A C T U A R I A L RESEARCH CLEARING HOUSE 1 9 8 8 VOL.

A P P L I C A T I O N S OF O P E R A T I O N S - R E S E A R C H T E C H N I Q U E S IN I N S U R A N C E

A r n o l d F. Shapiro ABSTRACT
A l i t e r a t u r e review w M c o n d u c t ~ i ~. p a r t of a s t u d y of a p p l i c a t i o n s of

ol~rationm-rexarch techniquu

in i n s u r a n c e .

Thim p a l ~ r p r o v i d e s a urmory

overview of s o m e of t h e l i t e r a t u r e reviewed.

2 3

APPLICATIONS OF O I N t R A T I O N S - ~ C H

TKCHNIQUES IN INSURANCE 1

The Pennsylvania State University University Park, P A 1 6 8 0 Z , USA

ABSTRACT. A literature ~evlew w u conducted u part of a study of appllcatiom of opeeatlom-research techniques ks insurance. This Paper Wovhles a cursory overview of some of the literature reviewed. I. INTRODUCTION "['ais paper is based upon a project which arose because of a coocern voiced by act,-udal students and man7 Ixactlcbq[ actuaries that they ]sad a woblem coacaptua/isio8 practical applicatio~ for Operatksas Research, OR, in tbe~ daL! 7 work, particu~17 applicatksas that are unique to the insurance business, l"ais obviously is a serious woblem. While some aspects of OR may be aesthetic, the primary thrust of t ~ subject is the development of tools and techniques to solve wactical problems." This focus can be lost if the area is viewed as abstract or primarily academic.
e

The purpose of~he project was to help alleviate this peoblem by providing a review of applicatioos of OR techniques in insurance. A major eesou~ce, in this regard, was Journals in o p e r a t i o l research, actuarial science, insurance, sad related fields d bushums. This pa w gives a clwsory overview of soma of the llterature reviewed.

In,;- research was funded in part by a llz~nt from the ActuarbLl Educatloa and Research F,-,~_ ~ o m e would regard this methodolory-oriented view of OR u too narrow. Jewel/ (1980, p. I13), for example, would Ix-afar to s t r e u the system building opportunities and areas for coostructive Intezaction within insurance, rather than tools and teclmiques of OR,

M ~ e d . :~byO. ~

I ~ . ) . I~wmz~mdP,Uk Theory,lZt~-I~. ~ ~ .

24

A. F, SHAPIRO

2. PREVIOUS STUDI1tS

T h a n have bee number of studies of the application of OR tochnklues. lbutmpies include Turban (197Z), Fabessi and Valonte (1976), and Lodhetter and Con (1977). These studies roixwtod on the results of surveys of inrso companies includiqb but not Umltod to, some hunmers. T1bere alw have bmm number of studies width have dmdt 8pelflcld]y with the inteutction of OR teludqons aud the lusuruce Industry. Them Irenoradly have taken two distinct routes: studies of OR appiicatines dlNctod at am lesu~ a c e audience and studies of OR appllctions in ~ c e directod at an OR audience. Commoal7 referred to examples of the fresher are the studies of Zubey (196S, 1969), Denambeqi (1968), and W a d , I t . el. (1970). Examples of the later m found in Suttoa (196.5), Halmstad (1974), and Jewel1 (1974). A number of these studies roportod that there w u not iPmat deal that specificU7 related OR to Inmaranc. situations, and many of the applications alludod to had not yet been implemented. Part of the focus of this study is to see hew t h , literatur, hes clumsod in these reserds.

3. GA)d]~ THEORY
Game theory involves competition or conflict between two or more decision makers, aad is concmmod with prescribin8 best strategies. Specific insurance app]/catio~ include topics such u insurance purchases, tales, underwritlnS, managememt, expense allocation, reinsurance, and premium determination.
3.1. insurance ConmunFi~

WUlinms (1960) discussed the use of pure strategies in iPtme theory for the evaluation of iuu~anc, couumptina alternatives, The analysis w u based on in utility uesciatod with the decision of whether or not to buy fire insurance. 1"he notice of a wots'y factor w u used to help explain imsurmuce purchases. Further budsht into the development of this mode] w u contained in Williams u d Dickeron (1966) mud u empirical investigation of the model w u reported in Neter and Williams (1973).
3.Z. O p t i m a l

Inmmutc, Conflict

While two--peno~ zm~sum 8ames have serious shortcomings in many Wactica] situations, owin8 wimarUy to ,1.. htck of total knowiedge and zero-sum co* d i t i m ~ they can Wovide usof ,,I modals in the proper circumstances. They set the stqi* for ]dUlor (197Z), for example, who used expected utility maximization to show the conflict between the bunu-er and lmnu-ed is~ofaur as the optimal type of insurance contract is concerned. 3.3. Insurance Sales Game

Bran (1966) envisioned aa insurance sales pine phtyed by three players: the

- 25

AIIq'LICATIONSOF OPERATION~RE.SEARCH TECHNIQUESIN INSUIL41~E

bsursr, tiM, a l ~ t , aad the b q w . The t i n t IdaSe of the p u e Isvolved a ceeUt k a b e t m m the humm'and the q~mt aplast the bulmr, wldl, the last Isvolved a ~ b . t w m the b m a l aml t h . qpmt.

ementhl] feature ot the t i n t phase was u attempt to uudmlso t h , x p o c t a b ~r a policy u i , , w ~ tiM, m l ~ l ~ w u the i - - m h m coet o t l W tlum commladon, and the ~ expoctatkm was t h , ]x~duct of the
~ c t r e v o h ~ about the Woi~rtSoe et the ~ which d m k l be allocated to omubek~4 and w u viewed u two-peme cool~aUve p e t .

