Bank monitoring incentives under moral hazard and adverse selection
BBank monitoring incentives under moral hazard and adverse selection ∗ Nicolás
Hernández Santibáñez † Dylan
Possamaï ‡ Chao
Zhou § January 17, 2019
Abstract
In this paper, we extend the optimal securitisation model of Pagès [50] and Possamaï and Pagès [51] between aninvestor and a bank to a setting allowing both moral hazard and adverse selection. Following the recent approachto these problems of Cvitanić, Wan and Yang [14], we characterise explicitly and rigorously the so-called credibleset of the continuation and temptation values of the bank, and obtain the value function of the investor as well asthe optimal contracts through a recursive system of first-order variational inequalities with gradient constraints. Weprovide a detailed discussion of the properties of the optimal menu of contracts.
Key words: bank monitoring, securitisation, moral hazard, adverse selection, principal–agent problem
AMS 2000 subject classification : 60H30, 91G40
JEL classifications:
G21, G28, G32
Principal–Agent problems with moral hazard have an extremely rich history, dating back to the early static models ofthe 70s, see among many others Zeckhauser [68], Spence and Zeckhauser [63], or Mirrlees [42, 43, 44, 45], as well as theseminal papers by Grossman and Hart [26], Jewitt, [33], Holmström [30] or Rogerson [57]. If moral hazard results fromthe inability of the Principal to monitor, or to contract upon, the actions of the Agent, there is a second fundamentalfeature of the Principal-Agent relationship which has been very frequently studied in the literature, namely that ofadverse selection, corresponding to the inability to observe private information of the Agent, which is often referredto as his type. In this case, the Principal offers to the Agent a menu of contracts, each having been designed for aspecific type. The so-called revelation principle , states then that it is always optimal for the Principal to propose menusfor which it is optimal for the Agent to truthfully reveal his type. Pioneering research in the latter direction weredue to Mirrlees [46], Mussa and Rosen [47], Roberts [55], Spence [62], Baron and Myerson [7], Maskin and Riley [38],Guesnerie and Laffont [27], and later by Salanié [59], Wilson [67], or Rochet and Choné [56]. However, despite theearly realisation of the importance of considering models involving both these features at the same time, the literatureon Principal-Agent problems involving both moral hazard and adverse selection has remained, in comparison, ratherscarce. As far as we know, they were considered for the first time by Antle [2], in the context of auditor contracts, andthen, under the name of generalised Principal-Agent problems, by Myerson [48] . These generalised agency problemswere then studied in a wide variety of economic settings, notably by Dionne and Lasserre [17], Laffont and Tirole [35],McAfee and McMillan [39], Picard [54], Baron and Besanko [4, 5], Melumad and Reichelstein [40, 41], Guesnerie, Picardand Rey [28], Page [49], Zou [69], Caillaud, Guesnerie and Rey [10], Lewis and Sappington [36], or Bhattacharyya [8] .All the previously mentioned models are either in static or discrete–time settings. The first study of the continuous timeproblem with moral hazard and adverse selection was made by Sung [64], in which the author extends the seminal finitehorizon and continuous–time model of Holmström and Milgrom [31]. A more recent work, to which our paper is mostlyrelated has been treated by Cvitanić, Wan and Yang [14], where the authors extend the famous infinite horizon model ∗ The authors gratefully acknowledges the support of the French Ministry of Foreign Affairs and the Merlion programme. † University of Michigan, [email protected]. ‡ Columbia University, IEOR, [email protected]. This author gratefully acknowledges the support of the ANR project Pacman,ANR-16-CE05-0027. § Department of Mathematics, National University of Singapore, Singapore, [email protected]. Research supported by NUS GrantR-146-000-179-133, Singapore MOE AcRF Grants R-146-000-219-112 and R-146-000-255-114. There were earlier attempts in this direction, but providing a less systematic treatment of the problem; see the income tax model ofMirrlees [46], the Soviet incentive scheme study of Weitzman [66], or the papers by Baron and Holmström [6] and Baron [3]. We refer the interested reader to the more recent works of Faynzilberg and Kumar [22], Theilen [65], Jullien, Salanié and Salanié [34],Gottlieb and Moreira [25]. a r X i v : . [ q -f i n . E C ] J a n f Sannikov [60] to the adverse selection setting. If one of the main contributions of Sannikov [60] was to have identifiedthat the continuation value of the Agent was a fundamental state variable for the problem of the Principal, [14] showsthat in a context with both moral hazard and adverse selection, the Principal has also to keep track of the so-called temptation value , that is to say the continuation utility of the Agent who would not reveal his true type. Althoughclose to the latter paper, our work is foremost an extension of the bank incentives model of Pagès and Possamaï [51],which studies the contracting problem between competitive investors and an impatient bank who monitors a pool oflong term loans subject to Markovian contagion (we also refer the reader to the companion paper by Pagès [50] for theeconomic intuitions and interpretations of the model).The home loan crash of 2008 has strongly highlighted the inherent weaknesses of the securitisation agreements createdduring the 2000s, and was at the heart of the decision from the US government to impose tight deadlines for theadoption of new and tighter regulations for credit risk retention. Among these, one of particular interest for us is theDodd–Frank Act, which prescribes that sponsors retain at least five percent of the credit risk in most securitisationtransactions. The purpose of [50, 51] was to study optimal securitisation when the sponsor remains involved with itsretail originations, and can engage in unobservable actions that result in private benefits at the expense of performance.The assumption that the bank itself can have impact on the default rate of the pool over time is a metaphor for thedistinction between its exogenous base quality and the endogenous default probability obtained after monitoring. Moralhazard then emerges because the bank has more "skin in the game" than the investors, and has the opportunity, exante and ex post , to exercise a (costly) monitoring of the non–defaulted loans. This is a stylised way to sum up allthe actions that the bank can enter into to ensure itself of the solvability of the borrowers. There is much that thebank can do to improve performance over the life of a transaction. First, a strong quality control process helps lendersexercise due diligence in evaluating borrowers’ current income, and keep track of those who might be getting closer todefault. This is a surveillance action which has to be undertaken continuously, and not only prior to the inception ofthe contract. Second, the bank can efficiently assist troubled borrowers by acting early and firmly, before mortgagesbecome seriously delinquent. The selection of bank employees in charge of these actions is also usually assumed topotentially affect loss severity by as much as 30%. For instance, Agarwal et al. [1] have put into light important andsystematic changes in the default rates of state–chartered banks’ real estate loans, when a so–called "rotation" policybetween federal and state supervisors at predetermined time periods is put into place. This clearly shows that banksare perfectly able to enter into corrective actions in the event of delinquencies, when they have incentives to do so.The findings of [50, 51] were that since the investors cannot observe the monitoring effort of the bank, they proposedCDS type contracts offering remuneration to the bank, and giving it incentives through postponement of payments andthreat of stochastic liquidation of the contract (similarly to the seminal paper of Biais, Mariotti, Rochet and Villeneuve[9]). In the present paper, we assume furthermore that there are two types of banks, which we call good and bad,co–existing in the market, differing by their efficiency in using their remuneration (or equivalently differing by theirmonitoring costs). Even if the investor is supposed to know the distribution of the type of banks, that is to say theprobability with which the bank he is currently discussing with is good or bad, he cannot know for sure what her typeis. Again, this is a stylised way to express the fact that "skin in the game" might significantly vary from one bankto the other. The fact that we consider only two types is mainly for simplicity and tractability, and because fullymultidimensional screening problems are already extremely hard to solve in static one–period models, and except forspecific models (see Section 2.1 for details), nothing more than existence of an optimal contract can be hoped for, seefor instance Carlier [11].Mathematically speaking, we follow the general dynamic programming approach of Cvitanić, Possamaï and Touzi[13], as well as that on adverse selection problems initiated by [14]. Intuitively, these approaches require first, usingmartingale (or more precisely backward SDEs) arguments, to solve the (non-Markovian) optimal control problem facedby the two types of banks when choosing contracts. This requires obviously, using the terminology introduced above,to keep track of both the continuation value and the temptation value of the banks, when they choose the contractdesigned for them or not. The problem of the Principal rewrites then as two standard stochastic control problems,one in which he hires the good bank, and one in which he hires the bad one. Each of these problems uses in turn theaforementioned two state variables (and these two only, because the horizon is infinite and the Principal is risk-neutral),with truth-telling constraint, asserting that the continuation value should always be greater than the temptation value.This leads to optimal control problems with state constraints, and thus to Hamilton-Jacobi-Bellman (HJB for short)equations (or more precisely variational inequalities with gradient constraints, since our problem is actually a singularstochastic control problem) in a domain, which, following [14], we call the credible set. This set is defined as the setcontaining the pair of value functions of the good and bad bank under every admissible contract offered by the investor.The determination of this set is the first fundamental step in our approach. Following the the orignal ideas of [14], weprove that the determination of the boundaries of this set can be achieved by solving two so-called double-sided moralhazard problems, in which one of the type of banks is actually hiring the other one. Fortunately for us, it turned out2o be possible to obtain rigorously explicit expressions for these boundaries by solving the associated system of HJBequations and using verification type arguments. We also would like to emphasise that unlike in [14], there is certaindynamic component in our model, since we have to keep track of the number of non-defaulted loans, through a timeinhomogeneous Poisson process. This leads to a dynamic credible set, as well as, in the end, to a recursive system ofHJB equations characterising the value function of the Principal.After having determined the credible set itself, we pursue our study by concentrating on two specific forms of contracts:the shutdown contract in which the investor designs a contract which will be accepted only by the good bank, andthe more classical screening contract, corresponding to a menu of contracts, one for each type of bank, which providesincentives to reveal her true type and choose the contract designed for her. These two contracts correspond simply tothe offering, over the correct domain of expected utilities of the banks (so as to satisfy the proper truth–telling andparticipation constraints), of the best contracts that the investor can design independently for hiring the good and thebad bank.Since we characterise, under classical verification type arguments, the value function of the investor through a system ofHJB equations, we also have classically access to the optimal contracts through this value function and its derivatives.This allows us to provide an associated qualitative and quantitative analysis. It turns out that the optimal contractsdesigned for the good and the bad bank share the same attributes, and are close in spirit to the ones derived in thepure moral hazard case in [51]. On the boundaries of the credible set, the value function of the bad bank plays the roleof a state process. The payments of the optimal contracts are postponed until the moment the state process reaches asufficiently high level, depending on the current size of the project. Similarly, when one of the loans in the pool defaults,the project is liquidated with a probability that decreases with the value of the state process. If the value function ofthe bad bank at the default time is below some critical level, the project will be liquidated for sure under the optimalcontracts. On the other side, if the value function of the bad bank is high enough at the default time, the project willbe maintained. In the interior of the credible set, the continuation value and the temptation value of the banks are thestate processes for the optimal contracts. It is possible to identify zones of good performance inside of the credible set,where the agents are remunerated and the project is maintained in case a default occurs. It is also possible to identifyzones of bad performance , where the agents are not paid and the project is liquidated in case of default. In the rest ofthe credible set the optimal contracts provide intermediary situations.The rest of the paper is organised as follows. In Section 2, we present the model, we define the set of admissible contractsand we state the investor’s problem. In Section 3, we recall the results obtained in [51] for the case of pure moral hazard,which will be useful later on for us. In Section 4, we formally study the credible set and obtain an explicit expressionfor it. In Section 5, we study both the optimal shutdown and screening contract, describing their characteristics andthe behaviour of the banks when they accept these contracts. The Appendix contains all the technical proofs of thepaper. Notations:
Let N denote the set of non–negative integers. For any n ∈ N \{ } , we identify R n with the set of n − dimensional column vectors. The associated inner product between two elements ( x, y ) ∈ R n × R n will be denotedby x · y . For simplicity of notations, we will sometimes write column vectors in a row form, with the usual transpositionoperator (cid:62) , that is to say ( x , . . . , x n ) (cid:62) ∈ R n for some x i ∈ R , ≤ i ≤ n . Let R + denote the set of non–negativereal numbers, and B ( R + ) the associated Borel σ − algebra. For any fixed non–negative measure ν on ( R + , B ( R + )) , theLebesgue–Stieljes integral of a measurable map f : R + −→ R will be denoted indifferently (cid:90) [ u,t ] f ( s )d ν s or (cid:90) tu f ( s )d ν s , ≤ u ≤ t. This section is dedicated to the description of the model we are going to study, presenting the contracts as well asthe criterion of both the Principal and the Agent. As recalled in the Introduction, it is actually an adverse selectionextension of the model introduced first by Pagès in [50] and studied in depth by Pagès and Possamaï [51]. Notice that in this respect the study in [14] was more formal, and our paper provides, as far as we know, the first rigorous derivation ofthis credible set. .1 Preliminaries We consider a model in continuous time, indexed by t ∈ [0 , ∞ ) . Without loss of generality and for simplicity, therisk–free interest rate is taken to be . Our first player will be a bank (the Agent, referred to as "she"), who hasaccess to a pool of I unit loans indexed by j = 1 , . . . , I which are ex ante identical. Each loan is a perpetuity yieldingcash flow µ per unit time until it defaults. Once a loan defaults, it gives no further payments. As is commonplace inthe Principal–Agent literature, especially since the paper of Sannikov [60], the infinite maturity assumption is here forsimplicity and tractability, since it makes the problem stationary, in the sense that the value function of the Principalwill not be time–dependent. We assume that the banks in the market are different, and that two types of banks coexist,each one being characterised by a parameter taking values in the set R := { ρ g , ρ b } with ρ g > ρ b . We call the bankgood (respectively bad) if its type is ρ g (respectively ρ b ). Furthermore, it is considered to be common knowledge thatthe proportion of the banks of type ρ i , i ∈ { g, b } , is p i ∈ (0 , .Denote by N t := I (cid:88) j =1 { τ j ≤ t } , the sum of individual loan default indicators, where τ j is the default time of loan j . The current size of the pool is,at some time t ≥ , I − N t . Since all loans are a priori identical, they can be reindexed in any order after defaults.The action of the banks consists in deciding at each time t ≥ whether they monitor any of the loans which have notdefaulted yet. These actions are summarised by the functions e j,it , where for ≤ j ≤ I − N t , i ∈ { g, b } , e j,it = 1 if loan j is monitored at time t by the bank of type ρ i , and e j,it = 0 otherwise. Non-monitoring renders a private benefit B > per loan and per unit time to the bank, regardless of its type. The opportunity cost of monitoring is thus proportionalto the number of monitored loans. Once more, more general cost structures could be considered, but this choice hasbeen made for the sake of simplicity.The rate at which loan j defaults is controlled by the hazard rate α jt specifying its instantaneous probability of defaultconditional on history up to time t . Individual hazard rates are assumed to depend on the monitoring choice of thebank and on the size of the pool. In particular, this allows to incorporate a type of contagion effect in the model.Specifically, we choose to model the hazard rate of a non–defaulted loan j at time t , when it is monitored (or not) bya bank of type ρ i as α j,it := α I − N t (cid:0) (cid:0) − e j,it (cid:1) ε (cid:1) , t ≥ , j = 1 , . . . , I − N t , i ∈ { b, g } , (2.1)where the parameters { α j } ≤ j ≤ I are positive constants representing individual “baseline” risk under monitoring whenthe number of loans is j , and ε > is the proportional impact of shirking on default risk. We assume that the impactof shirking is independent of the type of the bank. There are two main reasons for this choice. First of all, it is well–known that as soon as the dimension of the type is greater or equal to , we enter into the field of multidimensionalscreening, which, already for static one period models is notoriously hard to analyse, and deriving meaningful economicinterpretations is most often elusive (see the seminal paper of Rochet and Choné [56] or the more recent contributionof Figalli et al. [23] for more details). Notwithstanding this difficulty, we also found out that differentiating betweenthe banks in this regard created degeneracy in the model. We refer the reader to Section F.2 in the Appendix for amore detailed explanation.For i ∈ { b, g } , we define the shirking process k i as the number of loans that the bank of type ρ i fails to monitor at time t ≥ . Then, according to (2.1), the corresponding aggregate default intensity is given by λ k i t := I − N t (cid:88) j =1 α j,it = α I − N t (cid:0) I − N t + εk it (cid:1) . (2.2)The banks can fund the pool internally at a cost r ≥ . They can also raise funds from a competitive investor (thePrincipal, referred to as "he") who values income streams at the prevailing risk–less interest rate of zero. We assumethat both the banks and the investor observe the history of defaults and liquidations, as well as the parameters p b and p g , but the monitoring choices and the type of the bank are unobservable for the investor. Before going on, let us now describe the stochastic basis on which we will be working. We will always place ourselveson a probability space (Ω , F , P ) on which N is a point process with intensity λ t , which is defined by (2.2) when all As already pointed out in the seminal paper of Biais, Mariotti, Rochet and Villeneuve [9], see also [51], the only quantity of interesthere is the difference between the discounting factors of the Principal and the Agent. λ t = α I − N t (cid:0) I − N t (cid:1) . We denote by F := ( F Nt ) t ≥ the P − completion of the natural filtration of N . We call τ the liquidation time of thewhole pool and let H t := { t ≥ τ } be the liquidation indicator of the pool. We denote by G := ( G t ) t ≥ the minimalfiltration containing F and that makes τ a G − stopping time. We note that this filtration satisfies the usual hypothesesof completeness and right–continuity.Contracts are offered by the investor to the bank and agreed upon at time . As usual in contracting theory, thebank can accept or refuse the contract, but once accepted, both the bank and the investor are fully committed to thecontract. More precisely, the investor offers a menu of contracts Ψ i := ( k i , θ i , D i ) , i ∈ { g, b } specifying on the one handa desired level of monitoring k i for the bank of type ρ i , which is a G − predictable process such that for any t ≥ , k it takes values in { , . . . , I − N t } (this set is denoted by K ), as well as a flow of payment D i . These payments belong to set D of processes which are càdlàg, non–decreasing, non–negative, G − predictable and such that there exists some β > E P (cid:20) e βτ (cid:18) (cid:90) τ e − rs d D is (cid:19) (cid:21) < + ∞ . (2.3)We do not rule out the possibility of immediate lump–sum payments at the initialisation of the contract, and thereforethe processes in D are assumed to satisfy D − = 0 . Hence, if D (cid:54) = 0 , it means that a lump–sum payment has indeedbeen made. Notice also that since the intensities of N and H under P are bounded, we know that τ has at least someexponential moments under P , meaning that any bounded payment belongs to D .The contract also specifies when liquidation occurs. We assume that liquidations can only take the form of the stochasticliquidation of all loans following immediately default. Hence, the contract specifies the probability θ it , which belongsto the set Θ of [0 , − valued, G − predictable processes, with which the pool is maintained given default ( d N t = 1 ), sothat at each point in time, if the bank has indeed chosen the contract Ψ i d H t = (cid:40) with probability θ it , d N t with probability − θ it . With our notations, given a contract Ψ i , the hazard rates associated with the default and liquidation processes N t and H t are, if the bank does choose the contract Ψ i , λ k i t and (cid:0) − θ it (cid:1) λ k i t , respectively. The above properties translate into P (cid:2) τ ∈ (cid:8) τ , ..., τ I (cid:9) (cid:3) = 1 , and P (cid:2) τ = τ j |F τ j , τ > τ j − (cid:3) = 1 − θ iτ j , j ∈ { , . . . , I } . For ease of notations, a contract
Ψ := ( k, θ, D ) will be said to be admissible if ( k, θ, D ) ∈ K × Θ × D . As is commonplacein the Principal–Agent literature, we assume that the monitoring choices of the banks affect only the distribution of thesize of the pool. To formalise this, recall that, by definition, any shirking process k ∈ K is G − predictable and bounded.Then, by Girsanov’s theorem, we can define a probability measure P k on (Ω , F ) , equivalent to P , such that N t − (cid:82) t λ kt d s is a P k − martingale. More precisely, we have on G t d P k d P = Z kt , where Z k is the unique solution of the following SDE Z kt = 1 + (cid:90) t Z ks − (cid:18) λ ks λ s − (cid:19)(cid:0) d N s − λ s d s (cid:1) , ≤ t ≤ τ, P − a . s . Then, if the bank of type ρ i chooses the contract Ψ i , her utility at t = 0 , if she follows the recommendation k i , is givenby u i ( k i , θ i , D i ) := E P ki (cid:20) (cid:90) τ e − rs (cid:0) ρ i d D is + Bk is d s (cid:1)(cid:21) , (2.4)while that of the investor is v (cid:0) (Ψ i ) i ∈{ g,b } (cid:1) := (cid:88) i ∈{ g,b } p i E P ki (cid:20) (cid:90) τ (cid:0) I − N s (cid:1) µ d s − d D is (cid:21) . (2.5)The parameter ρ i actually discriminates between the two types of banks through the way they derive utility from thecash–flows delivered by the investor. Hence, for a same level of salary, the good bank will get more utility than a badbank. Such a form of adverse selection is also considered in the paper of Cvitanić, Wan and Yang [14]. Notice that dedependence of the value functions of both the bank and the investor depend on the contract through the process D i ,but also through the stopping time τ , whose distribution depends both on θ i and the effort choice k i of the bank.5 .3 Formulation of the investor’s problem We assume that the bank of type ρ i has an outside opportunity to the contract which provides her reservation utility R i . The investor’s problem is to offer a menu of admissible contracts (Ψ i ) i ∈{ g,b } := ( k i , θ i , D i ) i ∈{ g,b } which maximiseshis utility (2.5), subject to the three following constraints u i ( k i , θ i , D i ) ≥ R i , i ∈ { g, b } , (2.6) u i ( k i , θ i , D i ) = sup k ∈ K u i ( k, θ i , D i ) , i ∈ { g, b } , (2.7) u i ( k i , θ i , D i ) ≥ sup k ∈ K u i ( k, θ j , D j ) , i (cid:54) = j, ( i, j ) ∈ { g, b } . (2.8)Condition (2.6) is the usual participation constraint for the banks. Condition (2.7) is the so–called incentive compati-bility condition, stating that given ( θ i , D i ) , the recommended effort k i is an optimal monitoring choice for the bank oftype ρ i . Finally, Condition (2.8) means that if a bank adversely selects a contract, she cannot get more utility than ifshe had truthfully revealed her type at time . Following the literature, we call such a contract a screening contract.In the sequel, we will start by deriving the optimal contract in the pure moral hazard case, then we will look into the so–called optimal shutdown contract, for which the investor deliberately excludes the bad bank, before finally investigatingthe optimal screening contract. We will invoke some results from [50] in this paper, for this reason we will require laterthe assumptions of their main result, Theorem 3.15, which are the following. Assumption 2.1.
Let α I be the harmonic mean of the ( α j ) ≤ j ≤ I , ( i ) µ ≥ α I . ( ii ) We have for all j ≤ I , rB (1 + ε ) ≤ ( µε − B ) εα j . ( iii ) Individual default risk is non–decreasing with past default, α j ≤ α j − , for all j ≤ I. Let us start by discussing Assumption 2.1. Concerning ( i ) , under monitoring, the expected duration until the nextdefault in a pool of j loans is /λ j . Hence, the average revenue from the pool over that period will be given by µ/λ j ,of which /I is ascribed to the original loan. The payoff of a loan corresponds then to summing this quantity over j ,and the obtained result must be above the initial unit cost for the loan to be worth anything at all under monitoring.Assumption 2.1 ( ii ) imposes an upper bound on the bank’s discount rate, and basically states that it should not be solarge that the cost of the rent extracted by a monitoring bank outweighs the pecuniary gains stemming from the use ofthe monitoring technology. Finally Assumption 2.1 ( iii ) simply models a contagion effect, translating the fact that pastdefaults impact positively the likelihood of a further default to happen.A second important point in the model is the liquidation policy of the contract. Even though liquidations are inefficientin the first–best situation without moral hazard nor adverse selection (see [50]), they are necessary in the second–bestin order to restore incentives to monitor when performance is poor. However, liquidation can take many forms, andfor instance liquidating all loans with state–dependent probability is not necessarily better than partially liquidatingthe pool with fixed probability. Another option would be to downsize the pool by a potentially larger number of loans,possibly state–dependent as well, whenever a default occurs. Given that in practice liquidations are rarely decided insuch a random fashion, it is of the utmost importance to verify that such liquidation policies cannot improve on socialwelfare. In the case of pure moral hazard, [50, Proposition 6] has shown that stochastic liquidation was optimal amongall policies under an assumption which is met if changes in default intensities for the loans are gradual. Since thecapital structure of subprime mortgage–backed securities is typically split up into a large number of tranches, where itis then reasonable to assume that default intensities are constant, such an assumption will be verified in the real–worldapplications of the model.Another very important assumption here is the fact that we consider a "full commitment" dynamic contracting problembetween the investor and the bank. In other words, both parties are fully committed to the long–term dynamic contractat the onset of the relationship. However, one of the central features of banks, namely the fragility of their capitalstructures, stems from the fact that there is usually a limited commitment in the relationship between clients and banks,since, as highlighted by the seminal paper of Diamond and Dybvig [16], bank cleints can withdraw funds from banksat any time. We are perfectly conscient of this fact and have chosen to postpone the discussion of how to integratenon–commitment in our model to Section 6.2 below, since the mathematics behind are similar.6n our model, adverse selection stems only from the monitoring costs of the two types of banks. A possible extensionof our model could rely on a further differentiation between the work of the banks, i.e. when both good bank and badbank work, the good one would be more efficient in the sense that the associated default intensity is strictly smallerthan that of the bad bank. This would also be a possible way to model the fact that the pool of loans of each bankcould have different qualities. However, as will be explained in Appendix F.2, such a feature would actually make theproblem degenerate, in the sense that the upper boundary of the credible set that will be defined in Section 4 becomesinfinite. This would be a rather undesirable feature of the model and would create unwanted discontinuities. We givepotential solutions to extend the model in this direction in Appendix F.2, but leave the exact study to future research. In this section, we assume that the type of the bank is publicly known and is fixed to be some ρ i , i ∈ { g, b } , which makesthe problem exactly similar to the one considered in [51] (up to the modification of some constants). In particular, theinvestor only offers one contract. The results we obtain here, in particular the dynamics of the continuation utilities ofthe banks, will be crucial to the study of the shutdown and screening contracts later on. Therefore, they will be usedthroughout the paper without further references.In this setting, the utility of the investor, when he offers a contract ( k i , θ i , D i ) ∈ K × Θ × D is given by v pm0 ( k i , θ i , D i ) := E P ki (cid:20) (cid:90) τ (cid:0) I − N s (cid:1) µ d s − d D is (cid:21) , (3.1)for which we define the following dynamic version for any t ≥ v pm t ( k i , θ i , D i ) := E P ki (cid:20) (cid:90) τt ∧ τ (cid:0) I − N s (cid:1) µ d s − d D is (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . As usual, the so–called continuation value of the bank (that is to say her future expected payoff) when offered ( θ i , D i ) ∈ Θ × D plays a central role in the analysis. It is defined, for any ( t, k ) ∈ R + × K by u it ( k, θ i , D i ) := E P k (cid:20) (cid:90) τt ∧ τ e − r ( s − t ) (cid:0) ρ i d D is + k s B d s (cid:1)(cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . We also define the value function of the bank for any t ≥ U it ( θ i , D i ) := ess sup k ∈ K u it ( k, θ i , D i ) . Departing slightly from the usual approach in the literature, initiated notably by Sannikov [60, 61], we reinterpretthe problem of the bank in terms of BSDEs, which, we believe, offers an alternative approach which may be easier toapprehend for the mathematical finance community. Of course, such an interpretation of optimal stochastic controlproblem with control on the drift is far from being original, and we refer the interested reader to the seminal papersof Hamadène and Lepeltier [29] and El Karoui and Quenez [19] for more information, as well as to the recent articlesby Cvitanić, Possamaï and Touzi [12, 13] for more references and a systematic treatment of Principal–Agent typeproblems with this backward SDE approach. Before stating the related result, let us denote by ( Y i , Z i ) the unique(super–)solution (existence and uniqueness will be justified below) to the following BSDE Y it = 0 − (cid:90) τt g i ( s, Y is , Z is )d s + (cid:90) τt Z is · d (cid:102) M is + (cid:90) τt d K is , ≤ t ≤ τ, P − a . s ., (3.2)where M t := (cid:18) N t H t (cid:19) , (cid:102) M it := M t − (cid:90) t λ s (cid:18) − θ is (cid:19) d s, K it := ρ i D it , f i ( t, k, y, z ) := ry − Bk + kα I − N t εz · (cid:18) − θ it (cid:19) ,g i ( t, y, z ) := inf k ∈{ ,...,I − N t } f i ( t, k, y, z ) = ry − ( I − N t ) (cid:18) α I − N t εz · (cid:18) − θ it (cid:19) − B (cid:19) − . Y i and Z i which are respectively G − progressively measurable and continuous, and G − predictable, and satisfy E P (cid:20) sup ≤ t ≤ τ e ( βε − r ) t (cid:12)(cid:12) Y it (cid:12)(cid:12) (cid:21) < + ∞ , and E P (cid:20) (cid:90) τ e ( βε − r ) s (cid:107) Z is (cid:107) d s (cid:21) < + ∞ , (3.3)where β is the exponent associated to D i by (2.3). We have the following proposition, which is basically a reformulationof [51, Proposition 3.2]. The proof is postponed to Appendix A Proposition 3.1.
