Remember that AE and AD are the gains realized by new stock or
bondholders at t = + 1 when the firm’s true value is revealed. They depend on
a, b, S and the decision to issue and invest. If there is an equilibrium in which
equity is issued, there is a price Pk at which investors can rationally expect
A E = 0. For debt, the equilibrium firm value is P; and investors expect
AD = 0. Given a, b and S, A E and AD hare the same sign, but IA El > [A DI.
However, there is no equilibrium price PL at which the firm can issue stock.
It prefers stock to debt only if P; is high enough that A E < AD. This occurs
only if AE < 0, implying a sure capital loss for new stockholders. Therefore,
there can be no price P; at which (1) the firm is willing to issue stock rather
than debt and (2) investors are willing to buy.
To put it another way: suppose the firm announced at t = – 1 that it would
issue debt if it issued any security. It could not change its mind and issue
equity at t = 0, because investors would assume this meant A E < 0 and refuse
“See, for example, Galai and Masulis (1976). The option pricing framework of course rests on
more restrictive assumptions than those used so far in this paper. We return to these assumptions
below.
lXThis is Rendleman’s conclusion (1980). As noted above, he does not work out a full
equilibrium solution.
XC. Myers und N.S. Majiuf; Investment undfinuncingpolicy with drfferential inform&on 209
to buy. On the other hand, a firm which announced a policy of equity
financing at t = – 1 would be forced to change its mind, and to issue debt at
t = 0 if it issued at all. Equity would be issued at t = 0 only if debt were ruled
out at t = – 1; yet we showed above that precommitting to equity financing is
always inferior to precommitting to debt.
Thus, our model may explain why many firms seem to prefer internal
financing to financing by security issues and, when they do issue, why they
seem to prefer bonds to stock. This has been interpreted as managerial
capitalism – an attempt by managers to avoid the discipline of capital markets
and to cut the ties that bind managers’ to stockholders’ interests. In our model,
this behavior is in the stockholders’ interest.
3.4. Equity issues in asymmetric information models
The chief difficulty with this analysis of the debt-equity choice is that we
end up leaving,no room at all for stock issues. We could of course recreate a
role for them by introducing agency or bankruptcy costs of debt, as discussed
in, for example, Jensen and Meckling (1976), Myers (1977), and Smith and
Warner (1979). But it is also possible to rationalize equity issues in models
based on information asymmetries alone.
Our proof that debt dominates equity uses the standard option-pricing
assumption that percentage changes in value are lognormally distributed with a
constant variance rate known by everyone. However, suppose there is a large
information asymmetry about the (future) variance rate. If investors under-
estimate the variance rate, the firm will be tempted to issue debt, but if they
overestimate it, the firm will be tempted to issue equity, other things equal.
Thus, a decision to issue equity may not signal a sure capital loss for new
stockholders, but simply that the firm is safer than prospective bondholders
think. Thus equity issues are not completely ruled out in equilibrium.
Giammarino and Neave (1982) set up a model in which the managers and
investors share the same information about eoerything except risk. In this case,
equity issues dominate debt issues, because the only time managers want to
issue debt is when they know the firm is riskier than investors think. Investors,
realizing this, refuse to buy. Only equity, or perhaps a convertible security, is
issued in equilibrium.
Firms actually seem to favor debt over equity issues, not the reverse. We
believe asymmetric information about firm value is a stronger determinant of
financing behavior than asymmetric information about risk, and we will so
assume in subsequent comments, although future empirical research could of
course prove us wrong. On the theoretical side, an obvious next step is to
analyze the debt-equity choice in a version of our model which explicitly
210 S.C. Myers and N.S. Majiuj Investment andfinancingpolicy with dtflerential information
allows information asymmetry on the two dimensions of firm value and firm
variance.‘9
4. Assumptions about management’s objectives
We have shown that ample financial slack allows the firm to avoid external
financing and to disentangle investment decisions from conflicts of interest
between old stockholders and new investors. However, this result depends on
management’s acting in the interest of passive stockholders. We will now
consider how rational stockholders react to the firm’s investment decision. We
show that, in frictionless capital markets, their reaction does not depend on
whether the investment is financed with internal or external funds.20
4.1. The irrelevance of financing
Take the simplest case, in which the firm can only issue stock. When the firm
has inadequate slack (S < I), we showed that the firm may pass up valuable
investment opportunities. This loss would be avoided if old stockholders could
be compelled to buy and hold the new issue - in other words, to accept the
new asset in their own portfolios. In general, this will not be their optimal
portfolio strategy, however, so new shareholders enter, creating the conflict.
Now suppose the firm has ample slack (S = I). Old stockholders arrive at
t = 0 with shares representing a portfolio of three items: an asset in place, a
growth opportunity and cash. If the growth opportunity is taken, the cash
vanishes, and the portfolio changes to two assets in place. The old stockholders
‘buy’ all of the new asset via the firm’s internal financing. However, there is
nothing to force them to hold it. The same portfolio motives that would
prevent them from buying all of a new issue should prompt them to sell part of
their shares if the firm uses its cash to buy a risky real asset.
There is no deadweight loss so long as the firm buys this asset whenever it
has positive NPV (b > 0). However, suppose managers start to worry about the
price old shareholders trade at when they rebalance their portfolios after an
internally-financed investment is made. Table 2 sets out equilibrium conditions
for this case. The left-hand block (case I) shows old shareholders’ payoffs if the
firm has no slack. We assume old shareholders could buy all of the new issue.
Therefore, we earmark C = Z dollars of cash and other securities and take it as
potentially available for investment. However, their optimal portfolio calls for
19Note that the general version of our model, as described in eqs. (1) and (2). allows asymmetric
information about any feature of the joint distribution (k, 3). But addressing the choice among
financing instruments requires more specific assumptions.
2o We thank George Constantinides for suggesting this possibility.
S.C. Myers and N.S. Majiuf. Investmenr andJinancingpolicy with differential information 211
investing al in the new issue. The resulting equilibrium conditions are slight
generalizations of those given in section 3 above (we previously took a = 0).
In the right-hand block (case II), the firm holds the same amount of cash on
behalf of old shareholders. If the firm invests this cash, they recoup part of it
by selling shares to raise (1 – cr)Z. Their fractional ownership thus ends up as
(Z”’ – (1 – cy)Z)/P”. Note that P”, the market price of the firm conditional
on investment, includes the investment I. It’s convenient to substitute PAi, =
P,’ – I.
At equilibrium, PALt = x(M”) + B(M”), where M” indicates the states in
which investment by the firm is in the old shareholders’ interest given the price
P,.$ facing them when they sell.
The equilibrium conditions for the two cases shown in table 2 are identical.
The firm’s investment decision is independent of whether cash starts out in the
shareholders’ bank accounts or the firm’s The firm passes up good investment
opportunities in the same states, so the ex ante loss L is the same for the two
cases 21 So
P’ = i;;,.
are the market prices conditional on the decision to invest:
The choice between debt and equity financing should not matter either.
Suppose the starting position is case I in table 2. The firm borrows C = Z
dollars from its stockholders. That transforms case I into II, if the debt is
default-risk-free. The final equilibrium investment decision and stock price are
unaffected.
If the debt carries default risk, old shareholders are exposed to the firm’s
business risk through their new debt securities as well as their stock. Therefore,
when the firm invests, they will raise (1 – a)Z by selling a mixture of debt and
equity securities – the same fraction of their holdings of each. However, the
same final equilibrium is reached again.
If the risky debt is sold to outsiders, old shareholders would buy part of the
debt issue, and sell some of their shares. However, as long as capital markets
*‘If old shareholders are willing to hold all of any new investment – i.e., if a = 1 in table 2’s
expressions – the firm always invests if b > 0. This is, of course, the ex ante optimal policy; the
problem is enforcing it. Old shareholders could enforce it by purchasing 100 percent of any new
issue (case I) or by not selling any of their shares (case II).
However, note that the incentive for old shareholders to buy all of a new issue is strongest if
they act in concert. Management looks at the overall 0~. An investor who holds, say, one percent of
the firm’s stock, and who acts alone, buying one percent of the new issue, will reap only one
percent of his action’s rewards. If arranging a group action is costly, then individual investors’
incentives will not make a= 1 overall.
In case II, Q = 1 if old shareholders do nof trade when the firm invests. Financial slack helps by
making sure that old shareholders buy all of the new project, at least temporarily. Trading costs
then limit the extent of selling. If their portfolios are ‘sticky’, the conflict of interest between old
and new shareholders is reduced. However, any investor who sells out will not face the full cost of
his actions, since management’s decision depends on old stockholders’ overall participation in the
new project.
212 XC. M_tjers and N.S. Majluf, Investment andfinancingpoliq with d@erentiul informution
Table 2
Equilibrium conditions for the issue-invest decision with and without financial slack.
(1)
Firm has no slack (S = 0), Firm has S = I. However, if
and must issue the amount I it invests, shareholders sell a
in order to invest. Share- portion of their holdings to
holders have cash C = I and recover (1 – a)f in cash.
invest al in new issue. Managers and investors have
Managers know a. the value the same information as in
of assets-in-place, ‘and h, the
net present value of the in-
vestment opportunity, but in-
vestors do not.
Value to old
shareholders
( vo’d )
No issue: C+a=l+u
case I.
s+u=r+u
Issue:
(1 -a)[+
P”-(1 -a)1
P”
(I+u+h)
(1 -a)I+ z(l+a+h)
P” +a/
or(l-a)[+- p,, il (r+u+h)
net
P” +a[
I+u<(l-cr)l+ z(l+u+h) I+u<(l-a)l+=
P” +I
(I+u+h)
“.3
At equilibrium:” P’=A(M’)+B(M’)
aThe equilibrium values of the firm (P’ in case I and P&, in case II) are identical. Thus, the
investment decision can be independent of whether financing comes from internal funds or a stock
issue. Here we assume that investors rebalance their portfolios when the firm reveals its investment
decision.
are frictionless, and all traders understand what is going on, the final result is
the same.
We thus obtain an (MM) proposition of financial irrelevance, where all the
action comes from the firm’s decision to invest. If this tack is taken, our
model’s empirical implications change. We could not explain firms’ demands
for slack, their apparent preference for internal financing, or for debt over
equity issues. A fall in stock price on announcement of a stock issue would be
explained as an information effect. That is, the issue would not matter in itself,
but only as a signal of the decision to invest.
However, before we turn to the empirical evidence, we will devote a few
more words to the competing descriptions of management objectives and
shareholder responses when managers know more than shareholders do.
4.2. Ex ante optimal policies
Old shareholders are better off ex ante, and on average ex post, if manage-
ment takes all positive-NPV projects. Perhaps compensation packages have
SC. Myers und N.S. Majluf, Investment andfinancingpolicy with differentiul information 213
features that prompt managers to follow this rule. Social conventions or
corporate cultures that encourage managers to maximize ‘long-run value’ may
have the same effect. Also, following the rule may be in managers’ self-interest:
a manager who does not allow conflicts between old and new shareholders’
interests to block positive-NPV projects could demand a higher salary ex ante
than one who does.
However, management must take some responsibility for financing. Con-
sider the extreme instruction: ‘Take all positive-NPV projects and issue
securities at any price.’
The ‘wrong’ price for a security issue does not affect firm value. It just
transfers value from some securityholders to others. Nevertheless, the instruc-
tion is not credible. Public stockholders would not support it, because it would
leave them unprotected against sweet deals given to insiders or their friends.22
Of course this sort of sweet deal is illegal. An outside investor hurt by one of
them could sue, and probably win if the r&pricing were obvious and the
motive clear. The law requires a manager to worry about the terms of
financing; we think it encourages the manager to look at financing from the
viewpoint of the passive investor.
Consider the altered instructions: ‘Take all positive-NPV projects, and issue
securities at a fair price conditional on market information only.’ In other
words, managers should use their special information about investments, but
ignore it when it comes to financing. 23 However, these instructions are still not
fully credible, not only because of the mental discipline required, but also
because managers’ personal interests are likely to be more closely aligned with
old stockholders’ interests than with new stockholders’.
Consider a manager who is also a stockholder. If he always buys and holds a
pro rata share of any new issue, and maximizes the intrinsic value of his
holdings, then his interests will be aligned with all securityholders’, and he will
maximize the intrinsic value of the firm.
However, most managers would not want to buy a pro rata share of every
new issue, even if the issue is fairly priced from their point of view. The reasons
why can be traced to the same portfolio considerations which prevent old
stockholders from buying all of every new issue – loss of diversification and, in
extreme cases, limits to personal wealth. If the manager does not buy all of
22ECxi~ting securityholders could protect against this ripoff by taking a pro rata share of each new
issue. But this would be cumbersome at best. It would also invite a different kind of ripoff, in
which outside securityholders buy an overpriced issue while insiders and their friends sell or sell
short.
“This suggests the idea that managers could avoid conflicts between old and new shareholders
by conceuling the firm’s investment decision. Take case II in table 2, where the firm has ample
slack. Suppose its investment decision is not revealed until t = + 1. Then the firm’s actions prompt
no trading at r = 0, and good investment opportunities are not bypassed. In case I. on the other
hand, the investment decision cannot be concealed because a stock issue necessarily comes first.
214 XC. Myers and N.S. Majlu/, Investment andfinancingpolicy with differential information
every new issue then his interests as a shareholder are ‘those of an (informed)
old shareholder.
There is still another complication: a manager-shareholder who has inside
information will be tempted to trade on it. If the outside market is (semi-strong
form) efficient, the manager will want to sell half the time and buy the other
half. In particular, he will want to abstain from half of new issues, and buy
more than a pro rata share of the others. He will also want to buy or sell if the
firm does not issue. The potential trading profit will depend on the issue-invest
decision, although apparently not in any tractable way. We doubt the managers’
interests will be aligned with any outside investor’s if the managers are given
free rein to trade on personal account.
4.3. Empirical evidence
It is easy to see why managers should take all positive-NPV projects, but
hard to build a completely convincing theory explaining why they would
always do so. We think it more likely that managers having superior informa-
tion act in old stockholders’ interest. We also think that existing empirical
evidence supports our view.
If management acts in old shareholders’ interests, our model predicts that
the decision to issue and invest causes stock price to fall. If management took
all and only positive-NPV projects, even when issuing and investing reduces
the intrinsic value of the ‘old’ shares, the same decision would either increase
stock price or leave it unchanged. The decision to invest would reveal the
existence of an attractive project (i.e., b > 0). This is good news, unless
investors knew for sure that the firm would have that investment opportunity.
It cannot be bad news in any case.
Recent papers by Korwar (1982), Asquith and Mullins (1983) and Dann and
Mikkelson (1984) show significant negative average price impacts when a new
stock issue is announced. ‘Information effects’ are an obvious explanation.
However, as far as we know, ours is the only complete model explaining how
such an information effect could occur in a rational expectations equilibrium.
Of course our model predicts that stock prices will always fall when
investors learn of a new stock issue. But the model holds everything but the
issue-investment decision constant. In particular, it ignores the flow to inves-
tors of other information about the firm’s prospects. This flow creates a
random error in any measurement of how a stock price changes in response to
a specific event.
If our interpretation of these results is accepted, we can set aside, at least for
the present discussion, models assuming managers simply ‘accept all positive-
NPV projects’.
If managers act in old shareholders’ interests, do they assume those share-
holders are passive or active? Do they just maximize the existing shares’ value,
S.C. Myers and N.S. Majluf, Inoesrment andjnuncrngpoky wrth differentiul informution 215
or do they work out how rational shareholders’ portfolio choices depend on
their decisions?
These questions can also be answered empirically. If managers assume active
shareholders, then only the investment decision matters. Good investments are
foregone even when the firm has plenty of cash to pay for them. However, if
managers assume passive stockholders, then financing matters and firms will
adapt their financing policy to mitigate the loss in value from foregone
investment opportunities. For example, managers will try to build up financial
slack on order to avoid situations in which a security issue is required to
finance a valuable investment opportunity. If information asymmetries relate
primarily to firm value, rather than risk, managers will favor debt over equity
financing if external capital is required.
In our framework, the ‘passive investor’ assumption gives a variety of
interesting hypotheses about corporate financing. That is why we use the
assumption for most of this paper’s formal analysis.
We noted that Dann and Mikkelson (1984) found a significant negative
average price impact when stock issues are announced. They also looked at a
sample of debt issues, and found no significant price impact. Our model may
be able to explain this difference.24
The ‘passive investor’ assumption implies that stock price falls when stock is
issued. However, stock price should not fall if default-risk-free debt is issued,
because the ability to issue risk-free debt is equivalent to having ample
financial slack, and having ample slack insures that the firm will take all
positive-NPV projects. Thus, in our model the only information conveyed by
the decision to issue risk-free debt and invest is that the firm has a positive-NPV
project. This causes a positive price change unless the project’s existence was
known beforehand.
Under the ‘active investor’ assumption, the decision to invest would be bad
news, and the choice of debt over equity financing would not make the news
any better. Choosing this assumption would give us no way to explain Dann
and Mikkelson’s results.
Of course, the debt issues examined by Dann and Mikkelson were not
literally default-risk-free. But if the probability of default on these issues was
small, their negative ‘information effect’ should likewise be sma11.25
5. Extensions and implications
Having explained our model formally, and having discussed its assumptions
and some of its empirical implications, we can now turn to a few extensions,
24Miller and Rock’s model would predict the same negative stock price impact regardless of the
type of security issued.
*‘You would expect that the riskier the instrument issued, the greater the issues’ impact on the
market value of the firm. However, we have not been able to prove that this positive relationship is
always monotonic.
216 XC. Myers und N.S. MujluJ Investment andfinuncingpolicy with dlyerential information
qualifications, and further observations. 26 We specifically address two ques-
tions:
(1) What happens when the information asymmetry is temporary and what
happens when it is permanent but the firm has no immediate need for funds,
except to build up slack?
(2) What does our model say about mergers?
Discussing these questions leads us to other issues, for example, the impli-
cations of managers’ superior information for dividend policy.
5.1. An easy way out
There is of course an easy way out of the problems described in this
paper – an easy way to avoid any loss of market value: just issue stock at
t = – 1, when managers and the market are assumed to share the same
information. That is one lesson of our model. If managers know more than the
market does, firms should avoid situations in which valuable investment
projects have to be financed by stock issues. Having slack solves the problem,
and one way to get slack is to issue stock when there is no asymmetric
information.
That is not an easy way out, however, if the information asymmetry is
permanent. Suppose managers are always one period ahead of the market. At
t = – 1, for example, managers would know Kand B, but investors would not.
Investors at t = ‘- 1 would see x and B as random variables. At t = 0, they
would find out the means xand 3 (and the underlying distributions of 2 and
B) but by that time managers would know the realizations a and b.
Assume the firm has insufficient slack to undertake the project, and also, to
keep things simple, that the amount of slack is fixed unless equity is issued to
increase it and the investment required to undertake the project is known.
Consider the decision to issue E = I – S dollars of stock at t = – 1. If the firm
does not issue. its true value at t = – 1 is
V,,,(noissue) =A+B +S – L,
where L is the ex ante loss in firm value attributable to insufficient slack. That
is, L reflects the probability the firm will choose to pass up a positive-NPV
investment at t = 0, and the loss in value if it does. Of course, investors do not
know how big L is, because they do not know the distributions of 2 and 8.
However, they do know that L goes to zero if the firm issues stock at t = – 1
26‘Our model’ includes the assumption that managers act in old stockholders’ interests, and that
those stockholders are passive in the sense discussed above.
XC. Myers and N.S. MajluJ Investment andfinancingpolicy with diferential information 211
in order to raise the additional cash E = Z – S needed to assure investment at
t = 0.
This brings us back to the same problem we started with in section 2. We
have an ‘asset-in-place’ worth x+ B + S – L and an ‘investment opportunity’
worth L. Managers know these values but investors have only probability
distributions. Thus, the firm’s decision to issue and the price investors are
willing to pay are governed by eqs. (1) and (2). Managers may or may not issue
stock at t = – 1: it depends on the price they can issue it for. If investors are
too pessimistic, relative to what managers know, managers may accept the
ex ante loss L and take a chance that the firm will be able to issue and invest
at t = 0 if the NPV of its investment opportunity turns out positive.
We will not here pursue analysis of the optimal issue strategy in this
dynamic setting. However, we have shown that the problems addressed in this
paper do not go away when the firm has no immediate real investment
opportunity. Given asymmetric information, a firm with valuable future real
investment opportunities is better ofI with slack than without it. Moreover, it
should build up slack through retention rather than stock issues. This is
consistent with actual retention policies of most public firms, which limit
dividends so that they will rarely have to go to the market for fresh equity.
Thus we add one item in favor of the list of possible arguments for dividend
payouts low enough to avoid reliance on external equity financing. On the
other hand, dividends would alleviate the problems posed in this paper if they
help signal the true value of 2, thus reducing uA. However, this is not
necessarily an argument for high average payout; it merely supports payout
policies with a high correlation of changes in dividends and changes in the
value of assets in place. This could explain why dividend payments respond to
changes in earnings, not market value, if book earnings primarily reflect the
performance of assets-in-place.
At this point, we revert to our original three-date model, in which asymmet-
ric information is important only at t = 0.
5.2. Mergers
Our model’s main message is this: given asymmetric information, a firm with
insufficient financial slack may not undertake all valuable investment oppor-
tunities. Thus, a firm that has too little slack increases its value by acquiring
more.
One way to do this is by merger. In our model, a merger always creates value
when one firm’s surplus slack fully covers the other’s deficiency.27 Of course
271f the merged firms’ total slack does not fully cover their investment requirements, the merger
may or may not increase value. See Majluf (1978, pp. 239-256).
218 XC. Myers and N.S. Majluf, Investment andfinancingpolicy with differential information
this gain is only one of dozens of possible merger motives. But we have nothing
to say here about other benefits or costs, so we will assume them away here.
It turns out that the same conditions that create a potential gain from
transferring surplus slack between merger partners will also complicate the
merger negotiations, and in some cases rule out any possibility of their
successful completion. Consider a firm with an existing business, a good
investment opportunity, but insufficient slack to pay for it. It seeks a merger
with a cash-rich firm. The would-be buyer only knows the distributions of A
and & not the true values a and b.
Let Q’ be the proposed merger price. That is, if the merger offer is accepted,
the shareholders of the cash-poor firm receive Q’ in cash or shares.28 If the
offer is turned down, that firm’s shareholders forego the investment and are left
with S + a. Thus, given a and b, the offer will be accepted if Q’ 2 S + a.
However, the cash-rich firm will only offer Q’ = S + A< N ‘) + B( N ‘), where
A( N’) and B(N) are the expectations of A and b conditional on observing
that the cash-poor firm is willing to go through with the deal.
Under these assumptions, the merger would never occur. The cash-poor firm
can always do better by issuing stock directly to investors, because P’ always
exceeds Q ‘. 29
In our model, the decision to sell shares always carries negative information,
regardless of whether the shares are sold to investors generally or to a specific
acquiring firm. The buyers or buyers discount the shares so that cost equals
expected payoff. If the firm issues E = Z – S, old shareholders retain a stake,
but if their firm is sold they are completely disengaged from it. The decision to
sell all of the firm via merger, rather than issue the fraction E/( P’ + E), drives
down market price below P’, because the firm has chosen to sell more stock
than absolutely necessary to cover the investment I. [We assume that (1) the
acquiring firm’s slack exceeds the selling firm’s deficiency (I – S), (2) the
acquiring firm has other assets, and (3) everyone knows what these assets are
worth.]
“We assume for simplicity that the true value of any shares used to finance the merger is
independent of a and h. A more elaborate analysis is needed if they are not independent. A
further complication is introduced if (1) the merger is financed by shares and (2) the buying firm’s
management has superior information about what those shares are worth.
29A proof follows. Define a * ( N ‘) as the breakeven value of a. the value at which the cash-Door
firm is-just indifferent to being acquired at the equilibrium price Q’. Note that Q’ = a*( ,V’)*+ S.
Refer again to the requirement for the firm to issue stock (1’). E/P’( S + a) 2 E + b. I/ p’ were
equal to Q’, the firm would issue and invest at a*( N’) for any b > 0. That is, if P’ = Q’= s +
a*( N’), p r
;(s+“)=s+a:(N3 (S+a*(N’))=E 0. Thus,
x(M’)+B(M’)>x(N’)+B(N’) and P’>Q’.