3A. Life ][nmmuzce Unden~thql

I.,mab.e (19SO) ~ t b . w o b l s m ef Inmrance uadeewritlnl~ within tiM,

centezt of a two-IMMmom~m-COOlperative p i n e b.twen~ u ~ (pla w 1) and potential poUcybolden (player Z), who m la e i t h ~ a helLIthy ca" noabe~th~ state. The minimax rltm4on, which p r e s u ~ mal@vohmt appl/clmts, and the Bayes crltm4oa, wldch assumes the appUcant has adoptml a fixed priori strate~r, were coasldered.
$.S. Executive Game

One of tiM, . . r l l ~ r models of a p r o . for a prOlX~ty s~l nabIUty iasm~mce company was the executive game foe o f f i c e aad middle m a a a ~ m ~ t suggested by U c G ~ . . u (1960).
The inputs to the mode] lncludmi basic dscidom~ usumptiom4 and data while the outputs of the modal w m reports to the playm5 aad aa analysis of the effect of their decisioas. Assumptions ~cluded those that w m known to the players aad t h o ~ which described the ~virmtment and w m aot known to the p~ym's. The data included such iafoematice u ecoaomlc activity, underwrltb~ experience fluctuations, and cost of traiahq agents. Two limitations of the l a m e were that no Inventmemts were involvQd and ratin~ bureaus did not play a r o l l
3.6. Cost A l l ~ s t ~

Lamai~ (19114) discussmi the applicatien of p m e t]~m 7 to the problem of s21ocatln8 mtlmmm a m c ~ Uul departments of an Immraace comi~n 7 when coop~ a t l o a leads to eccmomiss of s c a ~ .
He t i n t showed that the oet a/]ocatioa woblem was identical to the value of a i [ a m with t r a m 4 m b l e utiUtimk aad than d/scussm/the attributen ot four coet aUocatica methods b e s e t ~ p m e theory. The criteria advocatml m oll,ct i v ~ ratiomd, In the m that no departments subsidize anothm', moaotoaic costs, la that an departments coatribute to an i n c r . a s , la i~ol~l coets, a~l additivttT, la the m that a subdivision of a department does aot a f f e c t the cost aJ/ocatioa.
3.7. R . c i w o c ~ R~msuNmce Treaties

A simple modal foe smdTsing a reciFrocJd t~dnsursace treaty w u diJcuss.d

- 26 -

A. Ir. SHAPIRO

by Boech (1960) within the context of two-pero coopeeatlve p m e . AJte: viewing basic 8=doms rolktln risk situations sod ~bowhqj that the quota s h a ~ treaty will best roduct the variance of the portfolio of u c h company when the utility index depends only on the free reserves and the variance of the risk distribution, he 8on* on to discum the nature of the quota ceded and the free rmmrvem tnuufe:rod.
3.8. The Atum7 Against Natm~ Bis~dmann (1976) investitated s a m e which pitted the actuary sSainat nature in the ccmtext of estimatin8 pm~ premium from sample information. The situation involved the actuar 7 who chose psbr of estimators, nature who chose parametric family of distribution functions and structure functiou, and quadratic loss function. Both the non-parametric c u e , where the parametric famfl 7 w u not known, and the parametric case, whe:e the parametric famU 7 w u known, w m conside~L

4. STOCHASTIC DOMINANCE Stochutlc dominance (SD) provides means for preference ordering when unce:tln alternatives are involved. The three orders of SD are based on lncrnasls~ly restrictive rots of utility functiomc lncrestng, risk adverse, and dec~asing absolute risk adverse. P o t u t l a l a r e u of application include optimal insurance c o v e : q u and reinsurance. 4.1. Insurance and Risk Sharing l)ohe:ty (1977) used second order SD to develop model for investtgatb~ the appropriateness of deductible from the vantagepoint of an individual risk-adverse lnsurecL He late: extended this approach (Dohert7, 1980) to model the ix-ef e : a c n of both the insured and the insure: insofar u the use of deductibles, colMurance, the franchise form, and first loss.
4.2. A Reinstance ~Hmple

Gandhi et. sL (1981), discussed a simple r e t n s u r u c e application where the mature: of stock company must ~ between two p0etfolioe, o n e c o o t a i n t n o reinsurance, and the other not. The paper described in detaU the characteristics of first, secom~ and third-order stoch~tic dominance, and discussed the superiority of stochastic dominance over mean-variance, coefficient of variation, and expected utility models.
S. LINEAR PROGRAMMING

Linear programming has l o q been reco~sized as one of the moat important tet'hnlques of OR because of its versatUity and power in resolving problems involvi~ the allocation of scarce resources. Specific applications in instance include instance purchase optimization, insurer profits maximization and the bounding of stop-lo-.a premiums.

