ECON5323 Organisational Economics
Perverse incentives
, Recap
• Our basic principal agent model had the following relationship between
output 𝑦 and effort 𝑒
𝑦=𝑒
• And the principal’s payoff was
𝑈! = 𝑦 − 𝑤
• What was desired by the principal? 𝑦
• What was observable and contractable? 𝑦
• What are the determinant(s) of 𝑦 ? Only 𝑒, the agent’s effort
, Misaligned incentives
• Most real-world relationships do not have such a straightforward
relationship between the effort placed by the agent, what can be observed
by the principal and what he wants
• We will now study a model in which what is desired and what is
observable are not the same
• i.e. the principal cannot reward the agent for exactly what he wants him to
do, he has to rely on some other measure of performance that is not
perfectly aligned with his own objectives
• Lead to “hidden action” or moral hazard problem. We will get back to it later
, Misaligned incentives
• There are now two tasks that the agent can do, task 1 and task 2, with
the agent choosing effort separately for each task: e1 for task 1 and e2
for task 2
• Agent’s payoff is now
1
𝑈" = 𝑤 − (𝑒#$ + 𝑒$$)
2
• So, both tasks are costly for the agent
• The principal benefits from the first task, but not the second:
𝑈! = 𝑦 − 𝑤
𝑦 = 𝑒#
• But 𝑒# cannot be separately rewarded because the principal does not
observe 𝑦, he only observes an “aggregate” performance measure
𝑥 = 𝑒# + 𝒅𝑒$
• We can think of 𝑑 as the degree of distortion of the performance measure:
the higher it is, the further 𝑥 is from 𝑦
Perverse incentives
, Recap
• Our basic principal agent model had the following relationship between
output 𝑦 and effort 𝑒
𝑦=𝑒
• And the principal’s payoff was
𝑈! = 𝑦 − 𝑤
• What was desired by the principal? 𝑦
• What was observable and contractable? 𝑦
• What are the determinant(s) of 𝑦 ? Only 𝑒, the agent’s effort
, Misaligned incentives
• Most real-world relationships do not have such a straightforward
relationship between the effort placed by the agent, what can be observed
by the principal and what he wants
• We will now study a model in which what is desired and what is
observable are not the same
• i.e. the principal cannot reward the agent for exactly what he wants him to
do, he has to rely on some other measure of performance that is not
perfectly aligned with his own objectives
• Lead to “hidden action” or moral hazard problem. We will get back to it later
, Misaligned incentives
• There are now two tasks that the agent can do, task 1 and task 2, with
the agent choosing effort separately for each task: e1 for task 1 and e2
for task 2
• Agent’s payoff is now
1
𝑈" = 𝑤 − (𝑒#$ + 𝑒$$)
2
• So, both tasks are costly for the agent
• The principal benefits from the first task, but not the second:
𝑈! = 𝑦 − 𝑤
𝑦 = 𝑒#
• But 𝑒# cannot be separately rewarded because the principal does not
observe 𝑦, he only observes an “aggregate” performance measure
𝑥 = 𝑒# + 𝒅𝑒$
• We can think of 𝑑 as the degree of distortion of the performance measure:
the higher it is, the further 𝑥 is from 𝑦
