Uncertainty and Information
Dr Bryony Reich
Topic 11:
Designing Incentives
To avoid distracting others, please turn off or silence your
phone and put it in your bag. Laptops/tablets should be in
airplane mode and not distract others in the class.
Adverse Selection
• These notes draw from Varian “Intermediate Economics” Chapter 36
Designing Incentives
• Designing incentives is really asking how do I design a system, based on economic benefits and costs, to
get an individual to do something.
• How should you design workers’ compensation to optimise workers’ productivity?
• Should the design of workers’ compensation include “perks”, such as the ability to work from home or
free meals in the office?
• How do you design incentives to get a salesperson to optimally balance quantity of sales with the quality
of customer service?
• How do you design incentives to give patients information about doctors and have doctors provide
top quality care?
• Economic theory provides a framework for understanding the optimal design of incentives.
Designing Incentives
• A simple example.
• A landowner owns a plot of land and wants to hire someone to farm the land for him. What sort of
compensation system should he set up?
• Question: Suppose he pays the farmer a lump sum fee, some £X. How much crop will he produce?
• An incentive system will often link, in some way, the farmer’s compensation to the amount of crop his
produces.
Designing Incentives
• Let 𝑥 ≥ 0 be the choice of effort expended by the farmer where a higher 𝑥 means higher effort.
• Let 𝑦 = 𝑓(𝑥) be the output produced by the farmer when he expends effort 𝑥, where 𝑓 ′ ⋅ > 0.
• Suppose the price of output is 1 so that 𝑦 also denotes the price received on the market for the total
output.
• Denote by 𝑠(𝑦) how much the farmer is paid if he produces output 𝑦.
• The landowner’s profit is his revenue from selling the output 𝑦 minus his costs which here are only labour
costs, 𝑠(𝑦). Thus, his payoff is 𝑦 − 𝑠(𝑦).
• Effort is costly for the worker. We let 𝑐(𝑥) denote the cost of effort to the worker, where 𝑐 ′ 𝑥 > 0,
𝑐 ′′ 𝑥 > 0. The worker’s payoff is his compensation minus his effort costs, 𝑠 𝑦 − 𝑐(𝑥).
• The worker’s payoff if he does not take the job is given by 𝑢 ≥ 0.
Designing Incentives
• This describes what is known as a principal-agent problem.
• The principal (the landowner) wants to the agent (the farmer) to undertake some action/s on his behalf.
• However, the agent’s incentives may not be aligned with the principal’s incentives.
• The principal wants to design an incentive system so the agent acts in the way that the principal wants
him to.
• What constraints does the landlord face in designing the incentive scheme?
• There are two constraints that arise here (and typically):
1. The participation constraint – ensuring the farmer will choose to take the job “participate in the employment
contract”.
2. The incentive compatibility constraint – ensuring the agent will choose the amount of effort the landowner
wants him to take.
Designing Incentives
• What is the participation constraint in this example?
• The utility the farmer gets from taking this job must be at least as much as the utility he would receive
elsewhere.
• His payoff from taking the job is 𝑠 𝑦 − 𝑐 𝑥 = 𝑠 𝑓(𝑥) − 𝑐(𝑥).
• Thus the participation constraint is 𝑠 𝑓(𝑥) − 𝑐(𝑥) ≥ 𝑢
• Of course, this needs to hold at the level of effort 𝑥 he will exert if he takes the job.
Designing Incentives
• What is the optimization problem for the landowner now?
• The landowner wants the farmer to exert a level of effort that maximises the landowner’s payoff subject
to the participation constraint, that is, subject to ensuring that the farmer will take the job:
𝑚𝑎𝑥𝑥 𝑓 𝑥 − 𝑠(𝑓(𝑥))
𝑠𝑢𝑏𝑗𝑒𝑐𝑡 𝑡𝑜 𝑠 𝑓 𝑥 − 𝑐 𝑥 ≥ 𝑢
• The landowner is maximising when he sets 𝑠 𝑓 𝑥 − 𝑐 𝑥 = 𝑢 meaning the farmer is just indifferent
between taking the job or not.
• Substitute this into the objective function to get
𝑚𝑎𝑥𝑥 𝑓 𝑥 − 𝑐 𝑥 − 𝑢
• Let’s suppose there is an amount of effort denoted 𝑥 ∗ > 0 which is the unique value of 𝑥 that maximises
this problem such that the marginal product of effort is equal to the marginal cost 𝑓′(𝑥 ∗ ) = 𝑐′ 𝑥 ∗ .
