentative and IReceivedby end agreed 2002 - Housing 1,349 World Bank 996 105 Health 60 ADB 503 75 Education 179 CRF (both years) 75 38 Damsafety & 91 NCCF (both years) 314 207 infrastructure capacity Industry 128 Agriculture 86 Other 15 Total 2,4 15 ITotal 3,591 765 62 Annex VII: BriefOverview Of IndianMortgageMarket Inrecent years the MinistryofFinance andthe Reserve BankofIndiahave beentaking stepsto develop a consumer finance industry in India. At present the housing finance industry is estimated to be disbursing approximately US$5 billion annually and i s growing at 40% per annum, with expectations that this will continue for at least a decade. This growth estimate is supported by a leveling out o f real estate prices and declining interest rates that have increased affordability, and by the growing presence o f housing finance intermediaries. In addition, the central government has provided a direct tax rebate on housing loans to individualhouseholds. A number of institutional features continue to inhibit development of housing finance, not the least o f which are penal stamp duty rates in some states and the varying quality o f land record keeping. These inefficienciesare now beinggradually addressed. The longest established direct lender i s HDFC, with approximately 46% of the market. However, its influence has declined as other lenders have entered the market, including LIC, the nationalized banks, ICICI and a large number o f smaller housing finance companies (HFCs), though many o f the latter are expected to revert to purely agency roles. Refinancing is provided through a range o f government sponsored organizations, with the largest, the national Housing bank (NHB) also acting as regulator. Commercial banks are now required to earmark 3% of their incremental deposits, or approximately US$1billionannually for the housingsector. In 2001 HDFC financed 1.9 million houses. A crude scaling up points to a 2-3% annual increment to the housing stock through mortgage financing. Ifmortgages granted inthe last four years are added, this points to an initial potential catastrophe insurance market o f at least 5% o f the insurable housingstock. Average loans vary between Rs 25,000 and Rs 90,000 dependingon the institution and market segment involved and approximately 75% o f loans are made to individual borrowers, with 50% being inurbanareas. While demand remains strongest inthe area around Mumbai, it i s growing rapidly inother parts o f India, the tribal areas excepted. 63 Annex VIII: US ConsumersUnionPerspectiveOnNaturalDisaster Insurance Principles Congress should not enact any legislationthat provides relief to the insurance industryunless the legislationmeets the following principles to ensure that it also benefits consumers andtaxpayers. Adequate InsuranceProtectionat AffordableRates Any proposal must ensure that adequate insurance be available at affordable rates to all consumers, especially inhigh-riskareas. Low and moderate income homeowners should be protected from loss o f insurance coverage. Deductibles, co-insurance and surcharges may all be ways to ensure that insurance i s available but should not beusedto rendercoverage levels meaningless. Strong MitigationMeasuresto Reducethe Costs of Disasters Any proposal must have as its focus mitigation and must provide for effective measures to reduce losses. All stakeholders must be included in mitigation efforts - central, state and local governments, businessesand consumers, and, most importantly, the insurance industry. The proposal shouldpromote buildingandrelocationefforts away from high-riskareas. The proposal must include measures to assist homeowners, especially low-income, in implementingdamage-reductionmeasures. Retentionof Riskinthe PrivateMarket Any program musthave as its goal retaining as much of the risk inthe private market as possible, taking into consideration the capacity o f the market and the type o f risk involved. The property/casualty insurance industryhas over $300 billion in surplus, the excess of assets over liability. Hurricane Andrew, the most costly disaster, caused $15.5 billion in insured losses. Clearly, the industry has a great deal o f capacity that should be drawn uponbefore calling on the public to help. Minimizationof the Effects of Cross-Subsidizationto Help Ensurethat those in High-Risk Areas are the PrimaryPayers 0 Cross-subsidization o f risks should be limited to help ensure that those living inhighrisk areas pay their fair share for their protection. Pricing according to risk promotes building away from high risk areas, a key goal that should be apart o f any program. Inhighriskareas, the various catastropherisks couldbepooledtogether, e.g., earthquake and hurricane, to help minimize rate disparities among different areas and to capitalize on the pooling o frisks as much as possible. 64 AppropriateState andFederalOversight Federal oversight o f the insurance industryi s essential ifthe federal government provides financial backup to the industryor states. While the federal government must oversee the industry if it provides financial support, states must retainthe ability to provide the appropriate protections for their residents. DemonstratedBenefitsto the FederalGovernment'sDisasterReliefExpenditures The FederalEmergency Management Agency provides an average o fover $2billion each year in disaster recovery and relief (1989-1997 average). The federal government as a whole provides even more relief. Any proposal should help reduce those costs to the federal government and taxpayers and should have a reasonable plan to accomplish this goal. Questions to be Answered Before any proposal is enacted, Congress should have before it the necessary information to ascertain the extent o f the problem andthe effect o f any solutions proposed. For example, what i s the capacity o f the insurance and reinsurancemarkets today? What i s the relationship between federal disaster aid and private insurance -- does disaster insurance decrease the costs o f federal disaster relief? What is the effect o f the various state actions on limiting losses o f private insurers? H o w best can insurers be involved in the mitigation efforts to reduce costs? What are the costs o f the various proposals to the federal treasury? to taxpayers? to consumers? to states? to the industry? What type o f coverage i s adequate to meet consumers' needs indisaster-prone areas? 65 Appendix I:InternationalExperienceWith Catastrophe Funds Overview - Even if the basic conditions for the mergence o f an insurance market exist (see Appendix II), are two rationales for govemment intervention in catastrophe insurance there markets. The first emphasizesthe highcost andlimitedsupplyofprivate capital.75 According to the proponents o f this view, a shortage o f risk-bearingcapital leads to an inadequate supply o f insurance capacity, which keeps prices high relative to projected losses for low frequency high severity events, which i s in turn socially sub-optimal. In 1994, for instance, catastrophe reinsurance premiums were more than seven times the expected loss although that multiple has dropped to between four and five more recently.76 Proponents o f this view also contend that government, with its vast capacity to tax and borrow, has an advantage over private insurers in bearing catastrophe risk because it does not need to hold explicit capital to pay off claims and avoid bankruptcy.77 To free insurers from the costly burden o f holding huge amounts o f capital, proponents suggest that the govemment act as a residual provider o f reinsurance for so-called mega-catastrophes. The govemment could set premiums below those charged by private insurers, thus lowering the cost o f insurance while protectingtaxpayers from losses. The second view emphasizes that the biggest barrier to an adequate supply o f insurance, especially immediately after a catastrophe, i s insurers' heightened uncertainty about the frequency and size o f future losses. After Hurricane Andrew, the Northridge earthquake, and the World Trade Center attacks, insurers were not certain that they could assess the risks they were beingasked to assume. Without such knowledge, they were unwilling to commit capital by underwritingthe coverage. Intime, insurers are usually able to recalibrate their estimates and reenter the market. Thus, proponents o f this view contend that the govemment needs to intervene to supply insurance while insurers reassessrisk after a disaster, but they argue for a temporary govemment Actual experience has been that mounting uninsured losses from natural disasters have pressed governments in disaster prone countries and regions to look for practical solutions for catastrophe risk management, spurring the formation o f national and regional catastrophe insuranceprograms. To date, 12 national catastrophe risk management programs have been established and are operating successfully in 10 different countries, with the sole purpose o f providing affordable catastrophe insurance coverage for homeowners. While design and coverage features provided by these insurance programs vary, the underlyingrationale for their introduction has been the same - to address the challenges faced by the private insurance markets in insuring the risk o f natural disasters. Table A 1.1 lists the most well known o f these programs, which include TCIP inTurkey, FONDENinMexico, the FHCFinFlorida, the HHRFinHawaii, CEA inCalifornia, EQC in New Zealand, NatCat in France, and Norway's Norsk Naturskadepool. The two most recent o f these -TCP and the Taiwan Pool (Box A 1.1), have been established in the last four 75For example, see D.M. Cutler and R.Zeckhauser (1999). 76Premiums for the highest layers o f coverage (the lowest probability layers) were between 20 and 30 times expected losses in 1994, according to estimates. K.Froot (2001) and Figures 3 and 4. However, research emphasizes the imprecision o f the estimates o f actuarial losses for the least likely events. See J. Moore (1999), available at http://fic.wharton.upenn.eddfic/. 77Statement o f Lawrence H.Summers, Deputy Secretary, Department o fthe Treasury, before the House Banking and Financial Services Committee, April 23, 1998. 78Proposalon Federal Reinsurance for Disasters, Congressional Budget Office, September 2002. 66 years, while it appears that N e w Zealand is winding down its scheme, reflecting the maturity and depth o f market solutions now available inthat country. Name of the Fund Country Year Established & RiskCovered Turkey Catastrophe Insurance Pool Turkey 20001Earthquake (TCIP) Catastrophe Naturelles (CatNat) France 19821All NaturalDisaster except for Windstorm, ice and snow Japanese Earthquake Reinsurance Japan 1966/ Earthquake, tsunami, and volcanic Company (JER) damage Earthquake Commission (EQC) New Zealand 1994/ Earthquake, tsunami, volcanic damage, landslide. Norsk Naturskadepool Norway 1980/ Floods, storms, earthquakes, avalanches, tidal waves Consorcio de Compensaci6n de Seguros I Spain II1954/ Earthquakes, tidal waves, floods, volcanic eruptions, and cyclonic storms. Taiwan ResidentialEarthquake Taiwan 20021Earthquake Insurance Pool (TREIP) Florida Hurricane Catastrophe Fund USA 1993/ Windstorm during a hurricane (FHCF) Hawaii Hurricane ReliefFund(HHRF) USA 1993/ Windstorm during a hurricane California Earthquake Authority (CEA) USA 19961Earthquake Natural catastrophe risk is unique due to its highly systemic nature. Since 1989, there have been 15 natural disasters inthe United States alone, resulting inUS$ 43 billion of insured losses, and it is no longer unusual for the global insurance industry to sustain losses from a single catastrophic event in excess of US$ 1 billion. The management of these catastrophe risk exposures is highly capital intensive and it is hard if not impossible to diversify away these exposures at the level of primary insurers. Inthe aftermath o f natural disasters, private insurance markets have tended to ration or, in some cases, discontinue offering their catastrophe insurance coverage for homeowners or small business unless some sort of a risk sharing arrangement with the govemment is put inplace.79 ''Apart from catastrophe insurance programs presented inTable 1below, some countries have opted for public sector managed and financed disaster funds with the primary objective of providing ex post disaster assistance to a) low-income households, and b) to carry out immediate repairs (but not necessarily replacement) o f damaged infrastructure assets inthe wake of natural disasters. An overview o f these disaster relieffunds i s beyond the scope o fthis paper. 67 Box A 1.1-Public/PrivateCatastrophe Funding kishCatastropheI eve1earth quake coverageandthat Itis alsoplannedthatthe initial Design Features of Catastrophe Insurance Programs - A survey of the 12 major national programs reveals some major similarities8' Most programs (1) tend to focus on providing coverage against a specific natural hazard; (ii) tend to have a regional focus; (iii) mainly cover for dwellings and contents; (iii) have premium rates which tend to reflect the characteristics o f the risk, with an element o f solidarity involved which effectively provides for cross-subsidies from betterrisksto worse; (iv) as a rule, these programs receive no direct government subsidies; (v) mitigation is not typically a major focus, although some programs encourage retrofitting and safer construction practices by offering premiumdiscounts; and finally (vii) sales and servicing Guy Carpenter, World Catastrophe ReinsuranceMarket (2002). 