Regulation Analyzed in two contexts: ◮ Procurement: The firm supplies a good to the government ◮ Regulation: The firm supplies a good to consumers on behalf of government Economics of Information and Contracts ⋆ Screening: Applications ⋆ Defense contracts Utilities (electricity, gas), telecommunications, railroads There are many natural monopolies in utilities, telecommunication, infrastructure, and transportation industries Levent Koçkesen ◮ Koç University ◮ Due to large fixed costs, average cost is decreasing even at large production levels It may be socially efficient to have a monopoly Government needs to regulate to prevent them acting as a monopolist But it doesn’t know the relevant characteristics (e.g., technology) and/or does not observe actions (e.g., cost reducing effort) of the firm Levent Koçkesen (Koç University) Screening: Applications 1 / 27 Regulation ◮ a: a fixed fee b: fraction of costs born by the firm (measures the power of incentives) Possible contracts ◮ ⋆ fixed-price (price-cap): b = 1 ◮ incentive (cost-sharing): 0 < b < 1 ⋆ 40.9% 13.6% 31.4% cost-plus incentive fixed-price Regulation: ◮ historically: cost-of-service more recently: price-cap and incentive regulation firm does not bear any cost low powered: firm does not have an incentive to lower cost ◮ ⋆ 2 / 27 Procurement: 1960 US ◮ cost-plus (cost-of-service): b = 0 ⋆ Screening: Applications Regulation Typically, government uses accounting (cost) data and reimburses a fraction of firm’s monetary expenditures c Its aim is to obtain the highest societal welfare in providing the good We will model it as follows: government pays c and a net transfer t to the firm t = a − bc ◮ Levent Koçkesen (Koç University) government does not reimburse any cost; pays only a fixed fee high powered: firm is the residual claimant for its cost savings Levent Koçkesen (Koç University) Screening: Applications 3 / 27 Levent Koçkesen (Koç University) Screening: Applications 4 / 27 Regulation Regulation Government: provide incentives to reduce cost and extract firm’s rent Fixed-price ◮ Incentives: Good ◮ Rent extraction: Bad ⋆ ⋆ ⋆ ⋆ Early models Firm is the residual claimant Provides efficient effort ◮ ◮ Rent is non-negative when cost is high Exogenous reduction in cost benefits the firm The bible: Laffont, J.-J. and J. Tirole (1993), A Theory of Incentives in Procurement and Regulation, Cambridge: MIT Press. Cost-plus ◮ ◮ Baron, D. and R. Myerson (1982), “Regulating a Monopolist with Unknown Costs,” Econometrica, 50, 911-30. Laffont, J.-J. and J. Tirole (1986), ”Using Cost Observation to Regulate Firms,” Journal of Political Economy, 94, 614-41. Incentives: Bad Rent-extraction: Good Optimal contract ◮ ◮ Perfect information about technology: fixed-price Imperfect information: incentive (cost-sharing) contract Levent Koçkesen (Koç University) Screening: Applications 5 / 27 Regulation Levent Koçkesen (Koç University) Screening: Applications 6 / 27 Perfect Information Our model: a simple version of Laffont and Tirole (1986) Regulator knows θ (therefore, can deduce e = θ − c) The firm produces a single indivisible output Accounting cost is c = θ − e Its problem is: ◮ ◮ ◮ min{θ − e + t} (e,t) θ ∈ {θL , θH }, prob(θL ) = p e is effort which costs the firm ψ(e) = e2 /2 We assume ∆θ = θH − θL ≤ 1 − p Regulator wants to maximize S − P ◮ ◮ s.t. t − e2 /2 ≥ 0, e≥0 