Debt Covenants
Garleanu and Zwiebel
• Model looking at assignment of control rights
under asymmetric information.
• A motivating factor is that debt covenants are
commonly very “tight” giving control rights to
uninformed parties.
• They are frequently renegotiated, relaxing the
covenants in exchange for concessions
(interest rate increase, maturity reduction).
GZ
• Debt covenants serve to protect debt holders
from transfers of value to equity based on
decisions made by management.
– Some transfers are the result of efficient decisions.
• Coase: with costless renegotiation and symmetric
information the initial assignment of control
rights is irrelevant.
• With symmetric information and costly
renegotiation one would minimize the need to
renegotiate.
– Suggests “looser” covenants than are observed.
GZ – Model
• E – Entrepreneur: wealth constrained, risk neutral
• L – Lender: risk neutral
• I – Required Funding: raised in a competitive
lending market at t = 0 (L breaks even) with
discount rate 0
• Project: initial investment I at t = 0 return R with
certainty at t = 2
• At t = 1 a decision may be made requiring no cash
outlay that will alter the t = 2 return on the
project and result in a transfer to the
entrepreneur from the lender
GZ – Model
• Timeline:
t=0
initial
Investment,
contracting
state
revealed
G, B
renego.
transfer = t
t=1
t=2
action
A or NA
payoff
• State can be (G)ood or (B)ad with prob p, 1-p (at t=0) revealed prior
to renegotiation stage (non-contractible)
• Action results in “addition” to final payoff {y, -y} (state dependent)
and transfer (x) – type (E’s private information), L knows x ϵ [a, b]
with a cdf F() but may acquire information for a cost (c0 or c1)
GZ – Model
• Payoffs
State
A
NA
prob
G
R – D + y + x, D – x
R - D, D
p
B
R – D + x, D – y – x
R - D, D
1-p
• No action implies the original investment only
• Action gets an added value and a transfer (think
asset substitution)
• The added value is positive in state G and negative in
state B
• Note, E always wants A and L always want NA
GZ – Model
• Contract:
– E signs contract {r, D} = {decision maker, face value} where
r ϵ {E, L} and D = t=2 payment to L
• Renegotiation:
– A concession t is paid by one party to acquire the control
right from the other party
– In all situations L makes a take-it-or-leave-it offer to E and L
also pays the fixed cost cr of renegotiation
– There is only scope for renegotiation if r = L and G is
realized or if r = E and B is realized
GZ – Equilibrium
• Benchmark – symmetric information
– Prop 1: With symmetric information at t=0, r = L whenever
p < ½ and r = E whenever p > ½
– This arrangement minimizes renegotiation costs
• The main result compares this benchmark to the initial
contract that is part of a Pure Strategy Perfect Bayesian
Equilibrium in which L becomes informed at t = 1 (if
there is a scope for renegotiation)
– Parameter values for which this is true are when c1 is small
and c = c1 + cr < y
– Consider L becoming informed at t = 0 as well but this is
not optimal if c0 is high and not highlighted
GZ – Equilibrium
• Begin at t = 1:
– Define a state as (r, s), r ϵ {E, L} and s ϵ {G, B}
– Of the 4 states, (E, G) and (L, B) have no scope for
renegotiation since the efficient action is the
preferred action of the controlling stakeholder (L
remains uninformed)
– (E, B) and (L, G) have scope for renegotiation as
the preferred action by the party in control is
inefficient (therefore L becomes informed)
• Recall L has all the bargaining power.
GZ – Equilibrium
• Original payoff structure
• Post renegotiation payoffs
State
r=E
r=L
prob
G
R – D + y + x, D – x
R – D, D + y
p
B
R – D + x, D – x
R – D, D
1-p
– In state (L, G) L asks for x + y from E to take action A
rather than his preferred NA
– In state (E, B) L gives only x to E so E will take action
NA rather than A
GZ – Equilibrium
• Contract at t = 0
– In any PSPBE there is at most one contract
associated with each choice of r.