68 are typically carried out through the established distribution networks o f private primary insurance companies and their agents. Table A 1.2 below provides a convenient overview o f key design choices available to policy makers and insurance practitioners involved in the creation o f national catastrophe insurance programs. A more detailed discussion o f these design options follows. Table A 1.2: Catastrophe Program Design Variables Management and Governance Less than a third o f catastrophe insurance programs mentioned in Table A 1.1 are managed by the government, with NatCat o f France and CEA o f California being the primary examples. Nevertheless, it should be noted that even inprivately runprograms, government influence and control remain strong through some form o f government representation on their Boards, which ultimately makes catastrophe insurance programs accountable to the public. In most cases an independent professional fundpool manager has been retained to carry out its day-to-day operations. Typically, the primary functions o f the fund manager include but are not limited to (a) collection o f premium, (b) claims management; (c) asset management and (d) placement o f reinsurance. Investment f i c t i o n s are carried out in accordance with the guidelines established bythe GoverningBoards o fDirectors. Insome cases, such as the FloridaHurricane Insurance Pool, catastropheprograms have their own direct distribution channels in addition to those o f participating private insurers. Most o f these entities tend rely heavily on the distribution and servicing capabilities o f primary insurers. For instance, inthe case o f the Turkish TCIP, the pool manager is the country's largest reinsurer, MilliRe, which markets earthquake coverage through the distribution channels o f the Turkish insurers. Incase o f claims adjustment, the TCIP relies on independent loss adjusters mobilized by insurance companies responsible for handling respective claims. The pool managers are typically compensated for their services with a management fee which varies widely - from 0.8 percent o f the net written premium inthe case o f FHCF to 2 percent inthe case o f JER. Insome cases the management fee is contingent upon achieving certain performance benchmarks such as a certain level o f insurancepenetration for the pool's major business line. Besides direct involvement in the operations o f a pool or through representation on the Board, government has another important role to play, to be a reinsurer o f last resort. Inthe case o f NatCat (Figure A 1.l), for instance, the French government provides a sovereign guaranty to the 69 state-owned reinsurer CCR for all claims in excess o f its claims paying capacity. In New Zealand, for EQC, the government guarantees to its policyholders that it would assume the financial responsibility for meeting the EQC's residual claims that are over and above its claims paying capacity. FigureA 1.1: FrenchNatCatSystem I FrenchGov't ........................................................................................................................ Unlimited Private Reinsurers Guarantee ........................................................................................... .......... Br- .. CCR" i *%. ' P & C Premium Extra-ChargeCatNat Decidedbythe government I PropertyandCasualtyContracts I @ CCR -Caisse Centrale de Reassurance, the public reinsurer. The French government offers the CCR a non- limited guarantee, meaning that the government i s the reinsurer of last resort. Source: The Public Private Sector Risk-Sharing inthe French "Cat. Nat System" by Marcellis-Warin and Michel- Kerjan, November 2001. The primary sources o f hnding for catastrophe pools are: insurance premium from the homeowners joining the system; reinsurance premium -in cases when pools act as reinsurers themselves, reinsurance coverage from their own reinsurers; pool's own surplus capital; assessments on private insurance companies; commercial backstop facilities, contingent credit lines, and direct government contributions in excess o fprograms' claims paying capacity. Funding Inaddition, over the last few years, some catastrophe insurance pools, such as CEA, have also obtained access to international capital markets by issuingcatastrophe insurance bonds. FigureA 1.2 below provides an example o f a mixed structure o f funding for catastrophe risk for JEE. 70 FigureA 1.2: JapaneseEarthquakeReinsuranceProgram 638-6 1.2373`4 4,500 Liability of J.E.R.;(Al) + (A2) +(A3) =Y 75,000m. +Y 219,40Om.+ Y 85,630m Liability of private ins. and The Toa Re.; (Bl)+(B2) =Y 281,800m. +Y 85,500m TI Liability of Government; (Cl) + (C2) =Y 501,200m. +Y 3,251,470m Insurance Vehicle In addressing the inherent underlying constraints o f the domestic private insurance market in case o f catastrophe insurance coverage, countries have opted for specialized direct catastrophe insurance or reinsurance vehicles. This choice inmany respects has been predeterminedby the development o f the local insurance market and its willingness to retain any catastrophic risk underwrittenby the program. A combination o f the two approaches is also possible, with the Florida Hurricane Fundbeingthe prime example. Currently, o f the twelve catastrophe programs listedinTable A 1.3, four programs (FHCF, CCR, JER, and Norsk Naturskadepool) are designed around the reinsurance concept and inthe remaining eight programs the government plays a very critical role by providing an "implicit" or an explicit guaranty to honor all claims against the pools, which inessence amounts to an excess o f loss reinsurance contract. TableA 1.3: InsuranceVehicles Fund Insurance Vehicle Characteristics insurers to-underwrite Cat Risks. FHCF/CCW Norsk Reinsurance Pool Reinsurance provided at bothbelow Naturskadepool market rate and withminimum volatility inreinsurance prices JER Insurance Company Riskspreading among insurance companies who are shareholders o f JER and also reinsurance capacity provided bythe Government o f Japan. Coverages While all the catastrophe insurance programs listed above offer coverage for buildings and usually contents, only one third covered the risk o f business interruption. Several o f the surveyed insurance programs also included emergency living expenses in the immediate aftermath o f a disaster intheir coverage. 71 While all programs offer personal catastrophe risk coverage, only a few cover commercial risks. One o f the reasons behind such a strong focus o f these institutions on providing residential coverage i s their explicit social commitment to ensuring that adequate catastrophe insurance coverage exists for the population. Besides, commercial/ industrial risks as a rule are well covered even in the least developed markets and thus are rarely a subject o f a public policy concern. Nevertheless, there i s certainly scope for extending catastrophe insurance coverage providedby catastrophepools to SMEs, which often are underinsured. Rates As the primary objective o f most catastrophe insurance programs is to ensure the availability o f affordable insurance coverage for homeowners, their premium rates for the worst risks tendto be capped at some level. Some programs, such as the HHRF and Norsk Naturskadepool, charge a flat rate irrespective o f location or construction quality o f covered properties; this o f course takes the "solidarity" principle to the extreme and offers no mitigation incentives. While the advantage o f having the flat rate i s its administrative simplicity, the majority o f programs charge variable rates that depend on a property's risk zone and the type o f construction. All inall, about a half o f the programs had risk based premiums and none are subsidized. On average, all programs appear to collect enough premium to cover claims and expenses. Table A 1.4: Rates Charged and Mitigation Incentives level changes. HHRF 1IUS$1.50 per US$l,OOO. IRate credits available for roof-wall and roof foundation clips and storm shutters. CEA Ranges from 1.1% and capped at Depending on its date o f construction, a house 5.25%. that has been retrofitted may be entitled to a 5% premium discount. Voluntaryvs. compulsory Most o f -the programs are voluntary, with only three being compulsory or semi-compulsory (TCIP, FHCF, and JER). Inthe case o f compulsory programs, compliance is generally low, with around 20 percent o f insurable housing stock covered in the case o f TCIP and JER. Yet, the 72 level o f insurance penetration achieved under the compulsory programs i s undoubtedly considerably higher than under the programs with voluntary participation. In the case o f India, for instance, where insurance coverage for natural disasters is optional, the insurance premium for natural disasters represents less than one percent o fthe total premiumwritten. Reinswance Catastrophic events are the greatest single threat to the solvency o f insurers. Rating agencies generally require that insurers have enough capital to pay for at least a 100-year loss event. To attain the top rating, insurers may need to maintain enough claims paying capacity for surviving a loss from a 250-year loss event. Reinsurance i s the traditional method used by insurers to boost their claims paying capacity, with capital markets becoming a growing source o f reinsurers' own capacity. A recent upward trend in reinsurance pricing (see Chart 2) has also spurred a series of reinsurance initiatives at various levels o f sub-national and national governments. Since the insurance premium charged to property owners is, to a greater or lesser extent, a fimction o f global reinsurance prices, some national governments are becoming more concerned with the availability and affordability o f such reinsurance coverages. While some programs such as FHCF, CEA, TCIP, HHRF, and Norsk Naturskadepool rely on private reinsurance markets for their reinsurance coverage, others are directly backed by their governments, as is the case with CCR. There are also cases when both private and government reinsurance capacities are used(JER). Another critical consideration in the design and management o f a reinsurance program i s the level of reinsurance to be purchased. This decision has an impact on the expected survivability o f a catastrophe insurance pool, on the speed at which it would accumulate its surplus and on the affordability o freinsurance or insurance coverage it provides. For instance, despite being among the safest insurance programs inthe world, CEA is one of the most expensive ones as well, as it has made a decision to maintain enough claims paying capacity for surviving a 1 in 800 year event. TCIP i s on the other end o f the spectrum as the least expensive catastrophe insurance program inthe world, buyingjust enough reinsurance to survive a 1in 170year event, which i s on the lower end o fthe investmentgrade scale for commercialinsurers. 73 Appendix 11: InsuranceMarket Economics Insurance i s an intangible and purchasers o f insurance are engaging in an act o f faith; they are giving up altemative current consumption to cover the small possibility that they will suffer a loss which i s large enough to significantly destabilize their or their dependents' fbture consumption pattern. The purchasers o f insurance must recognize and fear the potential for loss, not have attractive or easily accessible altemative means o f dealing with that loss, and have a perception that the loss has a not insignificant chance o f occurring. Inaddition, they must trust the,insurance company to still be inexistence when a claim occurs and to handle the claim fairly. Finally, the cost o f insurance should not involve a significant reduction in current consumption. For many these conditions do not exist and insurance is seen as a deadweight cost if no loss occurs. Inaddition, the marketmustbepreparedto providethe service at apricewhich is lessthanthe consumer's assessment o f the value o f removing the risk, if a market clearing equilibrium i s to exist. Another necessary condition i s that the relevant actuarial and socio/ legal infrastructure has to be inplace. A general model o f this framework has recently been developed by Vate and Dror (Figure A 2.1).*' FigureA 2.1: The Limits ofInsuranceMarkets The actuarial conditions are the best researched and require that the risk should appear to be random and thus not subject to the influence o f the insured, except possibly in a mitigating sense (for which the insuredwould ideally be rewarded). Inaddition, the insurer's aggregate retained risk should not have characteristics that invalidate the law o f largenumbersand the central limit theorem (for example, having a non-infinitesimal probability o f generating large losses relative *' Vate.And Dror (2003). 74 to premium income and capital) and should be definable and measurable to the satisfaction o f those pricing the contract. Most models in insurance economics assume that the consumer i s primarily motivated by a desire to reduce the chance o f lost hture consumption (including shelter) according to a concave utility h c t i o n and possibly a distorted assessmentof the probability o f loss. These models also assume that the insurance provider prices at a level to cover input costs, including allowances for the cost of the capital required,which inturnis assumedto be sufficient to reduce the probability o f insolvency to an acceptable level. These respective pricing algorithms may or may not lead to a market clearingprice (Box 2.1). Political determinants o f insurance consumption revolve around issues o f culture (including religion), property rights and rights o f redress, the definition o f public goods and the role o f the state and o f alternative risk management techniques. Until recently the state owned the major insurance and reinsurance activities inmany countries, and this tendedto restrict innovation and ultimately the energy applied to growing the insurance markets. While the actuarial and economic limits on the definition o f insurability have been gradually expanding in many industrial markets, insurance law i s often not only out o f date andhighlyrestrictive indeveloping markets but also tends to favor the insurer over the insuredinthe event o f a dispute. Inaddition to the factors already mentioned, it appears that humanbeings are not consistent in their assessment o f different types of risk, or over time, and tend to place different weights on severity and probability when determining their level o f risk aversion. A recent econometric study of subsidized flood insurance inthe United States indicated that the existence of a recent event i s an important determinant o f the willingness to buy.82 This is consistent with many similar studies o fmultiperil crop insurance. The role o fprice i s less obvious: non-life insurance appearsto have some o f the characteristics o f a normal good as opposed to life insurance, which i s clearly a luxury good.83 For example, the flood insurance study cited earlier indicates that the demand for flood insurance contracts is relatively insensitive to price changes, but that the amount o f coverage purchased is sensitive. Grace and others (2002) found evidence that the demand for catastrophe insurance has greater elasticity relative to price than normal householders coverages. These studies are mostly relevant to industrial societies, where insurance tends to go with credit creation and it could be argued that they are not applicable to poorer communities. However, studies and anecdotal evidence point to a strong desire to manage risk even amongst the poor and a willingness to sometimes pay heavily to use whatever mechanisms are available. Micro- insurance in particular i s a growing phenomenon and a number o f experiments on various continents appear to be showing some promise, although it i s still early days. Even the poor, however, demonstrate differing approaches according to the nature o f the risk concerned, with impact on earning ability and the perceived ability to control the risk being important considerations. For example, one study shows that in Cambodia farmers are more risk averse to loss o f health (and will thus buy stand alone insurance) than they are to loss o f livestock, the other major potential ~atastrophe.'