For any ( θ i , D i ) ∈ Θ × D , the value function of the bank has the dynamics, for t ∈ [0 , τ ] , P − a . s . d U it ( θ i , D i ) = (cid:18) rU it ( θ i , D i ) − Bk (cid:63),it + λ k (cid:63),i t Z it · (cid:18) − θ it (cid:19) (cid:19) d t − ρ i d D it − Z it · d (cid:102) M it , where Z i is the second component of the solution to the BSDE (3.2) . In particular, the optimal monitoring choice ofthe bank is given by k (cid:63),it = ( I − N t ) { Z it · (1 , − θ it ) (cid:62)
Definition 3.1.
We call V it the feasible set for the expected payoff of banks of type ρ i , starting from some time t ≥ ,that is to say all the possible utilities that a bank of type ρ i can get from all the admissible contracts offered by theinvestor from time t on. Our next result gives an explicit form of the the feasible set V it , which turns out to be independent of the type of thebank. The proof is relegated to Appendix A, and requires the introduction of k SH , the strategy of a bank which doesnot monitor any loan at any time, i.e. k SH s := I − N s for every s ≥ Lemma 3.1.
For i ∈ { g, b } and for any t ≥ , we have that V it = V t , with V t := (cid:20) B ( I − N t ) r + λ k SH t , + ∞ (cid:19) . In this section we come back to the case in which there are two types of banks in the market, and study the so–calledcredible set, which is formed by the pairs of value functions of the banks under the admissible contracts. As in [14],we do not expect all the points in the feasible set to correspond to a pair of reachable values of the banks under someadmissible contract. We will therefore follow the approach initiated by [14] and we will characterise the credible set.We emphasise an important difference with [14] though, in the sense that in our context, the credible set becomesdynamic as it depends on the current size of the pool. In this section we work with generic contracts ( θ, D ) ∈ Θ × D ,not necessarily designed for a particular type of bank. We introduce some notations first. Let (cid:98) λ SH j be the default intensity under k SH when there are j loans left, that is tosay (cid:98) λ SH j := α j j (1 + ε ) . We assume here, as is commonplace in the Principal–Agent literature, that in the case where the bank is indifferent with respect toher monitoring decision, that is when Z it · (1 , − θ it ) (cid:62) = b t , she acts in the best interest of the investors, and thus monitors all the I − N t remaining loans. (cid:98) λ SH j = λ k SH t = α I − N t ( I − N t )(1 + ε ) , for every t ≥ such that I − N t = j . Define then for any integer j between and I , the set (cid:98) V j := (cid:2) Bj/ (cid:0) r + (cid:98) λ SH j (cid:1) , ∞ (cid:1) . Observe that the feasible set V t = (cid:20) B ( I − N t ) r + λ k SH t , + ∞ (cid:19) , satisfies V t = (cid:98) V I − N t for every t ≥ , so the only dependence of the feasible set in time is due to the number of loansleft. The rigorous definition of the credible set is the following. Definition 4.1.
For any time t ≥ , we define the credible set C t as the set of ( u b , u g ) ∈ V t × V t such that there existssome admissible contract ( θ, D ) ∈ Θ × D satisfying U bt ( θ, D ) = u b , U gt ( θ, D ) = u g and ( U bs ( θ, D ) , U gs ( θ, D )) ∈ V s × V s for every s ∈ [ t, τ ) , P − a.s.Given a starting time t ≥ and u b ∈ V t , define the set of contracts under which the value function of the bad bank attime t is equal to u b A b ( t, u b ) := (cid:8) ( θ, D ) ∈ Θ × D : U bt ( θ, D ) = u b (cid:9) . We denote by U t ( u b ) the largest value U gt ( θ, D ) that the good bank can obtain from all the contracts ( θ, D ) ∈ A b ( t, u b ) .We also denote the lowest value by L t ( u b ) . Next, define C t := (cid:8) ( u b , u g ) ∈ V t × V t : L t ( u b ) ≤ u g ≤ U t ( u b ) (cid:9) . We will prove in Proposition 4.4 below that C t = C t for every t ≥ , and that the dependence on time of the credibleset, exactly as for the feasible set, only comes from the value of I − N t . In particular, this allows us to call respectivelythe functions L t and U t the lower and upper boundary of the credible set when there are I − N t loans left. The aimof the next sections is prove all these claims and to obtain explicit formulas for the boundaries. We start with someuseful technical results concerning specific contracts for which the banks do not monitor the loans at all. Consider any starting time t such that I − N t = j and any θ ∈ Θ . The continuation utility that the banks get fromalways shirking (without considering the payments) is u gt (cid:0) k SH , θ, (cid:1) = u bt (cid:0) k SH , θ, (cid:1) = E P k SH (cid:20) (cid:90) τt ∧ τ e − r ( s − t ) Bk SH s d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . (4.1)This quantity is obviously non–decreasing in θ , so that (4.1) attains its minimum value under any contract with θ ≡ ,which is equal to c ( j,
1) :=
Bj/ (cid:0) r + (cid:98) λ SH j (cid:1) . The following proposition provides the value of (4.1) when the pool isliquidated exactly after a fixed number of defaults m . Proposition 4.1.
Fix some t ≥ and let j := I − N t . For m ∈ { , . . . , j } , let θ m ∈ Θ be such that the pool is liquidatedexactly after the m th default occurring after time t , that is θ ms := (cid:40) , t ≤ s ≤ τ N t + m , , s > τ N t + m . The utility that the bank of type ρ i gets from shirking is c ( j, m ) := Bjr + (cid:98) λ SH j + j − (cid:88) i = j − m +1 Bir + (cid:98) λ SH i j (cid:89) (cid:96) = i +1 (cid:98) λ SH (cid:96) r + (cid:98) λ SH (cid:96) . In particular, under any contract such that θ ≡ , (4.1) attains its maximum value, which is equal to C ( j ) := c ( j, j ) = Bjr + (cid:98) λ SH j + j − (cid:88) i =1 Bir + (cid:98) λ SH i j (cid:89) (cid:96) = i +1 (cid:98) λ SH(cid:96) r + (cid:98) λ SH (cid:96) . (4.2)9 .3 Lower boundary of the credible set The lower boundary of the credible set is the simpler of the two boundaries and it can be computed directly. We willsee that it is a piecewise linear function corresponding to two lines with different slopes. All proofs for this section arecollected in Appendix C. The next proposition states the main inequalities that determine the lower boundary.
Lemma 4.1.
For any t ∈ [0 , τ ] and any admissible contract ( θ, D ) ∈ Θ × D , the value functions of the good and thebad banks satisfy, P − a.s. U gt ( θ, D ) ≥ U bt ( θ, D ) , (4.3) U gt ( θ, D ) ≥ ρ g ρ b U bt ( θ, D ) − ( ρ g − ρ b ) ρ b C ( I − N t ) , (4.4) where the function C is defined in (4.2) . Using Lemma 4.1, we prove the following characterisation of the lower boundary of the credible set.
Proposition 4.2.
For any t ≥ , and any u b ∈ V t , the lower boundary of the credible set is given by L t ( u b ) = u b , c ( I − N t , ≤ u b ≤ C ( I − N t ) ,ρ g ρ b u b − ( ρ g − ρ b ) ρ b C ( I − N t ) , C ( I − N t ) ≤ u b < + ∞ . In particular, the dependence in t of L t ( u b ) only comes from the number of non–defaulted loans at time t and we candefine for any j ∈ { , . . . , I } , the quantity (cid:98) L j ( u b ) given by (cid:98) L j ( u b ) := u b , c ( j, ≤ u b ≤ C ( j ) ,ρ g ρ b u b − ( ρ g − ρ b ) ρ b C ( j ) , C ( j ) ≤ u b < + ∞ , for which we have (cid:98) L I − N t ( u b ) = L t ( u b ) . Remark 4.1.
Of course, the computations of this section depend on our modelling choices, and are unlikely to bedirectly adaptable to other situations. There is however a generic way of finding the lower boundary ( as well as theupper one ) which we give details in the next section. It amounts to solving a fictitious contract situation where the goodbank hires the bad one and minimises ( maximises for the upper boundary ) her utility over her monitoring choices, andover all contracts for which the bad bank receives a fixed utility u b . The dynamic value function of this control problemis exactly L t ( u b ) , since it corresponds to the minimal utility that the good bank can have when the bad one receives u b . The upper boundary of the credible set is not as simple to obtain as the lower boundary and we have to solve a specificstochastic control problem to identify it. Notice that this approach is similar to the one used in [14].Let us fix any contract ( θ, D ) ∈ Θ × D . We remind the reader that thanks to Proposition 3.1, we know that thereexist G − predictable integrable processes ( h ,g ( θ, D ) , h ,g ( θ, D )) satisfying the second integrability condition in (3.3)and such that d U gs ( θ, D ) = (cid:0) rU gs ( θ, D ) − Bk (cid:63),gs ( θ, D ) (cid:1) d s − ρ g d D s − h ,gs ( θ, D ) (cid:0) d N s − λ k (cid:63),g ( θ,D ) s d s (cid:1) − h ,gs ( θ, D ) (cid:0) d H s − (1 − θ s ) λ k (cid:63),g ( θ,D ) s d s (cid:1) , s ∈ [0 , τ ] , (4.5)where the optimal monitoring choice k (cid:63),g ( θ, D ) is given by k (cid:63),gs ( θ, D ) = ( I − N s ) { h ,gs ( θ,D )+(1 − θ s ) h ,gs ( θ,D ) E P (cid:20) (cid:90) τ e ( βε − r ) s | h s | d s (cid:21) < + ∞ . We abuse notations and define, for every
Ψ := (
D, θ, h ,g , h ,g , h ,b , h ,b ) ∈ D × Θ × H , the processes U g (Ψ) and U b (Ψ) which satisfy the following SDEs d U gs (Ψ) = (cid:0) rU gs (Ψ) − Bk (cid:63),gs (Ψ) (cid:1) d s − ρ g d D s − h ,gs (cid:0) d N s − λ k (cid:63),g (Ψ) s d s (cid:1) − h ,gs (cid:0) d H s − (1 − θ s ) λ k (cid:63),g (Ψ) s d s (cid:1) , (4.7) d U bs (Ψ) = (cid:0) rU bs (Ψ) − Bk (cid:63),bs (Ψ) (cid:1) d s − ρ b d D s − h ,bs (cid:0) d N s − λ k (cid:63),b (Ψ) s d s (cid:1) − h ,bs (cid:0) d H s − (1 − θ s ) λ k (cid:63),b (Ψ) s d s (cid:1) , (4.8)where we defined k (cid:63),gs (Ψ) := ( I − N s ) { h ,gs +(1 − θ s ) h ,gs
In the model, there is no need to consider h ,g and h ,b as positive processes and we do this just fortechnical reasons. Intuitively, the optimal contracts should satisfy this additional constraint because the investor doesnot benefit from earlier defaults and if a contract increases the banks’ continuation utilities after one of the defaults,the banks should increase the default intensity as much as possible. Remark 4.3.
It is immediate from the definition that given
Ψ := (
D, θ, h ,g , h ,g , h ,b , h ,b ) ∈ D × Θ ×H , the dynamicsof U g (Ψ) only depends on ( D, θ, h ,g , h ,g ) , while the dynamics of U b (Ψ) only depends on ( D, θ, h ,b , h ,b ) . We will thussometimes also use the notations U g (Ψ) , U b (Ψ) , k (cid:63),g (Ψ) and k (cid:63),b (Ψ) when Ψ ∈ D × Θ × H . For fixed ( t, u b , u g ) ∈ R + × V t , we define the set of contracts A ( t, u b , u g ) as the set of Ψ := (
D, θ, h ,g , h ,g , h ,b , h ,b ) ∈D × Θ × H such that (4.7) and (4.8) have at least one weak solution , which satisfies the first integrability conditionin (3.3), and in addition U is − (Ψ) = h ,is + h ,is , U is − (Ψ) − h ,is ≥ B ( I − N s ) r + λ SH s , ∀ s ∈ [ t, τ ] , U it (Ψ) = u i , i ∈ { b, g } . What we claimed above is that all processes ( D, θ ) ∈ D× Θ can be obtained from a contract ( D, θ, h ,g , h ,g , h ,b , h ,b ) =:Ψ ∈ A (0 , u b , u g ) , meaning that we are not enlarging at all the class of admissible contracts in our reformulation. Indeed,we already know by the results from Section 3, that the continuation utilities of the good and the bad bank given acontract ( D, θ ) ∈ D × Θ were completely characterised as being the unique solutions of the corresponding BSDEs (3.2)satisfying in addition (3.3). If we take some Ψ ∈ A (0 , u b , u g ) , then it is immediate that the processes U g (Ψ) , U b (Ψ) solve the corresponding BSDEs (3.2), since the dynamics is the correct one by definition, we have U gτ (Ψ) = U bτ (Ψ) = 0 ,and all the required integrability conditions are satisfied. By uniqueness of the solution to the BSDEs, we thus musthave U g (Ψ) = U g ( θ, D ) , and U b (Ψ) = U b ( θ, D ) .To describe the stochastic control problem for the upper boundary of the credible set, we need to introduce additionalnotations. For any starting time t ∈ [0 , τ ] and for every u b ≥ B ( I − N t ) / (cid:0) r + (cid:98) λ SH I − N t (cid:1) , we let A b ( t, u b ) be the set ofquadruplets Ψ = (
D, θ, h ,b , h ,b ) ∈ D × Θ × H such that (4.8) has at least one weak solution, which satisfies the firstintegrability condition in (3.3) as well as U bs − (Ψ) = h ,bs + h ,bs , U bs − (Ψ) − h ,bs ≥ B ( I − N s ) r + λ I − N s s , ∀ s ∈ [ t, τ ] , U bt (Ψ) = u b . We will abuse notations and also call elements of A b ( t, u b ) contracts. The upper boundary U t solves the followingcontrol problem U t ( u b ) = ess sup ( k g , Ψ) ∈ K ×A b ( t,u b ) E P kg (cid:20) (cid:90) τt ∧ τ e − r ( s − t ) (cid:0) ρ g d D s + Bk gs d s (cid:1)(cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) , In general, all the processes in Ψ could for instance functionals of the paths of U g (Ψ) and U b (Ψ) , in which case wellposedness of theSDEs has to be assumed as part of the definition. U br (Ψ) = u b + (cid:90) rt (cid:0) ru bs − Bk (cid:63),bs (Ψ) + h ,bs λ k (cid:63),b s + h ,bs (1 − θ s ) λ k (cid:63),b s (cid:1) d s − ρ b d D s − (cid:90) rt h ,bs d N s − (cid:90) rt h ,bs d H s , r ∈ [ t, τ ] . Indeed, the above stochastic control problem corresponds to the highest value that the good bank can obtain from anyadmissible contract, while ensuring that when the bad bank takes said contract, she receives exactly u b , which is exactlythe definition of the upper boundary of the credible set. Another way to interpret this problem is that it correspondsto the (fictitious) situation where the good bank hires the bad one, when the latter wants to receive a utility of u b ,and maximises her utility among all contracts ensuring that this constraint is satisfied. The importance of the resultsof Proposition 3.1 is that it allows us to obtain easily the dynamic behaviour of the continuation utility of the badbank for any initial utility, which in turns allows us to express simply the constraint in the problem for the good bankthrough the set A b and the state variable U b (Ψ) .The next subsections are devoted to first obtaining the HJB equation associated with the above problem, its resolutionand then finally to the proof of a verification theorem adapted to our framework. Notice that the above is actually asingular stochastic control problem, since the control D is a non–decreasing process, which is not necessarily absolutelycontinuous with respect to the Lebesgue measure. We refer the reader to the monograph by Fleming and Soner [24] formore details. In particular, this implies that the HJB equation associated to the problem will be a variational inequalitywith gradient constraints. Exactly as in the case of the lower boundary, we expect that the time dependence of of the upper boundary only comesfrom the current number of remaining loans. In such a case, the HJB equations that will describe the behaviour of theupper boundary necessarily form a recursive system, with the upper boundary when j loans are left depending on theone with j − loans left. We will write down this system, solve it explicitly, and prove a verification theorem ensuringthat our initial guess was indeed correct.Fix some ≤ j ≤ I , and define for every k = 0 , , · · · , j , (cid:98) λ kj := α j ( j + kε ) . The system of HJB equations associated tothe previous control problem is given by (cid:98) U ≡ , and for any ≤ j ≤ I and u b ≥ Bjr + (cid:98) λ SH j min (cid:40) − sup ( θ,h ,h ) ∈ C j (cid:40) (cid:98) U (cid:48) j ( u b ) (cid:0) u b − Bk b + ( h + (1 − θ ) h ) (cid:98) λ k b j (cid:1) + (cid:98) λ k g j θ (cid:98) U j − ( u b − h ) − ( (cid:98) λ k g j + r ) (cid:98) U j ( u b ) + Bk g (cid:41) , (cid:98) U (cid:48) j ( u b ) − ρ g ρ b (cid:41) = 0 , (4.9)with the additional boundary condition (cid:98) U j ( Bj/ ( r + (cid:98) λ SHj )) =
Bj/ ( r + (cid:98) λ SHj ) , and where we defined for simplicity k b := j { h +(1 − θ ) h < (cid:98) b j } , k g := j { (cid:98) U j ( u b ) − θ (cid:98) U j − ( u b − h ) < (cid:98) b j } , as well as C j := (cid:26) ( θ, h , h ) ∈ [0 , × R : h + h = u b , h ≥ B ( j − r + (cid:98) λ SH j − (cid:27) . Remark 4.4.
Notice that if our guess on the time dependence of the upper boundary is correct, we must have for any s ≥ , U s = (cid:98) U I − N s . Then. the incentive compatibility condition for the good bank is implicit in the HJB equation.Indeed, at every s ≥ we have (cid:98) U I − N s (cid:0) U bs (Ψ) (cid:1) − (cid:98) U I − N s − (cid:0) U bs − (Ψ) (cid:1) = (cid:0) (cid:98) U I − N s − − (cid:0) U bs − (Ψ) − h ,bs (Ψ) (cid:1) − (cid:98) U I − N s − (cid:0) U bs − (Ψ) (cid:1)(cid:1) ∆ N s − (cid:98) U I − N s − − (cid:0) U bs − (Ψ) − h ,bs (Ψ) (cid:1) ∆ H s , which implies that on the upper boundary h ,gs (Ψ) = (cid:98) U I − N s − (cid:0) U bs − (Ψ) (cid:1) − (cid:98) U I − N s − − (cid:0) U bs − (Ψ) − h ,bs (Ψ) (cid:1) and h ,gs (Ψ) = (cid:98) U I − N s − − (cid:0) U bs − (Ψ) − h ,bs (Ψ) (cid:1) . Therefore h ,gs (Ψ) + (1 − θ gs ) h ,gs (Ψ) = (cid:98) U I − N s − (cid:0) U bs − (Ψ) (cid:1) − θ gs (cid:98) U I − N s − − (cid:0) U bs − (Ψ) − h ,bs (Ψ) (cid:1) . At the points where (cid:98) U (cid:48) j ( u b ) > ρ g /ρ b , the first term of the variational inequality (4.9) must be equal to zero, so theupper boundary must satisfy the following equation r (cid:98) U j ( u b ) = sup ( θ,h ,h ) ∈ C j (cid:110) (cid:98) U (cid:48) j ( u b ) (cid:0) ru b − Bk b + ( h + (1 − θ ) h ) (cid:98) λ k b j (cid:1) + (cid:0) (cid:98) U j − ( u b − h ) θ − (cid:98) U j ( u b ) (cid:1)(cid:98) λ k g j + Bk g (cid:111) . (4.10)12e will refer to this equation as the diffusion equation. • Step 1: case of 1 loan, solving the diffusion equation
Before dealing with the variational inequality (4.9), we will solve the diffusion equation (4.10). When j = 1 , it reducesto r (cid:98) U ( u b ) = (cid:98) U (cid:48) ( u b ) (cid:0) ru b − Bk b + u b (cid:98) λ k b (cid:1) − (cid:98) U ( u b ) (cid:98) λ k g + Bk g , (4.11)with k b = { u b < (cid:98) b } , k g = { (cid:98) U ( u b ) < (cid:98) b } . Remark 4.5.
Notice that the boundary condition (cid:98) U (cid:0) Br + (cid:98) λ (cid:1) = Br + (cid:98) λ is implicit in the equation. Our first result is the following, whose proof is deferred to Appendix D.
Lemma 4.2.
There is a family of continuously differentiable solutions to the diffusion equation (4.10) , indexed by someconstant C > , which are given by (cid:98) U C ( u b ) := C r + (cid:98) λ r + (cid:98) λ (cid:18) u b − Br + (cid:98) λ (cid:19) + Br + (cid:98) λ , u b ∈ (cid:20) Br + (cid:98) λ , x C ,(cid:63) (cid:19) ,C (cid:98) b (cid:98) λ − (cid:98) λ r + (cid:98) λ (cid:18) r + (cid:98) λ r + (cid:98) λ (cid:19) r + (cid:98) λ r + (cid:98) λ (cid:18) u b − Br + (cid:98) λ (cid:19) r + (cid:98) λ r + (cid:98) λ , u b ∈ (cid:2) x C ,(cid:63) , (cid:98) b (cid:1) ,C u b , u b ∈ (cid:2)(cid:98) b , + ∞ (cid:1) , where x C ,(cid:63) := (cid:18) C (cid:19) r + (cid:98) λ r + (cid:98) λ (cid:98) b r + (cid:98) λ r + (cid:98) λ + Br + (cid:98) λ . • Step 2: case of 1 loan, solving the HJB equation
In this case the variational inequality (4.9) reduces to min (cid:26) r (cid:98) U ( u b ) − (cid:98) U (cid:48) ( u b ) (cid:0) ru b − Bk b + u b (cid:98) λ k b (cid:1) + (cid:98) U ( u b ) (cid:98) λ k g − Bk g , (cid:98) U (cid:48) ( u b ) − ρ g ρ b (cid:27) = 0 . (4.12)We already found the solutions of the diffusion equation inside of this variational inequality and now we will take care ofthe whole HJB equation. We expect the upper boundary to saturate the second term in the variational inequality for bigvalues of u b , so we will search for a solution of (4.12) satisfying the following condition: there exists x (cid:63) ∈ [ B/ ( r + (cid:98) λ ) , ∞ ) such that (cid:98) U (cid:48) ( x (cid:63) ) = ρ g ρ b and (cid:98) U (cid:48) ( u b ) > ρ g ρ b , for u b < x (cid:63) . (4.13)At first sight it could seem that by doing this we face the risk of not finding the correct solution of the dynamicprogramming equation. Nevertheless, this is not the case and we will prove later a verification result which assures usthat the solution that we find under this condition corresponds indeed to the upper boundary of the credible set. Theproof of the following Lemma will be given in Appendix D. Lemma 4.3.
