Estimation of Liabilities Due to Inactive Hazardous Waste Sites. by Raja Bhagavatula, Brian Brown, and Kevin Murphy

Transcription

1 Estimation of Liabiities Due to Inactive Hazardous Waste Sites by Raja Bhagavatua, Brian Brown, and Kevin Murphy

2 ESTIMATION OF LIABILITIES DUE TO INACTIVE HAZARDOUS WASTE SITJZS Abstract: The potentia iabiity associated with inactive hazardous waste sites can be arge for both poicyhoders and insurance companies. Our paper outines severa methods that can be used to estimate and monitor insurance company and/or poicyhoder iabiities associated with inactive hazardous waste sites. We have outined severa pubicy avaiabe data eements which can be hepfu in evauating environmenta iabiities. None of the procedures described in this paper provide the method to anayze environmenta iabiity exposures. For financia reporting purposes, company management needs to evauate the detais of its own exposures and judge the utimate cost based on current facts and financia reporting principes. Additionay, this paper summarizes the ega issues invoved in environmenta coverage disputes between insureds and insurance companies. For the past ten years issuers of CGL poicies and their poicyhoders have engaged in a protracted strugge to determine whether or not environmenta iabiities are entited to defense and indemnity under CGL poicies. This paper discusses major coverage issues such as what constitutes a suit, whether it resuts in damages, whether it was sudden and accidenta, etc., upon which the primary batte ines between insurers and insureds are drawn. Athough the ega andscape of environmenta insurance coverage is becoming cearer, many of these and other issues have not been decided in a number of jurisdictions. Biography: I I Raja Bhagavatua is a Feow of the Casuaty Actuaria Society, a Member of the American Academy of Actuaries and a Consuting Actuary in the New York office of Miiman & Robertson, Inc. Raja serves on the CAS Reserve Committee and has a Masters degree in Pure Mathematics from Osmania University, India. She has worked on consuting assigmnents invoving mergers and acquisitions, reserving, pricing and financia anaysis for insurers and sef insurers. Many of her assignments invove estimation of environmenta iabiities for insurance companies and insureds. Brian Brown is a Feow of the Casuaty Actuaria Society, a Member of the American Academy of Actuaries and a Consuting Actuary in the Miwaukee office of Miiman & Robertson, Inc. Brian serves on the CLRS committee and has a Bacheors degree in Economics from Iinois State University. Brian s area of expertise is property and casuaty insurance, especiay in ratemaking, oss reserve anaysis, and actuaria appraisas for mergers and acquisitions. Brian is an active member of the Miiman & Robertson research group on environmenta iabiities and has previousy authored three papers. 302

3 Kevin Murphy is a partner in the Chicago office of Latham & Watkins, He graduated from the aw schoo of the University of Chicago in He chairs the Chicago office s environmenta iabiity department and speciaizes in environmentaaw and itigation with a focus on hazardous waste sites. He has represented poicyhoders in compex environmenta insurance coverage itigation in numerous jurisdictions. Acknowedgements: The authors woud ike to acknowedge the encouragement, advice and review provided by Gary Josephson and Aan Kaufman of Miiman & Robertson, Inc. We woud aso ike to acknowedge Chap Cook and Ron Swanstrom of the CAS Reserve Committee for their comments on presentation. We woud ike to thank Chery Bamett, Pat Best, Marion Marsden, Tim McCarthy, and Brian Meas of M&R, who did much of the eg work in the preparation of this paper. 303

4 ESTIMATION OF LIABILITIES DUE TO INACTIVE HAZARDOUS WASTE SITES INTRODUCTION Property and casuaty insurance companies are under increasing pressure to set aside arge sums for cean-up costs and other damages associated with inactive hazardous waste sites. A significant portion of this potentia iabiity arises from commercia genera iabiity (CGL) poicies issued between ten and thirty years ago or more. The cean-up cost iabiities arise from the Comprehensive Environmenta Response, Compensation and Liabiity Act (CERCLA) passed in This act is commony known as Superfund and it provided a financia mechanism for funding the cean-up of inactive hazardous waste sites. This act was reauthorized and amended in 1986 by the Superfund Amendments and Reauthorization Act (SARA) and is again up for reauthorization in Transporters and generators of hazardous waste as we as owners of dump sites are potentiay responsibe parties (PRPs) for ceaning up waste sites. Superfund empoys the foowing ega bases:. Strict iabiity;. Joint and severa iabiity; and. Retroactive iabiity. The potentia iabiities that arise from Superfund coud be staggering for both insurance companies and PRPs. To put the potentia cost in perspective, the Environmenta Protection Agency (EPA) estimates that the cean-up costs for the approximatey 1,300 sites currenty on the nationa priorities ist (NPL) may be $30 biion to $40 biion. 304

5 This figure is expected to increase significanty as more of the 37,000 potentia sites are added to the NPL ist. Additionay, a University of Tennessee study estimates that environmenta cean-up costs coud exceed $1.O triion *. Attention from severa forces such as the Securities Exchange Commission (SEC), reguators and rating agencies regarding the reporting of environmenta iabiities has recenty increased due to the magnitude of the potentia iabiities. In 1992 the Genera Accounting Office (GAO) recommended that the SEC require insurers to discose in their annua reports the number and type of environmenta caims they have received and an estimated range or minimum amount of associated caims and expenses. The 10-K s of industria companies in genera state that their poution iabiities are covered by insurance, and therefore, have no effect on their bottom ines. However stock insurers often state that environmenta caims fied to date are not covered by the poicies in question and are ony posting modest amounts reative to the potentia exposure. Therefore, there is a concern that neither companies nor insurers are recording environmenta iabiities. In an attempt to improve this situation, the SEC issued Staff Accounting Buetin No. 92 in Juy 1993 requiring companies to discose iabiities both gross and net of anticipated insurance recoveries. The K s issued by industria companies and insurers may shed some ight on the insurance recoveries anticipated by insureds as compared to iabiities acknowedged by insurance companies. In the remainder of this paper, we wi:. Describe methods which can be used by insurance companies to anayze their David Foppert Pressure Mounts for Cean-up Reserving Best s Review, November 1993 *Hazardous Waste Remediation Project Study of the University of Tennessee, December