S.C. Myers und N.S. Majuf. Inoestment andfinuncingpoliqy with differential rnformution 219
Negotiated mergers thus seem to be ruled out (in this simple case) regardless
of financing, because the cash-poor firm can always do better by issuing stock.
How can mergers be explained under the premises of this paper?
There are two possible explanations. First, there may be partial or total
disclosure of internal information during negotiation.30 Second, the merger
may go through if the buyer rather than the seller takes the initiative. In our
model, firms with plenty of slack should seek out acquisition targets which
have good investment opportunities and limited slack, and about which
investors have limited information. Such firms sell at a discount from their
average potential value AS B + S. 3i A tender offer made directly to the
slack-poor firm’s shareholders, at a price above A+B +S – L, but below
x+B +S, makes both the bidder and the target’s shareholders better off
ex ante, although neither buyer nor sellers know the true value a + b + S. A
cash tender offer conveys no bad news about a + b + S, so long as the target’s
management are not accomplices. Perhaps this explains why most mergers are
initiated by buyers. A firm that actively seeks to be bought out may end up a
wallflower. The more actively management seeks to sell, the less an outsider
will assume their firm is worth.
6. Conclusion
We have presented a model of the issue-invest decision when the firm’s
managers have superior information. We can sum up by reviewing some of the
model’s most interesting properties:
(1) It is generally better to issue safe securities than risky ones. Firms
should go to bond markets for external capital, but raise equity by retention if
possible. That is, external financing using debt is better than financing by
equity.
(2) Firms whose investment opportunities outstrip operating cash flows,
and which have used up their ability to issue low-risk debt, may forego good
investments rather than issue risky securities to finance them. This is done in
the existing stockholders’ interest. However, stockholders are better off
ex ante – i.e., on average – when the firm carries sufficient financial slack to
undertake good investment opportunities as they arise.
The ex ante loss in value increases with the size of the required equity issue.
Thus, increasing the required investment or reducing slack available for this
investment also increases the ex ante loss. In addition, numerical simulations
‘“The cash-poor firm would prefer to negotiate with a firm that is not a competitor. A
competitor might back out of the negotiations and take advantage of information acquired in
them. This hazard is less in a ‘conglomerate’ merger.
31 We assume the target firm has not yet declared its issue-invest decision.
220 S.C. Myers and N.S. Majlu/. Investment andfinancingpoiicy with di$erential information
indicate the loss decreases when the market’s uncertainty about the value of
assets in place is reduced, or when the investment opportunity’s expected NPV
is increased.
(3) Firms can build up financial slack by restricting dividends when
investment requirements are modest. The cash saved is held as marketable
securities or reserve borrowing power.
The other way to build slack is by issuing stock in periods when managers’
information advantage is small; firms with insufficient slack to cover possible
future investment opportunities would issue in periods where managers have
no information advantage. However, we have not derived a generally optimal
dynamic issue strategy.
(4) The firm should not pay a dividend if it has to recoup the cash by
selling stock or some other risky security. Of course dividends could help
convey managers’ superior information to the market. Our model suggests a
policy under which changes in dividends are highly correlated with managers’
estimate of the value of assets in place.32
(5) When managers have superior information, and stock is issued to
finance investment, stock price will fall, other things equal. This action is
nevertheless in the (existing) stockholders’ interest. If the firm issues safe
(default-risk-free) debt to finance investment, stock price will not fall.
(6) A merger of a slack-rich and slack-poor firm increases the firm’s
combined value. However, negotiating such mergers will be hopeless unless the
slack-poor firms’ managers can convey their special information to the pro-
spective buyers. If this information cannot be conveyed (and verified),
slack-poor firms will be bought out by tender offers made directly to their
shareholders.
Of course,’ the six items stated just above depend on the specific assumptions
of our model and may not follow in other contexts. We have only explored one
of many possible stories about corporate finance. A full description of corpo-
rate financing and investment behavior will no doubt require telling several
stories at once.
References
Akerlof, GA., 1970, The market for ‘lemons’: Quality and the market mechanism, Quarterly
Journal of Economics 84.488%500.
Asquith, P. and D.W. Mullins, 1983, Equity issues and stock price dilution, Working paper, May
(Harvard Business School, Cambridge, MA).
Bhattacharya, S., 1979, Imperfect information, dividend policy and the ‘bird in the hand fallacy’,
Bell Journal of Economics 10, 259-270.
Bhattacharya, S. and J.R. Ritter, 1983, Innovation and communication: Signalling with partial
disclosure, Review of Economic Studies 50, 331-346.
32 However, there is no mechanism in our model to insure that such a policy would be faithfully
followed.
SC. Myers und N.S. MajluJ Investment andjnancingpolicy with differential informution 221
Campbell, T.S., 1979, Optimal investment financing decisions and the value of confidentiality,
Journal of Financial and Quantitative Analysis 14, 913-924.
Campbell, T.S. and W.A. Kracaw, 1980, Information production, market signalling, and the theory
of financial intermediation, Journal of Finance 35, 863-882.
Dann, L.Y. and W.H. Mikkelson, 1984, Convertible debt issuance, capital structure change and
financing-related information: Some new evidence, Journal of Financial Economics, this issue.
Downes, D.H. and R. Heinkel, 1982, Signalling and the valuation of unseasoned new issues,
Journal of Finance 37, l-10.
Galai, D. and R. Masulis, 1976, The option pricing model and the risk factor of stock, Journal of
Financial Economics 3, 53-82.
Giammarino, R.M. and E.H. Neave, 1982, The failure of financial contracts and the relevance of
financial policy, Working paper No. 82-3 (Queen’s University, Kingston, Ont.).
Grossman, S.J., 1981, An introduction to the theory of rational expectations under asymmetric
information, Review of Economic Studies 48, 541-559.
Grossman, S.J. and O.D. Hart, 1981, The allocational role of takeover bids in situations of
asymmetric information, Journal of Finance 36, 253-270.
Hess, A.C. and P.A. Frost, 1982, Tests for price effects of new issues of seasoned securities, Journal
of Finance 36, 11-25.
Jensen, M.C. and W. Meckling, 1976. Theory of the firm: Managerial behavior, agency costs and
capital structure, Journal of Financial Economics 3, 305-360.
Korwar, A.N., 1981, The effect of new issues of equity: An empirical examination Working paper
(University of California, Los Angeles, CA).
Leland, H. and D. Pyle, 1977, Information asymmetries, financial structure and financial inter-
mediaries, Journal of Finance 32, 371-387.
Majluf, N.S., 1978, Study on mergers: A rationale for conglomerate mergers, Unpublished Ph.D.
dissertation (MIT, Cambridge, MA).
Miller, M.H. and K. Rock, 1982, Dividend policy under asymmetric information,Working paper,
Nov. (Graduate School of Business, University of Chicago, Chicago, IL).
Myers, S.C., 1977, Determinants of corporate borrowing, Journal of Financial Economics 5,
147-175.
Myers, S.C. and N.S. Majluf, 1978, Stock issues and investment policy when firms have informa-
tion investors do not have, Working paper (Sloan School of Management, MIT, Cambridge,
MA).
Rendleman, R.J., 1980, Information asymmetries and optimal project financing, Working paper
(Graduate School of Business, Duke University, Durham, NC).
Ross, S.A., 1978, Some notes on financial-incentive signalling models, activity choice and risk
preferences, Journal of Finance 33, 777-792.
Ross, S.A., 1977, The determination of financial structure: The incentive-signalling approach, Bell
Journal of Economics 8, 23-40.
Smith, C.W., and J.B. Warner, 1979, On financial contracting: An analysis of bond covenants,
Journal of Financial Economics 7, 117-161.
Stiglitz, J.E., and A. Weiss, 1981, Credit rationing in markets with imperfect information, Part I,
American Economic Review 71, 393-410.
• ~
El.SEVIER Journal or Financial Economics 42 (1996) 159-185
JOURNAL OF
Flnancial
ECONOMICS
Timing, investment opportunities, managerial discretion,
and the security issue decision
Kooyul Jung3, Yong-Cheol Kimb, Rene M. Stulz*·c,d
•M.J. Neeley School of Business. Texas Christian UniL-ersity, Fort Worth, TX 76129, USA
bCo/lege of Commerce and Industry. Clemson Unitiersity, Clemson, SC 29634, USA
•Max M. Fisher College of Business. Ohio State University, Columbus, OH 432 /0, USA
dNational Bureau of Economic Research, Cambridge, MA 0283/, USA
(Received January 1995; final version received November 1995)
Abstract
This paper investigates the ability of the pecking-order model, the agency model, and
the timing model to explain firms’ decisions whether to issue debt or equity, the stock
price reaction to their decisions, and their actions afterward. We find strong support for
the agency model. Firms often depart from the pecking order because of agency consider-
ations. We fail to find support for the timing model.
Key words: Security issue; Managerial discretion; Equity; Debt; Investment opportunities
JEL classification: G3
2
1. Introduction
Why is it that some firms raise new funds by issuing equity and others issue
debt? There are three important explanations for this choice in the literature:
• Corresponding author.
We are grateful for useful comments to Steve Buser, K.C. Chan, David Denis, Harry DeAngelo,
Thomas George, David Mayers, Kathy Kahle, Steve Kaplan, Tim Opler, John Persons, Patricia
Reagan, Jay Ritter, David Scharfstein, Clifford Smith (the editor), Rick Smith, Chester Spatt, Robert
Vishny, an anonymous rereree, and to the participants in the 1993 NBER Summer Institute, the 1995
American Finance Association meetings, and at finance seminars at Arizona State University, Hong
Kong Institute of Science and Technology, University of British Columbia. Ohio State University
and University of St. Gallen. The first author acknowledges financial support from the Charles
Tandy American Enterprise Center at Texas Christian University.
0304-405X/96/$15.00 © 1996 Elsevier Science S.A. All rights reserved
Pll S0304-405 X(96)00881- I
160 K. Jung et al./Journal of Financial Economics 42 (1996) 159-185
(1) the pecking-order model, (2) the agency model, and (3) the timing model. The
pecking-order model is based on the view that information asymmetries be-
tween new investors and managers who maximize the wealth of existing share-
holders make equity issues more costly than debt issues and therefore imply
a financing hierarchy. 1 Firms therefore prefer issuing debt to issuing equity, and
experience a negative stock price reaction if forced to issue equity. The agency
model relies on the argument that managers sometimes pursue their own
objectives, such as firm growth, at the expense of shareholders. If management
pursues growth objectives, equity issues are valuable for shareholders when
undertaken by firms that have good investment opportunities, but not other-
wise. The timing model has evolved from the striking finding of Loughran and
Ritter (1995) and Spiess and Affleck-Graves (1995) that firms experience long-
term underperformance after they issue equity. As argued by Stein (1995), if
equity is overpriced and the market underreacts to equity issues, then manage-
ment maximizes the wealth of existing shareholders by issuing equity.
A theory of the corporate security issue choice should explain (l) why firms
choose to issue a particular security, (2) how the market reacts to that choice,
and (3) the actions of the firm after the issue. The pecking-order model is well-
articulated and addresses each one of these questions. The agency model is
much better developed as an explanation of the cross-sectional variation in
capital structures (see Harris and Raviv, 1991; Smith and Watts, 1992) than as
a model of security issue choice. The timing model addresses the three questions,
but it relies on the assumption that the market fails to incorporate all the
information communicated by a security issue. In this paper, we develop
a unified analysis of the implications of the agency model to address the three
questions that have to be answered to provide a satisfactory model of security
issue choice. We then proceed to investigate how well the pecking-order model,
the agency model, and the timing model explain the data.
Our results strongly support the agency model. We find that firms issuing
equity are of two types: (i) firms with valuable investment opportunities that
seek financing to grow profitably and (ii) firms that do not have valuable
investment opportunities and have debt capacity. Without _agency ~
osts of
managerial discretion, one would not expect the latter firms to issue eqmty. The
agency model predicts that equity issues by such firms are bad n~ws _for
shareholders, since they enhance managerial discretion when ?1anagers objec-
tives differ from shareholders’ objectives. We find that, controlling for ~t~er firm
and issue characteristics firms without valuable investment opportumttes have
a more negative stock ~rice reaction to equity issues than firms with better
1Scc Myers (1984). Information asymmetries between management and outside investors do not
necessarily imply a financing hierarchy. Examples of models which emphasize informational asym-
metries but do not obtain a pecking-order result arc Brennan and Kraus (1987) and Noe (1988).
K. Jung et al. I Journal of Financial Economics 42 (I 996) 159-185 161
investment opportunities. We provide other evidence supporting the view th
at
some firms issue equity to benefit management rather than shareholders.
In
particular, we show that firms without valuable investment opportunities is
su-
ing equity invest more than similar firms issuing debt, that firms with l
ow
managerial ownership have worse stock price reactions, and that the wo
rst
stock price reactions occur for firms without valuable investment opportunit
ies
issuing equity to finance capital expenditures.
Even though firms issuing equity perform more poorly than firms issuing deb
t
on average, our cross-sectional regressions show that the subsequent evoluti
on
of the stock price does not explain the firms’ security issue choice. The reason
for
this is that the cross-sectional standard deviation of post-issue cumulat
ive
abnormal performance is extremely large, so that extremely large samples a
re
required to obtain statistically significant results. One interpretation of th
is
result is that our sample of 192 primary equity issues and 276 bond issues is t
oo
small to obtain a powerful test of the timing model. Alternatively, one mig
ht
argue that there is too much variation in long-term performance followi
ng
equity issues for it to be an important determinant of management’s decisio
n.
We provide evidence that the second interpretation should be taken seriou
sly
using a sample that is similar in size to the samples used in long-term perfo
r-
mance studies.
We proceed as follows. In Section 2, we provide a more detailed analysis of the
agency argument and its implications for the interpretation of the stock pri
ce
reaction to equity issues. In Section 3, we introduce our sample and discuss t
he
characteristics of firms issuing debt and equity. Section 4 provides estimates
of
an issue choice model. In· Section 5, we investigate how the stock price reactio
n
relates to firm characteristics. Section 6 shows that debt- and equity-issui
ng
firms have distinct investment -patterns following the new issue. Concludi
ng
remarks are presented in Section 7.
2. Models of the security issue decision
In this section, we analyze the role of agency costs in the security issue
decision and compare the predictions of the agency cost model to the pred
ic-
tions of other models in the literature. To understand the role of agency costs
in
the security issue decision, it is best to investigate a special case of Myers an
d
Majluf (1984). In their model, management has better information than inves-
tors about assets in place and about the firm’s investment opportunities.
If
management can issue securities at a higher price than they are truly wor
th
given its information, it chooses to do so to maximize the wealth of the existi
ng
shareholders. Riskless debt cannot be sold for more than it is worth, but ris
ky
debt and equity can. When the firm announces issues of risky securities, therefo
re,
outsiders adjust their valuation of the firm to reflect the new information. Th
is
162 K. Jung el al. /Journal of Financial Economics 42 (1996) /59-1115
adjustment is trivial if the securities issued are not very sensitive to firm value,
but is significant in the case of equity. The valuation impact of equity issues
increases their cost and induces firms to issue equity only as a way of raising
funds when debt financing would be extremely costly because the firm has
exhausted its ability to sell low-risk debt. For these results to hold, though, it is
crucial for outsiders to be less well-informed than management about the
components of firm value.
Suppose now that outsiders know the value of assets in place in the Myers and
Majluf model. Then, as recognized by Myers and Majluf, the model coJlapses:
the firm always invests if it has a positive NPV project and, in their set-up,
always issues equity to finance it. With agency costs, this special case remains
interesting. To see why, consider an aJJ-equity firm that is highly unlikely to have
profitable investment opportunities. If the management of that firm always
maximizes shareholder wealth, an equity issue undertaken to fund a project is
good news. It means that the firm has unexpectedly obtained a positive NPV
investment project. In the presence of agency costs of managerial discretion,
however, an equity issue that enables management to invest is not necessarily
good news and can be bad news altogether. A management investing in negative
NPV projects would rather finance that investment with equity; debt financing
for a negative NPV project eventuaJly reduces resources under management’s
control since the present value of the debt payouts exceeds the present value of
the project’s payoffs.
Jensen ( 1986) and Stulz ( 1990) show that leverage limits management’s discre-
tion and hence reduces the agency costs of managerial discretion. First, manage-
ment has less control over the firm’s cash flows since these cash flows have to be
used to repay creditors. Second, management is monitored by creditors who
want to make sure that they will be repaid. However, leverage also has adverse
effects on firm value. A firm with good projects but high leverage is Jess able to
take full advantage of these projects. For instance, the impact on investment of
an adverse liquidity shock increases with the amount of leverage. Consequently,
firms with good projects want to limit their leverage and, if levered, are more
likely to choose equity financing. Bernan~e, ?~rtler, and Gilch~ist (1993) review
the literature on the relation between hqu1d1ty shocks and mvestment, and
Lang Qfek and Stulz (1996) show that investment is negatively related to
lever~ge for,low-q firms. The agency costs that arise becau~e ~levered firm may
be unable to pursue the investment policy that would max1m1ze the value o~ an
all-equity firm are called here the agency costs of debt (see Jensen and Mec~lmg,
1976; Myers, 1977). Smith and Watts (1992) provide extensive cross-sect1onal
evidence of such agency costs, showing a negative relation between investm~nt
opportunities and leverage, and Titman and Wessels (1988) document a negative
relation between R&D and leverage.
In Fig. 1, we show the optimal amount of leverage for given investment
opportunities. The optimal amount of leverage is the amount at which the
K. Jung el al./Joumal of Financial Economics 42 (1996) 159-185
Agency
costs
0 Leverage
Fig. I. Optimal leverage and agency costs of debt and managerial discretion.
163
This figure shows optimal leverage as a function of the marginal agency costs of debt (DD) and the
marginal agency costs of managerial discretion (MM) for a given investment opportunity set. An
improvement in investment opportunities shifts the marginal agency costs of debt curve to D’D’
and the marginal agency costs of managerial discretion curve to M’ M’, so that optimal leverage falls
from L to L’.
marginal agency costs of debt equal the marginal agency costs of managerial
discretion. Based on our previous discussion, the marginal agency costs of debt
should increase with leverage and the marginal agency costs of managerial
discretion should fall with leverage. We show how a shift in investment oppor-
tunities leads to a decrease in the optimal amount of leverage: for each level of
leverage, an increase in investment opportunities (1) increases the marginal
agency costs of debt because the firm has more to Jose from financial distress and
(2) decreases the marginal agency costs of managerial discretion because the
objectives of management and shareholders become more congruent when
investment opportunities become better.
Since equity provides unrestricted funds, why is it that management ever
chooses to issue debt? Issuing equity has both direct and indirect consequences
for management. The direct effect is an increase in managerial discretion, which
management values. However, the indirect effect can be quite adverse for
management depending on the firm’s situation. If the firm does not have
valuable investment opportunities, an equity issue means that the agency costs
of managerial discretion increase, providing greater incentives for outsiders to
try to affect management’s actions. In particular, control activities, such as
takeovers, active monitoring by large shareholders, monitoring by board mem-
bers, and proxy fights, all become more advantageous for shareholders and
outside investors. Issuing equity inappropriately can therefore increase the
probability that management will lose control through corporate control ac-
tions unless it is well protected from such actions. Zwiebel (1994) presents
a model in which management issues debt because of a threat from the ma k t
for corporate control. Hoshi, Kashyap, and Scharfstein (1993) h r e
related model in which the better firms choose financing Wt.th 1 . av~ a ow momtonng,
164 K. Jung el al./Journal of Financial Economics 42 (1996) 159-185
intermediate-quality firms choose financing with high monitoring, and the worst
firms choose financing with low monitoring. (In their paper, financing with low
monitoring is public debt and financing with high monitoring is bank financing;
here, financing with low monitoring is equity and financing with high monitor-
ing is public debt.) In addition, equity financing reduces the fraction of
votes controlled by management and its allies unless they increase their
investment in the firm (see Stulz, 1988). Consequently, equity financing both
increases the benefits from outside intervention and makes outside intervention
easier.
Taking into account the agency costs of managerial discretion makes the
information content of new security issues more complicated. To understand
this information content, it is best to focus on the cross-sectional relation
between stock price reactions and a firm’s investment opportunities, since the
agency costs of managerial discretion are inversely related to the quality of the
firm’s investment opportunities. If there is no uncertainty about the value of
a firm’s investment opportunities, the issuing decision is straightforward in the
pecking-order model. If a firm has sufficiently good investment opportunities, it
issues equity if it cannot issue debt and the issue is not very informative
about the value of assets in place. In contrast, if the firm has no valuable
investment opportunities, it never issues equity. For firms with sufficiently good
investment opportunities, the interests of management and shareholders should
coincide so that they will follow the pecking-order model. Firms that can finance
with low-risk debt do so; otherwise they either issue equity or do not invest
at all if equity is too underpriced. For firms that have no valuable investment
opportunities, however, there are good reasons to expect departures
from the pecking-order model if management pursues objectives of its own.
In particular, management may issue equity to keep the firm growing even
though the firm has no positive NPV investment opportunities. For such
firms, an equity issue reveals to outsiders that management has to raise
funds to finance its plans, that it has decided to proceed with poor
investments, and, finally, that it views the risks to its position from doing all
this to be worth taking. If the equity issues are equally unanticipated,
the news for outsiders is worse for the firm with no valuable investment
opportunities.
At this point, it is useful to summarize the view that ag:ncy c.o~ts matter ~or
security issues by showing how these costs affect the firm s d~c1s10n r:gardmg
which security to issue (the issue decision), ~ow stoc~ pnce reacuons are
consistent with the existence of such costs (the information content), and how
the firm’s behavior after the issue is affected by these costs (the ex post actions): If
the threat of outside intervention is held constant, agency considerations imply
that managers favor equity over debt, so that firms for which the agency costs
of managerial discretion are important issue equity even though share-
holders would be better off with a debt issue or no issue at all. However, an
K. Jung et al./Joumal of Financial Economics 42 (1996) 159-185 165
equity issue that is not in the interests of shareholders will have a
negative
impact on shareholder wealth to the extent that it is not anticipated
because
the funds are likely to be invested poorly and because manageme~t is not
as constrained by monitoring from outside investors as was
expected.
Finally, whether they have good investment opportunities or not,
the firms
that issue equity do so to have the flexibility to grow and should
therefore
grow more than debt-issuing firms. This should be true even for firms
that have
debt capacity but no valuable investment opportunities, since in th
ese firms
management chooses to issue equity to have more freedom to inve
st in poor
projects.
It is important to note that the implications of agency costs do not m
ake the
considerations emphasized in the pecking-order model irrelevant. Irres
pective of
the importance of the agency costs of managerial discretion, there will
always be
some level of undervaluation of the existing shares at which ma
nagement
chooses not to issue. For firms whose agency costs of managerial disc
retion are
small enough, it may be that the pecking-order model applies exa
ctly. The
pecking-order model based on information asymmetries assumes that
manage-
ment maximizes shareholder wealth whereas the agency cost view ass
umes that
management pursues objectives of its own. As emphasized by Dy
bvig and
Zender ( 1991) and others, the pecking-order model makes an ad hoc as
sumption
about management’s objectives that would not be appropriate if sha
reholders
could choose a compensation policy for management such that the ex a
nte value
of the firm is maximized. Since both the pecking-order model and th
e agency
model rely on ad hoc assumptions about managerial objectives, only
empirical
evidence can allow us to evaluate the economic relevance of each mod
el for the
security issue decision.
Models with information asymmetries that assume away the agency c
osts of
managerial discretion are most successful at explaining the negative st
ock price
reaction to equity issues. As modified by Cooney and Kalay (1993), t
he Myers
and Majluf model can explain that high-growth firms issuing equi
ty would
have a more positive stock price reaction than low-growth firms
. Hence,
relating the stock price reaction to investment opportunities is not
sufficient
to make the case for the agency model of security issues. This is
why it is
also important to consider the choice decision and the post-issue a
ctions of
the firm.
