A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Vito FRAGNELLI Dipartimento di Scienze e Tecnologie Avanzate Università del Piemonte Orientale [email protected] Maria Erminia MARINA Dipartimento di Economia e Metodi Quantitativi Università di Genova [email protected] Second FIMA International Conference 2008 ”Energy and Environment: new challenges to mathematical modelling and applications” January 21-24, 2008 - Champoluc 1 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Summary Setting Problem Co-Insurance Games General Results Constant Quotas Proportional Allocation Case Study Concluding Remarks 2 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks 3 Setting A large risk, e.g. environmental pollution risk, has to be insured It may depend on firms that in their production processes may have as a side effect the release of polluting wastes, can alter the standard environment, influence the possibilities of using it damaging public or private goods, compromise directly or indirectly the human health, contaminate biological resources and ecosystems Consequences may be described as damages to persons and/or materials or interruption of various activities (industrial, agricultural or recreational) A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Possible environmental pollution risks: air pollution emissions harmful gases exhaust fumes stenches waste disposals chemical productions water pollution effluents, when discharged in the rivers soil pollution rubbish solid wastes industrial (chemical manures or pesticides) marine pollution wastes from coastal firms sewage oil tankers accidents acoustic pollution high noise level vibrations third party liability for damages caused to persons and/or materials → insurance policy (Bazzano, 1994) clean up costs for removal of pollutants 4 5 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Losses for environmental pollution can be very heavy Costs refunded after main accidents in OCSE countries 1976-1988 Naval transportation accidents are not included Year 1976 1978 1982 1984 1985 1986 1987 1988 1988 1988 Place Seveso (Italy) Los Alfaques (Spain) Livingstone (USA) Denver (USA) Kenora (Canada) Basilea (Switzerland) Herborn (Germany) Floreffe (USA) S. Basile (Canada) Piper Alpha (Northern Sea) Cost ∗ 103 15 41 20 7 16 8 67 39 111 Cause (pollutant) Chemical plant (dioxin) Tanker truck explosion (propylene) Derailment (toxics) Tank (gasoline) Spill (PCB) Fire with river Rhein pollution Tanker truck Tank explosion (oil) Fire (toxic wastes) Explosion (gas) * in millions of euros SCOR NOTES (1989) Asbestos pollution damages refused by Lloyd’s in about twenty years: 15 billions of euros (some trials are still lasting) Pool of companies 6 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Division Plan for the Italian Pool for Environmental Risk Insurance 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 COMPANY ALLIANZ SUBALPINA LE ASSICURAZIONI DI ROMA ASSICURAZIONI GENERALI ASSIMOCO ASSITALIA-LE ASSICURAZIONI D’ITALIA AUGUSTA ASSICURAZIONI AURORA ASSICURAZIONI AXA ASSICURAZIONI BAYERISCHE RUCK (*) BERNESE ASS.NI-COMP. ITALO-SVIZZERA BNC ASSICURAZIONI COMPAGNIA ASSICURATRICE UNIPOL COMPAGNIA DI ASSICURAZIONE DI MILANO IL DUOMO ERC - FRANKONA AG (*) F.A.T.A. LA FONDIARIA ASSICURAZIONI GAN ITALIA