~Other studies have shown examples o f market failure for the Browne and Hoyt (2000). 83 Lester and Galabova 2002. 84 Brownet al. (2000). 75 poor in industrial countries, which have in some cases led to government intervention in markets.85 Box A 2.1 -InsuranceMarket Economics wealthless thana unito separateriskaversion m is assumedto be indepe determined. The comb would be expected, the probability loss distrib institutionis prepared to This latter formulati The situation in developing countries has often been exacerbated by the way inwhich insurance markets have developed. Typically insurance is first consumed by the major industrial enterprises, often under pressure from international partners applying modem risk management techniques (airline hulls and liability are the classic example). Government and semi- government infrastructure sometimes follow (although with varying degrees of efficiency), and finally the inevitable growth o f motor car fleets usually leads to compulsory personal thirdparty liability insurance. Often personal business and particularly compulsory insurance is handled badly, with slow and sometimes corrupt claims handling, which creates the impression that it is a tax (at best) or an opportunity for graft by government employees and others (at worst). Either way, personal lines and small business insurance have inmany developing countries gotten offto a bad start incomparison with the development o f insurance markets inmost industrial countries over the last 200 years. 85See, for example, Peacock et al. (1997). 86The impact o funcertainty onreinsurance pricing canbe substantial - see Froot (1999). " SeeVateandDror,ibid.,page150,forthetheoreticalbasisofthisformulation. 76 TableA 2.1: InsuranceMarket DevelopmentPaths Industrial Developing 0 Friendlysociety, farmer mutuals -full 0 Compulsorymotor insurance trust from day one. 0 State insurers subject to non-market 0 Steady evolution-large mutuals, influences industrialinsurance, government 0 Poor claims payingrecord insurers 0 Seenastax 0 Demutualization, market conduct law, 0 Poor regulation, norecourse privatization Lossoftrust Giventhe limitations on dataplaguingthe insurance sector, the mostuseful aggregatemeasureof consumption at the country level is found by charting insurance consumption per capita against GDP per capita; these data have been recorded by Swiss Re for many years. If logarithmic charts are used, an immediate snap shot indication o f the elasticity o f insurance consumption relativeto economic growth i s produced (Chart A 2.1). This points to a global elasticity o f approximately 1.3 for non-life insurance (countries subject to Sharia law have been removed from the database because o f their particular and still evolving approach to insurance). Inother words a 1%increase inGDP per capita i s roughly matched by a 1.3% increase in premium spending per capita. Outliers on the low side include the higher income countries where strong social insurance systems are provided through state mechanisms (mainly the Scandinavian countries), and a number o f Asian countries, including India for reasons discussed earlier. Outliers on the high side tend to be industrial countries with strong and litigious liability environments or developing countries with long histories of private market development. ChartA 2.1: NonLifeInsuranceElasticityof Premiumper Capitavs. GDP per Capita 0 2 4 e I O 12 Ln GDP per Capita 77 Appendix 111: HazardAnd Vulnerability Models Hazards EarthquakeHazard Model A seismic risk assessmento ftwo states, Maharashtra and Gujarat, was conducted for this report, which involved the compilation o f an earthquake catalog, identification o f seismic sources, generation o f stochastic events and computation o f site-specific ground motion. Most o f the input data came from secondary sources such as published research undertaken within the country by its premier educational and research institutions. Inaddition, reputable international sources were used as necessary. As part o fthe calibration model andvalidation ofthe model, scenario analyses o fthe most recent catastrophic events inthe region were undertaken. The modeled and observed isoseismals were then compared. Historical earthquake catalog: The historical catalog compiled by RMSI serves as the basis for the earthquake model. The major source for this catalog i s the one published by ISET.88This catalog covers a perioddatingback from the history up to 1979. To meet the requirementso fthe present model, a new catalog was compiled taking ISET catalog as the starting point. The data beyond 1979 and up to the year 2001 were augmented usingother sources, including USGS and N O M . Study o f tectonics: To gather informed data on geology and fault system o f the area, the seismo- tectonics o fthe regions under consideration were reviewedusingexisting seismic zonation inthe publishedresearch papers, Indian codes and technical journals. The fault and geological data was obtained from the Seismotectonic Atlas o f India8'. The data from the atlas was processedto preparea detailed map o fthe active faults inthe region. Seismic sources: Seismic sources are geographical areas that have experienced seismic activity in the past and serve as potential sources of earthquakes in the future. Seismic sources are delineated based on tectonic or geophysical features and homogeneity o f seismic activity. For each seismic source, past earthquake activity was assumed to be a reliable predictor o f future activity. In a study carried out under GSHAP, eighty-six seismic sources were identified for developing the predictive model for India. The present model adopts the findings o f the study and considers only those sources falling within the boundaries o f the two states o f interest and also within a 200 km buffer outside the state boundaries. The selected sources and along with the maximum magnitude ineach source are shown inFigureA 3.1. 88ISET, "Catalogue of Earthquakes inIndia & Neighbourhood," Roorkee (1983). 89GSI, Seismotectonic Atlas o f India and Its Environs, Calcutta (2000). 78 FigureA 3.1: ModeledSources with MaximumMagnitudes The sources identified above are modeled by a series o f line sources o f uniform seismicity distributedevenly within the area source. The total seismicity o f the component line sources is equal to the seismicity o f the entire area source. Orientation o f the line source is done with respect to the main fault within the area source. The various events inthe catalog were assigned to the sources chosen for the analysis on a one-on-one basis. Small size events (called floating earthquakes) that could not be associatedwith any major source were assigned to "background" sources. Two background sources were delineated, each fully covering a separatestate. Earthquake rates of occurrence: Once the seismic sources were defined, it was assumed that future activity would be limited to those seismic sources and follow a pattem similar to past activity. The Poisson model i s the most common way o f representing the seismic activity o f an earthquake source. The basic assumption o f the Poisson model i s that the parameters governing earthquake occurrence are independent o f time, magnitude and space. Inother words, the model considers how often events occur on the average (average rate o f occurrence) and treats the probability o f future earthquakes as independent o f any previous earthquakes. The input required for this model is the average rate o f occurrence o f each magnitude o f interest. The average rate o f occurrence o f earthquakes i s commonly estimated using an exponential distribution for earthquake magnitude (the ratio o f the number o f small events to the number o f large events) expressedas a relationship betweenthe frequency andmagnitude o f earthquakes. This relationship, often described as the Gutenberg-Richter relationship, i s givenby the following equation: L o g N= a + PM 79 Where Nis the cumulative numbero f events greater thanmagnitude Manda and p are based on a regression analysis. For each source, the constants a and p o f the recurrence relationship are obtained by a regression analysis o f the historicalrecord o f earthquakes. Ground motion: The majority o f damage caused by earthquakes, especially to buildings, can be directly attributed to the effects o f ground shakinginducedby the passage o f seismic waves. The estimation o f the ground shaking expected at each location i s therefore findamental to the calculation o f the resultinglosses. Once the parameters o f each earthquake inthe stochastic set are defined, the intensityo f ground shaking i s calculated for each earthquake at each location o f exposure. The intensityo f an earthquake i s modeled from: 0 the attenuation of the ground shaking intensity, which depends on its magnitude, depth and earthquakemechanism; and the local modifications to the shakingthat are causedby the prevailing soil conditions. For a given earthquake, the attenuation, or rate o f decay, o fpeak ground acceleration (PGA) was estimated from the epicenter to the site o f interest based on the Joyner and Boore (1993-1995) attenuation equation. PGA to MMI conversion: Once the PGA had been obtained, it was converted to Modified Mercalli Intensity (MMI). The MMI i s a measure o f the local damage potential o f the earthquake. For the same PGA, distant earthquakes have longer duration and lower frequency content than nearby earthquakes and are therefore more damaging. Limited studies were performed to determine the correlation between structural damage and ground motion in the region. To convert PGA to MMI ,the present study employs Trifinac - Brady's relationship modifiedwhile calibratingwith recent events. Local soil correction: Local soil conditions can significantly impact earthquake ground motion and resultingstructural damage. Soil maps were procured from NBSS&LUP and processed to arrive at the soil classes and shear wave velocities within the region o fthe two states. The MMI at block centroid was thencorrected for the local soil effect. Validation: The PGA/ MMI values were computed for some o f the historical events at the centroids o f the blocks to calibrate and validate the hazard model. Comparisons between observed and modeled MMIs are given in Tables 1.4 and 1.5 in Annex Ifor the 2001 Bhuj earthquake inGujarat and 1993 Latur earthquake inMaharashtra. Cyclone Hazard Model - The model i s based on a stochastic module consisting o f thousands o f simulated events representative o f the characteristics o f the historic storms. The complex cyclone model comprises three separate, but related, sub-models: 1) a wind model, 2) a storm surge model and 3) a rainfall model. Each o f the three will produce a hazard that can be viewed separately from others. However, their combined effect i s a subject matter o f the vulnerability model. The following three states were considered for cyclone modeling - Andhra Pradesh, Orissa and Gujarat. Historical cvclone catalog: A historical cyclone catalog i s available for the period 1891 - 2000. However, information on central pressure, wind speed and bearing at every 6-hourly time-steps i s complete only for the period 1956 - 2000, a substantially shorter period than is ideally 80 desirable." The catalog was compiled by R M S I based on data and information published by IMD, NCDC, JTWC and other international sources. The compilation process involved sourcing, cleaning and filling the gaps by informedjudgment. Line pates: Depending on location of initial landfall of the historical storms and orientation of coast, coastal gateshegments were set up accordingly. These are line gates o f the same size, approximately 50 nauticalmiles (NMi)inlength, following the coast closely. Rates o f occurrence: The annual rates o f occurrence o f historical storms were calculated at each gate as a ratio o f total number o f storms to a time window o f historical data. The rates are smoothed at each gate to account for those gates where there was no storm inhistory. Cumulative distribution fknctions (CDF): Based on the landfall data given in the catalog, the probability distributions o f the cyclone parameters like central pressure, forward speed and track angle were then defined. For each o fthe parameters, the CDF was generated separately for each o f the states. These distributions were then sampled during the simulation process to generate stochastic events. Once the rates and distributions hadbeen finalized (referred to as targets), the CDFs were dividedinto bins for randomsampling. Stochastic events: From each o f the bins, using a uniform random sampling technique, an equal number of random samples were drawn for all the parameters and each sample variable was assigned an individual probability. Usingrandom numbers, each simulated or stochastic event was associatedwith a central pressure, forward velocity andtrack angle. The landfall location o f the event was assignedrandomly at a gate. Radius to maximum winds (Rmax) was assigned based on centralpressure derivedfrom a study by Bell (197) of Western Pacijic basin cyclones as this basin has similar characteristics to the North Indian Ocean. So, a stochastic event at landfall is finally deJined by central pressure, forward speed, track bearing, landfall latitude, landfall longitude and Rmax. Pattern matching: Each stochastic event was then matched with a historical event using a pattern recognition technique for its track and filling rate. The historical events were translated and rotated around the coastline to reflect the characteristics o f stochastic storms. The filling rate was verified by equations givenby Kaplan and DeMaria (1995). The number of storms required in the model was worked out experimentally to obtain the model 'sJit to the targets and loss convergence. Wind Model - The gradient wind field for stochastic events is defined by Georgiou's equation and surfacewind field is basedon calibration from keyhistoricalevents. Roughness: The land use and land cover (LULC) data for the three states were derived from high-resolution 25-meter remote-sensing data available with RMSI. Based on the land use, the roughness values were assigned with the help o f the classification given by Cook." The roughness data and assessment o f roughness change with direction were aggregated at block 90 The relevant statistical rule o f thumb i s that to estimate the returnperiod of an event, given a stable process, requires 5 to 10times the lengtho f the return period. 