Topics in Financial Economics Term 2

1
Topics in Financial Economics Term 2

• Lecture 4 Still more on payout policy
• Does it matter at all? Can firms influence it?
  still under construction
• Payouts and value
• Introduction to agency

2
Update on payout policy

• DeAngelo and DeAngelo (2006) criticise MM on the
  grounds that MM rule out any impact of dividends
  by assumption
• The MM irrelevance results arise in frictionless
  markets with fixed investment policy all
  financial structures and all payout policies are
  optimal, because all imply the same stockholder
  wealth.
• The only determinant of value in this model is
  the firms investment policy but the MM
  assumptions effectively force the firm to pay out
  100 of free cash flow in every period (after the
  fixed allocation to investment has been removed)
• DD modify the model to allow variable payouts
  as a result, some payout policies do not
  distribute 100 of FCF to current shareholders.
  Managers must make separate decisions regarding
  payout and investment policies to maximise
  shareholder value.
• There are then two possibilities
• Payouts and investment are both irrelevant,
  because costless contracting forces managers to
  make optimal choices (fully distributing FCF in
  effect)
• Both are relevant (if some choices are better
  than others)

3
More on DD

• Hold NPV of investment policy fixed, and let
  managers alter time-path through 0-NPV projects
• Payout optimality is that current shareholders
  (eventually) get the full value of current FCF.
  (Note the choice of payout policy may be
  indeterminate if more than one is optimal without
  being irrelevant which requires all to be
  optimal)
• The same is true of investment policy
• We will revisit this once we have done agency
  costs.
• The Dividend puzzle of Black (1976) suggests
  that taxes on payouts should lead firms to avoid
  dividends and other payouts.
• It is based on introducing taxes into the MM
  model (reverse of the tax shield argument for
  high debt)
• Essence is that when total wealth is fixed, it
  makes sense to put it where it bears least tax
• But if payouts go to (near) 0, stockholder wealth
  suffers, equity values will fall and WACCs will
  rise sharply

4
Taxes and dividends, continued

• If investment at time 0 is financed by shares
  sold at that date, dividends are paid both to
  those shares and to others outstanding at that
  time. Let V0 be the present value of payouts D1
  and D2 and qV0 be the share received by the t0
  shares. Also let r0 and r1 be the (one-period)
  interest rates at dates 0 and 1. The feasibility
  condition on payouts is
• Now suppose investors pay taxes of t1 and t2 on
  dividends in periods 1 and 2. Feasibility now
  requires
• Therefore, low future payouts prevent the firm
  from raising new capital at t0. But positive
  payouts will only disappear if the tt are close
  to 1. This resolves the dividend puzzle

5
DD, 3

• This suggests the need to take tradeoffs between
  e.g. issue costs (which encourage retained
  earnings) and agency costs (which well see
  discourage them)
• The life-cycle picture supports this young
  firms generally have more investment
  opportunities than access to internal capital,
  and thus issue equity but pay few dividends.
  Later on, they can pay dividends and repurchase
  stock.

6
Repurchases updated

• Brav et. al. (2005) update the Lintner survey on
  motives behind payouts, including repurchases.
• The empirical facts are
• Young firms pay few dividends
• Many firms prefer repurchases to dividends as
  ways to distribute residual FCF
• But many firms still pay substantial dividends
• The results indicate a pecking order of
  decisions
• Maintaining dividend levels is first-order, along
  with investment decisions
• Other aspects of payout policy (allocating FCF
  net of investment, liquidity and maintained
  dividend needs) are secondary
• Target payout ratios no longer drive decisions
• Firms that have not paid dividends will do so
  when a) earnings suddenly (and sustainably)
  increase and/or b) when institutional investors
  demand it.
• Repurchases are also second-order
• Managers view them as more flexible (they dont
  create precedents)
• Repurchases are thus increased when firms stock
  price is too low
• EPS also matters repurchases rise when
  investment opportunities fall, when equity
  flotation is adequate and when they wish to
  offset option dilution (earnings dilution by
  stock option compensation or employee stock
  plans). Some simply feel that fewer shares
  higher EPS others understand that the money used
  to buy back shares must otherwise earn less than
  WACC.
• Tax considerations do not feature heavily. A
  recent cut in US dividend taxation led 6 of
  non-payers to start paying dividends, but other
  increases had begun before the tax cut or
  involved stocks held by institutions
• More generally (Jagnnathan et. al. 2000),
• dividends tend to increase steadily over time
  while repurchases are very pro-cyclical
• Dividends are paid by firms with high permanent
  operating cash flows repurchases are used by
  firms with higher temporary (or non-operating)
  cash flows
• Repurchases are far more volatile