6 environmenta iabiities;. Outine pubicy avaiabe data that can hep actuaries and caim administrators in the evauation of environmenta iabiities;. Describe procedures that anaysts are ikey to appy based on pubic data as we as methods that management might want to incude as part of its overa evauation of a company s environmenta iabiities; and. Discuss insurance coverage issues (this ega anaysis is attached as Appendix A). Any reference to environmenta iabiities in the foowing sections shoud be interpreted as iabiities arising out of inactive hazardous waste sites. We acknowedge that other iabiities may be cassified as environmenta iabiities (e.g. oi spis); however, these categories are outside the scope of our paper. None of the procedures described in this paper provide the method to anayze environmenta iabiity exposures. For financia reporting purposes, company management needs to evauate the detais of its own exposures and judge the utimate cost based on current facts and financia reporting principes. Management shoud aso consider the provisions under the Superfund Reform Act of 1994 which are ikey to have a significant impact on these iabiities. Evauating Environmenta Liabiities Traditiona actuaria reserve projection techniques are not directy appicabe in evauating environmenta iabiity exposures for severa reasons. First, it is difficut to assign osses to an accident or poicy year. If a firm dumped at a particuar site between 1950 and 1990, the assignment of damages to years is uncertain. Second, insurance companies and insureds are invoved in extensive itigation with regard to coverage 306

7 issues. Finay, we ack historica data and there may be changes in the state and federa aws under which these caims may be utimatey resoved. We wi discuss a number of methods to project environmenta iabiities in this paper. Specificay we wi discuss the foowing methods which we beieve can be used to project environmenta iabiities: 1. A curve fitted to caendar year emergence; 2. A caendar year oss deveopment method; 3. An industry benchmark method; 4. A market share mode; and 5. An exposure mode. The first two methods are oss deveopment methods, the ony difference between the two methods being how the deveopment factors are derived. In method 1, we rey on a curvefit of the insurance company s interna data, whie in method 2, we anayze this data and an externa data source to seect deveopment factors. Method 3 provides benchmarks an individua company may use to compare itsef to peer companies and the industry. These benchmarks provide guidance on the reative eve of the company s reserves and payments as compared to the industry and peer companies. The benchmarks that are used for comparison incude: reserves as a mutipe of annua payments or annua incurred and indicated market share based on payments and incurred osses to date. Methods 4 and 5 are exposure-based methods. Method 4 requires an estimate of the iabiity for the U.S. insurance industry and assumes that an individua company s share is represented by its genera iabiity premium market share. Method 5 provides a systematic process of estimating these iabiities using insurer and EPA data. 307

8 LOSS DEVELOPMENT METHODS Methods 1 and 2 are oss deveopment methods. These methods treat the osses arising out of inactive waste sites as if they were due to one accident year and measure the deveopment of these osses in tota. As we mentioned previousy, it may be difficut or impossibe to assign individua environmenta caims to accident years. Aso, underying causes of deveopment are caendar year events which have the same effect on a od accident years regardess of accident year age. For exampe, in the case of cean-up costs for inactive waste sites, the underying cause of deveopment is the passing of CERCLA in The purpose of the two deveopment approaches is to use a methodoogy which is generay used for actuaria projections, unti such time as a company has sufficient data to utiize more refined approaches. The assumption underying the projections is that there is a reationship between environmenta osses reported and the utimate osses. The approaches differ with respect to the source of the deveopment factor, with one inferred from the patterns in the actua data, and the other derived from an externa - and presumaby sufficienty comparabe - source. CURVJ3 FITTING TO CALENDAR YEAR EMERGENCEMETHOD 1 I In expaining why we might want to rey on caendar year emergence, it may be usefu to outine what we wi ca the ife cyce of atent caims. This ife cyce can he broken down into the foowing segments: Event: Something happens to expose an individua/property to a hazardous agent (e.g. the initia dumping of waste into a site which does not immediatey resut in any property damage); 308

9 Emosure: Once the event occurs, the exposure to the hazardous agent takes pace often over a ong and undetermined period of time (e.g. chemicas from the site sowy enter the ground water system); EmerQence : The effects of the exposure are known (e.g., it becomes cear that the ground water system is pouted). In this stage caims are made or PRPs notified; and ~5% Payments are made to cean up sites as we as ega fees incurred to determine coverage issues. The attached Exhibit 1 dispays a graph for a hypothetica ife cyce for atent caims. Much of the activity that ed to waste site caims occurred between 1950 and This is the event stage. Stage two, the exposure stage, probaby overapped with the event stage but may have initiay agged the event stage by severa years (as the chemicas dumped did not immediatey eak from the site). The next stage, the emergence stage, probaby agged the exposure stage by severa years (especiay the emergence of the cean-up costs of inactive waste sites, which is governed argey by Superfund egisation). Superfund did not become aw unti Therefore, we woud expect the emergence curve to start ow but increase dramaticay after We woud expect the expenditure curve to ag the emergence curve by severa years and to increase ess dramaticay than the emergence curve due to the fact that severa coverage/iabiity issues are deaying actua payments. Additionay the expenditure curve wi be extended after the site is ceaned up because annua maintenance costs are significant and may he expected to continue for 30 or more years. Whie the expenditure curve ony refects payments in Exhibit 1, the expenditure horizon coud be separated into two steps: (1) Loss reserves estabished; and (2) Caim payments 309

