...

Moral Hazard
Swapnendu Banerjee
Jadavpur University
Kolkata, INDIA
 Asymmetric information: A primer
✘ When one party knows more than the other.
✘ Pre- contractual asymmetric info:
Adverse Selection: When the Principal and the Agent sign a
contract. The Principal doesn’t know the Agents’ types. Asymmetric
Information exits prior to signing of the contract. (Hidden
Information). Nature’s move is there. Information incomplete.
Example: Market for lemons, Job Market, etc.
Way- out:
Signalling
Screening
 Asymmetric information: A primer
✘ Post- contractual asymmetric info:
Moral Hazard => When the Principal and the Agent sign a contract
and after signing the agent takes an action that the Principal cannot
observe. (Hidden Actions). No nature’s move. Information complete.
Asymmetric information starts after the contract is signed.

Examples: Landlord-tenant (agriculture), Employer-Employee, etc.

Way- out: The Principal needs to design proper incentives to elicit
the desired effort.
 Moral hazard- Model with Discrete outcomes and Effort
✘ Employer- Worker framework.
✘ Employer is the principal and the worker is the agent.
✘ Employer is risk neutral and the agent is risk averse.
✘ The agent has an utility function 𝑢 𝑤, 𝑒 = 𝑤 − (𝑒 − 1), where ‘e’ is effort
and w is wage.
✘ 𝑢𝑤 > 0, 𝑢𝑤𝑤 < 0, 𝑢𝑒 < 0.
✘ Suppose only two effort levels are possible, e=1 (low effort) & e=2 (high
effort)
✘ Reservation utility u=1 (Utility from outside option)
✘ Agent’s effort is assumed to help increase the revenue of the employer.

4
✘ When 𝑒 = 1

10 𝑤𝑖𝑡ℎ 𝑝𝑟𝑜𝑏 2/3

Revenue (R)= Expected Revenue 50/3
30 𝑤𝑖𝑡ℎ 𝑝𝑟𝑜𝑏 1/3
✘ When 𝑒 = 2

10 𝑤𝑖𝑡ℎ 𝑝𝑟𝑜𝑏 1/3

Revenue (R)= Expected Revenue 70/3
30 𝑤𝑖𝑡ℎ 𝑝𝑟𝑜𝑏 2/3
{Revenue lottery with 𝑒 = 2} FOSD {Revenue lottery with 𝑒 = 1}

✘ Generally e is unobservable to the principal

✘ 𝑒 is non-verifiable to any third party (or the court)
✘ Therefore, e is non-contractible, cannot write contracts contingent on e.
✘ Revenue is verifiable and hence contractible.
5
 But first look at Observable effort
✘ The first best situation.
✘ Suppose 𝑒 is contractible.
✘ Suppose the principal wants to implement e=2
Q: What will be the optimal contract?
A: The employer will set a wage ‘w’ such that 𝑤 − 𝑒 − 1 ≥ 1
⇒𝑤≥4
Therefore, optimal wage is 𝑤 = 4
This wage once again will be paid irrespective of high or low revenue.
The corresponding net profit (𝜋)= 70/3 – 4 = 58/3

6
✘ Suppose the principal wants to implement e = 1
Q: What will be the optimal contract?
A: The employer will set a w such that 𝑤 − 𝑒 − 1 ≥ 1
⇒𝑤≥1
Therefore, optimal wage is 𝑤 = 1

This wage once again will be paid irrespective of high or low revenue.

The corresponding net profit (𝜋) = 50/3 – 1 = 47/3

7
✘ Under the FB contract principal offers the agent a fixed wage irrespective
of the state of nature.
✘ Principal is insuring the agent from any unobservable risk => full insurance
✘ This is Pareto optimal since the agent is risk averse (agent hates income
fluctuation).
✘ When 𝑒 = 2, 𝑤 = 4, 𝜋 = 58/3
✘ When 𝑒 = 1, 𝑤 = 1, 𝜋 = 47/3
Q: What will be the optimal contract?
A: It is optimal for the employer to implement 𝑒 = 2.
The optimal FB contract is {𝑒 = 2, 𝑤 = 4}

8
 Unobservable or non-contractible effort
✘ When effort is non-contractible, one has to write a contract contingent
on outcome (revenue).
✘ Thus the principal can at most offer a contract such that, “I will pay
𝑤30 if revenue is 30, I will pay 𝑤10 if revenue is 10”; 𝑤30 > 𝑤10.
✘ Paying fixed wage is not possible now, the principal needs to make the
agent face risk => the agent’s income (wage) will now fluctuate.
✘ The principal once again will maximize expected net revenue subject
to two constraints:
1) The Participation Constraint.
2) The Incentive Compatibility Constraint.

