Linear Prices Equilibria in Nonexclusive Insurance Market

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Linear Prices Equilibria in Nonexclusive Insurance Market

Frdric Loss e e Gwenal Piaser e

Abstract Key words: Common Agency, Insurance, Moral Hazard, Competition, Linear Prices Equilibria, Nonexclusive Contracts

1 2

Introduction The Model

We consider a competitive insurance market, derived from Mossins (1968) classical model of insurance. There is a continuum of identical agents of measure 1 and a nite but large number of identical insurance companies, indexed by i with i I = {1; . . . ; N }. Each agent faces a xed-damage accident denoted L. The accident is distributed by a Bernoulli distribution: there are two states of the world, either no accident or accident. Without insurance and if there is no accident, w corresponds to the consumption of an agent, while it is equal to w L if there is an accident. However, an agent can also subscribe insurance contracts. Insurance contracts are non exclusive (i.e., each agent can buy several contracts from dierent insurance companies) and private information to the agent. The consumption of an insured agent when there is no accident is w P , where P = N pi corresponds to i=1 the sum of the premia pi paid by the insured agent to the insurance companies. Moreover, if there is an accident the consumption of an agent is w L + R P , where R = N ri corresponds to the sum i=1
We would like to thank Andrea Attar and Anne Marie Tauzin. All errors are ours. CNAM - Laboratoire dconomtrie. e e IPAG, Business School, Paris.

of each of the repayments ri paid by the insurance companies. The probability of accident (e) is a function of the eort e devoted by an agent to avoid the accident. (e) is decreasing at an increasing rate in e, that is, (e) < 0 and (e) > 0. We assume that e [e; e], where e, e respectively denote the lowest and the highest eort that an agent can exert. Eort is unobservable by any insurance company. Finally, we assume that the utility function of an agent is separable in consumption and eort, and is event-independant.1 The expected utility of an agent is thus equal to: U (w, L, R, P, e) = (e)u (w L + R P ) + (1 (e)) u (w P ) e, (1) where u(.) is a von Neumann-Morgenstern utility function, with u > 0 and u < 0. Without loss of generality we assume that u(0) = 0 and that w > L. The insurance market is competitive. We thus assume that the number of insurance companies N is large and each insurance company has no market power. The insurance companies can oer any type of contract (ri , pi ). However, it is impossible to write a contract contingent on an agents eort or on his insurance contracts with other insurance companies since eort and insurance contracts are both private information to the agent.2 The expected prot of an insurance company associated with an insurance contract (ri , pi ) is thus equal to: i (ri , pi , e) = pi (e)ri . Following the literature on common agency we allow insurance companies to oer menus of contracts. Each insurance company oer a subset of all possible contracts i.e; oer a subset of R2 in which the agent choose at most one contract. To make the problem tractable, we do assume that menu are closed subsets of R2. Hence the set of strategy of each insurance company i is denoted by Mi , which the set of all closed subset of R2. Given the oered menus, the agent must choose a portfolio of contracts and an eort. We do oblige the agent to choose a contract in every menu, he can choose only one contract from one insurance company or no contract at all. Hence, if a insurance company proposes the menu Mi , the real choice set of the agent with respect to this company is Mi (0, 0).
1 This means that the occuring event does not aect the utility derived from consumption (i.e., the accident does not alter tastes). 2 However, each insurance company does know all the contracts that it has signed with an agent.

Thus, for the agent, a strategy is a function that maps any set of menu (M1 , , Mi , , MN ) iI Mi to an element of the set iI Mi (0, 0)E.