The unique solution of the
HJB equation (4.12) which satisfies condition (4.13) is given by, defining x (cid:63) := x ρ g /ρ b ,(cid:63) (cid:98) U (cid:63) ( u b ) := (cid:98) U ρ g /ρ b ( u b ) = (cid:18) ρ g ρ b (cid:19) r + (cid:98) λ r + (cid:98) λ (cid:18) u b − Br + (cid:98) λ (cid:19) + Br + (cid:98) λ , u b ∈ (cid:20) Br + (cid:98) λ , x (cid:63) (cid:19) ,ρ g ρ b (cid:98) b (cid:98) λ − (cid:98) λ r + (cid:98) λ (cid:18) r + (cid:98) λ r + (cid:98) λ (cid:19) r + (cid:98) λ r + (cid:98) λ (cid:18) u b − Br + (cid:98) λ (cid:19) r + (cid:98) λ r + (cid:98) λ , u b ∈ (cid:2) x (cid:63) , (cid:98) b (cid:1) ,ρ g ρ b u b , u b ∈ (cid:2)(cid:98) b , + ∞ (cid:1) . (4.14)As an illustration, in Figure 1 we show the credible set which corresponds to the region delimited by its upper andlower boundaries. In this example, we considered r = 0 . , B = 0 . , ε = 0 . , α = 0 . , ρ g ρ b = 2 . • Step 3: solving the HJB equation in the general case r + b λ x ? b b Br + b λ b b ρ g ρ b b b u g = u b b L ( u b ) b U ? ( u b ) u b u g Figure 1: Credible set with one loan left.In the general case, when j > , we can reduce the number of variables and rewrite the diffusion equation (4.10) in anequivalent form r (cid:98) U j ( u b ) = sup ( θ,h ) ∈ (cid:98) C j (cid:110) (cid:98) U (cid:48) j ( u b ) (cid:0) ru b − Bk b + [ u b − θ ( u b − h )] (cid:98) λ k b j (cid:1) + (cid:0) (cid:98) U j − ( u b − h ) θ − (cid:98) U j ( u b ) (cid:1)(cid:98) λ k g j + Bk g (cid:111) , (4.15)where we recall that k b = { u b − θ ( u b − h ) < (cid:98) b j } , k g = { (cid:98) U j ( u b ) − θ (cid:98) U j − ( u b − h ) < (cid:98) b j } and the set of constraints is now given by (cid:98) C j := (cid:40) ( θ, h ) ∈ [0 , × R + , u b ≥ h + B ( j − r + (cid:98) λ SH j − (cid:41) . (4.16)When we proved that the lower boundary of the credible set is reachable we used contracts of maximum duration,which maintain the pool until the last default. This gives us the intuition that the longer the contract lasts, the smallerthe difference between the utilities of the banks will be. Therefore the upper boundary of the credible set, where thedifference between both utilities is maximal, should be reachable with contracts of minimum duration, which terminatethe contractual relationship immediately after the first default. In the model this means that θ is equal to zero andthe resulting HJB equation for the upper boundary has the same form as the one in the case with one loan left. Weexpect then that the solution of the diffusion equation will be the of the same form as (4.14). The object of the nextproposition is to prove our guess rigorously. We postpone the proof to Appendix D. Proposition 4.3.
For any j ≥ , the function (cid:98) U (cid:63)j defined by (cid:98) U (cid:63)j ( u b ) := (cid:18) ρ g ρ b (cid:19) r + (cid:98) λ SH jr + (cid:98) λ j (cid:18) u b − Bjr + (cid:98) λ SH j (cid:19) + Bjr + (cid:98) λ SHj , u b ∈ (cid:20) Bjr + (cid:98) λ SHj , x (cid:63)j (cid:19) ,ρ g ρ b (cid:98) b (cid:98) λ SH j − (cid:98) λ jr + (cid:98) λ SH j j (cid:18) r + (cid:98) λ SHj r + (cid:98) λ j (cid:19) r + (cid:98) λ jr + (cid:98) λ SH j (cid:18) u b − Bjr + (cid:98) λ SH j (cid:19) r + (cid:98) λ jr + (cid:98) λ SH j , u b ∈ (cid:2) x (cid:63)j , (cid:98) b j (cid:1) ,ρ g ρ b u b , u b ∈ (cid:2)(cid:98) b j , + ∞ (cid:1) , (4.17) where x (cid:63)j := (cid:18) ρ b ρ g (cid:19) r + (cid:98) λ SH jr + (cid:98) λ j (cid:98) b j r + (cid:98) λ j r + (cid:98) λ SH j + Bjr + (cid:98) λ SH j , is a solution of the HJB equation (4.9) . According to the maximisers in equation (4.15) we define the following controls δ j ( u b ) := { u b ≥ (cid:98) b j } u b ( r + (cid:98) λ j ) ρ b , θ j ( u b ) := 0 ,h ,b,j ( u b ) := u b − B ( j − r + (cid:98) λ SH j − , h ,b,j ( u b ) := B ( j − r + (cid:98) λ SH j − ,k b,j ( u b ) := j { u b < (cid:98) b j } , k g,j ( u b ) := j { (cid:98) U (cid:63)j ( u b ) < (cid:98) b j } . (4.18)14efore stating the verification result for the upper boundary, we make a comment about the domain of the functions (cid:98) U (cid:63)j . Rigorously speaking, it is possible for the utilities of the banks to be zero but this happens only at time τ when allthe pools are liquidated. The domain of (cid:98) U (cid:63)j is the set (cid:98) V j but in the proof of the verification theorem it will be implicitlyunderstood that (cid:98) U (cid:63)j (0) = 0 . In any case, we do not need the functions (cid:98) U (cid:63)j to be defined at zero because Itô’s formulawill be used on intervals which do not contain τ . Theorem 4.1.
Consider any starting time t ≥ . For any u b ≥ B ( I − N t ) r + (cid:98) λ SH I − Nt , let the process ( u bs ) s ∈ [ t,τ ] be the uniquesolution of the following SDE u bv = u b + (cid:90) vt (cid:16)(cid:0) r + λ k b,I − Ns s (cid:1) u bs − Bk b,I − N s ( u bs ) − ρ b δ I − N s ( u bs ) (cid:17) d s − (cid:90) vt u bs − d N s , v ∈ [ t, τ ] . (4.19) Then, under the contract Ψ (cid:63) := ( D (cid:63) , θ (cid:63) , h ,b,(cid:63) , h ,b,(cid:63) ) ∈ D × Θ × H defined for s ∈ [ t, τ ] by d D (cid:63)s := δ I − N s ( u bs )d s, θ (cid:63)s ≡ , h ,b,(cid:63)s := h ,b,I − N s ( u bs ) , h ,b,(cid:63)s := h ,b,I − N s ( u bs ) , the value function of the bad bank is U bt (Ψ (cid:63) ) = u b and the one of good bank is U gt (Ψ (cid:63) ) = (cid:98) U (cid:63)I − N t ( u b ) . Moreover, Ψ (cid:63) ∈ A b ( t, u b ) and for any other contract which belongs to A b ( t, u b ) , the value function of the good bank under such acontract is less or equal to (cid:98) U (cid:63)I − N t ( u b ) . In particular, this implies that (cid:98) U (cid:63)I − N t ( u b ) = (cid:98) U I − N t ( u b ) = U t ( u b ) . To conclude the section, we state that C j is indeed equal to the credible set with j loans left and therefore the functions (cid:98) U j and (cid:98) L j correspond to its upper and lower boundaries. Proposition 4.4.
For every t ≥ , C t = C t . In this section we study two kind of contracts that the investor can offer to the bank, the shutdown contract, whichcorresponds to a single contract designed to be accepted only by the good bank and the screening contract, correspondingto a menu of contracts, one for each type of agent, providing incentives to the bank to accept the contract designed forher true type.
In the so–called shutdown contract, the investor designs a contract Ψ g = ( k g , D g , θ g ) only for the good bank and makessure that the bad bank will not accept it. Under these conditions the utility of the investor at time t = 0 is v g , Shut0 (Ψ g ) = p g E P kg (cid:20) (cid:90) τ µ ( I − N s )d s − d D gs (cid:21) . (5.1)So the investor will offer a contract which maximises (5.1) subject to the constraints u g ( k g , θ g , D g ) ≥ R g , sup k ∈ K u b ( k, θ g , D g ) ≤ R b , u g ( k g , θ g , D g ) = sup k ∈ K u g ( k, θ g , D g ) . Recalling the dynamics (4.5)–(4.6), we can rewrite the investor’s maximisation problem as follows v Shut0 := sup ( θ g ,D g ) ∈A g Shut p g E P k(cid:63),g ( θg,Dg ) (cid:20) (cid:90) τ µ ( I − N s )d s − d D gs (cid:21) , where A g Shut := (cid:110) ( θ g , D g ) ∈ Θ × D : U b,c ( θ g , D g ) ≤ R b , U g ( θ g , D g ) ≥ R g (cid:111) . Remark 5.1.
We will use the notation U b,c ( θ g , D g ) for the value function that the bad bank gets if she does not revealher true type and accepts the contract designed for the good bank. We make a distinction between this process and U b ( θ b , D b ) , which corresponds to the value function that the bad bank obtains if she accepts the contract designed for herby the investor. We make the same distinction between the associated processes h ,b,c ( θ, D ) , h ,b,c ( θ, D ) and h ,b ( θ, D ) , h ,b ( θ, D ) .
15s in the previous section, we will define a simple set of contracts and consider the value functions of the agents asdiffussion processes controlled by ( D, θ, h ,g , h ,g , h ,b,c , h ,b,c ) . As explained before, by doing so we do not look at alarger class of "contracts".Define for any ( t, u g , u b,c ) ∈ [0 , + ∞ ) ×C t , (cid:98) A g ( t, u g , u b,c ) to be the set of Ψ g = ( D g , θ g , h ,g , h ,g , h ,b,c , h ,b,c ) ∈ D× Θ ×H such that (4.7) and (4.8) have at least one weak solution, which satisfy the first integrability condition in (3.3), and inaddition, for any s ∈ [ t, τ ] U gs − (Ψ g ) = h ,gs + h ,gs , U gs − (Ψ g ) − h ,gs ≥ B ( I − N s ) r + λ I − N s s , U gt (Ψ g ) = u g ,U b,cs − (Ψ g ) = h ,b,cs + h ,b,cs , U b,cs − (Ψ g ) − h ,b,cs ≥ B ( I − N s ) r + λ I − N s s , U b,ct (Ψ g ) = u b,c . We will also consider in the sequel the following standard control problem, for any ( u b,c , u g ) ∈ C (cid:98) v g ( u b,c , u g ) := sup Ψ g ∈ (cid:98) A g (0 ,u g ,u b,c ) p g E P k(cid:63),g (Ψ g ) (cid:20) (cid:90) τ µ ( I − N s )d s − d D gs (cid:21) . We abuse notations and also call elements of (cid:98) A g ( t, u g , u b,c ) contracts. In this section, we characterise the value function of the investor when he offers only shutdown contracts. We will startby computing the value function on the boundaries of the credible set, before explaining how it can be characterised bya specific HJB equation in the interior of the credible set, under reasonable assumptions.
Value function of the investor on the lower boundary
Recall the lower boundary with j loans left (cid:98) L j ( u b,c ) = u b,c , c ( j, ≤ u b,c ≤ C ( j ) ,ρ g ρ b u b,c − ( ρ g − ρ b ) ρ b C ( j ) , C ( j ) ≤ u b,c < ∞ . Consider any starting time t ≥ . For u b,c ∈ C t , we denote by V L ,g ( u b,c ) the value function of the investor on the lowerboundary, that is V L ,gt ( u b,c ) := ess sup Ψ g ∈ (cid:98) A g ( t, (cid:98) L I − Nt ( u b,c ) ,u b,c ) E P k(cid:63),g (Ψ g ) (cid:20) (cid:90) τt (cid:0) µ ( I − N s )d s − d D gs (cid:1)(cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . (5.2)The following two propositions are proved in Appendix E and give explicitly the value of V L ,gt ( u b,c ) . Proposition 5.1.
For every u b,c ∈ C t , if u b,c ≥ C ( I − N t ) then the value function of the investor on the lower boundaryis given by V L ,gt ( u b,c ) = I − (cid:88) i = N t µ ( I − i ) (cid:98) λ SH I − i − (cid:18) u b,c − C ( I − N t ) ρ b (cid:19) . Proposition 5.2.
Fix some t ≥ . For every u b,c ∈ C t , with c ( I − N t , ≤ u b,c < C ( I − N t ) , let ν ( u b,c ) be the uniquesolution of the following equation in ν (cid:18) B ( I − N t ) r + (cid:98) λ SH I − N t − u b,c (cid:19) + I − (cid:88) i = N t +1 (cid:90) ∞ s i ( ν ) (cid:18) B ( I − i ) r + (cid:98) λ SH I − i e − rx (cid:19) f τ i ( x )d x = 0 , where f τ i is the density of the law of τ i under P k SH and where s i ( ν ) := , ν ≤ µ ( r + (cid:98) λ SH I − i ) B (cid:98) λ SH I − i , r ln (cid:18) νB (cid:98) λ SH I − i µ ( r + (cid:98) λ SH I − i ) (cid:19) , ν ≥ µ ( r + (cid:98) λ SH I − i ) B (cid:98) λ SH I − i . hen the value function of the investor in the lower boundary is given by V L ,gt ( u b,c ) = µ ( I − N t ) (cid:98) λ SH I − N t + I − (cid:88) i = N t +1 (cid:90) ∞ s i ( ν ( u b,c )) µ ( I − i ) (cid:98) λ SH I − i f τ i ( x )d x. Remark 5.2.
Observe that the function V L ,gt computed in Propositions 5.1 and 5.2 depends on t only through thequantity I − N t . Define, for any j = 1 , . . . , J the map (cid:98) V L ,gj ( u b,c ) := j (cid:88) i =1 µi (cid:98) λ SH i − (cid:18) u b,c − C ( j ) ρ b (cid:19) , u b,c ≥ C ( j ) ,µj (cid:98) λ SH j + j − (cid:88) i =1 (cid:90) ∞ s I − j ( ν ( u b,c )) µi (cid:98) λ SH i f τ I − i ( x )d x, u b,c ∈ ( c ( j, , C ( j )) . We have therefore, that V L ,gt ( u b,c ) = (cid:98) V L ,gI − N t ( u b,c ) . Value function of the investor on the upper boundary
The next proposition states that the upper boundaryof the credible set is absorbing in the following sense: if under any contract the pair of value functions of the banksreaches the upper boundary at some time, the pair will stay on the upper boundary until the pool is liquidated.
Proposition 5.3.
Fix a triplet ( t, u g , u b,c ) ∈ [0 , + ∞ ) × C t such that u g = (cid:98) U I − N t ( u b,c ) . For any contract Ψ g =( D g , θ g , h ,g , h ,g , h ,b,c , h ,b,c ) ∈ (cid:98) A g ( t, u g , u b,c ) , we have U gs (Ψ g ) = (cid:98) U I − N s ( U b,cs (Ψ g )) for every s ∈ [ t, τ ) . The next proposition states an important property satisfied by the contracts which make the continuation utilities ofthe banks lie in the upper boundary of the credible set.
Proposition 5.4.
Fix a triplet ( t, u g , u b,c ) ∈ [0 , + ∞ ) × C t such that u g = (cid:98) U I − N t ( u b,c ) . For any contract Ψ g =( D g , θ g , h ,g , h ,g , h ,b,c , h ,b,c ) ∈ (cid:98) A g ( t, u g , u b,c ) , we have ( i ) θ gs = 0 for every s ∈ [ t, τ ) such that U b,cs (Ψ g ) < b s . ( ii ) If U b,cs (Ψ g ) ≥ b s for some s ∈ [ t, τ ) then k (cid:63),b,cs (Ψ g ) = 0 and U b,cs (Ψ g ) ≥ b s for every s ∈ [ s , τ ) . We are now ready to give the value function of the investor on the upper boundary of the credible set. In the last regionof the upper boundary, in which both the good and the bad agent are monitoring all the loans, it coincides with thevalue function of the sub–problem studied in[50], denoted by v bj . For the sake of presentation, we recall the results of[50] in Appendix E.1. Proposition 5.5.
Under Assumption 2.1, we have that for any t ≥ and any u b,c ∈ (cid:98) V I − N t , the value function of theinvestor on the upper boundary, defined by V U ,gt ( u b,c ) := ess sup Ψ g ∈ (cid:98) A g ( t, (cid:98) U I − Nt ( u b,c ) ,u b,c ) E P k(cid:63),g (Ψ g ) (cid:20) (cid:90) τt (cid:0) µ ( I − N s )d s − d D gs (cid:1)(cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) , (5.3) verifies V U ,gt ( u b,c ) = (cid:98) V U ,gI − N t ( u b,c ) , where for any j = 1 , · · · , I (cid:98) V U ,gj ( u b,c ) := µj (cid:98) λ SH j + (cid:98) C j (cid:18) u b,c − Bjr + (cid:98) λ SH j (cid:19) (cid:98) λ SH jr + (cid:98) λ SH j , u b,c < x (cid:63)j ,µj (cid:98) λ j + (cid:18) v bj ( (cid:98) b j ) − µj (cid:98) λ j (cid:19)(cid:18)(cid:98) b j r + (cid:98) λ j r + (cid:98) λ SH j (cid:19) − (cid:98) λ jr + (cid:98) λ SH j (cid:18) u b,c − Bjr + (cid:98) λ SH j (cid:19) (cid:98) λ jr + (cid:98) λ SH j , x (cid:63)j ≤ u b,c < (cid:98) b j ,v bj ( u b,c ) , u b,c ≥ (cid:98) b j , with v bj given by (E.1) and (cid:98) C j := (cid:18) µj (cid:98) λ j − µj (cid:98) λ SH j + (cid:18) ρ b ρ g (cid:19) (cid:98) λ jr + (cid:98) λ j (cid:18) v bj ( (cid:98) b j ) − µj (cid:98) λ j (cid:19)(cid:19)(cid:18) ρ b ρ g (cid:19) − (cid:98) λ SH jr + (cid:98) λ j (cid:18)(cid:98) b j ( r + (cid:98) λ j ) r + (cid:98) λ SH j (cid:19) − (cid:98) λ SH jr + (cid:98) λ SH j . The authors only look at the contracts for which the agent performs the maximum effort, that is, monitors all the loans at every time. alue function of the investor in the credible set We define, for any t ≥ and any ( u b,c , u g ) ∈ (cid:98) C I − N t , the valuefunction of the investor in the credible set by V gt ( u b,c , u g ) := ess sup Ψ g ∈ (cid:98) A g ( t,u g ,u b,c ) E P k(cid:63),g (Ψ g ) (cid:20) (cid:90) τt (cid:0) µ ( I − N s ) ds − d D gs (cid:1)(cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . (5.4)The system of HJB equations associated to this control problem is given by (cid:98) V g ≡ , and for any ≤ j ≤ I , on (cid:98) V j × (cid:98) V j max sup C j ∂ u b,c (cid:98) V gj (cid:0) ru b,c − Bk b,c + ( h ,b,c + (1 − θ ) h ,b,c ) (cid:98) λ k b,c j (cid:1) + ∂ u g (cid:98) V gj (cid:0) ru g − Bk g + ( h ,g + (1 − θ ) h ,g ) (cid:98) λ k g j (cid:1) + (cid:0) (cid:98) V gj − ( u b,c − h ,b,c , u g − h ,g ) − (cid:98) V gj ( u b,c , u g ) (cid:1)(cid:98) λ k g j − (cid:98) V gj − ( u b,c − h ,b,c , u g − h ,g )(1 − θ ) (cid:98) λ k g j + µj , − ρ b ∂ u b,c (cid:98) V gj − ρ g ∂ u g (cid:98) V gj − = 0 , (5.5)where we defined k b,c = j · { h ,b,c +(1 − θ ) h ,b,c < (cid:98) b j } , k g = j · { h ,g +(1 − θ ) h ,g < (cid:98) b j } and the set of constraints C j = (cid:26) ( θ, h ,b,c , h ,b,c , h ,g , h ,g ) ∈ R : θ ∈ [0 , , u g = h ,g + h ,g , u b,c = h ,b,c + h ,b,c , h ,g ; h ,b,c ≥ B ( j − r + (cid:98) λ SH j − (cid:27) . The boundary conditions of (5.5) are given, for every u b,c ∈ (cid:98) V j , by (cid:98) V gj ( u b,c , (cid:98) U j ( u b,c )) = (cid:98) V U ,gj ( u b,c ) , (cid:98) V gj ( u b,c , (cid:98) L j ( u b,c )) = (cid:98) V L ,gj ( u b,c ) . (5.6)The last step is now to use classical arguments to prove that (cid:98) V gj is a viscosity solution of the above PDE for every j = 1 , . . . , I and that the functions are sufficiently smooth (at least weakly dfferentiable) in order to obtain the optimalcontract as the maximisers above. This program can in principle be carried out using standard arguments in viscositytheory of Hamilton–Jacobi equations. However, given the length of the paper, we believe that it would not serve aspecific purpose and decided to just describe the main steps that lead to this result. We list them below ( i ) For j = 1 , we can use the abstract results of [20] to prove that (5.4) coincides with the strong formulation of itself.This fact allows to prove directly that the value function (cid:98) V g is concave and therefore differentiable almost everywhere. ( ii ) For j > , let us define the penalised Hamiltonians for the diffusion equation, with j loans left. Given (cid:98) V gj − , definefor instance H nj ( u b,c , u g , v, p b,c , p g ) = sup C j p b,c (cid:0) ru b,c − Bk b,c + ( h ,b,c + (1 − θ ) h ,b,c ) λ k b,c j (cid:1) + p g (cid:0) ru g − Bk g + ( h ,g + (1 − θ ) h ,g ) (cid:98) λ k g j (cid:1) + (cid:0) (cid:98) V gj − ( u b,c − h ,b,c , u g − h ,g ) − v ( u b,c , u g ) (cid:1)(cid:98) λ k g j − (cid:98) V gj − ( u b,c − h ,b,c , u g − h ,g )(1 − θ ) (cid:98) λ k g j + µj + n ( − ρ b p b,c − ρ g p g − + . (5.7)Let v nj be the value function of the penalised version of our problem, in which payments are absolutely continuouswith bounded density. Then it can be argued as in [21] that v nj is a viscosity solution to H nj ( u, v, p ) = 0 , withappropriate credible set and boundary conditions. ( iii ) Note that H nj is convex in p , as a supremum of linear functionals and composition of convex functions. Moreover,for any R < + ∞ we have H nj ( u, v , p ) − H nj ( u, v , p ) ≥ − (cid:98) λ SH j ( v − v ) , ∀ ( u, p ) and R ≥ v ≥ v ≥ − R.H nj is also locally Lipschitz and H nj ( u, v, p ) −→ ∞ as | p | → ∞ for any u > Bjr + (cid:98) λ SH j . Finally, noticing that interiormaximisers take place in the interior of the credible set (to be more precise, boundary maximisers correspond tocontracts leading the agents to the absorbing boundaries of the credible set) and by using the envelope theorem, wecan show that H nj is actually strictly convex on the interior of the credible set and therefore satisfies ∀ R > , ∃ α R > , (cid:18) ∂H nj ∂p ( u, v, p ) − ∂H nj ∂p ( u, v, q ) , p − q (cid:19) ≥ α R | p − q | , | p | , | q | , | u | ≤ R, for any u. By Theorem 3.3 in Lions [37], it follows then that v nj ∈ W , ∞ loc and v nj is SSH (semi–super harmonic).18 iv ) Arguing again as in [21], it can be proved the sequence v nj converges to the value function (cid:98) V gj of our problem, whichis therefore SSH and a viscosity solution to (5.5) with boundary condition (5.6).Finally, since (cid:98) V gj is differentiable almost everywhere, we can define the optimal contract through the maximisers in theHamiltonian (5.5). Then, using the classical result (see for instance [32] for related arguments) that the domain in whichthe diffusion equation is not saturated is bounded, it follows that the optimal controls ( h ,g,(cid:63) , h ,g,(cid:63) , h ,b,c,(cid:63) , h ,b,c,(cid:63) ) arebounded and the corresponding SDEs admit weak solutions d U (cid:63),gs = (cid:0) rU (cid:63),gs − Bk (cid:63),gs ( U (cid:63),b,cs , U (cid:63),gs ) − ρ g δ (cid:63),gI − N s ( U (cid:63),b,cs , U (cid:63),gs ) (cid:1) d s − h (cid:63), ,gI − N s ( U (cid:63),b,cs , U (cid:63),gs ) (cid:0) d N s − λ k (cid:63),g (( U (cid:63),b,c ,U (cid:63),g )) s d s (cid:1) − h (cid:63), ,gI − N s ( U (cid:63),b,cs , U (cid:63),gs ) (cid:0) d H s − (1 − θ (cid:63),gI − N s ( U (cid:63),b,cs , U (cid:63),gs )) λ k (cid:63),g (( U (cid:63),b,c ,U (cid:63),g )) s d s (cid:1) , d U (cid:63),b,cs = (cid:0) rU (cid:63),b,cs − Bk (cid:63),b,cs ( U (cid:63),b,cs , U (cid:63),gs ) − ρ b δ (cid:63),gI − N s ( U (cid:63),b,cs , U (cid:63),gs ) (cid:1) d s − h (cid:63), ,b,cI − N s ( U (cid:63),b,cs , U (cid:63),gs ) (cid:0) d N s − λ k (cid:63),b,c (( U (cid:63),b,c ,U (cid:63),g )) s d s (cid:1) − h (cid:63), ,b,cI − N s ( U (cid:63),b,cs , U (cid:63),gs ) (cid:0) d H s − (1 − θ (cid:63),gs ( U (cid:63),b,cs , U (cid:63),gs )) λ k (cid:63),b,c (( U (cid:63),b,c ,U (cid:63),g )) s d s (cid:1) . Indeed, this can be proved by noticing that in–between two jump times, the above are actually first–order ODEs,which admit weak solutions in appropriately exponentially weighted L space (to make sure that bounded functionsare integrable over the credible set), thanks to Carathéodory’s theorem for ODEs. Thus, we have the equivalence v Shut0 = sup u b,c ≤ R b , u g ≥ R g (cid:98) v g ( u b,c , u g ) = sup u b,c ≤ R b , u g ≥ R g p g (cid:98) V gI ( u b,c , u g ) . Recall that in the screening contract the investor designs a menu of contracts, one for each agent, and his expectedutility is given by v (cid:0) (Ψ i ) i ∈{ g,b } (cid:1) = (cid:88) i ∈{ g,b } p i E P ki (cid:20) (cid:90) τ (cid:0) I − N s (cid:1) µ d s − d D is (cid:21) . (5.8)In this case, we will have to keep track of the value functions of both banks, when they choose the contract designedfor them, as well as when they do not truthfully reveal their type. We will denote by v the maximal utility that theinvestor can get out of the screening contract. v := sup ( θ g ,θ b ,D g ,D b ) ∈A Scr p g E P k(cid:63),g ( θg,Dg ) (cid:20) (cid:90) τ µ ( I − N s )d s − d D gs (cid:21) + p b E P k(cid:63),b ( θb,Db ) (cid:20) (cid:90) τ µ ( I − N s )d s − d D bs (cid:21) , where A Scr := (cid:110) ( θ g , θ b , D g , D b ) ∈ Θ × D : U i ( θ i , D i ) ≥ R i , U j ( θ j , D j ) ≥ U j,c ( θ i , D i ) , ( i, j ) ∈ { g, b } , i (cid:54) = j (cid:111) . Different from the study of the shutdown contract, where the investor contracts only the good bank, in order to obtainthe optimal screening contract we need to characterise also the value function of the investor when he contracts thebad bank. We will therefore follow Section 5.1.1, but by replacing the good bank by the bad bank. Hence, we definesimilarly, for any ( t, u b , u g,c ) ∈ [0 , + ∞ ) × C t the set (cid:98) A b ( t, u g,c , u b ) . We also introduce the following stochastic controlproblem for any ( u b , u g,c ) ∈ C I (cid:98) v b0 ( u b , u g,c ) := sup Ψ b ∈ (cid:98) A b (0 ,u g,c ,u b ) p b E P k(cid:63),b (Ψ b ) (cid:20) (cid:90) τ µ ( I − N s )d s − d D bs (cid:21) . The aim of the next sections is to compute the function (cid:98) v b0 ( u g,c , u b ) , representing the utility of the investor when hiringthe bad bank. We start by studying it on the boundary of the credible set. We denote by V L ,b ( u g,c ) the value function of the investor in the lower boundary, when hiring the bad bank, definedby V L ,bt ( u b ) := ess sup Ψ b ∈ (cid:98) A b ( t, (cid:98) L I − Nt ( u b ) ,u b ) E P k(cid:63),b (Ψ b ) (cid:20) (cid:90) τt µ ( I − N s )d s − d D bs (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . (5.9)The first result is that the value function of the investor on the lower boundary of the credible set is the same whenhiring either the bad or the good bank. This is mainly due to the fact that both banks shirk on the lower boundary.19 roposition 5.6. For every u b ∈ C I − N t , we have V L ,bt ( u b ) = V L ,gt ( u b ) . Let us now consider the upper boundary. We denote by V U ,b ( u b ) the value function of the investor on the upperboundary when hiring the bad agent. V U ,bt ( u b ) := ess sup Ψ b ∈ (cid:98) A b ( t, (cid:98) U I − Nt ( u b ) ,u b ) E P k(cid:63),b (Ψ b ) (cid:20) (cid:90) τt µ ( I − N s )d s − d D bs (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . (5.10)We have the following result. Proposition 5.7.