With the timing model, managers issue equity when they know th
at it is
overvalued. Since the market underreacts to equity issues, firms issui
ng equity
perform poorly in the long run as the market corrects the overvalua
tion that
exists at the time of issue. A market underreaction to equity issues c
ould play
a role both in the agency model and in the pecking-order model. The
question
we therefore want to address is whether timing is a first-order consid
eration in
the security issue choice decision. Importantly, none of the models ex
plain the
long-run post-issue abnormal returns.
166 K. Jung el al. /Journal of Financial Economics 42 (I 996) 159- I 85
3. The sample
To obtain our sample of new bond issues and primary stock offerings, we use
the Registered Offerings Statistics File from 1977 to 1984. For the stock
offerings, we use the Corporate Financing Directory published by the Invest-
ment Dealer’s Digest to exclude all issues that involve secondary stock offerings
and all shelf offerings. We restrict the sample to firms whose stock returns are
available on the Center for Research in Security Prices (CRSP) tape for the
whole calendar year before the announcement date. The announcement dates
come from the Wall Street Journal Index. We use as our event date the first
mention of a security issue before the offering date and exclude security issues
for which such announcements are not available. We exclude utilities and
banking firms to conform to the earlier literature. We also eliminate firms that
have confounding announcements, such as dividend or earnings announce-
ments.
We compute abnormal returns using a method similar to the one used by
Asquith and Mullins (1986). For each calendar year in the sample we rank
securities in the CRSP daily file according to their beta estimated using the
market model. We then divide the securities into ten portfolios based on
estimated betas. For each firm issuing a security, we compute the abnormal
return over a two-day period that includes the day of the Wall Street Journal
announcement and the day preceding the announcement. The abnormal return
is defined to be the return of the issuing firm minus the return of the portfolio to
which the firm belongs, although all of our results hold if we compute abnormal
returns as market model residuals.
Table 1 provides a summary of the abnormal return data for the stock and
bond issues. The results are similar to those reported in earlier papers in that
equity announcements have a significant negative stock price reaction and debt
announcements have an insignificant stock price reaction. 2 Table 1 also reports
various characteristics affirms issuing debt and equity. The median debt-issuing
firm has a stock market capitalization about four times larger than the median
equity-issuing firm and raises about four times ~or~ funds t~rough the issue.
The equity-issuing firms are riskier than the debt-1ssumg firms m that they have
both a higher beta and greater stock return volatility. The.leverage measure ~hat
uses the market value of common stock in the denommator does not differ
between firms issuing debt and those issuing stock, whereas the leverage
2For instance, Mikkelson and Partch (1986) find an average abnormal return for stock issues of
– 3.S6o/o and straight debt of – 0.23%. Eckbo (1986) finds a similar result for debt issues. Asquith
and Mullins (1986) find an abnormal return for primary stock issues for industrial firms of – 3.0%
whereas Masulis and Korwar(l986) find a stock price reaction of – 3.25%. Barclay and Litzenber-
ger (1988) find an abnormal return of – 2.44% for the three hours surrounding the announcement
on the Broad Tape.
K. Jung et al./Joumal of Financial Economics 42 (1996) 159-185 167
Table 1
Abnormal returns and firm characteristics for 192 equity and 276 bond issues from 1977 to 1984
The abnormal returns are computed for the day of the WSJ announcement and the previous day.
Amount equals the gross proceeds of the issue in millions of dollars. LTD is the book value of the
firm’s long-term debt. Cash flow is operating income before depreciation minus total taxes adjusted
for changes in deferred taxes, minus gross interest expense and minus dividends paid on common
and preferred stock, divided by total assets (TA). Market-to-book is the ratio of firm market value
(market value of equity plus TA minus book value of equity) to TA. All accounting data are for the
end of the fiscal year before the issue. The leading indicators are the six-month leading indicators.
The volatility of the firm’s stock return and the firm’s beta are obtained using the CRSP daily data
file for the period ( – 240, – 40). Difference is the mean of a variable for stock issues minus the mean
of the same variable for bond issues; the p-value is for the null hypothesis that the difference is zero
assuming unequal variances for the two subsamples.
Stock issues Bond issues
Mean Median Mean Median Difference
Abnormal return -2.70% -2.63% -0.09% – 0.15% – 2.62*
Amount 47.98 28.25 140.00 100.00
– 92.01 *
Market value of equity 682.74 186.02 2941.70 883.62
-2258.97*
(MVCS)
0.13 -0.09* Proceeds/MVCS 0.15 0.13 0.24
Dividend yield 2.06 1.43 3.96 3.69
– 1.90*
LTD/MVCS 0.65 0.42 0.72
0.41 -0.Q?
LTD/TA 0.29 0.28 0.2
3 0.21 0.06*
Cash flow 0.09 0.09 0.10
0.09 -0.Ql
Cash + Liquid assets/TA 0.06 0.04 0.06 0.04 0.00
Market-to-book 1.48 1.25 1.13
1.02 0.35*
Leading indicators O.Q3 0.03 0.00
0.00 0.03*
1 ]-month prior cumulative 13.95% 15.07% – 1.63%
– 3.26% 15.58*
excess return
0.24*
Beta 1.39 1.35 1.1
5 1.06
Volatility 7.27% 6.28%
4.67% 3.20% 2.60*
3-year raw returns 59.47% 37.86%
76.20% 52.49% – 16.73
5-year raw returns 98.88% 57.12% 146.
75% 98.56% -47.87**
Size-matched 3-year -7.89% -13.90% – 5.16%
-3.64% -2.74
cumulative returns
– 34.72
Size-matched 5-year – 32.69% -46.81% 2.
03% -18.60%
cumulative returns
*(**)denotes significance at the 0.01 (0.05) level.
measure that uses the book value of total assets in the denominator is higher for
firms that issue equity. Therefore, book leverage is more supportive of the
pecking-order story than a market measure of leverage.
The pecking-order model predicts that firms are more likely to issue equity
when the stock price experiences positive abnormal returns before the issue.
Measuring the cumulative excess return of the issuing firm’s common stock as in
168 K. Jung et al./Journa/ of Financial Economics 42 (1996) 159-185
Asquith and Mullins (1986), we find that firms that issue common stock have
experienced significant positive abnormal returns for the 11 months before the
stock issue, whereas firms that issue bonds experience insignificant negative
cumulative abnormal returns on average. Mikkelson and Partch (1986) obtain
a similar result on a smaller sample of bond offerings. The result for debt is
inconsistent with the conjecture of Lucas and McDonald (1990) that firms
issuing risky debt should have positive cumulative abnormal returns on average
if debt is viewed as equity with less risk. The firms issuing equity and those
issuing debt have similar cash flows before the issue. We also investigate, but do
not report here, earnings to total assets, earnings before interest and taxes
(EBIT) to total assets, and net operating income to total assets. In all cases, the
mean for equity-issuing firms is larger, but the difference in means is significant
only for net operating income. Finally, the firms issuing debt have a substan-
tially higher dividend yield than the firms issuing equity.
Using the ratio of firm market value (defined as the market value of equity
plus the book value of total assets minus the book value of equity) to the book
value of assets (market-to-book) as a proxy for investment opportunities as in
Smith and Watts (1992), firms issuing equity have better investment opportuni-
ties than firms issuing debt at the time of the announcement. In addition, firms
issuing equity (but not those issuing debt) are more likely to do so when the
leading indicators suggest good economic conditions and therefore good invest-
ment opportunities; Choe, Masulis, and Nanda (1993) observe the same result.
Finally, the cumulative abnormal returns before the issue (discussed in the
previous paragraph) are consistent with an improvement in the investment
opportunities of firms issuing equity before the issue.
The timing model relies on the observation that equity-issuing firms perform
poorly following the issue. Since this long-term performance is poor on average,
it is consistent with the view that firms time their issues to coincide with periods
when their equity is overvalued. Cheng (1994) provides further support for this
view by showing that debt-issuing firms do not ha~e poor long-term abnormal
returns and that firms issuing equity that do not mvest the proceeds have the
worst abnormal returns. In Table 1, we provide evidence on the long-term
performance of the firms in our sample. The cumulative returns are buy-and-
hold returns. We show both raw returns and excess returns obtained by
subtracting from the return of the issuing firm the return of a matching firm of
similar size that has not issued equity in previous sample years. We also
compute net-of-market returns but do not report them here since they lead to
the same conclusions as the results we report. Our procedures are the same as
the ones used by Loughran and Ritter (1995). . . .
The raw returns are significantly positive both for bond- and eqmty-1ssumg
firms. The difference between the raw returns of bond-issuing firms and equity-
issuing firms is significant at the five-year horizon with a t-statistic of 2.06, but is
not significant at the three-year horizon. The substantial worsening of the
K. Jung et al./Journa/ of Financial Economics 42 (1996) 159-185 169
performance of equity-issuing firms over the last two years of the five-y
ear
horizon is surprising. Turning to excess returns, we find that equity-issuing fi
rms
have significant negative excess returns on a five-year horizon at the 0.10 le
vel.
The excess returns are negative but not significant at the three-year horiz
on.
Irrespective of the horizon, though, the mean excess return is large in absol
ute
value for equity-issuing firms and consistent with previous evidence on
the
underperformance of equity-issuing firms. Bond-issuing firms have positi
ve,
although not significant, average excess returns at the five-year horizon. Us
ing
nonparametric statistics (rank and sign tests), excess returns are significan
tly
negative for equity-issuing firms but not for debt-issuing firms. There is
no
significant difference in the means of excess returns between bond- and equ
ity-
issuing firms, but the medians are significantly different. That such large dif
fer-
ences in means are not significant is consistent with the view articulated
in
Kothari and Warner (1995) that long-term returns have considerable cro
ss-
sectional variation so that statistical tests using such returns have low pow
er.
The limited significance of our results using long-term returns is no doubt pa
rtly
explained by the fact that the number of equity issues used here is less th
an
one-tenth of the number used in Loughran and Ritter (1995). In additi
on,
however, our sample contains larger and more established firms since it only
has
Compustat firms. Brav, Geczy, and Gompers (1994) argue that underperf
or-
mance is more pronounced for small issuing firms.
4. An empirical analysis of the security issue choice
In this section, we investigate an empirical model of security issue choice fo
r
our sample firms. This model uses standard variables from the literature
to
predict the security issue choice plus a proxy for investment opportunities a
nd
measures of long-term post-issue abnormal returns. Since the agency costs
of
debt are higher for firms with better investment opportunities, one expects
the
probability that a firm will issue equity to increase with investment opportu
ni-
ties if management maximizes shareholder wealth. Firms with high agency co
sts
of managerial discretion will issue equity when they have poor investm
ent
opportunities, but such firms are expected to be a subset of the sample so tha
t in
a logistic regression model they will be firms that are not expected to issue equ
ity
and hence issue against type. If our proxy for investment opportunities simply
proxies for firm overvaluation, as partisans of the timing model might arg
ue,
then inclusion oflong-term abnormal returns should account for overvaluati
on.
Further, if the timing model plays an important role in the issuing firm’s
decision, long-term cumulative excess returns should significantly affect
the
firm’s issuing decision because the timing model relies on the argument t
hat
management knows when future performance will be poor and issues acco
rd-
ingly. Using actual long-term returns as a proxy for management’s expectati
ons
170 K. Jung et al./Journal of Financial Economics 42 (/996) 159-185
of long-term returns amounts to assuming that management has perfect
foresight.
The literature on the determinants of firms’ capital structures is extensive, but
some variables are pervasive in the existing empirical work. Masulis (1988) and
Harris and Raviv (1991) contain references to empirical studies that use these
variables as well as references to theoretical papers that motivate their use. In
this paper, we focus on a small number of determinants of leverage that are
commonly considered by empiricists and reflect certain key ideas:
1. Taxation. Because of the deductibility of interest payments, a number of
papers argue that the gain from debt financing relative to equity financing
increases with the firm’s tax rate. The literature has shown that the firm’s tax
status affects the issue decision (see MacKie-Mason, 1990). As a proxy for these
benefits, we use tax payments divided by the book value of total assets for the
year preceding the issue.
2. Costs of financial distress. As debt and firm risk increase, financial distress
and bankruptcy become more likely. As a risk proxy, we use stock return
volatility measured over the 200 days preceding the issue. Profitability is
measured as cash flow divided by total assets, and leverage is measured as
long-term debt divided by total assets. We use other proxies for risk (beta
instead of volatility), profitability (earnings measures), and leverage (market
value of equity instead of total assets) but do not report the results because our
conclusions are insensitive to the choice of proxies for bankruptcy risks and
costs.
3. Asymmetric information. Following Myers and Majluf (1984), it is well-
established that issuing equity is more expensive when there is asymmetric
information between firm insiders and outsiders. Therefore, firms for which this
information asymmetry is large should issue debt if they can or abstain from
raising funds altogether. As emphasized by Korajczyk, Lucas, and McDonald
(1991), firms should time equity issues for periods when the information asym-
metry is smaller. Following Lucas and McDonald (1990), firms are more likely
to have good projects and hence raise funds if their returns before the issue are
high (measured here by net-of-market returns over the 200 days before the issue)
and leading indicators of economic activity are favorable. Firms that issue when
they have slack are also more likely to do so because of low information
asymmetries. We measure slack by cash and liquid assets normalized by total
assets.
In some of our regressions, we also control for the amount raised through the
security issue since net proceeds have been found to affect the stock price
reaction in some studies. Presumably, the amount raised by the firm and the
type of security issued are jointly endogenous variables. This suggests that
logistic regressions that do not include the amount raised as an explanatory
vari~ble have the interpretation of reduced-form equations, whereas equations
that mclude tbe amount raised suffer from a simultaneous equation bias. A more
K. Jung et al./Journal of Financial Economics 42 (1996) 159-185
171
important reason to consider regressions without the amount rai
sed as an
explanatory variable is that such regressions can be used by investors
to forecast
whether a firm will issue equity or debt, whereas regressions that incor
porate the
size of the issue cannot (since they incorporate information not availa
ble before
announcement of the type of security issued).
Regression (1) in Table 2 shows that investment opportunities play a s
ubstan-
tial role in the new issue decision. With our logistic model, an equity
issue takes
the value one and a debt issue takes the value zero. Therefore,
a positive
coefficient indicates that a firm is more likely to issue equity. Mark
et-to-book
has a positive coefficient that is highly significant. Further, market-to
-book has
substantial explanatory power in that, if it is omitted, the pseudo-R
2 falls by
almost one-third. Other variables indicative of good investment opp
ortunities
are significant also. Past cumulative excess returns and leading
indicators
have positive coefficients with p-values of less than 0.01. Cash fl
ow is not
significar.t, but some variables emphasized by other capital structur
e theories
are significant. The coefficient on tax payments divided by total
assets is
negative as expected and highly significant. Leverage, as measured by
long-term
debt to total assets, is insignificant. This result holds when we use
alternate
leverage measures and is not surprising considering the earlier liter
ature. For
instance, Baxter and Cragg (1970) do not find a significant leverage
coefficient
either, although Marsh (1982) uses deviations from target leverage in
his regres-
sions and finds that firms with high leverage relative to a target are m
ore likely
to issue equity. Since leverage and volatility are correlated, we omit v
olatility in
a regression not reproduced here; doing so does not make the coe
fficient on
leverage significant. Finally, we would expect slack to have a positive
coefficient,
but instead it has an insignificant negative coefficient. In regression (
2), we add
total assets as an explanatory variable. Total assets could be a pro
xy for the
degree of information asymmetry, since large firms are followed more
closely by
analysts and have more stringent reporting requirements. The coe
fficient on
total assets is significantly negative, indicating that large firms are les
s likely to
issue equity. All our other inferences remain unchanged by the additi
on of total
assets, except that stock return volatility ceases to have a significant ef
fect on the
probability of issuing equity.
In regression (3), we add post-issue cumulative excess returns as an ex
plana-
tory variable. The timing model suggests that the coefficient on
post-issue
cumulative excess returns should be significantly negative, so that firm
s expect-
ing poor performance would be more likely to issue equity. We repo
rt only the
regression with the five-year size-adjusted excess returns. We estimate
the same
regression using three-year size-adjusted excess returns, three-year an
d five-year
raw returns, and three-year and five-year net-of-market returns, but
the coeffi-
cient on long-term returns is never significant. This finding has tw
o possible
interpretations, however. First, it could mean that timing consideratio
ns are not
important in firms’ decisions. Second, there could be so much varia
tion in the
172 K. Jung et al./Journal of Financial Economics 42 (1996) 159-185
Table 2
Determinants or firm type
Logistic regressions in which the dependent variable takes the value one for equity issues and zero
otherwise. The sample has 276 debt issues and 192 equity issues from 1977 to 1984. Market-to-book
is the ratio of firm market value to total assets (TA). Cash flow is operating income before
depreciation minus total taxes adjusted for changes in dererred taxes, minus gross interest expense
and dividends paid on common and preforred stock, divided by TA. All book values are obtained
from Compustat for the year prior to the issue announcement. The volatility orthe stock return is for
the period ( – 240, – 40). MVCS is the market value of equity. The post-issue cumulative abnormal
return is the excess return of issuing firms over firms with similar size before the issue. The pseudo-R
2
equals 1 – (log-likelihood at convergenceflog-likelihood at zero); p-values for the chi-square statistic
are in parentheses.
Regression (1) (2) (3) (4) (5)
Intercept -3.27 – 2.57 – 3.23 -2.50 3.16
(0.01) (0.01) (0.01) (0.01) (0.01)
Tax payments/TA – 11.99 – 12.97 -9.31 -9.09 – 20.37
(0.01) (0.01) (0.02) (0.03) (0.01)
Long-term debt/TA 0.81 0.25 0.97 1.83 – 1.02
(0.36) (0.78) (0.31) (0.06) (0.32)
Market-to-book 2.13 1.96
2.06 1.68 2.20
(0.01) (0.01) (0.00) (0.00) (0.00)
Cash flow 0.11 0.23
-0.58 -2.07 0.96
(0.96) (0.93) (0.83) (0.47) (0.75)
Stock return volatility 5.40 2.99 5.98 13.24 – 5.86
(0.08) (0.35) (0.07) (0.01) (0.12)
6-month leading indicators 12.42 12.23 12.20 13.64 13.72
(0.01) (0.01) (0.01) (0.01) (0.01)
Past JI-month cumulative 2.33 2.29 2.10 2.74 1.53
excess return (0.01) (0.01) (0.01) (0.01) (0.01)
Cash and liquid assets/TA -265 -2.10 -2.94 – 1.26 – 1.30
(0.18) (0.29) (0.17) (0.60) (0.57)
Total assets -0.01
(0.01)
Gross proceeds/MVCS -5.04
(0.00)
Log of (Amount/MVCS) – 1.32
(0.0)
Post-Issue 5· year -0.0I
exce:ss returns (0.75)
Pseudo-R2 0.26 0.28 0.26 0.33 0.41
% correct 75.4% 75.8% 74.6% 79.5% 80.8%
K. Jung et al./Joumal of Financial Economics 42 (1996) 159-185 173
cross-sectional post-issue performance of firms that timing considerations are
only identifiable in large samples.
To investigate whether our lack of support for the timing model is due to our
sample size, we estimate a logistic regression using a sample more comparable in
size to the samples used in other long-run performance studies. Our expanded
sample includes 2,272 equity issues and 2,617 bond issues from 1970 to 1991 and
is constructed from the Registered Offerings Tapes and the Investment Dealer’s
Digest. This sample includes non-Compustat firms as well as Compustat firms.
We compute five-year post-issue buy-and-hold raw returns and size-adjusted
excess returns as we did for our original sample. The average return measures
are similar to those obtained in the long-run performance studies in that
long-run returns following equity issues are significantly negative and large in
absolute value and long-run returns following debt issues are insignificantly
different from zero. In a logistic regression with the post-issue cumulative
returns as the only dependent variable in addition to the constant, the post-issue
cumulative returns have a significant negative coefficient irrespective of how
they are computed, so that firms with poor post-issue returns are more likely to
issue equity. However, post-issue returns seem to explain very little: the pseudo-
R2 is on the order of 0.01 irrespective of how the post-issue returns are
computed. The regression with raw returns classifies 63.2% of the observations
correctly. The percentage of correct classifications falls to 54.3% for size-match-
ed excess returns. Even with a very large sample, therefore, it still turns out that
the timing model is not very helpful in understanding new issue decisions.
Interestingly, however, when we add to these regressions the cumulative abnor-
mal return for the year before the issue, this variable has an extremely significant
positive coefficient and the pseudo-R
2 increases strongly. In the regression using
raw returns, the pseudo-R
2 increases to 0.21 and the fraction of issues predicted
correctly increases to 72.8%; in the regression using size-adjusted returns, the
pseudo-R2 increases to 0.09 and the fraction predicted correctly increases
to 70.5%.
In regressions (4) and (5), we add measures of the size of the security issue
normalized by the market value of the firm’s equity as an explanatory variable.
These measures of the relative size of the security issue have no impact on the
effect of investment opportunities on the new issue decision. Not surprisingly,
given the statistics of Table 1, the relative size of the issue is negatively related to
the probability of issuing equity. Two firm characteristics seem to have effects
that depend on the relative size variable: leverage becomes significant for one
relative size measure and volatility ceases to be significant for the other. The size
measures have a substantial impact on the explanatory power of the regressions.
In regressions not reproduced here, we add total assets and the market value
of equity as separate explanatory variables. The addition of these variables
does not affect the conclusions drawn from Table 2, but their coefficients are
significantly negative. We re-estimate regressions (4) and (5) adding long-term
174 K. Jung et al./ Journal of Financial Economics 42 (1996) 159 1115
post-issue abnormal returns as explanatory variables but do not report the
results in the table. In the regression with the ratio of proceeds to pre-issue
market value of equity, long-term post-issue performance has a positive insigni-
ficant coefficient. In the regression with the log of the amount of the issue, the
coefficient on long-term performance is negative and significant at the 0.10 level.
The coefficient on market-to-book is 2.11 instead of 2.20 and its significance
level is unchanged. In this case, the percentage of correct predictions is 81.5%
instead of 82.1 %. There is therefore no convincing evidence that expectations
oflong-term cumulative excess returns play an important role in the firm’s issue
decision.
Although our regressions are parsimonious, they correctly classify a fraction
of the decisions similar to the fraction correctly classified in earlier papers. For
instance, the frequently cited paper by Marsh (1982) correctly classifies 75% of
the decisions, whereas our regressions in Table 2 correctly classify from 74% to
81 % of the decisions.
With this logistic model, we have firms that issue equity even though they
resemble firms that issue debt. One way to see this is to compare these firms to
the firms that issue debt and the firms that issue equity when predicted to do so.
To classify firms, we use regression (1) of Table 2. For that equation, the
threshold that minimizes the sum of the probability of a type I and the
probability of a type II error is 0.42. We find that 46 firms issue equity against
type using that threshold. In all characteristics except the ratio of proceeds to
the market value of equity, the firms that issue equity when predicted to issue
debt are indistinguishable from debt-issuing firms. In contrast. these equity-
issuing firms have many characteristics that are significantly different from firms
that issue equity and are predicted to do so. The firms issuing equity against type
pay more taxes relative to assets than other equity-issuing firms, so that one
would expect the tax deductibility of interest to be valuable for them. These
firms have less leverage than firms predicted to issue equity, although not
significantly so. They issue at times when leading indicators are neutral. Their
past abnormal returns are insignificantly different from zero. Their volatility is
closer to the volatility of firms issuing debt. Finally, these firms have muc~
poorer investment opportunities than firms predicted to issue equity. Their
mean and median market-to-book ratio is only trivially different from the. mean
and median market-to-book ratio of firms issuing debt. There are no signdicant . . f
cash flow differences among the three sets of firms. Given the charactens~tc.s_o
these firms, it is difficult to argue that they would benefit from the flexibility
resulting from issuing equity instead of debt.