The constraint must bind: t = e2 /2 min{θ − e + e2 /2} S is (constant) consumer surplus from consuming the good P = c + t is total payment, t is net transfer to the firm e≥0 If the solution is in the interior (e∗ > 0) Therefore, regulator wants to minimize P Payoff of the firm P − c − e2 /2 = t − e2 /2 −1 + e∗ = 0 ⇒ e∗ = 1 Reservation payoff of the firm is zero Comparing with the boundary (e = 0) shows that the solution is in the interior c is observable but not θ or e Levent Koçkesen (Koç University) Screening: Applications 7 / 27 Levent Koçkesen (Koç University) Screening: Applications 8 / 27 Perfect Information Unobservable θ and e Effort is efficient What happens if regulator offers first-best? marginal cost = ψ ′ (e∗ ) = e∗ = 1 = marginal benefit ◮ cL = θL − 1, cH = θH − 1 and tL = tH = 0.5 For the efficient type (θL ) and ∗ t = 0.5 ◮ firm receives no rent ◮ Regulator pays the firm a fixed total payment P = θ − 0.5 Induces the firm to minimize c + e2 /2 This is a price-cap (fixed-fee) regulation Effort required to obtain cost cL is 1 Effort required to obtain cost cH (e′ ) is 1 − ∆θ < 1: c H = θH − 1 = θL − e ′ ◮ Since transfers are the same, she prefers to mimic the inefficient type Usual recipe for optimal contract: t = a − (c − c∗ ) ◮ Efficient type (θL ): ◮ Inefficient type (θH ): ⋆ where ⋆ a = ψ(e∗ ) = 0.5 ⋆ c∗ = θ − e∗ = θ − 1 Levent Koçkesen (Koç University) Screening: Applications ⋆ 9 / 27 Unobservable θ and e min Levent Koçkesen (Koç University) Screening: Applications 10 / 27 1. IRL does not bind: tL − e2L /2 ≥ tH − (eH − ∆θ)2 /2 > tH − e2H /2 ≥ 0 p(tL + θL − eL ) + (1 − p)(tH + θH − eH ) 2. ICL binds, otherwise subject to tL − e2L /2 > tH − (eH − ∆θ)2 /2 > tH − e2H /2 ≥ 0 tL − e2L /2 ≥ 0 and the regulator can decrease tL tH − e2H /2 ≥ 0 3. IRH binds, otherwise (by (1)) can decrease tL and tH by the same small amount tL − e2L /2 ≥ tH − (eH − ∆θ)2 /2 tH − e2H /2 ≥ tL − (eL + ∆θ)2 /2 4. eH ≤ eL + ∆θ or cH ≥ cL : ICL and ICH imply Note that for type θL , effort required to achieve cost cH = θH − eH is equal to eH − (θH − θL ). Similarly, for high type’s effort to achieve low cost. We implicitly assume eH − ∆θ ≥ 0, which we will need to verify once we solve the problem. Levent Koçkesen (Koç University) provides less than first best effort (to reduce efficient type’s rent) obtains no rent Solving the Problem Regulator chooses a contract {(tL , cL ), (tH , cH )} to minimize p(tL + cL ) + (1 − p)(tH + cH ) subject to IR and IC constraints. Since c = θ − e, the problem can be written as (ti ,ei ) provides first best effort obtains positive rent Screening: Applications 11 / 27 e2H /2 − (eL + ∆θ)2 /2 ≤ tH − tL ≤ (eH − ∆θ)2 /2 − e2L /2 Result follows from simple algebra 5. ICH can be ignored: follows from steps (2) and (4) Levent Koçkesen (Koç University) Screening: Applications 12 / 27 Solving the Problem Solving the Problem Therefore, the problem is equivalent to min (ti ,ei ) Substituting in the constraints, the problem is equivalent to p(tL + θL − eL ) + (1 − p)(tH + θH − eH ) min {p(e2H /2 + e2L /2 − (eH − ∆θ)2 /2 − eL ) + (1 − p)(e2H /2 − eH )} subject to eL ,eH ≥0 tH − e2H /2 = 0 or tL − e2L /2 = tH − (eH − ∆θ)2 /2 max −{p(e2H /2 + e2L /2 − (eH − ∆θ)2 /2 − eL ) + (1 − p)(e2H /2 − eH )} eL ,eH ≥0 Note that tL − e2L /2 = ∆θ(eH − ∆θ/2) eH ≥ ∆θ implies that type θL obtains a rent Levent Koçkesen (Koç University) Screening: Applications 13 / 27 Solving the Problem Levent Koçkesen (Koç University) Screening: Applications 14 / 27 Solving the Problem Critical points are given by The objective function is strictly concave and the constraint functions eL and eH are concave Therefore, if a solution exists Kuhn-Tucker conditions will uniquely identify it −p(eL − 1) + λL = 0 −(p∆θ + (1 − p)(eH − 1)) + λH = 0 λL , λH , eL , eH ≥ 0, λL eL = λH eH = 0 First equation and λL ≥ 0 ⇒ eL > 0 ⇒ λL = 0, which implies The Lagrangean is given by eL = 1 L = −{p(e2H /2 + e2L /2 − (eH − ∆θ)2 /2 − eL ) + (1 − p)(e2H /2 − eH )} + λL eL + λH eH Similarly the assumption ∆θ ≤ 1 − p implies that eH = 1 − p ∆θ 1−p and eH ≥ ∆θ, as required. Levent Koçkesen (Koç University) Screening: Applications 15 / 27 Levent Koçkesen (Koç University) Screening: Applications 16 / 27 Properties of the Solution Financial Contracts and Credit Rationing Familiar “no distortion at the top” and “under-provision for inefficient types” result Optimal regulation: offer a menu so that ◮ Efficient firm chooses a price-cap regulation ⋆ ⋆ ◮ A constant payment PL = tL + θL − eL tL = aL − bL c, aL = tL + θL − eL and bL = 1 ◮ ◮ tH = aH − bH c, aH = tH + eH (θH − eH ) and bH = eH = 0.5 Levent Koçkesen (Koç University) Screening: Applications If there is excess demand of funds, the interest rate should increase to equate demand with supply The reason might be adverse selection As the interest rate increases only riskier projects apply Most popular article: Inefficient firm chooses a cost-sharing arrangement ⋆ Why are certain worthy projects not funded? 17 / 27 A Simple Model Stiglitz, J. and A. Weiss (1981) Levent Koçkesen (Koç University) Screening: Applications 18 / 27 Symmetric Information There is a population (unit mass) of risk-neutral borrowers Each borrower owns a project that requires an investment of 1 Two types of borrowers: risky (θr ) and safe (θs ) ◮ The problem of the bank facing borrower of type i is max prob(θs ) = q Di ≥0 Each project has an uncertain return X ∈ {0, Ri }, Ri with prob. pi ◮ pi (Ri − Di ) ≥ 0 Assume ps > pr and Rr > Rs We will look into two scenarios: For now assume pi Ri = m > 1 1. pi Ri = m > 1: Both projects worth investing 2. pr Rr < 1 < ps Rs : Only safe project worth investing Solution is simple: lend all α and set Single bank with total amount of funds max{q, 1 − q} < α < 1 Net expected payoff of the bank if it lends to type i and requires a repayment of Di is given by pi Di − 1 ◮ pi Di − 1 subject to θr = (pr , Rr ) and θs = (ps , Rs ) Di = Ri Expected payoff of the bank is α(pi Ri − 1) = α(m − 1) > 0 Assumes limited liability: if the return is zero, the borrower defaults and bank gets zero repayment Levent Koçkesen (Koç University) Screening: Applications 19 / 27 Levent Koçkesen (Koç University) Screening: Applications 20 / 27 Adverse Selection Adverse Selection Assume that a bank contract is simply D D > Rs ⇒ only risky apply ⇒ best to set D = Rr D = Rs < Rr ⇒ risky borrowers would like to pay more D to get a loan But as D increases borrower pool becomes riskier and the bank’s payoff decreases payoff = (1 − q)(pr Rr − 1) = (1 − q)(m − 1) D ≤ Rs ⇒ both apply ⇒ lend all α at D = Rs payoff = α[q(ps Rs − 1) + (1 − q)(pr Rs − 1)] = α[qm + (1 − q)pr Rs − 1] ◮ ◮ ◮ Adverse selection D = Rs some borrowers cannot get credit credit