- 27 -

APPLICATIONS OF OPIERA'r}oNS-RIESEAR~I TE(IlNIOUE5 IN INSURANCE

S.l. Life ~ c e

A permminl problem la the sale of life insurance b the opttmaJ combination of various types of life insurance policies and 8 1 t e r u t e investments. Schleef (1~J0) showed how a Une~ programming model could be used to belp resolve this psq)blQm.

The objective of the model w u to maximize the premeut v81ue, adjusted for the marginal t u rote, of future cash flows due to cult v i l e recovtrl, ~ ud other invutmentL The toetaints of the model included a budget c o s t r U t , which provided for the pitmen of premium and loan interest and recognized altem~ttlve investment cash flows, a death benefit coczstraint, which provided for the desired hrvtl of death benefits, and noa-4sesative cocmts~tints en the alternative investment fund, the cash v81um, the loan balance, the face amount, and all dectstoa variables. 5.2. Profits of Property and LiatdUty
Hofflander and Drsndell (1%9) used a linear programming model to mazimise the profits of a property and liabilit7 insurer subject to operational and regulatory constraints. The purpose of their model was to find the optimum allocation of u s e t s so that profits could be maximized. 5.3. Stop-Loss Premiums One way of reinsurh~ is through a stop-loss arrangement, under which the reinmsrer pays that portion of the loss in excess of a specified amount. A concm~, in this ~ l , is the maximum value for stop-loss la~emium. TaTlor (1977) used a linear ps~Fsmmis~ technique to help isolate the upper bounds oa stop-loss premiums. His approach was to ~ r a ] / z e from an fllustratioa which validates that if there w u a discrete claim distribution, the maximal stop-loos dlstritmtlms has its mass concentrated at the two extreme points of the range. 6. GOAL PROGRAMMING The essential feature of goal programming is that it provides an opportunity to assign priorities to conflicting objectives and then minimizes d e v l a t i o u from those objectives. The method provides an opportunity for three types of 8sudysts: input requirements, given a 8oal; relative 8oal attainment, given resources; and test for an optimum solution. An important benefit of the model is Its simulation capabilities. 6.1. Insurance Agency Management G l e u o a and T.my (1977) examined goal programming as an q e u c y decisionm&b;,,~ tool in the context of property and casualty insurance agency decisions regarding number of insurers to represent, cost reduction efforts, and expanded

28

A. F SHAPIRO

commercial lines. Rather thaa deal with aU pomdble 8oala that might coofNmt such an aiPmcy, they Umtt themselves to commou 6.2. Profit Maximistion in Property and Lisbilitlas Companies Klock aad Lee (1974) aqFaod that a linear woFammh~i approach to peofitability, even if it is extended to include other utility function characteristics, while theoretically cms~ct, is imwactical. They sulisasted 8onl ProlFmminii s a,, alternative. The authors applied a 8oal-peosrammin 8 model to the H-D model (HofflaJ~em & D~andeU, 1969), with some modifications, to trandorm it from a linear proiffammlng format. They identified six 8oal c o u t r a i n t s which m , in dectsasinlj Wiority: liquidity, fimmcial strength, Wofit, distribution of asset portfolio, distribution of premium portfolio, and use of capacity. The system comtreints involve policTboldm' surplus, capital, premiums written as multiple of policyholders' surplus, toted assets as multiple of Ix~mium volume, Ix~minm hahmcas as prOlX~tina of premium volume, dlstrlbutlon of assets, Ix~oportiou of misceilaneous liabilities, Ix'oportinu of Ions rmmrvas, IX'Oporticm of unearned premium rqmarvas, total assets, and equality of assets am/lfabf/ithm plus net worth. OsuD of the obvious questions to be resolvod Immfar as 8oal ProlFamminli is concerned, is the similarity between it and linear ProiFamml~i. In this regard, Draadell (1977) demonstrated that the goal proiFamming model of Klock and Lee produced results which were equlvadent to the H-D model The basis of the comparison w u a formulation of ~ into liquidity, stability, and profit. 6.3. Capital Budl~tinli in Property Liability Companies Law,once and Rstvas (198Z) formulated the capital budgeting in a IX'Opert7 liability insurance comlxm7 as a 0-I ~ IX~NFamming problem involving multiple S~xds and indivisibls activities. They viewed the model as being more rewesamtative of reality since it was not comtrainml by a single objective or continuous activities, while bein easy for decision makers to understand and use. For the purpose of examples, seven projects wens cemsidetsd which cover three time periods and numlx~ of regions of op~atinna. The constraints of the model included seven sets of 8oal constraints corr~pondlnli to the objectives. Addltimudly, thm~ wore constzaints to reflect indivisibility, single period nature, independence and/or depeodonce of the projects.
7. Q U A D R A T I C PROGRAMMING

C~adrntic Wogramml~ is concerned with optimizing quadratic objective

function subject to linear constrsints. Areas of application include portfolio ana-

lysis, from both a financial and insurance perspective, and international reinsur~ce.