Under Assumption 2.1, for any t ≥ and any u b ∈ (cid:98) V I − N t , we have that V U ,bt ( u b ) = (cid:98) V U ,bI − N t ( u b ) ,where for any j = 1 , · · · , I (cid:98) V U ,bj ( u b ) := µj (cid:98) λ SH j + ˜ C j (cid:18) u b − Bjr + (cid:98) λ SH j (cid:19) (cid:98) λ SH jr + (cid:98) λ SH j , u b < (cid:98) b j ,v bj ( u b ) , u b ≥ (cid:98) b j , with v bj given by (E.1) and ˜ C j = (cid:18) v bj ( (cid:98) b j ) − µj (cid:98) λ SH j (cid:19)(cid:18)(cid:98) b j ( r + (cid:98) λ j ) r + (cid:98) λ SH j (cid:19) − (cid:98) λ SH jr + (cid:98) λ SH j . We define, for any t ≥ and any ( u b , u g,c ) ∈ (cid:98) C I − N t , the value function of the investor in the credible set when hiringthe bad bank by V bt ( u b , u g,c ) := ess sup Ψ b ∈ (cid:98) A b ( t,u g,c ,u b ) E P k(cid:63),b (Ψ b ) (cid:20) (cid:90) τt (cid:0) µ ( I − N s ) ds − dD bs (cid:1)(cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . (5.11)The system of HJB equations associated to this control problem is given by (cid:98) V b ≡ , and for any ≤ j ≤ I max sup C j ∂ u b (cid:98) V bj (cid:0) ru b − Bk b + ( h ,b + (1 − θ ) h ,b ) (cid:98) λ k b j (cid:1) + ∂ u g,c (cid:98) V bj (cid:0) ru g,c − Bk g,c + ( h ,g,c + (1 − θ ) h ,g,c ) (cid:98) λ k g,c j (cid:1) + (cid:0) (cid:98) V bj − ( u b − h ,b , u g,c − h ,g,c ) − (cid:98) V bj (cid:1)(cid:98) λ k b j − (cid:98) V bj − (cid:0) u b − h ,b , u g,c − h ,g,c (cid:1) (1 − θ ) (cid:98) λ k b j + µj , − ρ b ∂ u b (cid:98) V bj − ρ g ∂ u g,c (cid:98) V bj − = 0 . (5.12)With k b = j · { h ,b +(1 − θ ) h ,b < (cid:98) b j } , k g,c = j · { h ,g,c +(1 − θ ) h ,g,c < (cid:98) b j } and the same set of constraints C j as in the system ofHJB equations associated to the functions (cid:98) V gj ( u b,c , u g ) . The boundary conditions of (5.12) are given, for every u b ∈ (cid:98) V j by (cid:98) V bj ( u b , (cid:98) U j ( u b )) = (cid:98) V U ,bj ( u b ) , (cid:98) V bj ( u b , (cid:98) L j ( u b )) = (cid:98) V L ,gj ( u b ) . Similarly to the shutdown contract, we can argue that the functions (cid:99) V bj are viscosity solutions to the system (5.12),differentiable almost everywhere and the maximizers in the Hamiltonian define an admissible contract. This impliesthe equivalence v = sup { R b ∨ u b,c ≤ u b ,R g ∨ u g,c ≤ u g } (cid:98) v g ( u b,c , u g ) + (cid:98) v b0 ( u b , u g,c ) = sup { R b ∨ u b,c ≤ u b ,R g ∨ u g,c ≤ u g } p g (cid:98) V gI ( u b,c , u g ) + p b (cid:98) V bI ( u b , u g,c ) . In this section we describe the optimal contracts for the investor when he designs a contract for the good or the badbank. We explain in detail the optimal contracts on the boundaries of the credible set, which can be obtained explicitlyfrom the value function of the investor. In the interior of the credible set, we discuss the properties we expect theoptimal contracts to have given the verification results described in the previous sections.20 .3.1 Optimal contracts on the boundaries of the credible set
We start with the upper boundary of the credible set. The following result is a direct consequence of the proofs ofProposition 5.5 and 5.7, and the optimal contract for the pure moral hazard case. In the last part of the upper boundary,in which both agents monitor all the loans, the optimal contract coincides with the one of the subproblem studied in[50]. Their main results can be found in Appendix E.1.
Proposition 5.8.
Under Assumption 2.1, consider for any t ≥ and ( u b , U t ( u b )) ∈ C t the process ( u bs ) s ≥ t as thesolution of the following SDE on [ t, τ )d u bs = (cid:0) ( ru bs − Bk b,(cid:63)s + λ k b,(cid:63) s ( h ,b,(cid:63)s + (1 − θ (cid:63)s ) h ,b,(cid:63)s ) (cid:1) d s − ρ b d D (cid:63)s − h ,b,(cid:63)s d N s − h ,b,(cid:63)s d H s , (5.13) with initial value u b at t , and with D (cid:63)s := { s = t } ( u b − γ bI − N t ) + ρ b + (cid:90) st δ I − N r ( u br )d r, θ (cid:63)s := θ I − N s ( u bs ) , h ,b,(cid:63)s := h ,b,I − N s ( u bs ) , k b,(cid:63)s := k b,j ( u bs ) , for s ∈ [ t, τ ) and j = 1 , . . . , I , where γ bj is given by (E.2) and δ j ( u ) := { u = γ bj } (cid:98) λ j (cid:98) b j + rγ bj ρ i , θ j ( u ) := { u ∈ [ (cid:98) b j , (cid:98) b j − + (cid:98) b j ) } u − (cid:98) b j (cid:98) b j − + { u ∈ [ (cid:98) b j + (cid:98) b j − ,γ bj ) } ,h ,b,j ( u ) := { u ∈ [ c ( j, , (cid:98) b j ) } u + { u ∈ [ (cid:98) b j , (cid:98) b j − + (cid:98) b j ) } ( u − (cid:98) b j − ) + { u ∈ [ (cid:98) b j + (cid:98) b j − ,γ bj ) } (cid:98) b j , k b,j ( u ) = j { θ j ( u ) h ,b,j ( u )+(1 − θ j ( u )) u< (cid:98) b j } . Then, the contract Ψ (cid:63) = ( D (cid:63) , θ (cid:63) , h ,b,(cid:63) , h ,b,(cid:63) ) is the unique solution of problems (5.3) and (5.10) . Let us comment the optimal contract for the investor on the upper boundary of the credible set. It is the same if hedesigns a contract for the good or the bad bank. The state process ( u bs ) s ≥ t defined by (5.13) corresponds to the valuefunction of the bad bank under the optimal contract. The optimal contract offers no payments to the banks when u bs is smaller than γ bI − N s . In this case the continuation utility of the bad bank is an increasing process and eventuallyreaches the value γ bI − N s , if no default happens in the meantime. Payments are postponed until this moment. If theinitial value for the bad agent u b is greater than γ bI − N t , a lump-sum payment is made at t − in order to have u bt = γ bI − N t .When u bs = γ bI − N s , the banks receive constant payments which keep the value function of the bad bank constant at thislevel. Concerning the liquidation of the project, if, at the default time τ j , it holds that u bτ j < (cid:98) b j , then the project isliquidated. In case of u bτ j ∈ [ (cid:98) b j + (cid:98) b j − , γ bj ) , the project will continue with probability θ j ∈ (0 , which will be closerto one as u bτ j gets closer to γ bj . If u bτ j ≥ γ bj , the project will be maintained. Finally, the bad bank will monitor all theloans only when her value function is greater than (cid:98) b I − N s , whereas the good bank will monitor when the value of thebad bank is greater than x (cid:63)I − N s . Figure 2 depicts the optimal contract of the investor on the upper boundary of thecredible set, denoting (cid:98) B j := (cid:98) b j + (cid:98) b j − . u bs c ( I − N s , x ?I − N s b b I − N s b B I − N s γ bI − N s k gs = I − N s k gs = 0 k gs = 0 k gs = 0 k gs = 0 k bs = I − N s k bs = I − N s k bs = 0 k bs = 0 k bs = 0 θ s = 0 θ s = 0 θ s ∈ (0 , θ s = 1 θ s = 1d D s = 0 d D s = 0 d D s = 0 d D s = 0 d D s > Figure 2: Optimal contract on the upper boundary.For the lower boundary of the credible set, we have the following result.
Proposition 5.9.
Under Assumption 2.1, consider for any t ≥ and ( u b , L t ( u b )) ∈ C t the process ( u bs ) s ≥ t as thesolution of the following SDE on [ t, τ )d u bs = (cid:0) ( ru bs − Bk b,(cid:63)s + λ k b,(cid:63) s ( h ,b,(cid:63)s + (1 − θ (cid:63)s ) h ,b,(cid:63)s ) (cid:1) d s − ρ b d D (cid:63)s − h ,b,(cid:63)s d N s − h ,b,(cid:63)s d H s , (5.14)21 ith initial value u b at t , and with D (cid:63)s := { s = t } ( u b − C ( I − N s )) + ρ b , θ (cid:63)s := { u bs ≥ C ( I − N s ) } ,h ,b,(cid:63)s := u bs − C ( I − N s − { u bs ≥ C ( I − N s ) } , h ,b,(cid:63)s := C ( I − N s − { u bs ≥ C ( I − N s ) } , k b,(cid:63)s = ( I − N s ) { h ,b,(cid:63)s +(1 − θ (cid:63)s ) h ,b,(cid:63)s Figure 3: Optimal contract on the lower boundary.
Figure 4 represents the optimal contracts on the boundaries of the credible set as well as the movements of the valuesof the banks along these curves. The green zone corresponds to the region where the contract offers payments to theagents and the project is maintained if there is a default. The red zone corresponds to the region where there are nopayments and the project is liquidated immediately after a default. Intermediate situations correspond to the yellowzone. We remark that the banks are paid only on the green zone.
Bjr + b λ jj x ?j b b j C ( j ) γ jBjr + b λ jj b b jρ g ρ b b b j C ( j ) u g = u b b L j ( u b ) b U ?j ( u b ) u b u g Figure 4: Optimal contract on the boundaries of the credible set.Let us now consider the whole credible set and explain how the green and red zones on the boundaries propagatetowards the interior region, given that the optimal contracts for problems (5.4) and (5.11) correspond to the maximisersin the Hamiltonian of the systems (5.5) and (5.12). Recall that payments only take place when the value function of the22nvestor saturates the gradient constraint. Therefore, if at some point of the credible set the banks are paid, this willalso be the case under movements in the direction ( ρ b , ρ g ) . The interpretation of this property is that the green region,where the banks are paid and the project is maintained after a default, is formed by the points where the banks havea good performance and they are rewarded. A movement in the direction ( ρ b , ρ g ) correspond to a better performanceof both banks, so it seems unnatural to deprive them of the reward. We can do the opposite interpretation for the redregion, consisting of the points where the banks receive no payments and the project is liquidated after a default. Inconsequence, under the optimal contracts, it is possible to identify red and green areas in the credible set, where thecharacteristics described in the boundaries will remain, and that will be delimited by some curves similar to those shownin Figure 5 below. Mathematically, these curves are delimiting the region where the gradient constraint is saturated. Bjr + b λ jj x ?j b b j C ( j ) γ jBjr + b λ jj b b jρ g ρ b b b j C ( j ) u g = u b b L j ( u b ) b U ?j ( u b ) u b u g Figure 5: Optimal contract on the credible set.
Any real–world application of our model requires to discuss the practical implementability of the contract. Fortunatelyfor us, the form of the menus of contracts we obtained is completely similar to the one obtained in [50, 51], in the sensethat all rely on a probation zone, where stochastic liquidation may occur, and a zone of good performance, where theliquidation never occurs. The only difference is of course that in [50, 51] these zones are simply intervals, while they aremore complex regions of the plan in our case, since we have to keep track of both the continuation and the temptationvalues of the Agent. Nonetheless, the practical implementation proposed by Pagès [50, Proposition 7] can readily beadapted to our context. Given the length of the paper, we leave the exact detail to the reader, and simply recall howthe implementation works.First, a natural way of implementing the contract is to replicate dynamically both the continuation and the temptationvalues of the Agent by use of two cash reserve accounts. The accounts should be managed by an independent trust, andactually serves to both provide protection to the investors, and to manage exactly the performance–based compensationscheme described in the optimal contract. The current balances reveal outright performance of both type of banks, andcan be used to determine the amount and timing of fees that are released. Then, the implementation basically takesthe form of a whole loan sale with monitoring retained. The reserve accounts then offer protection in the form of ABScredit default swaps (ABCDS), and serve as instruments to tie the amount and timing of compensation to performance(meaning that payments are made from the cash reserve only when the continuation and temptation values of theAgent are in the domain where the gradient constraint is satisfied). The reserve account reveals the level of underlyingperformance, which reduces the rent of the monitoring bank and allows it to retain risk at a lower cost than if it werefunded with deposits.
In a standard Principal–Agent problem, it is assumed that the Agent possesses a minimum level of utility that must beprovided by the Principal in order to make him accept the contract. This reservation value represents the utility that23he Agent would obtain if the contract offered by the Principal was not sufficiently attractive and he made use of anoutside option (see Condition (2.6) in Section 2.3).In this section we provide an endogenous characterisation of the reservation utilities of the banks by assuming that ifthey do not enter in a contractual relationship with the investor, they can manage the project by themselves. Whenthe outside option of a bank is to manage the pool of loans on its own, we can find the explicit value of its reservationutility. Moreover, we outline an extension of our model to the case in which the bank can break the contract at any timeif it can do better by itself. Different from the full-commitment problem studied in the previous sections, the ability ofthe bank to break the contract makes the investor offer only the so called renegotiation-proof contracts, which keep theutility of bank above a dynamic reservation utility until the end of the contract.If the bank of type ρ i manages the project, it receives the cash flows from the loans and does not face the threat ofliquidating the whole pool when one of the loans defaults. Consequently, its reservation utility R i is given by thefollowing expression R i = sup k ∈ K E P k (cid:20) (cid:90) τ I e − rs ( ρ i µ ( I − N s ) + Bk s )d s (cid:21) . (6.1)The value of R i can be obtained as an application of the results from the previous sections, since (6.1) corresponds tothe utility of the bank under a contract with no liquidation at all, θ ≡ , and with absolutely continuous payments, d D t = µ ( I − N t )d t . Its explicit value is provided in Proposition 6.1. Proposition 6.1.
Define the recursive sequence of numbers (cid:98) R ij = max (cid:26) ρ i µjr + (cid:98) λ j − r (cid:98) R ij − r + (cid:98) λ j , ρ i µj + jBr + (cid:98) λ SHj − r (cid:98) R ij − r + (cid:98) λ SH j (cid:27) , j ∈ { , . . . , I } , with (cid:98) R i = 0 . The endogenous reservation utility of the bank of type ρ i is given by R i = (cid:98) R iI . Moreover, the optimalaction in Problem (6.1) is constant in every interval ( τ I − j , τ I − j +1 ) and it is equal to k (cid:63),i ≡ if the maximum in thedefinition of (cid:98) R ij is attained at the first term, and to k (cid:63),i ≡ j if the maximum is attained at the last term. Suppose that the bank of type ρ i can decide at any time to break the contract with the investors and manage the loansby itself. By doing so, the bank’s utility at time t would be R it := ess sup k ∈ K E P k (cid:20) (cid:90) τ I t e − r ( s − t ) ( ρ i µ ( I − N s ) + Bk s )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . Notice that the previous expression depends on t only through the number of loans left at the time. It is straightforwardthen that R it = (cid:98) R iI − N t , for every t ∈ [0 , τ I ] .In this setting, a shutdown contract ( D, θ ) is one which is never broken by the bank of type ρ g and is rejected bythe bank of type ρ b , who prefers to run the project on its own. That is, U gt ( D, θ ) ≥ R gt for every t ∈ (0 , τ ) and U b ( D, θ ) < R b . To find the optimal shutdown contract, we need to characterize first the new credible set which includesadditional state constraints for the good bank. Let us mention immediately that the right of the bank to break thecontract generates differences between the credible sets associated to the shutdown and the screening problem, whichare no longer equal.Define the renegotiation–proof feasible set for the good bank with j loans left (cid:101) V gj = (cid:98) V j ∩ [ (cid:98) R gj , ∞ ) , j = 1 , . . . , I. Definition 6.1.
For any time t ≥ , we define the shutdown renegotiation–proof credible set (cid:101) C t as the set of ( u b , u g ) ∈ (cid:98) V I − N t × (cid:101) V gI − N t such that there exists an admissible contract ( θ, D ) ∈ Θ × D satisfying U bt ( θ, D ) = u b , U gt ( θ, D ) = u g and ( U bs ( θ, D ) , U gs ( θ, D )) ∈ (cid:98) V I − N s × (cid:101) V gI − N s for every s ∈ [ t, τ ) , P − a.s. Given a starting time t ≥ and u b ∈ (cid:98) V I − N t , define the set of contracts which are not broken by the good bank andunder which the value function of the bad bank at time t is equal to u b , A SH ,b ( t, u b ) = (cid:8) ( θ, D ) ∈ Θ × D : U bt ( θ, D ) = u b , U gs ( θ, D ) ≥ R gs , for every s ∈ [ t, τ ) (cid:9) .
24e denote by U SH t ( u b ) the largest value U gt ( θ, D ) that the good bank can obtain from all the contracts ( θ, D ) ∈A SH ,b ( t, u b ) and by L SHt ( u b ) the lowest one. Again, these sets can be proved to depend on t only through the value of I − N t so defining U I − N t ( u b ) := U SHt ( u b ) and L I − N t ( u b ) := L SHt ( u b ) we finally have C j := (cid:8) ( u b , u g ) ∈ (cid:98) V j × (cid:101) V gj : L j ( u b ) ≤ u g ≤ U j ( u b ) (cid:9) . As depicted in Figure 4, the upper boundary in the problem with full commitment is absorbing and it generates amovement of the utilities of the banks in the direction ( ρ b , ρ g ) . We conclude that the upper boundary in this extensionis the same as before and it is given by U I − N t ( u b ) = (cid:98) U I − N t ( u b ) , for every u b such that (cid:98) U I − N t ( u b ) ≥ (cid:98) R gI − N t . On the other hand, since the former lower boundary (cid:98) L I − N t ( u b ) generates a movement in the direction ( − ρ b , − ρ g ) , itcannot be used to obtain L I − N t ( u b ) which is the solution to the following control problem L SH t ( u b ) = ess inf ( k, Ψ) ∈ K ×A SH ,b ( t,u b ) E P k (Ψ) (cid:20) (cid:90) τt e − r ( s − t ) ( ρ g d D s + Bk s d s ) (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) , subject to the dynamics, for t ∈ [ r, τ ] U br (Ψ g ) = u b + (cid:90) rt (cid:0) ru bs − Bk (cid:63),bs (Ψ) + h ,bs λ k (cid:63),b s + h ,bs (1 − θ s ) λ k (cid:63),b s (cid:1) d s − ρ b (cid:90) rt d D s − (cid:90) rt h ,bs d N s − (cid:90) rt h ,bs d H s , with k (cid:63),bs (Ψ) = ( I − N s ) { h ,bs +(1 − θ s ) h ,bs < (cid:98) b I − Ns } , and where A SH ,b ( t, u b ) is defined similarly as A b ( t, u b ) in Section 4. Once the boundaries are determined and thecredible set is found, a system of recursive HJB equations can be associated to the Principal’s problem, as in theoriginal problem, and the same kind of study explained in Section 5 follows.The optimal screening renegotiation–proof problem can be studied analogously, by defining the corresponding credibleset, which is no longer equivalent to the credible set for the shutdown problem but will also keep the upper boundaryfrom the original problem with full commitment of the banks. References [1] S. Agarwal, D. Lucca, A. Seru, and F. Trebbi. Inconsistent regulators: evidence from banking.
The Quarterly Journal ofEconomics , 129(2):889–938, 2014.[2] R. Antle.
Moral hazard and auditor contracts : an approach to auditors’ legal liability and independence . PhD thesis,Stanford university, 1980.[3] D.P. Baron. A model of the demand for investment banking advising and distribution services for new issues.
The Journalof Finance , 37(4):955–976, 1982.[4] D.P. Baron and D. Besanko. Monitoring, moral hazard, asymmetric information, and risk sharing in procurement contracting.
The RAND Journal of Economics , 18(4):509–532, 1987.[5] D.P. Baron and D. Besanko. Monitoring of performance in organizational contracting: the case of defense procurement.
TheScandinavian Journal of Economics , 90(3):329–356, 1988.[6] D.P. Baron and B. Holmström. The investment banking contract for new issues under asymmetric information: delegationand the incentive problem.
The Journal of Finance , 35(5):1115–1138, 1980.[7] D.P. Baron and R.B. Myerson. Regulating a monopolist with unknown costs.
Econometrica , 50(4):911–930, 1982.[8] N. Bhattacharyya. Good managers work more and pay less dividends – a model of dividend policy. Technical report,University of British Columbia, 1997.[9] B. Biais, T. Mariotti, J.-C. Rochet, and S. Villeneuve. Large risks, limited liability, and dynamic moral hazard.
Econometrica ,78(1):73–118, 2010.[10] B. Caillaud, R. Guesnerie, and P. Rey. Noisy observation in adverse selection models.
The Review of Economic Studies ,59(3):595–615, 1992.[11] G. Carlier. A general existence result for the principal–agent problem with adverse selection.
Journal of MathematicalEconomics , 35(1):129–150, 2001.
12] J. Cvitanić, D. Possamaï, and N. Touzi. Moral hazard in dynamic risk management.
Management Science , 63(10):3328–3346,2017.[13] J. Cvitanić, D. Possamaï, and N. Touzi. Dynamic programming approach to principal–agent problems.
Finance andStochastics , 22(1):1–37, 2018.[14] J. Cvitanić, X. Wan, and H. Yang. Dynamics of contract design with screening.
Management Science , 59(5):1229–1244,2013.[15] R.W.R. Darling and É. Pardoux. Backwards SDE with random terminal time and applications to semilinear elliptic PDE.
The Annals of Probability , 25(3):1135–1159, 1997.[16] D.W. Diamond and P.H. Dybvig. Bank runs, deposit insurance, and liquidity.
Journal of Political Economy , 91(3):401–419,1983.[17] G. Dionne and P. Lasserre. Dealing with moral hazard and adverse selection simultaneously. Cahier de recherche 8559,Département de sciences économiques, université de Montréal, 1985.[18] O. El Euch, T. Mastrolia, M. Rosenbaum, and N. Touzi. Optimal make–take fees for market making regulation. arXivpreprint arXiv:1805.02741 , 2018.[19] N. El Karoui and M.-C. Quenez. Dynamic programming and pricing of contingent claims in an incomplete market.