7
If dividends dont measure firm values, what does?

• Contexts
• valuing mergers and acquisitions targets
  correctly
• IPO investment bank evaluation
• analysts who do not believe in efficient markets
• Commonly used methods
• discounted cash flow (DCF) firm as a project
• comparable MA deals - estimates based on ratios
  like premium over stock market value two months
  before acquisition announcement
• comparable publicly traded firms similar, uses
  e.g. price/earnings ratios
• Liquidation value of firm is broken up and sold
  piecemeal

8
Discounted cash flow

• Firm has value because it generates cash for
  shareholders, so stream of cash flows and the
  discount rate determine how much it is worth
  paying for a firm
• Step 1 Finding the cash flows
• Cash flow after taxes, not net income
• Timing of cash flows is critical
• Only incremental cash flows (those which occur on
  the margin because you invest in the firm)
• Be consistent in the treatment of inflation
• Result free cash flow

9
More on DCF

• Step 2 discount using WACC
• The formula based on MM with taxes, bankruptcy
  costs
• weighted average of after-tax cost of debt rD(1 -
  tc) and after-tax cost of equity rE

10
Comparable mergers and acquisitions

• In a competitive market similar firms should sell
  for similar amounts
• Why? If there is a wide difference in selling
  prices buyers will tend to hold back until a
  cheap firm comes along
• There are some major problems
• Sometimes very hard to find recent, similar MA
• Typically, hard to find reliable information
• Theoretical considerations (real option value of
  waiting, durable goods monopoly, expectation
  frenzies/crashes, etc.
• Similarity may need to extend from target to
  acquiring entity and/or acquisition procedure.

11
Finding comparable MA

• Points of comparison
• type of business activity in which the firm is
  engaged
• size of the business
• form of ownership - closely held or publicly held
• capital structure
• degree of profitability
• competitive position within the industry
• historical growth rate
• physical facilities
• Look at ratios
• E.g. pre-announcement ratio of selling price to
  stock price time pre-announcement stock market
  price
• Other statistics concerning selling price of firm
  - multiple of sales, multiple of book, multiple
  of earnings, multiple of cash flow, etc. can be
  used to get some idea of what the firm's value is
  in a similar way. Weighting of comparable
  transactions should be based on how comparable
  the firms are.

12
A case study BP
13
The big picture
The end result?
14
(No Transcript)
15
Activities that affect firm values

• Dividends change value of the asset
• Repurchases change the number of shares
  outstanding (exposed to market)
• Divestiture sell assets to another firm that can
  use them more efficiently
• Spin-off grant independence to a division
• Equity carve out make division
  semi-independent parent retains controlling
  interest
• Tracking stock division completely controlled
  (does not have own board)

16
Differences
Best
Worst
17
Introduction to agency theory

• A one-way flow of power and information
• Concerned with optimal contracts can the
  principal get the (better-informed) gent to do
  what he (the principal) wants?
• As usual, two complications hidden information
  (agent knows more) and hidden actions (costly or
  impossible to see what agent does)
• Two channels through which these play out
• Signing the contract (selection)
• Complying with the contract (incentives)
• The results show that there is always an agency
  cost

18
Principal-agent theory

• Interaction between 2 parties one acts for the
  other
• Not always clear who is the agent depends on
  incentives. Agent is the one who can gain by
  cheating, and therefore needs incentives to
  fulfil the bargain.
• Problem 1 hidden action (moral hazard)
  principal cannot (freely) monitor/verify agents
  action
• Problem 2 hidden information (adverse
  selection) principal cannot test whether agents
  action was justified
• Also known as contract theory
• Assumes (uninformed) principal moves first
• Distinct from signalling, where informed party
  moves first

