risk management july 2004, issue 2 - member | soa · by david ingram _____2 2004 enterprise risk...

Chairperson’s Cornerby David Ingram __________________________2

2004 Enterprise Risk ManagementSymposiumby Valentina Isakina ________________________4

Risk Management Task ForceSubgroup Updates _____________________8

Risk Management Skills and Aptitudesby H. Felix Kloman ________________________12

Risk Management at the Federal Reserveby Henry McMillan _______________________ 15

Influenza Pandemics: Are We Ready For the Next One? by Max J. Rudolph ________________________20

Getting to Know CTEby David Ingram _________________________33

July 2004, Issue No. 2

Newsletter of the Risk Management Section

Published in Schaumburg, ILby the Society of ActuariesManagement

Risk

Table of Contents

Chairperson’s Corner

T he Risk Management Task Force(RMTF) has set a new standard for vol-unteerism in our profession. Over a

third of the 300 members of the RMTF have par-ticipated in committees and working groups.That standard needs to be carried over to thenew Risk Management Section so that the actu-arial profession, the risk management disci-pline and the individuals who participate inthese developments will all reach their maxi-mum potential.

One unique character-istic of actuaries is ourlong-term perspective.We start our careers byundertaking one of thelongest educational andtraining processes of itskind. We regularlymake long-term finan-cial projections that runfor decades. We believeboth professionally andpersonally that thereare dependable cause-and-effect relation-ships that hold over long periods of time. Webelieved that if we put in the work to completeour exams that we would be able to have chal-lenging, interesting and well-paying careers,and our life experience has shown that to betrue.

That has been true especially in the traditionalactuarial areas of product development ratemaking and valuation. Every year brings newchallenges, but usually those challenges can bemet by work that is an incremental extension ofrecent experience built upon the basic actuarialeducation.

Risk management is not such an incrementalarea. Risk management, as defined by financialeconomists and practiced by bankers, is aprocess for risk traders who only retain a smallfraction of the risks that pass through theirhands. Insurance companies are just the oppo-site: they consist of risk collectors who only lay

off a small fraction of their risks. Some of themethods and models of banks and financialeconomists may produce incorrect signals whenapplied to insurance companies.

On the other hand, some of the fundamentalideas of risk management have tremendous ap-peal. Creating a control cycle for risk, attendingto net risk after accounting for correlation andlooking to the financial markets for informationon risk are just a few of the powerful ideas that

are fundamental to riskmanagement.

However, the details of ap-plying these ideas to insur-ance companies are farfrom easy or obvious. Atthe same time, these andother risk managementideas are only slowly be-coming well known in theinsurance community.

The process of buildingboth theoretical and prac-tical bridges between in-

surance company risk and banking riskmanagement is a monumental undertaking. Intwo instances, I have participated in prolongedbrainstorming sessions to identify manageablesteps in that process. In both cases (at the timeof the formation of the Risk Management TaskForce in 2001 and a few months ago at the firstmeeting of the Risk Management SectionCouncil) we developed lists of 20 to 30 projectsthat needed to be done.

From those extended lists, there are now 15projects at various stages of completion. By thetime those have been completed, 15 more willhave started, but only if we get enough help. Weneed volunteers; not just a handful of volun-teers, not a dozen, but hundreds! If you havenever volunteered, then you are exactly who weneed, because fresh perspectives and energyare in high demand.

Risk Management ◗ July 2004

The Risk Management Section Needs You!by David Ingram

Successful volunteer

efforts create recognition

in the profession, a

benefit in and of itself.

“

”

◗

Chairperson David Ingram,

FSA, MAAA, is a consulting

actuary at Milliman USA

in New York. He can be

reached at david.ingram

@milliman.com.

Volunteering is a long-term investment, but onewith a high number of valuable payoffs. The fol-lowing is a partial list of the benefits of volun-teering:1. The challenge of working on a problem of

your own choosing.2. Exposure to diverse perspectives from co-

volunteers who work with different prod-ucts and various jobs all over the worldwith company actuaries, regulators, con-sultants, analysts and others.

3. Learning new things as co-volunteersshare thoughts from a new or classic paperon a subject that interests you.

4. Networking with other volunteers to great-ly expand the list of people who you mightbe in contact with the next time you needhelp.

5. Learning broadly about the wide range ofindustry activities long before they aredocumented in formal meetings or publi-cations.

6. The lowest-cost way to gain these benefits.Almost all meetings are by phone and theSOA even pays for the calls.

7. Leading-edge practices developed by thevolunteer groups from combinations of thepractices and ideas of the participants.

8. As this volunteer activity produces resultsthat are recognized throughout the finan-cial services industry, the value of the actu-arial credentials will be greatly increased.

9. Learning to bring in and use resourcesfrom outside industry and to incorporatethem into actuarial practices. Often, thoseresources were completely unknown to thevolunteers before the project.

Volunteering forces participants into a differentmode of interacting, a more collaborative mode.Improving skills to “play well with others” willpay career dividends. Successful volunteer ef-forts create recognition in the profession, a ben-efit in and of itself. And you never know whenyou will be grateful that having your name on acommittee report opens a door that you didn’tknow existed when you did the work. As thesevolunteer committees complete different tasks,others will look upon their findings and methodsand realize that there are applications far be-yond the originators’ dreams.

If you are feeling in even a little of a rut, volun-teering can provide the satisfaction of accom-plishment.

More than one of the RMTF volunteers hasfound new positions, whether within the samecompany or at a different firm, at least in part be-cause of what they learned, who they met andtheir new skills and positive attitudegained from their volunteer accom-plishments.

For over 20 years, I lived in an oldneighbor-hood of stately Victorianhouses. The cost of a typical housewas quite modest, but the upkeep wasvery, very high. Most of the home-owners in the neighborhood paidtheir pittance for a slightly run downgiant of a house and let it slide slowlyat first, but more and more with time into furtherdisrepair. Once in a while, someone wouldcome into the neighborhood and take one of theworst houses, and with lots of sweat equity andnot a small amount of cash, return it to a reason-able semblance of its former glory. They werealways disappointed that the housing marketdid not support them. The top three factors inhouse valuation there as everywhere else areLocation, Location and Location. To get thevalue out of their efforts, many, probably most,of the rest of the neighborhood would have to dolikewise.

We all own (or want to own) houses in the actuar-ial and risk management neighborhood. Thosehouses will all be worth more if we all get out andpaint the walls and fix the roofs. Or better yet,read about what is going on in the volunteercommittees and join today!✦

July 2004 ◗ Risk Management

◗

Section Council Elections

DON’T FORGET TO VOTE IN THE SECTION COUNCIL ELECTIONS—JULY 12 THROUGH AUGUST 13!!

The following persons are candidates for the Risk Management Section Council:Clifford W. Angstman, Berkshire Life Insurance Company of America, Pittsfield, Mass.Douglas W. Brooks, Sun Life of Canada, Toronto, OntarioAsutosh Chakrabarti, Actuarial Consortium LLC, Monmouth, NJS. Evaronda Chung, IntAct Consulting, Inc., Toronto, OntarioLeonard Mangini, AXA CS Life Re, New York, NYJohn W.C. Stark, Anthem Health Plans of Virginia, Inc., Richmond, Va.Ken Seng Tan, University of Waterloo, Waterloo, OntarioFred Tavan, The Canada Life Assurance Company, Toronto, Ontario

For further information go to www.soa.org and click on the election information link.

Editorial

Nearly 400 chief risk officers and enter-prise risk management experts gath-ered in Chicago on April 26 and 27,

2004 to discuss the latest developments on en-terprise risk management (ERM) and sharetheir thoughts on how ERM enables companiesto optimize the overall corporate risk exposureand leverage business opportunities more prof-itably.

Sponsored by the Society ofActuaries and CasualtyActuarial Society, the ERMSymposium brought together en-terprise risk managers from a va-riety of disciplines andindustries from all over theworld. Guided by its missionstatement, “ From Cutting EdgeTheory to State of the ArtPractice,” the symposium drew

top ERM experts to Chicago to share their expe-rience with colleagues and challenge the audi-ence with thought-provoking issues to addressin the future. The Georgia State University’sThomas P. Bowles Symposium and ProfessionalRisk Managers’ International Association(PRMIA) were co-sponsors of this event andgenerously contributed to the program develop-ment.

For the two days of the symposium—over the course of five general sessions and 35 concurrent sessions—the attendees discussedand deliberated over the most critical aspects of implementation issues and obstacles for creation of the most effective, yet customized,enterprise risk management framework. Topicsdiscussed spanned across various industries—insurance, banking, energy, retirement systemsand beyond and various other disciplines—from technical risk measuring applications to broad operational risk issues and implemen-tation. In the following paragraphs, we offer you a glimpse of the symposium topics. Thecomplete program and presentations are available online at: http://www.casact.org/coneduc/erm/2004/handouts/.

CRO Forum The symposium was launched with a CROforum moderated by Harry Panjer.

With such panelists as Robert Mark, presidentand CEO, Black Diamond Inc.; James Lam,president, James Lam & Associates; and LucHenrard, general manager and chief risk offi-cer, Fortis Group, the forum was a sure hit.These experts were at the cornerstone of the im-plementation of ERM systems/processes ingeneral and their respective firms in particular,and have written authoritative books and/or ar-ticles on ERM. Each of them was able to bring aprofound insight to the session.

First, Bob Mark described the components of aproactive risk management function:

• Policies/procedures, including setting risktolerances and constraints

• Best practice methodologies, including calculating VaR, stress testing, setting of models and the tie-in to performancemeasurement

• Developing the infrastructure, including anassessment of people skills, operations, dataand technology.

Second, James Lam discussed the history andproliferation of the CRO position. In particular,his description of the “good CRO” attributesresonated with the audience:

• Organizational/leadership skills• Communication skills—to simplify without

being simplistic• Technical skills—need two out of three in

credit/market/operational risk• Balance business and risk requirements• Courage to push back/say no• High IQ and high emotional quotient• Ultimate test: the ability to integrate risk

management into the day-to-day businessoperations.


2004 Enterprise Risk Management Symposium—FromCutting Edge Theory to State of the Art Practice by Valentina Isakina

◗

Valentina Isakina is

finance practice area

actuary on staff at Society

of Actuaries. She can be

reached at 847.706.3584

or at [email protected]

James also gave his 10 predictions for the futureof ERM, including the line “Economic capital isin, VaR is out!”

Luc Henrard presented to the audience his ex-perience with the risk management of a finan-cial conglomerate and the challenges inbridging the gap between banking and insur-ance risk management.

In Fortis’s framework, risks include insurancerisks, investment risks and operational risks.Return measures include ROA (return on as-sets), ROE and return on risk-adjusted capital(RORAC)/risk-adjusted return on capital(RAROC). Luc also stressed the importance ofeconomic capital (EC) as an economic measurefor solvency capital. In his organization, EC iscalculated by turning embedded value modelsinto risk models.

All the panelists emphasized the importanceof buy-in from senior management and sug-gested that the actuarial curriculum be ex-panded to include all-around financial riskmanagement, thereby integrating actuarialand economic sciences.

Closing RemarksThe closing remarks session was one of the mostremarkable sessions at the symposium, andthose who attended this last session were richlyrewarded. Three expert panelists presented tothe audience their impressions of the sympo-sium, summarized the results achieved andchallenged the participants with thought-provoking issues to address for the next gather-ing of the symposium in 2005.

1. Observations on ERM Practice:First, Prakash Shimpi, visiting fellow from theLondon School of Economics & PoliticalScience, senior fellow at the Wharton Schooland president at Fraime LLC offered his obser-vations on the ERM practice. He pointed outthat the roots of ERM are in both asset-liabilitymanagement (ALM) and corporate finance andidentified the following key challenges that facepractitioners striving to implement ERM intheir organizations:

Alignment of management According to Prakash, breaking down the silosthat have arisen due to history and jargon acrossdifferent groups of professionals is one of the

most critical steps in moving the ERM profes-sion forward, but it also seems to be the biggestchallenge. Learning from each other’s bestpractices and speaking one language makes theERM process more effective and brings value inability to communicate between different pro-fessional groups and provide management asolid and consistent message.

Problem misspecificationFor actuaries or other risk management profes-sionals, it is very important in their modelingwork to keep in mind that the relationships usedas proxies for the underlying problem remainjust that—proxies. There are plenty of favoriterisk metrics used in a variety of industries, andsome of them are more valuable than others.However, according to Prakash, the lesson re-mains—it is essential to understand the realproblem and not just the proxies.

Margins for errorAs with any new discipline, ERM is dealing withvarious limitations—from systems to data avail-ability—while being asked to provide answersfor increasingly complex questions. Prakashsuggested that for the process to be successful, itis important to re-examine the traditional statis-tical methods and consider alternative tech-niques that help mitigate data limitationsthrough margins in the analysis.

Spurious precisionSpurious precision, stated Prakash, is a naturalconsequence following from the previous twoobservations. Being correct to the third decimalplace on something with many assumptionsbuilt in is not necessarily valuable information.Prakash argued that in ERM, it is more valuableto have a more complete representation of thefirm’s risks at lower overall precision than tohave surgical precision on some risks and no in-formation on others.

AnalyticsThe last observation Prakash made was to iden-tify the greatest limitations in the progress ofERM—the lack of solid analytics. He arguedthat in the absence of strong analytics is it diffi-cult, if not impossible to demonstrate value—“Analytics should enable us to convert the rawrisk data of a firm into knowledge about how therisks impact the firm and the economic value ofthe instruments and strategies that can be used


continued on page 6 ◗

“

”

Learning from each

other’s best practices

and speaking one lan-

guage makes the ERM

process more effective

and brings value in

ability to communicate

between different pro-

fessional groups and

provide management a

solid and consistent

message.

◗


as mitigants. Such analytics should become anintegral, not separate, part of the routine corpo-rate planning process of a firm.”

2. Summary Remarks on the ERMSymposium 2004 Chicago

Next, Thomas Ho of the Thomas Ho Companytook the microphone and provided a brilliantsummary of the story that the general sessionsconveyed in the two days of the symposium.

For Thomas, the overwhelming message thatcame out of the conference was the intensesearch for the definition of enterprise risk man-agement—the scope of ERM, the responsibilityof ERM, how to define the destiny of ERM.First, Shaun Wang challenged the audience toinvent a new paradigm of ERM, and stated thatERM is a manifold, a system that relates all thesmall parts to form the whole. Other symposiumspeakers described ERM from other perspec-tives, and, to help describe these perspectives,Thomas invited the audience to join him on atour of a headquarters of a securities firm.

First stop—the trading floor In the CRO Forum, Bob Mark illustrated howthe CRO could have lowered the VaR tolerancelevel in the months prior to the Russian CrisisAugust 98—tying the risk measures to perform-ance is key to the success of risk management.This view was extended by Leo Tilman in thesecond general session, “State of RiskManagement Practices.” Giving specific deriv-atives examples, Leo observed that ERM mustbe aware of the external forces in the market-place and that the CRO should take a holisticapproach and have resources beyond capitalmarkets instruments.