SIAMJournal on Control and Optimization , 33(1):29–66, 1995.[20] N. El Karoui and X. Tan. Capacities, measurable selection and dynamic programming part II: application in stochasticcontrol problems. arXiv preprint arXiv:1310.3364 , 2013.[21] R. Élie, L. Moreau, and D. Possamaï. On a class of non–Markovian singular stochastic control problems.
SIAM Journal onControl and Optimization , to appear, 2017.[22] P.S. Faynzilberg and P. Kumar. On the generalized principal–agent problem: decomposition and existence results.
Reviewof Economic Design , 5(1):23–58, 2000.[23] A. Figalli, Y.-H. Kim, and R.J. McCann. When is multidimensional screening a convex program?
Journal of EconomicTheory , 146(2):454–478, 2011.[24] W.H. Fleming and H.M. Soner.
Controlled Markov processes and viscosity solutions , volume 25 of
Stochastic modelling andapplied probability . Springer–Verlag New York, 2nd edition, 2006.[25] D. Gottlieb and H. Moreira. Simultaneous adverse selection and moral hazard. Technical report, The university of Pennsyl-vania and escola Brasileira de economia e finanças, 2011.[26] S.J. Grossman and O.D. Hart. An analysis of the principal–agent problem.
Econometrica , 51(1):7–45, 1983.[27] R. Guesnerie and J.-J. Laffont. A complete solution to a class of principal–agent problems with an application to the controlof a self-managed firm.
Journal of Public Economics , 25(3):329–369, 1984.[28] R. Guesnerie, P. Picard, and P. Rey. Adverse selection and moral hazard with risk neutral agents.
European EconomicReview , 33(4):807–823, 1989.[29] S. Hamadène and J.-P. Lepeltier. Backward equations, stochastic control and zero–sum stochastic differential games.
Stochas-tics: An International Journal of Probability and Stochastic Processes , 54(3-4):221–231, 1995.[30] B. Hölmstrom. Moral hazard and observability.
The Bell Journal of Economics , 10(1):74–91, 1979.[31] B. Holmström and P. Milgrom. Aggregation and linearity in the provision of intertemporal incentives.
Econometrica ,55(2):303–328, 1987.[32] R. Hynd. The eigenvalue problem of singular ergodic control.
Communications on Pure and Applied Mathematics , 65(5):649–682, 2012.[33] I. Jewitt. Justifying the first–order approach to principal–agent problems.
Econometrica , 56(5):1177–1190, 1988.[34] B. Jullien, B. Salanié, and F. Salanié. Screening risk–averse agents under moral hazard: single–crossing and the CARA case.
Economic Theory , 30(1):151–169, 2007.[35] J.-J. Laffont and J. Tirole. Using cost observation to regulate firms.
The Journal of Political Economy , 94(3, part 1):614–641,1986.[36] T.R. Lewis and D.E.M. Sappington. Optimal capital structure in agency relationships.
The RAND Journal of Economics ,26(3):343–361, 1995.[37] P.-L. Lions.
Generalized solutions of Hamilton–Jacobi equations , volume 69 of
Research notes in mathematics . PitmanAdvanced Publishing Program, Boston, London, Melbourne, 1982.[38] E. Maskin and J. Riley. Monopoly with incomplete information.
The RAND Journal of Economics , 15(2):171–196, 1984.[39] R.P. McAfee and J. McMillan. Bidding for contracts: a principal–agent analysis.
The RAND Journal of Economics ,17(3):326–338, 1986.[40] N.D. Melumad and S. Reichelstein. Centralization versus delegation and the value of communication.
Journal of AccountingResearch , 25:1–18, 1987.
41] N.D. Melumad and S. Reichelstein. Value of communication in agencies.
Journal of Economic Theory , 47(2):334–368, 1989.[42] J.A. Mirrlees. Population policy and the taxation of family size.
Journal of Public Economics , 1(2):169–198, 1972.[43] J.A. Mirrlees. Notes on welfare economics, information and uncertainty. In M.S. Balch, D.L. McFadden, and S.Y. Wu,editors,
Essays on economic behavior under uncertainty , pages 243–261. Amsterdam: North Holland, 1974.[44] J.A. Mirrlees. The optimal structure of incentives and authority within an organization.
The Bell Journal of Economics ,7(1):105–131, 1976.[45] J.A. Mirrlees. The theory of moral hazard and unobservable behaviour: part I (reprint of the unpublished 1975 version).
The Review of Economic Studies , 66(1):3–21, 1999.[46] James A Mirrlees. An exploration in the theory of optimum income taxation.
The review of economic studies , 38(2):175–208,1971.[47] M. Mussa and S. Rosen. Monopoly and product quality.
Journal of Economic Theory , 18(2):301–317, 1978.[48] R.B. Myerson. Optimal coordination mechanisms in generalized principal–agent problems.
Journal of Mathematical Eco-nomics , 10(1):67–81, 1982.[49] F.H.Jr. Page. Optimal contract mechanisms for principal–agent problems with moral hazard and adverse selection.
EconomicTheory , 1(4):323–338, 1991.[50] H. Pagès. Bank monitoring incentives and optimal ABS.
Journal of Financial Intermediation , 22(1):30–54, 2013.[51] H. Pagès and D. Possamaï. A mathematical treatment of bank monitoring incentives.
Finance and Stochastics , 18(1):39–73,2014.[52] A. Papapantoleon, D. Possamaï, and A. Saplaouras. Existence and uniqueness for BSDEs with jumps: the whole nine yards.
Electronic Journal of Probability , 23(121):1–68, 2018.[53] S. Peng. Probabilistic interpretation for systems of quasilinear parabolic partial differential equations.
Stochastics andStochastic Reports , 37(1-2):61–74, 1991.[54] P. Picard. On the design of incentive schemes under moral hazard and adverse selection.
Journal of Public Economics ,33(3):305–331, 1987.[55] K.W.S. Roberts. Welfare considerations on nonlinear pricing.
The Economic Journal , 89(353):66–83, 1979.[56] J.-C. Rochet and P. Choné. Ironing, sweeping, and multidimensional screening.
Econometrica , 66(4):783–826, 1998.[57] W.P. Rogerson. The first–order approach to principal–agent problems.
Econometrica , 53(6):1357–1368, 1985.[58] M. Royer. Backward stochastic differential equations with jumps and related non–linear expectations.
Stochastic Processesand their Applications , 116(10):1358–1376, 2006.[59] B. Salanié. Sélection adverse et aversion pour le risque.
Annales d’Économie et de Statistique , 18:131–149, 1990.[60] Y. Sannikov. A continuous–time version of the principal–agent problem.
The Review of Economic Studies , 75(3):957–984,2008.[61] Y. Sannikov. Contracts: the theory of dynamic principal–agent relationships and the continuous–time approach. In D. Ace-moglu, M. Arellano, and E. Dekel, editors,
Advances in economics and econometrics, 10th world congress of the EconometricSociety, volume 1, economic theory , number 49 in Econometric Society Monographs, pages 89–124. Cambridge UniversityPress, 2013.[62] A.M. Spence. Multi–product quantity–dependent prices and profitability constraints.
The Review of Economic Studies ,47(5):821–842, 1980.[63] M.A. Spence and R. Zeckhauser. Insurance, information, and individual action.
The American Economic Review , 61(2):380–387, 1971.[64] J. Sung. Optimal contracts under adverse selection and moral hazard: a continuous–time approach.
Review of FinancialStudies , 18(3):1021–1073, 2005.[65] B. Theilen. Simultaneous moral hazard and adverse selection with risk averse agents.
Economics Letters , 79(2):283–289,2003.[66] M.L. Weitzman. The new Soviet incentive model.
The Bell Journal of Economics , 7(1):251–257, 1976.[67] R. Wilson.
Non linear pricing . Oxford University Press, 1993.[68] R. Zeckhauser. Medical insurance: a case study of the tradeoff between risk spreading and appropriate incentives.
Journalof Economic Theory , 2(1):10–26, 1970.[69] L. Zou. Threat–based incentive mechanisms under moral hazard and adverse selection.
Journal of Comparative Economics ,16(1):47–74, 1992. Proofs for the pure moral hazard case
We provide in this section all the proofs of the results of Section 3. We start with the
Proof of Proposition 3.1.
Using the martingale representation theorem (recall that D is supposed to be integrable and that k isbounded by definition), we deduce that for any k ∈ K there exist G − predictable processes h ,i,k and h ,i,k such that, P − a . s . d u it ( k, θ i , D i ) = (cid:0) ru it ( k, θ i , D i ) − Bk t (cid:1) d t − ρ i d D it − h ,i,kt (cid:0) d N t − λ kt d t (cid:1) − h ,i,kt (cid:0) d H t − (1 − θ it ) λ kt d t (cid:1) , ≤ t < τ, (A.1)Let us then define Y i,kt := u it ( k, θ i , D i ) , Z i,kt := ( h ,i,kt , h ,i,kt ) (cid:62) , M t := ( N t , H t ) (cid:62) , (cid:102) M it := M t − (cid:90) t λ s (1 , − θ is ) (cid:62) d s, K it := ρ i D it , so that we can rewrite (A.1) as follows Y i,kt = 0 − (cid:90) τt f i ( s, k s , Y i,ks , Z i,ks )d s + (cid:90) τt Z i,ks · d (cid:102) M is + (cid:90) τt d K is , ≤ t ≤ τ, P − a . s . In other words, ( Y i,k , Z i,k ) appears as a (super–)solution to a BSDE with (finite) random terminal time, as studied for instanceby Peng [53] or Darling and Pardoux [15]. Notice that by direct computations, it is immediate that it is equivalent to look for asolution ( Y i , Z i ) of BSDE (3.2) or to look for solution to the following BSDE (cid:101) Y it = ξ τ − (cid:90) τt ˜ g i (cid:0) s, (cid:101) Z is (cid:1) d s + (cid:90) τt (cid:101) Z is · d (cid:102) M is , ≤ t ≤ τ, P − a . s ., (A.2)where we defined (cid:101) Y it := e − rt Y it + (cid:90) t e − rs d K s , (cid:101) Z it := e − rt Z it , t ≥ , ξ τ := (cid:90) τ e − rs d K s , ˜ g i ( s, z ) := ( I − N s ) (cid:18) α I − N s εz · (cid:18) − θ is (cid:19) − B e − rs (cid:19) − . By direct computations, it is easy to see that ˜ g i satisfies, for any ( t, z, z (cid:48) ) ∈ R + × R × R ˜ g i ( t, z ) − ˜ g i ( t, z (cid:48) ) ≤ = γ t ( z, z (cid:48) ) λ t ( z − z (cid:48) ) · (1 , − θ it ) (cid:62) , where γ t ( z, z (cid:48) ) := ε { ( z − z (cid:48) ) · (1 , − θ it ) (cid:62) > } , verifies ≤ γ t ( z, z (cid:48) ) ≤ ε. Since in addition ˜ g i ( t, is bounded, (2.3) holds, γ t ( z, z (cid:48) ) isbounded and non–negative, the intensity of (cid:102) M i is also bounded, as well as its jumps, we deduce that all the assumptions ofTheorems 3.5 and 3.24 in [52] hold in our setting, proving wellposedness of (3.2) in the space described by (3.3), and that wecan apply a comparison theorem. Therefore, we deduce immediately that for any k ∈ K Y i,kt ≤ Y it = Y i,k (cid:63),i t , P − a . s ., where we defined k (cid:63),it := ( I − N t ) { Z it · (1 , − θ it ) (cid:62)
First of all, it is clear that the bank of type ρ i can get arbitrarily large levels of utility (it suffices for theinvestor to set d D is := n d s for n large enough, starting from time t ). The bank’s maximal level of utility is therefore + ∞ , whichcorresponds to a utility equal to −∞ for the investor. Then, coming back to the definition of the bank’s problem, or to theBSDE (3.2), it is clear, for instance by using the comparison theorem for super solutions to (3.2) (see [58, Theorem 2.5]), that inorder to minimise the utility that the bank obtains, the investor has to set D i = 0 . Moreover, since by definition we must alwayshave Y it ≥ and Y iτ = 0 , and since the totally inaccessible jumps of Y (recall that D is assumed to be predictable) are given by ∆ Y it = − Z it · ∆ M t , we must have that Y it − = Z it · (1 , (cid:62) , and Y it − ≥ Z it · (1 , (cid:62) , t > , P − a . s ., (A.3)Indeed, the support of the laws of τ and the τ j under P is [0 , + ∞ ) . This implies in particular that we must have Z it · (0 , (cid:62) ≥ ,which in turn implies that the generator g i is then non–increasing with respect to θ i , and thus that the minimal utility for the We emphasise that since the filtration G is augmented and generated by point processes, the predictable martingale representation holdsfor any of the probability measures ( P k ) k ∈ K , see for instance [18, Lemma A.1]. ank is attained, as expected, when θ i = 0 . Then, if ( θ i , D i ) = (0 , (which is obviously in Θ × D ) starting from time t , it isclear that the bank will never monitor and will obtain U it (0 ,
0) = B ( I − N t ) E P I − N · (cid:20) (cid:90) τt e − r ( s − t ) d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = B ( I − N t ) r (cid:16) − E P I − N · (cid:104) e − r ( τ − t ) (cid:12)(cid:12)(cid:12) G t (cid:105)(cid:17) = B ( I − N t ) r (cid:18) − (cid:90) + ∞ λ I − N t t e − xr − xλ I − Ntt d x (cid:19) = B ( I − N t ) r + λ I − N t t . Notice that this corresponds to the investor getting µ ( I − N t ) E P I − N · [ τ − t |G t ] = µ ( I − N t ) λ I − N t t . B Short–term contracts with constant payments
In this section we first analyse the optimal responses and the value functions of the banks at a starting time t ≥ , under contractswith constant payments of the form d D s = c d s , where c is any G t − measurable random variable, and with θ ≡ , so that the poolis liquidated immediately after the first default. Then, we extend the study to the case in which the payments are delayed andthey happen only after a certain time t (cid:63) > t . B.1 Contracts with no delay
Proposition B.1.
For any t ≥ , consider the contract ( θ, D ) ∈ Θ × D such that θ s = 0 , d D s = c d s, ∀ s ≥ t, where c is any G t − measurable random variable. For i ∈ { g, b } , define ¯ c i := b I − N t ( r + λ t ) ρ i . The optimal effort of the bank of type ρ i and her expected utility under the contract are • If c ≤ ¯ c i then k (cid:63),is ( θ, D ) = k SH s , ∀ s ∈ [ t, τ ) and U it ( θ, D ) = ρ i c + B ( I − N t ) r + λ k SH t . • If c ≥ ¯ c i then k (cid:63),is ( θ, D ) = 0 , ∀ s ∈ [ t, τ ) and U it ( θ, D ) = ρ i cr + λ t .Proof. ( i ) If the bank of type ρ i always monitors, we have u it (0 , θ, D ) = E P (cid:20) (cid:90) τt e − r ( s − t ) ρ i c d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = ρ i cr + λ t . Hence, the continuation utility is constant in time and if the payment c is exactly equal to u i ( r + λ t ) /ρ i , for some u i ≥ , thenthe bank receives exactly u i . In this case, the strategy of always monitoring is incentive compatible if and only if u i ≥ b I − N t .The minimum payment such that the bank of type ρ i will always work is therefore c i = b I − N t ( r + λ t ) ρ i . ( ii ) If the bank of type ρ i always shirks, her continuation utility is constant and equal to u it (cid:0) k SH , θ, D (cid:1) = E P k SH (cid:20) (cid:90) τt e − r ( s − t ) ( ρ i c + B )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = ρ i c + B ( I − N t ) r + λ k SH t . Then, if for some u i ≥ one takes c equal to u i (cid:0) r + λ k SH t (cid:1) − B ( I − N t ) ρ i , the bank receives u i . Therefore k SH is incentive compatible if and only if u i < b I − N t . Nevertheless, since the payment c mustbe positive, u i must be greater than B ( I − N t ) / ( r + λ k SH t ) . The supremum of the payments such that the bank of type ρ i willalways shirk is therefore equal to b I − N t (cid:0) r + λ k SH t (cid:1) − B ( I − N t ) ρ i = b I − N t ( r + λ t ) ρ i = c i . .2 Contracts with delayed payments Proposition B.2.
For any t ≥ , consider the contract ( θ, D ) ∈ Θ × D such that θ s = 0 , d D s = c s ≥ t (cid:63) d s, ∀ s ≥ t, where c is any G t − measurable random variable and t (cid:63) > t is a fixed constant. For i ∈ { g, b } , define the time ¯ t i ( c ) := t + 1 r + λ t log (cid:18) ρ i cb I − N t ( r + λ t ) (cid:19) . The optimal effort of the bank of type ρ i and her expected utility under the contract are • If c ≤ ¯ c i , then k (cid:63),is (0 , D ) = k SH s , ∀ s ∈ [ t, τ ) and U it (0 , D ) = exp (cid:16) − ( r + λ k SH t )( t (cid:63) − t ) (cid:17) ρ i cr + λ k SH t + B ( I − N t ) r + λ k SH t . • If c > ¯ c i , t (cid:63) ≤ ¯ t i ( c ) , then k (cid:63),is (0 , D ) = 0 , ∀ s ∈ [ t, τ ) and U it (0 , D ) = exp (cid:16) − ( r + λ t )( t (cid:63) − t ) (cid:17) ρ i cr + λ t . • If c > ¯ c i , t (cid:63) > ¯ t i ( c ) , then k (cid:63),is (0 , D ) = k SH s { s< ¯ t i ( c ) } , ∀ s ∈ [ t, τ ) and U it (0 , D ) = exp (cid:16) − ( r + λ k SH t )( t (cid:63) − t ) (cid:17)(cid:18) ρ i cb I − N t ( r + λ t ) (cid:19) r + λk SH tr + λ t b I − N t ( r + λ t ) r + λ k SH t + B ( I − N t ) r + λ k SH t . Proof. ( i ) If the bank of type ρ i always works, at any time t ≤ s < t (cid:63) , her continuation utility is, noticing that since θ = 0 , wehave that ( λ u ) u ≥ t is constant u is (0 , , D ) = E P (cid:20) (cid:90) τt (cid:63) ∧ τ e − r ( u − s ) ρ i c d u (cid:12)(cid:12)(cid:12)(cid:12) G s (cid:21) = e − ( r + λ t )( t (cid:63) − s ) ρ i cr + λ t = u it (0 , , D )e ( r + λ t )( s − t ) . Therefore, at s = t (cid:63) the continuation utility of the bank is u it (cid:63) (0 , , D ) = u it (0 , , D )e ( r + λ t )( t (cid:63) − t ) . Next, for any s > t (cid:63) , thecontinuation utility of the bank will be u is (0 , , D ) = E P (cid:20) (cid:90) τs e − r ( u − s ) ρ i c d s (cid:12)(cid:12)(cid:12)(cid:12) G s (cid:21) = ρ i cr + λ t . Then, we see that once the bank starts being paid, her continuation utility becomes constant and it must be equal to u it (cid:63) (0 , , D ) .Then, if for some u i ≥ , one chooses c equal to u i e ( r + λ t ) t (cid:63) ( r + λ t ) ρ i , (B.1)the continuation utility of the bank will be an increasing process with initial value u i . Therefore, is incentive compatible if andonly if u i ≥ b I − N t . The minimum payment and delay such that the bank always works are t (cid:63) = 0 and c i = b I − Nt ( r + λ t ) ρ i . ( ii ) If the bank of type ρ i always shirks, at any time t ≤ s < t (cid:63) , her continuation utility is u is (cid:0) k SH , , D (cid:1) = E P k SH (cid:20) (cid:90) τt (cid:63) ∧ τ e − r ( u − s ) ρ i c d u + (cid:90) τs B e − r ( u − s ) ( I − N t )d u (cid:12)(cid:12)(cid:12)(cid:12) G s (cid:21) = e − (cid:0) r + λ k SH t (cid:1) ( t (cid:63) − s ) ρ i cr + λ k SH t + B ( I − N t ) r + λ k SH t . Therefore u is (cid:0) k SH , , D (cid:1) = e (cid:0) r + λ k SH t (cid:1) ( s − t ) (cid:18) u it ( k SH , , D ) − B ( I − N t ) r + λ k SH t (cid:19) + B ( I − N t ) r + λ k SH t , and the continuation utility is an increasing process. Recall that k SH is incentive compatible if and only if u is (cid:0) k SH , , D (cid:1) < b I − N t for every s ≥ t . However, if t (cid:63) is large, there will exist t w such that u it w (cid:0) k SH , , D (cid:1) = b I − N t and the bank will start to work.More precisely, t w depends on the initial value u it ( k SH , , D ) and is given by t w := t + 1 r + λ k SH t log (cid:18) b I − N t ( r + λ k SH t ) − B ( I − N t ) u it ( k SH , , D )( r + λ k SH t ) − B ( I − N t ) (cid:19) . Notice that t w ≥ t if and only if b I − N t ≥ u it ( k SH , , D ) . Therefore, k SH is incentive compatible if and only if t (cid:63) < t w . Under thiscondition, at t = t (cid:63) the continuation utility of the bank is u it (cid:63) (cid:0) k SH , , D (cid:1) = e ( r + λ k SH t )( t (cid:63) − t ) (cid:18) u it ( k SH , θ, D ) − B ( I − N t ) r + λ k SH t (cid:19) + B ( I − N t ) r + λ k SH t < b I − N t . Once the bank starts being paid her continuation utility is constant and equal to u is (cid:0) k SH , , D (cid:1) = E P k SH (cid:20) (cid:90) τs e − r ( u − s ) ( ρ i c + B ( I − N t ))d u (cid:12)(cid:12)(cid:12)(cid:12) G s (cid:21) = ρ i c + B ( I − N t ) r + λ k SH t . o if for some u i ≥ the payment c is equal to e ( r + λ k SH t )( t (cid:63) − t ) (cid:0) u i ( r + λ k SH t ) − B ( I − N t ) (cid:1) ρ i , (B.2)the expected payoff of the bank at time t is u i . The supremum of the delays and payments such that the bank always shirks arerespectively t w and e ( r + λ k SH t )( t w − t ) (cid:0) b I − N t ( r + λ k SH t ) − B ( I − N t ) (cid:1) ρ i = b I − N t ( r + λ t ) ρ i = c i . ( iii ) Finally, consider the case when t (cid:63) is greater than t w . Under this contract, the bank will shirk until time t w and will workafterwards. Indeed, from the previous analysis we know that this strategy is incentive compatible. At time t w we have that u it w ( k SH , , D ) = b I − N t and for s ∈ [ t w , t (cid:63) ) the continuation utility is given by u is (0 , , D ) = E P (cid:20) (cid:90) τt (cid:63) ∧ τ e − r ( u − s ) ρ i c d u (cid:12)(cid:12)(cid:12)(cid:12) G s (cid:21) = e − ( r + λ t )( t (cid:63) − s ) ρ i cr + λ t = e ( r + λ t )( s − t w ) u it w ( k SH , , D ) = b I − N t e ( r + λ t )( s − t w ) . Therefore, at t = t (cid:63) the continuation utility of the bank is u it (cid:63) (0 , , D ) = b I − N t e ( r + λ t )( t (cid:63) − t w ) , and for any s > t (cid:63) , the continuation utility of the bank is constant and equal to u is (0 , , D ) = E P (cid:20) (cid:90) τs e − r ( u − s ) ρ i c d u (cid:12)(cid:12)(cid:12)(cid:12) G s (cid:21) = ρ i cr + λ t . So if for some u i ≥ the payment c is equal to b I − N t ( r + λ t )e ( r + λ t )( t (cid:63) − t ) ρ i (cid:18) u i ( r + λ k SH t ) − B ( I − N t ) b I − N t ( r + λ t ) (cid:19) r + λ tr + λk SH t , (B.3)the expected payoff of the bank at time t is u i . The minimum payment and delay such that the bank shirks first and worksafterwards are t (cid:63) = t w and c i = b I − Nt ( r + λ t ) ρ i . Notice that t i ( c ) in the statement of the Proposition is the corresponding expressionfor t w as a function of the payments c .We conclude this section with the following result, saying that every point in the upper boundary of the credible set can beattained by short–term contracts with delay. Proposition B.3.