Designing Incentives
• The landowner would like to incentivize the farmer to exert effort 𝑥 ∗ .
• Notice that the optimal effort the landowner would like to incentivize is when the marginal effort, the
extra output from more effort, is equal to the farmer’s marginal cost of effort.
• The intuition for this is that because the landowner has to pay the farmer enough to take the job, the
landowner has to pay him for his costs of effort and therefore they enter the landowner’s maximization
problem.
Designing Incentives
• What is the incentive compatibility constraint here?
• For the landowner to incentivize a level of effort 𝑥 ∗ the landowner should set a compensation scheme 𝑠 𝑦
such that the farmer’s payoff from effort 𝑥 ∗ is higher than his payoff from other effort levels
𝑠 𝑓 𝑥 ∗ − 𝑐 𝑥 ∗ ≥ 𝑠 𝑓 𝑥 − 𝑐 𝑥 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑥
• This is the incentive compatibility constraint.
• We said the incentive scheme has to satisfy two conditions:
1. It has to give the farmer utility 𝑢.
2. It has to ensure his utility from 𝑥 ∗ is higher than his payoff from other effort levels.
• There are different types of compensation schemes that can implement this
Designing Incentives
• We said the incentive scheme has to satisfy two conditions:
1. It has to give the farmer utility 𝑢.
2. It has to ensure his utility from 𝑥 ∗ is higher than his payoff from other effort levels.
• Pay to play
• The landowner rents the land for some price 𝑅 the farmer then produces crops on the land and keeps the
revenue himself.
• The compensation scheme is therefore 𝑠 𝑓 𝑥 = 𝑓 𝑥 − 𝑅.
• The farmer chooses 𝑥 to maximize 𝑠 𝑓 𝑥 − 𝑐 𝑥 = 𝑓 𝑥 − 𝑅 − 𝑐(𝑥) which is maximized for 𝑥 ∗ !
• The landowner must set 𝑅 such that the payoff the farmer receives is 𝑠 𝑓 𝑥 ∗ − 𝑐 𝑥∗ = 𝑓 𝑥∗ − 𝑅 −
𝑐 𝑥 ∗ = 𝑢.
• So 𝑅 = 𝑓 𝑥 ∗ − 𝑐 𝑥 ∗ − 𝑢.
• Where do we see this kind of incentive scheme in real world contracts?
Paying to Play
Source: filmforum.org
Designing Incentives
• We said the incentive scheme has to satisfy two conditions:
1. It has to give the farmer utility 𝑢.
2. It has to ensure his utility from 𝑥 ∗ is higher than his payoff from other effort levels.
• Wage labour
• The landowner pays the farmer a constant wage 𝑤 per unit of labour, plus a lump sum 𝐾.
• The compensation scheme is therefore 𝑠 𝑓 𝑥 = 𝑤𝑥 + 𝐾.
• Set the wage equal to the marginal product of the farmer at 𝑥 ∗ that is 𝑓′(𝑥 ∗ ).
• Then the farmer solves 𝑠 𝑓 𝑥 − 𝑐 𝑥 = 𝑓 ′ 𝑥 ∗ 𝑥 + 𝐾 − 𝑐(𝑥) and will choose effort such that 𝑓 ′ 𝑥 ∗ = 𝑐′ 𝑥
which holds at 𝑥 = 𝑥 ∗ .
• Then to satisfy the participation constraint, 𝑠 𝑓 𝑥 ∗ − 𝑐 𝑥 ∗ = 𝑓 ′ 𝑥 ∗ 𝑥 ∗ + 𝐾 − 𝑐 𝑥 ∗ = 𝑢. Hence the
landowner sets 𝐾 = 𝑢 + 𝑐 𝑥 ∗ − 𝑓 ′ 𝑥 ∗ 𝑥 ∗ .
Designing Incentives
• We said the incentive scheme has to satisfy two conditions:
1. It has to give the farmer utility 𝑢.
2. It has to ensure his utility from 𝑥 ∗ is higher than his payoff from other effort levels.
• Take it or leave it
• The landowner pays the farmer a lump sum 𝐵 ∗ if the farmer works 𝑥 ∗ and 0 otherwise.
• The compensation scheme is therefore 𝑠 𝑓 𝑥 = 𝐵 ∗ if 𝑥 = 𝑥 ∗ and 𝑠 𝑓 𝑥 = 0 if 𝑥 ≠ 𝑥 ∗ .