91 N.J.Cook(1986). 81 centroids (as per the methodology in Cook (1986), chap.9). A simple tool is written to implement the methodology, which entails an aggregation of 8 or 12 directional roughnesses over a circular spread o f 200 km. ToPowaPhv: Topographical features are not considered in the model as the terrains under consideration are nearly flat. Gust factor: The gust factor was determinedbased on the local turbulence (local roughness, per Cook methodology). Site Coefficient: The site coefficient i s calculated as multiplication o f roughness factor and gust factor. Historical storms reconstruction: The important historical cyclones are calibrated using the available stations' data. The gradient height wind speed was determined using Georgiou's equation (1985). Based on the available data o f surface wind speeds, the relationbetween the gradient wind speeds andthe surfacewind speedswas worked out usingregressionanalysis. Stochastic storms: A tool in Excel was developed to generate the windfield for the stochastic storms. The windfield o f cyclones was computed using a simple windfield model, which entailed the following three-step approach: A. The gradient wind speed is obtained from Georgiou's equation as givenabove. B. Surface wind speeds are determinedfrom gradient wind usingthe relationship arrivedat from the historical analysis. C. Surface wind speedwas converted to Peak gust usingfollowing relation: Peak gust =Vs Site Coeff * compute the storm surge height along the coasts o f three states - Andhra Pradesh, Orissa and Storm Surge Model - A nomogram-based surge model developed in India was adopted to Gujarat. For a landfalling storm, surge height was computed taking central pressure, Rmax, forward velocity and orientation o fthe track as the inputs. Methodolow: In the first step, a preliminary estimate o f the peak surge height Sp i s obtained from the nomogram o f peak surge for different pressure drops (DP) and radius o f maximum wind (Rmax) for a standard basin and standard storm motion (storms crossing normal to the coast). In the second step, correction factor F for the bathymetry was obtained from the nomogram o f shoaling factors. To correct for the effect o f a non-standard storm track, a factor FMwas obtained from the third nomogram of vector storm motion. The product of Spy F and FM obtained from the first, second and third nomograms respectively give the final corrected estimate o f the peak surge at a location. The mean astronomical tide i s added to the peak surge height to estimate the surge tide at coastal locations. The surge tide at the coastal location is attenuated with distance inland to estimate the surge tide at inland location usingthe attenuation fknction. The difference between the surge tide and the elevation o fthe location inland gives the flood depth. 82 Rainfall Model - Rainfall associated with a tropical cyclone i s dependent on its size, forward speed, direction and intensity. Rainfall i s directly related to a storms' size and inverselyrelated to its speed, i.e. slower moving storms yield more rainfall at a point than faster moving storms and large sized storms produce more rainfall than relatively small sized storms. For cyclones of IndianOcean origin, it has been observed that rainfall is more inthe left forward sector for the westerly moving cyclones, inthe forward sector for cyclones moving ina northerly direction and inthe rightforward sector for cyclones headinginanortheasterly andeasterly direction(Mandal, 1990).92 However, this i s a generalized picture o f rainfall distribution around a cyclone and the pattern can vary significantly from cyclone to cyclone. To overcome this rainfall variability from system to system and to obtain a general picture o f rainfall distribution around the cyclone o f different intensities, the compositing o f rainfall suggested by Frank, whose methodology for rainfall estimation i s based on a study o f 87 U S hurricanes inthe Atlantic.93 The keyparameters o f that model are (i) hourly precipitation rate; (ii) translational speed and, (iii) o f the size cyclone. The above parameters were considered to model the rainfall distribution o f tropical cyclones in the Indianregion as well. The hourly precipitation rates were computed considering a study o f rainfall distribution around tropical cyclones in the Indian seas by Jayanti & Sarma by compositing rainfall data.94 The study considered 270 pre monsoon and post monsoon cyclonic disturbances o f differentintensities. The study observed that for cyclones makinglandfall along the east coast, the maximumrainfall concentrated in a circular region o f 50 km radius with a significant rainfall region extending up to 200 km. Beyond 200 km and up to 500 km rainfall i s observed to be too insignificant to cause any damage or to contribute towards flooding. To keep things simple, significant rainfall region i s assumedup to 300 Kmfrom the center inthe present model. Jayanti's study has providedthe rainfall rate o ftropical cyclones inthree stagesbased on wind speed. All cyclones with wind speed greater than 47 knots were clubbed inone category. This drawback has been removed by considering the rain rate associated with different stages o f tropical cyclones o f higher intensity. For this purpose, Frank's studyhas beentaken to consider proportionate rate o f rainfall associated with high intensitytropical cyclones (Suffir Simpson's Cat 1+2 and Cat 3+4+5) over the Indianregion. Rainfall i s estimated at block centroid at hourly interval o f storm progress for the period the area is affected by significant rainfall zone (300 km annulus circle). To compute the total rainfall for a block, the rainfall associatedwith each time step i s finally integrated over the exposed area o f the block for the significant rainfall duration. Validation: The model i s validated against historical events wherever observed values are available for wind speed, rainfall and storm surge. (Wind model results o f peak gust wind speeds are compared in Tables 1.6 and 1.7 (in Annex I)for two famous historical events - the 1977 cyclone o f Andhra Pradesh and the 1999 cyclone o f Orissa, and indicate the problems in capturing the idiosyncratic nature o fthis hazard). Flood Hazard Model - As mentioned earlier, the scope o f flood analysis i s limited only to riverine floods, which cause most o f the flood damage. Flooding due to cyclonic storm surge along the coast i s modeled separately and inland flash floods are excluded. A comprehensive 92G.S. Mandal(l990). 93W.M. Frank (1977). 94N.JayanthiandA.K.S. Sarma(1987). 83 river flood model would include all components from rainfall to runoff to river flow to flood inundation. However, considering the nature o f the study and the constraints o f available data the modeling scope was limitedto river flow to flood inundation. Methodology: U.S. Army Corps o f Engineers' software package "Hydrologic Engineering Center - River Analysis System'' (HEC-RAS) (version 3.0.1) for floodplain mapping was used for analyzing the flood prone areas including flood-protected areas. The stepwiseprocedure i s outlined as follows: 1. Fit Gumbel's extreme value probability distribution to the historical annual peak discharges observed at a gauging station. 2. Annual peak discharges for different retum periods were calculated. The return periods taken are 5, 10, 50, 100, 500 and 1000 years. These are the stochastic events for the model. 3. HEC-RAS i s run for each o f the stochastic discharges to obtain water surface profiles along and across various reaches of the river network. Suitable assumptions were made with regard to Manning's roughness coefficient, distributiono f flood flow rates ineacho f the channels, initial conditions and boundary conditions to carry out the computational runs. 4. Using the block boundary map and D E W T I N as inputs, average depth over block and extent o f flooded areas were obtained by post-processing the HEC-RAS results in ArcView GIS software. Data Requirements: The data used in this project were classified into three types: hydraulic, hydrologic and spatial data. Hydraulic data: Steady 1-D flow models require at a minimum, three forms o fhydraulic data: 1) stream geometry, 2) streambed resistance factors, and 3) flowhtage boundary conditions. The river network is taken from the topographic maps and cross sections are extracted from TIN. Stream cross-sections along the network make up a significant portion o f the overall geometry data. Bed resistance factor is taken as Manning's n. A value o f 0.35 i s assumed for the main stream and 0.1 for the flood plains. Peak flow data at river gauging stations i s taken from the publications o fUNESCO and CWC to the extent available. Hydrologic data: Since the model addresses river flow to flood inundationprocess only there i s no hydrologic data was required. Spatial Data: Visualization o f floods in ArcView GIS required a detailed representation o f the terrain to accurately depict flood inundation. DEM(Digital Elevationmodel) or T I N (Triangular Irregular Network) can be used to develop the terrain model. T I N was used in this model for better representationo fthe terrain and was extracted from topographic data. Validation: A detailed validation at the hazard level was not undertaken due to lack o f detailed hazard data o f historical events. For example, data on flood depths at different locations and extent o f flooding are requiredto validate the results o f the model which i s not available even for one historical event. The model was only validated for flood depth at the gauging station given the discharge o f the historical event at that point. However, detailed validation was undertaken at the loss level based on the available extensive historical loss data. 84 Assets at Risk The exposed assets considered in this study fall broadly under public and - private domains, with the latter consisting only o f residential dwellings. As to the public assets, the following infrastructure elements were considered: 1. Educational institutions: schools and colleges 2. Medical facilities: hospitals and health centers 3. Roads and bridges Exposure was calculated interms o f replacement cost indollars at 2002 prices. The distribution o f exposure by block, by district and by asset class can b e obtained from the main study. A quick comparison of the exposures from all perils in the four states is given in Table A 3.1 and the same is illustrated inChart A 3.1. TableA 3.1: ExposureValue Summary (US$ Million) Source: RMSDelhi 85 ChartA 3.1: ExposureValue Summary (US$) Housing has the highest exposure followed by roads and bridges, education and medical. This trend is consistent across the four states, except inOrissa, as housing accounts for over 60% of the total exposure for the selected states. The low values o fhousingand infrastructure exposures in Orissa compared to other states can be explained by its smaller population, a lower level of economic development and, thus o f the exposed asset base. Maharashtra clearly stands out in terms o f exposedvalue for all asset classes. Consistent with worldwide experience, India has large concentrations o f population along the coast that are highly vulnerable to the risk o f a cyclone. The results o f the cyclone hazard model suggest that wind speed, rainfall and storm surge are maximum along the coast. While maximum wind speeds prevail inthe districts along the coast, storm surges dominate the coastal blocks. In Andhra Pradesh, 44% o f housing value lies inthe nine coastal districts from Nellore to Srikakulam, which are highlyexposed to cyclones. InOrissa, 27% o fhousing value lies insix coastal districts from Ganjamto Balasore. However, once the proximity o f the districts Gajapati, Khurda and Cuttack (a part) to the coastline and thus to cyclones i s taken into consideration, the value o f housing stock exposed to potential loss goes as high as 47% in Orissa. Finally, in the state o f Gujarat, two-thirds o fhousingassets are inthe coastal districts. Methodology: Assessment o f exposures at block level for buildings and other assets inIndia has been a challenge due to the lack of detailed primary data. As a result, the exposure values at block level were estimated either from available secondary data sources or derived from the distribution patterns o f population at a district and block levels. The methodology for quantifyingrisk exposures for the selected states involved extensive literature surveys and site visits to the states o f Orissa and Andhra Pradesh to carry out data collection and ground validation o f the model assumptions. The values for all types o f exposures were derived as the value o finventory times the average cost per unit. 86 The floor area and unit cost estimates were made based on information available in event reconnaissance reports, reconstruction reports and from the ground validation exercises. To estimate the cost per unit o f floor area, cost information from the public works department was used as well. The results o fthese calculations along