Fix some t ≥ . For any point ( u b , u g ) in the upper boundary of the credible set C t , there exists a G t − measurable payment c , and t (cid:63) ≥ t such that the contract ( θ, D ) with θ s = 0 , d D s = c s ≥ t (cid:63) d s, ∀ s ≥ t, is such that U bt ( θ, D ) = u b and U gt ( θ, D ) = u g .Proof. ( i ) Let c > ¯ c b > ¯ c g and t (cid:63) ≤ ¯ t b ( c ) < ¯ t g ( c ) . Then k (cid:63),b ( θ, D ) = k (cid:63),g ( θ, D ) = 0 and the values of the banks are U gt ( θ, D ) = ρ g cr + λ t e − ( r + λ t )( t (cid:63) − t ) , U bt ( θ, D ) = ρ b cr + λ t e − ( r + λ t )( t (cid:63) − t ) . Therefore the utilities satisfy U gt ( θ, D ) = ρ g ρ b U bt ( θ, D ) , with U gt ( θ, D ) ∈ (cid:20) ρ g ρ b b I − N t , ∞ (cid:19) , U bt ( θ, D ) ∈ [ b I − N t , ∞ ) . ( ii ) If c > ¯ c b and ¯ t b ( c ) < t (cid:63) ≤ ¯ t g ( c ) , we have that the good bank will always work and the bad bank will start working at time ¯ t b ( c ) . Their value functions are U gt ( θ, D ) = ρ g cr + λ t e − ( r + λ t )( t (cid:63) − t ) , U bt ( θ, D ) = e − ( r + λ k SH t )( t (cid:63) − t ) (cid:18) ρ b cb I − N t ( r + λ t ) (cid:19) r + λk SH tr + λ t b I − N t ( r + λ t ) r + λ k SH t + B ( I − N t ) r + λ k SH t , so they belong to the curve U gt ( θ, D ) = ρ g ρ b b λk SH t − λ tr + λk SH t I − N t (cid:18) U bt ( θ, D ) − B ( I − N t ) r + λ k SH t (cid:19) r + λ tr + λk SH t (cid:18) r + λ k SH t r + λ t (cid:19) r + λ tr + λk SH t , and take values in the sets (recall the definition of x (cid:63)j in Proposition 4.3) U gt ( θ, D ) ∈ (cid:20) b I − N t , ρ g ρ b b I − N t (cid:19) , U bt ( θ, D ) ∈ [ x (cid:63)I − N t , b I − N t ) . iii ) If c > ¯ c b and t g ( c ) < t (cid:63) , the good bank will start working at time ¯ t g ( c ) and the bad bank will start to work at time ¯ t b ( c ) .Their value functions are U gt ( θ, D ) = e − ( r + λ k SH t )( t (cid:63) − t ) (cid:18) ρ g cb I − N t ( r + λ t ) (cid:19) r + λk SH tr + λ t b I − N t ( r + λ t ) r + λ k SH t + B ( I − N t ) r + λ k SH t ,U bt ( θ, D ) = e − ( r + λ k SH t )( t (cid:63) − t ) (cid:18) ρ b cb I − N t ( r + λ t ) (cid:19) r + λk SH tr + λ t b I − N t ( r + λ t ) r + λ k SH t + B ( I − N t ) r + λ k SH t , so they belong to the line U gt ( θ, D ) = (cid:18) ρ g ρ b (cid:19) r + λk SH tr + λ t (cid:18) U bt ( θ, D ) − B ( I − N t ) r + λ k SH t (cid:19) + B ( I − N t ) r + λ k SH t , with U gt ( θ, D ) ∈ (cid:20) B ( I − N t ) r + λ k SH t , b I − N t (cid:19) , U bt ( θ, D ) ∈ (cid:20) B ( I − N t ) r + λ k SH t , x (cid:63)I − N t (cid:19) . B.3 Initial lump–sum payment
Take any point ( u b , u g ) in the credible set at time t . We know that there exists an admissible contract ( θ, D ) , such that U bt ( θ, D ) = u b and U gt ( θ, D ) = u g . Consider the payments D (cid:96) which differ from D only at time t , where a lump-sum payment ofsize (cid:96) > is made. This added lump-sum payment will not change the banks’ incentives and the new value functions at time t will be U gt ( θ, D (cid:96) ) = u g + ρ g (cid:96), U bt ( θ, D (cid:96) ) = u b + ρ b (cid:96). Hence, the new pair of values of the banks belong to the line with slope ρ g ρ b which passes through the point ( u b , u g ) . Since in oursetting there is no upper bound on the payment, by increasing the value of (cid:96) it is possible to reach every point of the ray whichstarts at ( u b , u g ) and goes in the positive direction. B.4 Credible region under contracts with delay
From the previous subsection we know that for every point ( u b , u g ) on the upper boundary there exists a pair ( c, t (cid:63) ) , with c > ¯ c b ,such that under the contract ( θ ≡ , d D s = c { s ≥ t (cid:63) } d s ) we have U bt ( θ, D ) = u b and U gt ( θ, D ) = u g . As explained in Section B.3,if we consider the contract ( θ, D (cid:96) ) with an additional initial lump–sum payment, the incentives of the banks will not change andthe new value functions of the agents will be U bt ( θ, D (cid:96) ) = u b + ρ b (cid:96) , U gt ( θ, D ) = u g + ρ g (cid:96) . Therefore under short–term contractswith delay which reach the upper boundary and lump–sum payments, all the subregion of the credible set delimited by the linesshown in Figure 6 can be reached. We will not enter into details but it can be proved that under all the short–term contractswith delay (not only the ones who reach the upper boundary) and lump-sum payments, the subregion of the credible set whichcan be reached is exactly the same. When there is only one loan left, this region is equal to the whole credible set but when j > the credible set is strictly bigger due to the pair of utilities that can be achieved in situations when θ (cid:54)≡ . u b u g u g = u b L b U ?j ( u b ) Br + b λ SH j Br + b λ SH j b jρ g ρ b b j x ?j b j L : u g = ρ g ρ b u b + Br + b λ SH j (cid:16) − ρ g ρ b (cid:17) . Figure 6: Credible region under short-term contracts with delay and lump-sum payment.32
Technical results for the lower boundary
We begin this section with the
Proof of Proposition 4.1.
Observe first that we have E P k SH (cid:2) e − r ( τ Nt +1 − t ) (cid:12)(cid:12) G t (cid:3) = (cid:90) ∞ e − rx (cid:98) λ SH I − N t e − (cid:98) λ SH I − Nt x d x = (cid:98) λ SH I − N t r + (cid:98) λ SH I − N t , and for any (cid:96) ∈ { N t + 1 , . . . , I − } E P k SH (cid:2) e − r ( τ (cid:96) +1 − τ (cid:96) ) (cid:12)(cid:12) G t (cid:3) = (cid:90) ∞ e − rx (cid:98) λ SH I − (cid:96) e − (cid:98) λ SH I − (cid:96) x d x = (cid:98) λ SH I − (cid:96) r + (cid:98) λ SH I − (cid:96) . Thus, the utility that the bank gets from shirking (without considering the payments in the contract) is u t ( k SH , θ,
0) = E P k SH (cid:20) (cid:90) τt e − r ( s − t ) B ( I − N s )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = E P k SH (cid:20) (cid:90) τ Nt +1 t e − r ( s − t ) B ( I − N t )d s + N t + m − (cid:88) i = N t +1 (cid:90) τ i +1 τ i e − r ( s − t ) B ( I − i )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = B ( I − N t ) r E P k SH (cid:2) − e − r ( τ Nt +1 − t ) (cid:12)(cid:12) G t (cid:3) + N t + m − (cid:88) i = N t +1 B ( I − i ) r E P k SH (cid:2) e − r ( τ i − t ) − e − r ( τ i +1 − t ) (cid:12)(cid:12) G t (cid:3) = B ( I − N t ) r + (cid:98) λ SH I − N t + N t + m − (cid:88) i = N t +1 B ( I − i ) r E P k SH (cid:20)(cid:0) − e − r ( τ i +1 − τ i ) (cid:1) i − (cid:89) (cid:96) = N t e − r ( τ (cid:96) +1 − τ (cid:96) ) (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . Therefore, by independence we have u t ( k SH , θ,
0) = B ( I − N t ) r + (cid:98) λ SH I − N t + N t + m − (cid:88) i = N t +1 B ( I − i ) r + (cid:98) λ SH I − i i − (cid:89) (cid:96) = N t (cid:98) λ SH I − (cid:96) r + (cid:98) λ SH I − (cid:96) = B ( I − N t ) r + (cid:98) λ SH I − N t + I − N t − (cid:88) i = I − N t − m +1 Bir + (cid:98) λ SH i I − N t (cid:89) (cid:96) = i +1 (cid:98) λ SH (cid:96) r + (cid:98) λ SH (cid:96) . We proceed with the
Proof of Lemma 4.1.
The value functions of the banks under
Ψ := ( θ, D ) are given by U gt (Ψ) = E P k(cid:63),g (Ψ) (cid:20) (cid:90) τt e − r ( s − t ) ( ρ g d D s + Bk (cid:63),gs (Ψ)d s ) (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) , U bt (Ψ) = E P k(cid:63),b (Ψ) (cid:20) (cid:90) τt e − r ( s − t ) ( ρ b d D s + Bk (cid:63),bs (Ψ)d s ) (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . Thus, we first have, P − a.s. U gt (Ψ) ≥ E P k(cid:63),b (Ψ) (cid:20) (cid:90) τt e − r ( s − t ) ( ρ g d D s + Bk (cid:63),bs (Ψ)d s ) (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) ≥ E P k(cid:63),b (Ψ) (cid:20) (cid:90) τt e − r ( s − t ) ( ρ b d D gs + Bk (cid:63),bs (Ψ)d s ) (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = U bt (Ψ) . But we also have U gt (Ψ) ≥ E P k(cid:63),b (Ψ) (cid:20) (cid:90) τt e − r ( s − t ) ( ρ g d D s + Bk (cid:63),bs (Ψ)d s ) (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = U bt (Ψ) + ( ρ g − ρ b ) E P k(cid:63),b (Ψ) (cid:20) (cid:90) τt e − r ( s − t ) d D s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = ρ g ρ b U bt (Ψ) − ( ρ g − ρ b ) ρ b E P k(cid:63),b (Ψ) (cid:20) (cid:90) τt e − r ( s − t ) Bk (cid:63),bs (Ψ)d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . Observe next that sup k ∈ K E P k (cid:20) (cid:90) τt e − r ( t − s ) Bk s d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = E P k SH (cid:20) (cid:90) τt e − r ( t − s ) Bk SH s d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) , because the left–hand side is the value function of a bank who is offered a contract with no payments. Therefore, we have that U gt (Ψ) ≥ ρ g ρ b U bt (Ψ) − ( ρ g − ρ b ) ρ b E P k SH (cid:20) (cid:90) τt e − r ( s − t ) Bk SH s d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) ≥ ρ g ρ b U bt (Ψ) − ( ρ g − ρ b ) ρ b C ( I − N t ) , because the utility that the banks get from shirking is non–decreasing with respect to the process θ and its maximum value isequal to C ( I − N t ) , attained when θ ≡ (see (4.2)).We continue this section with the roof of Proposition 4.2. Thanks to Lemma 4.1, it suffices to prove the existence of contracts under which the value functionsof the banks satisfy the equalities. • Step 1:
First, fix some t ≥ , take any u b ∈ [ c ( I − N t , , C ( I − N t )] and fix m ∈ { , . . . , I − N t − } such that c ( I − N t , m ) ≤ u b ≤ c ( I − N t , m +1) . Next, take θ t ( u b ) ∈ [0 , such that u b = c ( I − N t , m )+ θ t ( u b )( c ( I − N t , m +1) − c ( I − N t , m )) . Then, there is a contract ( θ, D ) ∈ Θ × D such that U gt ( θ, D ) = U bt ( θ, D ) = u b . Such a contract can be defined as follows d D s := 0 , θ s := { t ≤ s ≤ τ Nt + m } + (1 − θ t ( u b )) { τ Nt + m
Fix again some t ≥ , and choose now any u b ≥ C ( I − N t ) and define u g := ρ g ρ b u b − ( ρ g − ρ b ) ρ b C ( I − N t ) . Let (cid:96) t := ( u b − C ( I − N t )) /ρ b and consider the admissible contract satisfying, θ s = 1 , d D s = (cid:96) t { s = t } , for every s ≥ t . The optimalstrategy for both banks under this contract is to always shirk and then U bt ( θ, D ) = E P k SH (cid:20) (cid:90) τt e − r ( s − t ) (cid:0) ρ b d D s + Bk SH s d s (cid:1)(cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = ρ b (cid:96) t + C ( I − N t ) = u b ,U gt ( θ, D ) = E P k SH (cid:20) (cid:90) τt e − r ( s − t ) (cid:0) ρ g d D s + Bk SH s d s ) (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = ρ g (cid:96) t + C ( I − N t ) = u g . We conclude this section by proving some useful results that will be used in Section 5.1.1 in the study of the value function of theinvestor on the lower boundary. We show that there are several ways of reaching the lower boundary and that all the contractswhich can achieve it have the same structure as the ones used in the proof of Proposition 4.2.
Lemma C.1.
Consider any ( t, u b , u g ) ∈ [0 , τ ] × (cid:98) V I − N t × (cid:98) V I − N t such that in addition u b = u g . Any contract Ψ = ( θ, D ) ∈ Θ × D such that U bt (Ψ) = u b and U gt (Ψ) = u g , has no payments on [ t, τ ] and consequently both banks always shirk under Ψ .Proof. Looking at the proof of (4.3) we deduce that necessarily k (cid:63),gs (Ψ) = k (cid:63),bs (Ψ) , dD s = 0 , ∀ s ≥ t. Since there are no payments,we have that k (cid:63),gs (Ψ) = k (cid:63),bs (Ψ) = k SH s for s ∈ [ t, τ ] and indeed have U gt (Ψ) = U bt (Ψ) = E P k SH (cid:20) (cid:90) τt e − r ( s − t ) B ( I − N s )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . Lemma C.2.
Consider any ( t, u g , u b ) ∈ R + × (cid:98) V I − N t × (cid:98) V I − N t such that in addition u g = ρ g ρ b u b − ( ρ g − ρ b ) ρ b C ( I − N t ) . Under any contract
Ψ = ( θ, D ) ∈ Θ × D such that U bt (Ψ) = u b and U gt (Ψ) = u g , the pool is not liquidated until the last default ( τ = τ I ) and both banks always shirk on [ t, τ ] .Proof. Looking at the proof of (4.4), we deduce that necessarily k (cid:63),gs (Ψ) = k (cid:63),bs (Ψ) = k SH s , θ s = 1 , for every s ≥ t. Thus, thevalue functions of the banks are given by U gt (Ψ) = ρ g E P k SH (cid:20) (cid:90) τ I t e − r ( s − t ) d D s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) + C ( I − N t ) , U bt (Ψ) = ρ b E P k SH (cid:20) (cid:90) τ I t e − r ( s − t ) d D s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) + C ( I − N t ) . D Technical results for the upper boundary
Lemma D.1.
For every j ≥ , x (cid:63)j > ρ b ρ g (cid:98) b j . roof. For any j ≥ , define the functions g, h : R −→ R by g ( x ) := x r + (cid:98) λ SH jr + (cid:98) λ j (cid:98) b j r + (cid:98) λ j r + (cid:98) λ SH j + Bjr + (cid:98) λ SH j , h ( x ) := (cid:98) b j x. Then g is strictly convex in R + and we have that g (1) = h (1) = (cid:98) b j and g (cid:48) (1) = h (cid:48) (1) = (cid:98) b j . Thus, h is the tangent line to g at x = 1 so g ( x ) > h ( x ) for every x (cid:54) = 1 and therefore x (cid:63)j = g (cid:18) ρ b ρ g (cid:19) > h (cid:18) ρ b ρ g (cid:19) = ρ b ρ g (cid:98) b j . Proposition D.1.
For every j ≥ , the function (cid:98) U (cid:63)j defined by (4.17) satisfies (cid:98) U (cid:63)j ( x ) x ≤ ρ g ρ b , ∀ x ≥ Bjr + (cid:98) λ SH j . Moreover, equality holds if and only if x ≥ (cid:98) b j .Proof. Define A ( x ) := (cid:98) U (cid:63)j ( x ) x . If x ≥ (cid:98) b j − then A ( x ) = ρ g /ρ b . If now x ∈ [ x (cid:63)j , (cid:98) b j ) , we have A ( x ) = ρ g ρ b ( (cid:98) b j ) (cid:98) λSHj − (cid:98) λ jr + (cid:98) λSHj (cid:18) r + (cid:98) λ SHj r + (cid:98) λ j (cid:19) r + (cid:98) λ jr + (cid:98) λSHj x (cid:18) x − Bjr + (cid:98) λ SHj (cid:19) r + (cid:98) λ jr + (cid:98) λSHj . This function is decreasing so that A reaches its maximum value over [ x (cid:63)j , (cid:98) b j ) at x (cid:63)j . Next, we have A ( x (cid:63)j ) = (cid:98) b j x (cid:63)j < ρ g ρ b ⇐⇒ x (cid:63)j > ρ b ρ g (cid:98) b j , and the last inequality holds as a consequence of Lemma D.1. Finally, if x ∈ (cid:2) Bjr + (cid:98) λ SH j , x (cid:63)j (cid:1) then A ( x ) = 1 x (cid:18) ρ g ρ b (cid:19) r + (cid:98) λ SH jr + (cid:98) λ j (cid:18) x − Bjr + (cid:98) λ SH j (cid:19) + 1 x Bjr + (cid:98) λ SH j . This function is increasing, hence A ( x ) ≤ A ( x (cid:63)j ) < ρ g ρ b , ∀ x ∈ (cid:104) Bjr + (cid:98) λ SHj , x (cid:63)j (cid:105) . Corollary D.1.
Let j ≥ and (cid:98) U (cid:63)j , (cid:98) U (cid:63)j − defined by (4.17) , and assume that (cid:98) λ k g j ≤ (cid:98) λ k b j . Then, for any u b ≥ h ,b + B ( j − r + (cid:98) λ SHj − wehave (cid:98) U (cid:63)j − ( u b − h ,b ) (cid:98) λ k g j − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) λ k b j ( u b − h ,b ) ≤ . Furthermore, equality holds if and only if u b − h ,b ≥ (cid:98) b j , u b ≥ (cid:98) b j and (cid:98) λ k b j = (cid:98) λ k g j .Proof. Under the conditions of the corollary, the following allows us to conclude immediately (cid:98) U (cid:63)j − ( u b − h ,b ) u b − h ,b ≤ ρ g ρ b ≤ (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) . Corollary D.2.
For j ≥ , let (cid:98) C j and (cid:98) U (cid:63)j be defined by (4.16) and (4.17) respectively. If ( θ, h ,b ) ∈ (cid:98) C j is such that u b − θ ( u b − h ,b ) ≥ (cid:98) b j then (cid:98) U (cid:63)j ( u b ) − θ (cid:98) U (cid:63)j − ( u b − h ,b ) ≥ (cid:98) b j . As a consequence, in the context of equation (4.15) , for every ( θ, h ,b ) ∈ (cid:98) C j wehave k g ≤ k b and (cid:98) λ k g j ≤ (cid:98) λ k b j .Proof. First observe that u b − θ ( u b − h ,b ) ≥ (cid:98) b j implies u b ≥ (cid:98) b j . Then we have (cid:98) U (cid:63)j ( u b ) − (cid:98) b j ≥ ρ g ρ b ( u b − (cid:98) b j ) ≥ (cid:98) U (cid:63)j − ( u b − h ,b ) u b − h ,b ( u b − (cid:98) b j ) . Also, θ ≤ u b − (cid:98) b j u b − h ,b and thus (cid:98) U (cid:63)j ( u b ) − θ (cid:98) U (cid:63)j − ( u b − h ,b ) ≥ (cid:98) U (cid:63)j ( u b ) − (cid:18) u b − (cid:98) b j u b − h ,b (cid:19) (cid:98) U (cid:63)j − ( u b − h ,b ) ≥ (cid:98) b j . e now proceed with the Proof of Lemma 4.2.
We start with the region u b < (cid:98) b , (cid:98) U ( u b ) < (cid:98) b . For these points, we have that k b = k g = 1 , so (4.11) canbe solved easily and leads to, for some C ∈ R (cid:98) U ( u b ) = C (cid:18) u b − Br + (cid:98) λ (cid:19) + Br + (cid:98) λ . If u b < (cid:98) b and (cid:98) U ( u b ) ≥ (cid:98) b , then k b = 1 , k g = 0 and we can solve (4.11) to obtain for some C ∈ R (cid:98) U ( u b ) = C (cid:18) u b − Br + (cid:98) λ (cid:19) r + (cid:98) λ r + (cid:98) λ . Finally, when u b ≥ (cid:98) b and (cid:98) U ( u b ) ≥ (cid:98) b the optimal strategies are k b = k g = 0 and we have for some C ∈ R , (cid:98) U ( u b ) = C u b . Weare interested in smooth solutions of (4.11). Denote by (cid:98) U (1)1 , (cid:98) U (2)1 and (cid:98) U (3)1 the following functions (cid:98) U (1)1 ( u b ) := C (cid:18) u b − Br + (cid:98) λ (cid:19) + Br + λ , (cid:98) U (2)1 ( u b ) := C (cid:18) u b − Br + (cid:98) λ (cid:19) r + (cid:98) λ r + (cid:98) λ , (cid:98) U (3)1 ( u b ) := C u b . We will determine the relations between the constants which allow the smooth fitting of (cid:98) U . First we impose (cid:98) U (2)1 ( (cid:98) b ) = (cid:98) U (3)1 ( (cid:98) b ) and we get C (cid:18)(cid:98) b r + (cid:98) λ r + (cid:98) λ (cid:19) r + (cid:98) λ r + (cid:98) λ = C (cid:98) b . It can be checked that this relation between C and C ensures also that ( (cid:98) U (2)1 ) (cid:48) ( (cid:98) b ) = ( (cid:98) U (3)1 ) (cid:48) ( (cid:98) b ) . Next, define x as the pointsuch that (cid:98) U (1)1 ( x ) = (cid:98) b , i.e. x = (cid:98) b C (cid:18) r + (cid:98) λ r + (cid:98) λ (cid:19) + Br + (cid:98) λ . Also, define x as the point such that (cid:98) U (2)1 ( x ) = (cid:98) b , i.e. x = (cid:18) (cid:98) b C (cid:19) r + (cid:98) λ r + (cid:98) λ + Br + (cid:98) λ . We impose x = x and we get (cid:98) b C (cid:18) r + (cid:98) λ r + (cid:98) λ (cid:19) = (cid:18) (cid:98) b C (cid:19) r + (cid:98) λ r + (cid:98) λ , and this relation ensures also that ( (cid:98) U (1)1 ) (cid:48) ( x ) = ( (cid:98) U (2)1 ) (cid:48) ( x ) . Expressing both C and C in terms of C we get (cid:98) U (3)1 ( u b ) = C u b ,and (cid:98) U (1)1 ( u b ) = C r + (cid:98) λ r + (cid:98) λ (cid:18) u b − Br + (cid:98) λ (cid:19) + Br + (cid:98) λ , (cid:98) U (2)1 ( u b ) = C (cid:98) b (cid:98) λ − (cid:98) λ r + (cid:98) λ (cid:18) r + (cid:98) λ r + (cid:98) λ (cid:19) r + (cid:98) λ r + (cid:98) λ (cid:18) u b − Br + (cid:98) λ (cid:19) r + (cid:98) λ r + (cid:98) λ . We pursue with the
Proof of Lemma 4.3.
For C > , define the following modification (cid:98) U C ,(cid:63) of (cid:98) U C (cid:98) U C ,(cid:63) ( u b ) := (cid:98) U C ( u b ) , u b ≤ x C ,(cid:63) ,ρ g ρ b ( u b − x C ,(cid:63) ) + (cid:98) U C ( x C ,(cid:63) ) , u b ≥ x C ,(cid:63) , where x C ,(cid:63) := inf (cid:26) u b ∈ (cid:20) Br + (cid:98) λ , + ∞ (cid:19) : (cid:0) (cid:98) U C (cid:1) (cid:48) ( u b ) ≤ ρ g ρ b (cid:27) . The function (cid:98) U C ,(cid:63) is continuously differentiable, solves the diffusion equation in [ B/ ( r + (cid:98) λ ) , x C ,(cid:63) ) and satisfies (cid:0) (cid:98) U C ,(cid:63) (cid:1) (cid:48) = ρ g /ρ b in ( x C ,(cid:63) , ∞ ) . In the following we will study for which values of C this function indeed solves the HJB equation. First of all, if C r + (cid:98) λ r + (cid:98) λ ≤ ρ g ρ b , we have that x C ,(cid:63) = Br + (cid:98) λ , (cid:98) U C ,(cid:63) ( u b ) = ρ g ρ b (cid:32) u b − Br + (cid:98) λ (cid:33) + Br + (cid:98) λ , (cid:16) (cid:98) U C ,(cid:63) (cid:17) (cid:48) ( u b ) ρ b − ρ g = 0 , so that we need to check that for every u b in [ B/ ( r + (cid:98) λ ) , ∞ ) r (cid:98) U C ,(cid:63) ( u b ) − (cid:0) (cid:98) U C,(cid:63) (cid:1) (cid:48) ( u b ) (cid:0) ru b − Bk b + u b (cid:98) λ k b (cid:1) + (cid:98) U C,(cid:63) ( u b ) (cid:98) λ k g − Bk g ≥ . Take u b > (cid:98) b . Then k g = k b = 0 , and we have r (cid:98) U C ,(cid:63) ( u b ) − (cid:0) (cid:98) U C ,(cid:63) (cid:1) (cid:48) ( u b ) (cid:0) ru b − Bk b + u b (cid:98) λ k b (cid:1) + (cid:98) U C ,(cid:63) ( u b ) (cid:98) λ k g − Bk g = r (cid:18) ρ g ρ b (cid:18) u b − Br + (cid:98) λ (cid:19) + Br + (cid:98) λ (cid:19) − ρ g ρ b (cid:0) ( r + (cid:98) λ ) u b (cid:1) + (cid:98) λ (cid:18) ρ g ρ b (cid:18) u b − Br + (cid:98) λ (cid:19) + Br + (cid:98) λ (cid:19) = ( r + (cid:98) λ ) Br + (cid:98) λ (cid:18) − ρ g ρ b (cid:19) < . Hence (cid:98) U C ,(cid:63) is not a solution of (4.12). − If (cid:16) ρ g ρ b (cid:17) r + (cid:98) λ r + (cid:98) λ < C ≤ ρ g ρ b , then x C ,(cid:63) = (cid:98) b r + (cid:98) λ r + (cid:98) λ (cid:0) C ρ b /ρ g (cid:1) r + (cid:98) λ (cid:98) λ − (cid:98) λ + Br + (cid:98) λ . Take u b > (cid:98) b , then k g = k b = 0 and r (cid:98) U C,(cid:63) ( u b ) − (cid:0) (cid:98) U C ,(cid:63) (cid:1) (cid:48) ( u b ) (cid:0) ru b − Bk b + u b (cid:98) λ k b (cid:1) + (cid:98) U C,(cid:63) ( u b ) (cid:98) λ k g − Bk g = ( r + (cid:98) λ ) (cid:18)(cid:98) b C r + (cid:98) λ (cid:98) λ − (cid:98) λ (cid:18) ρ b ρ g (cid:19) r + (cid:98) λ (cid:98) λ − (cid:98) λ (cid:98) λ − (cid:98) λ r + (cid:98) λ − ρ g ρ b Br + (cid:98) λ (cid:19) ≤ ( r + (cid:98) λ ) (cid:18)(cid:98) b ρ g ρ b (cid:98) λ − (cid:98) λ r + (cid:98) λ − Br + (cid:98) λ ρ g ρ b (cid:19) = 0 . The inequality is strict if C < ρ g ρ b so the only value of C such that (cid:98) U C ,(cid:63) solves the HJB equation is C = ρ g ρ b . − For large values of C , i.e. C > ρ g ρ b , we have that x C ,(cid:63) = + ∞ and then (cid:98) U C ,(cid:63) = (cid:98) U C . We exclude this case becausethese functions do not satisfy condition (4.13).We end this section with the Proof of Proposition 4.3.
