Microeconomics IV

Section 3
Aleksei Beliaev
March 30, 2023

1 Lending with adverse selection (PS3Q3)

There is a continuum of risk neutral borrowers with no personal wealth and limited liability.
A proportion v of borrowers (called type 1) have sure projects with return h for an investment
of 1. A proportion 1 − v of borrowers (called type 2) have (stochastically independent) projects
with return h only with probability θ ∈ (0, 1) and return 0 with probability 1 − θ, for an
investment of 1. If he does not apply for a loan, the borrower has an outside opportunity
utility level of u.
There is a single risk neutral bank available for loans which has a financing cost of r. The
bank offers contracts to maximize its expected profit. For simplicity, we assume that all projects
are socially valuable, i.e.
θh > r + u.
1. Explain why there is no loss of generality in considering the menus of contracts of the
form {(R1 , P1 ), (R2 , P2 )} where Pi is the probability of obtaining a loan and Ri is the
repayment to the bank when the investment succeeds if the borrower announces that he
is of type i.
Solution. It is a principal-agent problem with adverse selection, it is straightforward to
decipher the roles of all players, and the types of the agents.
Since there are only two types, by the Revelation Principle an optimal contracting problem
can be equivalently rewritten as to consider direct truthful mechanisms, that is, menus
of two contracts {(R1 , P1 ), (R2 , P2 )} satisfying incentive compatibility:
P1 (h − R1 ) ≥ P2 (h − R2 ), (IC1 )
P2 θ(h − R2 ) ≥ P1 θ(h − R1 ). (IC2 )

2. Write the maximization program of the bank which chooses the menu {(R1 , P1 ), (R2 , P2 )}
to maximize its expected profit under the borrower’s participation and incentive con-
straints (for simplicity assume that if a borrower applies for a loan he loses his outside
opportunity u).
Solution. If the bank wishes to give a loan with positive probability to both types (case 1),
then it solves:
max vP1 (R1 − r) + (1 − v)P2 (θR2 − r)
{(R1 ,P1 ),(R2 ,P2 )}

s.t. (IC1 ), (IC2 ), and

P1 (h − R1 ) ≥ u (IR1 )
P2 θ(h − R2 ) ≥ u (IR2 )

1
 Alternatively, it might offer a loan only accepted by type 1 (case 2), then it solves:
max vP1 (R1 − r) s.t. (IR1 ).
(R1 ,P1 )

Note: One can easily show that the bank would not wish to offer a loan only accepted by
type 2. To see that, suppose it does: (P2∗ , R2∗ ) are such that the quantity (1−v)P2 (θR2 −r)
is maximized subject to (IR2 ). By monotonicity, the constraint is binding, and so
P2∗ (h − R2∗ ) > P2∗ θ(h − R2∗ ) ≥ u,
so the contract offer to the type 2 agents also satisfies (IR1 ), and the type 1 agents will also
take it (bringing the situation to case 1). Note that the bank would not want to restrict
the type 1 agents to taking this loan, since otherwise it will be short of an additional
profit of vP2∗ (R2∗ − r) > 0.
3. Show that the optimal contract entails a non-random allocation of loans (i.e. Pi is either
0 or 1, i = 1, 2). Characterize the optimal contract. Discuss its properties. Draw
indifference curves (in the space (R, P )) and look at the Spence-Mirrlees condition. To
answer this question, you will need the more general definition of the Spence-Mirrlees
condition. Let U be the borrower’s utility, then the single crossing condition says that
( / ) has a constant sign.
∂ ∂U ∂U
∂θ ∂P ∂R
Solution. Consider two cases.
Case 1: both types are served. Dividing (IC2 ) through by θ, and combining with (IC1 ),
we obtain
P1 (h − R1 ) = P2 (h − R2 ). (1)

Moreover, note that, since θ < 1,

P1 (h − R1 ) = P2 (h − R2 ) > P2 θ(h − R2 ) ≥ u,
so in optimum, (IR2 ) should be binding, and (IR1 ) should be slack. From (IR2 ), we have
R2 = h − Pu2 θ . From (IR2 ) and (1) jointly, we have R1 = h − Pu1 θ .
Note that we have used up all the constraints. Substituting the expressions for R1 , R2
into the bank’s expected profit, we reformulate the problem:
max vP1 (h − r) + (1 − v)P2 (θh − r) − vu/θ − (1 − v)u.
P1 ,P2