with the floor area and the unit costs usedto calculate the Andhra Pradeshhousing exposure are tabulated below: Table A 3.2: HousingReplacementCost (Rs) Vulnerability To determine the degree o f loss to housing andinfrastructure resulting from exposure to a hazard o f a given severity, the study developed vulnerability functions covering the four states o f Andhra Pradesh, Gujarat, Maharashtra and Orissa. An outcome of this work is a set o f vulnerability hnctions for different hazards which show how structural damage varies with exposure to different levels o f hazard such as ground motion, wind speed or flood. This section provides abrief summary o fthis work. Methodolonv: Methodology adopted for the vulnerability modeling i s based on available loss/ inventory data complemented by engineering judgment and competent engineering and actuarial analyses. Development o f loss functions for buildings and other exposures inIndiaposes several challenges. The low availability o f sufficient loss data and presence o f large numbers o f non- engineered structures makes the task o f estimating vulnerability functions highlychallenging. Earthauake vulnerabilitv: The vulnerability relativities between different classes o f buildings were derivedbased on comparison o f performance duringpast events (mainly the 2001 Gujarat earthquake), seismic base shear coefficients, construction quality, etc. In general, for buildings, age and height parameters were omitted to simplify the vulnerability model. However, the building vulnerability functions were modified to account for high-rise apartment structures prevalent in major cities. In case o f roads and bridges, the vulnerability relativities between different classes o f roads and bridges were derived based on comparison o f performance during past events (mainly the 2001 Gujarat earthquake), construction quality and relevant engineering 87 studies such as HAZUS.'' The final curves for both residential construction and infrastructure were validated against the 2001 Gujarat earthquake loss data. Cyclonewind vulnerabilitv:The vulnerability relativitiesbetween different classesofbuildings were derived based on a component-based methodology. The relativities were further improved upon by incorporating information on performance o f structures duringpast events (mainly the 1999 Orissa Super Cyclone) and construction quality. In general, for buildings, age and height parameters were omitted due to non-availability o f detailed data. However the building vulnerability functions for high-rise buildings were modified to account for their prevalence in major cities. The final curves were validated against loss data for the 1977 and 1990 Andhra Pradesh cyclones and the 1999 Orissa super cyclone. Vulnerability fbnctions for RC and Brick buildings with different roof types and the respective general vulnerability functions are shown inChartA 3.2. Itwas assumedthat roads andbridgesareunaffectedbywinds duringacyclone. ChartA 3.2: GeneralBuildingVulnerabilityCurves - -+RCBuildingwithTileroof RC Buildingwith RC slab roof -General RC Building -e- Brick Buildingwith RCslab roof *Brick buildingwithTile roof -General Brick MasonryBuilding Peakgust(mph) Storm surge and rainfall vulnerability: To avoid double countingo f losses due to wind, surge and rain sub-perils, the cyclone model assumed that surge and rain perils affect only that part o f a structure left undamaged by the preceding winds. Also between surge and rain, surge affects a structure before rain. For loss validation purposes, the surge and rain losses due to cyclones were segregated from the overall loss figures by making reasonable and logical assumptions. The vulnerability relativities between different classes o f buildings were derived based on 95FEMA's StandardizedEarthquakeLoss Estimation Methodology. 88 information from engineeringstudies, performance during past events and construction quality. In general, for buildings, age and height parameters were omitted to simplify the vulnerability model. It was also assumed that rainfall-induced damage i s caused by post flooding only. Also both intensity and duration o f rainfall have been accounted for in the vulnerability model. The model assumptions used for infrastructure, wherever relevant, were similar to the buildings section described above. Flood vulnerability: The vulnerability curves based on depth o f flooding and vulnerability relativities between different classes o f buildings were derived based on comparison o f performance during past events and construction quality. Similar model assumptions for infrastructure were used. The final curves were validated against loss data for the 1986 Godavari (Andhra Pradesh) flood event. 89 Appendix IV: EleventhFinance Commission: Chapter ix -CalamityRelief 90 91 44 92 45 wiH be for deafing with calamities of rem severity and wlll be 9.20 Thesjzed thsfund wouldbe Re.700erores,tabebulR managedatIhenelianallevelbyasubzommmee oftheNatlonal up over the period 1995-2(300, with an initialcorpusaf Rs.200 Development Councll. This committee heeded by the Union cmreslowMchtheCentmwklcontribuIe Rs.150croresandthe Agriculture Ministercouldcomp&fi the Dy. Chairman, Planning States Rs.50 cmresin the propoltlan of 75:25.In addition, for Commission,andtwo UnionMinistersandfhle ChiefMiniaterato each of the fie years frwn 18&96 t~ 1999-2000 the be nominated by the Prime Minlsler annually by rotation. The conhibutionsdthe CentPeandtheStakswouldbeRs.75 mom Deparlmenlof Agkulture shouldprovidethe secretariatlorihls and Rs.25 croms respectlvety. The contribution by both tha fund The nominatimof the Chief MlniSters should be done In . CsntreandtheStateeswbuklbemadeannualtylnthetwaginnlngd Marchof eachyear for the nextfinancialyear. thefinandalyear. Contributionof 'Statestnter-sewouldbeInths 9.19 TheNatlonalFundfor CalamityRelief(NFCR),wltl be same proportion as their estlmeted total lax receipts attec operated by the Mhl&ry 01 Agricdture, Government af India devolution. The sham of each of the Slates, eo Indicated RI but it Will be maintained outside the Publlc Account of the Annexure IX.4, has been Included in the reassessment af Governmentof Indiaas recommendedbyusfor CRFsd Stales. expenditureof the ?tabs The Ministryof Financewill prescribe guidolis for this as we have recmended It should do In the cum af the CRF, The 921 Ws~lhatwnh~hesettlngupofthrNationalFundlor accounts 01 the NFCR shell be audlted annually by the CalamityRetlef H Would now be possibleto t W s calamitiesof rare severllymom effectively.what ie more, we hap4 that the expendlm,normsetc.forthiafund shouldbeworksdw1bythe Comptroller and Auditor aenerel. The admissible ltems 01 syslem recommendedby us would slgo helpcreateasense of Committeeof Expertswhchwe have"mendedabove fora nathal solldarlty in a common endeavour which woufd then slmilarpurposeinthe caseof CRFs. ablde beyondthe periodof dotress. 93 Appendix V: HedgingCatastropheRiskOfResidentialMortgage Loan Portfolios Overview: Mortgage lenders without any risk management program inplace on their portfolio o f loans are affected by losses arising from catastrophic events. We first demonstrate the potentially serious effects of a natural catastrophe on an unhedgedmortgage portfolio, including lower credit rating, loss o f equity, and insolvency o f the lender. We next discuss ways inwhich lenderscan minimize catastrophic lossesbypreparing for these events. Natural Disasters and Homeowner Insurance: Natural disasters can have devastating consequences including loss o f life, property, jobs, and businesses. The cost o f managing these events puts fiscal strains on government budgets, often leading to (or increasing) a budget deficit, particularly in smaller, developing countries. Homeowner property insurance policies do not generally include natural catastrophe coverage; additional coverage needs to be purchased for catastrophic events such as earthquake, flood, and cyclone. We use an example to reveal the potential scale o f natural catastrophe related losses to a mortgage lender, and demonstrate that lack o f catastrophe insurance can lead to homeowners' loss o f equity and have severe consequenceson a mortgage lender's portfolio. Consider a scenario with one mortgage lender, ABC Bank ("ABC"), which owns a portfolio of US$lOO million residential mortgage loans in a seismically active region. The makeup o f this portfolio consisted o f loans originating in the last 25 years. ABC did not have any risk management program inplace and assumed full portfolio risk, including that o f natural disasters. Because ABC did not require borrowers to take out earthquake insurance, borrowers did not purchase any. In addition, a disproportionate number o f properties (70%) in ABC's portfolio were located in the high risk areas o f the region. ABC's underwritingpolicy allowed for a maximum three month delinquency period for the lifetime o f the loan, after which ABC could proceed with foreclosure. No major catastrophes had emerged inthe region inthe last 25 years andABC's portfolio hadnot suffered any losses. A major earthquakewith highintensityhitsthe regioncausing extensive property damage. 2,000 borrowers who incurred losses less than 5% o funpaidprincipal balance (UPB), stayed current on their mortgage. However, anumberofborrowers out ofthe remaining3,000 became delinquent intheir mortgage payments; though after three months a few borrowers resumed payments, the rest defaulted on their loans. Duringthe three month delinquency periodABC lost bothprincipal payment andinterest income on a substantial numbero f loans. ABC portfolio pre-event average Loan-to-Value ratiog6 (``LTV',) was 80%, but the extensive damage to properties made the post-event L T V jump to 190%. The substantial number o f defaults on ABC's and other lenders' portfolios in the region increased the supply o f properties inthe market. As lenders triedto liquidatethe repossessedproperties, buyersbecame scarceand therealestatemarket dropped by 30%. ABC tried to raise funds to keep its business solvent, but there were no willing lenders in the market since ABC's credit rating had deteriorated and dropped significantly followhg defaults on its portfolio. Because o f its poor credit rating and lack o f sufficient funds, ABC could not 96Ratio of unpaid principal balance to property value. 94 restore the properties or wait for the market to stabilize. ABC had to sell the properties at rock bottom prices to recover some o fthe losses. As mentioned earlier, ABC portfolio had ahighconcentration o fproperties inthe highrisk areas o f the region. The heavy concentration o f portfolio in one region made the loans riskier than a geographically diversified portfolio, which would have spread the risk. The portfolio was a mixture o f loans with different maturities, anywhere from one month to 25 years. Table A 5.1 compares and contrasts the impact o f high, moderate and low risk concentrations on a portfolio o f mortgage loans. From Table A 5.1,one can observe that although an average L T V o f 80% i s reasonable for a portfolio o f mortgage loans, ABC would need a risk management program to sustain its business after a natural catastrophe. ABC should have determined its risk tolerance and established a policy where no more than a certain percentage o f loans were concentrated in one region. 95 TableA 5.1: Assumptions and Calculationsfor ABC Bank I f Post Event LTV 2120%, Borrower Defaults Land to property Ratio 60.0% Portfolio US$ 100.000.000 INumber of Loans I 5.0001 Average UPB $20,000 LandValue $15,000 LTV 80.0% PMLConcentrationFactor" 10.5% Post Event Default 30.0% Default Due to HighRisk 21.0% Gross Loss $12,600,000 ForeclosureRevenue $6,615,000 Net Loss $5,985,000 PMLConcentrationFactor" 7.0% Post Event Default 10.0% Default Due to Moderate Risk 2.0% Gross Loss $1,200,000 Foreclosure Revenue $630,000 Net Loss $570,000 Low Risk PMLConcentrationFactor" 3.5% Post Event Default 1.O% Default Due to L o w Risk 0.1% Gross Loss $60,000 Foreclosure Revenue $31,500 Net Loss $28,500 Total Net Loss $6,583,500 SoZvency Issues of ABC: ABC incurredUS$13.9 million o f gross losses, andwas able to recover US$7.3 million from landvalue o f foreclosed properties. ABC was unable to maintain its 8% or US$8 million regulatory capital requirement based on the Base1 Capital Accord. ABC was forced to raise fbnds to cover a shortfall o fUS$6.6 million inlosses, which proved to be difficult due to its poor credit rating as a consequence o f defaults. Despite earnings o f US$3 million, 96 giventhe regulatory capital requirement o fUS$8 million andnet losses o fUS$6.6 million, ABC became insolvent. Table A 5.2 illustrates that even with an average pre-eventL T V o f 50.5%, due to the post-event scarcity o f buyers and decrease in land value, average post-event LTV would have increased to 120%. ABC would have benefited from borrower paid earthquake insurance coverage and minimized losses. TableA 5.2: Pre-eventvs. Post-event LTV I fPostEvent LTV 2 120%, Borrower Defaults Average Land to property Ratio 60% Portfolio US$ 100,000,000 Number o f Loans 5,000 Average Property Value $39,604 Average UPB $20,000 Average LandValue $23,762 IAverage LTV IAverage % PropertyDepreciation II 50.5% I 30%( Average Post Event Property Value $16,634~ Average Post Event LTV 120% Preparationfor CatastrophicRisk and Loss Minimization: We now recommend methods to help lenders manage portfolio risk and minimize exposure to low frequency-high severity events. To ensure that layers o f risk above tolerance are transferred, to the extent available, lenders should either require borrower paid earthquake insurance coverage in their underwriting practice, or transfer this risk via insurance or capital markets. Lenders should conduct studies to have a better understanding o f the inherent risk in their portfolio. Freddie Mac, an agency investor in the secondary market, requires condo owners in California to purchase earthquake insurance based on the risk level o f the area and zip codes. The risk level i s calculated by applying the Earthquake Insurance Requirement Matrix prepared by Risk Management Solution, a risk modeling firm in California. Lenders/ investors who do not have this requirement in Californiamay endup with riskier condominium loans on their portfolio. RiskAnalysis: Lenders needto understand and analyze