ECON5323 Organisational Economics
Noisy Incentives & Relative Performance Evaluation
, Recap
• In all previous models we’ve studied, there was a precise link between the
agent’s effort and (observable) output
• in the basic model, the only source of inefficiency was that the agent was
“lazy”
• in misaligned incentives and multitasking models, both output
(unobservable) and the performance measure (observable) were fully
determined by agent’s effort, only source of inefficiency was how strongly
each of the two efforts was incentivised
• What if the link between effort and output is not so precise?
, Randomness
• Realistically, in addition to the agent’s effort, there is often some
randomness involved that also influences output
• Restaurant manager and poor sales - how much of this is because he is a
bad manager and how much is due to factors outside of his control, such as
performance of other employees, weather, nearby road work, local law
changes, etc.
• Student and poor research paper - how much is due to low effort and how
much due to power/internet being down, etc.
• Randomness can also be in the positive direction
• One way to model this: output is 𝑦 = 𝑒 + 𝜀, where 𝜀 is a random factor we
call “noise”
• In these situations, the principal cannot directly reward the agent for his
efforts
, Our noisy incentives model
• We will be adapting our baseline model to incorporate this “random
noise”
• No imperfect performance measures added into the model (yet) - we
want to isolate the effect of this randomness in determining the
optimal contract
• We want our model to have the following features:
1. Incentive schemes must be based on imperfect measures of effort, i.e.
𝑦=𝑒+𝜀
2. Agents are risk-averse, i.e. they dislike risky outcomes
Noisy Incentives & Relative Performance Evaluation
, Recap
• In all previous models we’ve studied, there was a precise link between the
agent’s effort and (observable) output
• in the basic model, the only source of inefficiency was that the agent was
“lazy”
• in misaligned incentives and multitasking models, both output
(unobservable) and the performance measure (observable) were fully
determined by agent’s effort, only source of inefficiency was how strongly
each of the two efforts was incentivised
• What if the link between effort and output is not so precise?
, Randomness
• Realistically, in addition to the agent’s effort, there is often some
randomness involved that also influences output
• Restaurant manager and poor sales - how much of this is because he is a
bad manager and how much is due to factors outside of his control, such as
performance of other employees, weather, nearby road work, local law
changes, etc.
• Student and poor research paper - how much is due to low effort and how
much due to power/internet being down, etc.
• Randomness can also be in the positive direction
• One way to model this: output is 𝑦 = 𝑒 + 𝜀, where 𝜀 is a random factor we
call “noise”
• In these situations, the principal cannot directly reward the agent for his
efforts
, Our noisy incentives model
• We will be adapting our baseline model to incorporate this “random
noise”
• No imperfect performance measures added into the model (yet) - we
want to isolate the effect of this randomness in determining the
optimal contract
• We want our model to have the following features:
1. Incentive schemes must be based on imperfect measures of effort, i.e.
𝑦=𝑒+𝜀
2. Agents are risk-averse, i.e. they dislike risky outcomes