10 made. The curve fitting to caendar year emergence method extrapoates the utimate caim costs based on fitting an S curve to the cumuative caendar year incurred osses. Exhibit 2 dispays cumuative incurred environmenta osses by accident year and caendar year for a hypothetica insurance company, ABC Insurance Company, based on the insurance company s assignment of osses to accident year. As the exhibit shows, the accident year osses do not dispay a norma deveopment pattern for a property/casuaty coverage as no payments or case reserves were estabished prior to year end 1989 for accident years 1970 through However, it appears that the caendar year cumuative osses, in tota, may be extrapoated based on an S curve. Exhibit 3 dispays the actua and fitted points and the estimated curve. The footnotes on Exhibit 3 eaborate on the mathematica form of the curve. (However, it shoud be noted that there is considerabe uncertainty invoved in estimating the shape of the curve at this time due to the fact that few of the waste sites have been ceaned up.) This method impies that currenty reported incurred osses wi increase from $128.8 miion currenty to $600.4 miion.... A second version of the curve fitting to caendar year emergence which may be usefu in the future is an extrapoation based on actua payments. At this point in the environmenta caim cyce so few payments have been made that this procedure is not practica. -,- CALENDAR YEAR LOSS DEVELOPMENTMJ3THOD 2 For this method, ABC s reported osses to date are projected to utimate using deveopment factors from an externa source that reasonaby matched ABC s deveopment to date. 310

11 This method is iustrated on Exhibit 4. The method reies on the incurred environmenta osses, from Exhibit 2, by accident year and caendar year for ABC Insurance Company. The accident year osses do not dispay a norma pattern of deveopment for a casuaty coverage, however; it appears that the caendar year incurred oss totas at the bottom of the exhibit show a deveopment pattern. We seected Reinsurance Association of America (RAA) 1993 data as the externa source of data which might reasonaby match ABC s oss deveopment to date. Exhibit 4 compares the environmenta caendar year period to period deveopment factors from Exhibit 2 to the incrementa RAA factors. The R4A data is provided on an accident year basis and the factors on Exhibit 4 dispay the incrementa change in the RAA accident year osses from one year to the next. By posting the caendar year deveopment factors for ABC s environmenta caims against the incrementa (age to age) accident year R4A factors, we are attempting to match ABC s age-to-age factors against the RAA factors to estimate the equivaent maturity of ABC s environmenta caims. Based on Exhibit 4, we woud estimate that ABC s environmenta caims (in tota) are at a maturity equivaent to an accident year at 36 months of maturity. Therefore, one approach to deveop utimate environmenta osses for ABC Insurance Company is to mutipy the environmenta osses to date by a 36 month to utimate oss deveopment factor from RAA data. The foowing chart dispays the cacuation. ABC-Utimate Environmenta Losses (1) ABC Incurred Losses - Ai Years (2) 36 Month to Utimate Factor Based on RAA Data* (3) Utimate Environmenta Lasses (1)x(2) (4) Environmenta IBNR Reserves** (3)-() $128, , ,854 * Based on our review of RAA GL data for combhd treaty and facutetive business excuding environmenta and asbestos caims *+ Incuding suppementa deveopment on case reserves. 311

12 The resuts obtained using this method have to be monitored cosey. The foowing discussion is hepfu in understanding why we beieve a factor of 3.6 may be too ow for an insurance company with significant exposures and some of the imitations of this method. 1. The caim paying and reserving activity for environmenta caims has just begun for many companies and it is ikey to extend over a period in excess of 50 years. Using what has emerged in a horizon of ess than 10 years to project what may be expected in the next 40 years is best characterized as the tai wagging the dog. It is important to note that in using the RAA patterns we are not stating that the environmenta oss deveopment patterns are simiar to excess reinsurance patterns. Those patterns were seected because they provided a reasonabe match to ABC s deveopment to date, and we beieve that environmenta patterns, ike excess reinsurance patterns, have a ong tai. I 2. As is discussed ater, our crude estimates of environmenta osses for the U.S. insurance industry indicate a ratio of utimate osses to recognized osses (payments to date + case reserves + IBNR) of 4.7, which is in excess of 3.6. If ony reported osses were considered for the U.S. insurance industry, the ratio woud have been higher than 4.7. The 4.7 ratio is based on an estimate of $70 biion for the U.S. insurance industry utimate osses and recognized osses of $15 biion through (A specia report by A.M. Best s entited Environmenta/Asbestos Liabiity Exposure: A P/C Industry Back Hoe dated March 28, 1994 indicates that approximatey $15 biion has been recognized by the U.S. insurance industry through The U.S. insurance industry estimate of $70 biion is based on our anaysis outined in Attachment A.) q - 312

13 INDUSTRYBENCHMARKSMJZTHOD There are mutipe forces exerting pressure on an insurance company to recognize environmenta iabiities, e.g.., rating agencies such as Best s, SEC and reguators. However, the standards for estabishing appropriate environmenta iabiity reserves are sti deveoping. There is uncertainty associated with the estimation of utimate iabiities because historicay based actuaria approaches do not appy and exposure modes, when appied, may produce significanty different resuts with sma changes in assumptions. The Super-fund Reform Act adds another dimension of uncertainty in the estimation of these iabiities. The Superfimd Reauthorization Act, as proposed, has sweeping changes which coud have a significant impact on these iabiities. A arge portion of these iabiities may be addressed via a premium tax. Given these uncertainties, one approach to evauating environmenta iabiities may be to examine the reasonabeness of the reserves from a number of perspectives incuding comparison to industry averages and consistency over time. We have used actua data for Company A from its 1992 IO-K, adjusted by an arbitrary scae factor to obscure its identity, to iustrate benchmarks an insurance company might consider in evauating its environmenta iabiities. 313