Second stop—up one floor of these corporate headquarters On this floor, Thomas described how James Lamdrew from his experience in ERM to concludethat the CRO must be a leader who leads themain asset in ERM—people. James alsostressed a practical perspective, which is thealignment of incentives: GE’s success in the

Six Sigma campaign owes much to the compen-sation scheme, which attributes one-third of thecompensation to meeting the six sigma targets.Thomas also noted the contribution made byChuck Lucas, who stated that the CRO has theunique challenge of being a leader in the busi-ness sense and a leader in the technical sense.Chuck used the variable annuity product to illustrate how the CRO must understand thetechnical aspects of hedging the risk of the guaranteed death benefits and at the same timecan bet on the disappearance of the re-insurance market for the variable annuity guarantees.

Third stop—taking the elevator to thethird floorThomas described how the theme of economiccapital, first introduced by James Lam andDonald Mango, was appearing throughout thesymposium. Don extended Chuck and James’points to suggest that managing all the stake-holders of ERM is the greatest challenge to theCRO and the best way to resolve the issue is toassign economic capital to the units. However,Donald asserted that we need a more dynamicdefinition of this process and introduced theconcept of “renting” capital. In this analogy, theCRO was presented as a hotelier who rentsrooms out for use, keeping the optimal balanceof the needs of the stakeholders.

Further, Thomas noted Luc Henrard’s contribu-tion to the discussion of the challenges facingERM in setting a uniform standard across busi-nesses. Luc observed that the regulatory capitalrequirements for banks, insurance companiesand other entities are different and challengedthe audience to specify the economic capital ap-propriately for the management of risk acrossthe various industries. To continue the theme ofERM and regulation, Thomas made a connec-tion to the presentation by Darryll Hendricks.Darryll Hendricks described the complexity ofsetting a uniform required capital that is beingnegotiated among the regulators. While theEuropean regulators are making progressacross the Atlantic in aligning the regulationsacross different industries, the office of thecomptroller currency (OCC), sub-area council(SEC), NAIC, Federal Reserve Board (FRB),and other regulators on this side of the Atlanticare not converging and will have a complexproblem to solve. Thomas noted that the lunch-eon keynote speaker Zvi Bodie also showed con-

Editorial◗ continued from page 5

Wang challenged the

audience to invent a

new paradigm of ERM,

and that ERM is a mani-

fold, a system that

relates all the small

parts to form the whole.

“

”

◗

cern for the impact of ill-conceived regulationsin the arena of pension—the alternative pen-sion liability valuation accounting methodsconflict with economic principles. Zvi arguedthat the asset valuation accounting methodsmust adhere to the law of one price and urgedthat our 401(k) plans should be prompted byregulations to invest more in bonds and not eq-uities. Inappropriate regulations affect not onlycorporations in the small, but the economy inthe large.

Last stop—the Board RoomThomas described the Board’s concerns withthe shareholders’ value and how various speak-ers addressed this fundamental issue during theconference. Tom Wilson, for example, saw ERMmanaging all the business processes of the firm.While ERM provides the risk measurements,monitoring and analysis, Tom stressed how im-portant the “soft” aspects of ERM were in relat-ing the analytics to the business processes of thefirm. To illustrate his observation, Tom provideda complex and comprehensive map of theseprocesses in a firm. At the same time, Thomasnoted that David Ingram, in his 12 points of thebest practices of ERM, suggested that one ele-ment was missing in the map, which was: “The firm has a process to identify weaknessesof ERM.” Thomas noted that Dave’s presenta-tion spoke of ERM as a dynamic process, not amap. This process has to have a self-correctingmechanism, and this self-correcting mecha-nism is the essence of risk management.

3. The Foundations of Enterprise RiskManagement

The final speaker of the session was William H.Panning, EVP and Managing Director of WillisRe Inc. Bill focused on an issue that is crucial toERM but rarely discussed—time.

Bill stated that measures of risk such as (VaR),and its close relatives are like flashlights in acave, in that they illuminate some aspects ofreality but simultaneously conceal others thatfall outside their scope. In particular, thesemeasures focus on the distribution of possibleoutcomes at some particular point in time. Infact, however, firms such as insurers take onrisks for a variety of time horizons, and no sin-gle point in time is predominant. This has sev-eral implications.

First, Bill suggested that for insurance compa-nies, a more appropriate approach to risk meas-urement should be three-dimensional: “Insteadof thinking about risk as some feature of a prob-ability distribution at some particular point intime, an insurer shouldthink about a surface con-structed from a probabilitydistribution of outcomesone day from now, behindwhich is another distribu-tion for outcomes two daysfrom now, and so on, extend-ing out to infinity. The re-sult is a three-dimensionalsurface, where the dimen-sions are percentagechange in value on the hori-zontal axis, probability on the vertical axis andtime on the axis projecting into the future.“What is needed is a more adequate measure ofrisk that reflects the time dimension.

Second, Bill argued that looking at risk metricsfrom such time prospective brings the impor-tance of strategy into clearer focus. Enterpriserisk management is not just enterprise riskmeasurement. One way ERM can benefit seniormanagement is by assisting them in identifyingappropriate strategies for responding to eventsor conditions that potentially affect the value ofthe firm. Ultimately, by making appropriatestrategy choices, managers can change theshape of the three-dimensional surface.

Third, the ability to manage this three-dimen-sional surface should likewise enable man-agers to enhance shareholder value. Billstressed the importance of creating valuationmodels able to link the shape of the three-di-mensional risk surface to the market value ofthe firm. At the present time, such valuationmodels are relatively primitive.

In his summary, Bill identified the following op-portunities for managers to make ERM a clearvalue-adding tool for executive decisions:

• focus on the time dimension of outcome distributions;

• identify ways in which this return surface canbe altered by adopting appropriate corporatestrategies; and


◗


T he Risk Management Task Force con-sists of several subgroups, and othersub-subgroups or grouplets. Below are

brief summaries from six of the groups. More in-formation on all subgroups can be found on theSociety Web site.

Health Risk ManagementSubgroupThe Health Risk Management (HRM) subgroupof the RMTF was formed to ensure that thehealth area was represented as risk manage-ment at the SOA evolved. There are manyareas in health insurance that correspond tothose in life insurance but that do not have thesame degree of importance. One of the best ex-amples is interest rate risk. Another exampleis provider networks, which do not have a paral-lel in life insurance. At its inception, the HRMsubgroup took a poll to determine its direction.Three major areas emerged: the need for a spe-cialty guide; the need to focus on solvency is-sues unique to health coverage; and the need toimprove models used by health actuaries.

Rajeev Dutt of Milliman USA is in charge of thegroup putting together the specialty guide.This group has already developed a health riskmapping process which is available on theHRM section of the Risk Management Web site.

Trevor Pollitt of American Republic InsuranceCo. heads the charge on solvency. His group isputting together an RFP to have research done onthe risks that provider networks pose to solvency.

Doug Fearrington of Anthem is leading the ef-fort to improve our models. He is devoting hisefforts to applying stochastic techniques tohealth insurance models. He will be speakingon this topic at the spring meeting in Anaheim.If anyone is interested in joining any of thesegroups please contact me or the appropriategroup leader.John StarkActuarial DivisionAnthem Blue Cross and Blue Shield([email protected]

Modeling Equity Risk SubgroupBefore the Risk Management Section existed,there was a special task force on RiskManagement that was formed and headed up byDavid Ingram in 2001. One of the first activi-ties of this task force was to create some workinggroups with various risk topics to address on be-half of, and through the extensive resources of,the SOA membership.

The working group on modeling equity risk firstmet in May 2002 with the following mandate:

“Prepare a guide to modeling and managing eq-uity risk. Such a guide would include explana-tions of modeling tools and techniques,discussions of limitations and applications ofthese tools, review of equity-based products andrisks, commentary on tools and approaches tomanage equity risk and descriptions of hedgingtechniques.”

Accordingly the group set out to prepare an an-notated reading list, covering four main top-ics—theoretical background, modelingreturns, measuring risk and managing risk forequity-based products commonly offered in thelife insurance industry. The reading list has nowbeen prepared and edited and is available to theSOA membership on the Web site athttp://rmtf.soa.org/rmtf_em.html.

The work of modeling equity risk is an ongoingchallenge to the profession, and practitionersare urged to contact David Ingram or ValentinaIsakina at the Society of Actuaries([email protected]) if they wish to carry on thetorch. Topics would include more analysis ofequity models and their use, the generation oflong-term equity returns for models, historicalversus arbitrage-free models and comparison ofinsurance markets to the options and futuresmarkets. Possible initiatives could include thedevelopment of a prototype model for use by theSOA membership.

While many SOA members were correspon-dents for the working group on equity modeling,


Risk Management Task Force Subgroup Updates

◗

special thanks are owed to the following groupmembers who were active participants at vari-ous points in time in developing the work prod-uct: Michael Bean, Kitty Ching, Andy Chua,Martin Hall, Geoff Hancock, Charles Hill,Novian Junus, Martin le Roux, Rob Stone andDarin Zimmerman.—Josephine Marks

Risk Metrics SubgroupThe Risk Metrics Subgroup is led by Fred Tavan who can be reached [email protected]. It is made upof members from both the SOA and CAS. Itmeets every three weeks through a conferencecall organized by the SOA. The main objectiveof the Risk Metrics Subgroup is to identify dif-ferent metrics that can be used to quantify thelevel of various risks across an organization.The group’s scope goes beyond financial riskssuch as credit, market and insurance risks intooperational, reputational, regulatory and litiga-tion risks in order to take an enterprise view ofrisk. There are a variety of risk metrics avail-able on financial risks, and this is typically agood place for companies to start building theirERM program, however, the actuarial profes-sion also needs to educate itself on other types ofrisk that have an impact on financial institu-tions. One of the main goals of the Risk MetricsSubgroup is to facilitate this educationalprocess for members of the SOA and CAS.

The subgroup has developed a number of paperson various risk metrics, and these are posted onthe Web at http://rmtf.soa.org. Once at theWeb site the user needs to select Subgroup and

then Risk Management Metrics to see the list ofpapers.

The subgroup is currently working on develop-ing a risk metrics database that allows a user toquery and report on different risk metrics thatcan be used for a specific risk subcategory with-in a predefined risk management framework.The federal reserve risk management frame-work with AAA extension has been chosen asthe risk management framework for doing themapping, however, the database also containsthe COSO framework and theIAA framework. About 60 differ-ent risk metrics have been iden-tified so far and work is ongoingto add to this list. The databasewill also ultimately contain ref-erences to various sources ofreading material on how eachrisk metric can be applied inmeasuring risk within an organi-zation. These tools will allow auser to identify quickly whichrisk metrics can be relevant totheir work and then learn abouthow to start using them in prac-tice. The database is a work inprogress and thanks to all thevolunteers who have helped ingetting it off the ground. Special thanks go to thekey contributors to the database so far and theseindividuals are: Cliff Angstman (BerkshireLife), Ellen Hall (ING), David Ingram(Milliman), Julie Perks (Sun Life), David Ruhm(The Hartford) and Fred Tavan (Canada LifeRe.).



◗

• identify how the consequent changes in thisreturn surface affect shareholder value.

Next Steps for the SymposiumThe attendance, the quality of the sessions, theprofoundness of the issues discussed and thevariety of different perspectives offered at thesymposium—all confirmed that ERM is here tostay and that the subject is worthy of more sym-posiums to come.

The 2005 ERM Symposium OrganizingCommittee has been formed, and it is pleased toannounce that the tradition will continue: the2005 ERM Symposium will be held in Chicagoin late April.

Mark your calendars and bring your ideas!!! ✦

2004 ERM Symposium◗ continued from page 7

Credit Risk Management—Dave Ingram

Economic Capital—Jenny Bowen

Enterprise Risk Management—Mark Shaw

Equity Modeling—Josephine Marks

Extreme Value Models—Tom Edwalds

Health Risk Management—John Stark

Policyholder Behavior in the Tail—Jim Reiskytl

Pricing for Risk—Novian Junus

Risk-Based Capital Covariance— Jim Reiskytl

Risk Management Metrics—Fred Tavan

Subgroups

New volunteers are encouraged to join the sub-group. You may join the Risk Metrics Subgrouplistserve by contacting Julie Young at the SOA([email protected]). The listserve provides itsmembers with the conference call dates/times,meeting materials, and minutes to the meetings.—Fred Tavan

Economic Capital SubgroupThe Economic Capital Calculations andAllocation (ECCA) subgroup has been very pro-ductive in its two years of existence. It has beena vital forum for the exchange of views and newsregarding both the development of economiccapital (EC) as a concept and its acceptance inpractical settings.

Under the leadership of Hubert Mueller, the ECCA subgroup produced an ECCA spe-cialty guide, which can be accessed from theECCA Web page at http://www.soa.org/sections/rmtf/rmtf_ecca.html. The spe-cialty guide is an overview of several aspects of economic capital. It is a great startingpoint, but it also situates economic capital in thecontext of its many possible uses, going so far aseven offering alternative definitions that arecurrently in use.

The body of the document is very straightfor-ward, while the annotated bibliography pro-vides thoughtful guidance to those seeking morein-depth treatment of particular topics. Thebody of the document treats economic capitalprimarily from a practical viewpoint, while thebibliography includes resources that can offermore rigorous theoretical treatment. Many peo-ple contributed to the review of the literaturecovered in the bibliography. The annotationsoffer meaningful guidance regarding how help-ful each publication might be toward meeting par-ticular needs. Brett Roush not only contributed tothe annotations, but also brought the bibliographyto life as a high-quality, finished product.

It is no surprise that the ECCA specialty guidehas been included in the syllabus for the newSOA Risk Management exam, which will be offered this fall under the Investment Track.

We included non-actuaries in the ECCA sub-group, particularly representatives of all majorrating agencies, whose insights and updates ex-panded members’ perspectives. It also becameapparent that all rating agencies are very inter-ested in developments in this area.

We found that there were many people thirstingfor knowledge about practical approaches toECCA. We agreed that it was difficult to extractfrom the available literature specific method-ologies that could be generalized or even com-pared. It seemed as though each demonstrationwas either too specialized and included toomany non-transferable assumptions or was justtoo complicated to break down into comparableelements.

We attempted to enlist the help of practitionerswho would apply their ECCA methodology torelatively simple sample case studies, so thatthe differences could be compared and ex-plained. What we found was that most practi-tioners either had not yet found just the rightmethod to meet their own needs, or were not yetready to share them.

We have concluded that the discipline of ECCAis still too immature for us to find meaningful, de-pendable “best practices” that we can analyzeand share in an effort to help actuaries bridge thegap between theory and the “real world.”Therefore, we have decided to disband our sub-group for now, with the idea that in 12 to 18months there may have been sufficient progressin this area to support a survey of best practices.

As evidence that EC is a vital topic that contin-ues to gain attention, Hubert and Brett have pro-vided the following updates regarding recentregulatory efforts to determine risk-based capi-tal, using proprietary company models consis-tent with the determination of EC.—Jenny Bowen

Extreme Value Models (EVM) The main objective of the Extreme ValueModels (EVM) subgroup of the RiskManagement Task Force (RMTF) is to enhancethe knowledge base of the actuarial professionconcerning extreme event risks. Extremeevents have very low frequencies (e.g., once-a-century) but extraordinarily high costs. In


The Economic Capital

Calculations and

Allocation (ECCA)

subgroup has been

very productive in its

two years of existence.