GENERAL & COLOGNE RE (*) GIULIANA ASSICURAZIONI ITALIANA ASSICURAZIONI ITAS ASSICURAZIONI ITAS SOC. DI MUTUA ASSICURAZIONE LEVANTE NORDITALIA ASSICURAZIONI LIGURIA LLOYD ADRIATICO LLOYD ITALICO ASSICURAZIONI MAECI - SOC. MUTUA DI ASS.NI E RIASS.NI MAECI ASSICURAZIONI E RIASSICURAZIONI LA MANNHEIM QUOTA % 1.286 0.286 5.263 0.429 5.263 0.717 1.071 2.460 2.857 0.429 0.286 2.231 5.263 0.574 5.714 1.429 5.263 0.791 2.714 0.286 0.857 0.529 0.529 1.029 0.429 1.340 0.429 0.286 0.429 0.429 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 TOTAL COMPANY MEDIOLANUM ASSICURAZIONI METE ASSICURAZIONI MUNCHENER RUCK ITALIA (*) LA NATIONALE NATIONALE SUISSE NAVALE ASSICURAZIONI NEW RE (*) NUOVA MAA ASSICURAZIONI NUOVA TIRRENA PADANA ASSICURAZIONI LA PIEMONTESE SOC. MUTUA DI ASS.NI LA PIEMONTESE ASSICURAZIONI RISPARMIO ASSICURAZIONI RIUNIONE ADRIATICA DI SICURTA’ ROYAL & SUN ALLIANCE SAI SARA ASSICURAZIONI SASA SCOR ITALIA RIASSICURAZIONI (*) S.E.A.R. SIAT-SOCIETA’ ITALIANA ASS.NI E RIASS.NI SOCIETA’ CATTOLICA DI ASSICURAZIONE SOCIETA’ REALE MUTUA DI ASSICURAZIONI SOREMA (*) SWISSE RE - ITALIA TICINO TORO ASSICURAZIONI UNIASS ASSICURAZIONI UNIVERSO ASSICURAZIONI VITTORIA ASSICURAZIONI WINTERTHUR ASSICURAZIONI QUOTA % 0.429 1.143 3.286 0.429 0.429 0.963 2.571 0.429 1.743 2.143 0.429 0.429 0.286 5.263 0.857 5.263 0.429 0.429 2.571 0.286 0.429 1.186 1.429 2.571 7.714 0.306 2.857 0.686 0.429 0.840 0.857 100.000 (*) Reinsurance company A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks 7 Problem The pool of companies that insures the risk look for being as competitive as possible → two intertwined questions: • Which premium should they charge? (Economic question) • How should they allocate the risk and the premium in order to obtain a fair division? (Game theoretic question) Applications of game theory to insurance: Borch (1962a, 1962b) Lemaire (1977, 1991) Suijs, Borm, De Waegenaere, Tijs (1999) Suijs (2000) - Survey A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks 8 N = {1, ..., n} set of companies L set of random variables (insurable risks) functional providing a measure of risk according to company i ∈ N Hi : L → R Hypothesis 1 a) a loading for a degenerate risk is not justified Hi (w) = w, ∀ w ∈ R, ∀ i ∈ N b) translation invariance Hi (w + X) = w + Hi (X), ∀ w ∈ R, ∀ X ∈ L, ∀ i ∈ N Many classical principles satisfy these hypotheses: • the net premium principle H(X) = E(X), where E(X) is the expectation of X • the variance principle H(X) = E(X) + aV (X), where V (X) is the variance of X and a > 0 • the standard deviation principle H(X) = E(X) + β V (X), where β > 0 • the zero utility principle H(X) = H̄, where H̄ satisfies E[u(z + H̄ − X)] = u(z) and u is the utility function of the insurance company which has an initial surplus z in particular for exponential utility u(x) = 1a (1 − e−ax), with a > 0 we have H̄ = a1 lnE(eaX ) • the ε-percentile principle H(X) = min {x|F (x) ≥ 1 − ε}, where F is the distribution function of X 9 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks The n companies have to insure a given risk R and receive a premium π They look for ”fair” allocations of (π, R) For any subset (coalition) of companies S ⊆ N let D(S) be the set of feasible divisions of the premium π: di = π D(S) = (di )i∈S ∈ R|S| s.t. i∈S and let A(S) be