14 (Doar Amounts in Miions) (41, (5) and (6) from A.M. Best s repon entited Envimnmcnra/Asbestos Liabiiiry Exposure: A P/C hdurny Back Hoe. The foowing observations can be made about Company A reserve eves: (1) Company A s reserves appear to be ess adequate than industry reserves. (Line 7a versus 7b and 8a versus 8b) (2) Company A s share of osses paid has been 11.5 % (ine 9) and its share of osses incurred is approximatey 8% (ine 10). Its market share based on CL premium is 4.5 %, The payment and incurred ratios to date indicate that Company A s share of utimate osses might be higher than its 45% premium share, This suggests severa possibiities, two of which are as foows: 314

15 (a) Company A s GL market share may not be representative of its share of industry osses because of higher than average exposure to insureds with environmenta iabiity exposures. (b) Company A s share is higher initiay but wi drop down to its GL written premium market share because most of Company A s exposure is in states where the environmenta case aw is more deveoped than for an average state, or its imits are ower. (3) Company A s reserves can fund 2 years of payments, compared to industry reserve eves which provide for 6 years of payments. (Line 7a versus 7b for 1992) (4) Company A s reserves provide for 1.2 years of IBNR osses compared to an industry eve of 3 years (Line 8a versus 8b for 1992). IBNR provides for true unreported caims as we as adverse deveopment on reported caims, Due to the uncertainty associated with coverage issues, initia case reserves may be ow even for caims that sette for significant amounts. Whie reviewing the environmenta iabiity reserve eves for Company A it might be instructive to review them in the context of what might be needed if Company A seected a reserving approach based on anaysis of the U.S. insurance industry data. The foowing tabe dispays the estimated paid osses through year-end 1993 for the U.S. insurance industry and some critica observations that can be inferred from the U.S. insurance industry experience. 315

16 ESTIMATED PAYMENT PATTERN U.S. MSURANCE INDUSTRY ENVIRONMENTAL LOSSES Estimated insurance Caendar Year Industry Payments (in Miions) and Prior $ , Tota $5,408.S 11. Estimated Utimate U.S. Insurance Industry $70,000 Losses 12. Paid Loss Deveopment Factor at December 13 31, Expected percentage of Losses Paid at 7.7% December 31, Reserve to Average Caendar Year Paid Factor 65 at December 31, 1993 (Assuming Average Caendar Year Payment of $1 Biion) (1): (2) - (5): (6) - (9): (11): (12): (13): (14): Estimated based on subsequent payments Estimated From Rand Study entited Superftmd and Transaction Costs A specia report entited Environmenta Asbestos Liabiity Exposures: A P/C Industry Back Hoe by A.M. Best Company dated March 28, See Attachment A. z ()+(O) (O)+() (70, )/1,COO.-I -_I Some of the,u.s. industry statistics that are hepfu in the evauation of Company A s reserve eves are outined beow:. Percentage of osses expected to be paid through December 31, 1992 [(o) - (9) / (11) 6% 316

17 . Mutipe of payments indicated as of December 31, 1992 for industry reserves to be fuy funded assuming average caendar year payment of $1 biion [() - (10) 1 1,ooo 66 Assuming average annua payments of $100 miion for Company A and a mutipe of 66 as indicated above, Company A s utimate osses coud be $6.6 biion. Thus indicated reserves as of December 31, 1992 woud be $6.2 biion ($6.6 biion - $0.4 biion estimated paid through December 31, 1992). Assuming that 6% of utimate osses are paid through December 31, 1992, the utimate oss eve for Company A is expected to be $6.7 biion ($0.4 biion / 0.06). Thus indicated reserves as of December 31, 1992 are $6.3 biion ($6.7 biion - $0.4 biion). i Using either one of the above approaches, Company A appears to be significanty underreserved with respect to what might be utimatey needed. Another test that is hepfu to Company A woud be to compare itsef to its peers. The foowing chart dispays the reserves as of December 31, 1992 expressed as a mutipe of average caendar year payments for three stock insurance companies using data from 12/31/92 IO-K s SURVEY OF ENVIRONMENTAL LIASILITY RFSERVFs* FOR A SELECTED GROUP OF CO~IPANIIB (A Doar Amounts in Miions) Of P) (3) (4) (9 (6) * W(4) RcacnC Cnksdnr Year F%wIats AVMgO To Amud 1990 m B?z AImsI Reserve fivmmt ppvment gb 12/31/92 w.!2 Company $30 $55 SW $ company company source: K s Iwhdcr Asbestos and Other Toxic Ton Caims < 317

18 As in the case of Company A, these sampe companies are posting reserves ess than the 65 factor that our anaysis for the U.S. insurance industry impied. However, Companies 1 and 2 show higher reserve ratios and are arguaby more adequatey reserved than Company A. Whie some companies might be justified in using a factor ess than 65, a factor higher than 65 may be appropriate for companies that are paying environmenta caims at a rate significanty sower than industry eves. Given the ong term nature of these iabiities, an argument coud be made that a factor ess than 65 is reasonabe. For exampe, industry net payments of $1 biion per year in a perpetuity at 5% interest woud be funded by $20 biion, impying a factor of 20. The vaues described above coud be atered by mutipes (even orders of magnitude) based on court decisions on coverage terms, reinsurance treatment, etc..._ MARKET SHARE MODEL The market share mode requires an estimate of the tota cost to the insurance industry associated with inactive waste sites, Attachment A provides an iustration of how the tota industry costs may be estimated. The cost for a specific insurance company is estimated based on the company s share of the tota insurance industry cost. The specific cacuation is described beow: 1. Tota company and U.S. insurance industry genera iabiity (GL) and Commercia Muti Peri (CMP) direct premiums written in the time period are compied. We are ony interested in GL and CMP premiums because these coverages are expected to generate the majority if not a of the insurance industry osses associated with inactive hazardous waste sites. We are interested in the years because those years are expected to generate the majority of the environmenta osses. 318