It has been a vital

forum for the exchange

of views and news

regarding both the

development of

Economic Capital (EC)

as a concept and its

acceptance in practical

settings.

“

”

RMTF Subgroup◗ continued from page 9

◗

pursuing this objective, the EVM subgrouphas two goals: • to increase the actuarial profession’s aware-

ness of these extreme risks and of the pitfallsof using simplistic methods to assess theserisks; and

• to provide education and tools needed toquantify, manage and price the risks associ-ated with extreme-valued outcomes.

At the recent Bowles Symposium on Enterprise Risk Management, jointly spon-sored by Georgia State University, the Society ofActuaries and the Casualty Actuarial Society,the EVM subgroup organized an embeddedmini-seminar on Extreme Value Theory.Professor H. N. Nagaraja, Ph.D., of Ohio StateUniversity and Steve Craighead, FSA, of Nationwide Insurance presented threebreakout sessions, instructing attendees about actuarial applications of the theory oforder statistics, the three basic extreme valuemodels and their generalization and methods ofinference using extreme value models.Handouts from these sessions, including a spreadsheet illustrating the techniques covered, are available on the Web athttp://www.casact.org/coneduc/erm/2004/handouts/.

Other resources on extreme risks, including theessays entered into the X-treme Actuary contestsponsored by the EVM subgroup last summer,can be found on the EVM subgroup page of theRMTF section of the Society of Actuaries Website, http://rmtf.soa.org/rmtf_evm.html. —Tom Edwalds

Pricing for Risk The stated goals of the Pricing for Risk sub-group (PFR) are to:

• Evaluate the effectiveness of different pric-ing techniques as to their ability to captureand quantify the risks associated with thesale and administration of life and annuityproducts

• Document and provide guidance to actuar-ies for when a given technique or measuremay be appropriate and the limitations of itsuse

In fall of 2002 the PFR, under the leadership of Todd Henderson, performed a survey of howactuaries currently reflect risk in pric-ing. The actuaries surveyed are mem-bers of the Investment and IndividualLife and Annuity Product Developmentsections. The completed survey and re-sults can be found in the PFR section ofthe Risk Management Task Force(RMTF) Web site. The general observa-tion is that asset-related risk has a high-er tendency to be modeled, butliability-related risk is more often stresstested; relying more heavily on judg-ment in assessing risk.

The PFR’s current proposed objective is to de-velop a specialty guide that:

• Identifies the common profit measures usedin the insurance industry and the commonmethods for reflecting risk

• Provides actuaries with relevant informationon those risks that have not been traditional-ly considered in their pricing models

• Include recommended readings that provideindications of how to price each of the risksand how they should be incorporated direct-ly and interactively or independently withtraditional actuarial risk models

• Identifies gaps in current literature andpractice with regard to pricing for risk.

The PFR is currently discussing the scope of thespecialty guide; whether it is in the form of anannotated reading list or as comprehensive asthe ALM specialty guide. We are also lookinginto the possibility of funding a research projectto develop the specialty guide that the PFR willdirect.

The stated goals of the subgroup seem rather daunting and the current objective is a step toward achieving those goals. The results ofthe survey and the regulatory movement toward stochastic reserving and capital (C3 RBCPhase II, AG 39, etc.) indicate that the goals are relevant. If you feel the same and are interested in participating please e-mail [email protected]. ✦—Novian Junus


◗

◗


Risk Management Skills and Aptitudesby H. Felix Kloman

Reprinted by permission from RiskManagement Reports, Volume 31, Number5 May 2004. ©Copyright 2004, by H. FelixKloman and Seawrack Press, Inc.

Skills and Aptitudes

What are the skills we need in the fu-ture? I heard this question twice inthe past eight months, first from a

parent of a junior enrolled in a sailing programin Tenants Harbor, Maine, and later from a riskmanager in Seattle, Wash. The answer, I think,is found not so much in specific skills as in apti-tudes.

Take sailing, for example. The skills that weteach and that are required to handle a smallboat successfully in various types of weatherand sea conditions include being able to swim,knowing the parts of a boat and its rigging, un-derstanding how sails work, sailing a boat bothupwind and downwind, knowing what to do inthe event of a person overboard or a capsize,leaving and returning from docks and mooring,tying the correct knots and understandingweather, tides and currents. These are specificand necessary skills that are easily taught. Moreimportant, however, are the aptitudes that serveas the foundation for these skills. Independenceis the first: the willingness to step out on yourown. Patience is the second: understanding thata sailboat cannot go directly upwind, nor can itmove when there is no wind. And third is team-work: sailing and racing a small boat requiresexquisite timing and cooperation in order to dowell. Without these three aptitudes, a sailor lit-erally may be lost at sea.

My sailing analogy applies equally to the disci-pline of risk management. Again, independ-ence comes first. In its current evolution as anintegrated and strategic process throughout anyenterprise, its “champion” and guide must beindependent of conventional staff and operatingfunctions. Too many organizations attempt toforce risk management into finance where it be-comes both dependent and restricted.Independence begins with a fresh and broadview of “risk” itself. It is not, as too many safety,

finance and insurance practitioners construe it,merely a “chance of loss.” It must be viewed asencompassing the unexpected, both favorableand unfavorable. Risk is “a measure of the pos-sibility of unexpected outcomes.” Under thisdefinition risk management becomes “a disci-pline for dealing with uncertainty,” a far morestrategic approach than as construed by the nar-row confines of finance, insurance, safety, qual-ity control and business recovery planning.Risk management independence thus requiresa leader who has a direct reporting relationshipto both the CEO and the organization’s govern-ing board. Only in this way can that leader raiseunpopular and even dangerous risk issues,those risk issues that are truly material to the fu-ture of the organization.

As an example, the most pressing current issueis that of excessive executive compensation.Too many organizations have allowed their sen-ior management reward systems to skyrocketout of control, to obscene levels. CEOs are natu-rally unwilling to take action and compliantboards exacerbate the problem. The result: reg-ulators, shareholders and stakeholders loseconfidence in management. We need chief riskofficers who are both able and willing to addressthese and similar, larger strategic issues andwho, at the same time, can present these issuesintelligently and dispassionately to criticalboard committees. Otherwise, we will continueto focus on relatively minor risks to the exclu-sion of those that materially affect our futures.As David Godfrey, the CRO for SwissReinsurance Company, said recently, “Andfrom time to time you (the CRO) need the abilityto say, ‘I’m sorry, but I don’t agree with what yousay.’ If you (the CRO) only report to the CEO, it’svery difficult to go beyond that in order to ex-press disagreement, if the channels aren’t therealready to do so.” (See “ERM, Operational Riskand Risk Management Evolution,” in GARPRisk Review, March/April 2004).

Lesley Daniels Webster, the executive vicepresident and global head of market risk atJPMorganChase, echoed this theme when sheemphasized the necessity of a CRO having the

Felix Kloman is editor

and publisher of

Risk Management Reports.

He can be reached at

[email protected]


My second aptitude,

drawn from sailing, is

patience. When the

wind isn’t favorable,

you may have to

anchor and wait for

it to change.

“

”

“willingness and ability to criticize senior andoperating management when required”(speech at the GARP 2004 Conference,February 25, 2004).

The Economist stated the issue of trust and inde-pendence most succinctly in its April 22, 2000issue (I quoted it earlier in RMR April 2001):“There may be two good reasons for companiesto worry about ethical behavior. One is anticipa-tion: bad behavior, once it stirs up a public fuss,may provoke legislation that companies willfind more irksome than self-restraint. The other,more crucial, is trust. A company that is nottrusted by its employees, partners and cus-tomers will suffer.”

Independence of risk management is necessaryto permit and stimulate both strategic perspec-tive and the courage to speak out when required.It is an aptitude that transcends specific skills.

My second aptitude, drawn from sailing, is patience. When the wind isn’t favorable, youmay have to anchor and wait for it to change. Forcenturies the Chinese used bamboo as a comparable example. In a storm the bambooshaft bends but doesn’t break, springing back toits normal position when the winds subside.Patience implies a long-term view of an organi-zation and its future. One of the most perniciouscurrent problems is the overfocus, even para-noia, on “shareholder value” and near-termstock prices. We have succumbed to a massfrenzy trying to outdo each other in managedearnings and artificial stimulation of the dailyprices posted in New York, London, Frankfurtand Sydney. The patient CRO understands thelong-view of an organization’s responsibility toits stakeholders, including shareholders, onethat may reach out as far as 20 to 30 years.Patience means revising the goal of risk man-agement (and the organization itself) to “build-ing and maintaining stakeholder confidence.”“Shareholder value” is only a piece of this equa-tion, with all respect to the University ofChicago theories of economic practice.

If a CRO accepts this basic thesis, then it followsthat the three basic objectives of risk manage-ment must be:

• Credibility: Communicating the nature ofrisks, both favorable and unfavorable, with

stakeholders, and their responses, to enhancethe support of these groups for the organiza-tion.

• Resilience: Building an internal and exter-nal flexibility so that the organization can re-spond to whatever unexpected event mayoccur, and in many cases actually taking ad-vantage of a downside event to improve mar-ket position.

• Perspective: Countering the prevailingover-focus on the short-term. Here PeterSchwartz’s The Art of the Long View (Doubleday,1991) remains one of the best expositions oflong-term perspective.

Patience, however, has an Achilles heel. Most ofthe prevailing metrics for measuring the suc-cess or failure of a risk management function arecast in short-term numbers. VaR is one of these,and it, like many others, is flawed. No one hasyet developed a consistent and accepted metricfor measuring the longer-term results of riskmanagement. We need one and we may be con-demned to the short-term until and unless wecan create a new measure.

My third aptitude is teamwork. Because tacticalrisk management embodies so many differentskills, it makes good sense for its practitioners toreach out and try and understand the problemsand solutions of others. While we are makingsome progress within organizations towardbreaking down the artificial barriers that kept usfrom communicating with one another, too manyof our major associations of risk managementplayers continue to operate behind impregnablefortresses. Most are unable,even unwilling, to bring rep-resentatives of their counter-part groups to their annualconferences and local chap-ter meetings. The result is anappalling lack of knowledgeof the work of others. LastDecember, I asked an audi-ence of some 40 members ofthe Society for Risk Analysishow many had even heard ofGARP, Professional Risk Manager’sInternational Association (PRMIA) or RIMS.Two hands were raised. I questioned registrants


◗

at the February 2004 GARP meeting: few hadheard of SRA or RIMS. Then, at a RIMS chaptermeeting in Seattle in March, I asked the samequestion of over 100 registrants. Only one wasalso a member of GARP; none were members ofPRMIA or SRA. Many members of one associa-tion had not even heard of the other groups. Thisis the worst sort of parochialism.

Many specialist skills are required for riskanalysis, the first step in the process (the identi-fication of possible unexpected events; theirmeasurement in terms of likelihood, timing,consequences and public perception; and theirassessment relative to an organization’s objec-tives). They include scenario analysis, quanti-tative and probabilistic analysis, actuarialscience, data management, knowledge of thelaw, econometric modeling, intuition and theuse of heuristics and, of course, the value of ex-perience. Similarly, another set of skills is em-ployed in risk response, the second step in theprocess (controls adopted to balance upside anddownside risk; measuring and monitoring per-formance; and communicating with stakehold-ers). These skills include knowledge of safetyand quality systems (Six Sigma), audit and ac-counting controls, environmental controls, be-havioral economics (financial incentives andpenalties), contingency and crisis management(business recovery planning), and financing(credit, derivatives, hedging, pooling, use ofcapital markets, insurance and claims manage-ment). It is too much to ask any one person to befully conversant and expert in all these fields.This makes teamwork the mandatory aptitude.It is high time that the IIA, GARP, PRMIA,RMA, CAS/SOA, SRA, RIMS/InternationalFederation of Risk and Insurance ManagementAssociations (IFRIMA) and American Societyof Safety Engineers (ASSE), among others,cease their guild-like restrictiveness and reachout to their counterparts, expanding the scope ofour discipline.

Swiss Re’s David Bothwell addressed the ques-tion of skills in a similar fashion: “They (risk of-

ficers) have to have skills that are seen to be rel-evant and at a high level. They have to be seen tobe balanced, to look at the total picture, assess-ing the opportunity, which the deal-doer istelling you is the greatest thing since slicedbread, while at the same time balancing thatwith the broader picture. Risk managers have tobe able to articulate well their reasoning for aparticular position or view-point. Risk man-agers have to be consistent—or they will lose re-spect, but in the final analysis, they ultimatelyhave to be prepared to stand up and say no.”

The major challenge for any risk managementteam is the prevailing failure to communicateintelligently and coherently with all of ourstakeholder groups. Last month (April 2004), Idescribed the Bank of Montreal’s exceptionaleight-page description of its internal risk man-agement program. Too few organizations at-tempt even this. I know of no organization thatemploys a consistent and effective continuingtwo-way dialogue with its stakeholder groups onits analysis of risks and its responses. Perhapsimproved teamwork among the existing riskmanagement groups can develop a better meansof communication.

Academic institutions are a critical part of theteamwork equation. More are beginning tostretch their formerly narrow programs (fi-nance, insurance, public policy, engineering) toincorporate ideas and methods from the othersub-disciplines. I hope that many of the associ-ation-run certification programs will also ac-knowledge their competitors and expand theircurricula to include, as least nominally, otherideas and techniques.

Independence, patience and teamwork arethree critical aptitudes for those who purport topractice this evolving discipline of risk manage-ment. Within them one can develop other tech-nical skills; without them, these skills aremeaningless. ✦


Skills & Aptitudes◗ continued from page 13

◗

The major chanllenge

for any risk management

team is the prevailing

failure to communicate

intelligently and

coherently with all of

our stakeholder groups.

“

”

In January 2004, Chairman AlanGreenspan spoke to the AmericanEconomic Association about “Risk and

Uncertainty in Monetary Policy.” In thisspeech, Chairman Greenspan referred to risk management as a central element in the conduct of monetary policy in the United States. The complete text is athttp://www.federalreserve.gov/boardocs/speeches/2004/20040103/default.htm and is easy and insightful reading.

I attended the talk and thought it would be in-teresting to our section membership to hearhow the Fed Chairman views risk management.As we define and delineate the scope of riskmanagement in the actuarial profession andour traditional industries and seek to expandour influence beyond the traditional bound-aries, we need to remember that many other in-dustries also use “risk management” todescribe their activities. We should under-stand how we view risk management similarlyto and differently from others. In this note Iquote Chairman Greenspan liberally and pro-vide my interpretation of his comments.

Chairman Greenspan gave his talk in a hotelballroom in front of several hundred academic,government and private sector economists. Hebegan by recounting the “key developments ofthe past decade and a half of monetary policy inthe United States from the perspective of some-one who has been in the policy trenches.” Hediscussed major policy events and decisionsduring his tenure and during that of his prede-cessor, Chairman Paul Volcker.

Midway through his talk, he began to explain ingeneral why decisions were made—that theywere an application of risk management.

