Models of Asymmetric Information

1) Environment:

There is a lender where the cost of providing loans is C(L) = L1+β where L is the total amount lent
and β is a positive parameter representing the difficulty of raising funds1 . There is a pool of potential
borrowers represented by a demand function Ld (r) where loan demand is decreasing in the interest
rate r. In the most basic case, there is a single common default rate (δ) for all borrowers. If a loan is
defaulted on, the lender receives nothing.

2) Model without Asymmetric Information:

The lender’s profit function is:

π(L, r) = (1 + r)(1 − δ)L − L1+β
The first term is the revenue generated by the loans based on the interest rate and default rate. The
second term is the cost of raising funds, per the environment. If we maximize profits with respect to
L, we get:
(1 + r)(1 − δ) − (1 + β)Lβ = 0
 1
s (1 + r)(1 − δ) β
L =
1+β
This is a loan supply curve, describing the amount that will be lent given a particular interest rate. It
is also easy to see that Ls is strictly increasing in r, so this supply curve has the usual upward slope.
Thus, the equilibrium will be at the intersection between the Ls curve and the Ld curve, as shown in
the diagram below:

3) Adverse Selection:

The above version of the model assumed that the default rate was independent of the interest rate.
However, it is more realistic to recognize that borrowers who are more desperate for funds will accept
a higher interest rate, and are also more likely to default. Thus, we modify the environment such that
there is a default rate function δ(r) that is increasing and convex2 . The profit function of the lender
becomes:
π(L, r) = (1 + r)(1 − δ(r))L − L1+β
1
For example, this could be the interest rate paid out on deposits.
2
Technically, this is more strict than we need. A linearly increasing δ(r) would also work, and so would some slightly
concave functions.

1
Thus, the loan supply function is:
 1
s (1 + r)(1 − δ(r)) β
L =
1+β

Since the default rate is escalating, this means there could be a maximum value for Ls , above which
the supply curve becomes backward bending. To see this we can solve for the profit maximizing interest
rate, by setting up the first and second order conditions for a maximum:
∂π
= 0 L − δ ′ (r)L − δ(r)L − rδ ′ (r)L = 0
∂r
1 − δ(r)
F.O.C. = δ ′ (r)
1+r
∂2π
< 0 − δ ′′ (r)L − δ ′ (r)L − rδ ′′ (r)L − δ ′ (r)L < 0
∂r2
S.O.C. (1 + r)δ ′′ (r) + 2δ ′ (r) > 0
Per the above, since δ ′ (r) > 0 and δ ′′ (r) > 0, there is a value of r that maximizes profits3 . It may
also help to interpret the first order condition as saying that the ratio of the repayment probability to
the repayment value of the loans is equal to the marginal rate of default at the maximum interest rate
(call this r̄). At interest rates above r̄, profit falls because the numerator is shrinking faster than the
denominator is growing.

It is also important to note that r̄ occurs at the point where Ls is highest (which we will call Lmax )4 .
As such, the lender will only want to offer loans along the upward sloping portion of the supply curve.
This creates two cases. If Ls and Ld intersect at a level below r̄, then the loan market will clear as in
the diagram below:

On the other hand, if demand does not cross the upward sloping portion of the supply curve, loan
demand will exceed supply. The lender will only be willing to offer loans at the interest rate r̄, and only
some borrowers will be able to obtain a loan (see diagram below). This phenomenon is called credit
rationing.
3
Note that the second order condition could be satisfied with a linear δ function, or even a modestly concave one.
However, if δ were too concave, the conditions for a maximum r would no longer be met – or to put it more simply, the
supply curve would not be backward bending.
4
This was proven in class by maximizing Ls with respect to r, which yields the same equation as the first order condition
for profit maximization in r, showing that the interest rate that maximizes total lending is also the one that maximizes
profits.

2
Essentially, because the lender cannot observe the default rate of any particular borrower, it has to offer
loans on the basis of the aggregate default rate. But since the pool of borrowers becomes increasingly
risky as the interest rate they offer rises, there is a point beyond which they are unwilling to raise the
interest rate. If demand is high enough, there are too many borrowers willing to take loans at that
interest rate relative to the loan volume that is profitable for the lender. This creates a shortage of
loaned funds.