19
Agency reading

• Principal-agent theory
• Osborne and Rubinstein A course in game theory,
  ch. 10.
• D. Gale notes sec 3 http//www.econ.nyu.edu/user/g
  aled/finchap03.pdf
• Gale and Allen, Financial Innovation and Risk
  Sharing (1994) sec. 4.1
• Stole notes secs 1.1, 2.1 (more as interest
  dictates) http//gsblas.uchicago.edu/papers/lectur
  es.pdf
• Handa slides http//www.biz.uiowa.edu/class/6F215_
  handa/Pr2lec5.ppt
• Agency problems in financial economics
• Jensen, M. and W. Meckling (1976) Theory of the
  firm, managerial behaviour, agency costs and
  ownership structure, Journal of Financial
  Economics 3, 305-360 Link from Warwick machines
• Myers, S. (1977) Determinants of corporate
  borrowing, Journal of Financial Economics 5,
  147-175 Link from Warwick machines
• Grossman, S. and O. Hart (1983) An analysis of
  the principal-agent problem Econometrica 51,
  7-45 Link
• Shleifer, A. and R. Vishny (1997) A survey of
  corporate governance Journal of Finance, 52,
  737-783 Link from Warwick machines
• Murphy, K. (1999) Executive compensation, in
  Handbook of Labor Economics, O. Ashenfelter and
  D. Card (eds), 2485-2563

20
Simple model try a risky venture, split proceeds

• Agent takes action s in a (finite or interval)
  set A
• The state is w in a (finite) set W the state is
  determined by a
• The probability of w given a is p(wa)
• The revenue if w occurs is R(w)
• The agents utility from consuming c (bought with
  his share of revenue) and taking action a is u(c,
  a) we assume u(c, a) U(c) J(a)
• The principals utility from her share of revenue
  is V(c)
• Both utilities are strictly increasing, concave,
  twice continuously differentiable
• We assume the agents consumption is non-negative
  (liquidity or limited liability)

21
Pareto optimality

• Contract (a, h(.)), where a is the contracted
  action and h(w) gt 0 is the payment to the agent
  in state w.
• Principals problem choose contract to maximise
  principals utility subject to agents
  participationPP
• Theorem A contract is Pareto efficient iff it
  solves this problem
• If the (risk-sharing) payment h is always gt 0 and
  optimalB
• If a is interior and p and J are differentiable

22
Incentive efficiency

• If agents actions cannot be observed or
  verified, they must be incentive-compatibleIC
  
• A contract is incentive efficient if it is Pareto
  optimal relative to the set of incentive
  compatible contracts. It solves the original
  problem (PP) with the added IC constraint.
• Any solution to this enhanced problem (PP) is
  incentive efficient if the participation
  constraint always binds.
• To solve this problem
• Fix a and compute the payoff V(a) from providing
  incentives to do a.
• Now choose a to maximise V(a)
• Because U and V are concave, the first part is
  easier to solve than the whole, and gives many
  useful insights. If V is linear (the principal is
  risk-neutral) this just minimises the expected
  payment to the agent

23
Does moral hazard prevent first-best?

• If the principal is risk-neutral and the agent is
  strictly risk-averse, condition B implies that
  payment h is independent of state, hence of
  agents action. The agent will choose the
  cost-minimising action and first-best can be
  achieved only if the optimal action (for the
  principal) is also the cheapest (for the agent).
• If the principal is risk-averse and the agent is
  risk-neutral, condition B implies R(w)-h(w) is
  constant for interior solutions sell the firm
  to the agent, providing firm remains profitable
  (agent gets gt 0).
• In general, there is y s.t. R(w)-h(w) miny,
  r(w), and h(w) maxR(w) y, 0

24
If both parties are risk-averse

• The first-order condition for the first-best
  is
• The first-order condition for incentive
  compatibility is
• The first-best and incentive-efficient contracts
  only coincide if

25
An example 2 states W (win) or L (lose)

• We assume winning is worthwhile R(W) gt R(L)
• The necessary condition is
• If effort is productive (i.e. pgt0), this implies
  paying the agent the full marginal
  contributionR(W)-R(L)h(W)-h(L)
• This does not satisfy condition B unless the
  agent is risk neutral on the interval h(L),
  h(W).