19 2. Based on the information compied in Step 1 above, individua insurance company direct premium as a percentage of tota US, industry direct premium is cacuated for the time periods , , , , , , and Expected U. S. insurance industry environmenta osses are then aocated to the five-year intervas described above using a basis such as the foowing: a. Years of operation of the sitesj: This is a proxy for years of dumping and is expected to provide a measure of the iabiity due to hazardous waste sites under the exposure trigger; or b. Year of discovery of sites: This basis of aocation provides a measure of iabiity based on the discovery trigger. 4. Individua company osses are estimated as the product of the percentage estimated in Step 2 and US insurance industry environmenta osses estimated in Step 3 for each appicabe five-year interva. These estimates may need to be modified based on some additiona factors. For exampe, if, an insurance company insured a high percentage of Fortune 500 companies or companies most often isted as PWs, then, its exposure may exceed its market share as determined in Step The resut of Step 4 is an estimate of direct utimate osses. The net utimate osses may be estimated based on individua insurance company s reinsurance programs. Some statistics that might be hepfu in the estimation process incude net to direct ratios exhibited by reported osses to date and written premiums. The procedure described above appies to primary companies. For reinsurers, a simiar approach may be used with one modification. The modification occurs in The attached Exhibit 5 dispays an aocation of costs to 5 year interva for a seect number of NPL sites based on data pubished by the EPA (this data is discussed in a ater section of the paper).

20 Step 2 and invoves anayzing a reinsurer s assumed premium as a percentage of tota direct premium to determine its share of the market. Additionay, the reinsurer s market share may have to be modified downwards because it is expected that the reinsurer s share woud be ower than what its market share woud otherwise indicate. This is because osses due to waste sites are expected to be spread over many years and many insureds, and therefore, may not expose the reinsurer as much as the primary company. (Steps 4 and 5 above woud require a primary company s market share to be increased based on the same ogic.) Additionay, specia adjustments may be necessary for companies which write a significant amount of excess and caims made coverage. The foowing tabe dispays an estimate of the tota cost for Company A based on the method discussed above4. I,_, +i MARKET SHARE MODEL ($ Biions) II 1) Seected Insurance Industry Tota Costs Due to $70 Inactive Waste Sites I 112) Percentage of Primary GL Market Written by 1 6% Company A 3) Adjustment for Company A s Reative Exposure (10 % 110% Greater Due to Concentration of nsureds Which are Chemica Companies) 4) Estimate of Losses Ceded to Reinsurers.- = Therefore for Company A, the cost estimate associated with inactive waste sites is roughy $3.5 biion. 4For simpicity, the industry osses are not separated into 5 year periods. 320

21 EXPOSURE MODEL The exposure mode separatey estimates the costs for reported caims and incurred but not reported (IBNR) caims. We first discuss the cost estimation procedure for reported caims. The cost estimation procedure for IBNR caims is discussed ater in this section. The costs due to inactive waste sites can be divided into the foowing categories: Cean up costs; Remedia Investigation I Feasibiity Study costs (RI/FS); Third party caim costs; Aocated oss adjustment expense costs (ALAE); Decaratory judgment action costs (DJA); and Unaocated oss adjustment expense costs (ULAE). The data required for the anaysis incudes the foowing information from insurer records: Reported caims and notifications per site and per PRP. Coverage terms-retention, imits, appicabe excusions, etc. Insurer estimates of costs (in tota or in the categories isted above), ikeihood of exposure, ikey share of tota cean-up costs for each insured, etc. Reinsurance attachment points, imits, and poicy terms. The insurer information can be suppemented by EPA data avaiabe in the foowing five databases. Comprehensive Environmenta Response, Compensation and Liabiity Information System (CERCLIS); Site Enforcement Tracking System (SETS); Superfund Comprehensive Accompishment Pan (SCAP); Record Of Decision (ROD); and 321

22 State books. CERCLIS contains a significant amount of information on each site identified by the EPA (not just the NPL sites). The information is site specific and a few of the fieds isted on the databases are: Name of the site; Location of the site; The physica cassification of the site (e.g. ground water contamination, dioxin, housing area); Status (NPL, non-npl); and Discovery date of the site. I.,,- asa Whie there are over 250 fieds in CERCLIS, CERCLIS does not incude a ist of the parties who dumped at the site (PRPs), the expected future costs associated with ceaning up the site, the actua expenditures to date associated with the site, or information regarding the dates the site was used/cosed. That information comes from other sources. The SETS database contains a ist of PRPs identified by site. These PRPs may or may not have yet fied caims with their insurance carriers. To the extent that this ist agrees with the insurers caim notifications, it represents reported caims. To the extent that poicyhoders are incuded in the SETS ist but have not yet fied caims, these sites represent potentia IBNR reports. -= The next database, SCAP, contains actua expenditures by site. The expenditures are divided into approximatey 50 categories, which can be aggregated into two broad types of expenditures: 322