“As a consequence, the conduct of monetarypolicy in the United States has come to in-volve, at its core, crucial elements of risk man-agement. This conceptual frameworkemphasizes understanding as much as possi-ble the many sources of risk and uncertainty

that policymakers face, quantifying thoserisks when possible, and assessing the costs as-sociated with each of the risks. In essence, therisk management ap-proach to monetary poli-cymaking is anapplication of Bayesiandecision making.”

It is noteworthy thatChairman Greenspan em-phasizes risk managementas a conceptual frame-work, not as a task or mod-eling exercise. It is a wayof formulating and ad-dressing questions, a way of thinking that we,as actuaries, do as part of our professional ex-istence, with or without formally recognizingit.

Our Bayesian decision making may not be doneformally, with priors and posteriors and conju-gate distributions, but we all intuitively putmore or less weight on recent experience insome proportion to the extent of our prior expe-rience. And I presume that is what policymak-ers at the Fed do also.

Greenspan next discussed the notion of strate-gy as applied to risk management.

This framework also entails devising, in lightof those risks, a strategy for policy directed atmaximizing the probabilities of achievingover time our goals of price stability and themaximum sustainable economic growth thatwe associate with it. In designing strategies tomeet our policy objectives, we have drawn onthe work of analysts, both inside and outsidethe Fed, who over the past half century havedevoted much effort to improving our under-standing of the economy … A critical resulthas been the identification of a relativelysmall set of key relationships that, taken to-gether, provide a useful approximation of oureconomy’s dynamics….

◗


A Case Study in Risk Management: U.S. Monetary Policyby Henry McMillan


Henry McMillan, FSA, is

senior vice president of the

institutional products division

at Pacific Life Insurance

Company in Newport Beach,

Calif and a member of the Risk

Management Section Council.

He can be reached at

henry.mcmillan@

pacificlife.com

Risk management involves strategy, not merelyrisk assessment, which involves measurement.Objectives and measurement are all part of theprocess, and the process is inherently an ap-proximation.

However, despite extensive efforts to captureand quantify what we perceive as the keymacroeconomic relationships, our knowledgeabout many of the important linkages is farfrom complete and, in all likelihood, will al-ways remain so. Every model, no matter howdetailed or how well designed, conceptuallyand empirically, is a vastly simplified repre-sentation of the world that we experience withall its intricacies on a day-to-day basis.

We have enough problems trying to model poli-cyholder behavior. Imagine trying to think ofeverything that everyone does. You come to ac-cept that it’s not the size of your model, but thehuge uncertainty surrounding what you’re try-ing to model.

Earlier in his talk Greenspan had referenced acommonly used distinction between risk anduncertainty due to Frank Knight, an economistactive in the ’20s and ’30s.

The term “uncertainty” is meant here to en-compass both “Knightian uncertainty,” inwhich the probability distribution of outcomesis unknown, and “risk,” in which uncertaintyof outcomes is delimited by a known probabil-ity distribution. In practice, one is never quitesure what type of uncertainty one is dealingwith in real time, and it may be best to think ofa continuum ranging from well-defined risksto the truly unknown.

In our models we have to define and use a distri-bution, even though we may not know that it isthe correct distribution. We shouldn’t losesight of the fact that quantified may only meanthat some quantity has been determined, con-ditional on the assumed underlying distribu-tions, behavioral assumptions and institutionalarrangements, which may be unknown or un-certain in the above sense.

Greenspan then turns to how one makes deci-sions if everything is not known.

Given our inevitably incomplete knowledgeabout key structural aspects of an ever-chang-ing economy and the sometimes asymmetriccosts or benefits of particular outcomes, a cen-tral bank needs to consider not only the mostlikely future path for the economy but also thedistribution of possible outcomes about thatpath. The decision makers then need to reachjudgment about the probabilities, costs andbenefits of the various possible outcomes underalternative choices for policy.

One should not be surprised to hear an econo-mist use the phrase “costs and benefits” in apublic address. Indeed there was an almostpalpable relief among the audience whenGreenspan anchored his ship to the solid rockof economic orthodoxy.

Chairman Greenspan did not say that stochas-tic analysis is necessary for effective monetarypolicy. However, it is quite clear that single-path analysis is not sufficient for monetary pol-icy. Actuaries would probably agree that singlepath analysis is not adequate for pricing insur-ance products or for selecting investmentstrategies to back pension liabilities. WhatGreenspan is leading up to is that the extent ofuncertainty and risk causes us to choose ac-tions that would not be optimal if we knewmore—more about the true nature of the econ-omy and more about where the economy isheaded in the absence of the Federal Reserve’sactions.

A policy action that is calculated to be optimalbased on a simulation of one particular modelmay not, in fact, be optimal once the full ex-tent of the risk surrounding the most likelypath is taken into account. In general, differ-ent policies will exhibit different degrees of ro-bustness with respect to the true underlyingstructure of the economy.

For example, policy A might be judged as bestadvancing the policymakers’ objectives, con-

◗


U.S. Monetary Policy◗ continued from page 15

We shouldn’t lose sight

of the fact that quanti-

fied may only mean that

some quantity has been

determined, conditional

on the assumed underly-

ing distributions, behav-

ioral assumptions and

institutional arrange-

ments, which may be

unknown or uncertain in

the above sense.

“

”

ditional on a particular model of the economy,but might also be seen as having relatively se-vere adverse consequences if the true structureof the economy turns out to be other than theone assumed. On the other hand, policy Bmight be somewhat less effective in advancingthe policy objectives under the assumed base-line model but might be relatively benign inthe event that the structure of the economyturns out to differ from the baseline.

One insight that I take from these comments isthat you need to consider “what if” your modelis wrong. A particular investment or hedgingstrategy might appear very appealing in onemodel, but if the strategy is not robust, if it is toosensitive to parameter estimates or modelstructure, then it is not likely to turn out to be asgood as you thought it would be. For example, astrategy involving multiple long and shorts offutures and options contracts to offset interestrate risk and volatility could be quite wrong orineffective if the hedging ratios or positions aretoo highly dependent on a covariance matrixthat is estimated with error.

In the next few paragraphs Greenspan makesstatements suggestive of tail risk analysis thatis increasingly used among actuaries.

As this episode illustrates, policy practition-ers operating under a risk management par-adigm may, at times, be led to undertakeactions intended to provide insuranceagainst especially adverse outcomes.Following the Russian debt default in the au-tumn of 1998, for example, the FOMC easedpolicy despite our perception that the econo-my was expanding at a satisfactory pace andthat, even without a policy initiative, it waslikely to continue doing so. We eased policybecause we were concerned about the low-probability risk that the default might triggerevents that would severely disrupt domesticand international financial markets, withoutsized adverse feedback to the performanceof the U.S. economy.

The product of a low-probability event and apotentially severe outcome was judged a moreserious threat to economic performance thanthe higher inflation that might ensue in themore probable scenario.

Such a cost-benefit analysis is an ongoingpart of monetary policy decision making andcauses us to tip more toward monetary easewhen a contractionary event, such as theRussian default, seems especially likely or thecosts associated with it seem especially high.

A parallel to the inflation/liquidity crisisdilemma might be short-runGAAP earnings versus longerterm statutory solvency facingrisk managers choosing to usederivatives in an accountingchallenged world.

Perhaps another view is thatwe get so caught up in themore likely events that we un-derestimate or ignore the low-probability events that canand do occur. Greenspan fur-ther notes that human behavior changes in cri-sis events, something that should beacknowledged and incorporated into the riskmanagement process.

The 1998 liquidity crisis and the crises associ-ated with the stock market crash of 1987 andthe terrorism of September 2001 prompted thetype of massive ease that has been the historicmandate of a central bank. Such crises areprecipitated by the efforts of market partici-pants to convert illiquid assets into cash.When confronted with uncertainty, especiallyKnightian uncertainty, human beings invari-ably attempt to disengage from medium tolong-term commitments in favor of safety andliquidity. Because economies, of necessity, arenet long—that is, have net real assets—at-tempts to flee these assets cause prices of equi-ty assets to fall, in some cases dramatically. Inthe crisis that emerged in the autumn of 1998,pressures extended beyond equity markets.Credit-risk spreads widened materially andinvestors put a particularly high value on liq-uidity, as evidenced by the extraordinarilywide yield gaps that emerged between on-the-run and off-the-run U.S. Treasuries.

We typically allow for changing policyholderbehavior as gaps between credited rates andmarket rates expand, but do we also allow for

◗



changing behavior in crisis situations? It seemsthat knowing what is in the model, and what isleft out, combined with more qualitative think-ing is all that can be done and what must bedone.

Chairman Greenspan then turns to a discussionrelating models to risk management, and therole of policy rules in risk management.

The economic world in which we function isbest described by a structure whose parametersare continuously changing. The channels ofmonetary policy, consequently, are changingin tandem. An ongoing challenge for theFederal Reserve—indeed, for any centralbank—is to operate in a way that does not de-pend on a fixed economic structure based onhistorically average coefficients. We often fitsimple models only because we cannot esti-mate a continuously changing set of parame-ters without vastly more observations than arecurrently available to us. Moreover, we recog-nize that the simple linear functions underly-ing most of our econometric structures maynot hold outside the range in which adequateeconomic observations exist. For example, itis difficult to have much confidence in theability of models fit to the data of the moderateinflations of the postwar period to accuratelypredict what the behavior of the economy

would be in an environment ofaggregate price deflation.

Seems like there are a lot oflessons here for risk manage-ment in insurance companiesand pension plans. We mustexpect change in estimatedparameters and in behaviorand in legal environments andso on. Model estimates from aprevious era may not fit thecurrent era. Calibration of a

model to historical data may inadvertentlystraitjacket risk management if we do not allowthe future to look different from the past.

But if the past is not prologue, what are we todo? Greenspan offers his pragmatic approach.

In pursuing a risk management approach topolicy, we must confront the fact that only alimited number of risks can be quantified withany confidence. And even these risks are gen-erally quantifiable only if we accept the as-sumption that the future will, at least in someimportant respects, resemble the past.Policymakers often have to act, or choose notto act, even though we may not fully under-stand the full range of possible outcomes, letalone each possible outcome’s likelihood. As aresult, risk management often involves signif-icant judgment as we evaluate the risks of dif-ferent events and the probability that ouractions will alter those risks.

So does the future resemble the past? It’s a hardquestion for an economic policymaker. It isperhaps even harder for an individual riskmanager because the financial markets dependon the behavior of the economic policymakertoo. An investor’s world changes if the FederalReserve changes its behavior, regardless ofwhether the economic world is fundamentallychanged. That is, an investor must consider notonly whether the world has changed, but alsowhether the Federal Reserve’s view of theworld has changed.

Certainly company executives must make deci-sions before our models are final or perfected oreven ready for prime time. Judgments are partof real life, and must be factored into the riskmanagement process, even if only qualitativelyor judgmentally.

Greenspan notes that rule-based behavior hasvalue, but can be overvalued if evaluated in thecontext of theoretical models rather than actualpolicy worlds.

Some critics have argued that such an approachto policy is too undisciplined—judgmental,seemingly discretionary and difficult to ex-plain. The Federal Reserve, they conclude,



◗

should attempt to be more formal in its opera-tions by tying its actions solely, or in the weak-er paradigm, largely, to the prescriptions of asimple policy rule…. And the prescriptions offormal rules can, in fact, serve as helpful ad-juncts to policy, as many of the proponents ofthese rules have suggested. But at crucialpoints, like those in our recent policy history—the stock market crash of 1987, the crises of1997-98, and the events that followedSeptember 2001—simple rules will be inade-quate as either descriptions or prescriptions forpolicy. Moreover, such rules suffer from muchof the same fixed-coefficient difficulties wehave with our large-scale models.

So many investment houses attempt to modelthe Fed’s behavior, and thereby gain an edge intheir investment strategy. It’s not clear whetherthey want the Federal Reserve’s behavior to bemore transparent—thereby making the fore-casting of Federal Reserve behavior easier—orto be even more clouded—thereby making thevalue of a correct forecast just that much higher.

At any rate, Greenspan appears to feel thatrules are good, but meant to be broken.Judgment must rule the day.

To be sure, sensible policymaking can be ac-complished only with the aid of a rigorous an-alytic structure. A rule does provide abenchmark against which to assess emergingdevelopments. However, any rule capable ofencompassing every possible contingencywould lose a key aspect of its attractiveness:simplicity. On the other hand, no simple rulecould possibly describe the policy action to betaken in every contingency and thus provide asatisfactory substitute for an approach basedon the principle of risk management.

At this point Greenspan provides some ammu-nition for those of us who want some disciplinein risk management and its associated model-ing. In short, we must remember his audience:academic economists, a group that might makean SOA Section Council seem exciting.Greenspan has been saying that models andrules, by themselves, are not the answer. Oneshould not, therefore, conclude that no modelsand no rules are the answer. Greenspan wants

good models and good rules, but also judgmentand discretion. Undoubtedly, he feels that hehas something to contribute to the risk manage-ment process known as monetary policy.

As I indicated earlier, policy has worked off arisk management paradigm in which the riskand cost-benefit analyses depend on forecastsof probabilities developed from large macro-models, numerous sub-models, and judg-ments based on less mathematically preciseregimens. Such judgments, by their nature,are based on bits and pieces of history thatcannot formally be associated with an analy-sis of variance.

Yet, there is information in those bits andpieces. For example, while we have been un-able to readily construct a variable that cap-tures the apparent increased degree offlexibility in the United States or the globaleconomy, there has been too much circum-stantial evidence of this critically importanttrend to ignore its existence. Increased flexi-bility is a likely source of changing structuralcoefficients.

So, could we say that not only does he rely ondata, but his knowledge of how data are pro-duced and collected. From that knowledge heand others at the Federal Reserve use qualita-tive adjustments to quantitative models andrecommendations to arrive at the actual policythat is implemented. He sums up this portion ofhis talk—that it’s not that we’re complex, it’sthat the world is complex.

Our problem is not, as is sometimes alleged,the complexity of our policymaking process,but the far greater complexity of a world econ-omy whose underlying linkages appear to becontinuously evolving. Our response to thatcontinuous evolution has been disciplined bythe Bayesian type of decision-making inwhich we have engaged.

Chairman Greenspan continued his remarkson the role of forecasting, inflation targetingand future challenges for policy makers. Heclosed the presentation with a question and answer period.



◗

So many investment

houses attempt to

model the Fed’s

behavior, and thereby

gain an edge in their

investment strategy.

“

”

◗


Influenza Pandemics: Are We Ready for the NextOne? What Actuaries Can Learn From 1918 By Max J. Rudolph

This article summarizes a presentation given atthe 2004 ERM Symposium by the author alongwith David Ingram. Opinions expressed are thoseof the author and not his employer or any otherorganization he is affiliated with. It is not a refer-eed paper and does not substitute for doing yourown research on the topic. The author is indebtedto the researchers who wrote the books referenced,to SOA librarian Ellen Bull (an underutilizedbut wonderful resource for actuaries) and toDavid Ingram, whose interest in the subject andinitial research helped to jump-start this project.