29

AIqI~ACATIONS OF OPERATION~RESEAR('II TECHNIQUES IN INSURANCE

7.s. p w t t o u o
O u of t h e most commoa uses of quadratic p r o ~ m m l s q j is in the resolutims ot qu~tlmss Nlatod to p~.t|oUoL ~ ~ d rids ai~molch was th@ wot.k of )darkowits (1~2)~ who Nlllestmi ushq prolutblllty estimates of future mcurlty perfm~uusce to develop an efficient set of p0etfolioe which could them be mstcbod with u lawmtor's preference.
Siuurpe (1%3) discussed an efflcimst remlution of the problem which M u t l y

etmpUned t ~ mud3nds.
7.7.. Portfolio Theory and Noa-Lffe I u u r m ~ Markle and H o f ~ (1976) applied the Markowits model to Me-life lnmmess under the ammmption that the 8oal is to maximize returns for Kivex hrveis of risk subject to lestitutimssl solvency constraints. The objective functioa is the underw~tinli aad investment profits.
For this purpose, the ovm~ll return for a 84ven line was assumed to be sum a v o r q p ntturu for that l i n t oiler the period of investllp|t|on, u wore the expected returns from securities, ~

7.3. intm~atloual Reinsurance Loubeqp, (1983) used quadratic prolp'amminll to model intmmatiomJ rolsumrance operatinas within a mean-varinnce context. In add/tlon to addre~inli standard aqiumants against such an application for an insurer in 8eneral, he dealt with q u a t i m w such u the nonnoeqna~ty of exchange rates and the distribution and estlmtiau of tJ~ ~ t u r u of t]~ excmu of ] c ~ treaty. 7.4. Life lmmrance Company Investments The common aHa.oach to the analysis of s t t ~ n a t l v e investment opportunities is based on quadratic prc~amming medals because the objective functioa involves the variance of distributiou, a quadratic. Brodt (1983) showed how to develop linear prc~ammin t alternative based on the mean abeolute d4rvintioa of returns, rather than the v ~ c e of returns. The objective w u to minimize risk, u measured by mean absolute deviation, subject to butra-temporal and Lntm'-tempor~ constraints. Since the objective functlms w u nonlinear, goal proFamming was resorted to for an actual solution.

~luusce-constrained prosrammin has also been used to address this problem. However, the use of clsance~onstrainod prosrammint model to i n v u t l irate the risks associated with investment and underwrithql returns has been criticized by Kahane (1977), who arsued that full quadratic programminz model is less restrictive aad provides sdmUar results, given the u m e ~mumptioas.

3 0

A, F. SI'IA~RO

8. DYNAMIC PROGRAMMING
Dlmamic prosrammh~ extends the s i q l e - s t a p assumption of ~ srammin8 to a muitistaBe envlroament, and is coacernad with the overall effectivmne~ of ~qumstlal, intm.rehttnd decisions over the plannisq horlso~ Areas of application include life insurance prosramming, risk mana4~meut, aad no-claim Umits in automobile hssuronce. 8.1. Dynamtc Life Imurance Prosrammhqj One of the e~rly attempts to brbq dynamic peoFammin8 to bexr on the problem of chooeiag an optimal life l u u r u c e peosram was the study by Belth (1964). The study w u naive by c u r ~ n t standards, but the i n t ~ t i o n s h i p of the variables is u relevant today u when they were first doom. 8.Z. Risk )4wqremont Lilly and Glnason (1977) discussed the use of dynamic lX'Cqp'ammin8 u u aid to the risk ~ in equipment replacement decisions whas~ the ~ is to minimise potential losses. The problem is that the risk mam~er must balance the replacement of the equipment while it still has economic value against peteathtl direct and indirect losses involving business interruption loss, budUy injury and property damage loss, replacement cost of damage to manufacturer, workers' loss, and loss of current boiler. Backward dynamic programming w u used to resolve the in~blem, where the stages were the years of the planning horizon and the state within the stages were the years of ~ of the boiler. 8.3. Optimal No-Claim Limits The la~d~lem commoa to all motorists is the optimal choice of when to make a claim. Motorists have an informal N t of rules whick they follow, under which claims are made if the extent of the accident exceeds some minimal amount, and no claim is made otherwise. This is a standard type of I~oblem in Markov Chains, since it intmllzIatw insurance claims during a given period and insurance l~emiurns durbNK subaequont periods. Martin-Lof (1973) discussed insurance with a bonus system under which annual premiums depend on the bonus class to which the insured belongs. An optimal decision rule was dmvaloped based upon an adapted policy iteratims which reduced the Mze o! the system which needed to be solved st each iteratioa. SimLlaz problems mere resolved by ILaekm,a von Lanzenauer (1974) and Hastings (1976) who dealt with determi-;~.~ whether to flit a claim a f t ~ u accident when there was more than one possible accident during a given policy year.