risk o ftheir portfolio. The key to any analysis i s accurate andup to date information on borrowers and onproperties (such as type, age, proximity to default lines, and current market value). LTV determines the value o f borrower equity, the difference betweenthe market value o f property and the unpaid balance (UPB). The market value o f a property updates LTV, which i s the main driver in the borrowers' decision whether to default. The higher the LTV, the likelier the borrower default. However, in some cases, borrowers with high LTV may not default due to psychological attachment to their properties and separation from home or community. Estimated post-event L T V enables lenders 97 to make a sound assessment o f potential losses. Information on borrowers allows lenders to conduct annual forecasting and sensitivityanalysis usingsimulation models.97 Modeling Risk and Methodology: After a catastrophic event the following chain o f events lead to lender losses: To assess risk o f a portfolio, simulation models subject it to stochastic analysis by randomly applying two million events, such as earthquake, to each property with different severity and frequency. The model calculates probability distribution for each event to estimate property damage and translates it into dollar losses. Post-event borrower equity, the measure o f equity left in the property, i s derived usingpost-event value o f property and UPB. The borrower may decide whether or not to default based on his equity. To assess the likelihood o f borrower default, historical data i s usedto develop a default algorithm usingpost-event LTV. 98 99 These default probabilities are then applied at the loan level to estimate losses, given the severity o f damage to property and the resultingpost-event LTV. Proceeds from foreclosure are estimated andnet loss to the lenderis calculated. The output o f the analysis serves as a guideline for lenders inmeasuring annual expected losses, probability o f loss exceedance and PMLs."' lo' These measures enable lenders to take corrective actions andselect a P M L consistent with their risk tolerance and objectives. Lenders account for loss o f income such as interest income and administrative cost due to foreclosures. Contributing factors such as the state o f economy and interest rate movements are taken into account, since these are additional factors that influence a borrower's decision whether to default. Home price and interest rate sensitivity analyses prepare lenders for any potential downturn in the housing market and allow them to adjust the market value o f properties accordingly for any depreciation or appreciation. For portfolios concentrated in one region, lenders should keep abreast o f economic conditions at boththe micro and macroeconomic levels. Lenders can either restore and sell the property, or sell the property at its post-event condition. After a catastrophic event, ina depressedmarket and with a scarcity o fbuyers, lenders may have to write off a significant portion o f their portfolio due to property depreciation. For defaults following a major catastrophe, costhenefit analysis should beperformed to examine the viability o f restoring properties; in some circumstances a lender may not be able to afford restoration costs and has to recover losses based on land value alone. In situations where borrowers cannot 97 The lender can either develop the model internally or hire a modeling fmspecializing incatastrophe analysis to conduct the task. 98 Ifsuchdata isnotavailable atthe regionallevel, the lendercanuse datafromother regionswith same characteristics. 99 E.g., assumption for default algorithm o fABC: ifLTV >120%, borrower defaults. looAnnual expected losses =UPB x C (Frequency x Severity) / C Frequency. 101Probability o f losses exceeding certain loss thresholds. 98 makepayments for a short period of time, the lendermay consider other altematives to prevent default, such as a workout planto allow borrowers to delay payments or provide additional loan to the borrower for restoring the property. It should be noted that a mortgage portfolio is not automatically affected by damagesAosses to properties after a catastrophic event. For a mortgagee to suffer losses, the default should follow damage to property, and the mortgagor should be unable to make future payments. Having information on other assets and credit worthiness o f the borrower will enable the lender to make a better assessmentof the borrower's potential likelihood o f default. Mortgage default i s usually the last option for the borrower after exhausting all other assets. The risk curve in figure 5.1 shows probability of exceeding specific losses. Lenders use probability o f loss exceedanceto decide on the layers o frisk to be transferred. FigureA 5.1: RiskCurve 100 .-b - 80 E 2 60 n 40 s 20 0 Risk Management Alternatives: Risk modeling is a fundamental and effective tool to use when deciding on mechanisms to manage risk. Lenders should manage expected losses and finance or transfer risk above their tolerance, in accordance with their objectives. Risk tolerance o f abusiness usually depends on aportion o fearnings that shareholders are willing to lose. If losses are covered under borrower paid catastrophe coverage, the lender should consider indirect counter party risk, that is, the credit worthiness o f the insurer. Where borrower paid coverage i s not available, lenders may decide on an option that transfers their catastrophe risk andbest hedgestheir portfolio. Below are some altematives: Risk transfer: Lenders can purchase reinsurance or an option based insurance allowing them to transfer lossesto a thirdparty. Risk Financing: One altematives i s a contingent finance arrangement, which is essentially a put option allowing the insuredto issue debt at a pre-negotiated pricehate. This arrangement allows lender to raise funds to cover losses. Another altemative is finite risk, an arrangement with a thirdpartyto smooth lossesover aperiod o ftime 99 Risk SharingArrangements: Insome developed mortgage markets, mortgage lenders share risk of default with mortgage insurers. Default insurance or private mortgage insurance allows lenders to share risk o f default regardless o f the cause o f default. Conclusion: W e have observed that the portfolio o f mortgages is not immune to natural disasters. As illustrated by our example, a profitable mortgage lending business can become insolvent due to a natural disaster. France, Japan, N e w Zealand, United Kingdom and the states o f Hawaii, Florida and California have implemented programs to deal with catastrophic risk. Some countries such as the United States have the resources to provide government assistance for natural disasters. However, this may not be feasible for developing countries with limited resources. It may be prudent for countries with limited resources to require mandatory home buyer and/or catastrophic insurance to prevent a significant fiscal burden on the government and to keep their mortgage markets solvent. This policy has already been adopted in some Latin American countries. Borrowers, lenders and governments can all benefit from an insurance plan that reduces the risk o f default for all parties involved. 100 Bibliography Anand. Dr.H.,Ollssa Mission Report & Composite Plan o f Action, UnitedNations Disaster Management Team / UNIndia, 1999. (http://www.un.org,in/dmt/orissa/Reports.htm). Anderson, M.B., "Vulnerability to Disaster and Sustainable Development: A General Framework for Assessing Vulnerability." inM.Munasinghe and C. Clarke, eds., Disaster Prevention for Sustainable Development: Economic and Policy Issues. IBRD/ World Bank, 1995. pp.41-59. Andersen, T., et al., "Disaster Insurance and RiskManagement in South Asia: Contemporary Country RiskManagement Approaches." World Bank Technical Paper, 2002. Anderson, T. and Kalavakonda, V., "The Economic Consequences of Catastrophe Events: An exploratory study ofexposeddeveloping countries." World Bank, forthcoming. Arrow, K.J. and Lind, R. C., "Uncertainty and Evaluation o f Public Investment Decision." The American Economic Review, 1970,60: 364-378. ASCE, Standard 7 "MinimumDesigno f Loads for Buildingsand Other Structures." 1998. Benson, C., "The Economic Impacts o fNatural Disasters inFiji." OD1Working Paper 97. Overseas Development Institute, 1997. Benson, C., "The Economic Impacts o fNatural Disasters inPhilippines." OD1Working Paper 99, Overseas Development Institute, 1997. Benson, C., "The Economic Impacts o fNaturalDisasters inVietnam." OD1Working Paper 98, Overseas Development Institute, 1997. Bertrand, A. J.M., The Political Economy o f Large NaturalDisasters. Clarenden Press, 1994. Bhatt, Natural Disasters as National Shocks to Poor and Development. World Bank, 1999. Brown. W, Green, C., and Lindquist,G. "A Cautionary Note for Microfinance Institutions and Donors Considering DevelopingMicroInsurance Products." Microenterprise Best Practices Project, USAID, 2000. Browne, M.and Hop, R., "The Demand for FloodInsurance: Empirical Evidence." Journal o f RiskandUncertainty, 20:3;291-306,2000. Business Line, April 23,2001 (Financial Daily ownedby The HinduGroup o fpublications. Article taken from online document). Cook, N.J., The Designer's Guide to Wind Loading o fBuildingStructures. Butterworth- Heinemann, 1986. Cutler, D.M., and Zeckhauser, R., "The Anatomy o fHealth Insurance." NBERWorking Papers 7176, National Bureauo f Economic Research, 1999. Cutler, D.M. and Zeckhauser, R., "Reinsurance for Catastrophes and Cataclysms." inKennethA. Froot, ed., The Financingo f Catastrophe Risk (Chicago: University o f Chicago Press, 1999), pp. 233-269. Dheri,S. K.,"Managing Disasters inUrbanAreas." inSahni, et al., DisasterMitigation, Prentice- Hall, 2001. Doherty, N.,Integrated Risk Management. McGraw Hill,2000. Dunleavy, H.Jeanne, "Capital Considerations o f Insurers with Catastrophe Exposure." 1998 (http://www,casact,org/coneduc/specsem/98catast/dunleavey.ppt). FEMA's Standardized Earthquake Loss EstimationMethodology. Frank, W.M., "The Structure and Energeticso fthe Tropical Cyclone Part I: Storm Structure." Monthly Weather Review, vo1.105, pp. 1119-1135, 1977. Freeman, P. and Martin, L.,National Systems for Comprehensive Disaster Management: Financing Reconstruction. IDB, May 2002. Froot, K.,The Financingo f Catastrophe Risk.NBER, 1999, p.7. Froot, KennethA., "The Market for Catastrophe Risk: A Clinical Examination." Workmg Paper No. 7286 (Cambridge, Mass.: National Bureau o fEconomic Research, February 2001). 101 General InsuranceCorporation o f India (GIC), 29th Annual Report, 2000-2001. Giles, T. and Biggar, S., RiskAllocation, Burdenon the Central Budget and Fiscal Benefits. Charles River Associates, 2000 and 2001. Gibert R., "Doing More for Those Made Homeless byNatural Disasters." Disaster hsk Management Working Paper Series No. 1, World Bank, 2001. Grace. M.,Kleindorfer, P., andKlein, R., T h e Demand for Homeowners Insurancewith BundledCatastrophe Coverage." Paper Presentedat NationalBureau O fEconomic Research, InsuranceProject Workshop, January 28,2002. GSI, Seismotectonic Atlas o f India and Its Environs, Calcutta, 2000. HighPoweredCommittee onDisaster Management, "HPC Report at a Glance: Buildinga Culture o f Prevention." Department o f Agriculture and Cooperation, Government o f India, 2001. Hirshleifer, J., "Investment Decisions Under Uncertainty: Applications o f the State-Preference Approach." Quarterly Journal o f Economics, Nov. 1965. IFRC, World Disaster Report, 2001. Institute o f AdministrationNaturalDisasters Conference, Delhi,February 2002. InsuranceRegulatory and Development Authority (IRDA), FirstAnnual Report, 2000-2001. InsuranceRegulatory and Development Authority (IRDA), Second Annual Report, 2001-2002. ISET, "Catalogue o f Earthquakes inIndia & Neighbourhood." Roorkee, 1983. Jaffee, D.and Russell, T., "Catastrophe InsuranceWhen Capital i s Limited: A Comparison o f Public and Private Approaches." Paper prepared for Pacific RimInsurance Conference, September 1997. Jayanthi, N.and Sarma, A.K.S., "Rainfall Associated with Cyclonic Disturbances from Indian Seas duringPre and Post Monsoon Seasons.'' Pre-published scientific report no. 87/2, RMC, Madras, 1987. Kleindorfer, P., and Kunreuther, H., "Managing Catastrophe Risk." inRegulation, Volume 23, NO.4: 26-31. Lahiriet al., "Economic Consequences o f the Gujarat Earthquake." Gujarat Earthquake Assessment Mission, Asian Development Bank, 2001. Lavall, A., "The Impact o f Disasters on Development Gains: Clarity or Controversy." IDMDR Forum, July 1999. Lester, R.and Galabova, T., "Is Insurancea Luxury?" TheActuary, Institute o f Actuaries, 2002. Litan, R., "The Impacts o fNatural Disasters: A Framework for Loss Estimation." Committee on Assessing the Costs o fNaturalDisasters, National Research Council, National Academy Press, 1999. Major, J. A., "The UncertainNature o f Catastrophe Modeling." inNatural Disaster Management, Tudor Rose, London, UK, 1999. Mandal, G.S., "Various Aspects o fTropical Cyclones inthe NorthIndian Ocean." Ph.D. thesis, Jadavpur University, Calcutta, 1990. McCarten, W., "The Challenge of FiscalDiscipline inThe Indian States." inFiscal Decentralization and the Challenge o f HardBudget Constraints, Rodden, J. et al., MIT Press, 2003. Mechler, R., "Natural Disaster RiskManagement and FinancingDisaster Losses inDeveloping Countries." PhD.thesis, Universityo fKarlsruhe, 2002. Mishra, K.,"Monogram o f Cat-risk insurance inIndia retrospect andprospect." National InsuranceAcademy, 2002. 102 Moore, James F., "Tail Estimation and Catastrophe Security Pricing: Can We Tell What Target We Hit IfWe Are Shooting inthe Dark?"Workmg Paper No. 99-14 (Philadelphia: Wharton Financial Institutions Center, 1999), available at http://fic.wharton.upenn.edu/fic/. Outreville, J., "The Retention Capacity of Insurance Markets inDevelopingCountries." Handbook o f Insurance, Kluwer, 2000, p.749. Peacock et al., Hurricane Andrew, Ethnicity, Gender andthe Sociology of Disasters. Routledge, 1997. Proposal on Federal Reinsurance for Disasters, Congressional Budget Office, September 2002. Rao, C.H.G., Real EstatePractices, 1997. Report o fthe Expert Group on Natural Disaster Prevention, Preparedness & Mitigation having bearing on Housing and RelatedInfrastructure-Part I,1998. UNCTAD, "A Ten Country Analysis of Catastrophe Exposure, Insurance Sector and Country Financial Capacity To Bear Risk."UNCTAD Secretariat backgrounddocument, UNCTAD/SDD/INS/8, October 1995. U.S. Army Corps o f Engineers, Manual No. 1110-2-1420, "Flood-Runoff Analysis Proponent." 