Introduction

T ail risk has gotten the actuarial profes-sion’s attention. With supposedly 10Sigma events occurring almost every

year and contagion risk consistently underesti-mated, best estimate assumptions using a singleeconomic scenario are not sufficient. Manypractitioners have known this for years, buttechnology now allows actuaries to model andmanage these risks. These models generally usea combination of stochastic and deterministicscenarios. Interest rates and equity returns areoften modeled deterministically to address spe-cific risks and stochastically to cover the uni-verse of possibilities. Risks as diverse asearthquakes and anthrax attacks can be similar-ly modeled.

Mortality risk is generally tested deterministi-cally, with specific concerns stress tested. Forlife insurance products, stresses of 110 percentof base mortality are commonly used for exter-nal projects like asset adequacy testing. Recentdiscussion has focused on the very real risk ofbioterrorism, with geographical concentrationsof exposure creating the risk. Hopefully therewill never be enough “data points” to fit a distri-bution for these risks, but some hypotheses canbe modeled. Either a dirty bomb or water poi-soning are examples that could present a majordisruption to a state, province or other geo-graphical region, but it is unlikely that such anevent would be either national or worldwide inscope. Even a natural disaster like an eruptionof Mt. Rainier or a large earthquake in

California would impact a limited region. Onlyan outbreak of a disease that is both highly con-tagious and lethal could affect global popula-tions materially. One might think, with today’smedical tools and research labs, that such anoutbreak could be easily contained. Whilemedical knowledge has improved greatly sincethen, in 1918 such an outbreak of disease oc-curred that remains untreatable even today. Itwas influenza.

This article, while not a formal book review,takes much of its information from three books.All are very readable and provide slightly differ-ent perspectives. America’s ForgottenPandemic, by Alfred W. Crosby and first pub-lished in 1976, follows the story throughAmerica’s cities as the virus spread and mutat-ed. It was the definitive book on the topic untilrecently. The second book, Flu: The Story of theGreat Influenza Pandemic of 1918 and theSearch for the Virus that Caused it, by GinaKolata (science editor for the New York Times)and published in 1999, describes the pandemicbut also tells a true detective story. Severalgroups, working independently, raced to isolatethe virus that caused this particularly virulentstrain of influenza and their stories are told here.The third book was released in 2004 by John M.Barry, titled The Great Influenza: The Epic Storyof the Deadliest Plague in History. It includesboth advances made in the past 30 years as wellas some new investigative work that convincedme that this virus originally presented itself inthe United States. It will quickly become thestandard bearer for this topic and a must read foranyone determined to learn from the mistakesmade in the past.

Actuarial InterestWhat makes the 1918 influenza pandemic (epi-demic is within a country or region, pandemic isworldwide) so interesting to actuaries is the ex-cess mortality curve. While the other epidemicsof the 20th century were lethal to the old andmost, but not all, caused extra mortality for thevery young, only in 1918 did those aged 20-40experience material excess mortality. When awhole life policy is several years old and the pol-

Max J. Rudolph, FSA, MAAA,

is vice president and actuary

at Mutual of Omaha Insurance

in Omaha, Neb. He can be

reached at max.rudolph@

mutualofomaha.com

In listening to Chairman Greenspan I found somany insights and applications of his com-ments to my work that I thought you all mightfeel the same. In trying to summarize what Itake from his comments, here’s the Top Ten:

1. Risk management is a conceptual framework.

2. Use your models as tools, as means to anend, not the end.

3. The simpler the better. 4. But the world’s complex so even the sim-

plest models might need to be complicated. 5. Robust is better than precise.

6. Rules help, but judgment matters too. 7. Know what you know and know what you

don’t know and never confuse one with theother.

8. Don’t confuse the future with the past.9. Avoid a big mistake from which you can’t

recover.10. Never stop learning.

Now, I’m not going to run out and campaign tobe the next Federal Reserve Chairman. But ifthat’s your dream, go for it. ✦

icyholder’s attained age is high, reserves havebuilt up and the net amount at risk (NAAR) ismuch smaller than the face amount of the policy(face amount = reserve + net amount at risk).With group life, sold as annual renewable term(ART) to all employees at many firms, the netamount at risk is a large percentage of the faceamount. And reinsurers generally provide ARTcoverage to direct writers. As a profession wespend a lot of time studying tail risk of equity-based and casualty products. However, thegreatest tail risk of all for insurers might be aproduct most life insurers and reinsurers havebeen happy selling for years as a core product,life insurance.

How it could happenSeveral months ago the Extreme ValuesSubgroup of the SOA’s Risk Management TaskForce had a contest to describe an extremeevent. In my entry I described a college all-starsoccer camp held in the Midwest over semesterbreak in late December. One of the campersisn’t feeling well, but this camp increases hischance to make the U.S. National Team and it isworth playing even while feeling a little slug-gish. The campers sleep in bunk beds set up in agym. They spend all their time together, makingmany new friends. After the camp everyone fliesback to his home to finish the rest of break, andthen returns to their respective universities. Noone realizes that all of them are now infectedwith influenza and contagious. Before anyone

knows it an influenza outbreak has occurred.How can society stop this scenario? It can’t, andthat is what is so scary about influenza.

Pandemic BasicsEvery year approximately 30,000 Americansdie from the flu. Most are very old or very young,with immune systems that are either not asstrong as they once were oras they will become. Manyget sick and spend a fewdays in bed to rest and re-cover. Few in the prime oflife die from it. The generalpattern is high morbidityand low mortality.Occasionally, perhaps fivetimes per century, the virusmutates into a form that iseither more contagious orexperiences higher mortali-ty. About once every centuryor two it takes a form thathistory remembers. This last happened in 1918,when 600,000 Americans and up to 100 millionpeople worldwide died, most over the course ofless than three months. While history is verylikely to repeat itself, and the odds seem to in-crease over time, the more we know about thelast pandemic the better we can deal with the



◗


next one. Unfortunately, when I speak withpracticing physicians they have only cursoryknowledge of the 1918 outbreak, often referredto as Spanish flu, and seem to assume thattoday’s available technology and drugs make itunlikely to return.

Influenza pandemics repeat periodically.Records are not always accurate, and other dis-eases are sometimes confused with influenza.Pandemics were recorded in 1510, 1580, 1688,1693 (all ages were impacted), up to six in the1700s and at least four in the 1800s. The 1889-91 pandemic was especially strong and the mostnotable in the past three centuries except for1918, which also returned in less virulent formin 1919-20.

Influenza can be very contagious. In the air afterbeing exhaled, it can infect someone else up to aday later. On a hard surface it can remain potentfor two days. Door knobs and remote controlsbecome inadvertent killers. The only effectivemeasures, even today, are isolation and quaran-tine. Doctors and nurses put their lives at riskwhen taking care of these patients and, to theircredit, very few shirked their duties in 1918.Keep in mind that many of the top medical pro-fessionals had been called to duty for the war ef-fort and there was a shortage of qualifiedpractitioners at the time. This fact wasn’t helpedwhen national nursing organizations refused tosupport curriculums to develop practical nurs-es. Only recently has the medical profession de-veloped knowledge about this particular virusand is starting to understand what made it solethal. In 1918 they knew little more than whenplague spread centuries earlier about how totreat it.

Are influenza pandemics like earthquake riskthat increases with time since the last one, or ispandemic probability completely random withthe likelihood just as high in the year followingthe most recent outbreak? Since antibodies cre-ated in one pandemic give some protectionagainst the next one, it seems more like earth-quake risk, building over time. As the last majorpandemic happened over 80 years ago, it is ap-

propriate to allocate resources to better under-stand influenza today.

WavesThe 1918 flu was active for several years, butmutated its form over time and created several“waves” that each traveled across most of theglobe. The first wave in spring 1918 was verycontagious, especially for young and healthyadults, but not very lethal, and was given littlenotice outside of the U.S. military camps whereit thrived. It required few contacts with others tospread and took only four months to cover theglobe. Africa, South America and Canada wereinitially spared. Those who were victims of thefirst wave showed resistance to later mutationsof the virus, although little preparation antici-pated future waves.

Author John M. Barry has traced the origins ofthe first wave to Haskell County, Kansas. Thissmall farming community in southwest Kansaswas missed in earlier research as a starting pointfor the pandemic because of an incorrectly pub-lished date of the local epidemic. Barry’s re-search of local newspapers allowed him to trackeach of the outbreaks from this local epidemic,advancing to what is now Ft. Riley in Kansas,then on to other military bases and beyond.

The second deadly wave occurred in the latesummer and fall of 1918. Since the “hosts” ei-ther died quickly or recovered with antibodiesto fight off future contacts with the disease, in agiven city the flu arrived, flourished and wasgone. In cities this took six to eight weeks, whilein military camps the overcrowding compressedit to three to four weeks. Within a region, thevirus appeared to grow weaker over time. If acommunity could delay its arrival or slow thecontact rate just a bit it had a material impact.As an example of the intertwined histories of thewar and pandemic, the Armistice Day celebra-tions recognizing the end of hostilities inNovember 1918 touched off a second wave ofthe flu in many cities.

U.S. troops in Europe had half the rate of deathfrom influenza as those in the United States. It islikely that they were exposed to the first waveand had developed some immunity. While thedeaths in the western world were staggering,those in other parts of the world that had not pre-viously been exposed to influenza were horrific.


Influenza Pandemics◗ continued from page 21

◗

Only recently has the

medical profession

developed knowledge

about this particular

virus and is starting to

understand what made

it so lethal.

“

”

It is estimated that 20 million deaths occurredin India alone.

The third wave started in December 1918 andhad mutated such that schoolchildren weremost impacted.

Life in 1918—Dominated by WorldWar IThe times surrounding 1918 were dominated byWorld War I, once thought to be the war to end allwars. Influenza likely impacted this concept,but I’ll get back to that later. After the UnitedStates entered the war and troops started to trav-el across the Atlantic Ocean to the front lines,the country quickly mobilized. This transitionwas helped by laws to improve morale. In effect,you could not criticize the federal governmentrun by Woodrow Wilson. Media fell in line, sup-porting “our boys” and the Liberty Loan bonddrives that pressured all citizens to buy govern-ment bonds to pay for the war effort (war fundingwas outside the regular budget at the time andrequired funding from other sources). Meetings,rallies and door-to-door solicitations raisedmoney for the war effort while unknowinglyspreading the virus. Any negative event wasdownplayed or ignored. Information regardingthe influenza epidemic was suppressed, caus-ing concerns to be downplayed and the media tolose the trust of their customers. The virusmoved very quickly. With an economy shifted towar production, there was limited lasting im-pact on the economy. Since the virus hit bothsides it does not appear to have impacted theoutcome of the war, although the fall elections inthe United States were very close and Congresswas divided evenly between parties.

The name “Spanish flu” seems to result fromthe lack of information reported by media incountries involved in the war. Spain, as a neu-tral party, had a relatively free press and re-ported the impact. Other countries reportedthe impact in Spain even as they ignored theirown outbreak.

American medicine had progressed rapidly inthe early years of the 20th century and, for thefirst time, had become as capable as the re-search facilities in Europe. In Europe, re-searchers were busy trying to combat gasattacks from the war and did not have resourcesavailable. As it turned out, the medical profes-

sion was helpless beyond common sense tech-niques of hydration and rest, which made publicpolicy even more important. The lack of honestinformation allowed fear to run rampant. Somefamily members refused toenter homes where simplecare would have saved the pa-tient. The army’s surgeon gen-eral, William Gorgas, warnedof a high pneumonia risk inthe military because of over-crowding, especially in thewinter when heat became aluxury. He suggested a 10- to14-day quarantine of troopsarriving in Europe, shorterthan Canada’s 28-day isolation period. He wasignored. Even when quarantines were set up bythe U.S. military, officers were often exempt.Unfortunately for the doughboys, the influenzavirus can’t read the bars on their shoulders. Anincomplete quarantine is little better than noquarantine at all.

As war production ramped up, workers de-scended on cities. Crowding became worse inthese cities, with shift workers often sharingrooms (and beds) with workers from other shifts.

Historical PerspectivePolitics played several roles in the 1918 flu pan-demic. No major player in world politics died.This is one reason why the pandemic escapesmany history books of the era and is likelycaused by the nature of the mortality curve.While normal influenza outbreaks cause excessmortality primarily at very young and very oldages, in 1918 the primary group afflicted wasaged 20-40. Few people have distinguishedthemselves enough by this young age to meritinclusion in a history book. We know, becausehe survived and accomplished much later inlife, that Franklin Roosevelt, as assistant secre-tary of the Navy, became very sick with flu dur-ing a trip to Europe. Had he died we would notknow this today. Who knows how many futureworld leaders were lost? Several members of theParis Peace Conference delegations died. Themost notable American death was WillardStraight. When he died he received threecolumns on the front page of the New York Times.Both President Woodrow Wilson and GeneralJohn Pershing became sick.


◗



◗

The other role of politics did little to minimizethe pandemic’s impact. President Wilson nevermentioned influenza publicly; not once. While600,000 Americans died at home from thevirus, he focused on the war effort and ignoredthe war against disease being fought by thecivilian population. It is interesting to note thathe tried to keep the United States out of the con-flict, but once America entered he seemed to become obsessed by it.

Woodrow Wilson arrived at the Paris PeaceConference driven by strong principles. He wasreportedly so frustrated with the negotiators,particularly those from France, that he wasready to leave the negotiating table and return toAmerica when he became ill. While there is stillsome debate on the matter (some think he had astroke), his symptoms were consistent with in-fluenza. After this time, future PresidentHerbert Hoover, also in Paris, thought his mindlost “resiliency.” Wilson suddenly had new andstrange ideas. For example, he thought therewere French spies in his residence. He sudden-ly abandoned principles he had previously in-sisted upon and yielded to France on everythingof significance. Four months later PresidentWilson suffered a major debilitating stroke.Many think that the seeds of World War II wereplanted in the treaty that ended World War I.Would history have turned out differently ifWilson had not encountered flu in early 1919?We will never know.

Other reasons for the lack of historical per-spective include the passage of time and thefact that flu is common, with a person rarelydying from it. I think a separate reason is bring-ing this devastating event back into the fore-front—research into the cause is starting to besuccessful. In the 1920s much research was at-tempted, and side benefits included penicillinand genetics, but the solution never came. Thescientists were not able to identify what madeso many so sick in 1918.

Although sanitation had improved enough sothat large cities like London could support theirown population without immigration from thecountryside for the first time, epidemics were

not rare in the early 20th century. As recently as1911 a measles epidemic had broken out in theU.S. Army that killed 5 percent of those whocaught it. Various forms of plague had also bro-ken out since 1900 around the globe. The entryof the United States into war drove leading sci-entists of the day to predict that travel andcrowded conditions would lead to some form ofepidemic. Their goal was for this war to be thefirst where more died from battles than fromdisease by lowering the number of disease-re-lated deaths. Little did they know that war con-ditions, especially troop ships, wouldaccelerate the pandemic.

Funeral homes and cemeteries were quicklyoverrun, even as life insurance agents wereplacing ads encouraging business (apparentlyimproperly aligned incentives are not a newphenomenon). Fear and sickness caused manyof the people who normally performed theseburials to stop or run away. Often there was noone left to pay the funeral bills, and wood forcoffins became scarce. Women of “society”often took over when the government proved in-effective. Volunteers too quickly abandoned thetask. Families had to dig their own graves andpolice often had to collect the bodies using wag-ons reminiscent of plague.