Why do these firms issue equity against type? With the pecking-order model,
these firms should issue debt if information asymmetries are significant. Hence,
these firms might be issuing equity because they happen to have low information
asymmetries. Viswanathan (1993) models such deviations from the pecking-
order model. In this case, one would expect the information content of equity
K. Jung et al./Journa/ of Financial Economics 42 (1996) 159-185 175
issues to be low as well because it must be public knowledge that information
asymmetries are low since otherwise firms will face high costs of issuing equity
anyway. This would suggest that firms that issue equity against type would have
a small stock price reaction. It would not make sense for firms to issue against
type if information asymmetries are high because these firms have similar
characteristics to debt-issuing firms and therefore could issue debt. The peck-
ing-order model cannot explain why firms for which information asymmetries
are high would issue equity when they could issue debt. Equity issues by such
firms are consistent with the managerial discretion model, however. Investigat-
ing the stock price reaction to equity issues should therefore help us distinguish
between the two models.
5. The stock price reactions to security issues and investment opportunities
Among firms issuing equity, there are firms with good investment opportuni-
ties and limited debt capacity (provided that we can interpret firms with high
leverage to be firms with low debt capacity). One would expect these firms to
issue equity if they raise funds and that this action would be in the interest of
shareholders. Other firms have poor investment opportunities and look like
they could issue debt. The pecking-order model explanation for this behavior is
that information asymmetries for these firms are not important, suggesting that
the stock price reaction should be small. The agency model, in contrast, predicts
large stock price reactions if these issues are unexpected because the share-
holders of these firms would be better off having the firm either issue debt or not
raise funds. Since firms form a continuum across types, the agency cost model
would expect the firms for which issuing equity is the least likely to benefit
shareholders to have the largest fall in stock price at the announcement of an
equity issue, assuming that all issues are equally unanticipated. Earlier work by
Bayless and Chaplinsky (1991) demonstrates, using a different logistic model,
that firms issuing unexpectedly according to the logistic model have a greater
abnormal return in absolute value. This result holds for our logistic model also.
Table 3 provides estimates of the correlation between a firm’s type, defined by
the probability that a firm will issue equity based on the logistic model of the
previous section, and the firm’s abnormal return for each type of issue. The
correlation estimates for the equity issues are positive and significant; the
estimates for debt issues are negative but insignificant. These results are consis-
tent with the agency cost model but cannot be explained with the pecking-order
model.
We now turn to the relation between abnormal returns and a firm’s invest-
ment opportunities. With the managerial discretion model, equity issues are not
in the interest of shareholders for firms with poor investment opportunities. The
Pearson correlation between the stock price reaction to equity issues and the
176 K. Jung et al./Joumal of Financial Economics 42 (1996) 159-185
Table 3
Correlations between firms’ types and abnormal returns
Firm type is obtained from regression (1) of Table 2. Abnormal returns (ARs) are cumulative
abnormal returns for days ( – l, 0), with day 0 the day of the Wall Street Journal announcement of
the security issue.
Correlation measures
Correlation coefficient
between firm type and
abnormal returns
Spearman rank-sum
correlation between
firm type and
abnormal returns
Correlation between firm type
and A Rs for bond issues
(p-values)
-0.QJ
(0.65)
-O.Q7
(0.25)
Correlation between firm
type and ARs for equity
issues (p-values)
0.17
(0.02)
0.17
(0.02)
market-to-book ratio is 0.22 (p-value of less than 0.01) and the Spearman
rank-sum correlation is 0.18 (p-value of 0.01). When we divide the sample into
market-to-book deciles, we find that the highest market-to-book decile has
a mean abnormal return of – 0.22% whereas the lowest market-to-book decile
has a mean abnormal return of – 4.60%. Therefore, there is a robust relation
between stock price reactions to equity issues and market-to-book. For debt
issues, the correlation measures are respectively 0.11 (p-value of 0.07) and 0.10
(p-value of0.10). The relation between stock price reactions and market-to-book
is much weaker for debt issues. In a regression of abnormal returns on a con-
stant and market-to-book, the coefficient on market-to-book is 0.97 with a
t-statistic of 2.63 for equity issues and it is 0.51 with a t-statistic of 1.39 for debt
issues. These results are stronger than the results from earlier research which
either uses the market-to-book ratio or Tobin’s q. Barclay and Litzenberger
(1988) and Pilotte (1992) find insignificant results using conventional levels of
significance, but they have fewer issues than we do. Denis (1994) has a large
sample yet finds a weaker relation than here. However, our sample stops in 1984,
so that it is not affected by the subsequent change in reporting practices of the
Wall Street Journal. 3
Market-to-book is positively correlated with a variable emphasized in
models that focus on adverse selection, namely the runup in the firm’s stock
3 Before 1985, the WSJ reports on equity issues as a regular news item. From 1985, most of the
infonnation on new issues is reported in the ‘new securities issues column’ which contains mostly
ofl’ering information. Hence, the event dates since 1985 reflect issues that are more likely to be
anticipated because the announcement of an equity issue is typically made earlier (by days or weeks)
via news-wire services than the WSJ listing. This biases the abnormal return estimate.
K. Jung et al./ Journal of Financial Economics 42 (1996) 159-185 177
price before the issue. Market-to-book is also likely to be correlated with other
variables emphasized in the literature. Therefore, it is important to investigate
whether the relation between abnormal returns and market-to-book can be
attributed to its role as a proxy for other variables that may have nothing to do
with managerial discretion. We investigate this in Table 4 for stock issues. It is
immediately apparent that the coefficient on market-to-book is not affected by
the inclusion of the additional variables emphasized by the earlier literature. In
these regressions, though, the stock runup is not significant and the leading
indicators are not significant either. It seems therefore that market-to-book
dominates the variables emphasized in papers that focus on adverse selection.
When we regress the abnormal return on market-to-book and past cumulative
abnormal returns alone, the coefficient on past cumulative abnormal returns is
1.62 with a t-statistic of 1.52, while market-to-book has a aoefficient of
0.93 with a t-statistic of 2.52. The inclusion of market-to-book results in
a substantial weakening of the variables emphasized in papers that focus on
adverse selection.
Is market-to-book successful because it proxies for the firm overvaluation
that underlies the timing model? In regression (8) of Table 4, we include the
long-term cumulative excess return as an explanatory variable. Presumably,
firms that are more overvalued have more negative cumulative excess returns.
The coefficient on long-term cumulative excess return is insignificant. More
importantly, though, the coefficients on the other variables, especially our proxy
for investment opportunities, are not significantly altered. We also estimate
regressions (9) and (10) with the same long-term cumulative excess return as
a dependent variable. The cumulative excess return is never significant. Finally,
we estimate regression (7) using three-year and five-year raw returns, three-year
and five-year net-of-market returns, and three-year size-adjusted returns. Only
one coefficient is significant, but it has the opposite sign from the prediction of
the timing model that investors underreact to the announcement. The coefficient
on five-year raw returns is negative with a t-statistic of – 1.72. If taken
seriously, this estimate implies that the stock price reaction is closer to zero for
firms that underperform more after the issue. None of this evidence is supportive
of the view that the stock price reaction to an equity issue is a fraction of the
long-run cumulative excess returns.
We estimate similar regressions for debt issues, but do not report them here.
The only variable that is ever significant in these regressions is the amount of the
issue divided by the value of common stock, which has a coefficient of – 1.57
and at-statistic of – 1.97. The adjusted R
2 for these regressions is never greater
than zero.
Table 5 shows the abnormal returns for equity issues divided according to the
purpose of the issue. The results provided are consistent with the role of agency
costs in the new issue decision. An equity issue allows firms with poor invest-
ment opportunities to invest in poor projects and/or to reduce the disciplinary
178 K. Jung et al./Journal of Financial Economics 42 (1996) 159-185
Table 4
Cross-sectional regressions of equity issue abnormal returns on firm characteristics
Abnormal returns (ARs) are cumulative abnormal returns for days ( – I, 0), with day 0 the day of the
Wall Street Journal announcement of the security issue. The regression models are estimated using
weighted least squares with the weight for each issue being the inverse of the variance of the market
model residual for the firm issuing the security. The sample includes 189 equity issues from 1977 to
1984. The proceeds of an issue correspond to the gross proceeds in millions of dollars.
Market-to-book is the ratio of firm market value (market value of equity plus book value of total
assets minus book value of equity) to total assets (TA). Cash flow is operating income before
depreciation minus total taxes adjusted for changes in deferred taxes, minus gross interest expense
and minus dividends paid on common and preferred stock. All book values are obtained from
Compustat for the year before the announcement. The leading indicators are the six-month leading
indicators. The volatility of the stock return is computed for the period ( – 240, – 40). The
post-issue cumulative excess returns are five-year size-adjusted returns. T-statistics are given in
parentheses.
Regression (6) (7) (8) (9) (JO)
Intercept – 3.72 -3.86 – 4.15 -4.061 – 3.94
( – 3.64) ( – 3.75) ( – 3.82) ( – 3.77) ( – 2.83)
Market-to-book 0.97 1.01 0.96 0.95 0.97
(2.11) (2.20) (2.00) (2.08) (2.11)
Cash/TA – 6.78 -6.29 -8.27 -7.82 – 6.73
( – 1.63) ( – 1.50) ( – 1.85) ( – 1.82) ( – 1.61)
Tax payments/TA -6.09 -4.65 – 1.94 – 5.41 – 5.68
( – 0.73) ( – 0.55) (-0.22) ( -0.65) ( -0.67)
Long-term debt/TA – 1.14 – 1.51 – l.54 – 1.09 – 1.13
( -0.61) ( -0.80) ( -0.73) (-0.59) (-0.60)
Cash flow 5.09 3.72 5.03
6.17 4.87
(0.89) (0.64) (0.83) (1.06) (0.84)
Stock return volatility – 3.49 -0.55 3.23
-4.76 -2.74
( – 0.49) ( – 0.07) ( -0.39) (-0.66) ( -0.35)
Leading indicators 1.64 1.31
1.18 2.20 1.48
(0.32) (0.25) (0.22) (0.42) (0.28)
Past cumulative excess 1.68 1.79 1.31
1.59 1.71
return (1.53) (1.62)
(1.09) (1.44) (l.54)
Total assets 0.00
(l.07)
Post-Issue cumulative -0.08
excess return ( – 0.70)
Proceeds/Market value of 2.47
common stock (1.00)
Log of proceeds 0.05
(0.24)
Adjusted R1 0.04 0.04 0.04 0.04 0.03
K. Jung et al./Journa/ of Financial Economics 42 (1996) 159-185 179
Table 5
Abnormal returns of equity issues by purpose of issue
Abnormal returns (A Rs) are cumulative abnormal returns for days ( – 1, 0), with day O the d
ay oft he
Wall Street Journal announcement of the security issue. The purpose of the issue is obtai
ned from
the Wall Street Journal announcement. We do not reproduce results for cells smaller th
an 10 or
when the purpose could not be determined unambiguously.
Purpose Number of Abnormal t-statistic
issues return
To repay bank debt 26 -2.93 -4.54
Capital expenditures 40 -3.04 -5.16
To repay long-term debt 20 -4.15 – 6.16
To repay short-term debt 15 – 1.16 – 1.15
Working capital 51 -2.34 -4.43
role of debt. The stock price reactions for firms that plan to use the proceeds for
capital expenditures, firms that plan to replace long-term debt, and firms that
plan to replace bank debt are above the average stock price reaction of the
whole sample. At the 0.10 level, firms that plan to replace long-term debt have
significantly lower abnormal returns than firms that plan to use the proceeds to
replace short-term debt or to invest in working capital; further, at the 0.11 level,
firms that plan to use the proceeds for capital expenditures have significantly
lower abnormal returns than firms that plan to replace short-term debt. The
p-values for the other differences are much higher. We investigate whether there
is a relation between firm type and the abnormal return for a given issuing
purpose. The problem with this investigation is that the cell sizes become small.
Nevertheless, it is interesting that the 11 firms that are not of the equity-issuing
type and plan to use the proceeds for capital expenditures have an average
abnormal return of – 4.43% with at-statistic of – 5.52, whereas the 29 firms of
the equity-issuing type that plan to use the proceeds for capital expenditures
have an average abnormal return of – 2.52% with a t-statistic of – 3.41. The
difference between these two abnormal returns has a t-statistic of 1.75. This
evidence should be treated with caution given the cell sizes, but it nevertheless
provides support for the argument that outsiders view a firm that invests the
proceeds when it is not of the equity-issuing type negatively.
It is often argued that agency costs of managerial discretion are lower for
firms with high managerial ownership because management bears more of the
monetary consequences from pursuing its own objectives. We have managerial
ownership data available from Value Line for 100 equity-issuing firms. For this
smaller sample, we find that when we split the sample into high and low
ownership, the low-ownership sample has a mean abnormal return of – 3.71 %
and the high-ownership sample has a mean abnormal return of – 2.56%. The
difference between the two groups is 1.16% with a t-statistic of 1.72. This
180 K. Jung et aJ./Journa/ of Financial Economics 42 (1
996) /59-185
difference could be size-related since ownership is inversely related t
o size, but
when we split the sample according to firm size, there is no dif
ference in
abnormal returns.
6. Ex post characteristics of firms issuing debt and equity
So far, we have shown that the typical equity-issuing firm has good
invest-
ment opportunities compared with the typical debt-issuing firm, an
d that the
market reaction to an equity issue is positively related to the issu
ing firm’s
investment opportunities. It could be that firms issuing equity w
ith poor
investment opportunities do so because they believe that they are
worth less
than the market’s valuation since they are low market-to-book firms. I
f this were
the case, these firms should invest less than the other equity-issuing
firms. In
contrast, agency considerations predict that these firms issue equity
for invest-
ment purposes even though they have poor investment opportunitie
s.
In this section, we investigate whether the post-issue characteristics
of firms
issuing equity against type resemble those of debt-issuing firms of simi
lar type or
those of equity-issuing firms of different type. We provide this informa
tion for all
firms issuing a type of security and for subsamples of firms that issue a
s expected
and those that do not. To distinguish between firms that are expecte
d to issue
a security and those that are not, we proceed in the same way as discu
ssed at the
end of Section 4 by defining firms predicted to issue equity as all those
firms that
have a probability of issuing equity greater than 0.42 using regres
sion (1) of
Table 2. For each variable, we compute the change in the variable from
the fiscal
year before the issue to the fiscal year after the issue, expressed as a per
ce~tage of
the variable in the fiscal year before the issue. We reproduce the chan
ge m cash
flow and leverage, even though the type of security issued affects thes
e variables
directly, reducing cash flow and increasing leverage for debt-iss
uing firms
compared with equity-issuing firms.
The results of Table 6 are striking. Firms predicted to issue debt that
actually
issue equity invest more than the comparable debt-issuing firms: t
heir plant,
property, and equipment (P P&E), total assets, and capital expenditures a
ll grow
at a significantly higher rate. The differences in growth are economi
cally large:
a firm issuing equity against type has 20% more PP&E at the end of t
he year
following the security issue than a firm expected to issue debt. S
ince both
categories of firms have similar market-to-book ratios, these result
s are fully
consistent with the view that firms that issue equity against the peckin
g order do
so to pursue a more aggressive investment policy that is not in the
interest of
their shareholders. Compared to the firms expected to issue equity
, the firms
that issue equity when expected to issue debt have total assets tha
t grow at
a significantly lower rate, but their PP&E and capital expenditures ha
ve insigni-
ficantly different growth rates than firms that issue equity as expec
ted. EBIT
K. Jung et al./Journal of Financial Economics 42 (1996) I 59-185 181
Table 6
Percentage changes in firm characteristics according to firm type and security type for the three-year
period overlapping the security issue
The sample includes 283 debt issues and 189 equity issues from 1977 to 1984. Cash flow is operating
income before depreciation minus total taxes adjusted for changes in deferred taxes, minus gross
interest expense and minus dividends paid on common and preferred stock. TA denotes the book
value of assets. For each characteristic, we use Com pus tat to compute the percentage increase from
the year before the issue to the year after the issue. High-type firms are those expected to issue equity
based on regression (1) of Table 2.
Bond issues Equity issues Difference
(Number of firms) (Number of firms) (I-statistic)
PP&E 41.83% 68.19% – 26.36%
(267) (178) ( -4.42)
PP&E, low type 38.48% 58.98% -20.50%
(210) (46) ( – 1.77)
PP&E, high type 54.16% 71.39% – 17.24%
(57) (132) ( – 1.73)
Total assets 37.83% 65.60% – 27.77%
(269) (180) ( – 5.79)
Total assets, low type 32.70% 45.82% -13.13%
(211) (46) ( – 2.11)
Total assets, high type 56.50% 72.39% – 15.88%
(58) (134) ( – 1.70)
Net capital expenditures 51.57% 107.50% – 55.93%
(210) (177) ( – 3.57)
Net capital expenditures, 43.12% 93.92% – 50.80%
low type (206) (46) ( – 1.70)
Net capital expenditures, 82.67% 112.27% – 29.60%
high type (56) (131) ( – 1.20)
Long-term debt/TA 36.55% -8.18% 44.73%
(268) (181) ( – 4.52)
Long-term debt/TA, 40.67% – 5.45% 46.13%
low-type firms (210) (46) (3.20)
Long-term debt/TA, 21.65% -9.11% 30.76%
high-type firms (58) (135) (2.33)
Cash flow -17.50% 2.08% – 19.58%
(265) (178) ( – 1.90)
Cash flow, low-type firms – 21.85% 0.96% – 22.81%
(209) (45) ( – 2.71)
Cash flow, high-type firms – 1.25% 2.46% – 3.71%
(45) (133) ( – 0.19)
EBIT 15.45% 53.15% – 37.70%
(266) (181) (- 2.28)
182 K. Jung el al./Journal of Financial Economics 42 (1996) 159-185
Table 6 (continued)
Bond issues Equity issues Difference
(Number of firms) (Number of firms) (I-statistic)
EBJT, low type 6.84% 10.60% – 3.76%
(210) (46) (-0.11)
EBIT, high type 47.72% 67.65% – 19.92%
(56) (135) (- 1.01%)
Change in dividend yield 0.18 -0.10 0.28
(269) (183) (1.78)
Change in dividend yield, -0.02 -0.21 0.19
low type (209) (48) (0.77)
Change in dividend yield, 0.84 -O.Q7 0.91
high type (60) (135) (3.98)
Five-year size-matched 2.03 – 32.69% 34.72
excess return (242) (178) (1.33)
Five-year size-matched 11.04% – 39.20% 50.24%
excess return, high type (56) (135) (0.88)
Five-year size-matched -0.69% – 12.26% 11.58%
excess return, low type (186) (43) (0.31)
increases substantially for the firms expected to issue equity but not for the firms
that issue equity when expected to issue debt. We also report some evidence on
dividend policy. Firms issuing equity have a drop in dividend yield in contrast to
firms issuing debt. Though firms issuing equity against type form the subsample
with the largest drop in dividend yield, the difference between the change in
dividend yield for that subsample and for the subsample of firms issuing debt
according to type is not significant.
We explore the long-term stock performance of the issuing firms according to
their type and find that for all subsamples, equity-issuing firms have mean
cumulative excess returns that are much lower than debt issuing firms, although
the mean differences are not significant. The firm characteristics that proxy for
agency costs are not helpful in explaining the cross-sectional variation in
post-issue cumulative abnormal returns. Such a result is not surprising for those
who believe that markets are efficient. It points towards a risk-based explana-
tion of long-term abnormal returns.
7. Conclusions
In this paper, we investigate the empirical relevance of three explanations of
the security issue decision: the pecking-order model, the agency model, and the
timing model. Our results support the agency model. We show that the typical
K. Jung et al. /Journal of Financial Economics 42 (1996) J 59
– J 85 183
firm. issuing equity has valuable investment opportunities
and experiences
considerable asset growth from the year before the equity issue
to the end of the
year following the issue. Firms with the most valuable investm
ent opportunities
do not experience adverse stock returns when they issue equ
ity. We find that
some firms with poor investment opportunities issue equity
even though the
pecking-order model suggests that they should issue debt to r
aise funds. These
firms, otherwise similar to debt-issuing firms, experience sub
stantially higher
asset growth than debt-issuing firms. However, they regis
ter an extremely
significant drop in their share price when they issue. Though it
is true that these
firms reveal that they are overvalued when they issue, an expla
nation consistent
with this excessive valuation is that, given their investment o
pportunities, the
market did not expect these firms to issue equity and does not
expect the invest-
ments undertaken with the proceeds to increase shareholder w
ealth. The behav-
ior of the firms issuing equity against type is inconsistent with t
he pecking-order
model or asymmetric information models which assume that
managers maxi-
mize shareholder wealth. If the firms that issue against typ
e have valuable
investment opportunities that are not recognized by the financ
ial markets, they
should not be issuing equity since their equity is underprice
d and they could
issue debt. The evidence we present in this paper is also inco
nsistent with the
view that firms time equity issues to take advantage of equi
ty overvaluation
when they know that the firm’s equity will underperform in f
uture years.
An agency approach that emphasizes the costs of manager
ial discretion
provides a consistent framework in which evidence on the is
sue decision, the
stock price reaction, and the post-issue investment policy of th
e issuing firm can
be understood. In contrast to the agency model, models base
d on information
asymmetries alone are at best consistent only with our evide
nce on the stock
price reaction, and the timing model receives almost no suppo
rt in our sample.
References
Asquith, Paul and David W. Mullins, Jr., 1986, Equity issue
s and offering dilution, Journal of
Financial Economics 15, 31-60.
Barclay, Michael J. and Robert H. Litzenberger, 1988, Announcem
ent effects of new equity issues
and the use of intraday price data, Journal of Financial Econo
mics 21, 71-99.
Baxter, Nevins D. and John G. Cragg, 1970, Corporate choice
among long-term financing instru-
ments, Review of Economics and Statistics 52, 225-235.
Bayless, Mark and Suzanne Chaplinsky, 1991, Expectations o
f security type and the information
content of debt and equity offers, Journal of Financial Interm
ediation 3, 195-214.
Bemanke, Ben, Mark Gertler, and Simon Gilchrist, 1993, The f
inancial accelerator and the flight to
quality, Working paper (Princeton University, Princeton, NJ).
Brav, Alon, Christopher Geczy, and Paul A. Gompers, 1994,
The long-run underperformance of
seasoned equity offerings revisited, Working paper (University
of Chicago, Chicago, IL).
Brennan, Michael and Alan Kraus, 1987, Efficient financing und
er asymmetric information, Journal
of Finance 42, 1225-1243.
184 K. Jung et a/./Journal of Financial Economics 42 (1996) 159-185
Cheng, Li-Lan, 1994, Equity issue under-performance and the timing of security issues, Working
paper (Massachusetts Institute of Technology, Cambridge, MA).
Choe, Hyuk, Ronald Masulis, and Vik Nanda, 1993, Common stock offerings across the business
cycle: Theory and evidence, Journal of Empirical Finance I, 3-3 t.
Cooney, John W., Jr. and Avner Kalay, 1993, Positive information from equity issue announce-
ments, Journal of Financial Economics 33, 149-172.
Denis, David J., 1994, Investment opportunities and the market reaction to equity offerings, Journal
of Financial and Quantitative Analysis 29, 159-177.
Dybvig, Philip H. and Jaime F. Zender, 1991, Capital structure and dividend irrelevance with
asymmetric inforrnation, Review of Financial Studies 4, 201-220.
Eckbo, Bjorn E.,1986, Valuation effects of corporate debt offerings, Journal of Financial Economics
15, 119-151.
Harris, Milton and Artur Raviv, 1991, The theory of capital structure, Journal of Finance 46,
297-356.
Hoshi, Takeo, Anyl Kashyap, and David Scharfstein, 1993, The choice between public and private
debt: An analysis of post-deregulation corporate financing in Japan, Working paper (Mass-
achusetts Institute of Technology, Cambridge, MA).
Jensen, Michael C., 1986, Agency costs of free cash flow, corporate finance and takeovers, American
Economic Review 76, 323-329.
Jensen, Michael C. and William H. Meckling, 1976, Theory of the firrn: Managerial behavior, agency
costs and ownership structure, Journal of Financial Economics 3, 305-360.
Korajczyk, Robert, Debbie J. Lucas, and Robert L. McDonald, 1991, The effect of information
releases on the pricing and timing of security issues, Review of Financial Studies 4, 685-708.
Kothari, S.P. and Jerrold B. Warner, 1995, Measuring long-horizon security price performance,
Working paper (University of Rochester, Rochester, NY).