• If a contract {r, D} is accepted in equilibrium, then no
type of E would offer a contract {r, D’} with D’ > D
– Thus are at most two contracts, one with r = E and
one with r = L.
– Let SE c [a, b] denote the set of types E who offer
contract with r = E (show this is an upper interval
[x, b] of types – i.e. bad types)
GZ – Equilibrium
• Contract at t = 0
– If L is uninformed at t = 0 but will acquire information at t =
1 (if useful) then L must breakeven in expectation w/ EQ
contract
– If r=E then I = D - pE[x│x ϵ SE] - (1-p)(c + E[x│x ϵ SE]) 1
– If r=L then I = D + p(y-c) 2
• When D satisfies 1 or 2 we can find E’s final payoff
– UEr=E = R - I - E[x│x ϵ SE] - (1-p)c + p(x+y) +(1-p)x 3
– UEr=L = R - I + py - pc 4
– Note that E’s payoff increases in x when r = E and is
independent of x when r = L. Thus high x types want r = E.
– There is a cutoff x” where types below pool with r = L and
above pool with r = E. SE is of the form [x”, b].
GZ – Equilibrium
• To characterize the cutoff x” define G(u)
– G(u) ≡ E[x│x > u] – u
– Expresses the difference between average type in
a pool of [x”, b] and the lowest type in the pool
– The cost of adverse selection for a lower type to
pool with higher types
– Since a type x” must be indifferent between r = L
and r = E, equating eqns 3 and 4 tells us that x” is
given by G(x”) = (2p-1)c
GZ – Equilibrium
• Proposition 2 – main result
– Assume L pays c1 and learns x at t = 1 if there is scope
for renegotiation. Then a PSPBE always exists and
takes the form:
(i) if G(x) > (2p-1)c then all types offer r = L with D = I
– p(y-c)
(ii) if G(x) < (2p-1)c then all types offer r = E with
D = I + E(x) + (1-p)c
(iii) if G(x”) = (2p-1)c for some x” in [a, b] then high x
types offer r = E and low x types offer r = L where
DE = I + E[x│x > x”] + (1-p)c and DL = I – p(y-c)
GZ – Equilibrium
• Proposition 2 shows that the nature of the equilibrium
trades off cost of information acquisition and
renegotiation against the adverse selection cost for
different assignments of control.
• When p < ½, all types choose r = L, corresponds with
the constrained efficient symmetric information
outcome.
• When p > ½, the constrained efficient outcome implies
all types should choose r = E. Asymmetric information
however implies covenants are more restrictive, some
types choose r = L.
GZ – Empirical Implications
• We should see strict covenants and asymmetric
renegotiation. (They stretch this.)
• Higher renegotiation costs imply less restrictive covenants.
– Private vs Public debt or complex vs simple capital structure.
• Covenants are more restrictive when there is more
asymmetric information or complexity.
– Small firms, high growth, high leverage
• More fungible cash flow implies more uncertainty about
transfers and more restrictive covenants.
• Asymmetric information about investment (y) likely related
to asymmetric information about transfers (x) and thus
they suggest similar predictions.
Lemmon and Zender
• Examines a similar question but in the context of
a capital structure question.
• The model takes the Myers/Majluf environment
and adds an additional period.
– The asymmetric information remains and a liquidation
decision must be made.
– The initial use of debt financing, motivated by the
asymmetric information implies a cost in the
liquidation decision.
– Leverage choice and debt covenants help control this
cost.
LZ – Model
• At t=0 managers of different types seek funding, I, for
a project.
– Type t ϵ {G, B} with G > B and prob(G) = θ.
• At t=1 a signal is publicly observed that identifies the
strength of the market.
– Signal w ϵ {w1, w2} with w1 > w2 and prob(w1) = p
• w1 labeled a strong market and w2 a weak market
– Based on signal, firm may be liquidated (Q) or continued.
• At t=2 final payoffs
– Success (H, with prob = tw) or failure (L) if project is
continued.
LZ – Model
• Financing
– Entrepreneur is assumed to sell debt and equity at t=0
to raise I.