23 Remedia Investigation/Feasibiity study (RIIFS) expenditures; and Actua cean-up costs. The RI/FS expenditures represent the costs associated with investigating the site and determining how to best cean up the site. These costs are often significant. Both RI/FS costs and actua cean-up costs are not avaiabe for a sites. The next database, ROD, contains information on cean-up costs estimated by the EPA at individua sites. The record of decision (ROD) is a forma estimation procedure empoyed by the EPA. The foowing information is avaiabe on ROD: The date the ROD was estabished; Estimated initia cean-up costs; Estimated cost to monitor the site once the initia cean-up is compete; Number of years of annua maintenance; Whether the estimated costs are undiscounted or discounted: and Owner of the site (sometimes). The ROD database aso contains information on the physica condition of the site. In many cases the EPA deineates cost summaries by technoogy empoyed to cean up a site. Of the 1,300 sites on the NPL ist, about 600 have RODS. Of the remaining non-npl 323

24 sites RODS have ony been competed on a sma percentage of the popuation of sites, but it is anticipated that approximatey 60% of the 37,000 potentia sites wi not require a ROD as the site wi not need to be ceaned up. State Books The ast data source is the state books. The state books contain, among other things, the number of years the site was in operation, the year the site was cosed, nature of ground water contamination if appicabe and proximity of neighborhoods to the site. Descrbtion of ExDosure Mode - Known PRPs/Sites The mode estimates utimate osses associated with reported caims (situations where a PRP has notified the insurance company of its exposure at a site) for cean-up costs, RJ/FS costs, third party caim costs and ALAE. Estimates of costs for ULAE, Decaratory Judgment Actions (DJA) and IBNR are prepared separatey _ The key steps in the mode are as foows: 1. Identify reported caims for each PRP and site combination 2. Estimate costs by site from EPA data, insurer data and other sources 3. Aocate the costs by year for each site 4. Appy the PRP share to the step 3 resuts 5. Appy poicy imits and reinsurance retention by year/prp/site 6. Adjust for the probabiity that insurance coverage appies 7. Repeat steps -6 for each PRP/site combination and aggregate to obtain the tota insurer cost estimate for reported caims 324

25 This mode can be envisioned in the foowing manner by site. First, PRPs are identified by site. Step 2 invoves estimating the cean-up and RI/FS costs by site. If a ROD estimate from EPA is avaiabe this may be used, otherwise, cean-up and RI/FS costs can be estimated. (For exampe, we woud expect simiar sites in the same genera area to have simiar costs.) Next, costs are spread to year and PRP based on the assumed ega coverage theory and known or estimated PRP shares. These costs are then increased for deficiencies in EPA estimates, third party costs, ega expenses, ALAE, etc. Some costs (e.g., third party costs) may be estimated as a percentage of the cean-up costs on the assumption that these costs are ikey to be correated with cean-up costs. Next, specific coverage items are considered (sef insured retentions, aggregate imits and reinsurance). The resut of the first four steps is the anticipated cost to the insurance company assuming that a inactive waste site exposures are covered (i.e. insurance company does not win on any coverage defense issues). Lasty, the probabiities of coverage responding are appied to certain cost items (to cean-up costs but not ega costs), Site Identification and Cost Estimates Based on insurance company records, known PRPs and exposure years can be identified. The sites on which PRPs are exposed can be identified from both insurer records and EPA databases. For exampe, from EPA data sources, a record can be created to refect : The insured (PRP);. A cost estimate for the site (cean-up, etc.); and The number of years the site was in operation. Cost Aocation bv PRP and Year Next, costs by site need to be spread to year and PRP. There are severa ega theories that can be used to spread the oss estimate to individua 325

26 years. Potentia triggers are: Exposure; Manifestation; Continuous; or Actua injury. If the appicabe trigger were the exposure trigger, the osses might be spread equay to the years the site was used (years of operation of the site may be used as a proxy if more detaied information is not avaiabe). Simiary, oss estimates under aternative triggers can be cacuated. Next, the PRP share by site/year may be estimated as (a) /n where n is the number of PRPs on the site or (b) 1 /n adjusted to refect the reative size or degree of responsibiity for the PRP. A size adjustment woud be based on the theory that a arger PRP is more ikey to be abe to pay and may have contributed more to the environmenta impairment than a smaer PRP. One measure of degree of responsibiity might be how often the PRP is on an EPA site ist. Another measure of size is whether or not the PRP is a Fortune 500 company. For exampe, if 20 PRPs are named at a site, one estimate of a specific PRP s share for the site woud be 5%. However, a Fortune 500 chemica company shoud probaby be assigned a share greater than 5 %..., *.- Poicv Terms and Reinsurance In the next phase, poicy provisions and reinsurance are appied to estimate individua insurance company shares of these osses under the assumption that coverage appies. For exampe, if the above mentioned procedure resuted in $1,275,000 of osses per year for a specific PRP insured, and if the insurance company ony wrote poicy imits of $,OOO,OOO per year (in aggregate), then the insurance company s indemnity exposure woud be capped at $,OOO,OOO per year. 326

27 Probabiitv of Coverage The ast step woud be to incorporate the probabiity that coverage appies to the estimates by site/prp/year. This probabiity is based on the jurisdiction and the insurer s coverage defenses. The probabiity of coverage responding is a rather compex item which woud most ikey vary by: The coverage defenses postuated by the insurance company; The state; and The year (IS0 introduced a poution excusion in 1973 and a second stronger excusion in 1986, and many companies foow IS0 forms.) The probabiity of coverage responding may best be thought of as a matrix by year: PROBABILITIES OF INSURANCE COVERAGE RESPONDING Coverage Defense Cean-up Costs Not Damages as Defined in State Cean-Up Costs Excuded Due to Poution Coverages ony appies if Damage is not XX% XX% Expected or Intended I Owned Property Excusion XX% J XX% Late Notice of Occurrence 1 XX% XX% Tota Costs The above procedure is performed by siteprp/year combination and the resuts aggregated to determine the insurance company s potentia reported exposure for a PRP. A insured PRPs can then be aggregated to estimate the insurance company s potentia exposure. DJA, ULAE and IBNR costs are described in the next sections 327