In some families everyone was wiped out. Largenumbers of orphans and single-parent familieswere created in late 1918. Although the impacton servicemen is well documented, youngwomen were also heavily impacted. Pregnantwomen proved especially vulnerable. Somethought it might be the end of the world basedon biblical references in Revelations to adeathly pale horse.

ReactionsSince medical researchers at the time had notidentified the virus and how tiny it was, manycities passed laws requiring that masks be wornwhen out in public. They did not help. As long asthe masks were replaced frequently there wasno harm done, but the masks themselves startedto breed germs when moist for long periods. Notsurprisingly, the supply of masks ran low, creat-ing shortages and encouraging people to wearthem for longer periods. Another rumor had itthat chewing tobacco decreased your odds ofcontracting influenza.


While originally thought

to be purely a human

disease, influenza

actually passes back

and forth between

humans and several

species of animals.

“

”


In late September 1918, the surgeon generalprovided advice to avoid influenza. It includedgeneral common sense suggestions like avoid-ing needless crowding and washing your hands,but also to chew your food and avoid tight fit-ting clothes.

Victor Vaughan, head of the Army’s Division ofCommunicable Diseases, said “If the epidem-ic continues its mathematical rate of accelera-tion, civilization could easily disappear…fromthe face of the earth within a matter of a fewmore weeks.”

The draft was canceled in the United States—not to reduce the impact of influenza, but be-cause influenza had so impacted the militarycamps that no training for deployment couldoccur. It was the right thing to do, but for thewrong reason. Many lives were likely savednonetheless.

Children are often a good barometer of the na-tion’s pulse, taking a serious subject and puttingit in their terms. A jump rope verse heard acrossthe country reflects this.

I had a little birdAnd its name was Enza.

I opened the windowAnd in-flew-Enza.

The VirusPigs, ducks and humansWhile originally thought to be purely a humandisease, influenza actually passes back andforth between humans and several species of an-imals. Each one reacts differently to a specificstrain of the virus. Since farmers are in regularcontact with pigs and birds (primarily ducksand chickens), they are one of the conduits be-tween human and animal influenza. Another ismarkets where live animals are sold. It isthought that if an animal (or human?) catchesboth human and animal flu at the same time thatit mutates into something that can be morelethal to one or the other species. The 1918swine flu was new to the species, has been pres-ent ever since and was initially deadly to pigs. Itwas recently shown to have originated as anavian, or bird, flu, adding to the confusion. Thevirus strain itself is mild to pigs but, as hap-pened in 1918, in combination with B. influen-zae as a secondary invader it becomes highly

lethal. Antibodies from humans have been usedto protect pigs from the virus. An old wives talethat just might be true is that an outbreak of fluamong horses often is a leading indicator ofhuman flu epidemic.

Several recent influenza out-breaks have originated inChina. Live animal marketshave been involved in manyof these epidemics, but farm-ers also live in a high-risk en-vironment. For example,farmers in China use ducks togrow rice. When the fields areflooded, the ducks eat insectsand weeds while leaving therice alone. The ducks are re-moved while the rice grows,returning to clean the fields after the harvest.Having served their purpose, the domesticatedducks are then slaughtered for dinner. Thesefarmers also raise pigs. This combination ofswine, birds and humans has historically be-lieved to have been the source of the worst in-fluenza outbreaks. The viruses are able tocombine between species and sometimes createa lethal cocktail.

Medical BasicsWhile this paper is not being prepared for amedical journal, some basic information willprovide background.

Influenza consists of eight genes made of ri-bonucleic acid (RNA). Other viruses in thisfamily include HIV and coronaviruses (SARSand many colds). Viruses have no means toreplicate themselves. They use their genes tosubvert and direct host cells. Influenza moves inand out of these cells using sharp proteinshards, hiding from the immune system while inthe cell. The virus requires a specific enzymefound in humans only in the lungs. Birds are thenatural and original homes for influenza, exist-ing in their gastrointestinal tract. It is oftenfound in their droppings. While a human cancatch avian flu from birds directly when ex-posed to massive amounts of bird droppings, itcan’t spread to other humans because the bind-ing process between virus and host cell differsbetween species. The virus must first adapt tohumans, using an intermediary animal like

◗



pigs. When deoxyribonucleic acid (DNA)copies itself, it works very hard to do so correct-ly. There are redundancies to make sure thecopy is correct. When RNA copies itself, veryfew copies are exact because no redundanciesare built in. Because the influenza virus is a seg-mented genome, the genes can reassort them-selves, allowing the virus to move betweenspecies. It is much easier for RNA to mutate andfind a form that helps it to survive than it is forDNA to do so. This “mutant swarm” processmakes it inevitable that new forms of the in-fluenza virus will emerge. It is very hard to cre-ate a vaccine, trying to hit a moving target as thevirus mutates. In addition, influenza mutatesvery quickly to become drug resistant, oftenwithin days. While not meaning to detract fromthe seriousness of AIDS, the cumulative totaldeaths from HIV is lower than those recorded bythe 1918 influenza virus. Since research is con-tinuing on both and they are similar, perhapsjoint efforts will help control both viruses.

The virus contains two proteins: hemagglutinin(H) that helps flu to enter cells and neu-raminidase (N) that helps flu escape from thecell. As these proteins mutate, they are identi-fied by shape (each shape is identified by num-ber). H1N1 identifies the 1918 Spanish flu,along with epidemics that occurred in 1976,1977, 1986 and 1988. Other epidemics includ-ed 1956 (Asian flu—H2N2), 1968 and 1993(Hong Kong flu—H3N2), 1995 (H7N7), 1997(Chicken flu—H5N1) and 1999 (H9N2). Theinfluenza virus is unique in that, as a new virustakes hold, it seems to drive old forms of thevirus to extinction. There are 15 known shapesof hemagglutinin and nine of neuraminidase.

The 1997 virus generated another concern.Influenza vaccine is grown in eggs, and this par-ticular one was so strong that it killed the eggs. Ittook a year to overcome the problem. In a pan-demic you don’t have a year.

The similarities to the 1918 virus created theswine flu scare in 1976, where the United Statesencouraged immunizations for all.Unfortunately, the vaccine proved to have somenegative side effects. Luckily the influenza out-break did not occur. This highlights a very diffi-

cult issue for public policy makers. They use theprior year’s flu, especially those mutations thatappear in the spring, to make the next year’s vac-cine. If they successfully reduce a pandemic’simpact they are open to criticism for crying wolf.Resources and time are both limited, so theyhave to focus their efforts. Stockpiling vaccinedoesn’t work because the virus mutates quickly,making such efforts fruitless.

Moisture appears to play a part in the influenzastory, as it appears to die quickly in high hu-midity. One story I heard referred to the 1988Kansas basketball team, led by DannyManning and coached by Larry Brown.Apparently flu was going around campus andCoach Brown had his players sit in a humid en-vironment every day to breathe in the moistureand kill any flu virus. While you still have toplay the game, that team won the NCAA tour-nament as a number six seed.

SymptomsThere are stories of people that appeared tofaint, collapse and die during the pandemic.Soon-to-be patients would leave for work, seem-ingly healthy, and die the same day. In one to two hours a person could become prostrate withfever up to 105 degrees, general weakness andsevere headaches. Their eyes would burn andtheir ears would ache, with blood coming fromears and eye sockets. So much pressure wouldbuild up that blood would sometimes spurt sev-eral feet from a patient’s nose. Severe aches intheir muscles, joints, backs and heads madethem feel as if they had been beaten all over by aclub. Kidneys, liver, adrenal glands and testesall were impacted. Nurses reported that airpockets beneath the skin caused patients tocrackle like a bowl of Rice Krispies when theywere rolled over.

When autopsies were completed, the heart hadoften been impacted. There was often an in-flamed pericardium, the sac around the heart,and the heart muscle itself was flabby.

Doctors will tell you that there is no such thingas intestinal flu, but many of the cases in 1918packaged together influenza and dysentery.This is likely a case of the influenza weakeningthe immune system in a way that encouraged op-portunistic secondary diseases to take hold.


Doctors will tell you

that there is no such

thing as intestinal flu,

but many of the cases

in 1918 packaged

together influenza

and dysentery. This is

likely a case of the

influenza weakening

the immune system in

a way that encouraged

opportunistic secondary

diseases to take hold.

“

”

◗

Acute Respiratory Distress Syndrome (ARDS)was not identified as a specific sickness untilafter 1918, although the doctors of the timewould be very familiar with it because of itscommon occurrence among influenza deaths. Itis caused by an extreme stress on the lungs. Thiscan be caused by near drowning, smoke inhala-tion, inhaling toxic fumes (or poison gas) or in-fluenzal viral pneumonia. The immune systemburns the insides of the lungs, followed by sec-ondary invaders that attack the weakened de-fense systems. Someone diagnosed with ARDStoday has a 40-60 percent chance of survival.Without intensive care units, this rate dwindlesto near zero. There were no ICUs in 1918. Thearmy found what is now called ARDS in half theautopsy cases it reviewed from 1918. As thevirus mutated it became less prevalent and wasrarely found in the comparable 1919 autopsies.

As a comparison, bacterial pneumonia follow-ing influenza has 7 percent mortality rate.Antibiotics are used to treat these infections.Over 35 percent of pneumococcal infections areresistant to the antibiotic of choice. For staphy-lococcus aureus bacterium, which is very resist-ant to antibiotics, the death rate is over 40percent.

Lasting EffectsThe 1918 influenza strain appears to have leftresidual effects on many who recovered, leavingthem tired and susceptible to heart and neuro-logical problems. Many who had flu died earlyin 1919 of secondary causes and were notrecorded as flu deaths.

The influenza virus is hosted in the humanbody by the lungs. Yet the virus also appears tohave impacted the neurological system, caus-ing paralysis and mental illness. Hysteria,melancholia, delirium and insanity with suici-dal intent were not uncommon. Some of thiswas temporary. A study that followed influenzapatients that suffered from schizophreniashowed that most recovered completely withinfive years.

There appears to be a link to the brain as well,with those infected having a higher rate ofParkinson’s disease a decade later.

Worst Case—DeathMany of the worst cases resulted in what was es-sentially drowning. As the lungs filled with red-

dish fluid, less oxygen was transported to the ex-tremities. Nurses could tell which patients werenot going to make it by looking at their feet. Ifthey were black it was as if a death sentence hadbeen given. Faces turned a dark brownish pur-ple and patients coughed up blood while gasp-ing for breath.

ResearchOver the years much research time and effortwas devoted to Pfeiffer’s bacillus, a bacteriumoften present with influenza as a secondary in-vader. In some locations it was found in nearlyall of the cases and was thought to be the causeof influenza. This is partly because of the statureof the scientist who discovered the correlation,Richard Pfeiffer of Germany. Many researcherswere so sure of themselves that, if no influenzabacilli were found, then the disease presentcould not be identified as influenza. But this wasa bacterium, and the true cause of all influenzais known today to be a virus.

While little success was achieved in discover-ing the source of the virus, lots of good researchwas performed and some important discoverieswere made. Penicillin was found to kill bacteria,and Oswald Avery made the key observationthat DNA carried genetic information when hedetermined how a weak pneumococcus could bemade strong by adding a capsule around itsstructure.

Several individuals worked for many years toisolate samples of the virus. Since the Civil War,standard autopsy procedurefor military doctors has beento take thumbnail samples oftissue, soak it in formalde-hyde and store it in paraffinwax. The Armed ForcesInstitute of Pathology thenstores these samples. Severalof these samples survivedfrom 1918, and Dr. JeffreyTaubenberger painstakinglydissected them until he hadsequenced the genome. Dr. Taubenberger con-tinues to be a principal researcher of this topic.Dr. Johan Hultin, while studying at theUniversity of Iowa, also became involved in theissue. The Swedish-born scientist made trips toAlaska to find bodies buried in the permafrost

◗



◗


where the virus might have been frozen.Eventually the science caught up and he wasable to bring back small samples to study.Several other researchers have been involved aswell, but these two have been more focused onsaving lives than personal glory.

The mystery that still stymies scientists today isthe 1918 strain’s lethality. What made so manypeople die from this influenza and not from an-other one? Once this is identified, vaccines andother methods can be used to lessen its impact.Since flu shots create antibodies in a person,does getting an annual flu shot help to protectyou from the next pandemic? It is currentlythought that a mild flu earlier than the pandem-ic of 1889 provided those older than 40 withsome immunity to the 1918 virus.

Similar Topics TodaySARSSevere Acute Respiratory Syndrome (SARS)first appeared in late 2002 in China. It was notreported to the World Health Organization(WHO) for several months, which allowed it toget a foothold and spread through international

travel. While it is more lethal thaninfluenza, it is less dangerous. Itrequires close contact to spread. Italso lives primarily in the upperrespiratory tract, so is spread viacoughs and sneezes up to a weekafter symptoms develop. Thismakes it much easier to isolateand quarantine. Nevertheless, theease of its spread to Toronto showshow important it is to react quick-ly to any outbreak.

BioterrorismIt is difficult to read about influenza in today’senvironment and not think about the possibilityof terrorists discovering the secrets of influenzabefore “the good guys” do. There are many his-torical cases of bioterrorism, mostly involvinghurling infected animals over the wall of a town.In recent times the Japanese spread bubonicplague in China during World War II, in 1984 anOregon cult infected salad bars in a restaurant

with salmonella, in 1995 a cult group releasedthe nerve gas sarin on commuter trains in Tokyo,and in 2001 anthrax was spread through themail in the United States. Smallpox was usedagainst Native Americans, both accidentallyand on purpose (millions of Native Americanslived along the Mississippi River until a mes-senger carrying news of the Spanish arrival alsocarried smallpox). The risks today are anthrax,plague, smallpox and botulinum toxin. All canbe countered with either antibiotics or antitox-in. Poison gas has been used more frequently,notably in World War I and more recently inIraq.

Actuarial InvolvementThe NumbersThe resulting mortality curve for 1918 lookedlike a W, with some additional deaths at earlyages and little excess mortality at older ages.The unusual characteristic of this pandemicwas the high number of deaths among those inthe prime of life. Other flu outbreaks have hadfewer than 10 percent of the deaths at ages 16-40. In 1918 over half the deaths occurred in thisage range, with ages 21-30 the worst hit.

Specific industries that featured overcrowdingand moist conditions were particularly suscep-tible. In 1918 Met Life reported that 6.21 per-cent of all insured coal miners and 3.26 percentof all insured industrial workers died. Overallestimates of worldwide deaths have ranged from20-100 million, with the higher numbers nowconsidered more credible. The high end of therange translates into 5 percent of the world’spopulation dead from influenza within aboutthree months. Based on population increases,today as many as 350 million would die.Population morbidity was less than 30 percent,which is toward the high end but still within anormal range for influenza. Slightly higher per-centages were found in the military, likelycaused by the overcrowded conditions. In thearmy, one in 67 died over a 10-week periodstarting in mid-September 1918.