-31

AFIq.ICATIONS OPERATION~tE,~ARCI-ITECHNIOUIESIN INSURANCE OF 9. INVKNT~RY MODKL bsveatory modeis seek tim optima] bahmce betwmm the coet of holdksj Inventory and the coet of peocm4~ it. Specific hm~ance appUcatimss include the makstemance of cash u m e r v u and the Wocueement of ksmemsce ~ .

9.11mmNe Cash Reeerves

White and Norman (1%S) formulated the cmstrol of cads N u r v u resolved m/sq d y u m i c peosrammksg. 9.2. Optima] ksm~ance Coverage Smith (1968) envisiooed the optimal ksm~ance coverqle ks the coatext of -,, optima] ksve,~tory stocksKe under uncertakst7. Specifically, ff 8n [rotatable l o u (demand) is exceeded b7 the insurance covetsZe (ksventoey level), excessive lnsurmsce coat (ksvantor7 holdbqj co6t) is t~'urred. Convenaly, if kssurable lores is greater t h ~ insurance o v e r a p , unr~ecoverable lomws must be al~x'bed by the of an
Insurer as an ksvlmtm7 cmttt~l problem where s t o r a ~ and I P ~ t y costs m well defined. The optimum value f o r - t h e ehoet-term effective bank baksnce was

insured. The problem, then, becomes oo. of ~

tory) ksv.ls. 10. NETWORK MOD~-~

optima] Im~ance (ksvan-

The standard applications of network models for the admkslstratioa and coot'db~tioo of proqmcts am/ orgaa]satioas were mudly trsa_~ferable to the insurance anvlronment, as were tlm plann/qj am/ coatro] te,-h,,~mm of P E R T and CPM. Specific examples of applicatioa m cash flow management and Imsimms ksterruptioo downtime. 10.1. Cash Flow Management Managemlmt of a property-casua]ty insurance company has two primary respomdbUltias. The first rupoodbUity revolves about the Nksctloa of an appropriate composition foe tim insurance portfolio, Com~eratloas in this area include which lines to ~ tim amount of coverage to be provided ks each line, and marketksg mmideratimm. The other respouibilit7 revolves about the insurer's investment portfolio. Here, coustderations include the proper mix of securities, apSWOlX, t t , matmdty dates, the limit oa investment in any particular security, h aad so oa. Crum and Nye (1981) applied network model to the problem of manaSks8 the overall cashf~w of medium-size insurer. The thrust was to coordJaate the insurance underwrtth~ operations and the investment portfolio pursuits. lO.Z. Business ksterruptioa Downtime
Close (1982) ~ the application of PERT to the reduction of downtime due to business interruption. The system discussed w u applied to the situation

3 2

A. F SHAPIRO

where a f i n l e e occurred, and network was 8iven dsowlng the | t q e s betweeD the firs 1oe8 and the m u m p t i o n of the former level of production.
11. QUEUING MODELS

The numerous applications of queuin8 theory used to resolve waitin4j line problems of business m dis~ctly pplicJble to the JJuursnce industry. An application, bowevw, which is unique to insurance has to do with the ruin of an insurance enterprise. 11.1. ~ Ruin u QutuinE Problem

Under proper scenario, the solvency of an insurer may be viewed as queuing Woblem, whets the pt~habUity of not bein8 ruined by some time t, u(U,t), is essentially the same probability as that involving customer waiting less than some time U, 8/yen that the customer }oined the queue t periodl after the server w u f r o . A discussion of this aspect is 8iven by Seal (1978, Chape: Z).
lZ. MARKOV PROCESS

A Markov process provides dynamic system under which only the immediate past is relevant to the pt~liction of future behavior. Its application to life tables is an obvious one. 12.1. Workini Life Tables Since work/rig life tables evolve from the dynamics of Labor force participation, Hoem (1977) advocated that such tables be produced using the theory of continuous-time Markov chains. The trusitinus in this case are due to deh, accession to the inbor force, and separation from the labor force. His discussion of an application of such model is based on previous study by him and Fens (1976), which contain the formal details of the model Similarly, Braun (1978) emphasized stochastic stable population theory where the forces of fertility and mortality depend on age, parity, and place of residence. 13. SIMULATION Because of the easy access to computers, the use of simulation within insurance and related industries has become commonplace. Parsllel,-~ this application of computers and simulation, has been the need for the development of models which adequately encompass particulars of specific a r e a within the industry. A number of these models are discussed by Pitcco elsewhere in this volume. Representative examples of areas of application include variable life
insurance, merit-rating, and reinsurance.

