Advanced Microeconomics 6QQMN970: Exercise

Sheet 7
Niall Hughes

1. A risk neutral Principal wishes to pay a risk averse Agent based on the effort
e she exerts, so as to incentivise her to work hard. However the Principal only
observes the profits generated by the Agent, where profits are a noisy signal
of effort π = e +  with  ∼ N (0, σ 2 ). Assume the Agent has a cost of effort
2
c(e) = ce2 and a utility function UA (e, w) = −exp[−r(w − c(e))] with reservation
utility U which is what she gets from not working. The Principal has utility
UP (π, w) = E(π − w) with w being the amount he pays to the Agent. Hint: for
k2 σ 2
 ∼ N (0, σ 2 ) and some constant k, we have E(expk ) = exp 2

(i) Solve for the optimal symmetric information contract of the form w =
β0 + βπ
(ii) Now under asymmetric information if the Principal gives the Agent the
contract w = β0 + βπ, what is the amount of effort she will exert?
(iii) What is the optimal contract (of the form (β0 , β)) that the Principal will
offer the Agent under asymmetric information?
(iv) Interpret the results

(i) The Principal wishes to maximise his utility subject to the Agent partic-
ipating and so must ensure the Agent gets at least her reservation utility.
However, there is no incentive for the Principal to pay the Agent more
than w where −exp−rw ≡ U as this would be eating into his own profits.
The Principal solves:
max E[π − β0 − βπ]
e,β0 ,β

s.t. E[UA (e, w)] = U

1
We can sub in the form of the Agents utility, the wage structure, profits
and costs to get

ce2 ce2
E[UA (e, w)] = E(−exp−r[β0 +β(e+)− 2
]
) = (−exp−r[β0 +βe− 2
]
)E(exp−rβ ) = U
(1)
Then using the Certainty Equivalence hint
ce2 2 2
− rβ 2σ ]
E[UA (e, w)] = −exp−r[β0 +βe− 2 =U

Next we take the log of both sides so that it’s easier to work with.

ce2 rβ 2 σ 2
−r(β0 + βe − − ) = log(−U )
2 2

Next we note that

E(π − βo − βπ) = e − βo − βe

We are now in a position to set up the Lagrangian.

ce2 rβ 2 σ 2
L = e − β0 − βe + λ[−r(β0 + βe − − ) − log(−U )]
2 2

F OC
1
β0 : −1 − λr = 0 ⇒ λ = −
r
β : −e − λ[re − r2 βσ 2 ] = 0 ⇒ β = 0
1
e : 1 − β + λ(−rβ + rce) = 0 ⇒ e =
c
Hence the Principal will pay the Agent a fixed wage β0 and require an
effort equal to 1c . This is the efficient outcome because whereas the Prin-
cipal is risk neutral, the Agent is risk averse.
Thus the Principal takes all the risk upon himself and fully insures the
Agent who gets utility − exp[−r(β0 − 2c 1
)] = U = −exp−rw . Subbing the
values back into the constraint we find that β0 = − log(−U
r
) 1
+ 2c 1
= w + 2c .

Page 2
(ii) Without the ability to observe the Agents effort, the first best contract is
no longer desirable for the Principal. We know that
ce2 2 2
− rβ 2σ )]
E[UA (e, w)] = −exp[−r(β0 +βe− 2

but when β = 0 as it is in the first best case then the above simplifies to
ce2
−exp[−r(β0 − 2
)]

and as the Principal cannot observe effort, the Agent is free to choose
any effort so as to maximise her expected utility.Due to the negative
exponential this utility is maximized when the argument is minimized, so
we can write the problem as:

ce2
max[r(β0 − )] ⇒ e = 0
e 2
Clearly the Principal would be unhappy with this case of complete shirk-
ing and so would like to redesign the payment structure so as to encourage
the Agent to exert effort.If β > 0 then the Agents maximisation problem
becomes
ce2 rβ 2 σ 2 β
max[r(β0 + βe − − )] ⇒ e =
e 2 2 c
and so the Agent now has an incentive to exert effort the greater the piece
2 2
rate(β) is and the smaller the cost (c) is. The term − rβ 2σ reflects the
negative effect on the risk averse Agent of taking on more risk by moving
away from a fixed wage contract.

(iii) The optimal contract should incentivise the Agent to exert effort but as
minimium cost to the Principal. The Principal achieves this by maximis-
ing his expected utility subject to the Agent’s participation constraint
(PC) and her incentive compatability constraint (IC).

max E[π − (β0 + βπ)]

β0 ,β

ce2 2 2
− rβ 2σ )]
s.t. − exp[−r(β0 +βe− 2 ≥ U (P C)
β
e= (IC)
c

Page 3
 The PC binds as usual and the IC is the expression we found in the
previous question to incentivise the Agent. We can now sub e∗ = βc into
the Principals objective function and maximise it subject to the PC in a
Lagrangian.