The Agents Optimal Level of Eort

For any given insurance contract (R, P ) an agent exerts the eort which maximizes its expected utility: e = arg max U (w, L, R, P, e) .
e

Computing the rst order condition with respect to e, then when the agents maximization program admits an interior solution, e satises the following equality: (e ) [u (w L + R P ) u (w P )] = 1. (2)

However, there are cases for which the agents maximizing program admits a corner solution: e = e (the agent optimally exerts the lowest level of eort). For instance, consider that there is full insurance (R = L) then an agents consumption is the same whether there is accident or not. The agent has thus no incentive to exert an eort higher than e since, for a given probability of accident, his expected utility is decreasing with the eort that he exerts (see Equation (1)). Moreover, for any given P an agent also exerts e = e when R > L. Consider now that R < L, but R L. Then, by continuity, since for R = L, Equality (2) is not veried, it is also not veried for R < L, but R L. More generally, for any given P suppose that there exists a repayment level in case of accident, denoted R, such that Equality (2) is satised for e = e and R = R. Then, an agent optimally chooses to exert an eort e > e when R < R, while he optimally exerts the lowest eort for R R. Consider now the particular case where an agent is not insured, R = 0. Then, the rst order condition indicates that an agent optimally chooses to exert an eort e > e if and only if: u (w) u (w L) This inequality is satised when: u (w) > 1 and L is close to w, (e) (Conditions 1) 1 . (e)

since u(0) = 0. Let us now dene R(P ) as the implicit solution of Equality (Conditions 1) (if it exists). For any given P , R is thus such that (e) [u (w L + R P ) u (w P )] = 1. (3) 3

Applying the implicit function theorem to this last equality, we obtain that: u (w L + R P ) u (w P ) dR = > 0, dP u (w L + R P ) since u () > 0, but u () < 0. This implies that the minimal repayment R(P ) above which an agent chooses to exert the lowest level of eort e increases with P . Moreover, Conditions 1 imply that for any positive premium P 0, R(P ) ]0; L[. Let us now study the indirect utility function of an agent. For a given insurance contract (R, P ) the indirect utility function of an agent, denoted V (R, P ), for e = e can be written as: V (R, P ) = (e )u (w L + R P ) + (1 (e )) u (w P ) e . Any indierence curve is thus characterized by the relation: (e )u (w L + R P ) + (1 (e )) u (w P ) e = A, (4)

where A R. Using Equality (4), we obtain that the agents marginal rate of substitution (denoted M RSagent hereafter) between R and P is (e )u (w L + R P ) dP = . dR (e )u (w L + R P ) + [1 (e )] u (w P ) (5)

The marginal rate of substitution, M RSagent , is positive since the utility of the agent is increasing in R, but decreasing in P . Moreover, M RSagent (R = L) = (e ) = (e) since the agent exerts the lowest effort when fully insured. Besides, any indierence curve is well shaped, that is, increasing at a decreasing rate, when R R. Indeed, when R R, e = e which implies that M RSagent along the indierence curve only depends on the properties of the utility function u(.). However, for any R [0; R[ an indierence curve is not necessarily well shaped since M RSagent also depends on e which is aected by P and R. Consider any point (R, P ) with R < R. The marginal rate of substitution at that point M RSagent (R, P ), for a given repayment R is decreasing in P if and only if M RSagent (R,P ) < 0 or equivalently if M RSagent (R,P ) > 0, P w R=R and P =P since P and w play the same role in the agents utility function. Using relations (2) and (5), we obtain that u wP u wL+RP M RSagent (R, P ) (e ) e >0 + . w (e ) [1 (e )] w u wP u wL+RP (Condition 2) 4

Note that Condition 2 can be satised even with DARA utility functions u (wP ) u (wL+RP ) for which u wP u wL+RP , when e < 0 and (e (e ) )] e w )[1(e w ( ) ( ) is high enough. Consider now any point (R, P ) with R L. In that case, e = e and Condition 2 reduces to
u u wP u wL+RP wL+RP

wP u

which is satised when the agents VNM utility function is DARA. In the rest of the paper we are going to assume that any agent is characterized by a von Neumann-Morgenstern utility function which satises Condition 2. ON A UN PROBLEME : EN FAIT, e = e POUR R R(P ), SAUF QUE POUR R(P ) R < L AVEC UN FONCTION DUTILITE
u wP u wL+RP

DARA, ON AURA
u wP

u wL+RP

!!!!!

