Prerequisites Almost essential Adverse selection Frank Cowell: Microeconomics August 2006 Contract Design MICROECONOMICS Principles and Analysis Frank Cowell Purpose of contract design Frank Cowell: Microeconomics A step in moving from how we would like to organise the economy… …to what we can actually implement Plenty of examples of this issue: Purpose and nature of the design problem hiring a lawyer employing a manager construct a menu of alternatives… …to induce appropriate choice of action Key: takes account of incomplete information Informational issues Frank Cowell: Microeconomics Two key types of informational problem: Hidden action: the adverse selection problem concerned with unseen attributes… …and unseen effort Here focus on the hidden information problem The moral hazard problem concerned with unseen/unverifiable events… …and unseen effort Hidden information: each is relevant to design question each can be interpreted as a version of “Principal and Agent” How to design a payment system ex ante… …when the quality of the service/good cannot be verified ex ante Attack this in stages: outline a model examine full-information case then contrast this with asymmetric information Overview... Contract design Frank Cowell: Microeconomics Design principles Roots in social choice and asymmetric information Model outline Full information Asymmetric information The essence of the model Frank Cowell: Microeconomics The Principal employs the Agent to produce some output But Agent may be of unknown type The Principal designs a payment scheme takes into account that type is unknown… …and that one type of Agent might try to masquerade as another Provides an illustration of second best problem type here describes Agent’s innate productivity …how much output per unit of effort because of delegation under imperfect information may have to forgo some output … “Agency cost” Use a parable to explain how it works A parable: paying a manager Frank Cowell: Microeconomics An owner hires a manager A problem of hidden information It makes sense to pay the manager according to talent But how talented is the manager? Similar to adverse selection problem But here with a monopolist – the owner The nature of the design problem Owner acts as designer Wants to maximise expected profits Wants to ensure that manager acts in accordance with this aim “Mechanism” here is the design of contract (s) Frank Cowell: Microeconomics The employment contract: information Perhaps talent shows Perhaps it doesn’t Ability can be observed or … …costlessly verified Full-information solution Ability cannot be observed in advance of the contract Will low ability applicants misrepresent themselves? Will high ability applicants misrepresent themselves? The approach Examine full-information solution Get rules for contract design in this case Remodel the problem for the second-best case Modify contract rules Overview... Contract design Frank Cowell: Microeconomics Design principles A simple ownerand-manager story Model outline Full information Asymmetric information Model basics: owner Frank Cowell: Microeconomics Owner makes first move Has market power designs payment schedule for the manager makes a take-it-or-leave-it offer Can act as a monopolist Appropriates the gains from trade Gets profit after payment to manager: utility (payoff) to owner is just the profit pq - y p: price of output q: amount of output y : payment to manager Model basics: manager Frank Cowell: Microeconomics A manager’s talent and effort determines output: Manager’s preferences q = tz. q : output produced t : the amount of talent z : the effort put in u = y(z) + y u : utility level y : income received y() : decreasing, strictly concave, function equivalently: u = y(q / t) + y. Manager has an outside option u : reservation utility A closer look at manager’s utility The utility function (1) Frank Cowell: Microeconomics Preferences over leisure and income y Indifference curves Reservation utility u = y(z) + y yz(z) < 0 u≥u u 1– z The utility function (2) Frank Cowell: Microeconomics Preferences over leisure and output y Indifference curves Reservation utility u = y(q/t) + y yz(q/t) < 0 u≥u u q Model basics: information Frank Cowell: Microeconomics There are different talent types j = 1,2,… Profits (owner’s payoff) depend on talent: Type j has talent tj Probability of a manger being type j is pj Probability distribution is common knowledge Owner may or may not know type j of a potential manager pqj - yj. qj = tjzj: the output produced by a type j manager zj : effort put in by a type j manager Managers’ preferences are common knowledge Utility function is known Also known that all managers have the same preferences, independent of type Indifference curves: pattern Frank Cowell: Microeconomics Managers of all types have the same preferences Function y() is common knowledge uj = y(zj) + yj. uj = y(qj/tj) + yj. But utility level uj of type j depends on effort zj and payment yj. Take indifference curves in (q, y) space u = y(q/tj) + y. Clearly slope of type j indifference curve depends on tj. Indifference curves of different types cross once only The