1997. Vate, M.and Dror, D.,"To Insure or Not to Insure?Reflections on the Limitso f Insurability." in Social Re-Insurance, L O and World Bank, 2003. 103 GLOSSARY Acquisition Costs All expenses directly related to acquiringinsurance or reinsurance accounts, - Le., commissions paid to agents, brokerage fees paid to brokers, and expenses associated with marketing, underwriting, contract insurance andpremium collection. Aggregate Excess of Loss Reinsurance - A form of excess of loss reinsurance which indemnifies the ceding company against the amount by which all o f the ceding company's losses incurred during a specific period (usually 12 months) exceed either (1) a predetermined dollar amount or (2) a percentage o f the company's subject premiums (loss ratio) for the specific period. This type o f contract i s also commonly referred to as "stop loss" reinsurance or "excess o f loss ratio" reinsurance. Aggregate exceedance probability (AEP) A measure o f the probability that one or more - occurrences will combine in a year to exceed the threshold. See also "occurrence exceedance probability." Annual rate of occurrence Average number of occurrences per year. This statistic is reported - inRiskLink@ EP analyses. Not to be confused with the term "probability," which refers to the probability o f at least one event occurring ina year. Attenuation The reduction in ground motion with distance from an earthquake. The ground - motions resulting from an earthquake decay as they travel away from the fault. An attenuation equation is used to estimate this decay based on the magnitude o f the earthquake as well as the distance and depth to the source. Block An administrative sub-division of a district, which inturn is a sub-division o f a state. A - block i s larger thana postcode. Boundary files - Geographical mapping files that represent administrative or other regions for purposes o f creating maps to visualize risk information. Building inventory database Database representing the distribution o f types o f structures - withinthe built environment, for a givenregion such as a state or a country. Capacity - The largest amount o f insurance or reinsurance available from a company or the market in general. Also used to refer to the maximum amount o f business (premium volume) which a company or the total market couldwrite based on financial strength. Catastrophe Reinsurance - A form o f excess o f loss reinsurance which, subject to a specific limit, indemnifies the ceding company in excess o f a specified retention with respect to an accumulation o f losses resulting from a catastrophic event or series o f events arising from one occurrence. Catastrophe contracts can also be written on an aggregate basis under which protection i s afforded for losses over a certain amount for each loss in excess o f an aggregate amount for all losses inall catastrophesoccurring during a period o ftime (usually one year). Cede - To transfer to a reinsurer all or part of the insurance or reinsurance risk written by a ceding company. 104 CedingCommission- Incalculating areinsurance premium, an amount allowedby the reinsurer for part or all o f a ceding company's acquisition and other overhead costs, including premium taxes. Itmay also include aprofit factor. See Overriding Commission. CedingCompany (Also Cedent, Reinsured, Reassured) - The insurer which cedes all or part o fthe insurance or reinsurance risk it has written to another insurerheinsurer. Central pressure The lowest instantaneous atmospheric pressure at the center of a storm or - depression. Central pressure difference - The difference between the atmospheric pressure (central pressure) at the center or eye o f the storm and the pressure at the periphery o fthe stom. Centroid A point latitude and longitude which is meant to represent the center of a defined - geographical area. Cession- The amount of insurance risk transferred to the reinsurer bythe ceding company. Commission - In reinsurance, the primary insurance company usually pays the reinsurer its proportion o f the gross premium it receives on a risk. The reinsurer then allows the company a ceding or direct commission allowance on such gross premium received that i s large enough to reimbursethe company for the commissionpaid to its agents, plus taxes and its overhead. The amount o f such allowance frequentlydetermines profit or loss to the reinsurer. Cyclone Area of low-atmosphere pressure with winds blowing around it, counterclockwise in - the Northem Hemisphere, clockwise in the Southem Hemisphere. See "tropical cyclone" and "extra-tropical cyclone." Damage Any economic loss or destruction caused by earthquakes, windstorms, and other - perils. Damage ratio The repair cost o f a location represented as a percentage o f the value at that - location. Depth of earthquake The distance from the hypocenter of an earthquake to the surface of the - earth. Also called the hypocentral depthor focal depth. Directloss-Stock losses including destruction o finfrastructure, productive capital andhousing. Duration A qualitative or quantitative description of the length o f time during which ground - motion at a site shows certain characteristics, such as perceptibility or large amplitudes. Earned Premium - (1) That part o f the premium applicable to the expired part of the policy period, including the short-rate premium on cancellation, the entire premium on the amount of loss paid under some contracts, and the entire premium on the contract on the expiration o f the policy. (2) That portion o f the reinsurance premium calculated on a monthly, quarterly or annual basis which i s to be retained by the reinsurer should their cession be canceled. (3) When a premiumis paid inadvance for a certain time, the company is said to "eam" the premiumas the time advances. For example, a policy written for three years and paid for in advance would be one-third "eamed" at the end ofthe first year. 105 Earthquake - A sudden or abrupt movement along a fault or other pre-existing zone of weakness inresponseto accumulated stresses. Earthquake magnitude A scale defined by scientists to quantify the "dimension'' o f an - earthquake. There are a number o f different magnitude scales including local magnitude (ML), surface wave magnitude (Ms), and body-wave magnitude (mb). Each scale measures how fast the ground moves at some distance from the earthquake for a specific frequency band. Because they do not look at the entire frequency range o f an event, the different magnitude scales will produce similar, but possibly different magnitudes. This difference becomes more pronounced for large events (>6.5). For this reason, it i s very important to note which magnitude scale has been quoted for a given earthquake. Seismologists have recently developed a new scale, moment magnitude (Mw),which is calculated from the total energy released by the earthquake. The media often reports magnitudes using the "open-ended'' Richter scale, developed for a specific type o f seismograph that i s no longer in use. Richter magnitudes usually refer to local magnitudes, but should be viewed with caution unless additional informationi s provided. Economic loss Total losses from a catastrophe that include direct and indirect losses, as well as - insured losses and those losses paid by all other sources (such as property owners and the public sector). Elements at risk Population, buildings, and civil engineering works, economic activities, - public services, utilities and infrastructure, etc., that are at risk ina given area. Epicenter The surface o f the earth directly above the hypocenter o f an earthquake, where the - hypocenter (or focus) i s the point at which the fracture o f the earth's crust begins, thus triggering anearthquake. Representedby latitude and longitudecoordinates for risk modelingpurposes. Equalization reserve - Long-term reserve set aside by the insurer or reinsurer in order to equalize operating results from certain catastrophe risks. Event loss table (ELT) In its basic form an event loss table contains columns o f event ID, - event loss and event rate of occurrence. In its expanded form columns for associated uncertainties o f loss and rate are also provided. Event set The set o f discrete events used inprobabilistic risk modeling to simulate a range of - possible outcomes. Exceedance probability (EP) - Also known as "exceeding probability" or "EP," it i s the probability o f exceeding specified loss threshholds. Inrisk analysis, this probability relationship i s commonly represented as a curve (the EP curve) that defines the probability o f various levels o fpotential loss for a defined structure or portfolio o f assets at risk o f loss from natural hazards. Exceedingprobability See "exceedance probability." - Excess of Loss Reinsurance - A form o f reinsurance which, subject to a specified limit, indemnifies the ceding company against the amount o f loss in excess o f a specified retention. It includes various types o f reinsurance, such as catastrophe reinsurance, per risk reinsurance, per occurrence reinsurance and aggregate excess of loss reinsurance. See also Non-Proportional Reinsurance. 106 Exposure The total value or replacement cost o f assets (such as structures) that i s at risk from a - loss-causing event such as a catastrophe. Exposure data Information describing the exposures, used as an input for risk modeling. For - insured property exposure, this information includes: geographic location (e.g., state, county, postal code), physical characteristics (e.g., occupancy type, construction class, year built, height of structure, building/contents/time element contributions), replacement cost value (building/contents/time element), and financial structure (limits, deductibles, % insured, insurance-to-value). Fault -Break in the earth's crust along which movement occurs or has occurred. Sudden movement along a fault produces earthquakes. Slow movement produces seismic creep. Filling Weakening o f a storm such as a tropical cyclone as it moves inland. - Flash flood Floodingwith arapidwater rise. - Flat Rate - (1) A fixed rate not subject to any subsequent adjustment; (2) A reinsurance premiumrate applicable to the entire premiumincome derivedby the ceding company from the business ceded to the reinsurer, as distinguished from a rate applicable to excess limits. Forward velocity -The speed at which the center o f a low-pressure system moves forward. Also known as translational velocity (Vt). This is not the rotational velocity o f the winds around the center o fthe low-pressure system. Gate - For modeling purposes, short sections along a hurricane-prone coastline or along some other geographic feature through which stochastic storms such as hurricanes can be simulated. Generally these are 50 milesections o f coastline. Geocoding The process o f associating an address (such as a street or postal address) with an - estimate o fthe latitude and longitude that represents the location on the ground. Gradient wind A calculated wind speedthat represents the velocity o f air movement at altitude - inresponseto the dynamic pressure gradient that isassociatedwith anextra-tropical cyclone. Ground Up (From the) - A phrase referring to reinsurance losses subject to the contract under consideration before the application o f the retention, but afier reduction because o f any other reinsurance which inures to the benefit o f the coverage being considered. Also sometimes used to describe lossesbefore reduction for inuringreinsurance. Hazard A conditionthat may create or increase the chance o f loss from a peril. - High resolution - Adjective referring to data that i s at a highly detailed level o f geographic definition. Historical storm Any storm such as a hurricane, typhoon, or extra tropical cyclone, that has - already occurred. Hypocenter The point on the fault where rupture i s initiated at the start o f an earthquake, also - known as the focus. 107 Indirect loss - Flow losses including loss o f government revenues, reduction in GDP growth and opportunity costs. Insurance System under which individuals, businesses, and other organizations or entities, in - exchange for payment o f a sum o f money (a premium), are guaranteed compensation for losses resultingfrom certain causes under specified conditions. Insured loss The portion o f total economic loss from a catastrophe that is paid by insurance - policies, including payments made by insurance carriers based on recoveries from reinsurance contracts or other financial guarantees. This excludes deductibles paid by the policy holder as well as losses that are not covered by insurance (such as losses above insurance limits or losses for perils that are not insured). Intensity A measure o f the physical strength o f a damage causing event such as an earthquake - or windstorm. Common scales for intensity include the MMI scale for earthquakes, the Saffir- Simpson scale for tropical cyclones, the F-intensity for tornadoes, and the H-intensityfor hail. Landfall location - The point at which the eye o f a tropical cyclone (hurricane, typhoon, cyclone) first crosses over land.Expressedinterms o flatitude and longitude coordinates. Landslide - Massive down slope movement o f soil and rock materials, often generated by earthquakes. Layer - A horizontal segment of the liability insured, e.g., the second $100,000 of a $500,000 liability is the first layer ifthe cedunt retains $100,000, but a higher layer if it retains a lesser amount. Lifeline - The utilities, highway systems, and other systems that are needed to support a population. - the US. include : Residential Lines - Single-FamilyDwelling, Renters, Condos, and Mobile Line of business (LOB) A name or code used to specify a particular policy form. Examples in Home; Commercial Lines - General Industrial, General Commercial, and Multi-Family Commercial. Liquefaction - The temporary transformation o f a solid soil into a semi-liquid state when vibrated. Liquefaction i s most likely to occur in young, water- saturated sediments, particularly those with large amounts o f sand. Local soil conditions The potential for ground