By assumption, θr − h > u ≥ 0, so h − r > 0 as well, and the optimum is at P1∗ = P2∗ = 1,

which implies R1∗ = R2∗ = h − uθ . Thus, we obtain a pooling contract, at which the bank’s
expected profits equal
 u
v h−r− + (1 − v)(θh − r − u).
θ

Under this pooling contract, the type 1 agents’ information rent is P1∗ (h−R1∗ )−u = 1−θ
θ
u.
Case 2: only type 1 is served. The only constraint is (IR1 ). The obvious solution is
(P1∗∗ = 1, R1∗∗ = h − u), at which the bank’s expected profits are v(h − r − u).
Case 2 is better for the bank if and only if
 u
v(h − r − u) ≥ v h − r − − (1 − v)(θh − r − u)
θ
1−θ
⇔ (1 − v)(θh − r − u) ≤ v u,
θ

2
 which can be interpreted as the profits from type 2 being smaller than the rent paid to
type 1.
Finally, let us address the Spence-Mirrlees condition. The borrower’s utility is given by
U = P θ(h − R). The indifference curves are depicted below:

Here, the red line represents type 1’s indifference curve (θ = 1), and the blue line type
2’s indifference curve (θ < 1). As we can see, they do not cross.
Formally, we check the Spence-Mirrlees condition:
!
∂U    
∂ ∂P ∂ θ(h − R) ∂ R−h
= = = 0,
∂θ ∂U∂R
∂θ −P θ ∂θ P

which is unsigned. Thus, the Spence-Mirrlees condition does not hold, and there is no
way to screen the types.

2 The bribing game (PS3Q4)

We consider an administration which is supposed to deliver with some fixed delay a service
to the citizens (passport, permits, ...). With the normal functioning of the administration,
citizens derive a benefit u0 which depends on their valuation of time.
With some additional effort the official can deliver the service with a shorter delay. Let us
call q the decrease of delay that the official can provide at a cost (q − Q)2 /2 for him where Q
is a constant.
We assume that there is a proportion v (resp. 1 − v) of type 1 (resp. type 2) citizens who
derive a benefit from a decrease q of delay equal to θq (θq). Citizens are willing to bribe the
official to decrease delays.
Characterize the optimal bribing contract that the official will offer to the citizens.

Solution. Let (t1 , q1 ) and (t2 , q2 ) be the contracts offered where t1 (resp. t2 ) is the bribe
requested for a decrease of delay q1 (resp. q2 ).

3
The principal’s problem is:
(q1 − Q)2 (q2 − Q)2
   
max v t1 − + (1 − v) t2 −
{(t1 ,q1 ),(t2 ,q2 )} 2 2
s.t. θq1 − t1 ≥ u0 (IR1 )
θq2 − t2 ≥ u0 (IR2 )
θq1 − t1 ≥ θq2 − t2 (IC1 )
θq2 − t2 ≥ θq1 − t1 (IC2 )
Let ∆θ = θ − θ > 0, u1 = θq1 − t1 , u2 = θq2 − t2 , and rewrite the constraints as follows:
u1 ≥ u0 (IR1 )
u2 ≥ u0 (IR2 )
u1 + ∆θq2 ≥ u2 (IC1 )
u2 ≥ ∆θq1 + u1 (IC2 )
Let us figure out which constraints are binding and which are not. We show the following:
• q1 > 0 and q2 > 0.
Proof. By contradiction, suppose qi = 0 for some i. Then from (IRi ), −ti ≥ u0 , or
ti ≤ −u0 < 0. Thus, the principal loses money on type i, which is suboptimal.
• (IR2 ) holds as a strict inequality: u2 > u0 .
Proof. From (IR1 ) and (IC2 ), u2 ≥ ∆θq1 + u1 > u1 ≥ u0 .
• q2 > q1 .
Proof. From (IC1 )—(IC2 ), u2 ≥ ∆θq1 + u1 ≥ ∆θ(q1 − q2 ) + u2 , so ∆θ(q2 − q1 ) ≥ 0 and
thus q2 ≥ q1 . By contradiction, suppose q1 = q2 . Then (IC1 ) implies t2 ≥ t1 , and (IC2 )
implies t1 ≥ t2 , so t1 = t2 . Thus, every constraint except (IR2 ) holds with equality.
Let us show that this pooling contract is suboptimal. Pick ε > 0 small, and consider an
alternative menu of contracts:
   
ε ε ε ε
(t1 , q1 ) = t − , q − , (t2 , q2 ) = t + ,q + .
v vθ 1−v (1 − v)θ
It is straightforward that, if the pooling contract is feasible, then this new contract is
feasible as well: the new contracts bring the same utilities to the agents as the pooling
one, and the contract for type i is strictly worse for type i than the contract for type
j ̸= i. Now, note that the principal’s profit at the pooling contract is t − (q − Q)2 /2, and
if we calculate the difference in profits if the principal uses the new contracts instead, it
is given by !
 