Lang, Larry, Eli Ofek, and Rene M. Stulz, 1996, Leverage, investment, and firm growth, Journal of
Financial Economics 40, 3-30.
Loughran, Timothy and Jay R. Ritter, 1995, The new issue puzzle, Journal of Finance 50, 23-52.
Lucas, Debbie J. and Robert L. McDonald, 1990, Equity issues and stock price dynamics, Journal of
Finance 45, 1019-1043.
MacKie-Mason, Jeffrey K., 1990, Do taxes affect corporate financing decisions?, Journal of Finance
45,1471-1495.
Marsh, Paul, 1982, The choice between equity and debt, Journal of Finance 37, 121-144.
Masulis, Ronald, 1988, The debt/equity choice (Ballinger Publishing Company, Lexington, MA).
Masulis, Ronald W. and A.N. Korwar, 1986, Seasoned equity offerings: An empirical investigation,
Journal of Financial Economics 15, 91-118.
Mikkelson, Wayne and Megan Partch, 1986, Valuation effects of security offerings and the issuance
process, Journal of Financial Economics 15, 31-60.
Myers, Stewart, 1977, Determinants of corporate borrowing, Journal of Financial Economics 5,
147-175.
Myers, Stewart, 1984, The capital structure puzzle, Journal of Finance 39, 575-592. . .
Myers, Stewart and Nicholas Majluf, 1984, Corporate financing and investment deets1ons when
firms have information that investors do not have, Journal of Financial Economics 13, 187-221.
Noe, Thomas, 1988, Capital structure and signaling equilibria, Review of Financial Studies 1,
331-356.
Pilotte, Eugene, 1992, Growth opportunities and the stock price response to new financing, Journal
of Business 65, 371-395.
Smith, Qifford W. and Ross L. Watts, 1992, The investment opportunity set and corporate
financing, dividend, and compensation policies, Journal of Financial Economics 32, 263-292.
Spiess, D. Katherine and John Affleck-Graves, 1995, Underperformance in long-run stock returns
following seasoned equity offerings, Journal or Financial Economics 38, 243-267.
K. Jung el al./JournaJ of Financial Economics 42 (1996) 159-185 185
Stein, Jeremy C., 1995, Rational capital budgeting in an irrational world, Working paper (Mass-
achusetts Institute of Technology, Cambridge, MA).
Stulz, Rene M., 1988, Managerial control of voting rights: Financing policies and the market for
corporate control, Journal of Financial Economics 20, 25-54.
Stulz, Rene M., 1990, Managerial discretion and optimal financing policies, Journal of Financial
Economics 26, 3-27.
Titman, Sheridan and Roberto Wessels, 1988, The determinants of capital structure choice, Journal
of Finance 43, 1-20.
Viswanathan, P.V., 1993, Strategic considerations, the pecking order hypothesis, and market
reactions to equity financing, Journal of Financial and Quantitative Analysis 28, 213-234.
Zwiebel, Jeffrey, 1994, Dynamic capital structure under managerial entrenchment, Working paper
(Stanford University, Palo Alto, CA).
- 159-185_Page_01
- 159-185_Page_02
- 159-185_Page_03
- 159-185_Page_04
- 159-185_Page_05
- 159-185_Page_06
- 159-185_Page_07
- 159-185_Page_08
- 159-185_Page_09
- 159-185_Page_10
- 159-185_Page_11
- 159-185_Page_12
- 159-185_Page_13
- 159-185_Page_14
- 159-185_Page_15
- 159-185_Page_16
- 159-185_Page_17
- 159-185_Page_18
- 159-185_Page_19
- 159-185_Page_20
- 159-185_Page_21
- 159-185_Page_22
- 159-185_Page_23
- 159-185_Page_24
- 159-185_Page_25
- 159-185_Page_26
- 159-185_Page_27
ARTICLE IN PRESS
Journal of Financial Economics 84 (2007) 266–298
0304-405X/
$
doi:10.1016/j
$
This rese
Business Sch
o
Almeida, Luc
Friedman, M
Michael Ostr
at the Univer
Duke Univer
McGill Unive
Toronto, the
National Bur
�
Correspo
E-mail ad
www.elsevier.com/locate/jfec
Corporate financing decisions when investors take
the path of least resistance
$
Malcolm Baker
a,
b
, Joshua Coval
a,b
, Jeremy C. Stein
b,c,�
a
Harvard Business School, Boston, MA 02163, USA
b
National Bureau of Economic Research, Cambridge, MA 02138, USA
c
Harvard Economics Department, Harvard University, Cambridge, MA 02138, USA
Received 6 April 2005; received in revised form 13 February 2006; accepted 17 March 2006
Available online 23 January 2007
Abstrac
t
We argue that inertial behavior on the part of investors can have significant consequences for
corporate financial policy. One implication of investor inertia is that it improves the terms for the
acquiring firm in a stock-for-stock merger, because acquirer shares are placed in the hands of
investors, who, independent of their beliefs, do not resell these shares on the open market. In the
presence of a downward-sloping demand curve, this leads to a reduction in price pressure and, hence,
to cheaper equity financing. We develop a simple model to illustrate this idea and present supporting
empirical evidence.
r 2006 Elsevier B.V. All rights reserved.
JEL classification: G32; G34
Keywords: Mergers; Inertia; Equity issuance
– see front matter r 2006 Elsevier B.V. All rights reserved.
.jfineco.2006.03.005
arch is supported by the National Science Foundation and the Division of Research at Harvard
ol. Thanks to Lauren Cohen for supplying the data on insider ownership and to the referees, Heitor
ian Bebchuk, Jon Bernstein, Rob Daines, Stefano DellaVigna, Peter DeMarzo, Darrell Duffie, John
ark Garmaise, Peter Henry, David Laibson, Ulrike Malmendier, Rich Matthews, Stefan Nagel,
ovsky, Jay Ritter, Andrei Shleifer, Ilya Strebulaev, Ivo Welch, Jeff Zwiebel, and seminar participants
sity of California at Berkeley, Brown University, the University of Chicago, Columbia University,
sity, the Federal Reserve Bank of New York, Harvard Business School, the University of Illinois,
rsity, the University of Notre Dame, Stanford University, the University of Texas, the University of
University of Southern California conference on financial economics and accounting, and the
eau of Economic Research for helpful comments.
nding author.
dress: jeremy_stein@harvard.edu (J.C. Stein).
www.elsevier.com/locate/jfec
dx.doi.org/10.1016/j.jfineco.2006.03.005
mailto:jeremy_stein@harvard.edu
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 267
1. Introduction
Much of finance theory rests on the assumption that investors continuously monitor
their portfolios and condition their investment decisions on the most recently available
information. Even in models with transaction costs (e.g., Constantinides, 1986) or
behavioral biases (e.g., Barberis, Shleifer, and Vishny, 1998; Daniel, Hirshleifer, and
Subrahmanyam, 1998; Hong and Stein, 1999), where trade need not be continuous and
updating need not be fully rational, investors still can be thought of as processing new
information and reevaluating the decision of whether or not to trade on a constant basis.
While this assumption is convenient for modeling purposes, it is also unrealistic. A large
body of existing evidence suggests that people often behave in a way that might be
characterized as inertial, or as taking the path of least resistance. Inertial behavior can arise
from a variety of sources, including endowment effects (Thaler, 1980; Kahneman,
Knetsch, and Thaler, 1990, 1991), a tendency to procrastinate in decision making (Akerlof,
1991; O’Donoghue and Rabin, 1999), and the cognitive fixed costs associated with
reevaluating and re-optimizing an existing portfolio.
In this paper, we argue that investor inertia can exert a significant influence on financial
market outcomes. Our particular focus is on the consequences of inertia for mergers, and
the main idea can be illustrated with a simple example. Consider a firm A that intends to
acquire another firm T via a stock-for-stock merger, and suppose that the following two
conditions hold. First, there is a downward-sloping demand curve for firm A’s shares. This
downward-sloping demand curve arises not from asymmetric information, but from
irreducible differences of opinion among investors as to the value of A’s preexisting assets.
Second, and crucially, some investors in the target firm T are inertial in the following sense:
They will not make the active decision to buy shares in A in, say, a seasoned equity offering
(SEO). But, if they are granted these shares in a stock-for-stock merger, they will also not
make the active decision to sell them.
Under these conditions, an increase in the fraction of inertial investors makes the stock-
for-stock merger more attractive to firm A. Greater target-firm inertia means that more
firm A shares are simply absorbed by the current T investors and thus are not ever floated
on the open market. With a downward-sloping demand curve for firm A shares, this
implies a smaller negative price impact, which means that firm A does not have to give up
as many new shares in the merger. Said differently, a stock-for-stock merger changes the
default setting for inertial T investors relative to an SEO. It makes the default one in which
they are holders of A shares, which can be thought of as pushing out the overall demand
curve for firm A stock.
1
After fleshing out our idea with the aid of a simple model in Section 2, we examine some
of its empirical implications in Section 3. We begin by verifying that our premise of
investor inertia is relevant in the context of mergers. Using data on both individuals and
institutions, we look at investors’ propensity to hold on to shares that they are granted in
stock-for-stock mergers. We focus on situations in which a given investor in the target
1
Madrian and Shea (2001) and Choi, Laibson, Madrian, and Metrick (2002, 2004) demonstrate just how
powerful the effect of defaults can be in the context of retirement savings decisions. To take just one example,
when firms set the default in their 401(k) plans to automatic enrollment, few workers choose to opt out, resulting
in participation rates close to 100%. In contrast, if the default is no enrollment, so that a worker has to make an
active decision to participate in the plan, participation rates are generally much lower. In a related corporate
finance paper, Zhang (2004) argues that the endowment effect can explain initial public offering underpricing.
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298268
owns none of the acquirer before the deal, so that it can be inferred that he does not have a
high valuation for the acquirer. (The conceptually cleanest case is one in which the acquirer
is very large relative to the target, so the post-merger combined company is composed
almost entirely of acquirer-firm assets.) Even so, target investors have a remarkably high
likelihood of owning acquirer shares after the merger transaction closes. We estimate that
roughly 80% of individuals behave as sleepers, and simply accept shares they are given in a
merger. For institutions, the estimated fraction of sleepers is significantly smaller, at
around 30%, but still noteworthy.
Next, we test one of the theory’s central predictions. Given that institutional investors
are less prone to inertia than individuals, our model implies that the announcement return
to the acquirer in a stock-for-stock deal is more negative, all else equal, if the target firm
has a higher proportion of institutional shareholders. This is because institutions are
expected to dump more of the acquirer-firm shares they receive back onto the market.
Individuals, by contrast, tend to hang on to these shares, thereby mitigating price impact.
Using a variety of specifications, we find robust evidence for this hypothesis. We also
provide another clue that these return effects are the result of price pressure, as our model
suggests: The acquirer has more trading volume around the announcement date when the
target has a higher proportion of institutional shareholders.
To rule out alternative explanations, we verify some finer predictions of the model. Both
acquirer return and volume effects are largest when the overlap between target and
acquirer institutional ownership is small. Intuitively, non-overlapping owners of the target
are the ones most likely to unload their shares on announcement of a merger, as they have
demonstrated a lack of interest in holding acquirer assets. The results are also stronger
when various proxies suggest that the acquirer’s demand curve is steep. Finally, we show
that, consistent with our model, each of the above results is present only in stock-swap
mergers, and not in cash deals.
On a more speculative note, in Section 4 we make two empirical connections between our
theory and corporate financing decisions. The first has to do with the means of payment in a
merger. In a more general version of the model, acquiring firms prefer to use stock as opposed
to cash as consideration when the inertia of target shareholders is high. Consistent with this
prediction, we find a negative relationship between target institutional ownership and the
probability that a merger is conducted with stock. Like our other results, this effect is most
pronounced when the overlap between target and acquirer institutional ownership is small.
A second connection is with recent empirical work by Fama and French (2005). They
show that, although SEOs are relatively rare, total external equity financing (which, in
addition to SEOs, can come in the form of stock-for-stock mergers or stock-based employee
compensation) is substantial for most classes of firms. For example, over the period from
1999 through 2001, Fama and French find that Standard & Poor’s (S&P) 100 firms raised
an annual average of only 0.09% of assets via SEOs, but 1.05% via various forms of stock-
based compensation and 3.68% via mergers. In other words, the volume of equity finance
raised in mergers by these large firms was roughly 40 times that raised in SEOs.
As Fama and French (2005) point out, these stylized facts are not easily reconciled with
standard corporate finance theories, such as the asymmetric information-based approach
of Myers and Majluf (1984). Myers and Majluf have a good story for the relative scarcity
of SEOs taken in isolation, but they have little to say about why mergers would be a
dominant substitute. A direct application of Myers and Majluf logic would seem to imply
that stock-for-stock mergers face the same asymmetric-information problems as SEOs.
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 269
In contrast, our theory suggests an affirmative rationale for the primacy of stock-for-
stock mergers as compared with SEOs.
2
To be specific, imagine a firm with an exogenously
specified growth strategy that, over the next year, involves one acquisition, and one major
new greenfield investment (e.g., construction of a new plant). Suppose further that, because
of optimal capital-structure considerations, one of these two transactions needs to be
financed with an equity issue; i.e., either the merger has to be stock-for-stock or the
greenfield investment has to be accompanied by an SEO. Which outcome is more likely?
Because the SEO effectively amounts to a limiting case of our model with no investor
inertia, it is associated with a more negative price impact, all else equal, and hence tends to
be less attractive. Thus we would expect the firm to finance the merger with a stock swap,
but to finance the greenfield investment with cash. This is an outcome very much in the
spirit of Fama and French (2005).
2. The model
2.1. Investor beliefs
The model has three dates, labeled 0, 1, and 2. The focus is on the behavior of a potential
acquirer firm A, which is faced with an investment decision at time 1. As of time 0,
however, the prospect of investment is unanticipated by the market, so A’s stock is priced
solely on the basis of cash flows from assets already in place. These assets in place yield a
liquidating dividend of D at time 2.
Our first assumption is that there are differences of opinion among investors in firm A as
to the expected value of D. In particular, there is a continuum of A-specialists who have
values of E(D) uniformly distributed on the interval [F, F+H], where the parameter H can
be interpreted as a measure of the divergence of opinion. And while they are risk-neutral,
the A-specialists are constrained to not invest more than their total wealth of W. To ensure
the existence of interior solutions in what follows, we stipulate that W4F. Finally,
maintaining a short position over the interval from time 0 to time 2 is assumed to be
impossible.
3
Taken together, these assumptions have the effect of creating a downward-
sloping demand curve for firm A assets. However, we should emphasize that any set of
assumptions that produces downward-sloping demand is sufficient for our purposes.
4
For
example, we could alternatively allow short-selling but make all investors risk-averse.
2
Our basic line of reasoning suggests that stock-based employee compensation could also be preferred to SEOs.
If workers are subject to an endowment effect (so that, once granted stock, they are reluctant to sell it, even if they
would not have gone out and bought it on their own in the first place), a firm facing a downward-sloping demand
curve prefers to place stock with them than to sell it on the open market. This observation could help to resolve
the puzzle of why firms give stock to low-level employees, when incentive effects are likely to be minimal. See, e.g.,
Bergman and Jenter (2006) and Oyer (
2004).
3
Miller (1977) was the first to model the combined effects of differences of opinion and short-sales constraints.
Recent treatments include Chen, Hong, and Stein (2002), Scheinkman and Xiong (2003), and Hong, Scheinkman,
and Xiong (2006).
4
The empirical literature provides clear support for the premise of downward-sloping demand. Bagwell (1992)
and Hodrick (1999) look at demand curves in the context of Dutch auction share repurchases. Shleifer (1986),
Harris and Gurel (1986), and more recently Kaul, Mehrotra, and Morck (2000), Wurgler and Zhuravskaya
(2002), and Greenwood (2005) illustrate the price impact of uninformed demand, by examining index inclusion
and rebalancing decisions. Mitchell, Pulvino, and Stafford (2004) focus on price pressure in the context of mergers
and acquisitions.
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298270
Given the demand curve, the market value of the firm at time 0, P0, is determined by
setting P0 equal to the total wealth of those A-specialists with valuations in excess of P0. In
other words, the value of the firm is equal to the wealth of those investors who are buyers
in equilibrium. This condition is equivalent to
P0 ¼
W
H
F þ H � P0ð Þ or P0 ¼ðF þ HÞ
W
WþH
. (1)
From Eq. (1), along with our assumption that W4F, it follows that P0 always lies between
F and (F+H). The fraction of investors who are long the stock in equilibrium is given by
ðF þ HÞ=ðW þ HÞ. Also, we have the intuitive properties that dP0=dW 40 and
dP0=dH40. The latter is just a version of the Miller (1977) insight that, in the presence
of a short-sales constraint, prices are increasing in the heterogeneity of investor opinion.
To see the import of the downward-sloping demand curve, observe that dP0=dF ¼
W=ðW þ HÞo1. This means that, if W is held constant, the firm’s market value does not
go up one-for-one with an increase in expected cash flows. The intuition is that, as the firm
gets larger, shares must be absorbed by investors who are less optimistic. Moreover,
d
2
P
0
�
dFdHo0, so that an increase in heterogeneity amplifies the downward slope of the
demand curve.
Example. Suppose F ¼ 100, H ¼ 100, and W ¼ 300. Then firm A has a market value of
P0 ¼ 150 at time 0. The more optimistic half of the A-specialists (those with valuations
between 150 and 200) own all the shares, while the remaining half of the A-specialists
(those with valuations between 100 and 150) sit out of the market.
2.2. Stock-for-stock merger when all target shareholders are awake
At time 1, the manager of firm A announces that he has decided to acquire a target firm
T and to finance the acquisition with an equity issue. The purchase price of the target is K,
and the merger increases firm A’s terminal dividend by R4K, which implies that the
merger is positive-NPV (net present value) for firm A.
5
For simplicity, we assume there is
no disagreement among the A-specialists as to the value added by the acquisition, so that
once it is on the books, their expectations of terminal cash flow are uniformly distributed
on the interval [F+R, F+R+H].
We assume that none of the target shareholders is among the group of A-specialists.
That is, as of time 0, their expectations of firm A’s terminal dividend are relatively low.
Without loss of generality one can think of all of them as simply having E(D) ¼ F.
Empirically, this implies that, prior to the merger announcement at time 1, there is no
overlap between the investors in firms A and T. Nothing substantive changes if we allow
for partial overlap.
In a stock-for-stock merger, shares in the newly created merged firm M are placed
directly into the hands of firm T’s investors. Based on their pessimism about the prospects
for firm A’s preexisting assets, these T-investors also have low valuations of M [they
consider it to be worth only (F+R)] as compared with the pool of A-specialists, whose
valuations of M all exceed (F+R). Thus one might expect that the T-investors would
5
Although we assume that the A-manager is interested in the merger simply because it represents a positive-
NPV investment, our results would be similar if instead we assumed that the A-manager was motivated by either a
desire to increase the size of his empire or a belief that his stock was overvalued, as in Stein (1996) or Shleifer and
Vishny (2003), for example.
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 271
immediately take any shares in M that they receive in a stock-for-stock exchange and
sell them off in the open market, where these shares would have to be absorbed by
A-specialists.
This logic is correct, if all T-investors are aware of the merger and actively respond to it
in a fashion consistent with their low valuations. In this case, the stock-for-stock merger is
identical to the following decoupled transaction: Shares in the merged firm M are sold off
in an SEO (and bought by A-specialists) and the proceeds of this SEO are used to pay the
T-investors in cash for their shares. In either case, the absence of inertia implies that any
shares in M must ultimately wind up in the hands of those with the highest valuations for
M, and the T-investors are not among this group.
To finance the merger, new shares must be issued to raise an amount K. For the case in
which all T-investors are responsive, denote the post-merger market value of the firm at
time 1 by P1
R
. Any A-specialists who were long at time 0 cannot participate in the new
issue, because they already have all of their wealth invested in firm A shares. Thus the new
shares must be absorbed by those A-specialists who were previously on the sidelines. This
group has total wealth of (W – P0), and has valuations distributed uniformly on the
interval [F+R, P0+R].
6
Because the market value of the shares they absorb must equal K,
equilibrium requires that
K ¼ W�P0
P0�F
P0 þ R � P
R
1
� �
, (2)
which can be rewritten as
PR1 ¼ P0 þ R �
H
W
K. (3)
P1
R
is necessarily less than P0+R. The market value does not go up by the full amount of
the added cash flows R from the firm-T assets. In other words, the equity issue has a price-
pressure effect, the magnitude of which is increasing in the heterogeneity parameter H.
This reflects the fact that the issue must be absorbed by the relatively less optimistic A-
specialists, who were sitting out of the market prior to the issue.
Example. (continued) Keep F ¼ 100, H ¼ 100, and W ¼ 300. Suppose further that
K ¼ 100, and R ¼ 110. This yields P1
R
¼ 226.67, so that a 44.12% share in the firm is
issued in the merger (because 0.4412�226.67 ¼ 100). The market value of the stake held
by preexisting firm A shareholders drops from 150 to 126.67. Thus the merger is
accompanied by a negative price impact of 15.56%. After the merger, 83.33% of the A-
specialists have long positions.
2.3. Stock-for-stock merger when some target shareholders are asleep
To explore the consequences of investor inertia, we now posit that only a fraction ao1
of the T-investors are awake. As before, these awake T-investors sell off any shares in the
merged firm M that they receive in the merger. The remaining (1– a) of the T-investors
are asleep and simply hold on to the shares in M that they are given. Thus for the sleeping
T-investors, the default condition matters. They would not actively seek to buy shares
offered in an SEO, but they also do not actively seek to sell shares that are granted to them
6
For the equity issue to raise K from the pool of A-specialists, it must be that the total wealth of the previously
sidelined A-specialists, (W – P0), weakly exceeds K. This condition can be expressed as W(W – F)/(W + H) Z K,
and we assume it holds in what follows.
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298272
as part of a stock-for-stock merger. Or said differently, the sleeping T-investors always
take the path of least resistance, which is simply to do nothing.
While the inertial behavior of the sleeping T-investors clearly differs from that of
investors in standard finance models, for our purposes one need not interpret this inertia as
a manifestation of fundamentally irrational behavior. Perhaps some investors, most likely
individuals, view it as costly to keep track of the stocks in their portfolio on a constant
basis. Or even if they are aware of the merger, the benefits of rebalancing could be
outweighed by the search and transactions costs associated with finding a new stock to
buy, especially for those investors who have relatively small positions in the target. We find
some evidence that those individual investors with the smallest dollar positions in the
target behave in an especially inertial fashion, consistent with this meta-rational
hypothesis.
In the presence of sleepers, we assume that the awake T-investors are the only ones who
actively evaluate the bid from firm A. In doing so, they continue to place a reservation
value of K on their firm’s assets and recognize that they will immediately resell all shares
that they receive in the stock swap. If we denote by P1
S
the post-announcement market
value of the merged firm when there are sleepers, then the requirement that the bid be
satisfactory to the awake T-investors amounts to saying that the total value of shares
issued in the merger, evaluated at P1
S
, be equal to K. This condition is exactly the same as
the analogous one in the fully awake case.
Where things differ is in the determination of P1
S
. As a result of the sleeping T-investors
who hang on to their shares, only a fraction a of the shares issued in the merger ever comes
on the market, and hence only this fraction must be absorbed by the pool of previously
sidelined A-specialists. Thus Eqs. (2) and (3) are modified as follows:
aK ¼ W�P0
P0�F
P0 þ R � P
S
1
� �
, (4)
PS1 ¼ P0 þ R �
H
W
aK. (5)
Example. (continued) As before, keep F ¼ 100, H ¼ 100, W ¼ 300, K ¼ 100, and R ¼ 110.
Assume that the fraction of sleepers among T-investors is given by a ¼ 0.50. These
parameters yield P1
S
¼ 243.31, which implies that a 41.10% share of the firm is issued in the
merger, less than the 44.12% share issued in the case with fully awake investors. (The value
of the merger bid, evaluated at market prices, is 0.411�243.31 ¼ 100.) The market value
of the stake held by preexisting firm A shareholders drops from 150 to 143.31: The merger
is accompanied by a negative price impact of 4.46%. After the merger, 66.67% of the
A-specialists have long positions with an aggregate market value of 193.31, and sleeping
T-investors have a long position with a market value of 50.