motion amplification by the geologic materials - underlying a site. R M S classifies soils along a spectrum ranging from hard rock (least amplification) to soft soils such as bay mudor artificial landfill (most amplification). Location A place with a single buildingor structure. Where several buildings are next to each - other, eachwould be considered a separate location. Also see site. Loss The part o f the damage suffered by each party. For the insured, it is the deductible plus - any loss over the limit. For the insurer, it is generally the damage amount in excess o f the deductible, not exceeding the limit. For a reinsurer, it would be the reinsurer's portion o f the insurer's loss. 108 Loss Loading or "Multiplier" (Also Loss Conversion Factor) - A factor is applied to the anticipated losses (or loss cost) for an excess o f loss reinsurance agreement in order to develop the reinsurance premium (or rate.) This factor provides for the reinsurer's loss adjustment expense, overhead expense, and profit margin. Magnitude The measurement o f an earthquake's energy as determined by measurements from - seismographic records. There are a number of different magnitude scales that are used depending on how the seismic energy was measured, which usually yield values in the same range. See "earthquake magnitude". Maximumcredible earthquake Maximum credible earthquake is defined as the most severe - earthquake that is believed to be possible along a particular earthquake source or fault segment based on geological and seismographic evidence. Mean damage ratio (MDR) The amount o f damage, expressed as a percentage of the value, - that a typical building o f a specific class will incur for a given shaking intensity or wind peak gust. ModifiedMercalli Intensity (MMI) ModifiedMercalli Intensity is a subjective scale used to - describe the observed local shaking intensity and related effects of an earthquake. This scale ranges from I(barely felt) to XI1 (total destruction), with slight damage beginning at VI. In general, the MMIwill decrease with distance from the fault, except in regions with poor soils. Intensity is different from magnitude, which is a measure of earthquake "dimension" rather than effects. Modifier Any factor used to adjust the basic classification vulnerability attributes of a specific - risk. Natural hazard Any natural phenomenon that poses a hazard to society, the economy, or - financial assets. Examples include earthquakes, fires, windstorms, floods, extreme temperature, andother atmospheric phenomena. ObligatoryTreaty - A reinsurance contract under which business must be ceded in accordance with contract terms andmust be accepted bythe reinsurer. Occupancy - Categories of usage for a structure. Used as an input factor in estimating vulnerability to loss. Occurrence exceedance probability (OEP) - A measure of the probability that a single occurrence will exceed a certain threshold. See also "aggregate exceedance probability". One-minute wind speed The maximum averaged one-minute wind speed at 10 meters (30 - feet) above the ground. Used as one o f the criteria to rate storms on the Saffr-Simpson intensity scale. Orientation Orientation indicates the bearing of a fault relativeto due north. It is expressed as - a value between -90" (due west) and 90" (due east) relative to due North (0"). Peak ground acceleration (PGA) The maximum value of ground motion acceleration as - displayed o n an accelerogram. A measurement of the maximum pulse of ground shaking at a location. 109 Peak gust - The maximum 3-second sustained wind gust at 10 meters (30 feet) above the ground. Since the peak gust i s sustained for a relatively brief period o f time, it typically i s substantially higher than a 1-minutewind speed. Peril - The loss producing agent, such as a storm (humcane, tornado, other windstorm), earthquake, or flood. Pool (Also Association, Syndicate) - An organization of insurers or reinsurers through which pool membersunderwrite particular types o f risks with premiums, losses, and expenses shared in agreed amounts. Primary- Inreinsurance this term is applied to the nouns: insurer, insured, policy and insurance and means respectively: (1) the insurance company which initially originates the business, Le., the ceding company; (2) the policyholder insuredby the primary insurer; (3) the initial policy issuedby the primaryinsurer to the primary insured; (4) the insurance covered under the primary policy issued by the primary insurer to the primary insured (sometimes called "underlying insurance"). Probabilisticmodel A model that assesses the impact o f ahazardand assignsprobabilities to a - whole range o fpossible outcomes. Probability See annual rate ofoccurrence. - Probability of exceedence The probability that the actual loss level will exceed a particular - threshhold. Probability of non-exceedance The probability that the actual loss level will not exceed a - particular threshhold. Probable maximum loss (PML) - A general concept applied in the insurance industry for defining high loss scenarios that should be considered when underwritinginsurance risk. The exact probability or return period associated with a PML can vary based on the company's policies and objectives. Radiusto maximumwind (Rmax) A distance measured normal to the track of a storm to the - location where the winds experienced throughout the storm were highest. Rate- The percentage or factor applied to the ceding company's subject premiumto produce the reinsurance premium or the percent applied to the reinsurer's premium to produce the commission. Rate On Line - Same as payback, except that the price is quoted as a percentage o f the limit. Thus, a 20 percent rate on line would be equivalent to a five year payback. Regression Regression analysis is the study of the dependence of one variable (the dependent - variable), on one or more other variables (the explanatory variables), with a goal o f estimating and/or predicting the mean or average value o f the former interms o f the known or fixed values o fthe latter. Reinstatement- A provision inan excess o f a loss reinsurance contract, particularly catastrophe and clash covers, that provides for reinstatement o f a limit which i s reduced by the occurrence o f 110 a loss or losses. The number o f times that the limit can be reinstated varies, as does the cost o f the reinstatement. Reinsurance - The transaction whereby the assuming insurer, for a consideration, agrees to indemnify the ceding company against all, or a part, o f the loss which the latter may sustain underthe policy or policieswhich ithas issued. Reinsurance Premium - The consideration paid by a ceding company to a reinsurer for the coverage providedbythe reinsurer. Reinsurer- The insurer which assumes all or a part o f the insurance or reinsurance risk written by another insurer. Reserve - An amount which is set aside to provide for payment o f a future obligation. Retention- The amount of risk the ceding company keeps for its own account or the account of others. Retrocession - A reinsurance transaction whereby a reinsurer (the retrocedant) cedes all or part ofthe reinsurance risk it has assumedto another reinsurer (the retrocessionaire). Return period The expected length o f time between recurrences o f two events with similar - characteristics. The returnperiod can refer to hazard events such as hurricanes or earthquakes, or it can refer to specific levels o f loss (e.g. a $100million loss inthis territory has a returnperiod o f 50 years). Richter scale - The original magnitude scale developed by Charles Richter in 1935. Usually referred to as local magnitude, this scale i s still often usedby scientists for events less thanM7.0. The term is often misusedinthe media to refer to earthquakemagnitudes measured usingother scales. See "earthquake magnitude'' for more explanation o f earthquake measurement scales. Risk A measure o fpotential financial loss, commonly encompassing two factors: exposure or - elements at risk (amount o f value subjected to potential hazard), and specific risk (the expected degree o f loss due to a particular natural phenomenon). Also used more generally in insurance marketsto refer to a specific propertycoveredby aninsurance or reinsurance policy. Risk management Management of the varied risks to which a business firm or corporation - might be subject. It involves analyzing all exposures to gauge the likelihood o f loss and determining how to minimize losses by such means as insurance, self-insurance, reduction or elimination o frisk or the practice o f safety and security measures. Risk premium The portion o f the insurance rate or premium intended to pay for insuredloss - under the insurance policy, for the cost o f repairing or rebuildingthe damaged property. It does not include adjusting expenses, underwriting expenses, or profit, other contingencies and inflation, which insurers add to the loss cost to obtain a final rate. Risk models are often used to quantifyloss costs for insuredperils. Riverine- Geographical area covered by a river, as well as the area surrounding the river, that mightbeaffected byflooding and other water damage from the river. 111 Rupture length - The rupture lengthrepresents the total length o f a fault that shifts during an earthquake. While the hypocenter i s a point location, an earthquake is actually the result o f rupture across an area o f a fault. For large earthquakes this can result in movement continuing from the hypocenter to a considerable distance along the fault. Saffir-Simpson scale Scale commonly used to measure windstorm intensity. Uses a range o f 1 - to 5, with 5 being the most intense storms. Namedafter Herbert Saffr andRobert Simpson. Secondary characteristics - Characteristics o f a structure (other than the primary characteristics) that can be specified to differentiate vulnerability, such as year o f upgrade, soft story, setbacks and overhangs, torsion, and cladding. Secondary peril - Hazards that are an additional source o f potential loss, commonly associated with a primary hazard. Examples include storm surge that accompanies a hurricane, fires that accompany an earthquake, or flooding that accompanies a windstorm. Secondary uncertainty While primary uncertainty measures uncertainty inthe likelihood that - a particular event occurs, secondary uncertainty incorporates the distribution o f potential loss amounts for the event. Inother words, it recognizes that when an event occurs, there i s a range o f possible loss values. The inclusion o f secondary uncertainty produces smoother EP curves with longer tails; a longer tail on the curve indicates a positive probability that losses exceed a maximum event. Seismic source - A region or geologic feature considered to have the potential to generate earthquakes. Seismicity The occurrence o f earthquake activity. - Site Same as location. When defining exposure data, a site may represent multiplebuildingsin - close proximity that are o f similar construction, andhave a single deductible amount. Slosh - Sea, lake, and overland surge from hurricanes/windstorms. The U.S. National Meteorological Center's computer model for calculatinghow much surge a windstorm will cause at any place along a coast. Stochastic storm - A possible storm scenario created as part of a probabilistic model, the probability o fwhich has beenassignedusingprobability distributions from the historicalrecord. Storm surge - Quickly rising ocean water levels associated with windstorms, which can cause widespread flooding. Measured as the difference between the predicted astronomical tide and the actual height o f the tide when it arrives. This difference arises in response to the lower barometric pressure associatedwith tropical or extra-tropical cyclones, and the action o fthe wind inpilingupthe surface ofthewater. The amount ofsurge dependson a storm's strength, the path it is following, and the contours o f the ocean and bay bottoms as well as the land that will be flooded. Subduction zone - Areas along tectonic plate boundaries where one plate i s moving downward relative to the opposite plate. Also known as a Benioffzone. Surface friction - The slowing effect on wind speed caused by vegetation or structures above ground level. 112 Tail Commonly used to refer to the portion of the exceedance probability (EP) curve that - represents very low probability o f loss, but very highlevels o floss. Terrain The topography as represented by changes in elevation; terrain can have an effect on - many hazards, including localized wind speeds in storms and landslide susceptibility in earthquakes. Track The movement ofthe center of a low-pressure system suchas ahurricane. - Track angle The directioninwhich a stormtravels (theta). - Tropical cyclone A low-pressure system that develops in the tropics, inwhich the 1-minute - sustained surface wind has reached 74 miles per hour (119 km/hr) or greater. Called a "hurricane" in the Atlantic and eastern Pacific, a "typhoon" in the westem Pacific, and a "cyclone" inthe south Pacific and IndianOcean. Tropical storm - A low-pressure system that develops in the tropics, in which 1-minute sustained surface wind ranges from 39 to 73 mph(63 to 118 km/hr). Typhoon See "tropical cyclone." - Validation Processbywhich probabilistic models and assumptions are reviewed and compared - to empirical data (such as historically observed losses or insurance claims) to confirm that the model approach andassumptions generate reasonable estimates o fpotential loss. Vulnerability Degree of loss to a system or structure resultingfrom exposure to a hazard of a - given severity. Vulnerability curve A set of relationships that defines how structural damage varies with - exposure to differinglevels o fhazard (such as ground motion or wind speed). Wind speed The speedo fthe wind duringawindstorm. - Windfield The time-integrated pattem o f peak gust wind speeds experienced during the - passageo f a storm. Windstorm Generic term referring to low-pressure systems of various types that cause high - winds and resulting damages. These include tropical cyclones (hurricane, typhoon, cyclone), extra-tropical cyclones, tomados and other convective systems. WB179446 C:\Documents and Settings\WB179446\My DocumentsWinancing Rapid Onset Disastersin India Delivery Version 2 --part edited l08.doc - August 24,2003 12:12 PM 113