Let us now dene for a given eort e, P (e) as the half-straight line P = (e)R in the mark (R, P ), starting from (0; 0). Consider any eort e [e, e] and an indierence curve tangent to P (e) at the point (R, P ) with R < R. Let us constraint an agent to choose an eort e e. Given this constraint, the probability of accident is (e) (e). Then, the agents optimal choice of insurance would be characterized by R L and not R. Hence, we can state that for an insurance contract (R, P ) an agent exerts an eort e (R, P ) > e. This leads us to the following lemma. Lemma 1 Consider e [e, e] and an indierence curve tangent to P (e) at the point (R, P ) with R < R. Then, for an insurance contract (R, P ) an agent exerts an eort: e (R, P ) > e. Let us now look at the derivability of the indierence curves. Consider rst that R > R then any indierence curve exhibts the standard properties, since the eort is xed and always equals to e. This implies that for R R any indierence curve is always derivable in R. Consider now that R R and look at the rst order condition (cf. Equality 2). (e) > 0 implies that (e) is strictly increasing. Thus, for a given {w; L; R; P } Equality 2 admits a unique solution. Suppose that an indierence curve, denoted IC(A), is not derivable at point A: (RA ; PA ) (i.e., there is a kink at (RA ; PA )). Denote e t the agents optimal efLef fort at the left, but close to point A. Moreover, Denote e Right the agents optimal eort at the right, but close to point A. Then, by continuity both e t and e Lef Right should correspond to the agents optimal eort, and 5

should both satisfy Equality 2. We thus obtain a contradiction. Note that when RA = R (PA ) then e Right = e. This leads us to the following lemma. Lemma 2 For R 0, any indierence curve is always derivable in R. We are now going to look at the equilibria in the insurance market with linear primia and non exclusivity.

The Insurance Market Linear Equilibria with Non Exclusivity

Consider the half-straight line P (e) and denote by IC the indierence curve which is tangent to P (e) when there is full insurance, that is, for R = L. There are two potential cases. The rst one is when IC has no other intersection point with P (e) for any R < L. This case is illustrated in Figure 1 below in which Conditions 1 are satised and R(P ) is assumed to be convex in R. The second case is when IC has at least one other intersection point with P (e) for R < L. This case is illustrated in gure 2 below in which Conditions 1 are satised and R(P ) is assumed to be convex in R. The denition of an equilibrium with non exclusivity that we are going to use throughout the paper is the following: Given the set of contracts issued by the insurance companies each agent (privately) chooses which contracts to buy. This determines his consumption in the two states of the world: no accident or accident. Each agent also (privately) chooses the eort he exerts. Anticipating the choices of each agent, as a function of the set of contracts they trade, insurance companies strategically choose which contracts they issue so as to maximize their prots. Given the preceding denition, in order to characterize a equilibrium one should consider every principals deviation. Since the set of strategy of a principal is a set of subsets of R2 it is impossible to do such a thing. In the following lemma we show that considering take-it or leave-it oers is enough at the deviation stage. A take-it or leave-it oer is a point is R2 and the simplest menu that one can consider, and then the restriction to these simple deviations allows us to simplify drastically the analysis. Lemma 3 If a Principal has no protable deviation toward a take-it or leave it oer, then no deviation toward any menu is protable. Proof. Suppose not. Principal i has not protable deviation when he deviates toward a single contract, but there a menu which gives him a higher payo. 6