3 3

APPLICATIONSOF OPERATIONS-RESEARCHTECHNIOUES IN INSURANCE

13.1. Zquity Linkod L/re lasurance Policies Bremum and Schwarts (1979) used dmuLatlon to m q d m F~k-tsducin8 investmen strateSins amochtted with equity linked life Inmmutcs poUcteL T h e n are hnplicst|oa8 for the devuiopment of Bayesian aPSW~ch foe flndin8 estimates of the model's parameters. 13.2. MerJt-Ratlnj Plaas Ferreir (1974) modeled driver's annual premium as )knkov process and used compound Pokson accident model and Monte Carlo dmuLatiao to compote 50 year premium histories in oeder to investigate various merit-rating plans. 13.3. inwuret and Reinsurer interactioo Galits and Brown (1981) discussed the qualitative nature of simulation model for insurance and reiusu,rance operatioN. A l t h o n ~ the a~dstiomddps w m not specificit7 defined, important overall coasideratious were delineated. The basic components of the model were the surplus, capital, unearned wemium reserve, and the l o u reserves. The model used flow-driven system which fixed the varLables at some starting point, aad the flow of variables therefrom, in oed~ to determine the state of the initisJ variables at some Later time. 14. OR SYSTEMS AND MODELS OR authorities view the "systems appesmch', which coodintea overall reintlonships and interdisciplinary teams, as the fundamental thrust of OR. Representative applications include the s e a s of popuintion pinnnint, the insurance industry and wofkmon's compensation.

14.1. Popu~tU~, puuu,~

R e i ~ e (1970) d/JcuJed the role of model ~ i,, popoLatJoa p h n n i q in underdeveloped countries. The problem was that t h i n was a cooflict between natiomd and family goals, sparce relevant information, and limited resources. Within this framework, the role of OR was to analyze the decision process and organize profession] activities in this area.

14.Z. The Insur--ce Industry

Pentikainee (1983) discussed 8eneral-ptnl)oN model fo~ stud]d,,8 the solvency condition of broad range of insurers. 1"me model includes w o v i s i o ~ for state variables and parameters, such as portfolio mix, reserves, and assets, exosenous factors, such as business cycles and inflation, and business strateKies, such as retention levels, investment policy and sales effort. An important attribute of the paper is that it discussed many facets of the mode] which require consideratioa.

3 4

A.

F SIIAPIRO

14.3. Workmen'a Compensation Inmu.ance JowoU o t el. (.1974) discussed the muitldisciplinary nature of comprehensive proJoct lnvolvin8 workmen's compenution lnmu.ance. Each phase of tbo proJoct is discussed, u are the iaterreiationships of the phases. The study embodies the spirit of model bufldinj which OR authorities stress. 1S. COMMENT This article wovides on]7 a cursory overview of the literature pertalnin8 to applications of OR techniques in bssuraoco. Nonetheless, it is hoped that it stimuiates discuss/on and wovides direction sad insight into Eurther research in this area. To the extent it has met these criteria, it wiU have served its purpose.

16.

]SZBUOGR~JPHY

Belth, J. M. 1964, 3~ru,,,ic Life Insurance ProlFemmins,' Journal of Risk end Insurance, 31 No.t, pp. 539-556. Borch K. 1960, 'Reciprocal Reinsurance Treaties Seen As A Two-Person Co-operative Game,' Sksndinavisk Aktuarietidskrift, 43 No. 1-2, pp. 29-58. Botch, K. 1968, The Economics of Uncertainty, Princetoa University Press. B r a ~ , J. M. 1966, 'Prices And Commissions Based On The Theory Of Games,' Journal of Risk end Insurance. 33 No. Z, pp. 169-193. Braun, H. 1978, 'Stochastic Stable Population Theor~ hs Cootinuous Time,' Scandinavian Actuarial Journal, No. 4, pp. 185-203. Brennam, M. J. and F.. S. Schwartz. 1979, 'Alternate Investment Strateg4es For The Issuers Of Equity t~;nked Life Insurance Policies With An Asset Value Guarantee,' Journal of Business. 82 No.l, pp. 63-93. Brodt, A. L 1983, 'bibs-Mad Llto: A Multi-period Optimization Model For Life Insurance Company Investment Decisbsna,' Insurance: Mathematics & Economics, 2 pp. 91-102. Biilhmann, IL 1976, 'Minimax Credibility,' Scandinavian Actuarial Journal. No. 2, pp. 65-78. Close, D. IL 1982, 'PERT Networks' Use Can Help Reduce El Downtime,' Risk Management. 29 No. 3, pp. 26-30. (:rum, ]El. I.. and D. J. Nye. 1981, 'A Network Model Of Insurance Company C u b Flow Management,' Mathematical Programming Study. 18 pp. 80-101. Denenbeqi, IL ~ 1968, 'A Review Article - A Basic Look At Operations Research,' Journal of Risk and Insurance. $S No. 1, pp. 159-163. Dohmrt~, N..6,. 1977, 'Stock-tic Choice In Insurance And Risk Sharing,' Journal of Finance, 32 No. 3, pp. 921-926. Doherty, N. A. 1980, 'Stochastic Orderbsg Of l~sk/insurance Exchanges,' Scandinavian Actuarial Journal, No. 4, pp. 203-208. DrandeU, M. 1977, 'A Resource Association Model For Insurance Management Utilizing Goal ProiFamming,' Journal of Risk and Insurance, 44 No. 2, pp. 311-315.