1 1 1 1 1
∆π = + (q − Q)ε − + ε2 .
θ θ 2 vθ2 (1 − v)θ2

Note that the coefficient on the linear term is positive (if q > Q; the converse case is
considered later), and the coefficient on the quadratic term is negative. Hence, there
must exist such ε > 0 small that ∆π > 0. Therefore, the initial pooling contract is indeed
suboptimal.
Now, if q ≤ Q, pick ε > 0 small and offer another pooling contract (t′ , q ′ ) = (t + θε, q + ε)
instead. Type 1 is indifferent between the two contracts, and type 2 is strictly better off,
so the new contract is feasible. Moreover, ∆π = (θ − q + Q)ε − ε2 /2, which is positive for
some small ε > 0 as the coefficient on the linear term is positive.

4
 • (IC1 ) holds as a strict inequality: u1 + ∆θq2 > u2 .
Proof. By contradiction, suppose (IC1 ) holds with equality. Then (IC2 ) must hold with
a strict inequality. Indeed, by (IC1 ) and q2 > q1 , we have u2 = ∆θq2 + u1 > ∆θq1 + u1 .
Observe that ∆θq1 + u1 > u1 ≥ u0 .
Then pick ε ∈ (0, u2 − (∆θq1 + u1 )) and consider the menu {(t1 , q1 ), (t2 + ε, q2 )}. It is
feasible: (IR1 ) is unaffected, (IR2 ) still holds since ε < u2 − (∆θq1 + u1 ) < u2 − u0 , (IC1 )
still holds as it becomes slack, and (IC2 ) still holds since ε < u2 − (∆θq1 + u1 ). Moreover,
the new menu is a profitable deviation for the principal: it just requests a higher bribe
from type 2. We have a contradiction with optimality.
• (IR1 ) holds with equality.
Proof. By contradiction, suppose (IR1 ) holds as a strict inequality. Since (IR2 ) holds as
a strict inequality as well, increase both t1 and t2 by the same sufficiently small amount
ε > 0 such that both constraints still hold. Note that (IC1 )—(IC2 ) are unaffected, so this
action is feasible, and moreover, strictly profitable for the principal.
• (IC2 ) holds with equality.
Proof. By contradiction, suppose (IC2 ) holds as a strict inequality. Since (IR2 ) holds as
a strict inequality as well, increase t2 by a small ε > 0 such that both constraints still
hold. Note that (IR1 ) is unaffected, and (IC1 ) becomes slack, so it still holds. Thus, this
action is feasible, and it is strictly profitable for the principal.
All in all, we have shown that (IR1 ) and (IC2 ) are binding, and moreover, we can neglect
(IR2 ) and (IC1 ) as they are slack. Thus, use (IR1 ) and (IC2 ) to express t1 , t2 as functions of
q1 , q2 :
t1 = θq1 − u0 ;
t2 = θq2 − ∆θq1 − u0 ,
and substitute into the principal’s objective function:
(q1 − Q)2 (q2 − Q)2
   
v θq1 − + (1 − v) θq2 − ∆θq1 − − u0 → max
2 2 q1 ,q2

The FOCs with respect to q1 and q2 are given by, respectively,

v(θ − q1 + Q) − (1 − v)∆θ = 0,
(1 − v)(θ − q2 + Q) = 0,
and the solution is
1−v
q1∗ = Q + θ − ∆θ,
v
q2∗ = Q + θ.
Comparing with the first-best contracts (under perfect information):
q1∗ < q1F B = Q + θ,
q2∗ = q2F B = Q + θ,
t∗1 < tF1 B = θq1F B − u0 ,
t∗2 < tF2 B = θq2F B − u0 .
Thus, in the second best we have no distortion at the top (efficient delay for the impatient
type), and the bribes requested from both types are lower than in the first best.