If the agent do not buy any contract in Mi , then for principal i the menu Mi and the contract (0, 0) are equivalent. If the agent buy the contract (Pi , Ri ) in the menu Mi . Then For principal i oering the menu Mi and the menu {(Pi , Ri )} is equivalent: It is optimal for the agent in the two cases to have the same behavior. Let us now start with the rst case. Consider a situation in which each insurance company issues a continuum of linear contracts (ri , pi ) with pi = (e)ri . Then, given this type of insurance contracts maximizing the expected utility of an agent amounts to nding the highest indierence curve tangent to P (e). This is precisely IC since this indierence curve has no other intersection point with P (e) for any ri = L. Therefore, the agent optimally chooses to be fully insured and exerts an eort e = e. Given the agents optimal eort and insurance choice (e = e, ri = L) the expected prot of an insurance company is just equal to 0: i (ri = L, pi = (e)L, e = e) = (e)L (e)L = 0. This is an equilibrium if and only if there is no protable deviation for any insurance company. Consider that an insurance company dep viates and issues a unique contract Cd1 = (rd1 , pd1 ) with rd1 < (e), d1 and rd1 < L. Then, an agent can buy this contract, and, since the insurance contracts are not exclusive, can also complete his insurance coverage with other contracts bought from the non-deviating insurance companies. Thus, each agent faces a new continuum of linear contracts which belong to the half-straight line P (e) starting from Cd1 and parallel to P (e). There exists an indierence curve which is tangent to P (e) for a repayment ri = L, since for any given pi , the M RSagent of any indierence curve is equal to (e) when ri = L. We denote this indierence curve ICCd1 . Besides, Lemma (2) together with Condition 2 imply that if IC has no other intersection point with L(e) for ri < L, then ICCd1 also has no intersection point with L (e) for ri < L. Thus, if an insurance company deviates and issues contract Cd1 the agent buys a global insurance contract C1 for which he is fully insured. He thus exerts the lowest eort e = e which implies that the expected prot of the deviating insurance company is negative. Any deviation by an insurance company is thus not protable. This leads us to the following proposition. Proposition 4 Under Condition 2 when IC has no other intersection point with P (e) for any R < L, then there exists a linear equilibrium which is such that: - Each insurance company proposes a continuum of linear contract (ri , pi ) with pi = (e)ri , 7

- Each agent chooses to be fully insured ri = L, and exerts the lowest eort e = e, - Each insurance company makes an expected prot equal to 0. Proof. See Appendix. Proposition 4 is illustrated in Figure 3 below. Moreover, the global insurance contract C1 that the agent would buy if an insurance company deviated and issued contract Cd1 is also represented in Figure 3. Let us now study the second case. When IC has at least one other intersection point with P (e) for R < L, then the linear equilibrium studied in the previous case is no longer an equilibrium. Indeed, the agents optimal eort and insurance choice associated with a continuum of linear contracts (ri , pi ) with pi = (e)ri is no longer (e = e, ri = L). For instance, as illustrated in Figure 4 below the agent strictly prefers to buy insurance contract C which belongs to an indierence curve superior to IC. Consider now that there exists one indierence curve which is tane gent to P (e) for R < L. (ICI, ON CONSID`RE R < L, MAIS PEUT ETRE FAUDRAIT-IL MIEUX DISTINGUER ET EXPLIQUER SEPAREMENT LES 2 SOUS CAS : R < R ET R < R > L) Then, Lemma 1 implies that the agent chooses to exert an eort e > e, which is by denition impossible. This implies that the indierence curve which is tangent to P (e) is necessarily tangent to P (e) for R > L. Finally, since the indierence curve IC which is tangent to P (e) for R = L has at least one other intersection point with P (e) for R < L and since P (e) only admits an indierence curve tangent to it for R > L, by continuity Lemma 2 implies that there must exist at least one eort e [e; e[ for which the half-straight line P (e) admits one indierence curve which is tangent to it both for R L and for R > L. We thus obtain the following Lemma. Lemma 5 When IC has at least one other intersection point with P (e) for R < L, there exists at least one eort e ]e; e[ such that the halfstraight line P (e) admits one indierence curve, denoted ICe , which is tangent to P (e) both at the right and at the left of R = L. Moreover, P (e) is bellow P (e) (and never intersects P (e)) since e e. Proof. See Appendix. Consider that each insurance company issues a continuum of linear contracts (ri , pi ) with pi = (epi )ri where epi is an eort such that P (epi ) admits an indierence curve ICepi which is tangent to P (epi ) both at the right and at the left of ri = L. Then, maximizing the expected utility of 8