3 5

Alqq.JCATIONSOF OPERATIONS-RESEARCH TIECIINIOUESIN INSURANCE

Fobom#i, F. J. and J. Velento. 1976, 'Mathouatlcal P u q F t m m J s l In A n e d c a a Compuinw A Sample Survey,' InterhN:os, 7 No. lm Pp. 93-98. lrm~elra, J., Jr. 1974, q'ee L o q - t e m n Effects Of Idm4t-Ratin8 Plans Ou Individual Motorists,' Operations Research. ~t No. S, pp. 954-978. Freifeidar, L~ R. 1976, A Decision Thoo~JtJc A p p r o x h to Ineurenee Rmtomekin9 (Pennsylvania, Pa.t S. S. Huelme~ Foundation for Insuruco Education). Gall, s, L. and Id. Brown. 1981, 'AS/RI The Advanced Simulatioa Model Of lnmmmce And Relnmursnce C)pe~tkm~' Ge4~evs hpoa~ on Risk sod Insurance. I0 pp. 85-96. Gandhi, D. K., A. Sounders, and L G. Susars. 1981, 'Stochastic Dominance An AppUcatioo To The Insurance PoetfoUo Deciskm,' Oenove Pspeqs on Risk end Insurance. 21 pp. 51-62. G~ber, H. U. 1979, An Introduction to Mathematical Risk Theory (P-.~-Tlvanin, Pat.: ,5. S. Hueboe~ Foundatino fro" Insurance F.ducstloo). Gleason, J. M. and C. C. LIUy. 1977, 'A Goal ProFammtnG Model Foe Insurance Agency ManaEement,' Decision k J o n c e s . 0 No. 1, pp. 180-190. llanhlin8 yon Lansonawr, C. 19"74, 'Optimal Claim Decisions By P o l i ~ In Automobile Insurance With Merit-Rating Structures,' epees,Sons Roansrch. n No. S, pp. 979-990. Kshnstad, D. 0. 1974, 'Actum-ial Techniques And l"eeir Relations To Noninsursnce Models,' Opera,Jane Research, 22 No. S, pp.942-9S3. Kasti~,s, N. A. J. 1976, 'Optimal CLaimi~ On Vehicin Insurance,' Operational Research Guertoriy. 27 No. 4, pp. 805-813. H111iw, F. S. and G. J. L~berman. introduction to Operations Roansrch. 3M (Sea Francisco: Holdan-Day, Inc.). H o s e , J. M. 1977, 'A Markov Chain Model o f Wot4cr~ Life Tables,' Scandinavian Actuarial Journal. No. 1, pp. 1-20. Hose, J. M. and M. S. Foe8. 1976, A Merkov Chela Model of Working Ufo Tobies. Copenhagen Univ. Lab. of Actuarial Math., Woekin Palx~ No. 2. Hofflandmr A. E. and M. DrandeIL 1969, 'k LAnuz ProiFamming Model of ]hrofitabflity, Capacity, And Resuintion In Xnsumumce Management,' Journal of Risk and Ineurmnce. 1MSNo. I, pp. 41-54. Jewel1, W. S. 1974, 'Operations Research In The ~ e Industry: L A Survey o f Applications,' OperaUons Research. 22 No, S, pp. 918-928. Jewel/, W. S. 1980, ~dodels in Insu~tece: Psz~disms, Puzzles, Communications, and Revohttions,' Transactions of the 21st Intornmtlenel Congress of Actuaries, PP. M7-S141. JeweU, W. L , T. L. Johnston, and S. S. Leavitt. 1974, 'Operations Research In The Insurance IndusUT~ IL An Application in Claims Operatiom Of WoHunan's Compensatine Insurance,' Operations Reseerch. 22 No. S, pp. 929-941. Kahane, Y. 1977, ~[nsuruco Expomu~ Aud Investment Risks: A Onmment On The Ues Of Chance Coos,rained Pro,ramming,' Operations Research, 2S No. 2, pp. 330-337. Klock, D. R. and S. M. Lie. 1974, 'A Note On Decision Models For Insurers,' Journal of Risk and Insurance. 41 No. 3, pp. 537-$43. Eawraace, K. D. and G. R. Reeves. 1982, 'A Zero-one Goad Proorammi~ Model For Capita) Budsetins,' Computers and Operstions Research. O No. 4, pp.
303-309. Ledbett~r, W. N. and J. F. Cox. 1977, 'Are OR Techniques Being Used?' Industrial Engineering (February), pp. 19-21.