2.4. Empirical implications
The model’s most direct empirical implications can be summarized as follows.
Proposition 1. (Merger announcement effects) (i) All else equal, an increase in the fraction a
of awake target-firm investors strengthens the adverse impact of a stock-for-stock merger
announcement on the price of the bidding firm. (ii) An increase in the slope of the acquirer’s
demand curve, as measured by the degree of investor heterogeneity H, has a similar effect.
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 273
(iii) There is an interaction between these two variables: A steeper acquirer demand curve
amplifies the negative stock-price consequences of awake target-firm investors. Thus,
denoting the announcement price impact by DP, (i) dDP=dao0; (ii) dDP=dHo0; and (iii)
d
2DP
�
dadHo0.
Proposition 1 follows immediately from inspection of Eq. (5). Part (i) of the proposition
forms the basis for one of our main empirical tests below. Part (ii), which does not involve
the wakefulness parameter a, and which holds even in a world with no inertia, is the subject
of a recent paper by Moeller, Schlingemann, and Stulz (2004). They use the dispersion of
analyst forecasts as a proxy for H and find evidence consistent with the hypothesis that
dDP=dHo0. Taking a similar approach to measuring H, we also attempt to test part (iii)
of the proposition. Thus, consistent with our theoretical emphasis, our empirical work
centers on those effects that are most directly related to the wakefulness parameter a.
2.5. Merger arbitrage
In the model, the announcement and completion of a merger occur simultaneously, at
time 1. More realistically, there can be a substantial time lag between announcement and
completion, and completion may not be a sure thing when the deal is first announced. It is
easy to extend the model to incorporate these features. Proposition 1 above is unchanged,
though the model now also admits a role for merger arbitrageurs.
To see how such arbitrage might work, imagine that the announcement of a merger
occurs at time 1, but the transaction is not completed until time 11
2
. Moreover, because
completion is not ensured as of time 1, the stocks of firm T and firm A are not
interchangeable immediately post-announcement: They are no longer certain to both turn
into claims on the merged firm M. In this setting, it is possible that awake T-investors want
to sell their shares in T immediately, as of time 1. But the previously sidelined A-specialists
could prefer to buy A shares when their price falls at this time, instead of buying T shares.
This is because their primitive preference is for A assets, and there is a risk that the T shares
do not turn into a claim on any A assets, if the deal falls through. The arbitrageurs can
bridge this gap by buying T shares from the T-investors and for each share bought, shor
t-
selling K/(P1
S
– K) A shares to the A-specialists.
7
If the deal does go through, each side of
the arbitrage trade converts into the same number of shares in the merged firm M. If not,
arbitrageurs are left with an unhedged position in T and A shares.
As long as completion risk is small, the results for prices at time 1 remain approximately
the same as described. That is, the adverse impact to A’s stock price occurs primarily on
announcement of the deal, not on completion, and continues to be a function of the
number of sleepers among the T-investors. Intuitively, the more sleepers there are, the
fewer shares of T are unloaded onto the arbitrageurs at time 1, and hence the fewer shares
of A are short-sold by these arbitrageurs into the downward-sloping demand curve of the
A-specialists.
7
This calculation assumes that both firms have the same number of shares outstanding. We are now allowing
arbitrageurs to hold short positions between time 1 and time 1
1
2
even though we previously ruled out short-selling
by investors over the longer interval from time 1 to time 2. A loose rationalization might be that those players who
have the ability to short-sell are unwilling to take long-horizon unhedged short positions because of fundamental
risk. Alternatively, it is possible to reformulate the model so that downward-sloping demand curves arise from
other sources (e.g., risk aversion) and nobody faces any short-sales constraints.
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298274
The one thing we gain by explicitly considering the process of merger arbitrage is a more
precise set of predictions about trading volume as of the announcement date.
Proposition 2. (Abnormal trading volume around merger announcements) (i) All else
equal, an increase in the fraction a of awake target-firm investors leads both to more trading
volume in the target around the announcement date and (ii) to more trading volume in the
acquirer.
Part (i) of the proposition is self-evident: If all target-firm shareholders are asleep, none
of them sells on announcement of the deal, and there is no trading volume in the target.
Part (ii) is a bit more subtle and relies on the merger-arbitrage mechanism: The more
shares are dumped by awake target-firm shareholders, the more the arbitrageurs have to
step in and buy, and hence the more short-selling of the acquirer they end up doing to
hedge their positions.
8
These predictions for volume are a useful complement to our
predictions for acquirer stock returns in Proposition 1. If both Propositions 1 and 2 are
borne out in the data, it becomes more likely that the results for stock returns are driven by
the sort of price-pressure effects envisioned in our model, as opposed to some other
confounding factor.
Our account of the price-pressure consequences of merger arbitrage is complementary
to that of Mitchell, Pulvino, and Stafford (2004). Loosely speaking, these authors take
the volume of risk arbitrage selling in the acquirer as exogenous and show that this
selling has a negative impact on the acquirer’s announcement returns. Our model
endogenizes the level of risk arbitrage activity, tracing it back to the fraction of awake
T-investors.
2.6. Capital gains taxes
We have not yet addressed the following question: To what extent can what we call
investor inertia, or sleepiness, be thought of as simply a rational reluctance on the part of
target-firm shareholders to incur capital gains taxes by selling their shares? This question
can be addressed both theoretically and empirically. On the empirical side, our analysis in
Section 3.2.3 below demonstrates conclusively that inertia is not primarily driven by tax
considerations. In particular, we find that sleepiness is not any greater among individuals
with taxable accounts than among those with non-taxable accounts.
Even holding this empirical finding aside, it is not at all obvious that, as a matter of
theory, one can generate the sorts of patterns in acquirer announcement returns predicted
by Proposition 1 using capital gains tax effects alone. To see why, consider the simple
limiting case of a target firm in which all investors are heavily taxed individuals. It could be
true that these individuals are reluctant to sell any shares in the acquirer that they get as
part of a merger. But if they are fully rational, and understand this ex ante, they demand a
greater number of shares in the first place, to compensate them for the fact that they are
going to be forced to hold stock that they do not value highly. So on net, the acquirer is
made worse off as a result of the tax-related friction. In this simple example, as long as
8
An important caveat is that this short-selling occurs only if the terms of the merger involve an exchange ratio
that is fixed as of the announcement date. If, instead, the dollar value of the bid is fixed, and the exchange ratio is
left to float until completion, the arbitrageurs do not want to short the acquirer. See Mitchell, Pulvino, and
Stafford (2004) for an analysis of this issue.
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 275
everybody is rational, the acquirer must bear the cost of inefficiently placing shares in the
wrong hands.
Example. (continued) As before, keep F ¼ 100, H ¼ 100, W ¼ 300, K ¼ 100, and R ¼ 110.
Consider two cases. In the first, all target shareholders are awake, and there are no taxes.
This implies that 44.12% of the firm’s shares must be issued to raise 100. In the second
case, all target shareholders are again awake, but there are prohibitive capital gains taxes,
so that target shareholders are forced to hang on to any acquirer shares they receive in a
merger. Target shareholders value the combined post-merger firm at 210 and have a
reservation price of 100. So to get them to sell, they must be given 47.62% of the combined
firm (100/210 ¼ 0.4762), which is more than in the case with no taxes.
Thus, in this example, investor inertia delivers something that capital gains taxes alone
cannot. When it is granted to them as consideration in a merger, not only are inertial target
investors more likely to hang on to a stock that they would otherwise never have bought,
but also, crucially, they do not have to be compensated for doing so. This latter feature is
why acquirers can benefit from target-investor inertia, as in Proposition 1.
However, the above example is not the only relevant case. Consider an alternative with
some heterogeneity across investors in terms of their tax situations. Suppose 60% of target
investors are tax-exempt institutions, and 40% are taxable individuals. In this case, it could
be that only the tax-exempt institutions are pivotal in setting the terms of the deal. If so,
the acquiring firm does not have to compensate the individual shareholders of the target
for giving them a stock they do not want. In other words, once one introduces
heterogeneity, capital gains taxes can generate an effect that is isomorphic to the concept
of inertia in our model, though this depends on parameter values, e.g., on the proportion
of taxable versus non-taxable investors.
Again, however, the most clear-cut answer to the tax question is the empirical one. The
evidence strongly suggests that the inertia that we find among individual investors is
unrelated to tax considerations.
3. Empirical analysis
Our empirical work is divided into two parts. First, we simply show that the key pre-
mise of our model holds in the data; that is, target investors behave in an inertial
fashion around merger transactions. While we find that both individual and institutional
investors exhibit inertia, it is substantially more pronounced among individuals. Second,
using this distinction between individuals and institutions to create a proxy for the extent
of investor inertia, we test the cross-sectional implications of the model summarized in
Propositions 1 and 2. We show that acquirer announcement returns are more negative in
transactions in which the target firm has a greater proportion of institutional (i.e., awake)
shareholders and that acquirer abnormal trading volume is also greater around such
transactions.
3.1. Data
Our sample of mergers has 2,995 successfully completed transactions announced
between the second quarter of 1980 and the fourth quarter of 2000. Of these, most of our
analysis focuses on the subset of 1,851 stock-swap deals; the remaining cash deals are used
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298276
only as a control sample.
9
We require that each deal involves a public acquirer, with a
matched announcement return available from the Center for Research in Security Prices
(CRSP). Information on individual investor holdings comes from the records of a discount
brokerage firm (this is the Barber and Odean, 2000, data) and is available only from 1991
to 1996. We also make use of a variety of CRSP and Compustat variables, as well as
analyst forecast data from Institutional Brokers’ Estimate System (IBES).
Quarterly observations on institutional ownership come from the CDA/Spectrum
Institutional Holdings database, over the period from the first quarter of 1980 through the
fourth quarter of 2002. These data, which come from Securities and Exchange Commission
(SEC) 13-F filings, cover institutions with more than $100 million in assets, which means
that we effectively measure institutional ownership with some error; e.g., if a firm is owned
only by institutions with less than $100 million in assets, we code it as having zero
institutional ownership. Furthermore, this data set aggregates up mutual fund holdings to
the level of the fund family, not the individual fund, which prevents us from saying
anything about the inertial tendencies of individual fund managers. We have, however,
also looked at a different data set on mutual fund holdings to gain some insight into the
determinants of inertia at the individual-fund level.
Table 1 presents summary statistics for the mergers in our sample. All variables are
Winsorized at the first and 99th percentiles, both in Table 1 and in the analysis that
follows. However, our regressions yield essentially identical point estimates and modestly
larger standard errors if we do not Winsorize any of the dependent variables. In Panel A,
we look at target institutional ownership, which is calculated for the quarter prior to the
announcement of a merger and expressed as a percentage of shares outstanding. We also
look at non-overlapping target institutional ownership, defined as the fraction of the target
owned by those institutions who own no shares in the acquiring firm. (Matvos and
Ostrovsky (2006) highlight the importance of overlapping institutional ownership in
mergers.) As we discuss below, non-overlapping institutional ownership is probably the
best available proxy for the wakefulness parameter a.
Panel B gives several deal characteristics. Acquirer and target size are equal to price
times shares outstanding two days prior to announcement. Relative size is equal to target
size expressed as a percentage of total target and acquirer size. Acquirer and target leverage
are equal to interest-bearing debt (items 9+34 from Compustat) expressed as a percentage
of book assets (item 6) at the fiscal year-end prior to announcement. Acquirer and target
market-to-book are equal to book assets minus book equity (items 216-130+35) plus price
times shares outstanding (from CRSP), all divided by book assets. Same industry is an
indicator variable equal to one if the target and the acquirer are in the same Fama and
French (1997) industry.
Panel C presents stock market data, all taken from CRSP. Acquirer and target
announcement returns are cumulative returns in excess of the value-weighted market over
a five-day window surrounding the announcement of a merger. Acquirer announcement
volume is the average daily volume over a five-day window surrounding the announcement
9
Of our 1,851 stock deals, we identify 53, or just under 3%, where the acquirer has a pre-bid position in the
target. The number of these toeholds is too small for them to affect our overall analysis one way or another.
However, toeholds appear to behave as sleepy investors. In deals with toeholds, the return to the acquirer is less
negative, and trading volume is reduced, though given the small number of toehold observations these effects are
not reliably estimated.
ARTICLE IN PRESS
Table 1
Merger summary statistics
The sample includes successful stock-swap and cash mergers with Center for Research in Security Pricing
(CRSP) targets announced between the second quarter of 1980 and the fourth quarter of 2000, involving a CRSP
acquirer with a matched announcement return. Institutional ownership (IO) is summarized in Panel A. Target
IO
is totaled from the CDA/Spectrum Institutional Holdings database for the quarter prior to the announcement of
the merger and expressed as a percentage of shares outstanding. Non-overlapping target IO includes only those
institutions that own no shares of the acquiring firm. This is expressed as a percentage of both shares outstanding
and total target IO. The deal characteristics in Panel B are from CRSP and Compustat. Cash deal is equal to one
when the consideration is cash and zero when the consideration is stock. Acquirer and target size are equal to
price times shares outstanding from CRSP. Relative size is equal to target size expressed as a percentage of total
target and acquirer size. Acquirer and target leverage are equal to interest-bearing debt (9+34) from Compustat
expressed as a percentage of book assets. Acquirer and target market to book (M/B) are equal to book assets (6)
minus book equity (216�130+35) from Compustat plus price times shares outstanding from CRSP all divided by
book assets. Same industry is an indicator variable equal to one if the target and the acquirer are included in the
same Fama and French 48-industry grouping. The stock market data in Panel C are from CRSP. The acquirer
and target announcement returns are the return in excess of the value-weighted market over a five-day window
surrounding the announcement of the merger. The acquirer announcement volume is the average daily volume
over a five-day window surrounding the announcement of the merger expressed as a percentage of shares
outstanding. Normal volume is the average daily volume over a 60-day window starting 90 trading days before the
announcement of the merger and expressed as a percentage of shares outstanding. The acquirer demand curve
proxies in Panel D are from CRSP and Institutional Bankers’ Estimate System. The dispersion in analyst forecasts
is the standard deviation of all outstanding long-term growth forecasts. Idiosyncratic risk is the standard
deviation of the residuals from a regression of acquirer excess returns on the Fama and French benchmark factors
(RM, SMB, HML) and the matched 48 industry portfolio return. All factor and portfolio returns were obtained
from Ken French’s website. All variables are Winsorized at the first and 99th percentiles.
Summary statistics, second quarter of 1980 to fourth quarter of 2000
N Mean Median Standard
deviation
Minimum Maximum
Panel A. Institutional ownership
Target IO (percent of total) 2,995 26.32 20.42 22.75 0.00 85.19
Non-overlapping target IO (pct of total) 2,995 13.09 8.91 13.63 0.00 61.
30
Non-overlapping target IO (percent of IO) 2,816 51.07 50.66 30.87 0.00 100.00
Panel B. Deal characteristics
Cash deal 2,995 0.38 0.00 0.49 0.00 1.00
Acquirer size (millions of dollars) 2,995 7,192 1,094 20,900 5 146,000
Target size (millions of dollars) 2,995 490 105 1,196 0 7,6
80
Relative size (percent) 2,992 17.69 11.28 18.27 0.00 85.77
Acquirer leverage (percent) 2,816 21.91 19.29 16.65 0.00 78.56
Target leverage (percent) 2,508 21.75 17.29 20.36 0.00 84.93
Acquirer M/B 2,834 2.42 1.30 3.25 0.63 22.04
Target M/B 2,537 2.02 1.30 2.05 0.45 14.01
Same industry 2,995 0.57 1.00 0.49 0.00 1.00
Panel C. Stock market data
Acquirer announcement return (percent) 2,995 �1.17 �0.94 7.29 �24.08 20.82
Target announcement return (percent) 2,995 18.67 14.72 21.79 �24.50 97.37
Acquirer announcement volume (pct) 2,963 0.81 0.38 1.14 0.01 6.65
Normal volume (percent) 2,929 0.49 0.29 0.60 0.01 3.53
Panel D. Acquirer demand curve
Dispersion in analyst forecasts (percent) 1,599 3.28 2.50 3.07 0.00 60.
90
Idiosyncratic risk (percent) 2,853 8.06 6.71 4.66 2.54 27.59
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 277
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298278
of a merger, expressed as a percentage of shares outstanding. Normal volume is the aver-
age daily volume over a 60-day window starting 90 trading days before the announcement
of a merger.
Finally, Panel D shows two acquirer demand-curve proxies, constructed from data in
CRSP and IBES. Our first proxy follows Scherbina (2004) and Moeller, Schlingemann,
and Stulz (2004), who use dispersion in analyst forecasts as a measure of disagreement
about fundamental value. Our particular measure is the same adopted by Moeller,
Schlingemann, and Stulz in a similar context and is equal to the standard deviation of all
outstanding earnings forecasts of long-term growth. Our second proxy, idiosyncratic risk,
is the standard deviation of the residuals from a regression of acquirer excess returns on
the Fama and French factors (RM, SMB, HML) and the matched 48-industry portfolio
return. (All factor and portfolio returns were obtained from Ken French’s website.)
3.2. Measuring investor inertia
If all mergers were announced and completed instantaneously, as in our simple model, it
would be a straightforward matter to measure target investor inertia. Consider a merger in
which we have a set of target investors who have no initial position in the acquirer. By
revealed preference, these target-only investors do not place an especially high value on the
acquirer. If these investors were all awake, we would expect them to sell their shares
immediately upon announcement and simultaneous completion. In contrast, if the target-
only investors were asleep, we would expect them to do nothing. Thus a natural measure of
inertia would be the fraction of target-only investors doing nothing. This would
correspond exactly to the key variable (1– a) in our model.
In practice, there is a lag between the announcement and completion of a deal. Our aim
is to get a picture of the total selling activity by target-only investors over this interval.
Moreover, the passage of time raises subtle benchmarking issues. For example, suppose
that a particular deal takes a year to close and 10% of the target-only investors sell their
shares during the course of this year. Should we draw the conclusion that 10% are
effectively awake and responding to the merger? Not necessarily. Even in a year without a
merger, there is a baseline level of turnover. In other words, we expect to see some selling,
for example, because of liquidity demand, even if all target investors are completely
oblivious to the fact that a merger has occurred. So, to measure inertia correctly, we need
to calculate turnover above and beyond what would be expected absent a merger.
Fig. 1 illustrates our method for calculating investor inertia in a hypothetical merger
transaction. Time from announcement to completion is measured along the X-axis. The
merger is announced at the Y-axis and is completed by the end of the dashed line. At any
point in between, the dashed line represents the fraction of those pre-announcement target
investors with no initial position in the acquirer who continue to hold a long position,
either in the target (in the period prior to completion) or in the acquirer (in the period after
completion). For example, the figure shows that, at completion, 32% of the original target-
only investors are still holding their positions, which have now converted into shares of the
acquirer.
The upper solid line in Fig. 1, which starts out at 100% and declines gradually, is the
fraction of target-only investors we would expect to be continuing holders based on
normal turnover during a non-merger period. In this hypothetical example, the figure
shows that only 53% of investors are expected to still hold their target shares seven
ARTICLE IN PRESS
0
10
20
30
40
50
60
70
80
90
100
Time from merger announcement
a
b
Inertia = a/b
P
e
rc
e
n
t
Fig. 1. Calculating inertia. To calculate the fraction of investors that are passive with respect to a merger
transaction, we compare their holdings with benchmark holdings levels (in bold). The upper benchmark reflects
the fraction of investors that, during a non-merger period, held a given stock in month zero and continued to hold
it in subsequent months. The lower benchmark reflects the fraction of investors that, during a non-merger period,
did not hold a given stock in month zero but did hold it during subsequent months. These benchmarks can be
compared with the fraction of investors that, at the time of a merger announcement, held the target but not the
acquirer and who continue to hold the target prior to completion or who hold the acquirer after the merger is
completed.
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 279
quarters later, even if they pay no attention to the merger. The lower solid line, which
starts out at 0%, captures the idea that an investor who is awake and sells the target
immediately upon announcement might, over time, experience a change in view and re-buy
the target or the acquirer at some later date. The figure shows that 14% of the original
target-only investors in this example would be expected to own the acquirer seven quarters
later if they were to sell out at the time of the merger announcement, simply because they
revise their opinion of the acquirer.
To compute an adjusted measure of inertia that incorporates both of these benchmarks,
we define inertia at completion as the difference between the fraction of post-completion
holders and the lower benchmark (labeled a in Fig. 1), divided by the difference between
the upper benchmark and the lower benchmark (labeled b in Fig. 1). In other words,
inertia measures, in relative terms, how close post-completion holdings are to the upper
benchmark, as opposed to the lower benchmark. In the specific example shown in Fig. 1,
inertia at completion is 44.9% (44.9 ¼ 32–14 divided by 53–14).
In the institutional data, we calculate the upper and lower benchmarks by examining the
behavior of investors in the same set of target stocks in the period beginning 12 quarters
before merger announcement. For the purpose of computing the benchmarks, we again
focus on those target investors who have no initial position in the acquirer. The upper
benchmark is these investors’ propensity to close out their position in the target over
various horizons, and the lower benchmark is their propensity to establish a new position
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298280
in the acquirer over various horizons. We take a similar approach with individual
investors, calculating the benchmarks in 1991 for all deals that are announced in 1993 or
later.
3.2.1. Individual investor inertia
Table 2 presents our analysis of inertia among individual investors. This analysis is
restricted to the 305 stock-swap mergers that have at least one target-only investor during
the interval for which we have the brokerage-firm data. For each deal, we begin with all
investors, who, at the month-end prior to the transaction, hold the target but not the
acquirer. We then track the holdings of this set of individuals over the period from
announcement to completion. Each of the first nine columns in the table isolates deals with
a fixed length of time to completion (one month, two months, etc.) and the final column
presents aggregated results for all 305 deals.
For example, the first column of Table 2 shows that 21 mergers are completed in the next
calendar month after announcement. In these transactions, 11.4% of pre-merger
individual investors close out their positions in the target before completion. Thus, pre-
completion holdings are 88.6% (88.6 ¼ 100–11.4). Of the remaining 88.6%, another
26.3% close out their positions within three months after completion, leaving net post-
completion holdings at 65.3% [65.3 ¼ 88.6 (1–0.263)] four months after we start tracking
the holdings of the target-only investors in these 21 mergers.
This post-completion figure of 65.3% can then be compared with the upper four-month
benchmark of 87.3%, which means that, in a typical non-merger-affected, four-month
period for the full set of target stocks, 12.7% of individual investors close out their
positions. It can also be compared with the lower four-month benchmark of 0.3%, which
means that in a typical non-merger-affected, four-month period for the full set of target
stocks, only 0.3% of individual investors in the target who do not initially have a position
in the acquirer open a new position in the acquirer. Putting it all together, we calculate
inertia of 74.7% for the subset of mergers that take one month from announcement to
completion (74.7 ¼ 65.3�0.3 divided by 87.3�0.3).
The results are similar in each of the subsequent columns, representing deals that take
two or more months to move from announcement to completion. The final column
aggregates across these different samples, weighting each one by the number of target-only
investors. On average, for individuals, post-completion holdings are 64.1%, and inertia is
78.8%. Thus, loosely speaking, we estimate that about 80% of individuals effectively sleep
through mergers in which they are shareholders in the target firm.
This is a rough number, and one can certainly argue with the details of our
benchmarking methodology. However, our basic conclusion is likely to be robust: The vast
majority of individual target-firm investors do not react to a merger by unloading their
shares. Fig. 2 highlights the conclusions in Table 2 graphically. Across all merger horizons,
the ownership percentage falls at a similar rate to that of the upper benchmark. Essentially
all horizons are within 10% of this upper benchmark. Given that the lower benchmark
never gets above 1%, it is clear that the inertia among individual target investors is
substantial.