the non-empty set of the feasible decompositions of the risk R: Xi = R A(S) = (Xi )i∈S ∈ L|S| , s.t. i∈S According to the allocation (di , Xi)i∈S ∈ D(S) × A(S), the company i ∈ S receives the amount di and pays the random variable Xi Hypothesis 2 for each subset S ⊆ N it is possible to compute an optimal decomposition of the risk ∀ S ⊆ N there exists min Hi (Xi ) = P (S) (Xi )i∈S ∈A(S) i∈S P (S) can be seen as the evaluation that the companies in S (as a whole) give of the risk R 10 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Example 1 For each i ∈ N the variance principle holds: Hi(Y ) = E(Y ) + ai V ar(Y ) ∀ Y ∈ L, 0 < a1 ≤ ... ≤ an It is possible to prove (Deprez - Gerber, 1985) that P (N) = i∈N Hi (qi R); moreover as in Fragnelli-Marina (2003) we have: qi Hi R = E(R) + a(S)V ar(R) P (S) = q(S) i∈S a(N) 1 where qi = ai , q(S) = i∈S qi and a(S) = i∈S a1i ∀S⊂N ♦ A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Consider the allocations of (π, R) with a premium high enough to cover the risk For any subset of companies S ⊆ N let B(S) be the set of individually rational allocations: B(S) = {(di , Xi)i∈S ∈ D(S) × A(S)| di − Hi (Xi ) ≥ 0, ∀ i ∈ S} • B(S) = ∅ ⇐⇒ P (S) ≤ π • B(S) = ∅ ⇒ B(T ) = ∅, ∀ T ⊃ S. In fact P (T ) ≤ P (S) because Hi (0) = 0, ∀i ∈ N • We suppose that π > P (N), i.e. the companies all together may obtain a positive gain 11 12 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks For any subset of companies S ⊆ N s.t. B(S) = ∅ let O(S) be the set of allocations of (π, R) corresponding to optimal risk decompositions: O(S) = (di , Xi )i∈S ∈ B(S) | Hi (Xi ) = P (S) i∈S and let PO(S) be the set of Pareto optimal allocations of (π, R): PO(S) = {(di , Xi)i∈S ∈ B(S) | ∃ (di , Xi)i∈S ∈ B(S), s.t. di − Hi (Xi ) > di − Hi (Xi ), ∀ i ∈ S} Theorem 1 O(S) = PO(S) (see Proposition 3.5 in Suijs-Borm, 1999 and Proposition 1 in Fragnelli-Marina, 2003) • If B(S) = ∅ then if we take (Xi )i∈A(S) s.t. i∈S Hi (Xi ) = P (S) and define di = Hi (Xi ) + P (S)), ∀ i ∈ S then (di , Xi)i∈S ∈ PO(S) 1 |S| (π − Let Q(N) be the set of allocations s.t. there do not exist Pareto optimal allocations for S ⊂ N preferable to them and let CO(N) be the set of allocations s.t. any S ⊆ N cannot do better acting separately by itself: Q(N) = {(di , Xi )i∈N ∈ PO(N) | ∀ S ⊂ N s.t. B(S) = ∅, ∃ (di , Xi)i∈S ∈ PO(S) s.t. di − Hi (Xi ) > di − Hi (Xi ), ∀ i ∈ S} CO(N) = (di , Xi )i∈N ∈ PO(N) | ∀ S ⊆ N i∈S (di − Hi (Xi )) ≥ max {0, π − P (S)} Theorem 2 CO(N) = Q(N) (see Theorem 2 in Lari-Marina, 2000) 13 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Co-Insurance Games A cooperative game in characteristic function form with transferable utility (TU-game) is a pair (N, v) where: N set of players (companies) v : 2N → R characteristic function, with v(∅) = 0 v(S), S ⊆ N is the worth of coalition S Given a TU-game (N, v) the core is the set of allocations that give to each coalition S at least its worth: |N| xi ≥ v(S), ∀ S ⊆ N and xi = v(N)} Core(v) = {(xi )i∈N ∈ R s.t. i∈S i∈N When a game has non-empty core it is said to be balanced For our co-insurance problem we define a game whose characteristic function v is: v(S) = max {0, π − P (S)} ∀S⊆N The allocations of the co-insurance problem are related to the allocations of the co-insurance game: Theorem 3 CO(N) = ∅ ⇐⇒ Core(v) = ∅ Let us suppose that the players are ordered as: So P (N) ≤ P (N \ {n}) ≤ ... ≤ P (N \ {1}) j∈N P (N \ {j}) − (n − 1)P (N) ≥ P (N \ {i}), ∀ i ∈ N 14 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks General Results Lemma 1 π ≤ P (N \ {1}) ⇒ the game is balanced Lemma 2 π > i∈N P (N \ {i}) − (n − 1)P (N) ⇒ the game is not balanced Theorem 4 There exists π such that the game is balanced if and only if π ≤ π (see Theorem 1 in Fragnelli et al., 2000) Hypothesis 3 The cost function P satisfies the (reduced concavity) hypothesis: for each S ⊂ N s.t. P (S) < π , P (S) − P (S ∪ {i}) ≥ (P (N \ {i}) − P (N), ∀ i ∈ N \ S Theorem 5 Suppose that P satisfies Hypothesis 3, then π = j∈N P (N \ {j}) − (n − 1)P (N) If P satisfies Hypothesis 3 and the premium is precisely π = j∈N P (N \ {j}) − (n − 1)P (N) then: • the core is the singleton whose only element is the marginal solution x = (P (N \ {1}) − P (N), ..., P (N \ {n}) − P (N)) • the marginal solution is also the nucleolus of the game (Schmeidler, 1969) • the marginal solution is also the τ -value of the game (Tijs, 1981) 15 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Constant Quotas Let us suppose that there exist a convex function H and n real numbers q1 ≥ ... ≥ qn > 0, Y , Hi (Y ) = qiH qi (In Example 1 H(Y ) = E(Y ) + a(N)V ar(Y )) i∈N qi = 1 s.t.: ∀ i ∈ N, ∀ Y ∈ L An optimal decomposition can be obtained sharing the risk according constant quotas, represented by the numbers q1, ..., qn and the cost function P satisfies the reduced concavity hypothesis Proposition 1 The function P verifies Hypothesis 3 16 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Proportional Allocation The proportional (problem) allocation (qi π, qiR)i∈N has the following properties: • belongs to B(N) 1 − qn • belongs to CO(N) iff π ≤ qn H R q(N \ {n}) − H(R) = π 1 (P (N \ {n}) − P (N)) the corresponding game solution is If the premium is exactly π , then v(N) = q n qi (P (N \ {n}) − P (N) qn i∈N Player n gets exactly its marginal contribution, while each player i ∈ N \ {n} gets the marginal contribution of player n times the ratio among qi and qn If the premium is exactly π , then v(N) = j∈N (P (N \ {j}) − P (N)) the proportional solution is , while the marginal solution is (P (N \ {i}) − P (N))i∈N qi j∈N P (N \ {j}) − P (N) i∈N The proportional solution divides each marginal contribution proportionally among all the players (plus the ”refund” of the risk assumed) The marginal solution assigns to each player exactly his marginal contribution (plus the ”refund” of the risk assumed) A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks 17 Case Study We suppose that the 61 companies in Example 1 express their evaluation of a random variable X according to the variance principle Next we suppose that qi , i ∈ N are the quotas of the risk as in Table 2 and a(N) = 0.1 1 1 and V ar(R) = 2 µ µ and V ar(R) = 1.1025) (in millions of euros) and consequently: Finally we suppose that the distribution function of the risk R is F (x) = 1−e−µx, so E(R) = We assume that E(R) is 1.05 (µ = P (N) = 1.160250 π = 1.274612 π = 1.270816 1 1.05 18 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Comp,(i) 55 15 3 5 13 17 44 46 33 9 57 19 37 49 54 8 12 40 39 16 53 26 1 52 32 7 24 36 21 45 % 7.71 5.71 5.26 5.26 5.26 5.26 5.26 5.26 3.29 