Management and Markets

1
Management and Markets

• Agency Models and Incentives
• (usual thanks to George Baker)

2
Course Content (Strategy and Org Econ.)

• Overview Introduction to Strategy
• Market Competition (sessions 2-6)
• Theory of the Firm (sessions 7-8)
• Internal Organization Incentives and
  Performance Management (sessions 9-11)
• Summary and Application to Innovation

3
Conventional Principal-Agent Problem

• Principal and agent
• Principal owns the firm, gets output V
• Agent works, exerting effort e
• Solve for the (second best) optimal sharing rule
• This problem emphasizes tradeoff between effort
  inducement and risk aversion of agent

4

• Production function Output V
• V ge e
• Effort e is unobservable to the principal
• g is the marginal product of agent effort
• e represents the random factors that affect
  output

5

• Preferences -- Principal
• Principal cares only about money, and is risk
  neutral (maximizes expected value)
• Pays the agent S bV (salary and incentive pay
  based on rate b times outcome V)
• Keeps V-S-bV
• Chooses S and b to maximize EV-S-bV

6

• Preferences -- Agent
• Agent is risk averse, and cares about money and
  effort.
• Likes money
• Dislikes effort and variance in pay s2
• Has an alternative worth H
• EU ES bV - e2/2 - rs2pay
• r is a parameter capturing degree of risk
  aversion

7
Timing

• Principal offers agent a contract
• Agent rejects or accepts
• If rejects, agent gets H
• If accepts, proceeds to next step
• Agent exerts effort, generates value
• Principal and agent get their payoffs

8
Formal statement of the problem

• Principals problem
• MaxS,b E(V - S - bV)
• subject to
• (1) the contract must give the agent utility H
• Participation constraint
• (2) the agent will choose e to maximize his
  utility
• Incentive compatibility constraint

9
Solution Procedure

• Use IC constraint to solve for e(b)
• Use PC to solve for S(b)
• Maximize the principals expected profit as a
  function of b

10
Incentive Compatibility constraint

• Agent maximizes utility
• MaxeE(Sb(ge e)) - e2/2 - rs2pay
• e gb
• Higher bonus rate increases effort
• Higher marginal product increases effort

11
Participation Constraint

• ESbV - e2/2 - rs2pay H
• S H - EbV e2/2 rs2pay
• H - EbV (gb)2/2 rb2s2e

12
Principals problem

• Maxb EV - S - bV
• Maxb g2b - H - E(bV) (gb)2/2 rb2s2e -
  E(bV)
• Maxb g2b - H - (gb)2/2 - rb2s2e

13
Solution

• b

14
Risk and Incentives

• Standard agency model predicts negative
  relationship between risk and incentives
• What is the empirical evidence on this?
• Prendergast 2003, Tenuous trade-off
• Allen and Lueck 1992 on share-cropping
• Crops with more variable yield per acre are more
  likely to use share-cropping than fixed rent
  contract
• Martin 88 and Lafontaine 92 on franchising
• Proportion company-owned decreases with the
  variability of outlet sales
• Likelihood of franchising increases with the
  proportion of discontinued outlets

15
Incentive alignment and the agency problem

• Why are incentives weaker in organizations than
  in markets?
• Risk preferences
• Distortion in performance measurement

16
Distortion in performance measurement

• On the folly of paying for A while hoping for
  B. (Kerr 1975)
• Multi-task agency problems
• Holmstrom Milgrom 1991
• Pay for what can be measured
• Distortion
• Baker 1992, 2002
• Paying for A while hoping for B

17
Holmstrom and Milgrom (JLEO 91) Multitask
Principal-Agent Problems

• Agents task has several dimensions (e.g.
  Volume and quality) and at least one dimension
  easy and one hard to measure
• Incentive pay
• Induces effort and allocates risk (conventional
  treatment)
• Allocates attention among various dimensions (new
  here)
• When harder to measure other dimensions then less
  desirable to provide incentives on measurable
  dimensions
• Including attention leads to explanation for
  lower-powered incentives that are often observed
  in practice
• Job design is instrument to constrain agents
  attention allocation
• Lowers marginal opportunity cost by changing
  incentives to engage in competing tasks
• Useful when performance of agent on central task
  is hard to measure

18
Baker (1992 and 2002)

• Principals objective is non-contractible
• Private company
• Non-profit
• Government agency
• Public company
• How good a performance measure is the stock price
  for most employees?