3.2.2. Institutional investor inertia
Table 3 undertakes an analogous exercise for institutional investors. Given the broader
coverage of the CDA/Spectrum data, we are able to include 1,797 stock-swap mergers in
ARTICLE IN PRESS
Table 2
Turnover around merger announcements: individual holdings
We identify situations in a database of individual investor holdings at a large discount broker from 1991 to 1996
when the investor has a position in a target but not its acquirer in the month ending prior to the announcement of
a stock-swap merger in the Wall Street Journal. The first row shows the number of successful stock-swap mergers,
involving a public acquirer and with at least one matched individual investor with a position in the target but not
the acquirer. We split the sample according to the number of months elapsed between the announcement and
completion of a merger. The following ten rows show percentage turnover in this set of target positions over the
nine months following the merger announcement and in the three-month period that the merger is successfully
completed. Turnover in the quarter that the merger is completed occurs when one of the original investors still has
a position in the target in the month prior to completion but does not have a position in the acquirer three months
later. The next two rows compound the pre-completion and completion turnover to compute the percentage of the
original investors that still has a position in the target at the month end prior to completion and that has a
position in the acquirer three months later. The last two rows benchmark the post-completion holdings. The upper
benchmark holdings reflects the corresponding turnover for situations in the individual investor database where
an individual owns shares in a target but not its acquirer in 1991 for all deals that are announced in 1993 or later.
The lower benchmark starts with the same situations and tracks the purchase of the acquirer shares. The final row
reports the difference between post-completion holdings and the lower benchmark expressed as a percentage of
the difference
between the upper and lower benchmarks.
Total months to completion
1 2 3 4 5 6 7 8 9+ All
Number of successful stock swaps 21 76 56 44 39 32 12 8 17 305
Announcement+1 turnover (percent) 11.4 16.4 11.0 8.6 9.6 7.4 20.4 10.0 6.5 12.1
Announcement+2 turnover (percent) 7.2 5.5 5.7 7.0 6.4 6.0 1.2 5.0 6.1
Announcement+3 turnover (percent) 4.0 6.3 3.8 5.4 4.1 6.3 1.4 4.4
Announcement+4 turnover (percent) 6.4 5.8 2.3 3.6 4.0 5.2 5.0
Announcement+5 turnover (percent) 1.7 3.2 2.1 2.8 4.4 2.9
Announcement+6 turnover (percent) 3.0 3.2 2.9 7.7 4.8
Announcement+7 turnover (percent) 3.3 4.4 10.1 6.5
Announcement+8 turnover (percent) 1.5 3.7 3.4
Announcement+9 turnover (percent) 3.1 3.1
Completion turnover (percent) 26.3 21.1 11.3 11.5 7.7 5.6 17.1 4.7 6.9 14.0
Pre-completion holdings (percent) 88.6 77.6 80.8 75.6 74.9 75.2 63.4 71.1 56.4 74.7
Post-completion holdings (percent) 65.3 61.2 71.6 66.9 69.1 70.9 52.6 67.8 57.2 64.1
Upper benchmark holdings (percent) 87.3 85.7 83.9 82.3 80.9 79.7 78.4 76.4 70.7 81.7
Lower benchmark holdings (percent) 0.3 0.3 0.4 0.4 0.4 0.5 0.5 0.5 0.6 0.4
Inertia (percent) 74.7 71.3 85.3 81.2 85.3 88.9 66.9 88.7 80.7 78.8
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 281
this analysis. This is the subset of mergers that have at least one target-only investor
between the second quarter of 1980 and the fourth quarter of 2002. Because of the SEC 13-
F reporting requirements, we are forced to look at things on a quarterly, as opposed to
monthly, basis.
Aggregating across all transactions, we find that 30.0% of pre-merger institutional
investors hold on to their positions through completion, substantially less than the
corresponding figure of 64.1% for individuals. The baseline rate of turnover is also higher.
However, the first effect dominates and inertia is significantly lower, at 32.3% for
institutions, as compared with 78.8% for individuals. In contrast to the individual investor
turnover in Fig. 2, institutional holdings in Fig. 3 decline at a significantly faster rate than
ARTICLE IN PRESS
0
10
20
30
40
50
60
70
80
90
100
0 1 2 3 4 5 6 7 8 9 10
Months from merger announcement
P
e
rc
e
n
t
Fig. 2. Turnover around merger announcements: individual holdings. We identify situations in a database of
individual investor holdings at a large discount broker from 1991 to 1996 when the investor has a position in a
target but not its acquirer in the month ending prior to the announcement of a stock-swap merger in the Wall
Street Journal. The figure includes 305 successful stock-swap mergers that are completed within nine months
involving a public acquirer and with at least one matched individual investor with a position in the target but not
the acquirer. Each dashed line tracks the percentage of the original individual investors that still has a position in
the target at each month end prior to completion and that has a position in the acquirer two full months following
the completion of the stock swap. We split the sample, plotting one line for each number of months elapsed
between the announcement and completion of a merger. The solid lines benchmark the individual investor
holdings. The upper benchmark holdings reflects the corresponding turnover for situations in the individual
investor database when an individual owns shares in a target but not its acquirer in 1991 for all deals that are
announced in 1993 or later. The lower benchmark starts with the same situations and tracks the purchase of the
acquirer shares.
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298282
the upper benchmark across all merger completion horizons. By the conclusion of most
of the mergers, the institutional holdings lines are all closer to the lower benchmark than to
the upper one.
3.2.3. The determinants of inertia
Before proceeding to test Propositions 1 and 2, we examine some of the cross-sectional
determinants of inertia within the broad categories of individual and institutional
investors. For example, one might expect that capital gains taxes would encourage any
given investor to hold on to the stock of a target that has appreciated substantially. Or,
if the target is large relative to the acquirer, a target investor with no revealed preference
for the acquirer could nevertheless want to hold on to the shares of the merged com-
pany, because its value is determined to a large extent by the prospects of the original
target assets.
In Table 4, we present the pre- and post-completion turnover of target-only investors
calculated for various subsamples of stock-swap mergers. In Panels A and D, we address
ARTICLE IN PRESS
Table 3
Turnover around merger announcements: institutional holdings
We identify situations in the CDA/Spectrum Institutional Holdings database in which an institution has a
position in a target but not its acquirer in the quarter ending prior to the announcement of a stock-swap merger in
the Wall Street Journal. The first row shows the number of successful stock-swap mergers, involving a public
acquirer and with at least one matched institution with a position in the target but not the acquirer. We split the
sample according to the number of quarters elapsed between the announcement and completion of a merger. The
following five rows show percentage turnover in this set of target positions over the four quarters following the
merger announcement and in the quarter that the merger is successfully completed. Turnover in the quarter that
the merger is completed occurs when one of the original institutions still has a position in the target in the quarter
prior to completion but does not have a position in the acquirer in the quarter after completion. The next two
rows compound the pre-completion and completion turnover to compute the percentage of the original
institutions that still has a position in the target at the quarter end prior to completion and that has a position in
the acquirer at the quarter end following the completion of the stock swap. The last two rows benchmark the post-
completion holdings. The upper benchmark holdings reflects the corresponding turnover for situations in the
CDA/Spectrum Institutional Holdings database when an institution owns shares in a target but not its acquirer in
the quarter ending 12 quarters prior to announcement. The lower benchmark starts with the same situations
12
quarters prior to announcement and tracks the purchase of the acquirer shares. The final row reports the
difference between post-completion holdings and the lower benchmark expressed as a percentage of the difference
between the upper and lower benchmarks.
Total quarters to completion
0 1 2 3 4+ All
Number of successful stock swaps 85 862 609 166 75 1.797
Announcement+1 turnover (percent) 37.5 30.3 27.8 36.0 34.4
Announcement+2 turnover (percent) 27.3 22.5 8.3 23.8
Announcement+3 turnover (percent) 18.5 12.9 17.5
Announcement+4 turnover (percent) 11.0 16.7
Completion turnover (percent) 72.7 51.3 39.6 27.2 26.6 43.6
Pre-completion holdings (percent) 62.5 50.7 45.6 33.8 55.0
Post-completion holdings (percent) 27.3 30.4 30.6 33.2 23.4 30.0
Upper benchmark holdings (percent) 85.3 77.2 70.9 66.0 56.9 73.1
Lower benchmark holdings (percent) 5.2 8.6 10.0 11.1 13.3 9.5
Inertia (percent) 27.6 31.8 33.9 40.3 23.2 32.3
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 283
the tax hypothesis by splitting the individual and institutional-investor samples according
to the pre-announcement return of the target, calculated as the cumulative return over the
two-year period ending one month prior to announcement. Pre-announcement returns
appear to be weakly related to the inertia of individual investors: Target investors are
somewhat less willing to dump shares that have appreciated in value by more than 20%.
However, the differential across groups is modest.
10
Even mergers in which the target has
recently declined in value also involve considerable inertia. The inertia statistic of 76.5%
for this group is close to the full-sample value of 78.8%. With institutions, there is no
10
Two effects combine to make the inertia statistic greater among targets with large positive returns. First, post-
completion holdings are greater, consistent with a capital gains tax story. And second, the upper benchmark is
lower, consistent with the disposition effect (Shefrin and Statman, 1985). Thus, if anything, the modest differences
in inertia that we show likely overstate the pure impact of capital gains tax considerations.
ARTICLE IN PRESS
0
10
20
30
40
50
60
70
80
90
100
0 1 2 3 4 5 6 7 8 9
Quarters from merger announcement
P
e
rc
e
n
t
Fig. 3. Turnover around merger announcements: institutional holdings. We identify situations in the CDA/
Spectrum Institutional Holdings database in which an institution has a position in a target but not its acquirer in
the quarter ending prior to the announcement of a stock-swap merger in the Wall Street Journal. The figure
includes 1,789 successful stock-swap mergers that are completed within six quarters, involving a public acquirer
and with at least one matched institution with a position in the target but not the acquirer. Each dashed line tracks
the percentage of the original institutions that still has a position in the target at each quarter end prior to
completion and that has a position in the acquirer at the quarter end following the completion of the stock swap.
We split the sample, plotting one line for each number of quarters elapsed between the announcement and
completion of a merger. The solid lines benchmark the institutional holdings. The upper benchmark holdings
reflects the corresponding turnover for situations in the CDA/Spectrum Institutional Holdings database when an
institution owns shares in a target but not its acquirer in the quarter ending 12 quarters prior to announcement.
The lower benchmark starts with the same situations 12 quarters prior to announcement and tracks the purchase
of the acquirer shares.
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298284
evidence at all to suggest that capital gains taxes matter for our measure of inertia. Both
raw post-completion holdings and our inertia statistic are at their highest among those
firms with negative pre-announcement returns, the group for which tax considerations
would suggest that there should be the least inertia.
Using past two-year returns is a crude way to proxy for the tax status of any given
investor. It fails to take into account the investor’s basis in the stock, his holding period,
and the potential gains and losses on any other stocks in his portfolio. One way to do a
little better is to re-calculate the inertia statistic for only those individual positions that
appear to involve capital losses, i.e., those in which the post-announcement stock price is
below the investor’s initial basis in the stock. If tax considerations were important, we
would expect less inertia in this subsample. However, the inertia statistic (not tabulated) is
if anything a bit higher in the capital-loss subsample than its full-sample value, at 84.0%.
A somewhat sharper test of the tax hypothesis is contained in Panel C. Here we split the
individual-investor sample according to whether or not the investor in question has a
taxable account. As can be seen, the differences in inertia across the two subsamples are
ARTICLE IN PRESS
Table 4
Turnover around merger announcements: subsamples based on deal characteristics
We repeat the analysis in Tables 2 and 3 for subsamples of stock swaps. See Tables 2 and 3 for details. In Panels
A and D, we split the successful stock-swap mergers with at least one matched target position into three groups
according the return in the two years ending one month prior to the announcement of the merger in the Wall
Street Journal. In Panels B and E, we split the sample according to the relative size of the target and acquiring
firms. Relative size is equal to the target market capitalization (price times shares outstanding from the Center for
Research in Security Pricing) expressed as a percentage of the total market capitalization of the target and the
acquirer. In Panel C, we split the sample of individuals according to whether or not any of their accounts are
taxable.
Pre-completion Post-completion
Holdings Holdings Upper Lower Inertia
Panel A. Individuals, split on target two-year pre-announcement return
o0% 75.2 61.2 79.9 0.2 76.5
0–20% 83.3 71.3 78.9 0.3 90.3
420% 74.9 65.6 70.1 0.4 93.5
Panel B. Individuals, split on relative size
o5% 77.1 62.7 74.8 0.3 83.8
5–25% 73.7 63.7 77.9 0.3 81.7
425% 74.1 64.5 76.8 0.2 83.9
Panel C. Individuals, split on tax status
Taxable 65.2 58.9 76.7 0.4 77.2
Tax-exempt 70.5 61.9 78.2 0.4 79.6
Panel D. Institutions, split on target two-year pre-announcement return
o0% 53.5 33.1 73.4 9.4 37.0
0–20% 45.3 24.2 69.0 15.3 16.5
420% 57.0 30.6 74.0 9.2 33.0
Panel E. Institutions, split on relative size
o5% 54.6 31.6 74.5 9.2 34.4
5–25% 50.8 33.8 75.3 13.1 33.3
425% 58.2 27.8 73.4 9.4 28.8
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 285
negligible, with inertia statistics of 77.2% and 79.6% for taxable and non-taxable
investors, respectively. This runs strongly counter to the idea that inertia is a rational
response to capital gains tax considerations. One caveat, however, is that the types of
individuals who hold stocks in their retirement accounts could be different from those who
do not. If so, our comparison of inertia across taxable and non-taxable accounts could be
misleading, because it could be muddied by compositional differences in the investors
under consideration. Fortunately, there is an easy fix for this problem. We can simply redo
the taxable versus non-taxable account comparison, restricting the analysis to only those
investors who have both types of accounts. The results (not tabulated) are essentially
identical to those we report in Panel C of Table 4.
Panels B and E split the samples based on the relative size of the target, calculated as the
target’s market capitalization divided by the total market capitalization of the target and
acquirer. Individuals exhibit uniformly high inertia in deals of widely varying relative size.
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298286
Institutions appear slightly less passive when the target is large. Again, this result goes in
the opposite direction relative to a simple rational story: Mergers with large targets
represent situations in which the combined firm’s assets ought to appear most attractive to
those investors who initially found the target worth owning, so such deals should be
expected to generate more, not less inertia.
We have also tried a couple of other cross-sectional tests that are not shown in Table 4.
In one of these, we show that inertia among individuals is somewhat lower, at 72.3% (as
compared with the full-sample value of 78.8%) for those investors whose positions in the
target are above the median size of $2,700. This is consistent with individual-investor
inertia being in part a product of either the transactions costs or the fixed search costs
associated with portfolio rebalancing.
11
At the same time, inertia remains substantial even
for those individual investors whose stakes in the target are large, suggesting that not
everything can be explained simply by small fixed costs. For example, for target positions
over the 90th percentile value of $14,625, the inertia statistic is still 67.0%. One can
rationalize this observation by appealing to sufficiently large search costs, but we would
argue that in the limit this amounts to not much more than a relabeling of what we have
been calling inertial behavior.
Finally, using the CDA/Spectrum Mutual Fund database (as opposed to the Institutional
Holdings database), we uncover some systematic patterns in inertia across different types of
mutual funds. For example, the inertia statistic is greater for balanced funds, at 52.8%,
than it is for aggressive growth funds, at 38.2%. This is likely because the latter type of
funds have a narrower investment mandate and, e.g., are forced to sell a high-growth target
when it is acquired by a low-growth bidder. In addition, mutual funds in general tend to
display substantially more inertia when the target and bidder are closer to the same size.
Again, a plausible story is that size-based style mandates force a fund to unload a target
when it is in a very different market-capitalization bracket than the acquirer.
3.3. Acquirer returns and volume around merger announcements
Having established the main premise of the model (that investors exhibit inertia), we
now consider its empirical implications. The first part of Proposition 1 is that target-
investor inertia leads to less negative acquirer announcement returns. A corollary laid out
in Proposition 2 is that this effect works through volume, so that target-investor inertia
leads to lower acquirer announcement volume at the same time. The third part of
Proposition 1 is that the impact of target-investor inertia interacts with the slope of the
demand curve for acquirer stock. The steeper is this demand curve, the greater should be
the impact of inertia on acquirer announcement returns.
3.3.1. Institutional ownership and acquirer announcement returns
In Panel A of Table 5, we focus on stock-for-stock mergers and regress acquirer
announcement returns on both target institutional ownership and non-overlapping target
11
For example, given that our sample period is the first half of the 1990s, round trip commission costs could
potentially eat up 10% of the median position of $2,700. If one adds to this a modest search cost in the range of
$100–$200, this might be sufficient to deter a rational investor with a median-size stake from trying to replace his
shares in the acquirer with shares in another firm, even if his subjective alpha for the former was, say, �5% per
year over the next three years.
ARTICLE IN PRESS
Table 5
Acquirer announcement returns
Regressions of acquirer merger announcement returns on target institutional ownership (IO) and deal
characteristics. Panel A shows results for stock-swap mergers, and Panel B shows results for cash mergers. The
acquirer announcement return is the return in excess of the value-weighted market over a five-day window
surrounding the announcement of the merger. Target institutional ownership is totaled from the CDA/Spectrum
Institutional Holdings database for the quarter prior to the announcement of the merger and expressed as a
percentage of shares outstanding. Non-overlapping target IO includes only those institutions that own no shares
of the acquiring firm and is also expressed as a percentage of shares outstanding. M/B denotes the market-to-book
ratio.
The other deal characteristics are described in Table 1. All variables are Winsorized at the first and 99th
percentiles. Announcement year fixed effects are included in all four specifications. Heteroskedasticity-robust t-
statistics are reported in braces to the right of the corresponding point
estimates.
Announcement return (percent)
1 2 3 4
Panel A. Stock deals
Target IO �3.87 [�4.72] �2.59 [�2.17]
Non-overlapping target IO �8.17 [�5.44] �5.49 [�3.07]
log(acquirer size) 0.42 [2.92] 0.34 [2.27]
log(target size) �0.60 [�2.88] �0.64 [�3.57]
Acquirer leverage 0.80 [0.50] 0.91 [0.56]
Target leverage 0.95 [0.85] 0.93 [0.84]
Acquirer M/B �0.27 [�2.91] �0.27 [�2.91]
Target M/B 0.24 [1.48] 0.24 [1.46]
Target announcement return 0.06 [5.16] 0.06 [5.16]
Same industry �0.22 [�0.45] �0.26 [�0.53]
Year fixed effects Yes Yes Yes Yes
N 1,851 1,475 1,851 1,475
R
2
0.03 0.08 0.03 0.08
Panel B. Cash deals
Target IO �0.43 [�0.49] 1.51 [1.27]
Non-overlapping target IO 0.91 [0.68] 1.31 [0.75]
Additional controls No Yes No Yes
Year fixed effects Yes Yes Yes Yes
N 1,144 902 1,144 902
R
2
0.03 0.05 0.03 0.05
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 287
institutional ownership. The first regression uses the raw measure of institutional
ownership as our proxy for investor wakefulness a and has no other controls, other
than year fixed effects. This specification generates a coefficient of �3.87 on institutional
ownership, which is strongly statistically significant (t-statistic ¼ 4.72), consistent with
Proposition 1. In economic terms, this coefficient implies that a 2.0 standard deviation
increase in target institutional ownership reduces the acquirer’s announcement return by
1.76 percentage points, taking it from its unconditional mean value of �2.23% down to
�3.99%.
The next regression again uses the raw measure of institutional ownership but adds a
variety of controls described in Table 1: acquirer and target market capitalization (both
in logs); acquirer and target leverage; acquirer and target market-to-book; the target
announcement return; and an indicator variable equal to one if the two firms are in the
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298288
same industry. Many of these (the size, leverage, market-to-book, and relatedness
variables) are commonly used in regressions to explain acquirer announcement returns.
12
We add the target announcement return to the list because one potential competing
explanation for the effect of institutional ownership has to do with bargaining power.
Perhaps institutional blockholders in the target are able to extract a better price from the
acquiring firm, leading to lower acquirer announcement returns. If so, the target
announcement return should control for this effect. However, as it turns out, this variable
has a significant, but positive, relationship with acquirer returns.
In any case, adding the full battery of controls has only a modest impact on the co-
efficient on institutional ownership. It goes from �3.87 to �2.59 and remains stati-
stically significant, with a t-statistic of 2.17. In this specification, a 2.0 standard deviation
increase in target institutional ownership reduces the acquirer return by 1.18 percentage
points.
The next two columns of Panel A are analogous to the previous two, except that we
replace target institutional ownership with non-overlapping target institutional ownership.
This latter variable is arguably a more precise measure of the wakefulness parameter a in
our model, because we expect an alert target shareholder to be most likely to sell shares in
the acquirer if he did not own any such shares prior to the merger announcement and
hence has not demonstrated a high valuation for acquirer assets. This redefinition of target
institutional ownership leads to coefficient estimates that are markedly higher in absolute
value. They are now �8.17 and �5.49 in the no-controls and full-controls specifications,
respectively. We view this pattern as particularly supportive of our model, because it is
hard to think of alternative hypotheses that would suggest a similar outcome.
Not surprisingly, target institutional ownership and non-overlapping target institutional
ownership are highly correlated, with a univariate correlation coefficient of 0.76 for our
sample of stock deals. In an untabulated analysis, we run an alternative version of our
specifications in which we include both the level of institutional ownership and the level of
non-overlapping ownership simultaneously. In these regressions, the coefficient on the
latter variable can be interpreted as an interaction term. It is the impact on the institutional
ownership coefficient of increasing the fraction of non-overlapping institution ownership.
This coefficient is still statistically and economically significant in the more elaborate
specification. We have also tried adding a control for acquirer institutional ownership to
our regressions, to address the possibility that non-overlapping target ownership is
somehow just a proxy for this. However, the coefficients on non-overlapping target
institutional ownership are not affected by this modification.
One control that is not included in the specifications in Table 5 is ownership of the target
by insiders, which could potentially influence acquirer returns through some sort of
screening mechanism. For example, a target with high inside ownership might accept only
very synergistic bids, because insiders are likely to have to hold on to the stock of the
combined firm after completion. The practical problem is that we do not have data on
target inside ownership for many of the transactions in our data set. However, we do have
12
See, e.g., Lang, Stulz, and Walkling (1989) and Morck, Shleifer, and Vishny (1990). Other typical merger-
related controls are not included because of the nature of our sample. The acquirer attitude (see Schwert, 2000, for
a discussion of hostility) is always friendly in our sample of stock swaps, and by definition the form of payment
(see Andrade, Mitchell, and Stafford, 2001) is stock and the accounting treatment is pooling (see Martinez-Jerez,
2004).
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 289
such data for a selected subsample, compiled by Thomson Financial from SEC Form 3,
Form 4, Form 5, and Form 144. Within this subsample, which is approximately 70% of
the size of our full sample, the effect of adding target insider ownership to our regressions
appears to be relatively small. It causes the coefficients on target institutional ownership
and non-overlapping target institutional ownership to drop by 18% and 7%, respectively.
Panel B of Table 5 is an exact replica of Panel A, except that the sample includes cash
mergers instead of stock mergers. This is effectively a placebo check. According to our
theory, target institutional ownership should be irrelevant in cash deals. In contrast, some
competing explanations for the results in Panel A suggest a similar pattern across stock
and cash mergers. For example, if high institutional ownership of the target leads to low
bidder returns through some sort of enhanced-bargaining effect, this should work similarly
for both stock and cash deals. However, as can be seen in Panel B, there is no discernible
impact of target institutional ownership in the cash-merger sample. The coefficient of
interest is never close to statistically significant and is positive in three out of four cases.
Thus cash mergers seem to be fundamentally different from stock mergers on this
dimension, consistent with our model.
We have been focusing our analysis of returns on the short (five-day) window around the
merger announcement date, both because this gives us the most statistical power and
because our theory makes the most clear-cut predictions with respect to it. Our static
model offers less guidance regarding longer-run price impacts. In practice, there are likely
to be two competing effects. On the one hand, because less-inertial institutional investors
continue to sell more after the merger announcement and through completion, one might
expect the adverse price impact associated with institutional ownership to continue to grow
over the medium run. On the other hand, at some longer horizon, it is possible that sleepy
individual investors eventually wake up and sell their shares, too, which would work in the
opposite direction. Even if we do not have much to say about the longer run, a temporary
price-pressure effect nevertheless matters for existing shareholders of the acquirer, because
it directly influences the number of acquirer shares that have to be granted to the target.