2.86 2.86 2.71 2.57 2.57 2.57 2.46 2.23 2.14 1.74 1.43 1.43 1.34 1.29 1.19 1.14 1.07 1.03 0.96 0.86 0.86 qi 0.07714 0.05714 0.05263 0.05263 0.05263 0.05263 0.05263 0.05263 0.03286 0.02857 0.02857 0.02714 0.02571 0.02571 0.02571 0.02460 0.02231 0.02143 0.01743 0.01429 0.01429 0.01340 0.01286 0.01186 0.01143 0.01071 0.01029 0.00963 0.00857 0.00857 qi π 98,031 72,614 66,883 66,883 66,883 66,883 66,883 66,883 41,759 36,307 36,307 34,490 32,673 32,673 32,673 31,262 28,352 27,234 22,150 18,160 18,160 17,029 16,343 15,072 14,525 13,610 13,077 12,238 10,891 10,891 qi π 98,324 72,831 67,083 67,083 67,083 67,083 67,083 67,083 41,884 36,416 36,416 34,593 32,770 32,770 32,770 31,355 28,437 27,315 22,216 18,214 18,214 17,080 16,392 15,117 14,569 13,651 13,116 12,275 10,923 10,923 marg 98,717 72,978 67,189 67,189 67,189 67,189 67,189 67,189 41,872 36,391 36,391 34,565 32,739 32,739 32,739 31,323 28,401 27,279 22,179 18,178 18,178 17,045 16,357 15,084 14,536 13,620 13,085 12,245 10,896 10,896 Comp,(i) 61 60 18 6 58 14 22 23 4 10 25 27 29 30 31 34 35 38 41 42 47 48 51 59 56 2 11 20 28 43 50 % 0.86 0.84 0.79 0.72 0.69 0.57 0.53 0.53 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.31 0.29 0.29 0.29 0.29 0.29 0,29 qi 0.00857 0.00840 0.00791 0.00717 0.00686 0.00574 0.00529 0.00529 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00306 0.00286 0.00286 0.00286 0.00286 0.00286 0,00286 qi π 10,891 10,675 10,052 9,112 8,718 7,294 6,723 6,723 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 3,889 3,635 3,635 3,635 3,635 3,635 3.635 qi π 10,923 10,707 10,082 9,139 8,744 7,316 6,743 6,743 5,468 5,468 5,468 5,468 5,468 5,468 5,468 5,468 5,468 5,468 5,468 5,468 5,468 5,468 5,468 5,468 3,900 3,645 3,645 3,645 3,645 3,645 3.645 marg 10,896 10,680 10,057 9,115 8,721 7,296 6,724 6,724 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 5,452 3,889 3,635 3,635 3,635 3,635 3,635 3.635 Some allocations, in euros, for the problem (marg = qiP (N ) + P (N \ {i}) − P (N )) 19 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks Comp,(i) 55 15 3 5 13 17 44 46 33 9 57 19 37 49 54 8 12 40 39 16 53 26 1 52 32 7 24 36 21 45 % 7.71 5.71 5.26 5.26 5.26 5.26 5.26 5.26 3.29 2.86 2.86 2.71 2.57 2.57 2.57 2.46 2.23 2.14 1.74 1.43 1.43 1.34 1.29 1.19 1.14 1.07 1.03 0.96 0.86 0.86 qi 0.07714 0.05714 0.05263 0.05263 0.05263 0.05263 0.05263 0.05263 0.03286 0.02857 0.02857 0.02714 0.02571 0.02571 0.02571 0.02460 0.02231 0.02143 0.01743 0.01429 0.01429 0.01340 0.01286 0.01186 0.01143 0.01071 0.01029 0.00963 0.00857 0.00857 − P (N )) qi (π 8,529 6,318 5,819 5,819 5,819 5,819 5,819 5,819 3,633 3,159 3,159 3,001 2,843 2,843 2,843 2,720 2,467 2,369 1,927 1,580 1,580 1,482 1,422 1,311 1,264 1,184 1,138 1,065 948 948 − P (N )) qi (π 8,822 6,535 6,019 6,019 6,019 6,019 6,019 6,019 3,758 3,267 3,267 3,104 2,940 2,940 2,940 2,813 2,551 2,451 1,993 1,634 1,634 1,532 1,471 1,356 1,307 1,225 1,177 1,101 980 980 marg 9,216 6,681 6,125 6,125 6,125 6,125 6,125 6,125 3,746 3,242 3,242 3,076 2,909 2,909 2,909 2,781 2,516 2,414 1,956 1,598 1,598 1,497 1,436 1,323 1,275 1,194 1,146 1,072 953 953 Comp,(i) 61 60 18 6 58 14 22 23 4 10 25 27 29 30 31 34 35 38 41 42 47 48 51 59 56 2 11 20 28 43 50 % 0.86 0.84 0.79 0.72 0.69 0.57 0.53 0.53 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.43 0.31 0.29 0.29 