In the United States, the consensus is that 25percent of the population became infected withinfluenza and, of these, 2.5 percent died. Thisresulted in .06 percent of the population dyingfrom the virus (.25 * .025 = .00625). This left600,000 Americans dead and reduced the ex-


pected lifetime (taking the probability of surviv-ing 1918 for each attained age and multiplyingthe results) from 51 to 39. More Americans diedfrom the 1918 influenza pandemic than from all20th century combat deaths. Up to 10 percent ofthe world’s young adults died. One can only imag-ine the impact these lost lives could have had, forbetter or worse, during their lifetimes. CouldWorld War II been prevented, or did a brilliantGerman soldier die that would have changed theoutcome of the war? We will never know.

Looking at U.S. population trends shortly before1918, the average annual increase was about1.4 million (average of 1915-1917). In 1918 thepopulation decreased by about 60,000. Onecould conclude that excess mortality from allcauses was 1.5 million (an alternative could befewer children being born—this was not thecase in 1918 but did reduce in 1919). Wardeaths are listed as over 50,000 in battle andover 60,000 other deaths, which would includethose caused by influenza. It appears that astrong argument can be made that the 600,000deaths from flu estimate in the United States islow, especially since the number also includesdeaths from 1919 and 1920 when a less virulentform of influenza created later waves.

Remote AreasAreas not readily accessible and without do-mesticated farm animals were especially vul-nerable since many had not previously beenexposed to milder forms of influenza. The visit-ing postman wiped out some Alaskan villages,delivering flu along with the mail. In Tahiti over10 percent of the population died within amonth of a ship’s docking with flu aboard.Samoa provided an interesting case study. InWestern Samoa, 20 percent died after a boatdocked with flu. In American Samoa, the gover-nor quarantined boats with the flu and there waslimited impact. The locals still sing a tributesong to the governor for his strong and promptaction. Native Americans were also suscepti-ble, with 2 percent mortality. In Russia and Iranit is estimated that 7 percent died. In India itwas 5 percent. Funerals in India use cremationand return ashes of the deceased to the rivers.When they ran out of firewood, corpses soonclogged the rivers.

ModelsThis paper is based on a presentation given inApril 2004 at the Enterprise Risk ManagementSymposium. Dave Ingram also presented at asession titled “Pandemic History and FinancialImplications: Focus on Flu Epidemics.” Dave’sfocus was to present a simple teaching modelthat he developed while working on SARS. Sucha model can help someone cut through the hys-teria and understand where an outbreak may beheading. Parameterizing and modeling anemerging epidemic allows the modeler to pro-vide a leading indicator of the epidemic’sstrength while it is still early enough to putmeasures in place that reduce the impact.

Dave trued up his model to the 1918 U.S. pan-demic using the following parameters: popula-tion 100 million, contact rate 12 per day,transmission efficiency (probability of infectionper contact) 2.5 percent, recovery rate 25 per-cent per day and mortality .64 percent per day.This is consistent with the 25 percent morbidityand 2.5 percent mortality data described earlierthat resulted in a reported 600,000 U.S. deaths.With a population at year-end 2002 of 289 mil-lion, proportionate results would be 1.8 milliondeaths today in the United States. It is easy tothink of adjustments to the contact rate based ontoday’s global industrialized society that wouldquickly increase these results even if mortalityresulting from secondary infections were re-duced. Adding inoculations to the model and re-ducing the contact rate even slightly has a majorimpact on the results. Dave’s model can also beadjusted to include an initial number of infectedpersons in case the epidemic has already started.

The Centers for Disease Control and Prevention(CDCP) has built a model to project a pandemicin the United States. It uses an average of the1918, 1957 and 1968 outbreaks. While thesenumbers are large, if they were to use only the1918 statistics the deaths would be four times asgreat as those modeled and the shape of the ex-cess mortality curve will not match that of 1918.Users of this model should be careful to under-stand its limitations.

◗



The Centers for Disease

Control and Prevention

(CDCP) has built a

model to project a

pandemic in the United

States. It uses an

average of the 1918,

1957 and 1968

outbreaks.

“

”

◗


InsurersActuaries had their hands full in 1918. Forthose policies where the attained age was olderand the policy had been in force for severalyears, reserves had built up and the net amountat risk was lower. For younger policyholders,where little reserve had been set up, the hit washarder. About three quarters of the companieseither omitted or reduced their dividend.

As a result of the 1918 pandemic, theCommittee on Statistical Study of theInfluenza Epidemic created an opportunity toshow what statistics and their methods coulddo for preventive medicine. Insurers alsobegan funding research on the flu. Insuranceapplications picked up, not surprisingly, asthere were many examples where life insur-ance would have helped families cope withtheir losses. Underwriters were aware of theheart based after effects of this flu and includ-ed this in their analysis.

During a June 1919 American Institute ofActuaries conference, the dynamics of the timewere discussed. In addition to the mortality in-creased by the influenza pandemic, the war alsotook its toll. There were other repercussionsfrom the war effort. Some added to mortality,some encouraged a healthy lifestyle. They in-cluded dietary improvements in the generalpopulation as they planted “war gardens”, aswitch from wool to cotton based clothing (high-er mortality assumed from the change, not be-cause one was better than the other), landlordscutback of coal during the pandemic to “sup-port” the war effort, lengthened work hours andhigher accidents caused by skilled workers en-tering the armed forces.

Before the end of 1918, President of theActuarial Society of America Henry Moir esti-mated that the pandemic had cost 400,000American lives having an average age of 33. Theaverage age at death prior to 1918 had been 55-60. He estimated that 25 years were lost per vic-tim and a total of 10 million person years werelost to mankind.

Fast-forward to today, where many individualpolicies and most group life insurance are de-signed using annual renewable term features.Many reserves are 1⁄2 cx for these policies, astatutory accounting convention that treats eachpolicy year as an independent event. In addi-tion, reinsurance is often designed using ART.A company might determine that it can survive5 percent excess mortality due to offsetting in-benefit annuity mortality, but what if (like in1918 when older ages did not experience highermortality) the offset did not appear? And whatabout the counterparty risk of the reinsurer?What if the reinsurer pays 50 cents of every $1 itowes the direct writer? Even these reduced pay-ments could be delayed, especially if a bank-ruptcy results for the reinsurer. My opinion isthat the life insurance industry will survive thenext pandemic only if there is wide spread secu-ritization of term policies. At least life reinsurersand companies with large ART concentrationsshould strongly consider this option.

SummaryWhy the next pandemic may be mildThere are several reasons why the next influen-za pandemic may not be as serious as occurredin 1918.

• Today’s health teams have developedIntensive Care Units and other means of car-ing for patients that are much advanced from1918. Many deaths at that time were fromsecondary invaders that can be better treatedtoday. An example is bacterial pneumonia,which can be treated with antibiotics.

• There was a World War going on at the time,with nearly all regions of the world impacted.This made countries unwilling to imposetravel quarantines.

• Germ theory has advanced to where the nextvirus can more easily be analyzed.

• Antiviral drugs are now available.• International cooperation has improved dra-

matically, partly because of the 1918 pan-demic. Countries, cities and towns creatednew departments of health, including theU.S. National Institute of Health. The WorldHealth Organization (WHO) coordinatessurveillance of mutations to adjust eachyear’s vaccine.

• Many emergency hospitals created in 1918were converted into permanent ones.


As a result of the 1918

pandemic, the Commit-

tee on Statistical Study

of the Influenza

Epidemic created an

opportunity to show

what statistics and their

methods could do for

preventive medicine.

Insurers also began

funding research on

the flu.

“

”

Why the next pandemic may beworseThere are also several reasons why the next in-fluenza pandemic may be more serious than oc-curred in 1918.

• Travel—it is the primary concern. It is soeasy to travel internationally that any conta-gious virus would appear to hit everywhere atonce. A quarantine of one region would likelycome too late, as it did with SARS. It tookonly two weeks for the influenza virus to trav-el from Boston to Seattle in 1918. This viruswould have covered 90 percent of the earth inthat time period today.

• Time—it has been a long time since the lastmajor influenza pandemic and few are stillalive with antibodies in their system.

• Medical advances have not yet identified thevirus completely and don’t know why it wasso lethal, let alone how to counteract it.Within the next few years, additional re-search may turn this into a reason the nextpandemic will be mild. Offsetting this is theability of the virus to quickly mutate to a drug-resistant strain.

• ICUs will quickly be overrun with patients.Many hospitals have downsized as they pri-vatized and are not capable of supporting theneeds of 30 percent of the population simul-taneously. There will be a shortage of bedswhen they are needed most.

• Many bacteria are building resistance to cur-rently available antibiotics, making second-ary infections more dangerous.

• Regional conflicts throughout the world areconstantly flaring up. Some could expand.

• Influenza is showing signs of building resist-ance to antiviral drugs.

• More countries do not participate with theWHO than do, so coverage of future influen-za mutations is not complete.

• Politics and the courts—try to imagine tellingeveryone that they can’t leave a state. Thecourts would be flooded with cases to allowfreedom of travel and might have to allow it iflaws are not enacted in advance to allow andenforce a quarantine. The quick challenges tothe Patriot Act provisions are good exampleswhere the proper balance is a debatable ques-tion among reasonable people.

• More people live in urban areas. This meansthere are more contacts per person per day.On the flip side, there is also greater proba-bility of previous exposure to a similar virusthat provides partial pro-tection.

• There will be a greater im-pact on the economy, asrelatively more peopleless than 40 have key posi-tions in the information-based global economy.The financial markets arealso likely to react quicklyto an outbreak.

• A population exists todaywith compromised im-mune systems. Those withHIV, being treated with ra-diation/chemotherapy and transplant recip-ients will all be susceptible to influenza andits secondary invaders.

In May 1919, the Actuarial Society of Americapublished, in Transactions (Vol. XX, Part 1 No.61), a paper by James D. Craig and Louis I.Dublin titled, “The Influenza Epidemic of1918. “Their lead paragraph included a state-ment that could be made today, and I will in-clude it here. “With the recent development ofhygiene and sanitation as marked as it is, theworld felt safe against the possibilities of anynew conflagration from influenza. But the expe-rience of the last two years has demonstratedthat we are not so far advanced in our knowledgeof this disease, of its cause and of the methods ofits control as we thought we were. Epidemicsmay still occur with sufficient virulence to testthe resources and stability not only of life insur-ance companies but also of civilization itself.”Will history repeat itself?

Overall, if the scientists can develop counter-measures that can be quickly implemented thiswill be a great weapon in the battle against an in-fluenza pandemic. If not, public policy willdrive the results. Closing down public meetingplaces like schools and shopping malls, limitingcontact with others and temporary telecommut-ing will all help.

◗



◗


What can we do to protect our families?This is a tough question, and one that I havethought a lot about over the past year as I re-searched this topic. Until Dr. Taubenberger andhis extended team solve the riddle of why thisparticular strain of influenza was so virulent anddevelop the tools to create countermeasuresquickly, there is little beyond common sense.Getting a flu shot every year will help, even ifjust to ease identification of similar viruses likeSARS. I have to wonder if flu shots might haveantibodies that will also help protect from fu-ture, similar, strains. Perhaps something thatseems odd today, like getting a transfusion fromsomeone who lived through earlier mutations ofinfluenza, will be found to add the proper anti-bodies. The most difficult challenge will be po-litical. It is difficult to determine as an outbreakis occurring how bad it will be. Will our leadershave the nerve to quarantine large segments ofthe population in a democracy? And even if theyhave the nerve, will the courts allow it to happen(Tom Clancy’s fictional work Executive Ordersaddresses the quarantine issue when the Ebolavirus is released.) As the models show, reducingthe contact rate even slightly has a major impacton how bad a pandemic will become.

Risk to the Life Insurance IndustryAccording to the 2003 ACLI Life Insurers FactBook, which cites data from 2002, the total netamount at risk in the life insurance industry is$15.5 trillion and statutory surplus is $198 bil-lion. Surplus as a ratio of NAAR is 1.28 per-cent. Excess mortality, beyond conservativestatutory based assumptions of 1.28 percentwould bankrupt the industry if it were just onecompany. This, of course, does not include lifeinsurance benefits designed into a productlike a variable annuity. The 1918 pandemic re-portedly killed .6 percent of the population andhad limited excess mortality at high ages.There appears to have been underreporting ofthis cause of death in 1918 even in the UnitedStates, let alone elsewhere in the world. Evenso, the estimate is 5 percent dead worldwide,much higher than in the United States. There

are a number of reasons that likely account forthis anomaly. Sanitation matters greatly whensecondary diseases attack. So does poverty,since it leads to overcrowding.

Each year companies report results such asasset adequacy tests (U.S. regulators) and dy-namic capital adequacy tests (OSFI in Canada).These require deterministic scenarios, andmost companies include at least one based onhigher mortality for life business. Companiesshould consider the risk to their block if the1918 influenza pandemic recurs. Even withU.S. data, which is lower than most of the world,the risk to insurers writing mainly group life orother ART-based products is high. Insurersshould also stress their counterparty risk. Whatif the reinsurer pays less than 100 percentand/or the payments are delayed? Can the in-surer survive?

This naturally occurring catastrophic riskseems logical to be considered for a governmentbackstop, much as the man-made terrorismpayback coverage. Without it, the industry willnot survive the next influenza pandemic in itspresent form. Another possible solution wouldbe to tap into the financial markets.Securitization of mortality risk into a financialinstrument, similar to catastrophe bonds, wouldspread the risk across a diverse group of in-vestors that could absorb a large loss occasion-ally, if paid for taking the risk in the other years.

History often repeats itself, and influenza pan-demics are no exception. The key to dealingwith them is for politicians to treat them serious-ly and not just hope they will go away. Any re-ductions in contact rate can have major benefitsfor a region. It likely will take at least one moremajor outbreak for scientists to get their handsaround this disease, but in the future lies hope oftreatment and vaccine. In the meantime actuar-ies should consider various scenarios and deter-mine the expected impact of each. Only throughanalysis can you determine the interactions be-tween product lines to various levels of mortali-ty and shapes of the mortality curve. It’s time todo that analysis. ✦


History often repeats

itself, and influenza

pandemics are no

exception. The key to

dealing with them is for

politicians to treat them

seriously and not just

hope they will go away.

“

”

The risk measure conditional tail expec-tation (CTE) has been getting more andmore attention for measuring risk in any

situation with non-normal distribution of losses.Canadian and U.S. insurance regulators haveadopted CTE as a standard for regulatory capi-tal measurement. Academics have lauded CTEas a “coherent” statistic. Those outside the in-surance industry call it “Tail VaR” or “expectedtail loss” (ETL). Actuaries, who have alwaysbeen suspicious or even hostile to the usage ofvalue at risk (VaR) as a risk measurement stan-dard, have readily embraced CTE.

This article presents the observations of somerandom investigation into the nature of CTE.There are still many questions that need to beasked and answered before we can say that weare completely familiar with the nature andcharacteristics of CTE. There are six questionsthat will be discussed:• How does CTE compare to VaR?• How sensitive is CTE to the number of

scenarios used?• Is CTE consistent with existing RBC factors?• What is the impact of time period used for

calculating CTE?• How does CTE measure the difference

between regime switching and single regimelognormal scenarios?