• The amount 𝐵 ∗ is then determined by the participation constraint so that 𝑠 𝑓 𝑥 ∗ − 𝑐 𝑥 ∗ = 𝐵 ∗ − 𝑐 𝑥 ∗ = 𝑢.
• 𝐵 ∗ = 𝑐 𝑥 ∗ + 𝑢 so the compensation 𝐵 ∗ compensates the farmer for the cost of his effort and for his outside
option 𝑢.
Designing Incentives
• We said the incentive scheme has to satisfy two conditions:
1. It has to give the farmer utility 𝑢.
2. It has to ensure his utility from 𝑥 ∗ is higher than his payoff from other effort levels.
• Each of the three schemes we discussed are equivalent. They each provide the farmer with the incentive to
work 𝑥 ∗ and each give him utility 𝑢 given he works 𝑥 ∗ .
• What would a nonoptimal incentive scheme look like? “Sharecropping”
• Consider a very standard compensation scheme where the landowner and the farmer each get some
percentage of the output.
• Suppose the farmer’s compensation is 𝑠 𝑓 𝑥 = 𝛼𝑓(𝑥) + 𝐹 where 𝛼 < 1 and 𝐹 is a constant.
• The farmer’s payoff is 𝛼𝑓 𝑥 + 𝐹 − 𝑐(𝑥) and so he sets effort 𝑥 to satisfy 𝛼𝑓′ 𝑥 = 𝑐′(𝑥) rather than 𝑓′ 𝑥 =
𝑐′(𝑥) which gives effort 𝑥 ∗ .
Designing Incentives
• We said the incentive scheme has to satisfy two conditions:
1. It has to give the farmer utility 𝑢.
2. It has to ensure his utility from 𝑥 ∗ is higher than his payoff from other effort levels.
• What is common about the way these optimal incentive schemes are structured?
• The person who is making the effort decision should be the residual claimant to the output.
• The way the landowner makes his payoff as high as possible is by getting the worker to put in the profit
maximizing level of effort 𝑥 ∗ .
• This is the level of effort at which the marginal product of the worker’s extra effort is equal to the worker’s
marginal cost of that effort.
• To get the worker to exert this level of effort, the incentive scheme must set the marginal benefit to the worker
of extra effort equal to his marginal product of extra effort.
Designing Incentives
• Another way to put this, if you want the farmer to put in the optimal amount of effort you need to give him the
true value of his effort.
Chinese Economic Reforms
• Prior to 1979 Chinese rural communes were organized such that workers were paid a rough estimate of how
much they contributed to the commune income.
• In 1978 the Chinese government instituted major agricultural reforms. A very simplified version of those
reforms is the following.
• This system was such that any production above a fixed quota was kept by the household and could be sold on
private markets.
• Restrictions on private plots were also removed.
• This system is very much like the optimal incentive mechanism above. Each household makes a lump sum
payment and then keeps the excess which sets his marginal benefit equal to his marginal product.
• From 1978-1984 agricultural output increased 61% (economists have tried to estimate how much is due to the
change incentive scheme – estimate approx. 3/4th).
Why do we see Sharecropping?
• If this is optimal, why do we see sharecropping in the real world? (And versions of it in all kinds of incentive
schemes beyond agriculture and for many centuries.)
• What is left out of the model?
• We essentially assumed that the landowner could see effort. In most scenarios this is unrealistic.
• At best landowners may observe output which depends on effort but also on lots of other factors outside the
control of the agent, such as weather, quality of the inputs available, and so on.
• Making output a “noisy” signal of effort.
• This is a problem of asymmetric information: the worker knows his own effort, but the landowner does not.
• Output is risky – can depend on things outside the control of the worker.
• The landowner will need to adjust the incentive scheme accordingly.
Why do we see Sharecropping?
• Consider pay to play – the landowner rents the land and the farmer keeps his output to sell. Or the scheme
where the landowner pays the farmer a wage based on his output.
• Since output depends on weather, crop failure, and so on, if the farmer is paid according to his output and so
his compensation becomes risky.
• Typically, workers are more averse to these risks than landowners. The landowner will have to pay the worker to
face these risk costs.
• Sharecropping, in contrast, splits the output between the farmer and the landowner and therefore splits the
risk. It maintains some incentives for the farmer to work but reduces the risk he faces.