the agent amounts to choose one of the two levels of insurance for which ICepi is tangent to P (epi ). Since the agent is indierent between these two possibilities, we can consider that he chooses to be only partially insured. Then, Lemma 1 implies that he exerts an eort e > epi (> e). Thus, the expected prot of each insurance company issuing this type of contract is strictly positive since the agent exerts an eort strictly higher than the eort on which the insurance premium pi is based. This situation is an equilibrium if and only if any deviation by an insurance company is not protable. Consider that an insurance company devip ates and issues a single contract Cd2 = (rd2 , pd2 ) with rd2 < (epi ) and d2 rd2 < L. Then, each agent faces a new continuum of linear contracts which belong to the half-straight line P (epi ) starting from Cd2 and parallel to P (epi ). Lemma 2 together with Condition 2 imply that there is no indierent curve which is tangent to P (epi ) both for ri < L and for ri > L. Moreover, the indierence curve which is tangent to P (epi ) is necessarily tangent to P (epi ) for ri > L. Therefore, if an insurance company deviates and issues a contract Cd2 then an agent buys this contract and completes his insurance coverage until he reaches a global insurance contract C2 for which ri > L. Each agent thus exerts the lowest eort e = e which implies that the expected prot of the insurance which deviates is negative. Any deviation by an insurance company is thus not protable. This leads us to the following proposition. Proposition 6 Under Condition 2 when IC has at least one other intersection point with P (e) for ri < L, there exists at least one linear equilibrium which is such that: - Each insurance company proposes a continuum of linear contract (ri , pi ) with pi = (epi )ri , - Each agent chooses to be only partially insured ri < L, and exerts an eort e > epi , - Each insurance company makes a strictly positive expected prot. Proof. See Appendix Proposition 6 is illustrated in Figure 5 below. Moreover, the global insurance contract C2 that the agent would buy if an insurance company deviated and issued contract Cd2 is also represented in Figure 5. We are now going to study the second-best eciency of the linear equilibria.

Second Best Eciency of the Linear Equilibrium


pi (e)ri = B, 9 (6)

Any insurance companys isoprot curve is such that

where B R+ . Dierentiating (6) with respect to ri and to pi indicates that (e) + (e) de i ri dpi dr = , (7) dri 1 (e) de i ri dp
i where dpi indicates how the insurance premium must increase when the dr repayment increases marginally so that the insurance companys prot i stays constant. dpi corresponds to the marginal rate of substitution dr between ri and pi for an insurance company (denoted M RSInComp here after). Note that M RSInComp 0, since (e) < 0, de i 0 and de i > 0. dr dp An equilibrium (ri , p ) is going to be second-best Pareto ecient if i it is impossible to increase both the expected utility of an agent and the expected prot of an insurance company, or if it is impossible to increase the expected utility of an agent (the expected prot of an insurance company) without decreasing the expected prot of an insurance company (the expected utility of an agent). This is the case if and only if for (ri , p ) we have: i M RSInComp = M RSagent .

Using equations (5) and (7) we obtain that M RSInComp = M RSagent only when (e) + (e) de i ri dr 1 (e) dpi ri
de

(e)u (w L + ri pi ) . (e)u (w L + ri pi ) + [1 (e)] u (w pi ) (8)

Consider rst that IC has no intersection point with L(e) at the left of ri = L. In that case, we know that there exists a linear equilibrium in which each agent is fully insured ri = L and exerts the lowest eort e = e. This implies that de i = 0 = de i . Thus, the LHS of Equality dr dp 8 reduces to (e). Moreover, for ri = L and pi = (e)L the RHS of Equality 8 also reduces to (e). Equality 8 is thus satised in this case.The linear equilibrium is thus second best Pareto ecient. This leads us to the following proposition. Proposition 7 When IC has no intersection point with P (e) at the left of ri = L, then the linear equilibrium is second best Pareto ecient. Consider now that IC has at least one other intersection point with P (e) for ri < L. In that case, we know that there exists a linear equilibrium in which each agent is not fully insured ri < L, pays a premium 10