26
Optimal incentives, 1

• Assume the principal is risk-neutral (always
  prefers the action with higher expected revenue
  if costs are equal).
• There are a finite number of states and actions,
  strictly increasing revenue
• The monotone likelihood ratio property (MLRP)
  for any actions altb, the ratio p(w,b)/p(w,a) is
  non-decreasing in w. Also assume that the vectors
  p(.,b) and p(., a) are distinct (ratio sometimes
  rises). Expected revenue is increasing in a.

27
Optimal incentives, 2

• Because it only uses a one-sided IC constraint,
  is at least as big as V. Suppose it is bigger
  for some a, so the agent wants to choose more
  effort but the principal will always prefer
  more effort if it comes at the same price, so the
  a that maximises the two V-functions will be the
  same., so we can limit attention to the modified
  problem.
• Most incentive schemes are increasing a higher w
  brings a higher h though not necessarily a
  higher h(w)/R(w). Is this true of the optimal
  scheme?
• Rearranging the Kuhn-Tucker conditions for an
  interior solution to the modified problem, we
  get
• By the MLRP, the RHS is non-increasing in w so
  too is the LHS, hence the optimal h(w) is also
  non-decreasing

28
Examples

• 2 states W (win), L (lose)
• 2 projects or effort levels (A,B)
• let 0 lt pA p(W,A) lt pB p(W,B) lt 1
• Effort cost J(A) 0 lt J(B).
• U(0) reservation utility 0.
• To implement the inferior project (A), set
• h(L) hL h(W) hW 0
• Constraints to implement BICIR

29
Relation between IC, IR constraints

• Can rewrite IC as
• IR implies positive consumption in at least one
  state, so
• Therefore, IC implies IR holds strictly (doesnt
  bind)
• We can thus rewrite principals problem as

30
Optimum and project choice

• Clearly, optimality implies hL 0, and the
  optimum level of hW satisfies pB -
  pAU(hW)J(B)
• What makes managerial compensation hW higher?
• Higher effort cost J(B)
• Higher managerial risk aversion U(hW) U(hL)
• Lower managerial marginal productivity pB - pA
• Optimal project compare profits from both schemes

31
Incentive-efficiency example

• Set-up as before (2 states, 2 effort levels, 0
  reservation utility, risk-neutral agent,
  principal)
• Also assume
• R(L) 0 lt R(W) revenues in 2 states
• J(A) 0 lt J(B) effort cost to 2 actions
• 0 lt pA lt pB lt 1 probability of winning state
  for 2 actions
• Optimal implementation
• Project A hW hL 0
• Project B hL 0 hW J(B)/pB pA
• Principals payoff V(w)
• Project A pAR(W)
• Project B pBR(W) - J(B)/pB pA
• If it happens that v(W) V(L), the contract
  implementing project A solves the principals
  problem, but is not incentive efficient, because
  the project implementing B gives the agent
  strictly higher utility.

32
Corner solution example

• In previous example, the agent gets 0 in state L
  regardless of which action (project, effort
  level) he chooses. You might think this is due to
  risk neutrality, but corner solutions are
  possible even when U(0) ?. Suppose U(h) ha
  for 0 lt a lt 1.
• U(0) is still 0, so the optimal incentive scheme
  implementing A in the above example is still h(w)
  0
• To implement B, the optimal scheme is h(L) 0
  and h(W) J(B)/(pB pA)(1/a)
• Thus it is not always safe to simply assume
  interior solutions

33
Local conditions example

• It may be enough to check IC conditions locally
• Finite set of actions (effort levels) just check
  nearest ones
• Continuous set of actions check that derivative
  of agents payoff is 0 at specified action ( 2OC
  if needed)
• Not always safe
• 2 states, 3 projects (A, B, C), risk neutrality,
  0 reservation utility
• R(L) 0 lt R(W)
• J(A) 0 lt J(B) J(C)
• 0 lt pA lt pB lt pC
• Optimal scheme to implement C is hL 0 hW
  J(C)/pB pA
• Because B costs the same but offers lower
  probability of success, agent will never defect
  from C to B, but might defect to A if payment in
  state L is too low. To prevent this, we must
  assume a sort of diminishing marginal return to
  effort
• For more general conditions, see Stole notes