28 Decaratorv Judment Action (DJA) Costs DJA costs represent the costs associated with itigating coverage issues (e.g., whether a CGL poicy responds to Superfimd cean-up). The DJA costs may be estimated based on: Average DJA expenditures per site and PRP; Expected number of future caims (PRP/site notifications); A factor refecting the fact that over time as coverage issues become more we defined, costs may be reduced; and Infation in ega expenditures. The foowing tabe dispays a sampe cacuation for a hypothetica insurance company: ABC Insurance Company.1 (1) Average Historica DJA Costs Per Site per PRP $75O,ocQ (2) Estimated Future Site/PRP Combinations Invoving DJA Litigation 100 (3) Factor Refecting More Ceary Defined Case Law 50% (4) Infation Factor for Lega Fees 1.2 Estimated Future DJA Costs $45 Miion ULAE costs One method to estimate ULAE costs is to estimate:.r Average annua ULAE costs; The number of years in the future for which ULAE costs wi be incurred; and Infation in caims adjustment costs. For exampe, many insurance companies have estabished a specia work force of caims personne dedicated to handing ony environmenta caims. If we assume: (1) A unit generates annua saary and benefits of $350,000; (2) Wage and benefit infation of 5% per year; and 328

29 (3) Environmenta caims take 30 more years to be setted, then, the estimated ULAE reserve is equa to (35O,OOO)(1.O5)+(35O,OOO)(1.O5)2+...(35O,OOO)(1.O5)3o or approximatey $24.4 miion. IBNR Caims IBNR caims may resut from the foowing: (1) Known PRPs being named at future sites; and (2) Unknown PRPs being named at known and future sites. The cost of IBNR caims can be cacuated by PRP for known PRPs at future sites based on:. Anticipated number of sites where an insured (PRP) wi be named; Estimated cost of the sites (incuding cean-up; RI/FS costs, third party costs and ALAE costs); The PRP s share at IBNR sites (PRP shares at known sites may be used as a proxy); Insurance company coverage response probabiity (again information at known sites may be used as a proxy); and Coverage provisions and reiusurance. To iustrate, assume that PRPs have been notified by the EPA on 600 sites and utimatey we expect PRPs to be notified by the EPA at 3,000 sites. Therefore our IBNR caim universe for PRPs is 2,400 sites (i.e. the maximum number of additiona times that an insured coud receive a PRP etter is 2,400). Based on the 600 sites for which the 329

30 EPA has identified a ist of PFWs, a specific PRP is identified 60 times (10% of the time). Therefore, for the additiona 2400 sites we might assume that the PRP woud be named 240 times (2,400 times 10%). Next, based on evauating previous sites, we might estimate a cean-up cost of $33 miion for each newy identified site. Based on known sites, the PRP s average share is 5 %. Based on the specific insurance company s success in arguing that coverage does not appy and on the insurer s coverage and imits, we estimate that the insurance company may be responsibe for 40% of the tota costs. Therefore, one estimate of the insurance company s iabiity for a specific PRP s IBNR exposure is: M *.,._, ~ Insurance Company s Estimated Liabiity for Newy Identified Sites (1) Estimated Number of Future Sites 2,400 (2) Estimated PRP Exposure at Future Sites 10% (3) Estimated PRP IBNR Sites (1)x(2) 240 (4) Average Cost of Newy Identified Sites $33 Miion (5) PRP Share 5% (6) Insurance Company Coverage Probabiity* 40% (7) Third Party and ALAE Costs Factor 1.70 (8) Insurance Company Liabiity (3)x(4)x(5)x(6)x(7) $269 Miion *Incudes coverage provtstons (e.g. hmns, number of years insured).-.-. = - T This process can be repeated for a the insured PRPs to obtain a tota estimate of IBNR cost for known PRPs at future sites. Unknown PRPs at current and future sites may be refected using a judgmenta factor. These costs can then be aocated to year based on EPA information (e.g., years of operation of the future site universe). The IBNR estimates by year pus the estimates for reported caims equa the tota costs. 330

31 If the estimates of tota costs are summarized in five-year intervas, these vaues can be compared to the resuts of the market share mode discussed previousy. SUMMARY This paper has outined severa methods that can be used to estimate insurance company (as we as PRP) iabiities associated with inactive hazardous waste sites. Additionay we have outined severa pubicy avaiabe data eements which can assist in evauating environmenta iabiities aong with summarizing the current ega issues invoved in coverage disputes between insureds and insurance companies (Appendix A). The potentia iabiity associated with inactive hazardous waste sites is significant. Insurance companies and PRPs need to introduce procedures to attempt to monitor and quantify the potentia iabiity. None of the procedures described in this paper provide the method to anayze environmenta iabiity exposures. For financia reporting purposes, company management needs to evauate the detais of its own exposures and judge the utimate cost based on current facts and financia reporting principes. Management shoud aso consider the provisions under the Superfimd Reform Act of 1994 which are ikey to have a significant impact on these iabiities. 331

32 Exhibit 1 Lifecyce of Latent Caims 60% * Event + Exposure + Emergence + Expenditures Year Event - Assumes the event stage occurs between 1960 and Company A uniformy dumps at a particuar site between 1960 and Exposure - The chemicas start eaking in 1970 and are sti eaking. Therefore the exposure stage starts in 1970 and is sti occurring. Emergence -- The effects of the exposure are known. For one particuar site, this may be a point in time. However, it wi be a curve for a sites. Expenditures - Company A makes payments to cean up the site. Ceanup at the site begins in the year 2000 with ongoing maintenance continuing unc2040.