The proof is by induction. For j = 1 the result is proved in Step 2, so we take any j > and assumethat (cid:98) U (cid:63)j − solves its corresponding diffusion equation. We will need to consider three different cases to prove that (cid:98) U (cid:63)j solves theequation (4.15). In each one of them we prove that the supremum in the right–hand side of (4.15) is attained with θ = 0 , so thatthe diffusion equation takes the same form as the one in the case with one loan left. Then, it follows from the analysis in Step 2that its solution satisfies also the variational inequality (4.9). − Case : u b < (cid:98) b j , (cid:98) U (cid:63)j ( u b ) < (cid:98) b j .In this case for any ( θ, h ) ∈ (cid:98) C j , we have that k g = k b = j . To ease notations, define c j ( u b ) := (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:0) ru b − Bj + u b (cid:98) λ SH j (cid:1) .Then the term inside the supremum in (4.15) becomes c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ SH j + Bj + θ (cid:98) λ SH j (cid:0) (cid:98) U (cid:63)j − ( u b − h ) − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b )( u b − h ) (cid:1) , and the optimal choice of θ in this case is (uniquely) because thanks to Corollary D.1 we have (cid:98) U (cid:63)j − ( u b − h ) − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b )( u b − h ) < . − Case : u b < (cid:98) b j , (cid:98) U (cid:63)j ( u b ) ≥ (cid:98) b j .In this case k b = j for every ( θ, h ) ∈ (cid:98) C j . The term inside the supremum in (4.15) becomes c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ k g j + Bk g + θ (cid:0) (cid:98) U (cid:63)j − ( u b − h ) (cid:98) λ k g j − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) λ SH j ( u b − h ) (cid:1) . Define the following sets (cid:98) C j := (cid:8) ( θ, h ) ∈ (cid:98) C j : (cid:98) U (cid:63)j ( u b ) − θ (cid:98) U (cid:63)j − ( u b − h ) ≥ (cid:98) b j (cid:9) , (cid:98) C jj := (cid:8) ( θ, h ) ∈ (cid:98) C j : (cid:98) U (cid:63)j ( u b ) − θ (cid:98) U (cid:63)j − ( u b − h ) < (cid:98) b j (cid:9) , and note that k g = 0 for every ( θ, h ) ∈ (cid:98) C j and k g = j for every ( θ, h ) ∈ (cid:98) C jj . Also, the pair (0 , h ) belongs to (cid:98) C j for everyfeasible h . If ( θ, h ) ∈ (cid:98) C j we have c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ k g j + Bk g + θ (cid:0) (cid:98) U (cid:63)j − ( u b − h ) (cid:98) λ k g j − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) λ SH j ( u b − h ) (cid:1) = c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ j + θ (cid:0) (cid:98) U (cid:63)j − ( u b − h ) (cid:98) λ j − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) λ SH j ( u b − h ) (cid:1) ≤ c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ j , where the inequality is due to Corollary D.1. • If ( θ, h ) ∈ (cid:98) C jj we have c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ k g j + Bk g + θ (cid:0) (cid:98) U (cid:63)j − ( u b − h ) (cid:98) λ k g j − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) λ SH j ( u b − h ) (cid:1) = c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ SH j + Bj + θ (cid:0) (cid:98) U (cid:63)j − ( u b − h ) (cid:98) λ SH j − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) λ SH j ( u b − h ) (cid:1) < c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ SH j + Bj = c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ SH j + (cid:98) b j ( (cid:98) λ SHj − (cid:98) λ j ) ≤ c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ SH j + (cid:98) U (cid:63)j ( u b ) (cid:0)(cid:98) λ SH j − (cid:98) λ j (cid:1) = c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ j , where the first inequality is a consequence of Corollary D.1 and the second one holds because (cid:98) U (cid:63)j ( u b ) ≥ (cid:98) b j . So we conclude thatthe optimal value for θ in this case is also 0 (uniquely). − Case : u b ≥ (cid:98) b j , (cid:98) U (cid:63)j ( u b ) ≥ (cid:98) b j .Thanks to Proposition D.2 , we know that there are only three possibilities for the value of ( k b , k g ) . Define the sets (cid:98) C , j := (cid:110) ( θ, h ) ∈ (cid:98) C j : u b − θ ( u b − h ) ≥ (cid:98) b j , (cid:98) U (cid:63)j ( u b ) − θ (cid:98) U (cid:63)j − ( u b − h ) ≥ (cid:98) b j (cid:111) , (cid:98) C j, j := (cid:110) ( θ, h ) ∈ (cid:98) C j : u b − θ ( u b − h ) < (cid:98) b j , (cid:98) U (cid:63)j ( u b ) − θ (cid:98) U (cid:63)j − ( u b − h ) ≥ (cid:98) b j (cid:111) , (cid:98) C j,jj := (cid:110) ( θ, h ) ∈ (cid:98) C j : u b − θ ( u b − h ) < (cid:98) b j , (cid:98) U (cid:63)j ( u b ) − θ (cid:98) U (cid:63)j − ( u b − h ) < (cid:98) b j (cid:111) . Then, ( k b , k g ) = (0 , for every ( θ, h ) ∈ (cid:98) C , j , ( k b , k g ) = ( j, for every ( θ, h ) ∈ (cid:98) C j, j and ( k b , k g ) = ( j, j ) for every ( θ, h ) ∈ (cid:98) C j,jj .Also, (0 , h ) belongs to (cid:98) C , j for any feasible h . • If ( θ, h ) ∈ (cid:98) C , j then the term inside the supremum in (4.15) is, because of Corollary D.1, equal to (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) u b (cid:0) r + (cid:98) λ j (cid:1) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ j + θ (cid:98) λ j (cid:0) (cid:98) U (cid:63)j − ( u b − h ) − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b )( u b − h ) (cid:1) ≤ (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) u b (cid:0) r + (cid:98) λ j (cid:1) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ j , • If ( θ, h ) ∈ (cid:98) C j, j , then h < (cid:98) b j and u b − (cid:98) b j u b − h < θ ≤ (cid:98) U (cid:63)j ( u b ) − (cid:98) b j (cid:98) U (cid:63)j − ( u b − h ) . The term in the supremum in (4.15) is c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ j + θ (cid:0) (cid:98) U (cid:63)j − ( u b − h ) (cid:98) λ j − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) λ SH j ( u b − h ) (cid:1) < c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ j + (cid:18) u b − (cid:98) b j u b − h (cid:19)(cid:0) (cid:98) U (cid:63)j − ( u b − h ) (cid:98) λ j − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) λ SH j ( u b − h ) (cid:1) ≤ c j ( u b ) − (cid:98) U (cid:63)j ( u b ) λ j + ( u b − (cid:98) b j ) (cid:0)(cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) λ j − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) λ SH j (cid:1) = (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:0) ru b + u b (cid:98) λ j (cid:1) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ j . Both inequalities are direct consequences of Corollary D.1. • Finally, if ( θ, h ) ∈ (cid:98) C j,jj , note that h < (cid:98) b j , (cid:98) U (cid:63)j ( u b ) − (cid:98) U (cid:63)j − ( u b − h ) < (cid:98) b j and u b − (cid:98) b j u b − h ≤ (cid:98) U (cid:63)j ( u b ) − (cid:98) b j (cid:98) U (cid:63)j − ( u b − h ) < θ. Then, the term inside the sup in (4.15) becomes c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ SHj + Bj + θ (cid:98) λ SH j (cid:0) (cid:98) U (cid:63)j − ( u b − h ) − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b )( u b − h ) (cid:1) ≤ c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ SH j + Bj + (cid:98) U (cid:63)j ( u b ) − (cid:98) b j (cid:98) U (cid:63)j − ( u b − h ) (cid:98) λ SH j (cid:16) (cid:98) U (cid:63)j − ( u b − h ) − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b )( u b − h ) (cid:17) ≤ c j ( u b ) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ SH j + Bj + (cid:98) λ SH j (cid:18) (cid:98) U (cid:63)j ( u b ) − (cid:98) b j − (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) U (cid:63)j ( u b ) − (cid:98) b jρ g ρ b (cid:19) = c j ( u b ) − (cid:98) b j (cid:98) λ j + (cid:98) λ SH j (cid:18) − ρ b ρ g (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) U (cid:63)j ( u b ) + ρ b ρ g (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:98) b j (cid:19) = (cid:98) λ SH j (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:18) u b − ρ b ρ g (cid:98) U (cid:63)j ( u b ) (cid:19) + (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:18) ru b + ρ b ρ g (cid:98) λ SH j (cid:98) b j − Bj (cid:19) − (cid:98) λ j (cid:98) b j . he first inequality comes from Corollary D.1 and the second one from the fact that the map h (cid:55)−→ (cid:98) U (cid:63)j − ( u b − h ) / ( u b − h ) isnon–decreasing and constant for large values of h , which implies that (cid:98) U (cid:63)j − ( u b − h ) / ( u b − h ) ≤ ρ g /ρ b . Now we use the explicitform of (cid:98) U (cid:63)j and compute (cid:98) λ SH j (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:18) u b − ρ b ρ g (cid:98) U (cid:63)j ( u b ) (cid:19) + (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:18) ru b + ρ b ρ g (cid:98) λ SH j (cid:98) b j − Bj (cid:19) − (cid:98) λ j (cid:98) b j = ρ g ρ b ru b + (cid:98) λ SH j (cid:98) b j − ρ g ρ b Bj − (cid:98) λ j (cid:98) b j = ρ g ρ b ru b + Bj (cid:18) − ρ g ρ b (cid:19) < ρ g ρ b ru b . The term in the last line corresponds to (cid:0) (cid:98) U (cid:63)j (cid:1) (cid:48) ( u b ) (cid:0) ru b + u b (cid:98) λ j (cid:1) − (cid:98) U (cid:63)j ( u b ) (cid:98) λ j and therefore the optimal θ in this case is also .Observe that in this case every ( θ, h ) ∈ (cid:98) C , j such that u b − h ≥ (cid:98) b j is optimal.We next continue with the Proof of Theorem 4.1.
We divide the proof in steps. • Step : Let us prove first that the SDE (4.19) has a unique solution, keeping in mind that Ψ (cid:63) liquidates the poolimmediately after the first default. We consider two cases: if u b < (cid:98) b I − N t , by right–continuity we can find for every solution of(4.19) some ε ∈ (0 , τ − t ) such that u bs < (cid:98) b I − N t for s ∈ [ t, t + ε ] . Consequently u b solves the ODE du bs = (cid:0) ( r + (cid:98) λ SH I − N t ) u bs − B ( I − N t ) (cid:1) d s, s ∈ [ t, t + ε ] , whose unique solution is given by u bs = e ( r + (cid:98) λ SHI − Nt )( s − t ) (cid:18) u b − B ( I − N t ) r + (cid:98) λ SH I − N t (cid:19) + B ( I − N t ) r + (cid:98) λ SH I − N t , s ∈ [ t, t + ε ] . So, as long as there is no default and the project keeps running u bs will be deterministic until it reaches the value (cid:98) b I − N t . Thatwill eventually happen at time t (cid:63) ( u b ) := t + 1 r + (cid:98) λ SH I − N t log (cid:18) (cid:98) b I − N t ( r + (cid:98) λ I − N t ) u b ( r + (cid:98) λ SH I − N t ) − B ( I − N t ) (cid:19) , and we see from (4.19) that at time t (cid:63) ( u b ) we will have d u bs = 0 , so u bs = (cid:98) b I − N t for every s ∈ [ t (cid:63) ( u b ) , τ ) . In the second case,if u b ≥ (cid:98) b I − N t then (4.19) becomes d u bs = − u bs − d N s , s ∈ [ t, τ ] , and necessarily u bs = u b for every s ∈ [ t, τ ) . This proves theexistence and uniqueness of the solution of (4.19) in both cases. It is then immediate that the first integrability condition in(3.3) is satisfied. • Step : Now we turn to the values of the banks under Ψ (cid:63) . If u b ≥ (cid:98) b I − N t , we know from the previous analysis that u bs = u b ≥ (cid:98) b I − N t for every s ∈ [ t, τ ) , so in this case Ψ (cid:63) is a short–term contract with constant payment, see Section B.1. Using thenotations of that section, since the payment c = u b ( r + (cid:98) λ j ) ρ b is such that c ≥ ¯ c b ≥ ¯ c g both banks will always work, the value function ofthe bad bank is U bt (Ψ (cid:63) ) = ρ b c/ ( r + (cid:98) λ I − N t ) = u b and the one of the good bank is U gt (Ψ (cid:63) ) = ρ g c/ ( r + (cid:98) λ I − N t ) = ρ g /ρ b u b = (cid:98) U (cid:63)I − N t ( u b ) .In the case where u b < (cid:98) b I − N t , Ψ (cid:63) is a short–term contract with delay t (cid:63) ( u b ) and constant payment, see Section B.2. Using thenotations of that section, since c = ¯ c b the bad bank will always shirk and her value function is U bt (Ψ (cid:63) ) = ρ b c e − ( r + (cid:98) λ SH I − Nt ) t (cid:63) ( u b ) r + (cid:98) λ SH I − N t + Br + (cid:98) λ SH I − N t = u b . For the good bank we have two sub–cases. First, if u b ∈ [ x (cid:63)I − N t , (cid:98) b I − N t ) then ¯ t g ( c ) ≥ t (cid:63) ( u b ) , so the good bank will always workand her value function is U gt (Ψ (cid:63) ) = ρ g ρ b (cid:98) b (cid:98) λ SH I − Nt − (cid:98) λ I − Ntr + (cid:98) λ SH I − Nt I − N t (cid:18) r + (cid:98) λ SH I − N t r + (cid:98) λ I − N t (cid:19) r + (cid:98) λ I − Ntr + (cid:98) λ SH I − Nt (cid:18) u b − B ( I − N t ) r + (cid:98) λ SH I − N t (cid:19) r + (cid:98) λ I − Ntr + (cid:98) λ SH I − Nt = (cid:98) U (cid:63)I − N t ( u b ) . If u b ∈ (cid:2) Br + (cid:98) λ SH I − Nt , x (cid:63)I − N t (cid:1) then ¯ t g ( c ) < t (cid:63) ( u b ) , so the good bank will start working at time t (cid:63) ( u b ) and her value function is U gt (Ψ (cid:63) ) = ρ g ρ b r + (cid:98) λ SH I − Ntr + (cid:98) λ I − Nt (cid:18) u b − B ( I − N t ) r + (cid:98) λ SH I − N t (cid:19) + B ( I − N t ) r + (cid:98) λ SHI − N t = (cid:98) U (cid:63)I − N t ( u b ) . Step : Since U bt (Ψ (cid:63) ) = u b , we have Ψ (cid:63) ∈ A b ( t, u b ) . Consider now a contract Ψ = (
D, θ, h ,b , h ,b ) ∈ A b ( t, u b ) . Werecall that the value function of the bad bank under Ψ satisfies d U bs (Ψ) = (cid:16) rU bs (Ψ) − Bk (cid:63),bs (Ψ) + (cid:0) h ,bs + h ,bs (1 − θ s ) (cid:1) λ k (cid:63),b (Ψ) s (cid:17) d s − ρ b d D s − h ,bs d N s − h ,bs d H s , with k (cid:63),bs (Ψ) = { h ,bs +(1 − θ s ) h ,bs
Proof of Proposition 4.4.
Notice that the inclusion C t ⊆ C t holds by definition and therefore we only need to prove the reverseinclusion. We will make use of contracts with lump–sum payments to prove that every point in C t belongs to the credible set C t .We start by defining the line with slope ρ g /ρ b which passes through the point ( u b , u g ) = (cid:0) B ( I − N t ) r + λ SH t , B ( I − N t ) r + λ SH t (cid:1) M t ( u b ) := ρ g ρ b u b + Bjr + λ SH t (cid:18) − ρ g ρ b (cid:19) , and the sets C t := (cid:110) ( u b , u g ) ∈ V t × V t : M t ( u b ) ≤ u g ≤ U t ( u b ) (cid:111) , C t := (cid:110) ( u b , u g ) ∈ V t × V t : L t ( u b ) ≤ u g ≤ M t ( u b ) (cid:111) . From Section B.4 in the Appendix, we know that C t ⊆ C j . Indeed, every point from the upper boundary U t belongs to the credibleset, and if we perturb a contract Ψ = ( θ, D ) only by adding a lump–sum payment ε at time t , that is d D Ψ (cid:48) s = { s = t } ε + d D Ψ s ,then the values of the banks under Ψ (cid:48) are U gt (Ψ (cid:48) ) = u g + ερ g and U bt (Ψ (cid:48) ) = u b + ερ b , so ( U bt (Ψ (cid:48) ) , U gt (Ψ (cid:48) )) = ( u b , u g ) + ε ( ρ b , ρ g ) .We use this idea to prove also that C t ⊆ C j . From Proposition 4.2, we know that the graph of L t is contained in C t . Thereforeany point of the following form belongs to C t ( (cid:98) u b , (cid:98) u g ) = ( u b , u g ) + (cid:96) ( ρ b , ρ g ) , (cid:96) ≥ , u g = L t ( u b ) . (D.2)By the geometry of the lower boundary L t , the set of points of the form (D.2) is exactly C t . E Principal’s value function on the boundary of the credible set
E.1 The optimal full–monitoring contract in pure moral hazard
The full-monitoring problem studied in [51], considers that the only acceptable behaviour for the bank, from the social point ofview, is that she never shirks away from her monitoring responsibilities. In other words, only contracts with a recommendationof k = 0 are allowed. In this section there is no adverse selection, so there is only one type of bank, and the main result standsfor both i = b, g , a good bank or a bad bank. The value function of the investor in this sub-problem is given by V pm , t ( R ) := ess sup ( D i ,θ i ) ∈A ,i ( t,R ) E P (cid:20) (cid:90) τt ∧ τ ( I − N s ) µ d s − d D is (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) , where the set of admissible contracts A ,i ( t, R ) is defined for R ≥ b t , by A ,i ( t, R ) := (cid:110) ( θ i , D i ) ∈ Θ × D , s.t. ( θ i , D i ) enforces k = 0 and U it ( θ i , D i ) ≥ R (cid:111) . efine for x > the functions φ ( x ) := (cid:18) x x (cid:19) x − , ψ ( x ) := φ ( x ) − x (1 − x ) φ ( x ) . Let us then define some family of concave functions, unique solutions to the following system of ODEs (cid:16) ru + (cid:98) λ j (cid:98) b j (cid:17) ( v ij ) (cid:48) ( u ) + jµ − (cid:98) λ j (cid:32) v ij ( u ) − u − (cid:98) b j (cid:98) b j − v ij − ( (cid:98) b j − ) (cid:33) = 0 , u ∈ (cid:16)(cid:98) b j , (cid:98) b j + (cid:98) b j − (cid:105) , (cid:16) ru + (cid:98) λ j (cid:98) b j (cid:17) ( v ij ) (cid:48) ( u ) + jµ − (cid:98) λ j (cid:16) v ij ( u ) − v ij − ( u − (cid:98) b j ) (cid:17) = 0 , u ∈ (cid:16)(cid:98) b j + (cid:98) b j − , γ ij (cid:105) ,ρ i ( v ij ) (cid:48) ( u ) + 1 = 0 , u > γ ij , (E.1)with initial values γ i := (cid:98) b and v i ( u ) := v i − ρ i ( u − (cid:98) b ) , u ≥ (cid:98) b , v i := µ (cid:98) λ − (cid:98) b ( r + (cid:98) λ ) ρ i (cid:98) λ , and where for j ≥ , γ ij is defined recursively by r/ (cid:98) λ j − ∈ ∂v ij − ( γ ij − (cid:98) b j ) , (E.2)where ∂v ij − is the super–differential of the concave function v ij − . The main result of [51] is Theorem E.1.
Assume that the (cid:0)(cid:98) λ j (cid:1) ≤ j ≤ I satisfy the following recursive conditions for j ≥ r (cid:98) λ j − ≤ v ij − (cid:0)(cid:98) b j − (cid:1)(cid:98) b j − and (cid:16) ( v ij − ) (cid:48) (cid:16)(cid:98) b j − (cid:17)(cid:17) + (cid:98) b j − v ij − (cid:16)(cid:98) b j − (cid:17) ≤ ψ (cid:18) r (cid:98) λ j (cid:19) . Then, under Assumption 2.1, the system (E.1) is well–posed and we have V pm , t ( R ) = sup u t ≥ R v iI − N t ( u t ) , where ( u s ) s ≥ t is defined as the unique solution to the SDE on [ t, τ )d u s = (cid:16) ru s + λ I − N s (cid:98) b I − N s (cid:17) d s − ρ i d D (cid:63),is − (cid:16) { u s ∈ [ (cid:98) b I − Ns , (cid:98) b I − Ns − + (cid:98) b I − Ns ) } ( u s − (cid:98) b I − N s − ) + (cid:98) b I − N s { u s ∈ [ (cid:98) b I − Ns + (cid:98) b I − Ns − ,γ iI − Ns ) } (cid:17) d N s − (cid:16) { u s ∈ [ (cid:98) b I − Ns , (cid:98) b I − Ns − + (cid:98) b I − Ns ) } (cid:98) b I − N s − + ( u s − (cid:98) b I − N s ) { u s ∈ [ (cid:98) b I − Ns + (cid:98) b I − Ns − ,γ iI − Ns ) } (cid:17) d H s , with initial value u t at t , and where we defined for s ∈ [ t, τ ) and j = 1 , . . . , ID (cid:63),is := { s = t } ( u t − γ iI − N t ) + ρ i + (cid:90) st δ I − N r i ( u r )d r, θ (cid:63)s := θ I − N s i ( u s ) ,δ ji ( u ) := { u = γ ij } (cid:98) λ j (cid:98) b j + rγ ij ρ i , θ ji ( u ) := { u ∈ [ (cid:98) b j , (cid:98) b j − + (cid:98) b j ) } u − (cid:98) b j (cid:98) b j − + { u ∈ [ (cid:98) b j + (cid:98) b j − ,γ ij ) } . E.2 Proofs of the main results
We start this section with the
Proof of Proposition 5.1.
Consider any time t ≥ and take any u b,c ≥ C ( I − N t ) , as well as some Ψ g ∈ (cid:98) A g ( t, (cid:98) L I − N t ( u b,c ) , u b,c ) .From Lemma C.2, we know that the components of Ψ g must satisfy θ g ≡ and that both banks shirk under Ψ g . The paymentsdetermine the utility of the banks and the following holds by definition E P k SH (cid:20) (cid:90) τ I t e − r ( s − t ) d D gs (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = u b,c − C ( I − N t ) ρ b . Besides, the utility of the investor under the contract Ψ g is E P k SH (cid:20) (cid:90) τ I t ( µ ( I − N s )d s − d D gs ) (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = I − (cid:88) i = N t µ ( I − i ) (cid:98) λ SH I − i − E P k SH (cid:20) (cid:90) τ I t d D gs (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) . Now, observe that E P k SH (cid:20) (cid:90) τ I t d D gs (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) ≥ E P k SH (cid:20) (cid:90) τ I t e − r ( s − t ) d D gs (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = u b,c − C ( I − N t ) ρ b , nd the equality holds if and only if D g has a jump at time t of size u b,c − C ( I − N t ) ρ b and dD gs = 0 for every s > t . That means thatit is optimal for the investor to use a contract with an initial lump–sum payment and to pay nothing afterwards. Consequently,the value function of the investor on the lower boundary is given by V L ,gt ( u b,c ) = I − (cid:88) i = N t µ ( I − i ) (cid:98) λ SH I − i − (cid:18) u b,c − C ( I − N t ) ρ b (cid:19) . We continue this section with the
Proof of Proposition 5.2.