pi = (ePi )ri and exerts an eort e > epi . Let us make the hypothesis that this equilibrium is second best Pareto ecient. This implies that equality 8 is satised in equilibrium. Consider now that the expected prot function of an insurance company is modied and becomes i (, ri , pi , e) = pi (e)ri , with, for instance, > 1 (e.g. corresponds to a functioning cost of the insurance company). Modifying the insurance prot function does not modify the linear equilibrium since it is determined by the agents preferences which have not changed. However, now (e ) + (e ) de i ri dr . M RSInComp = 1 (e ) de i ri dp Therefore, if the equality M RSInComp = M RSagent was veried with i (ri , pi , e) = pi (e)ri , this equality is necessarily not veried with i (, ri , pi , e) = pi (e)ri for any = 1. This implies that the linear equilibrium (ri < L; pi = (epi )ri ) is not generically second best Pareto ecient. We thus have the following proposition. Proposition 8 When IC has at least one other intersection point with L(e) for ri < L, then the linear equilibrium is not generically second best Pareto ecient. Proposition 8 is illustrated in Figure 7 below in which the linear equilibrium (ri < L; pi = (epi )ri ) is not second best Pareto ecient. Any point in area B is preferred both by the agent and by the insurance company to the equilibrium. We have thus just shown that when insurance contracts are not exclusive the linear equilibrium in which insurance companies have a strictly positive expected equilibrium prot (even if the insurance market is competitive) is not generically second best Pareto optimal. This result has two important consequences. First of all, a laissez-faire policy is not the best choice for a government when insurance companies have a strictly positive expected prot in equilibrium. A government should intervene in the insurance market so as to increase the welfare of the economy. Moreover, our results also suggest that it might be better for the economy in terms of welfare to oblige people to sign only exclusive contracts with the insurance companies. Note that even if exclusive contracts might be preferable for the economy, insurance companies might prefer non-exclusive contracts since they can then obtain a strictly positive expected equilibrium prot which should not be the case in a perfectly competitive insurance market with exclusive contracts.

11

6 7 7.1

Conclusion Appendix Proof of Proposition 4

Proof. First consider that each insurance company oers a menu of linear contracts (ri , pi ) with pi = (e)ri . Since IC has no intersection point with P (e) for any ri = L, each agent chooses to be fully insured and exerts an eort e = e. Thus, each insurance company makes an expected prot equal to 0. Consider now that an insurance company deviates and oers a sinpd gle contract characterized by the couple (rd1 , pd1 ) with rd1 < (e) and 1 rd1 < L. We can dene the half-straight line P (e) starting from (rd1 , pd1 ) and parallels to P (e) and C1 the intersection point of P (e) with the vertical line passing through point (0, L). C1 is of course preferred to contract (ri = L, pi = (e)L), since the indierence curves are increasing in the repayment ri and decreasing in the premium pi . Let now IC T be the translation of IC in the vertical direction tangent to P (e) for ri = L. IC T has by denition no intersection point with L (e) for ri = L. Besides, Lemma 2 together with Condition 2 imply that the indierence curve which is tangent to L (e) at C1 is necessarily below IC T and has thus no other intersection with P (e) for ri = L. Therefore, C1 is preferred by the agent to any other point on the half-straight line P (e) including (rd1 , pd1 ). Hence, if an insurance company proposes contract (rd1 , pd1 ), the agent would buy it and complete this contract so as to reach contract C1 . The agent would then exert e = e, since he is fully pd insured with contract C1 . However, since rd1 < (e) the expected prot 1 of the insurance company which proposes contract (rd1 , pd1 ) is strictly negative and the deviation is not protable.

7.2

Proof of Lemma 5

Proof. For any e e the indierence curve tangent to P (e) for ri L exists and is unique, since for ri L the agent always exerts an eort e = e and each indierence curve is then well shaped (i.e. increasing and at a decreasing rate). We denote this indierence curve ICe . Consider now that e = e. Then, there is no indierence curve tangent to the half-straight line P (e) for ri L and which is tangent with P (e) for ri < L. Suppose the contrary, and suppose that there exists one indierence curve tangent to P (e) at point (ri , pi ) with ri < L. Lemma 1 implies that e (ri , pi ) > e which is impossible by denition. Moreover, there is also no indierence curve tangent to the half-straight line P (e) for ri L and which has more than one intersection point with P (e) 12

for ri < L. Suppose the contrary and consider Point A on the following picture. At point A, assume that e e. At the intersection point A, we necA essarily have: (e) > M RS(P oint A). Moreover, x the eort exerted by the agent to e . Then any indierence curve exhibits the standard A properties of an indierence curve associated with a WNM utility fundtion. In particular, this implies that M RS(P oint A) > (e ). ThereA fore, we have: (e) > M RS(P oint A) > (e ). This implies that A (e) > (e ) and thus e < e . A contradiction. A A Therefore, since for e = e, there is one indierence curve which is tangent to P (e) at point ri = L and which has at least one other intersection point with P (e) for ri < L, by continuity, Lemma 2 implies that there must exist at least one eort e such that the half-straight line P (e) admits an indierence curve ICe which is also tangent to P (e) for ri < L.