9
Suppose the principal wants to elicit e = 2 :
Agent’s expected utility from picking e =2 is
1/3 ( 𝑤10 − 1) + 2/3 ( 𝑤30 − 1)

Agent’s expected utility from picking e =1 is

2/3 ( 𝑤10 − 0) + 1/3 ( 𝑤30 − 0)

Participation Constraint ensures that the Agent gets her outside option.
Thus the Participation Constraint (under e =2):
1/3 ( 𝑤10 − 1) + 2/3 ( 𝑤30 − 1) ≥ 1

10
Incentive Compatibility Constraint ensures that the Agent picks e =2 over
e = 1 in self interest.

Thus the Incentive Compatibility Constraint will be:

1/3 ( 𝑤10 − 1) + 2/3 ( 𝑤30 − 1) ≥ 2/3 ( 𝑤10 − 0) + 1/3 ( 𝑤30 − 0)

This Incentive constraint is a new concept that we encounter in the ‘Theory

of Incentives’.

11
Principal’s Problem:
max 1/3 (10 − 𝑤10 ) + 2/3 ( 30 − 𝑤30 )
𝑤30 ,𝑤10

Subject to:
The Participation Constraint:
1/3 ( 𝑤10 − 1) + 2/3 ( 𝑤30 − 1) ≥ 1
The Incentive Compatibility Constraint :
1/3 ( 𝑤10 − 1) + 2/3 ( 𝑤30 − 1) ≥ 2/3 ( 𝑤10 − 0) + 1/3 ( 𝑤30 − 0)
Also assume Limited Liability:
𝑤30 ≥ 0 , 𝑤10 ≥ 0 ; Negative wages not allowed.
(Remember we are in a regime where the principal wants to elicit e = 2).
12
 𝑤30

Feasible IC
Contracts

𝑤10
0 6
PC

Solution: 𝑤30 = 9, 𝑤10 = 0

When eliciting 𝑒 = 2, the optimal contract is {𝑤30 = 9, 𝑤10= 0}.

Principal’s net expected payoff is

1/3 (10 −0) + 2/3 ( 30 −9) = 52/3 less than the FB payoff 58/3.

=> Eliciting high effort is costly.

The expected wage offered is 1/3X0+2/3X9 = 6.

The principal needs to offer a higher expected wage to elicit high

effort.

14
Suppose the Principal wants to elicit e = 1 (wants the Agent to sleep)
Principal’s Problem:
max 2/3 (10 − 𝑤10 ) + 1/3 ( 30 − 𝑤30 )
𝑤30 ,𝑤10

Subject to:
The Participation Constraint:
2/3 ( 𝑤10 − 0) + 1/3 ( 𝑤30 − 0) ≥ 1

No need for the Incentive Compatibility Constraint anymore.

No need to incentivize to elicit low-effort.

15
Making an agent sleep is easy.

Just pay a fixed wage across states of nature and only ensure that the PC
is satisfied.

The optimal contract is {𝑤30 = 1, 𝑤10 = 1} which is a fixed wage contract

similar to the First Best contract while eliciting low-effort.

So eliciting low effort doesn’t entail any extra cost to the principal.

Principal’s net expected payoff while eliciting e = 1 is 47/3.

16
Since 52/3 > 47/3, it is optimal for the principal to elicit e = 2 .

The optimal contract under non-contractibility is {𝑤30 = 9, 𝑤10 = 0}.

The agent in her own interest puts in e = 2 .