33 Exhibit 2 ABC Insurance Environmenta Company Caims Incurred Losses (SOOok) At At At 12/ J2Jg At At At 12/ / ,300 3,350 13,350 13, , ,230 11,400 11,400 27, ,000 7, ,876 14, ,690 11, ,400 19,000 23, ,000 7,300 29, & Subsequent Caendar Year Tota ,530 10,380 26,760 85, ,790 Caendar Year LDF NA NA

34 Exhibit I /--- i;,. LLL Caendar Year S - Curve: Y = 200,140 Arctan[X According to Makridakis and Wheewright, An S curve impies a sow start, a steep growth. and then a pateau. Forecastina Methods for Manaaemen_t, Paqe ti t 4.,&ii

35 Exhibit 4 Comparison of Deveopment Factors ABC Insurance Company Environmenta Caims RAA Data For Genera Liabiity Caendar Year ABC insurance Company Aae to Aae RAA Accident Year Age to Aae Factors / /91-12i / t

36 Exhibit 5 An Estimate of the Aocation of NPL Cean-Up Costs to S-year Periods For Seect NPL Sites Period Prior to &Subsequent Percentage 4?x!a 0.64% 0.22% 0.29% 0.37% 0.41% 2.58% 2.19% 0.59% 0.81% 1.75% 2.76% 6.32% 9.99% 12.99% 15.84% 18.04% 16.88% 5.25% 2.10% Note: In aocating costs to S-year period, we assumed an exposure trigger and used the years of operation of the site as a proxy for years of dumping. The exposure is based on an aocation of ROD cean-up cost estimates to year for those NPL sites with avaiabe ROD cost estimates.,, II!1

37 Attachment A Sheet 1 The approaches for esthnating insurance industry iabiities due to inactive hazardous waste sites are iustrated on sheets 2 through 7 of this attachment. We have used an estimate of $70 biion throughout this paper as an estimate of the tota iabiities for the U.S. insurance industry. It is important to recognize that these utimate oss estimates are highy uncertain. For exampe, a specia report entited BnvironmentabAsbestos Liabiity Exposure: A P/C Industry Back Hoe dated March 28, 1994 indicates expected environmenta iabiities of $255 biion. The best and worst case estimates in that report are $50 biion and $608 biion respectivey, showing the uncertainty associated with estimating these iabiities. This uncertainty stems from the fact that many of these cases have not been resoved in court yet. In addition, average ceanup costs, third party costs, PBP shares, insurer itigation costs and success of insurer coverage defenses are critica assumptions in the estimation process and am best guesses at this point. The approach described in sheet 2 expicity considers the various eements such as cean up costs, ALAB costs, etc. for which the insurance industry woud be responsibe with respect to inactive hazardous waste sites. The ony item that is not considered is the payment associated with natura resource damages. The PIG%, and hence, the immance industry, may be required to share in the cost of restoring natura resources damaged by poutants to their origina form. The cost for this eement is not considered because there is very itte information avaiabe on this issue. Sheet 2 provides the utimate oss estimate for the insumnce industry using a set of what might be considered reasonabe assumptions. The notes on sheets 3 and 4 expain some of the thought process tbat underies our assumptions. It is important to understand tbat there is uncertainty associated with each of those assumptions and more than one set of assumptions may be considered reasonabe. To iustrate this uncertainty, we have incuded resuts based on a variation of the critica assumptions. Sheet 5 provides resuts based on these aternate assumptions. Sheet 6 outiis an aternate method where the insurance industry utimate oss payments are estimated as a percentage of tota nationa expenditures reated to cean-up activity. Sheet 7 SummarizRs the resuts of various estimates of utimate environmenta iabiities. Based on review of resuts in sheets 2, 5, 6 and 7 we seected $70 biion as the utimate oss estimate for the U.S. insurance industry for the iustrations in the paper. 337

38 Estimated Utimate Insurance Industry Liabiity Due to Inactive Hazardous Waste Sites Doar Amounts in Miions Attachment A Sheet 2 (1) Expected number of utimate NPL sites (2) Estimated cean up cost per site (3) Estimated RI IFS cost per site (4) Estimated tota cean up and RJ I FS cost for NPL sites [ (1)x ((2)+(3)} ] (5) Estimated expected number of non-npl sites (6) Estimated cean up and RI IFS cost per non-npl site (7) Estimated tota cean up and RI I FS cost for non-npl sites [ (5)x(6) ] (8) Tota cean up cost at NPL and non-npl sites [ (4)+(7) ] (9) PRP share of (8) (10) Tota PRP cean up cost responsibiity [(8)x(9) ] (11) Thirdpartycosts [25%of(O)] (12) Insurance Industry portion of PRF share if coverage were to appy 100% of the time (13) Insurance Industry cost if coverage were to appy OO%ofthetime [((O)+())x(2)] $180,000 I,,., 50% :&I $90,000..I., $22,500-60% $67,500 (14) Probabiity that coverage appies (15) Insurance Industry Indemnity cost [ (13)x(14)] (16) ALAE / ULAE I DJA costs as a percentage of tota indemnity costs (17) ALAE/ULAF/DJAcosts [(15)x(16)] (18) Tota cost to the industry for Indemnity, ALAE, ULAF, DJA costs [ (15)+(17)] $33, Et 60% -..I -. * $20,250 -_ $54,