0.29 0.29 0.29 0.29 qi 0.00857 0.00840 0.00791 0.00717 0.00686 0.00574 0.00529 0.00529 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00429 0.00306 0.00286 0.00286 0.00286 0.00286 0.00286 0.00286 − P (N )) qi (π 948 929 875 793 758 635 585 585 474 474 474 474 474 474 474 474 474 474 474 474 474 474 474 474 338 316 316 316 316 316 316 − P (N )) qi (π 980 961 905 820 785 656 605 605 491 491 491 491 491 491 491 491 491 491 491 491 491 491 491 491 350 327 327 327 327 327 327 Some allocations, in euros, for the game (marg = P (N \ {i}) − P (N )) marg 953 934 879 796 762 636 586 586 475 475 475 475 475 475 475 475 475 475 475 475 475 475 475 475 338 316 316 316 316 316 316 A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks 20 Concluding Remarks • if π = j∈N P (N \ {j}) − (n − 1)P (N ) the proportional allocation is not in the core of the game because it assigns too much to last players, w.r.t. what is assigned to the first players • the unique core allocation, the marginal solution, is favorable to the first players, i.e. those who assume larger quotas of risk • the proportional solution is more ”leveled” so that in order to have a core allocation it is necessary to have a lower premium that produces an amount assigned to the last player by the proportional allocation at most equal to the amount of the marginal one A Game Theoretic Approach to Insurance Sharing in Environmental Pollution Risks 21 References Bazzano C (1994) I rischi dell’azienda, Pirola Editore, Milano, Italy. Borch K (1962a) Equilibrium in a Reinsurance Market, Econometrica 30, 424:444. Borch K (1962b) Application of Game Theory to Some Problems in Automobile Insurance, Astin Bulletin 2, 208:221. Deprez O, Gerber HU (1985) On Convex Principle of Premium Calculation, Insurance: Mathematics and Economics 4 : 179-189. Fragnelli V, Marina ME (2003) A Fair Procedure in Insurance, Insurance: Mathematics and Economics 33 : 75-85. Fragnelli V, Garcı́a-Jurado I, Méndez-Naya L (2000) On Shortest Path Games, Math Meth Op Res 52 : 139-216. Gerber HU (1980) An Introduction to Mathematical Risk Theory, The S.S. Huebner Foundation, Philadelphia, PA, USA. Goovaerts MJ, De Vylder F, Haezendonck J (1984) Insurance Premiums, North-Holland, Amsterdam, The Netherlands. Lari E, Marina ME (2000) Coassicurazione e principi di calcolo del premio, Preprint, Dipartimento di Economia e Metodi Quantitativi, Universitá di Genova. Lemaire J (1977) Echange de risques entre assureurs et théorie des jeux, Astin Bulletin 9 : 155-179. Lemaire J (1991) Cooperative Game Theory and Its Insurance Applications, Astin Bulletin 21, 17:40. Schmeidler D (1969) The Nucleolus of a Characteristic Function Game, SIAM Journal of Applied Mathematics 17 : 1163-1170. Suijs J (2000) Cooperative Decision Making under Risk, Kluwer Academic Publisher, Dordrecht, The Netherlands. Suijs J, Borm PEM (1999) Stochastic Cooperative Games: Superadditivity, Convexity, and Certainty Equivalents, Games and Economic Behavior 27 : 331-345. Suijs J, Borm PEM, De Waegenaere A, Tijs S (1999) Cooperative Games with Stochastic Payoffs, European Journal of Operational Research 113, 193-205. Tijs SH (1981) Bounds for the Core and the τ -Value, in Game Theory and Mathematical Economies (Moeschlin O, Pallaschke D eds.), Amsterdam, North Holland : 123-132.