These observations are all tied to examples.Since insurance products are very complex, it isdifficult to see whether the results of any one testwould apply to any other situation. Therefore,the examples used here will all be based on sim-ple random values, rather than to calculationsbased on the random values. The random valuesare developed as stock returns that would beused as inputs to calculations about a stock-linked product. These discussions will not re-solve any issues about CTE calculations for in-surance products, but through some sort ofrough transitive property, can form a part of astarting point for investigations into the ulti-mate nature of this new tool.

1. CTE vs. VaRWithout getting into the mathematics of coher-ence, a simple picture can show the appeal ofCTE as opposed to VaR. CTE is defined as theprobability weighted loss above a certain prob-ability level, while VaR is the loss at a certainprobability level. One definition of CTE is that

it is the average of all VaR values for probabili-ties above a specified level. Generally VaR isused by banks and most often it is used to meas-ure risk over a very short (less than one month)time frame. CTE is becoming the standard forinsurance company risk measurement, espe-cially regarding risk capital and is most oftenused to measure risk over multi-year timeframes that are needed to view insurance risks.The following two charts show the distributionof values for stochastic tests of two products.Product A has a normal distribution of gainsand losses, while the distribution of gains andlosses for Product B are not as predictable.

For Product A (Chart 1) and any other productwith a normal distribution of gains and lossesthere is a regular predictable relationship be-tween CTE and VaR. Using one measure or theother does not necessarily add any information.

◗


Getting to Know CTEBy David Ingram


Chart 1—Distribution of Gains and Losses

Chairperson David Ingram,

FSA, MAAA, is a consulting

actuary at Milliman USA

in New York. He can be

reached at david.ingram@

milliman.com.

◗


For Product B (Chart 2), VaR and CTE cansometimes show drastically different stories.The VaR for Product B on the graph above is anear zero value. There is a small loss comparedto the expected gain, but not much. The tail ofthe profit distribution starts to drop-off some-where above the 95th percentile. A CTE calcu-lation includes the impact of that drop-off,wherever it starts. A VaR measure will only re-flect the drop off if it is occurring at the percent-age chosen for the VaR measure.

Above, we noted that there is a predictable re-lationship between VaR and CTE for a normal-ly distributed gain/loss distribution. In fact,Chart 3 shows the values of 95VaR and CTE90as a percentage of mean for various normal dis-

tributions are both linear relationships, with the95VaR equal to the mean less 164.56 percent ofthe standard deviation and the CTE90 equal tothe mean less 176.3 percent of the standard de-viation. Looking at the graph in Chart 3, firstnotice that VaR and CTE are both positive whenthe standard deviation is less than about 65percent of the mean. The modified CTE usedfor U.S. NAIC risk-based capital would usezero for any positive values and would thereforeproduce small negative values a little further tothe left on the chart.

The conclusion here is that VaR and CTE arenot necessarily significantly different if gainsand losses are normally distributed. However,if gains and losses do not conform to a formula-ic distribution, the difference is unpredictable.In those situations, CTE can provide signifi-cantly different information from VaR.

2. Tails and Number of ScenariosAt this point, we will switch from looking at ageneral loss distribution to a specific realmodel situation. The latest CTE discussionsare revolving around variable products andguarantees in those products. As mentionedearlier, products and modeling results varywidely. However, significant insight into theworkings of CTE can be observed from lookingat the CTE of a simple portfolio of commonstocks. The portfolio returns can be generatedusing several different scenario generationmethods and the impact on CTE displayed.

The AAA committee report on C3 Phase II RBCsuggests that a regime switching lognormal(RS2LN) model adequately captures the de-sired characteristics for the proposed CTE cal-culations. We are going to start with a log-normally distributed set of returns for a stockportfolio and eventually compare that to aRS2LN set of scenarios. For both sets, we willcalibrate the models so that the mean is 11.9percent and the standard deviation is 17.2percent overall.

Getting to Know CTE◗ continued from page 33

Chart 2—Embedded Value

Chart 3—Normal Distribution

For the lognormal model, by stating those val-ues, the parameterization is complete. Thereare only two parameters, mean and standarddeviation. The ratio of standard deviation tomean is 1.45 percent and we would expectCTE90 for normally distributed losses to comein at 155 percent of the mean or an absolutevalue of 18.45 percent based on calculationsbehind the graph above. In fact, for our set of1,000 scenarios, we get a CTE90 of 18.35 per-cent, which is 154 percent of the mean return.

For 20 sets of 1,000 lognormal scenarios, theCTE90 averaged 18.40 percent. In addition,we found that the CTE90 has range of 18.33percent to 18.50 percent and that there is a 96percent correlation between the CTE90 and theworst value of the 1,000 scenarios. That is asmall cause for concern, since the worst value isprobably the least dependable statistic of theentire set (Chart 4).

The ratio of the largest and smallest CTE valuesis less than 101 percent. That would suggestthat the CTE90 calculation for 1,000 scenariosmay be sufficiently reliable.

To look for a smaller number of scenarios thatmay be reliable, we took the same set of ran-dom values and divided it into 40 sets of 500scenarios. The CTE90 results get much morediverse: (Chart 5)

The range of highest to lowest CTE value is 126percent for these 40 sets of random returns. Itwould seem that the empirical evidence sup-ports the idea that 1000 scenarios may be ade-quate, while 500 scenarios is not. The findingthat there was high correlation between worstscenario and CTE90 value suggests that a userof CTE should examine the worst scenario tomake sure that it is not causing an exaggerationof CTE value due to an extreme outlier in an in-sufficiently large scenario set.

3. Comparison to C1(CS)The RBC requirements for life insurance compa-nies already contain a component for commonstocks. That amount is 22.5 percent for a fully di-versified portfolio that qualifies for the lowestlevel of RBC based on the beta of the portfolio. Ifwe calculate from the 20 sets of 1,000 random sce-

narios, the present value of the CTE90 result onyear out, we get a result that ranges from 22.13 to22.29. That looks to be very consistent with the22.5 percent requirement that already exists.

That question and answer allow us to begin tounderstand the drastic difference between therisk of an investment holding like a commonstock portfolio and an insurance product.

4. Modified CTE and HoldingPeriodWhen the CTE concept is applied to calculateC3 RBC, there are three differences from any ofthe above discussion. First, the calculation isprojected over multiple years. For that pur-pose, we will use our 20 sets of 1,000 scenariosto now represent 1,000 scenarios of 20 years.

◗



Chart 5—CTE90 for 500 Scenario Sets

Chart 4—CTE90 for 1000 Scenarios

◗


Second, for RBC purposes, when calculatingthe CTE, a present value is calculated for eachpossible future period under each multi-yearscenario and the worst present value is chosen.Finally, in calculating the CTE for any one year,scenarios, which would otherwise become apart of the average, are set to zero first if they arenot in a loss position. In this section we willshow the impact of each of those modificationsto CTE related to our simple test case of a com-mon stock portfolio.

First, let us look at the impact of extending thecalculation to multiple years. The RBC forcommon stocks is equivalent to the CTE90 for aone-year holding period for our common stockmodel. One year is an appropriate standard forthis comparison because there is no commit-ment on the part of an insurance company tohold stocks for any future period. Variable an-nuity products and the guaranteed benefits thatare the concern of the C3 Phase II proposals doinvolve long-term commitments. For this ex-ample, we will use our stock model as if therewere a long-term commitment to remain in thestock market. Most people have seen materialduring the 1990s bull market that emphasizedthe favorable impact of diversification overtime. That material showed that risk decreasesover time for a buy and hold stock investment

strategy. However, the mechanics of the CTEmeasure produce the opposite effect. WithCTE as the risk measure, risk appears to go up over time for several years, and then declines (Chart 6).

Why does that happen? While the cumulativestandard deviation of returns gets smaller andsmaller over time because there are ultimatelyup scenarios that moderate the down scenarios,the CTE gets worse because the make-up of theCTE90 is continually changing with each pass-ing year. As the random scenarios generate newoutliers each and every year, the scenarios thatwere moderated with the passage of time getpushed out of the CTE by these new outliers.


Chart 6—S&P 500 Normal Scenarios

Table 3

Number of Worst 100 Next 100total years at 20 years Scenariosin worst 100

0 32

1 5 10

2 4 14

3 3 6

4 6 1

5 2 7

6 5 3

7 4 7

8 5 4

9 6 4

10 5 2

11 9 5

12 3 2

13 5 0

14 9 0

15 5 2

16 5 1

17 3 0

18 6 0

19 7 0

20 3 0

100 100

Table 3 shows that there are only three scenar-ios that were a part of the CTE90 for all 20 yearsand that five scenarios crept in only in the final year. In addition, the table shows that 68of the next 100 scenarios ranked by final valueof the stock fund, were a part of the CTE90value for as many as 16 of the 20 years.

Second, we look at the impact of choosing theworst present value (Chart 7). The worst present value rule takes the forces that we men-tion above and maximizes their impact. Underthis rule, the really extreme scenarios dominateno matter whether they are followed by betterscenarios or not.

Finally, with the lowest PV adjustment, at leastin this example, the need for the no-negativevalues adjustment is not needed. There are 344scenarios to choose from in determining theone-year CTE and since no negative value isever lost, that stable grows to over 600 possiblenegative values over the 20 years. That meansthat over 600 of the scenarios show that thestock portfolio is underwater at some time in the20 years. That is not a statistic that the mutualfund and brokerage industry would have beenexcited to share when they were talking abouthow well stocks do when held for the long run(Chart 8).

In fact, if the no-negative adjustment is appliedwithout the lowest PV adjustment, then there isan impact. With our scenario sets, there areless than 100 negative scenarios for holding pe-riods of over 15 years.

This example shows that, as measured by CTE,there is negligible time diversification in this stock portfolio. With the modifications toCTE that are imposed on its use for RBC, exten-sions to holding period are measured to add significant risk.

◗





Chart 9—S&P 500 1000 Scenarios

5. Regime Switching vs. SingleLognormal Scenarios and CTENow that we are good friends with CTE and itscharacteristics, we will look at the impact ofdifferent scenario generators. A regime-switching lognormal scenario generator(RS2LN) has been mentioned as a methodthat is sufficiently robust for the C3 Phase IIcalculations.

We created a set of RS2LN scenarios startingwith the parameters in Mary Hardy’s paper inthe April 2001 North American ActuarialJournal. Over 20 years and 1,000 scenarios,the mean is 11.9 percent and the standard devi-ation is 17.5 percent. If we go back to our nor-mal set of scenarios, we found that the one-yearCTE on the return was -18.45 percent. In fact, our scenarios have a CTE90 for a one-year re-turn of -19.98 percent. That means that the tailfor these scenarios is somewhat more robustthan the tail for normal scenarios.

That more robust tail should and does translateinto a somewhat larger CTE figure (Chart 9).

The difference is less severe after the adjust-ment for lowest PV. The 20-year holding periodunadjusted RS2LN value is 45 percent higherthan the normal value while the lowest PVvalue is only 12 percent higher (Chart 10). Infact, Chart 10 shows such a regular relationshipbetween RS2LN and lognormal, that the extraeffort does not seem to add any information. Aslightly higher CTE (92 or 93) would give thesame result under the lognormal scenarios asthe CTE90 under RS2LN.

6. ConclusionThis examination of CTE has been performedbutterfly fashion. I have skipped around look-ing at whatever caught my interest. If you aregoing to use CTE as a primary risk measure,then you should perform these tests and othersto assure yourself that you understand why theCTE performs as it does. In this analysis, youshould not get overwhelmed by the sheer vol-ume of numbers that are involved. Even thou-sands and thousands of scenario results willeventually succumb to rigorous analysis.

Once mastered, CTE can be a very useful toolbeyond its required use for RBC. As shownabove, CTE is consistent with historical riskmeasures such as the C1 requirement for com-mon stocks. CTE can also be used as a generalrisk measure and as criteria for replication (See“Why Write Variable Products When You CanPut the Money Directly into the Stock Market,”Ingram & Silverman, Risk & Rewards, October2003). ✦

◗



Chart 10—S&P 500 1000 Scenarios

Coming in September!

The SOA e-mail newsletter will debut this fall, bringing you news you can use!

Get the latest details about:

Stay tuned … more details to come this summer!

◗


Risk ManagementIssue Number 2 • July 2004

Published by the Society of Actuaries475 N. Martingale Road, Suite 600Schaumburg, IL 60173-2226phone: (847) 706-3500 fax: (847) 706-3599www.soa.org

This newsletter is free to section members. A subscription is $15.00 for nonmembers.Current-year issues are available from theCommunications Department. Back issues of section newsletters have been placed in the SOA library and on the SOA Web site:(www.soa.org). Photocopies of back issuesmay be requested for a nominal fee.

2003-2004 SECTION LEADERSHIP

EditorHenry M. McMillan, FSAPacific Life InsuranceNewport Beach, CA 92660phone: (949) 219-3320fax: (949) 219-8720e-mail: [email protected]

ChairpersonDavid Ingram, FSA

Vice-ChairpersonFrank P. Sabatini, FSA

TreasurerCharles L. Gilbert, FSA

SecretaryBeverly Margolian, FSA

Council MembersHenry M. McMillan, FSA Hubert B. Mueller, FSA Ruth Y. Sayasith, FSAKen Seng Tan, ASAShaun Wang, ASA

Society Staff ContactsClay Baznik, Publications [email protected]

Lois Chinnock, Sections Manager [email protected]

Staff ActuaryValentina Isakina, ASA

Newsletter DesignMary Pienkowski, Graphic [email protected]

Facts and opinions contained herein arethe sole responsibility of the persons expressing them and should not be attributed to the Society of Actuaries, itscommittees, the Risk Management Sectionor the employers of the authors. We willpromptly correct errors brought to our attention.

Copyright © 2004 Society of Actuaries.All rights reserved.Printed in the United States of America.

Articles Needed for Risk Management

Your help and participation is needed and welcomed. All articles will include a byline to give you full credit for your effort. If you would like to submit an article, please contact HankMcMillan, editor, [email protected].

The next issue of Risk Management will be published:

Publication Date Submission DeadlineNovember 2004 August 23, 2004

Preferred FormatIn order to efficiently handle articles, please use the following format when submitting articles:

Please e-mail your articles as attachments in either MS Word (.doc) or Simple Text (.txt) files. Weare able to convert most PC-compatible software packages. Headlines are typed upper andlower case. Please use a 10-point Times New Roman font for the body text. Carriage returns areput in only at the end of paragraphs. The right-hand margin is not justified.

If you must submit articles in another manner, please call Bryeanne Summers, 847.706.3573, at the Society of Actuaries for help.

Please send an electronic copy of the article to:

Henry M. McMillan, FSAPacific Life InsuranceNewport Beach, CA 92660phone: ( 949) 219-3320fax: (949) 219-8720e-mail: [email protected]

Thank you for your help.

• SOA activities• Educational opportunities• Department updates• Exam information

• National and global issues for actuaries• Business news• And much, much more!

◗

475 N. Martingale Rd., Suite 600Schaumburg, IL 60173

(847) 706-3500 main(847) 706-3599 faxwww.soa.org

risk management july 2004, issue 2 - member | soa · by david ingram _____2 2004 enterprise risk...

Documents