34
Mechanism design

• PA set-up is a special case of a more general
  problem of institutional design if the desired
  result depends on dispersed information and
  individual actions, can we design a game which
  gives the desired result as an equilibrium?
• A simple set-up many agents i, with types qi in
  Qi (known only to them) and actions ai. Let a in
  A and q in Q (without subscripts) be the actions
  and types for the whole population. The common
  knowledge distribution of types is p(q), and
  agent is utility is ui(a,q).
• The design problem is two-fold
• How to get agents to reveal the relevant
  information (Stage 1 agents send signals mi in
  Mi to principal)
• How to get agents to take the right action (Stage
  2 principal tells agents what to do m0 in M0)
• Each type of each agent has to choose a message
  strategy mi(qi) and a response function ai(m0,qi)
  to maximise its expected payoff conditional on
  its type. Effectively, we have a game between all
  the types of all the agents, and the equilibrium
  is called a Bayesian equilibrium.
• A mechanism (M, f) consists of a message space M
  for the agents and a decision rule f(m) in A for
  the planner

35
The revelation principle

• A direct mechanism has Mi Qi and M0 A in
  other words, agents report their types and the
  principal tells them exactly what to do. In a
  truthful strategy, agents report their true types
  and do exactly what the principal suggests.
• Theorem if (m, a) is a Bayesian equilibrium for
  some mechanism (M, f) there is an equivalent
  Bayesian mechanism for the associated direct
  mechanism.
• This lets us greatly simplify the institutional
  design problem by concentrating on direct
  mechanisms.
• A societal choice rule f associates a desired set
  of actions f(q) to each state of nature q.
• The rule f is incentive compatible iff for every
  agent telling the truth and obeying the
  principals suggestion is optimal assuming
  other agents are truthful as well.
• A direct mechanism has a truthful equilibrium iff
  the societal choice rule is incentive compatible.

36
So what?

• This is a weak notion of implementation
• Honesty and truth are only equilibria more
  restrictive conditions apply if we want them to
  be dominant strategies
• A direct mechanism may have other, untruthful
  equilibria
• The societal choice rule may embrace many
  different outcomes implementability just says
  that at least one can be achieved by a direct
  mechanism
• A large literature tries to improve on this by
  using stronger solution concepts and/or fancier
  game forms.
• The PA problem is a special case with only two
  agents
• the principal has no action to choose, the
  agent picks his effort level and the designer
  picks the contract (a,h).
• The societal choice rule is the
  incentive-efficient outcome.
• Optimal mechanisms usually leave the agent
  indifferent between effort levels, showing the
  multiple equilibrium problem
• To fix the problem in the last example, let the
  principal pick the incentive scheme first and
  have the agent respond with a best reply if the
  agent wishes to choose the effort level (A in the
  example) that is worse for the principal, the
  principal will instead pick a nearby scheme h
  which would make the agent strictly prefer B.
  (The original scheme was not subgame perfect)

37
No, I mean what has this got to do with finance?

• The analysis we did up till now treated the
  firm as a single entity that maximised something
  usually value
• Empirical data relating to leverage, dividends,
  executive compensation, takeovers, etc. does not
  really confirm this view
• Jensen and Meckling suggested viewing the firm as
  players of a game, with differing information,
  incentives and powers of action.
• We will look at several moral-hazard agency
  problems
• Risk shifting or asset substitution equity
  holders get paid whatever is left over after debt
  claims are met, and want the management to
  maximise this expected value. If the firm
  undertakes a very risky project, this may
  increase the expected payout to equity holders
  (truncated at 0) while also increasing the risk
  of bankruptcy (Think St. Peterburg paradox). This
  hurts bond holders, who are not compensated by
  the excess return if the firm wins
• Managerial effort if managers are paid salaries,
  they will not act in the interests of equity
  holders. But if they are paid as equity holders,
  they may undertake negative NPV projects (in the
  above example). If the interests of equity
  holders depart from NPV, the struggle for
  corporate control may distort the application of
  the NPV criterion
• Debt overhang if a firm has outstanding debt
  obligations, equity holders might not wish to
  undertake even safe, positive-NPV projects if the
  proceeds go to bond holders.

38
OK, but what has this got to do with the formal
analysis?

• The formal analysis can be used to examine the
  incentive effects of debt and equity
• The mechanism design analysis can be used to
  analyse and design executive compensation schemes
• We can also consider the extent to which specific
  institutions can be interpreted as optimal (or
  optimisable) contracts
• Debt
• Corporate control (LBOs MA, etc.)