17
 Moral Hazard with discrete effort and continuous outcomes (For
Advanced readers)
✘ Employer- Worker framework
✘ Employer is the principal and worker is the agent
✘ Employer is risk neutral and agent is risk averse
✘ The agent puts in non-observable effort e
✘ The principal observes profit 𝜋 ∈ 𝜋, 𝜋 (continuous variable)
✘ Profit is stochastically related to e by conditional density function 𝑓 𝜋 e >
0 for all 𝑒 ∈ 𝐸 and all 𝜋 ∈ 𝜋, 𝜋
✘ Let there be only two effort levels -𝑒𝐻 & 𝑒𝐿
✘ 𝐹 𝜋 𝑒𝐻 ≤ 𝐹 𝜋 𝑒𝐿 at all 𝜋 ∈ 𝜋, 𝜋 with strict inequality in some open set
𝜋 ∈ 𝜋, 𝜋 .

18
 Moral Hazard
✘ The agent has an utility function 𝑢 𝑤, 𝑒 , where e is effort and w is wage,
𝑢𝑤 > 0, 𝑢𝑤𝑤 < 0, 𝑢𝑒 < 0 for all 𝑤, 𝑒
✘ Let 𝑢 𝑤, 𝑒 = 𝑣(𝑤) − 𝑔(𝑒), where 𝑔(𝑒𝐿 ) < 𝑔(𝑒𝐻 )
✘ Reservation utility u = 𝑢
✘ The wage cannot be conditional on 𝑒 as it is unobservable and non-
verifiable
✘ Profit of the firm is observable and verifiable , therefore 𝑤 can only be
contracted on 𝜋.
✘ Agent’s effort is assumed to help increase the revenue of the employer.

19
 Objective of the principal
✘ Effort observable

Participation/
Individual
Rationality
20 constraint
 Optimal Contract
✘ Optimal wage would be fixed such that

✘ Optimal effort choice depends on the one which maximizes the

expected profits of the principal

21
Proposition 1: Observable effort

22
 Risk Neutral agent
✘ 𝑢 𝑤, 𝑒 = 𝑤
✘ Optimal wage will be
𝑤* = 𝑢 + 𝑔(𝑒)
Optimal effort choice depends on the one which maximizes the
expected profits of the principal
𝜋𝑓 𝜋 𝑒 𝑑𝜋 − 𝑢 + 𝑔(𝑒)

23
 Unobservable effort
✘ Problem: agent may take lower effort and claim higher wage or
agent may offer high effort but principal claim that the agent has put
lower effort and offers lower wage- possibility of reneging contract

✘ Way out? Elicit the desired effort through a carefully designed

incentive structure.

24
Objective of principal

IR
constraint

Incentive
Compatibility
(IC)constraint

25
 Revisit Risk Neutral agent
For risk neutral agents moral hazard is no problem. The principal can offer a
contract that elicits the first best effort without any cost.
Consider the contract 𝑤 𝜋 = 𝜋 − 𝛼 where 𝛼 is fixed.
This contract is basically ‘selling the project’ contract that makes the agent a
‘residual claimant’
✘ From the IC the agent choose e by maximizing
𝜋𝑓 𝜋 𝑒 𝑑𝜋 − 𝛼 − 𝑔(𝑒)
The optimal effort will be exactly the first best effort.
Optimal 𝛼 ∗ = 𝜋𝑓 𝜋 𝑒 ∗ 𝑑𝜋 − 𝑢 − 𝑔(𝑒 ∗) will give the agent exactly 𝑢 .
The above contract implements FB without any cost.
26
 Risk Averse Agent (Implementing High Effort (𝒆𝑯 ))
✘ Incentive compatibility constraint:

✘ Principal choses the wage such that expected payoff of the principal is
maximum given the IC and the PC constraint. Both will bind.
✘ The principal needs to pay higher expected wage to elicit 𝒆𝑯 => welfare loss
✘ The optimum wage schedule need not be monotonically increasing in 𝜋.
✘ It will be monotonically increasing in 𝜋 if the monotone likelihood ratio
property (MLRP) holds.
✘ For monotone likelihood ratio property see MWG page 484-485.

27
Implementing High Effort (𝒆𝑯 )

28
 Implementing low effort (𝒆𝑳 )
✘ Principal would offer fixed wage
✘ Therefore as good as first best situation under 𝒆𝑳 .

Intuition:
Implementing lower effort is easier and is therefore not a problem. It is similar
to the observable case.

For Detailed discussion:

See Mascolell, Whinston and Green, Section 14 B, Pages 478-488.

29