β β2 β 2 c β 2 rβ 2 σ 2
L(β0 , β, λ) = − (β0 + ) + λ − r(β0 + − − ) − log(−U ))
c c c 2 c2 2

Taking FOCs we find λ = − 1r and β = 1

1+rcσ 2

(iv) • 0 < β ≤ 1: The Principal always offers the Agent less than the full
worth of the work.
δβ
• δr
< 0: More risk averse agents face less powerful incentives.
δβ
• δσ
< 0: Riskier environments give rise to less powerful incentives.
δβ
• δc
< 0: Less powerful incentives when greater disutility of effort.

2. Consider the moral hazard model with 2 efforts (eH , eL ) and 2 outcomes (xS , xF ).
The costs of effort are cH and cL . Assume the probability of success after high
effort is higher than the probability of success after low effort i.e. qH > qL .
Assume a risk averse agent and risk neutral principal, and assume it is efficient
to implement the high action. Give a necessary and sufficient condition on qH
and qL such that the first best applies.

The first best will take place if (a) the high action is implemented and (b)
compensation does not vary depending on the outcome (remember the Agent
is risk averse while the Principal is not). The first statement is true when ICH
holds. Recall that ICH is

qH u(wS ) + (1 − qH )u(wF ) − cH ≥ qL u(wS ) + (1 − qL )u(wF ) − cL

That is when (qH − qL )(u(wS ) − u(wF )) ≥ cH − cL which will hold only if

wS > wF .
The second statement will not be violated only if qH = 1, that is exerting high
effort always leads to the high outcome. Therefore, the first best can occur only

Page 4
 if qH = 1. When this is the case, we can use that fact that P CH must bind to
find that the optimal wage solves u(wS ) = u + cH .
We can sub this into P CH to get (1 − qL )(u + cH − u(wF )) ≥ cH − cL . We can
rearrange this to get
cL + u − u(wF )
qL ≤
cH + u − u(wF )
we are free to set wF arbitrarily low as it will not violate any constraints and
will make ICH easiest to satisfy. This gives us the following condition on qL
cL + u
qL ≤
cH + u

3. Suppose we have 2 levels of effort eL , eH but three possible outcomes: Good,

Medium and Bad. Let c(eL ) = 0 and c(eH ) = 1. The probability of outcome
i from each effort level is given in the table below. Suppose the Agent is risk
averse with utility u(w, eL ) = ln(w) after low effort and u(w, eH ) = ln(w) − 1
after high effort.

Profits πi qiH qiL

Good 100 0.5 0.2
Medium 50 0.3 0.3
Bad 0 0.2 0.5

(i) Solve the Principal’s problem if he wants to induce low effort. What wage
does he set and what are his expected profits?
(ii) Solve the principal’s problem if he wants to induce high effort. What wages
does he set and what are his expected profits (as functions of the Lagranian
multiplies)?
(iii) Is the Principal better off under high effort or low effort? what conditions
on the Lagrange multiplies are necessary?
(iv) Now suppose the probabilities are as follows: What wages should the Prin-
cipal set if he wants the Agent to exert high effort? How does this differ
from the wage schedule in (ii) and why?

Page 5
 Profits πi qiH qiL
Good 100 0.55 0.3
Medium 50 0.25 0.4
Bad 0 0.2 0.3

(i) We can set-up the Lagrangean and solve for the optimal wage to exert
low effort, to find that it is simply that which makes P CL bind, which
is u(wL ) = eL = 0. In this case that is ln(wL ) = 0. We can take the
exponent of both sides to find expln(wL ) = exp0 and therefore wL = 1.
The Principal’s expected profits are

E(πL ) = (0.2)(100) + (0.3)(50) + (0.5)0 − expeL = 35 − 1 = 34

(ii) We can set-up the Lagrangean and solve for the optimal wage to exert
high effort, to find that it is that which makes P CH and ICH bind. The
optimality condition for wages in general is
1  qiL 
= λ + µ 1 −
u0 (wi ) qiH
In our specific case then this becomes:
 0.2  3
wG = λ + µ 1 − =λ+ µ
0.5 5
 0.3 
wM = λ + µ 1 − =λ
0.3
 0.5  3
wB = λ + µ 1 − =λ− µ
0.2 2

We can see that wages are increasing in outcomes.

The Principal’s expected profits are
3 3
E(πH ) = (0.5)(100 − λ − µ) + (0.3)(50 − λ) + (0.2)(0 − λ + µ)
5 2
= 65 − λ

(iii) We can see that profits from high effort are higher whenever 65 − λ > 34
or λ < 31.

Page 6
(iv) The optimality condition for wages in general is
1  qiL 
=λ+µ 1−
u0 (wi ) qiH

In our specific case then this becomes:

 0.3  5
wG = λ + µ 1 − =λ+ µ
0.55 11
 0.4  3
wM = λ + µ 1 − =λ− µ
0.25 5
 0.3  1
wB = λ + µ 1 − =λ− µ
0.2 2

We can see that wages here are not increasing in outcomes. Wages are
higher after bad outcomes than after medium outcomes. The reason wages
are not monotonic here is because qqiHiL
does not satisfy the monotone
likelihood ratio property.

Page 7