• Proportion of external equity is α
• Entrepreneur retains 1 – α of the equity
• External debt has face value F (not allowed to be contingent
upon signal) and an allocation of control over liquidation
(covenant dependent upon the signal, w’, where if w < w’
control is allocated to the lender)
• Financing may induce an agency problem with respect to the
time 1 decision (that may be controlled by the covenant).
• Examine renegotiation of the covenant
LZ – Model
• Incentives (t=1):
wMG(F)
w*
G
wMB(F)
w*B
– Note w’ will always be in the interval [w*G, w*B]
• w’ = w*G unrestrictive covenant and w’ = w*B a
restrictive covenant
– Assume that w1 > w*B > w2 > w*G
– This implies only the incentives of the bad
manager are an issue
LZ – Model
• Incentives (t=0):
– At time zero, the good entrepreneur prefers debt to
equity due to the adverse selection.
• Limited by the behavior induced for the bad entrepreneur
– A version of Myer’s idea of debt capacity comes out of
the extension to the model.
• Once F = Q there is no further motivation for the
entrepreneur to issue debt rather than equity, external
equity and added debt have the same informational
sensitivity. A “soft” capacity.
• We will see later that renegotiation provides a hard
constraint on debt at this level.
LZ – Equilibrium
• No Renegotiation:
– When we disallow renegotiation, or assume that it
is too costly to be effective, the optimal financing
scheme includes a “low” level of debt and an
unrestrictive covenant.
– This implies a high cost for adverse selection.
– The use of “high” debt and a restrictive covenant
(to mitigate the consequent incentive problem) is
not optimal when there is no ability to renegotiate
the covenant.
LZ – Equilibrium
• Costless Renegotiation:
– Two types of covenants are potentially useful.
– We first show that a (range of) high debt level(s) and a restrictive
covenant is equivalent ex ante to the use of low debt and an
unrestrictive covenant.
• Relies on renegotiation (relaxation) of the restrictive covenant in a strong
market for a higher debt burden (higher interest rate).
• Ex post, renegotiation is more costly the stronger is the market/economy as it
takes more concessions to separate from the bad firm. Ex ante, it is priced
into the initial debt contract.
• Debt capacity becomes a hard constraint – at F = Q good firms cannot separate
ex post.
– We then show that high debt and an unrestrictive covenant will be
equivalent only if the lender takes all the rents in the renegotiation
which is not natural in this circumstance.
• It matters who sits at the bargaining table.
LZ – Equilibrium
• Costly Renegotiation – main results
– When renegotiation is costly there is scope for a high
debt/restrictive covenant solution as well as a low
debt/unrestrictive covenant solution.
– Firms are more likely to use high debt and a restrictive
covenant:
•
•
•
•
The greater the adverse selection at t=0 (lower B or lower θ).
The stronger is the economic outlook (higher is p).
The larger is the firm’s debt capacity.
When the expected concessions in renegotiation are low
(the separation constrain is easier to satisfy).
Conclusions
• Both papers provide arguments for the use of
restrictive covenants and “asymmetric”
renegotiation.
• Both papers provide that greater asymmetric
information is associated with more restrictive
covenants.
• LZ are also able to tie covenants with debt levels.
• LZ indicates a possible new approach to testing
Capital Structure.
Chava and Roberts (2008)
• Does financing influence investment?
• In a perfect world, no.
– What about the world we live in?
• There are suggestions in the nature of contracts that it
does.
• Chava and Roberts highlights bond covenants.
– Covenants appear to be there to control conflicts between stock
holders and debt holders.
– If this is true, decision making will be affected when control
transfers between these parties.
– When covenants are violated lenders have leverage to exert
their influence on decision making.
– Is this a channel through which financing affects investment?
Covenant Tightness and Violations
Table 3
Fraction of observations is fraction of firm quarters in violation.
Tightness measured as number of firm specific standard deviations of the relevant
accounting variable.
Time to first violation measured normalizing all loans to unit time.
Chava/Roberts Figure 1