39
Risk shifting

• A firm owes its debt holders 5 Million, but its
  retained earnings are only 1 Million. It has one
  year to make good or else it will go bankrupt
• The firm has a risky prospect if it invests 1
  Million, it gets 25 Million with probability p,
  and 0 with probability (1-p). Interest rate is 4
• The present value of the project is - 1 M (p
  25 M)/(1r) which is negative if p is less
  than (1r)/25 do the project if p gt 4.16
• If the equity holders do nothing, they get 0 if
  they undertake the project they get (in
  expectation) p 20 Million/(1r) always do
  the project
• The bondholders get p 5 Million/(1r) if the
  project goes ahead otherwise they get 1
  Million/(1r) do the project if p gt 20

40
A different perspective (r 0)

• Do nothing
• Value of debt 1 Million
• Value of equity 0
• Total 1 Million
• Undertake risky project if p 2
• Value of debt 2 5 Million 100,000
• Value of equity 2 20 Million 400,000
• Total 500,000
• Note equity holders gain less than bond holders
  lose (like Prisoners Dilemma), so there is
  certainly scope for improvement.

41
A model of risk shifting

• Managers actions a ? A states w ? W (finite)
  managerial impact p(w,a) revenue R(w) gt 0.
• Utility functions
• Agent (manager) u(a,h) U(h) z(a)
• Principal V(R h)
• both twice continuously differentiable, strictly
  increasing, concave
• Risk shifting manager has convex payment scheme
  (h) and prefers riskier projects. Think of
  principal as bond holder(s) and agent as acting
  on behalf of equity holder(s)
• Fixed costs of financing are 0, finite set A of
  projects
• Consider a contract that obliges the principal to
  pay investment cost z(a) and to give the agent
  h(w)
• Manager picks a to maximise

42
More model

• Assume h(w) gt 0 (limited liability).
• All parameters, functions, contract, eventual
  state are common knowledge, but only manager
  knows true a.
• As a PA problem principal chooses contract
  (a,h)to maximise his expected return s.t.
  incentive-compatibility (IC) and participation
  (IR) constraints
• .. the principal must get as least as much as her
  outside option as well.
• If the principal is risk neutral and the agent
  strictly risk-averse, offer agent a fixed payment
• Agent will be indifferent to choice of a and
  might as well choose the a that maximises the
  principals payoff (but might not)

43
Model, 3

• If the agent is risk-neutral and the principal
  strictly risk-averse, optimal risk-sharing would
  place all risk on the agent (up to the limited
  liability constraint). Define h by choosing r gt 0
  s.t.
• With this contract, the manager chooses a to
  maximise
• If the principal must offer a contract specified
  by (a, r), he solves

44
A result!

• For any probability vector p (p1,,pW), define
  the cdf
• A distribution p is a mean preserving spread of
  p if it gives the same expected value of R and
  satisfies one of the following three (equivalent)
  conditions

45
So?

• If p(a,.) is a mean-preserving spread of p(b,.)
  then for any r the r-contract that stipulates a
  pays the agent more than the r-contract that
  stipulates b.
• In other words, the agent (entrepreneur) prefers
  mean-preserving spreads.
• The principal, by contrast, wants to maximise his
  payoff at any solution to the problem the IR
  constraint should be binding, so we can describe
  the principal as maximising
• So e.g. a risk-averse principal would always
  prefer a project with higher expected value, but
  may be constrained by agent's risk-shifting
  preferences.

46
Debt overhang example

• Principle is the same as risk-shifting debt,
  equity holders own different parts of revenue
  stream, so differ about best course of action.
• Firm has no cash, debt 10K, interest rate 0
• Project investment 3K certain return 12K
• Do nothing
• Firm goes bankrupt, no-one gets anything
• Equity holders undertake project
• Project present value -3K12K 9K
• Value to bondholders 10K
• Value to equity holders -3K 2K -1K
• So the equity holders wont undertake the project
• This would be true even if the firm had the 3K
  already in hand the share holders would rather
  have all of it as a dividend than settle for the
  2K left over after the debt is paid
• Project only happens if debt holder put up the
  money but they typically have inferior
  information (will it be this kind of (very good)
  project or the (very bad) project from last
  weeks risk-shifting example?