rationing Levent Koçkesen (Koç University) Screening: Applications 21 / 27 Second Best Levent Koçkesen (Koç University) max The bank’s problem is (xi ,Di ) 22 / 27 This suggests the reduced problem: xi is probability of getting a loan max Screening: Applications Solving the Problem A contract is {(xs , Ds ), (xr , Dr )} ◮ ◮ Can the bank do better by designing more sophisticated contracts? If (1 − q)(m − 1) < α[qm + (1 − q)pr Rs − 1] (xi ,Di ) xs ps (Rs − Ds ) ≥ 0 xr pr (Rr − Dr ) ≥ xs pr (Rr − Ds ) subject to xi pi (Ri − Di ) ≥ 0, 0 ≤ xi ≤ 1, xr ≥ xs i = s, r xi pi (Ri − Di ) ≥ xj pi (Ri − Dj ), 0 ≤ xi ≤ 1, i, j = s, r This should be familiar by now ◮ If bank offers the first best menu {(xs , Ds ), (xr , Dr )} = {(α, Rs ), (α, Rr )} ◮ ◮ Risky type mimics the safe type Screening: Applications i = s, r qxs + (1 − q)xr ≤ α i = s, r qxs + (1 − q)xr ≤ α Levent Koçkesen (Koç University) qxs (ps Ds − 1) + (1 − q)xr (pr Dr − 1) subject to qxs (ps Ds − 1) + (1 − q)xr (pr Dr − 1) IR constraint of the high type (θr ) does not bind IC constraint of the low type (θs ) does not bind Monotonicity in the allocation xr ≥ xs The original problem (P) is equivalent to the reduced problem (RP) 23 / 27 Levent Koçkesen (Koç University) Screening: Applications 24 / 27 1. Constraints of (P) imply those of (RP): From the two IC constraints: xr Rs − xs Rs ≤ xr Dr − xs Ds ≤ xr Rr − xs Rr max xs [q(ps Rs − 1) − (1 − q)pr (Rr − Rs )] + (1 − q)xr (pr Rr − 1) which implies xr ≥ xs 2. At the solution to (RP) IC constraint holds as equality: otherwise can increase Dr ◮ Here we use xr > 0. If xr = 0, then by Step (1) xs = 0, which cannot be a solution to (RP) IR and IC constraints of (RP) imply IR for type θr xr = 1 > α xs > 0 implies Together with monotonicity and budget constraint α − (1 − q) xs = <α q IC for risky type (Rr − Dr = xs (Rr − Ds )) and Ds = Rs imply xr pr (Rr − Dr ) ≥ xs pr (Rr − Ds ) ≥ xs pr (Rs − Ds ) ≥ 0 ◮ xs ,xr Since we assumed pi Ri = m > 1 q(ps Rs − 1) − (1 − q)pr (Rr − Rs ) ≥ 0 3. Solution of (RP) satisfy the constraints of (P) ◮ Suppose xs > 0. Then, Ds = Rs Substitute the IC and IR into the objective function Step (2) and xr ≥ xs imply IC for type θs ps [xs (Rs − Ds ) − xr (Rs − Dr )] = ps [xs Rs − xr Rs + xr Rr − xs Rr ] = ps (xr − xs )(Rr − Rs ) ≥0 Dr = xs Rs + (1 − xs )Rr > Rs = Ds If q(ps Rs − 1) − (1 − q)pr (Rr − Rs ) < 0 xs = 0 Levent Koçkesen (Koç University) Screening: Applications 25 / 27 Interpretation The bank trades off the rents extracted from risky borrowers with ability to lend all its funds When safe borrowers are numerous enough, wants to lend to them, which leads to rent for risky borrowers But rather than financing everybody by setting D = Rs , the bank gives risky borrowers preferential access by setting xr > xs in exchange for higher repayment Credit rationing disappears: The only ones not fully funded are safe borrowers, but they are indifferent about being funded Credit rationing reappears if we instead assume pr Rr < 1 < ps Rs ◮ ◮ Bank wants to turn away risky borrowers (i.e., set xr = 0), but cannot do it without also denying the safe ones (since xr ≥ xs ) Either everybody is funded (if (qps + (1 − q)pr )Rs ≥ 1) or nobody, i.e., there is a financial collapse Levent Koçkesen (Koç University) Screening: Applications 27 / 27 Levent Koçkesen (Koç University) Screening: Applications 26 / 27