Adverse selection can be reduced via screening, where the lender tries to assess how credit-worthy a
borrower is before offering them a loan. The goal here is to split the market and offer different interest
rates to borrowers with different default rates. If screening is good enough, credit rationing could be
reduced or even eliminated. However, screening is not costless – for example, imagine if screening
increased the value of β. The reduction in credit rationing could be offset by an increased cost of
providing loans, which means that adverse selection would still be distorting the market (just through
the cost of screening rather than through credit rationing). In practice, banks use a variety of screening
techniques to try and tailor their credit offers to each individual loan applicant, but errors in detection
and the cost of screening still mean that adverse selection is reducing the overall amount of loans
relative to a world with full information.

4) Moral Hazard:

Return to the basic model from section 2), but now allow for the behavior of borrowers to change after
the loan is extended. Assume that the loans are being extended to businesses based on their current
business model, and that this yields a return to the borrowers of rc and their ventures will fail at rate
δc . The borrowers, however, have an alternative business plan that is riskier but more profitable. The
alternate business model yields a return of ra and fails at rate δa , where ra > rc and δa > δc .

It is helpful to imagine what the lender’s supply curve would be if they knew which business plan would
be used by all borrowers. Thus, there would be two supply curves, each based on one particular default
rate, where Ls (δa ) < Ls (δc ) for all values of r. However, depending upon the interest rate charged, a
borrower might choose one or the other business plan. We can find that decision point by writing the
condition under which a borrower would use the alternate business plan:

(1 − δa )(ra − r) > (1 − δc )(rc − r)

(1 − δc )rc − (1 − δa )ra
r >
(1 − δc ) − (1 − δa )
Thus, if the interest rate charged by the lender is greater than the expression on the right hand side,
borrowers will switch to the alternate plan. We can call label this switching point ra . Again, there are

3
two cases. If the demand curve crosses Ls (δc ) below ra , then borrowers will remain on their current
plan:

On the other hand, if loan demand is high enough, then borrowers will switch to the alternate plan,
meaning that the lender will need to charge a higher interest rate and extend fewer loans to compensate
for the increased default rate:

Because the lender cannot prevent the borrowers from changing their business plan, if loan demand
is high enough, the lender will have to reduce their lending and raise the interest rate to compensate.
Sometimes moral hazard can be reduced via monitoring where the borrower is restrained from switching
to an alternate plan. However, monitoring is costly and not always fully effective, which is why it is
rarely used5 . The other major way to offset moral hazard is for the lender to require collateral . . .

5
Venture capital is the main exception, but here the monitoring is being compensated for by the venture capital firm
getting a share of profits that may vary with the borrowing firm’s performance.

4
5) Moral Hazard with Collateral:

The mechanism behind collateral is to increase the cost of default for the borrower, thus making it less
enticing to switch to a riskier business plan. Have the lender require collateral with value equal to a
fraction of the total loan amount, which we will denote as c. This means that the expected gain for the
borrower per unit of money is being reduced by the collateral. This modifies the borrower’s decision to
use the alternate business plan:

(1 − δa )(ra − r) − δa c > (1 − δc )(rc − r) − δc c

(1 − δc )rc − (1 − δa )ra + c(δa − δc )

r >
(1 − δc ) − (1 − δa )
Notice that as c increases, so does the cutoff value of r. Thus, to prevent borrowers from switching
to the alternate plan, they need to increase c until the equation above holds with equality (meaning
that the borrower no longer gains from switching). Thus, there will be some c such that ra (c) is at the
intersection between loan demand and the supply curve for the current business plan (Ls (δc ) = Ld ).

The above result assumes that all borrowers have collateral available in the required amount. If bor-
rowers lack suitable assets to offer as collateral, then a collateral requirement can push borrowers out
of the market. Also, there are many kinds of loans that cannot be secured with collateral since the
value of the collateral has to be assessed, which limits this method to large purchases (homes, cars,
machinery).

6) Conclusion:

The existence of adverse selection and moral hazard in credit markets reduces the amount of credit
available. These effects can be partially mitigated, but never entirely. The distortions of credit supply
created by asymmetric information are broadly referred to as financial frictions, and are large enough
to impact the economy on the macro scale. Credit models of the business cycle typically propose that
financial frictions escalate toward the end of an economic boom as credit demand begins to increase
faster than supply6 . This ultimately leads to a wave of defaults that triggers the recession phase of the
business cycle.

6
Note how in the models above, an increase in Ld will often make the asymmetric information effects worse, and will
never reduce them.