7.3

Proof of Proposition 6

Proof. Consider rst that each insurance company oers a menu of linear contracts (ri , pi ) with pi = (ePi )ri , where ePi is such that there exists an indierence curve, denoted ICePi which has two tangent points with P (ePi ). Since the agent is indierent between one of the two tangent points we can consider that he chooses the tangent point denoted (ri , p ) i with ri < L. Then Lemma 1 implies that he exerts an eort e (ri , p ) > i ePi . Thus, (ePi )ri > (e (ri , p ))ri and each insurance company makes i a strictly positive expected prot. Consider now that an insurance company deviates and oers conpd tract (rd2 , pd2 ), with rd2 < (ePi ). Let P (ePi ) be the half-straight line
T starting from (rd2 , pd2 ) and parallels to P (ePi ). Let ICeP be the vertical T translation of ICeP in the vertical direction. By denition ICeP could be tangent to P (eP i ) for ri < L and is necessarily tangent to L (ePi ) for ri > L. Besides, Lemma 2 and Condition 2 imply that for any given T repayment level ri L any indierence curve is necessarily below ICeP . Moreover, since for ri L the indierence curves are increasing at a decreasing rate, there exists an indierence curve which is tangent to P (ePi ) at a point C2 = (r , p ) with r > L. C2 is of course preferred by an agent to any other point on the half-straight line P (ePi ) including (rd2 , pd2 ). Hence, if an insurance company oers contract (rd2 , pd2 ) the optimal choice for an agent is to buy this contract and to complete it in order to get C2 . The agent thus makes an eort e = e. The expected prot of the insurance company is strictly negative since (e) < (ePi ), and the deviation is thus not protable.
2

13

References
[1] Ales, L., Maziero, 2009. Adverse selection and non-exclusive contracts, Working Paper Tepper School of Business, Carnegie Mellon University. [2] Arnott, R., Stiglitz, J., E., 1991. Price equilibrium, eciency, and decentralizability in insurance markets, Working Paper NBER, N .3642. [3] Attar, A., Campioni, E., Chassagnon, A., Rajan, 2007. Incentives and competition under moral hazard, Mimeo IDEI, Toulouse. [4] Attar, A., Campioni, E., Piaser, G., 2006. Multiple lending and constrained eciency in the credit market, Contributions to Theoretical Economics, 6(1), 1253. [5] Attar, A., Chassagnon, A., 2009. On moral hazard and nonexclusive contracts, Journal of Mathematical Economics, 45(9-10), 511525. [6] Bisin, A., Guaitoli, D., 2004. Moral hazard and nonexclusive contracts, RAND Journal of Economics, 35, 306-328. [7] Hellwig, M., 1983. On the moral hazard and non-price equilibria in competitive insurance markets, Discussion Paper N 109, Institut fur Gesellschafts. [8] Helpman, E., and Laont, J.-J., 1975. On moral hazard in general equilibrium, Journal of Economic Theory, 10, 8-83. [9] Kahn, C., Mookherjee, D., 1998. Competition and incentives with nonexclusive contracts, RAND Journal of Economics, 29(3), 443 465. [10] Martimort, D., Stole, L., 2002. The revelation and delegation principles in common agency games, Econometrica, 16591673. [11] Mossin, J., 1968. Aspects of rational insurance purchasing, Journal of Political Economy, 76(4), 553568. [12] Parlour, C., Rajan, R., 2001. Competition in loan contracts, American Economic Review, 91(5), 13111328. [13] Pauly, M., 1974. Overproduction and public provision of insurance, Quarterly Journal of Econcomics, 88, 44-62. [14] Peters, M., 2001. Common agency and the revelation principle, Econometrica, 69(5), 13491372. [15] Rotschild, M., Stiglitz, J., 1976. Equilibrium in competitive insurance markets: An essay on the economics of imperfect information, Quarterly Journal of Economics, 90(4), 629649.

14

You might also like