19
Performance Measurement

• In the standard model
• There is no way for a agent to do the wrong
  thing
• There are no dysfunctional consequences
• The principal knows what she wants the agent to
  do
• She just cant observe effort
• There is no notion of specific knowledge

20
A richer model

• Agents can do the wrong thing
• Sears Auto Centers
• Agents can game the system
• Joseph Jett at Kidder, Peabody
• Well intentioned systems can go wrong
• Education reform and standardized tests

21
Baker 2002 Model

• V is the principals objective
• It is known to the principal, but is not
  contractible
• P is a performance measure
• It is a contractible outcome, but it is not
  exactly what the principal wants
• Two questions
• How will firms use these performance measures in
  incentive contracts?
• What are the characteristics (statistical and
  otherwise) that make for a good P?

22
Analytical frameworkA multi-task model

• Organizational objective
• V fe e
• where
• e is a vector of tasks
• f is a vector of marginal products on these tasks
• e represents uncontrollables

23
A distorted performance measure

• P ge ?
• where
• e is the same vector of tasks
• g is a vector of marginal products on these tasks
  (which might be different from the fs)
• ? represents uncontrollables (which might be
  correlated with e)

24
Distortion

• Some elements of the vector g are different from
  those of the vector f
• That is, there are things that the agent can do
  that increase the performance measure, but not
  the organizations objective (and vice versa)
• Cooking the books
• RD spending
• Teaching to the test

25
Contracting with distortion

• Pay S bP
• Agent is risk neutral
• What will the optimal contract look like?
• The strength of optimal incentives
• Decreases with an increase in distortion (cos ?
  smaller)
• Decreases w/ an increase in risk aversion (h) and
  noise (s2)
• Increases with agents marginal product (f)

26
Comparative statics (cont.)

• What about the correlation of P and V?
• The optimal incentive contract is independent of
  the correlation between P and V!
• What matters is not whether the measure is
  correlated with value, but rather whether the
  margins are correlated

27
Job Training Partnership Act (JTPA)Heckman,
Heinrich, and Smith (1997)Courty and Marschke
(1997)

• True objective
• Increase the employability of the disadvantaged
• Performance measure
• Employment outcomes of those in the program
• Problems
• Cream skimming
• Strategic termination

28
Customer satisfaction Ittner, Larcker and Meyer
(1997)

• Profitable bank branches have high levels of
  customer satisfaction
• True objective
• Well-run bank branches
• Performance measure
• Customer satisfaction
• Result
• Within a year, the branch managers with the
  highest bonuses were those in the least
  profitable branches

29
Degradation

• Performance measures that are good predictors of
  performance degrade once they become the basis
  for incentive contracts
• NCLB (no child left behind) and test scores
• Koretz, Linn, Dunbar, and Shepard, 1991
• Test scores go up on high stakes tests, but not
  on otherwise similar low stakes tests

30
Controllability and distortion

• Paying for accounting profits or stock prices
• Responsibility accounting
• Rewarding for individual or group performance
• Rewarding for long-term or short-term performance
• Relative performance evaluation (filter out
  noise)
• What the are tradeoffs?

31
CEO
Division Manager
Division Manager
Sales Manager
Plant Manager
Plant Manager
32
Autonomous Strategic Process at Intel(Burgelman
91)

• Middle managers try to get organization to wrok
  on thing outside current strategy
• RISC chips as disguised co-processor
• Add-on boards via corporate venturing
• Need for bottom-up internal experimentation
• What incentive structure would you design to
  achieve this? (how implement?)
• How can firms balance induced and autonomous
  processes?
• Relate to Built to Last or GE Welch?

33
Questions

• Limitations of incentive contracts that we have
  have looked at so far
• Range of instruments explored too narrow. Arent
  other mechanisms useful (e.g. promotion,
  financial contracts, internal governance
  mechanisms) and how would their use affect the
  results of the models?
• Agents motivated by other than payhow does this
  affect value of these models?
• As agents differ on key motivating factors, can
  these contracts be effective in a
  one-size-fits-all implementation or is customized
  (and expensive) implementation needed?
• How can these contracts be used to motivate..? Do
  they introduce additional bad side effects?
• joint or team production?
• creative job workers?
• Is the tenure system efficient?

34
Conclusions positive implications

• Why are incentives in organizations weak?
• Most performance measures are distorted
• Organizations reduce the strength of optimal
  incentive plans to account for this distortion
• Organizations use strong incentives when
  performance measures are controllable and not too
  distorted

35
Conclusions normative implications

• Organizations provide rewards even when they are
  not designed
• Promotions
• Perks
• Status and prestige
• Attaboys

36

• Conclusions 2
• Organizations often get incentive provision
  wrong!
• The folly of paying for A while hoping for B
• Models like these can help those in organizations
  to understand what makes incentive systems work
  better
• Dont rely on correlated measures
• Analyze how agents can game the measures
• Anticipate performance measure degradation