single-crossing condition Frank Cowell: Microeconomics Preferences over leisure and output y High talent Low talent Those with different talent will have different sloped ICs in this diagram j=b j=a q qa = taza qb = tbzb Overview... Contract design Frank Cowell: Microeconomics Design principles Where talent is known to all… Model outline Full information Asymmetric information Full information: setting Frank Cowell: Microeconomics Owner may be faced with a manager of any type j But owner can observe the type (talent) tj j j j Therefore can observe effort z = q /t Owner prepares menu of such contracts in advance So the contract can be conditioned on effort Offer manager of type j the deal (yj, zj) Aims to maximise expected profits Manager then chooses effort in response Aims to maximise utility This choice is correctly foreseen by the owner designing the contract Full information: problem Frank Cowell: Microeconomics Owner aims to maximise expected profits Expectation is over distribution of types. Maximisation subject to (known) manager behaviour Participation constraint of type j. Choose yj, zj to max Sj pj [ptjzj - yj] subject to yj + y(zj) ≥ uj. Solve this using standard methods for constrained maximum Full information: solution Frank Cowell: Microeconomics Set up standard Lagrangean: First-order conditions: Lagrange multiplier lj for participation constraint on type j. Choose yj, zj, lj to max Sj pj [ptjzj - yj] +Sj lj [yj + y(zj) − uj] lj = pj - yz(z*j) = ptj yj + y(z*j) = uj Interpretation “Price” of constraint is probability of a type j manager MRS = MRT Reservation utility constraint is binding Full-information solution Frank Cowell: Microeconomics y a type’s reservation utility ub _ b type’s reservation utility a type’s contract b type’s contract _ua p y*a Both types get contract where marginal disutility of effort equals marginal product of labour y*b q q*b q*a Full information: conclusions Frank Cowell: Microeconomics “Price” of constraint is probability of getting a type-j manager The outcome is efficient: MRS = MRT …for each type of manager Owner drives manager down to reservation utility complete exploitation owner gets all the surplus Overview... Contract design Frank Cowell: Microeconomics Design principles Where talent is private information Model outline Full information Asymmetric information Asymmetric information: approach Frank Cowell: Microeconomics Full-information contract is simple and efficient However, this version is not very interesting. Problem arises when contract has to be drawn up before talent is known Agent may have an incentive to misrepresent his talents this will impose a constraint on the design of the contract Re-examine the Full-information solution Another look at the FI solution Frank Cowell: Microeconomics y a type’s reservation utility ub _ b type’s reservation utility a type’s contract b type’s contract a type’s utility with b type contract _ua p y*a An a type would like to masquerade as a b type! y*b q q*b q*a Asymmetric information again Frank Cowell: Microeconomics As we have seen a type would want to mimic a b type We can exploit a standard approach to the problem. Assume that the distribution of talent is known. For simplicity take two talent levels qa = taza with probability p b b b q = t z with probability 1-p The “second-best” model Frank Cowell: Microeconomics Participation constraint for the b type: Incentive-compatibility constraint for the a type: yb + y(zb) ≥ ub. Have to offer at least as much as available elsewhere ya + y(qa/ta) ≥ yb + y(qb/ta). Must be no worse off than if took b contract Maximise expected profits a a b b p[pq - y ] + [1-p][pq - y ]. Choose qa, qb, ya, yb to max p[pqa - ya] + [1-p][pqb - yb ] + l [yb + y(qb/tb) - ub] + m [ya + y(qa/ta) - yb - y(qb/ta)] Second-best: results Frank Cowell: Microeconomics Lagrangean is p[pqa - ya] + [1-p][pqb - yb] + l [yb + y(qb/tb) - ub] + m [ya + y(qka/ta) - yb - y(qb/ta)] FOC are: - yz(qa/ta) = pta - yz(qb/tb) = ptb + kp/[1-p] k := yz(qb/tb) - [tb/ta] yz(qb/ta) < 0 Results imply MRSa = MRTa MRSb < MRTb Two types of Agent: contract design Frank Cowell: Microeconomics a type’s reservation utility y b type’s reservation utility b type’s contract incentive-compatibility constraint b type’s contract a contract schedule y~ a y~ b q q~ b q~ a Second-best: lessons Frank Cowell: Microeconomics a-types b-types for high-talent people… …marginal rate of substitution equals marginal rate of transformation no distortion at the top for low-talent people… …MRS is strictly less than MRT Principal will make lower profits than in full-information case this is the Agency cost Summary Frank Cowell: Microeconomics Contract design fundamental to economic relations Asymmetric information raises deep issues: Second-best approach builds these issues into the problem Principal cannot know the productivity of the agent beforehand Agent may have incentive to misrepresent information important not to have a manipulable contract known distribution of types incentive-compatibility constraint Solution satisfies “no-distortion-at-the-top” principle gives no surplus to the lowest productivity type