3 6

A , F, S H A P I R O

Z,emahre, J. 1980 'A Game ~mcxry Lock At Life lasurance Underwritbqj,' Aatln Bulletin, 11 No. 1, Pl~ 1-16. Lemabre, J. 1984, 'An AppUcatlae Of Game ~ Coet AUocatlon,' Astin Bulletin, 14 No.J, pp~ 61-81. IJUT, C. C. sad J. M. Ghm~me. 1977, 'A Dynamic PNqpammla8 Approach To ECl~tpment Loss Aadd~j~' Journal of Risk and Ineursnce. 44 No. 2, pp. 30S-310. Loubeqre', H. 1983, 'A Poetfollo Model Of hsternatlaeml Reinsurance Operations,' Journal of a s k end Insuranco, 80 No.l, pp. 44-60. )ku4do, J. I . and A. F.. Hofflander. 1976, 'A Quadratic ProFammla 8 Model Of The Noa-Ute ~ , ' Journal of Risk end Insurance, 49 No. 1, pp. 99-120. Mukowits, H. M. 1952, ~xrtfoUo Selection,' The Journal at Finance, 12 (March), pp. 77-91. ldartla-Lof, A. 1973, 'A Method Foe Finding The Optimal Decislaa Rule For A Policy Hohier Of An Insurance With A Bonus System,' Sksndinsvlsk Aktuarietidskrift. H No. 1, pp. 23-29. McGulaaess, J. S. 1960, 'A Msnai~WLi Game For AN l a s u r u c e Company,' Operations Research, No. Z, PI~ 196-209. MUler, R. B. 1972, 'Insurance Contracts As Two-Pwmm Games,' Management Ik:ienco. 18 No. 7, pp. 444-447. NaJrrsl~e~ F...4. 1980, A Study Manuel For Operations Research, (Chicago: University of Chicsgo). Neter, J. and C. ,6,. Williams, Jr. 19"73, 'Performance Of The Expected UtUity Method And Two Other Normative Methods In lasursnce Decision Making,' Decision kionces. 4 pp. 517-532. Penti~slaen, T. 1983, 'On Model B u U d ~ foe las~rznce Industry,' European Journal of Operations Research, 13 No. 3, pp. 310-325. Relake, W. A. 1970, "The Role Of Operations R~sHrch In Population Plaanlag,' Operations Research, 18 No. 6, pp. 1099-1111. Schleef, IL 1980, 'Us/aS I J n e u ProlFammlag Foe Planning Life Insurance P u ~ ' Decision Ik:ience8. 11 No. 3, pp. 522-$34. Sea], H. I . 1978, Survival Probabilities, (New York: John WUey & Sons). Shapiro, A. F. saSJ. M. McAdcm. 1982, Operations-Research Applications for Pension Consultants and Administrators, (Washington, D. C.: American Society of Pansies Actuaries). Sh~rpe, W. F. 1%3, 'A SflnpiHiml Model Foe Portfolio AnalTsis,' Management Science, g NO. Z, pp. 277-293. Smith, V. L. 1968, 'Optimal Insurance Coverap,' Journal of Political Economy. 76 No. 1, Pl~ 6&-77. Suttoa, R. L. 1965, 'Operatiomll Resmurch On Actuarial Work,' Opsosrch. 2 No. 3-4, pp. 17-22. Tayloe, G. C. 1977, 'Upper Bounds 0 . Stop-loss Premiums Under Coastratnts On Claim Size Distribution,' k s n d i n e v t s n ActusrJsl Joumsl, Vol No. 2, pp. 94-105. No. Z, pp. 94-105. Turban, E. 1972, 'A Sample Survey of Operations-Research Activities at the Corporate Level,' Operations Research. 20 pp. 708-77.1. Wade, R. C., B. C. Rosomeatich, and E. Y. Kuag. 1970, Operations Research and Insurance Applications: An Annotated Bibliography. (Bryn Mawr, Pa.: McC,ahan Foundation). White, D. J. and J. M. Norman. 1965, 'Control Of Cash Reserves,' OperstionaI Research Quarterly, 16 No. 3, pp. 309-3?.8.

37

APPLICATIONSOF OPIERATIONs-RrSEAR('II'ITEC'HNIQUE5IN INSURAN("IE

WIUiems, C. A., Jr. 1960, 'Game Theory And lasm.uce CoasumFIoa,' Joumll of Risk Imd Insurance. T/No. 4, pp. 47-56. WlUlems, C. A., Jr. sad O. D. Dtcbnmm. 1966, 'Gsme Thew7 sad luunaco Consumptlcmm The Worry Fectoe Rev4slted.' Journal of Risk end Insureneo. n No. 3, PI~ 411-4ZS. Zubsy, E. A. ~1%5, T e a ~ l ~ t y Stud,/ Of Opm'atioas Research In ]~ursnce,' Journsl of Nsk end Ineurence, 112No.3, pi~ 325-336, Zulmy, K. A. 1%9t "Flue Role Of Opetatloea RewmJrch And LJfe Inmmuce,' Amorlcmn Society Of Chmrtorod Lifo Underwritors. aS NO, Ip pp. 56-61.

3 8