47
Model of debt overhang

• As before, manager chooses effort a result is
  probability distribution p(w,a) and the manager
  gets U(h(w)) z(a) in state w. As before, h(w) gt
  0 (limited liability) and we assume the agent
  (manager) maximises expected utility.
• Assuming the parties are risk neutral, we can
  write the contract problem as
• To model overhang, let D be debt, and consider
  two probability distributions p without
  managerial effort (cost 0) and p with effort
  (cost z).
• p is assumed to first-order stochastically
  dominate p (in other words, for every
  non-decreasing h(w), the expectation of h under
  p is greater than the expectation of h under p)
  Sp(w)h(w) gt Sp(w)h(w)

48
Debt overhang result

• The project is worthwhile from the firms point
  of view
• (F)
• When will the shareholders want the project?
• (SH)
• Clearly, neither condition implies the other, so
  there are plenty of examples where the project
  wont be undertaken.
• We can even make examples where the bondholders
  might forgive the debt in order to get the
  entrepreneur to put forth effort (HIPC programme)

49
Further clarification of overhang
Firm gets R-z
Managergets h-z

• Let W Ws ? Wb - firm is
• solvent in Ws (R(w)-h(w) gt D)
• bankrupt in Wb (R(w)-h(w) lt D)
• Shareholders get
• Bond holders get
• So SH, BH, firm may have different preferences
• If D is reduced, attractiveness of stoch. dom.
  project rises

Share holdersget max0,R-h-D
Bond holdersget minD,R-h
50
Debt and Equity as incentive mechanisms

• Grossman-Hart (1982) The manager of a
  debt-laden firm may work hard to stave off
  bankruptcy (direct costs, reputation damage,
  etc.) like a non-pecuniary aspect to success
  vs. failure.
• Jensen (1986) firms with abundant CFE may
  fritter it away on untested projects. This
  agncy problem can be controlled by having lots of
  debt to obligate the CFE.
• Applicable to LBOs (unfriendly take-overs
  financed by debt (esp. zero-coupon or junk bonds)
• Usually, new owners sold assets to reduce debt
• Alternative new owners had better information
  and thus preferred debt.
• These articles did not address risk shifting and
  overhang
• Easterbrook (1984) uses agency to explain
  dividends precommittment (formal or habitual) to
  paying big dividends ties the hands of
  management incentive effects of debt without
  risk-shifting or overhang

51
Compensation 1

• Jensen and Meckling (1976) treatment of agency
  points directly to managerial compensation
  schemes.
• Murphy (1999) has a good survey
• Agency perspective very useful, but standard
  model results not very helpful
• E.g. Holmstroms (1979) informativeness principle
  tie compensation to all information about
  performance (like linkage principle in auction
  theory). But few compensation schemes come
  anywhere near this.
• At most, there may be accounting information
  (like contribution statements)
• Relative performance only implicit in stock
  options

52
Compensation 2 in practice

• Murphy (1999) points to a lot of academic papers
• Recent changes
• Base salary and bonuses based on accounting
  information always important
• Stock options have become very important
• Levels vary considerably across countries,
  sectors and degree of publicness
• Highest in financial services and in US (2x
  international average) lowest in regulated
  utilities
• Options and bonuses more important in US/UK than
  continental Europe
• Generic package
• Base salary consultant/survey based
• Options long-term, subject to policy change,
  tax advantage
• Bonuses based on EBIT, target (floor/ceiling)
• Retirement packages

53
Compensation 3 other points

• What the market will bear is the market for
  management competitive? How strong are the
  externalities?
• Pay and promotion
• The basic agency problem has a single agent
• Possible to extend to multiple competing agents
• More interesting multiple interacting agents
• Company problem to sort agents out (get them to
  reveal private information about ability, risk
  preference, opportunity cost of effort,
  intelligence, etc.) to provide effort incentives
  to get them to work to encourage human capital
  formation to select candidates for promotion
• One solution the tournament model pay CEO a
  lot and VPs nothing make them compete to become
  CEO. One paradox pay inversely proportional to
  performance or potential future competition.
  Model widespread in accounting, bank, legal
  firms.
• Pay and corporate governance
• Golden handshakes,
• Golden parachutes
• Poison pills
• The problem is much richer than the standard
  analysis and there remains much to be done.