Consider any time t ≥ . Take any u b,c ∈ [ c ( I − N t , , C ( I − N t )) , and Ψ g ∈ (cid:98) A g ( t, u b,c , u b,c ) . FromLemma C.1, we know that the components of Ψ g must satisfy d D gs = 0 for all s ≥ t and that both banks will shirk under thiscontract. Then, θ g determines the continuation utilities of the banks in the following way u b,c = E P k SH (cid:20) (cid:90) τt e − r ( s − t ) B ( I − N s )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) , so in this case, the problem (5.2) reduces to ( P ) sup θ ∈ Θ E P k SH (cid:20) (cid:90) τt µ ( I − N s )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) , s.t E P k SH (cid:20) (cid:90) τt e − r ( s − t ) B ( I − N s )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = u b,c . Next, we rewrite the objective function in a more convenient way E P k SH (cid:20) (cid:90) τt µ ( I − N s )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = µ ( I − N t ) E P k SH (cid:2) τ N t +1 − t (cid:12)(cid:12) G t (cid:3) + I − (cid:88) i = N t +1 µ ( I − i ) E P k SH (cid:2) { τ>τ i } ( τ i +1 − τ i ) (cid:12)(cid:12) G t (cid:3) = µ ( I − N t ) (cid:98) λ SH I − N t + I − (cid:88) i = N t +1 µ ( I − i ) E P k SH (cid:20) E P k SH (cid:2) { τ>τ i } (cid:12)(cid:12) G τ i (cid:3) E P k SH (cid:2) τ i +1 − τ i (cid:12)(cid:12) G τ i (cid:3)(cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = µ ( I − N t ) (cid:98) λ SH I − N t + I − (cid:88) i = N t +1 µ ( I − i ) (cid:98) λ SH I − i E P k SH [ θ τ i |G t ] . We do the same with the constraint E P k SH (cid:20) (cid:90) τt e − r ( s − t ) B ( I − N s )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = E P k SH (cid:20) (cid:90) τ Nt +1 t B ( I − N t )e − r ( s − t ) d s + I − (cid:88) i = N t +1 { τ>τ i } (cid:90) τ i +1 τ i e − r ( s − t ) B ( I − i )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = B ( I − N t ) r + (cid:98) λ SH I − N t + I − (cid:88) i = N t +1 B ( I − i ) r E P k SH (cid:20) E P k SH (cid:104) { τ>τ i } (cid:0) e − r ( τ i − t ) − e − r ( τ i +1 − t ) (cid:1)(cid:12)(cid:12)(cid:12) G τ i (cid:105)(cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = B ( I − N t ) r + (cid:98) λ SH I − N t + I − (cid:88) i = N t +1 B ( I − i ) r + (cid:98) λ SH I − i E P k SH (cid:2) θ τ i e − r ( τ i − t ) (cid:12)(cid:12) G t (cid:3) . So we obtain the following expression for our problem ( P ) sup θ ∈ Θ µ ( I − N t ) (cid:98) λ SH I − N t + I − (cid:88) i = N t +1 µ ( I − i ) (cid:98) λ SH I − i E P k SH [ θ τ i |G t ] , s.t B ( I − N t ) r + (cid:98) λ SH I − N t + I − (cid:88) i = N t +1 B ( I − i ) r + (cid:98) λ SH I − i E P k SH (cid:2) θ τ i e − r ( τ i − t ) (cid:12)(cid:12) G t (cid:3) = u b,c . We do not know how to solve ( P ) directly, so we will define its dual problem, characterise its solution and show that the dualitygap is zero. In order to do that, we define the Lagrangian function L : Θ × R × Ω −→ R as follows L ( θ, ν, ω ) := − µ ( I − N t ( ω )) (cid:98) λ SH I − N t ( ω ) − I − (cid:88) i = N t ( ω )+1 µ ( I − i ) (cid:98) λ SH I − i E P k SH [ θ τ i |G t ]( ω )+ ν (cid:18) B ( I − N t ( ω )) r + (cid:98) λ SH I − N t ( ω ) + I − (cid:88) i = N t ( ω )+1 B ( I − i ) r + (cid:98) λ SH I − i E P kS H (cid:2) θ τ i e − r ( τ i − t ) (cid:12)(cid:12) G t (cid:3) ( ω ) − u b,c (cid:19) , and also define the dual function and the dual problem respectively as g ( ν, ω ) := inf θ ∈ Θ L ( θ, ν, ω ) , ( D ) sup ν ∈ R g ( ν, ω ) hen, we have the weak duality inequality (where val denotes the value of the optimisation problem) − val ( P ) = inf θ ∈ Θ sup ν ∈ R L ( θ, ν, ω ) ≥ sup ν ∈ R inf θ ∈ Θ L ( θ, ν, ω ) = val ( D ) . We rewrite the dual function as follows g ( ν, ω ) = − µ ( I − N t ( ω )) (cid:98) λ SH I − N t ( ω ) + ν (cid:18) B ( I − N t ( ω )) r + (cid:98) λ SH I − N t ( ω ) − u b,c (cid:19) + inf θ ∈ Θ I − (cid:88) i = N t ( ω )+1 (cid:90) Ω θ τ i ( (cid:101) ω ) (cid:18) ν B ( I − i ) r + (cid:98) λ SH I − i e − r ( τ i ( (cid:101) ω ) − t ) − µ ( I − i ) (cid:98) λ SH I − i (cid:19) d P SH t,ω ( (cid:101) ω ) , where P SHt,ω is a regular conditional probability distribution for the conditional expectation with respect to the raw (that is to saynot completed) version of G t . We have easily that it is optimal to set the optimal control θ ν to be θ ντ i ( (cid:101) ω ) := (cid:101) ω ∈ A iν ( (cid:101) ω ) , wherethe set A iν is defined by A iν := Ω , if ν < µB r + (cid:98) λ SH I − i (cid:98) λ SH I − i , (cid:26)(cid:101) ω : τ i ( (cid:101) ω ) − t > r ln (cid:18) νB (cid:98) λ SH I − i µ ( r + (cid:98) λ SH I − i ) (cid:19)(cid:27) , if ν ≥ µB r + (cid:98) λ SH I − i (cid:98) λ SH I − i . Therefore, for any ν ∈ R the dual function has the following form, using that the conditional law of τ i − t given G t is the sameas the law of τ i g ( ν, ω ) = − µ ( I − N t ( ω )) (cid:98) λ SH I − N t ( ω ) + ν (cid:18) B ( I − N t ( ω )) r + (cid:98) λ SH I − N t ( ω ) − u b,c (cid:19) + I − (cid:88) i = N t ( ω )+1 (cid:90) ∞ s i ( ν ) (cid:18) νB ( I − i )e − rx r + (cid:98) λ SH I − i − µ ( I − i ) (cid:98) λ SH I − i (cid:19) f τ i ( x )d x. (E.3)It is not difficult to see that g is a continuous and differentiable function. As we want to maximise g in the dual problem, wecompute its derivative with respect to ν and we get g (cid:48) ( ν, ω ) = B ( I − N t ( ω )) r + (cid:98) λ SH I − N t − u b,c + I − (cid:88) i = N t +1 (cid:90) ∞ s i ( ν ) B ( I − i ) r + (cid:98) λ SH I − i e − rx f τ i ( x )d x. Since ν (cid:55)−→ s i ( ν ) is non–decreasing for any i = 1 , . . . , I , g (cid:48) is non–increasing in ν . Furthermore, since u b,c ≥ c ( I − N t , , we havethe limit at + ∞ of g (cid:48) is non–positive, and that its value for small ν is positive because u b,c < C ( I − N t ) and B ( I − N t ( ω )) r + (cid:98) λ SH I − N t + I − (cid:88) i = N t +1 (cid:90) ∞ B ( I − i ) r + (cid:98) λ SH I − i e − rx f τ i ( x )d x = C ( I − N t ) . Therefore, there is a unique value of ν that makes g (cid:48) equal to . Now, we compute for any ν the value of the constraint from theprimal problem for the control θ νI − (cid:88) i = N t +1 B ( I − i ) r + (cid:98) λ SH I − i E P k SH (cid:2) θ ντ i e − r ( τ i − t ) (cid:12)(cid:12) G t (cid:3) = I − (cid:88) i = N t +1 (cid:90) ∞ s i ( ν ) B ( I − i ) r + (cid:98) λ SH I − i e − rx f τ i ( x )d x, so θ ν is feasible in problem ( P ) if and only if g (cid:48) ( ν, ω ) = 0 . Next, we compute for θ ν the value of the objective function in theprimal (minimisation) problem − µ ( I − N t ) (cid:98) λ SH I − N t − I − (cid:88) i = N t +1 µ ( I − i ) (cid:98) λ SH I − i E P k SH t (cid:2) θ ντ i (cid:3) = − µ ( I − N t ) (cid:98) λ SH I − N t − I − (cid:88) i = N t +1 (cid:90) ∞ s i ( ν ) µ ( I − i ) (cid:98) λ SH I − i f τ i ( x )d x. If this quantity is equal to g ( ν, · ) , the duality gap is zero. From (E.3) we see that this happens if and only if ν (cid:18) B ( I − N t ) r + (cid:98) λ SH I − N t − u b,c + I − (cid:88) i = N t +1 (cid:90) ∞ s i ( ν ) B ( I − i ) r + (cid:98) λ SH I − i e − rx f τ i ( x )d x (cid:19) = 0 ⇐⇒ νg (cid:48) ( ν, · ) = 0 . We conclude that if ν ∈ R is such that g (cid:48) ( ν ) = 0 then the control θ ν is optimal in the primal problem.We continue with the Proof of Proposition 5.3.
Define the process (cid:96) s = (cid:98) U I − N s ( U b,cs (Ψ g )) − U gs (Ψ g ) and note that (cid:96) s ≥ for every s ≥ . We will provethat (cid:96) t = 0 implies (cid:96) v = 0 for every v ≥ t . Assume thus that (cid:96) t = 0 . Following the same idea as in the proof of Theorem 4.1, we ave for v ≥ t(cid:96) v = I − (cid:88) i = N t (cid:90) τ i +1 ∧ vτ i ∧ v − (cid:16) rU gs (Ψ g ) − Bk (cid:63),gs (Ψ g ) + [ h ,gs + (1 − θ gs ) h ,gs ] λ k (cid:63),g (Ψ g ) s (cid:17) d s + I − (cid:88) i = N t (cid:90) τ i +1 ∧ vτ i ∧ v (cid:98) U (cid:48) I − i ( U b,cs (Ψ g )) (cid:16) rU b,cs (Ψ g ) − Bk (cid:63),b,cs (Ψ g ) + λ k (cid:63),b,c (Ψ g ) I − i ( h ,b,cs + (1 − θ gs ) h ,b,cs ) (cid:17) d s + I − (cid:88) i = N t (cid:90) τ i +1 ∧ vτ i ∧ v (cid:16) h ,gs + (cid:98) U I − i − ( U b,cs − (Ψ g ) − h ,b,cs ) − (cid:98) U I − i ( U b,cs − (Ψ g )) (cid:17) d N s + I − (cid:88) i = N t (cid:90) τ i +1 ∧ vτ i ∧ v (cid:16) h ,gs − (cid:98) U I − i − ( U b,cs − (Ψ g ) − h ,b,cs ) (cid:17) d H s + (cid:16) ρ g − ρ b (cid:98) U (cid:48) I − i ( U b,cs (Ψ g )) (cid:17) d D gs . Since the functions (cid:98) U i solve the system of HJB equations (4.9), and (cid:0) ρ g − ρ b (cid:98) U (cid:48) i ( U b,cs (Ψ g )) (cid:1) d D gs ≤ for every s , we have (cid:96) v ≤ I − (cid:88) i = N t (cid:90) τ i +1 ∧ vτ i ∧ v (cid:16) r (cid:98) U I − i ( U b,cs (Ψ g )) − rU gs (Ψ g ) − [ h ,gs + (1 − θ gs ) h ,gs ] λ k (cid:63),g (Ψ g ) s (cid:17) d s − I − (cid:88) i = N t (cid:90) τ i +1 ∧ vτ i ∧ v λ k (cid:63),g (Ψ g ) s (cid:16) θ s (cid:98) U I − i − ( U b,cs − (Ψ g ) − h ,b,cs ) − (cid:98) U I − i ( U b,cs (Ψ g )) (cid:17) d s + I − (cid:88) i = N t (cid:90) τ i +1 ∧ vτ i ∧ v (cid:16) h ,gs + (cid:98) U I − i − ( U b,cs − (Ψ g ) − h ,b,cs ) − (cid:98) U I − i ( U b,cs − (Ψ g )) (cid:17) d N s + (cid:16) h ,gs − (cid:98) U I − i − ( U b,cs − (Ψ g ) − h ,b,cs ) (cid:17) d H s = I − (cid:88) i = N t (cid:90) τ i +1 ∧ vτ i ∧ v (cid:0) r + λ k (cid:63),g s (cid:1)(cid:16)(cid:98) U I − i ( U b,cs (Ψ g )) − U gs (Ψ g ) (cid:17) + (cid:16) h ,gs − (cid:98) U I − i − ( U b,cs (Ψ g ) − h ,b,cs ) (cid:17) θ gs λ k (cid:63),g s d s + I − (cid:88) i = N t (cid:90) τ i +1 ∧ vτ i ∧ v (cid:16) h ,gs + (cid:98) U I − i − ( U b,cs − (Ψ g ) − h ,b,cs ) − (cid:98) U I − i ( U b,cs − (Ψ g )) (cid:17) d N s + (cid:16) h ,gs − (cid:98) U I − i − ( U b,cs − (Ψ g ) − h ,b,cs ) (cid:17) d H s . Recall from Remark 4.4 that on the upper boundary, we have h ,gs = (cid:98) U I − N s − ( U b,cs − (Ψ g )) − (cid:98) U I − N s − − ( U b,cs − (Ψ g ) − h ,b,cs (Ψ g )) , h ,gs = (cid:98) U I − N s − − ( U b,cs − (Ψ g ) − h ,b,cs (Ψ g )) , so that for i = N t the drift of the right–hand side is in [ τ i , τ i +1 ) and the jump at time τ i +1 is also . It is easy to see that thesame happens for every i ∈ { N t , . . . , I } and therefore (cid:96) v ≤ for every v ≥ which means (cid:96) v = 0 for every v ≥ t .We go on with the Proof of Proposition 5.4. ( i ) We have from the proof of Proposition 5.3 that the processes ( θ g , h ,b,c , h ,b,c ) are necessarilymaximisers of the system of HJB equations (4.9). We can go back to the proof of Proposition 4.3, which is based on CorollaryD.1, to observe that for u b,c < (cid:98) b j the optimal θ ∈ C j is uniquely given by θ = 0 . ( ii ) Observe that for every ( t, u b,c , u g ) ∈ [0 , τ ] × (cid:98) V I − N t × (cid:98) V I − N t and Ψ g ∈ (cid:98) A g ( t, u g , u b,c ) we have U b,ct (Ψ g ) ≥ E P k(cid:63),g (Ψ g ) (cid:20) (cid:90) τt e − r ( s − t ) ( ρ b d D gs + Bk (cid:63),gs (Ψ g )d s ) (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21) = ρ b ρ g U gt (Ψ g ) + E P k(cid:63),g (Ψ g ) (cid:20) (cid:90) τt e − r ( s − t ) Bk (cid:63),gs (Ψ g )d s (cid:12)(cid:12)(cid:12)(cid:12) G t (cid:21)(cid:18) − ρ b ρ g (cid:19) ≥ ρ b ρ g U gt (Ψ g ) . Then U b,cs (Ψ g ) = ρ g ρ b U gs (Ψ g ) implies that k (cid:63),gs (Ψ g ) = k (cid:63),b,cs (Ψ g ) = 0 , for every s ∈ [ s , τ ) , and in consequently U b,cs (Ψ g ) = ρ g ρ b U gs (Ψ g ) ≥ b s , for every s ∈ [ s , τ ) . We continue with the
Proof of Proposition 5.5.
We divide the proof in steps. • Step : We start with the region u b,c > (cid:98) b I − N t . Let Ψ g = ( D g , θ g , h ,b,c , h ,b,c ) ∈ A g ( t, u b,c ) be such that U b,ct (Ψ g ) = u b,c ≥ (cid:98) b I − N t , U gt (Ψ g ) = (cid:98) U I − N t ( u b,c ) . From Proposition 5.4 we know that U b,cs (Ψ g ) ≥ (cid:98) b I − N s , k (cid:63),b,c (Ψ g ) = 0 , s ∈ [ t, τ ) . herefore, Problem (5.3) is equivalent to V U ,gt ( u b,c ) = sup Ψ g ∈A g ( t,u b,c ) E P (cid:20) (cid:90) τt µ ( I − N s )d s − (cid:90) τt d D gs (cid:21) , s.t U b,cs (Ψ g ) ≥ (cid:98) b I − N s , s ∈ [ t, τ ) , E P (cid:20) (cid:90) τt e − r ( s − t ) d D gs (cid:21) = u b,c ρ b . This is exactly the problem considered in Pagès and Possamaï [51] (see Theorem 3.15). We conclude that V U ,gt ( u b,c ) = v bI − N t ( u b,c ) . • Step : For the rest of the upper boundary, observe that the system of HJB equations associated to (5.3) is is given by (cid:98) V ≡ , and for any ≤ j ≤ I min (cid:40) − sup ( θ,h ,h ) ∈ C U ,j (cid:40) (cid:98) V (cid:48) j ( u b,c ) (cid:0) ru b,c − Bk b,c + ( h + (1 − θ ) h ) (cid:98) λ k b,c j (cid:1) + µj + (cid:98) λ k g j θ (cid:98) V j − ( u b,c − h ) − (cid:98) λ k g j (cid:98) V j ( u b,c ) (cid:41) , (cid:98) V (cid:48) j ( u b,c ) + 1 ρ b (cid:41) = 0 , (E.4)for every u b,c ≥ Bjr + (cid:98) λ SH j , with the boundary condition (cid:98) V j ( Bj/ ( r + (cid:98) λ SH j )) = µj/ (cid:98) λ SH j , and where k b,c := j { h +(1 − θ ) h < (cid:98) b j } , k g := j { (cid:98) U (cid:63)j ( u b,c ) − θ (cid:98) U (cid:63)j − ( u b,c − h ) < (cid:98) b j } , and the set of constraints C U ,j determined by Proposition 5.4 is defined by C U ,j := (cid:26) ( θ, h , h ) ∈ [0 , × R : h + h = u b,c , h ≥ B ( j − r + (cid:98) λ SH j − , θ { u b,c < (cid:98) b j } = ( k b,c + k g ) { u b,c ≥ (cid:98) b j } = 0 (cid:27) . Then, for any u b,c < (cid:98) b j , the diffusion equation in (E.4) reduces to the ODE (cid:98) V (cid:48) j ( u b,c ) (cid:0)(cid:0) r + (cid:98) λ SH j (cid:1) u b,c − Bj (cid:1) − (cid:98) V j ( u b,c ) (cid:98) λ k g j + µj, (E.5)with the boundary condition (cid:98) V j (cid:0) Bjr + (cid:98) λ SH j (cid:1) = µj (cid:98) λ SH j . If u b,c < x (cid:63)j , we get that (cid:98) V j ( u b,c ) = µj (cid:98) λ SH j + C (cid:32)(cid:32) r + (cid:98) λ SH j (cid:98) λ SH j (cid:33) u b,c − Bj (cid:98) λ SH j (cid:33) (cid:98) λ SH jr + (cid:98) λ SH j , for some C ∈ R . If u b,c ∈ (cid:104) x (cid:63)j , (cid:98) b j (cid:17) , equation (E.5) is solved by (cid:98) V j ( u b,c ) = µj (cid:98) λ j + C (cid:32)(cid:32) r + (cid:98) λ SHj (cid:98) λ j (cid:33) u b,c − Bj (cid:98) λ j (cid:33) (cid:98) λ jr + (cid:98) λSHj , for some C ∈ R . The values of C and C for which the solution of equation (E.5) is continuous are C = µj (cid:98) λ j − µj (cid:98) λ SH j + (cid:0) ρ b ρ g (cid:1) (cid:98) λ jr + (cid:98) λ j (cid:0) v bj ( (cid:98) b j ) − µj (cid:98) λ j (cid:1)(cid:0) ρ b ρ g (cid:1) (cid:98) λ SH jr + (cid:98) λ j (cid:0) (cid:98) b j ( r + (cid:98) λ j ) (cid:98) λ SH j (cid:1) (cid:98) λ SH jr + (cid:98) λ SH j , C = (cid:18) v bj ( (cid:98) b j ) − µj (cid:98) λ j (cid:19)(cid:18)(cid:98) b j r + (cid:98) λ j (cid:98) λ j (cid:19) − (cid:98) λ jr + (cid:98) λ SH j . It follows from the properties of the map v bj , that the resulting function (cid:98) V j is a concave map with slope greater than − /ρ b andtherefore the family { (cid:98) V j } ≤ j ≤ I is a solution of the system of HJB equations (E.4). It can be proved similarly as in the proof ofTheorem 4.1 (see also Theorem 3.15 in [51]), that the verification result holds for this family of functions. We therefore omit theproof of this result.We continue with the Proof of Proposition 5.6.
By definition we have the set equality (cid:98) A g ( t, (cid:98) L I − N t ( u b ) , u b ) = (cid:98) A b ( t, (cid:98) L I − N t ( u b ) , u b ) . From Lemmas C.1and C.2 we know that for every Ψ b ∈ (cid:98) A b ( t, (cid:98) L I − N t ( u b ) , u b ) , both agents always shirk under Ψ b , therefore the objective functionsin the definitions of V L ,gt ( u b ) and V L ,bt ( u b ) are also the same and equality holds.We go on with the roof of Proposition 5.7. The proof is identical to the proof of Proposition 5.5, with the only difference that since the principalis hiring the bad agent, for u b < (cid:98) b j the ODE associated to the value function is (cid:98) V (cid:48) j ( u b ) (cid:0)(cid:0) r + (cid:98) λ SH j (cid:1) u b − Bj (cid:1) − (cid:98) V j ( u b ) (cid:98) λ SH j + µj, with the boundary condition (cid:98) V j (cid:0) Bjr + (cid:98) λ SH j (cid:1) = µj (cid:98) λ SH j .We end this section with the Proof of Proposition 5.9.
The payments and the value of θ (cid:63) in the case u b ≥ C ( I − N t ) are a direct consequence of the proof ofProposition 5.1. From the proof of Proposition 5.2 we have that if u b < C ( I − N t ) then θ (cid:63)s = (cid:110) s − t> r ln (cid:16) ν ( ub ) B (cid:98) λ SH I − Ntµ ( r + (cid:98) λ SH I − Nt ) (cid:17)(cid:111) , where ν ( u b ) the solution of the associated dual problem. Since the quantity inside of the logarithm decreases with time, we havethat θ (cid:63) is a process which starts at zero, jumps to one at some instant and keeps constant afterwards. This means that if θ (cid:63) jumps to one at some time s and the project is still running, necessarily the continuation utility of the bad agent is equal to C ( I − N s ) because the project will continue until the last default. F Extensions of the model
F.1 Endogenous reservation utility
Proof of Proposition 6.1.
Define the dynamic version of R i by R it := sup k ∈ K E P k (cid:20) (cid:90) τ I t e − r ( s − t ) ( ρ i µ ( I − N s ) + Bk s )d s (cid:12)(cid:12)(cid:12)(cid:12) F t (cid:21) . Observe that, thanks to our earlier results, we know that the previous expression depends on t only through the value of I − N t .Call then (cid:98) R iI − N t = R it , the value when there are I − N t loans left. The explicit value of (cid:98) R i and the optimal action of the agentof type ρ i in ( τ I − , τ ) were obtained in the study of short–term contracts with constant payments, in section B. Suppose nowthat j > and that the value of (cid:98) R ij − as well as the optimal action of the agent after default τ I − j are known.If the agent decides to monitor all the loans in ( τ I − j , τ I − j +1 ) , his expected utility will be given by u i (0) := ρ i µjr + (cid:98) λ j + (cid:98) λ j r + λ j (cid:98) R ij − . The process h ,i (0) associated to this action is given by h ,i (0) := u i (0) − (cid:98) R ij − = ρ i µjr + (cid:98) λ j − rr + (cid:98) λ j (cid:98) R ij − . Therefore, it is incentive compatible to monitor all the loans in ( τ I − , τ ) if and only if h ,i (0) ≥ b j ⇐⇒ ρ i µj − r (cid:98) R ij − ≥ b j ( r + (cid:98) λ j ) . Similarly, if the agent chooses to shirk in ( τ I − j , τ I − j +1 ) , his expected utility will be equal to u i ( j ) := ρ i µj + Bjr + (cid:98) λ SH j + (cid:98) λ SHj r + (cid:98) λ SH j (cid:98) R ij − . The process h ,i (0) associated to this action is given by h ,i ( j ) := u i ( j ) − (cid:98) R ij − = ρ i µj + Bjr + (cid:98) λ SH j − rr + (cid:98) λ SH j (cid:98) R ij − , and it is incentive compatible to not monitor any loan in ( τ I − , τ ) if and only if h ,i ( j ) < b j ⇐⇒ ρ i µj + Bj − r (cid:98) R ij − < b j ( r + (cid:98) λ SH j ) ⇐⇒ ρ i µj − r (cid:98) R ij − < b j ( r + (cid:98) λ j ) . .2 Unbounded relationship between utilities of the banks A possible extension of our model could rely on a further differentiation between the work of the two banks, i.e. when bothbanks work, the good one would be more efficient in the sense that the associated default intensity is strictly smaller than thatof the bad bank. We can do this by introducing an extra type variable with values m g and m b , with m g < m b and modelling thehazard rate of a non-defaulted loan j at time t , when it is monitored by a bank of type i as α j,it = α I − N t (1 + e j,it m i + (1 − e j,it ) ε ) . Then, if the banks fails to monitor k loans, the default intensity will be λ k,it = α I − N t (( I − N t )(1 + m i ) + ( ε − m i ) k t ) . We did not consider such a situation because it creates a degeneracy, in the sense that the credible set no longer has an upperboundary. Indeed, consider for simplicity the case j = 1 and take any u b ≥ b j , t (cid:63) ≥ and choose the corresponding payment c ( t (cid:63) ) := u b e ( r + (cid:98) λ ,b ) t (cid:63) ( r + (cid:98) λ ,b ) ρ b ≥ b b ( r + (cid:98) λ ,b ) ρ b ≥ b g ( r + (cid:98) λ ,g ) ρ g . Then, under the contract with delay and constant payments given by dD s = c ( t (cid:63) )1 { s>t (cid:63) } ds the bad bank will always work andher value function will be equal to u b (see section B.2). Notice that the optimal strategy for the good bank will be also to workat every time. Then, her value function is equal to u g := u b ρ g ( r + (cid:98) λ ,b ) ρ b ( r + (cid:98) λ ,g ) e ( (cid:98) λ ,b − (cid:98) λ ,g ) t (cid:63) . We see that by increasing t (cid:63) , it is possible to make u g as big as we want and keep fixed the value of the bad bank. This meansthat the credible set will have no upper boundary in the interval [ b b , ∞ ) . Moving to any j > and considering short-termcontracts with delay, with θ = 0 and the analogous payments, we observe the same degeneracy and the credible set will have noupper boundary in the interval [ b bj , ∞ ) .One way out of this problem would be to consider different discount rates for the banks, r b and r g , and assume that thedefault intensities are such that λ ,bt + r b ≤ λ ,gt + r g . However, this complicates things a lot because simple statements that weexpect to be true are very difficult to prove or need assumptions on the parameters of the problem. For example the inequality U gt ( D, θ ) ≥ U bt ( D, θ ) is no longer clear at all. We therefore refrained from going into that direction, and leave it for potentialfuture research.is no longer clear at all. We therefore refrained from going into that direction, and leave it for potentialfuture research.