In spite of the theoretical ambiguities, we have undertaken some further empirical work
that effectively redoes Table 5 for longer return horizons of up to a year after the initial
announcement window. The point estimates from these regressions suggest that the short-
run effect of target institutional ownership is partially reversed in subsequent months.
However, these point estimates are sufficiently imprecise that we can never come close to
rejecting the hypothesis that there are no further abnormal returns associated with target
institutional ownership after the initial five-day event window.
3.3.2. Institutional ownership and acquirer announcement volume
The model makes the ancillary prediction that, in stock deals, the impact of shareholder
inertia on acquirer announcement returns works through trading volume, as in
Proposition 2. In particular, target shareholders who are awake sell out on announcement.
Merger arbitrageurs buy these shares and short the acquirer, closing their positions when
the merger is successfully completed. Examining volume also serves as yet another check
on alternative hypotheses linking target institutional ownership and acquirer returns.
Again, if target institutional ownership affects acquirer returns through a mechanism such
as bargaining power, we would not expect it to also influence volume simultaneously.
Table 6 repeats the analysis in Table 5, replacing acquirer returns as the dependent
variable with the average daily turnover in the acquirer over the five-day window
ARTICLE IN PRESS
Table 6
Acquirer announcement volume
Regressions of acquirer merger announcement volume on normal volume, target institutional ownership (IO),
and deal characteristics. Panel A shows results for stock-swap mergers, and Panel B shows results for cash
mergers. The acquirer announcement volume is the average daily volume over a five-day window surrounding the
announcement of the merger expressed as a percentage of shares outstanding. Normal volume is the average daily
volume over a 60-day window starting 90 trading days before the announcement of the merger and expressed as a
percentage of shares outstanding. Target institutional ownership is totaled from the CDA/Spectrum Institutional
Holdings database for the quarter prior to the announcement of the merger and expressed as a percentage of
shares outstanding. Non-overlapping target IO includes only those institutions that own no shares of the
acquiring firm and is also expressed as a percentage of shares outstanding. M/B denotes the market-to-book ratio.
The other deal characteristics are described in Table 1. All variables are Winsorized at the first and 99th
percentiles. Announcement year fixed effects are included in all four specifications. Heteroskedasticity-robust
t-statistics are reported in braces to the right of the corresponding point estimates.
Acquirer volume (percent)
1 2 3 4
Panel A. Stock deals
Target IO 1.07 [10.88] 0.59 [4.57]
Non-overlapping target IO 1.70 [7.84] 0.86 [3.64]
Normal volume 1.35 [24.00] 1.27 [18.71] 1.34 [23.42] 1.26 [18.62]
log(acquirer size) �0.19 [�11.61] �0.18 [�10.90]
log(target size) 0.21 [10.16] 0.23 [12.44]
Acquirer leverage �0.26 [�1.56] �0.27 [�1.64]
Target leverage 0.26 [2.11] 0.26 [2.14]
Acquirer M/B 0.01 [0.55] 0.01 [0.71]
Target M/B 0.06 [3.24] 0.06 [3.27]
Target announcement return 0.57 [4.45] 0.59 [4.64]
Same industry 0.08 [1.61] 0.08 [1.52]
Year fixed effects Yes Yes Yes Yes
N 1,807 1,464 1,807 1,464
R
2
0.58 0.63 0.58 0.63
Panel B. Cash deals
Target IO 0.21 [2.31] 0.12 [0.75]
Non-overlapping target IO 0.27 [1.74] 0.11 [0.53]
Normal volume 0.99 [9.39] 0.88 [7.92] 0.99 [9.32] 0.88 [7.90]
Additional controls No Yes No Yes
Year fixed effects Yes Yes Yes Yes
N 1,122 898 1,122 898
R
2
0.44 0.42 0.44 0.42
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298290
surrounding the merger announcement. We keep all the same right-hand-side variables as
before and also add normal trading volume (defined as the average daily turnover in the
acquirer over the 60-day period starting 90 days before the announcement) as another
control.
Across all four specifications in the stock merger sample in Panel A, the results are
uniformly supportive of Proposition 2. The coefficients on raw target institutional
ownership are 0.59 and 1.07 in the specifications with and without controls, respectively,
and are strongly statistically significant in both cases. When we use non-overlapping target
institutional ownership instead, the coefficients again rise in absolute value, to 0.86 and
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 291
1.70, respectively. And as before, the implied economic effects are substantial. A 2.0
standard deviation increase in target institutional ownership increases the average
acquirer’s daily turnover during the announcement period from a mean of 1.00% to
between 1.26% and 1.48%, or by between 25% and 50%. In Panel B, we see that, for cash
mergers, target institutional ownership has little effect on turnover, just as it has no effect
on returns. The coefficients are in all cases much smaller than in Panel A and, with one
exception, not statistically significant. Again, this is just what one would expect based on
our model.
3.3.3. Demand-curve and relative-size interactions
The two key ingredients of our model are inertia among target shareholders and a
downward-sloping demand curve for acquirer shares. Thus, as formulated in the third part
of Proposition 1, we expect our results for target institutional ownership to be strongest
among acquirer firms with steeply sloped demand curves. To operationalize this
hypothesis, we employ two different proxies for the slope of the demand curve. The first
aims to measure the difference of opinion among investors with respect to acquirer value,
in the spirit of the parameter H in the model. Specifically, we follow Moeller,
Schlingemann, and Stulz (2004) and calculate the standard deviation of all outstanding
analysts’ forecasts for long-run growth.
The second proxy is the nonindustry idiosyncratic risk of the acquirer. We compute
this as the standard deviation of the residuals from a regression of acquirer excess returns
on the Fama and French factors (RM, SMB, HML) and the matched 48-industry port-
folio return. The premise here is as follows. In the presence of both differences of
opinion and risk aversion, an increase in idiosyncratic risk makes the demand curve
steeper, because it reduces the size of the position that any one investor with a given
valuation is willing to take on. Although this effect is absent from our model (which,
for simplicity, uses wealth constraints instead of risk aversion to generate the shape
of the demand curve), it is formalized in, e.g., Chen, Hong, and Stein (2002). Moreover,
Wurgler and Zhuravskaya (2002) provide empirical validation for the idea of using
idiosyncratic risk as a proxy for demand-curve slope, showing that the impact of S&P
500 index inclusion on stock prices is increasing in the idiosyncratic risk of the included
firm.
Idiosyncratic risk could be a proxy for other factors as well. For example, Dierkens
(1991) uses idiosyncratic risk as a measure of asymmetric information in a study of equity
issues. So our results with this variable are no doubt open to alternative interpretations.
Nevertheless, while an asymmetric-information story might easily explain why acquirers
with more idiosyncratic risk have more negative announcement returns on average, it is
less clear that such a story has anything to say about the higher-level interactions that we
focus on.
Table 7 presents regressions of acquirer announcement returns (in stock deals only) on
target institutional ownership, our measures of the slope of the demand curve facing the
acquirer, and the product of the two variables. In Panel A, we use the raw measure of
institutional ownership; in Panel B we use non-overlapping institutional ownership. In
either case, our theory suggests that the interaction term should attract a negative
coefficient. The first two regressions in each panel employ dispersion in analyst forecasts as
the proxy for the steepness of the acquirer’s demand curve, while the second two use
idiosyncratic risk.
ARTICLE IN PRESS
T
a
b
le
7
A
c
q
u
ir
e
r
a
n
n
o
u
n
c
e
m
e
n
t
re
tu
rn
s:
in
te
ra
c
ti
o
n
s
w
it
h
p
ro
x
ie
s
fo
r
d
e
m
a
n
d
-c
u
rv
e
sl
o
p
e
a
n
d
re
la
ti
v
e
si
z
e
R
e
g
re
ss
io
n
s
o
f
a
c
q
u
ir
e
r
m
e
rg
er
a
n
n
o
u
n
c
em
e
n
t
re
tu
rn
s
(i
n
st
o
c
k
d
e
a
ls
o
n
ly
)
o
n
ta
rg
et
in
st
it
u
ti
o
n
a
l
o
w
n
e
rs
h
ip
(I
O
),
a
c
q
u
ir
e
r
d
e
m
a
n
d
-c
u
rv
e
(D
C
)
o
r
re
la
ti
v
e
-s
iz
e
p
ro
x
ie
s,
in
te
ra
c
ti
o
n
s
b
e
tw
e
e
n
th
e
tw
o
,
a
n
d
d
e
a
l
c
h
a
ra
ct
e
ri
st
ic
s.
T
h
e
a
c
q
u
ir
e
r
a
n
n
o
u
n
c
e
m
en
t
re
tu
rn
is
th
e
re
tu
rn
in
e
x
c
e
ss
o
f
th
e
v
a
lu
e
-w
e
ig
h
te
d
m
a
rk
e
t
o
v
e
r
a
fi
v
e-
d
a
y
w
in
d
o
w
su
rr
o
u
n
d
in
g
th
e
a
n
n
o
u
n
c
e
m
en
t
o
f
th
e
m
er
g
e
r.
In
P
a
n
el
A
,
ta
rg
et
in
st
it
u
ti
o
n
a
l
o
w
n
e
rs
h
ip
is
to
ta
le
d
fr
o
m
th
e
C
D
A
/S
p
e
ct
ru
m
In
st
it
u
ti
o
n
a
l
H
o
ld
in
g
s
d
a
ta
b
a
se
fo
r
th
e
q
u
a
rt
er
p
ri
o
r
to
th
e
a
n
n
o
u
n
c
e
m
en
t
o
f
th
e
m
e
rg
er
a
n
d
e
x
p
re
ss
e
d
a
s
a
p
e
rc
e
n
ta
g
e
o
f
sh
a
re
s
o
u
ts
ta
n
d
in
g
.
In
P
a
n
e
l
B
,
n
o
n
-o
v
e
rl
a
p
p
in
g
ta
rg
et
IO
in
c
lu
d
e
s
o
n
ly
th
o
se
in
st
it
u
ti
o
n
s
th
a
t
o
w
n
n
o
sh
a
re
s
o
f
th
e
a
c
q
u
ir
in
g
fi
rm
a
n
d
is
a
ls
o
e
x
p
re
ss
e
d
a
s
a
p
er
c
e
n
ta
g
e
o
f
sh
a
re
s
o
u
ts
ta
n
d
in
g
.
T
h
e
fi
rs
t
tw
o
re
g
re
ss
io
n
s
in
e
a
c
h
p
a
n
e
l
u
se
d
is
p
e
rs
io
n
in
a
n
a
ly
st
lo
n
g
-t
e
rm
g
ro
w
th
fo
re
c
a
st
s
a
s
a
p
ro
x
y
fo
r
th
e
sl
o
p
e
o
f
th
e
a
c
q
u
ir
e
r
d
em
a
n
d
c
u
rv
e;
th
e
n
ex
t
tw
o
re
g
re
ss
io
n
s
u
se
id
io
sy
n
c
ra
ti
c
ri
sk
.
T
h
e
d
is
p
e
rs
io
n
in
a
n
a
ly
st
fo
re
c
a
st
s
is
th
e
st
a
n
d
a
rd
d
ev
ia
ti
o
n
o
f
a
ll
o
u
ts
ta
n
d
in
g
lo
n
g
-t
e
rm
g
ro
w
th
fo
re
c
a
st
s.
Id
io
sy
n
c
ra
ti
c
ri
sk
is
th
e
st
a
n
d
a
rd
d
e
v
ia
ti
o
n
o
f
th
e
re
si
d
u
a
ls
fr
o
m
a
re
g
re
ss
io
n
o
f
a
c
q
u
ir
e
r
e
x
c
e
ss
re
tu
rn
s
o
n
th
e
F
a
m
a
a
n
d
F
re
n
c
h
b
en
c
h
m
a
rk
fa
c
to
rs
(R
M
,
S
M
B
,
H
M
L
)
a
n
d
th
e
m
a
tc
h
e
d
4
8
in
d
u
st
ry
p
o
rt
fo
li
o
re
tu
rn
.
T
h
e
fi
n
a
l
tw
o
re
g
re
ss
io
n
s
in
e
a
c
h
p
a
n
el
fo
c
u
s
o
n
re
la
ti
v
e
si
z
e
,
d
e
fi
n
e
d
a
s
th
e
lo
g
o
f
th
e
ra
ti
o
o
f
ta
rg
e
t
to
a
c
q
u
ir
e
r
m
a
rk
et
c
a
p
it
a
li
za
ti
o
n
.
A
ll
fa
c
to
r
a
n
d
p
o
rt
fo
li
o
re
tu
rn
s
w
e
re
o
b
ta
in
e
d
fr
o
m
K
en
F
re
n
c
h
’s
w
e
b
si
te
.
T
h
e
o
th
e
r
d
e
a
l
c
h
a
ra
c
te
ri
st
ic
s
a
re
d
e
sc
ri
b
e
d
in
T
a
b
le
1
.
A
ll
v
a
ri
a
b
le
s
a
re
W
in
so
ri
ze
d
a
t
th
e
fi
rs
t
a
n
d
9
9
th
p
e
rc
e
n
ti
le
s.
T
h
e
d
e
m
a
n
d
c
u
rv
e
v
a
ri
a
b
le
s
a
re
st
a
n
d
a
rd
iz
e
d
to
h
a
v
e
z
e
ro
m
e
a
n
a
n
d
u
n
it
v
a
ri
a
n
c
e
.
A
n
n
o
u
n
c
e
m
en
t
y
e
a
r
fi
x
e
d
e
ff
e
ct
s
a
re
in
cl
u
d
e
d
in
a
ll
fo
u
r
sp
e
c
ifi
c
a
ti
o
n
s.
H
e
te
ro
sk
e
d
a
st
ic
it
y
-r
o
b
u
st
t-
st
a
ti
st
ic
s
a
re
re
p
o
rt
e
d
in
b
ra
c
es
to
th
e
ri
g
h
t
o
f
th
e
c
o
rr
e
sp
o
n
d
in
g
p
o
in
t
e
st
im
a
te
s.
D
is
p
e
rs
io
n
in
a
n
a
ly
st
lo
n
g
-t
e
rm
g
ro
w
th
fo
re
c
a
st
s
N
o
n
-i
n
d
u
st
ry
id
io
sy
n
c
ra
ti
c
ri
sk
R
e
la
ti
v
e
si
z
e
1
2
3
4
5
6
P
a
n
e
l
A
.
T
a
rg
e
t
IO
T
a
rg
e
t
IO
�
2
.9
8
[�
3
.2
3
]
�
1
.8
7
[�
1
.3
5
]
�
3
.5
2
[�
4
.5
2
]
�
2
.1
7
[�
1
.8
4
]
�
5
.1
0
[�
3
.4
9
]
�
4
.9
6
[�
2
.8
6
]
D
e
m
a
n
d
c
u
rv
e
/
si
z
e
(D
C
)
0
.4
6
[0
.8
6
]
0
.6
8
[1
.2
3
]
0
.4
7
[1
.3
8
]
0
.9
5
[2
.2
2
]
�
0
.2
7
[�
1
.7
0
]
T
a
rg
e
t
IO
�
D
C
�
2
.6
2
[�
1
.8
5
]
�
2
.6
2
[�
1
.7
9
]
�
3
.4
2
[�
3
.0
3
]
�
3
.2
4
[�
2
.7
3
]
�
1
.1
5
[�
2
.1
9
]
�
1
.2
6
[�
2
.1
4
]
A
d
d
it
io
n
a
l
c
o
n
tr
o
ls
N
o
Y
e
s
N
o
Y
e
s
N
o
Y
e
s
Y
e
a
r
fi
x
e
d
e
ff
e
c
ts
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
N
1
,1
8
6
9
9
9
1
,7
7
1
1
,4
5
4
1
,8
3
5
1
,4
7
5
R
2
0
.0
3
0
.0
9
0
.0
4
0
.0
8
0
.0
4
0
.0
8
P
a
n
e
l
B
.
N
o
n
-o
v
e
rl
a
p
p
in
g
ta
rg
e
t
IO
N
o
n
-o
v
e
rl
a
p
p
in
g
ta
rg
e
t
IO
�
7
.3
0
[�
4
.1
8
]
�
4
.8
0
[�
2
.3
3
]
�
6
.7
8
[�
4
.8
6
]
�
4
.7
6
[�
2
.7
7
]
�
7
.3
3
[�
2
.9
4
]
�
6
.5
8
[�
2
.3
2
]
D
e
m
a
n
d
c
u
rv
e
/s
iz
e
(D
C
)
0
.2
4
[0
.4
9
]
0
.4
6
[0
.9
0
]
0
.1
9
[0
.5
7
]
0
.6
0
[1
.4
2
]
�
0
.4
4
[�
3
.0
6
]
N
o
n
-o
v
e
rl
a
p
p
in
g
ta
rg
e
t
IO
�
D
C
�
4
.7
9
[�
1
.6
1
]
�
4
.9
9
[�
1
.6
1
]
�
3
.0
0
[�
1
.4
7
]
�
2
.8
4
[�
1
.3
5
]
�
0
.6
1
[�
0
.5
9
]
�
0
.7
1
[�
0
.5
8
]
A
d
d
it
io
n
a
l
c
o
n
tr
o
ls
N
o
Y
e
s
N
o
Y
e
s
N
o
Y
e
s
Y
e
a
r
fi
x
e
d
e
ff
e
c
ts
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
Y
e
s
N
1
,1
8
6
9
9
9
1
,7
7
1
1
,4
5
4
1
,8
3
5
1
,4
7
5
R
2
0
.0
4
0
.1
0
0
.0
3
0
.0
8
0
.0
4
0
.0
8
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298292
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 293
The results in Table 7 provide further corroboration of the model. The interaction of the
acquirer demand-curve proxies and target institutional ownership is negative in all eight
specifications shown in the table and significant (at the 10% level or better) in the four
specifications in Panel A that use raw institutional ownership. In Panel B, with non-
overlapping institutional ownership, the interaction coefficients are generally at least as
big, but the standard errors are larger, perhaps because non-overlapping ownership has
less cross-sectional variation.
Moreover, each of the eight specifications implies economically meaningful interaction
effects. The demand-curve proxies are standardized to have zero mean and unit variance.
So the fact that the coefficients on target institutional ownership and on the interaction
term are of the same magnitude means that a 1.0 standard deviation increase in the
steepness of the demand curve roughly doubles the effect of institutional ownership on
acquirer returns.
Table 7 also considers the interaction of target institutional ownership and relative size,
defined as the log of the ratio of target to acquirer market capitalization. The logic for this
test again comes directly from Eq. (5), with relative size serving as a proxy for the variable
K in our model: All else equal, a larger target implies a potentially larger demand shock to
be absorbed by the pool of A-specialists, thereby increasing the marginal impact of
investor inertia. As can be seen, the coefficient on this interaction term is negative, as
predicted, in all specifications and significantly so in those in Panel A that use raw
institutional ownership.
To sum up, higher target institutional ownership leads to more negative announcement
effects for the acquirer. Our additional tests help pin down the mechanism. Institutional
holders of the target exhibit less inertia and so are more likely to sell their shares on
announcement. This leads to more price pressure and a negative announcement effect that
is strongest when the target is large relative to the acquirer and for those acquiring firms
with steeply sloped demand curves.
4. Further implications
In addition to its implications for merger announcement effects, our model makes two
more speculative predictions for corporate finance. All else equal, inertia makes equity
more attractive than cash as consideration in a merger or acquisition, and it makes equity
issued in the context of a merger more attractive than equity issued in an SEO.
13
4.1.
Financing choice
The first hypothesis is that stock-financed mergers are more attractive to an acquiring
firm when a greater fraction of target shareholders are asleep. This arises naturally in a
more general version of the model, in which the acquiring firm can use either cash or shares
as payment. To make the financing choice interesting, we need to introduce a cost of using
13
In formulating these hypotheses, we are taking the identity of the target firm as exogenous; i.e., as determined
by factors outside of our model, such as the quality of the match between the acquirer and the target. It is
tempting to add the prediction that a firm with a greater fraction of sleeping shareholders is more likely to become
a target in the first place. However, this factor could be of second-order importance relative to considerations of
match quality. Also, the same inertia that improves the terms for the acquiring firm could also lead to target
management entrenchment, thereby discouraging merger bids.
ARTICLE IN PRESS
Table 8
Financing choice
Probit regressions of the form of payment (cash or stock) on target institutional ownership (IO) and deal
characteristics. Target institutional ownership is totaled from the CDA/Spectrum Institutional Holdings database
for the quarter prior to the announcement of the merger and expressed as a percentage of shares outstanding.
Non-overlapping target IO includes only those institutions that own no shares of the acquiring firm and is also
expressed as a percentage of shares outstanding. M/B denotes the market-to-book ratio. The other deal
characteristics are described in Table 1. The coefficients show the impact of a unit change in the independent
variable, evaluated at its mean, on the probability that the consideration is the acquiring firm’s stock. All variables
are Winsorized at the first and 99th percentiles. Announcement year fixed effects are included in all four
specifications. Heteroskedasticity-robust t-statistics are reported in braces to the right of the corresponding point
estimates.
Consideration ¼ stock
1 2 3 4
Target IO �6.83 [�1.65] �20.76 [�3.30]
Non-overlapping target IO �45.02 [�6.67] �61.95 [�7.21]
log(acquirer size) �3.03 [�4.29] �4.28 [�5.79]
log(target size) 6.34 [6.18] 7.04 [7.71]
Acquirer leverage �12.39 [�1.78] �11.03 [�1.59]
Target leverage �5.47 [�0.98] �6.19 [�1.10]
Acquirer M/B 4.82 [4.71] 4.92 [4.79]
Target M/B 0.92 [1.09] 0.86 [1.00]
Target announcement return �0.43 [�8.18] �0.41 [�7.88]
Same industry 13.87 [6.00] 12.58 [5.41]
Year fixed effects Yes Yes Yes Yes
N 2,995 2,377 2,995 2,377
R
2
0.04 0.15 0.05 0.16
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298294
cash, such as the expected costs of financial distress. With this extra ingredient, it is
straightforward to show that an increase in target shareholder inertia makes stock
financing relatively more attractive.
Using the entire sample of cash and stock mergers, we run a probit regression in which
the dependent variable is a dummy that equals one if the consideration in the deal is stock
(and zero if it is cash), and in which the independent variables include either target
institutional ownership or non-overlapping target institutional ownership, as well as the
controls used in previous tables. As can be seen in Table 8, either measure of institutional
ownership has a strong negative influence on the probability of a merger being done with
stock. We report the coefficients as percentage effects evaluated at the mean of each
independent variable. In the specifications with the full set of controls, these coefficients
imply that a 2.0 standard deviation increase in target institutional ownership reduces the
probability of a stock offering by 9.4% from a mean of 61.8%, and a 2.0 standard
deviation increase in non-overlapping target institutional ownership reduces the
probability by 16.9%. Both estimates are statistically significant, as is the difference
between the two. Given that institutional investors are less likely to be sleepers than
individuals are, this pattern fits with our hypothesis.
There are other interpretations for these results. Maybe some institutions simply have a
preference, as compared with individual investors, for receiving cash as opposed to stock in
a merger transaction, and acquirers cater to this preference to lower the required merger
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 295
premium. We cannot directly refute this alternative story, though it would seem to have a
hard time rationalizing the fact that our results are considerably stronger for the non-
overlapping measure of institutional ownership that is most closely tied to our theoretical
model. Moreover, because acquirer shares can always be sold, an institutional preference
for cash requires that institutions expect significant transaction costs when selling acquirer
stock. The focus of our model is precisely on the price pressure component of these
transaction costs.
4.2. Equity financing in mergers and SEOs
Our theory also has something to say about why equity financing might be more
common in the context of mergers than it is with greenfield investment (Fama and French,
2005). As outlined in the Section 1, this point is easiest to see by thinking of a firm with an
exogenously specified growth strategy that, over the next year, entails one acquisition and
one major new greenfield investment, with the two transactions being of roughly similar
size. Assume that, to keep its capital structure in balance, the firm needs to finance one of
these two investments with an equity issue. That is, either the merger has to be stock-for-
stock or the greenfield investment has to be accompanied by an SEO.
14
We have seen that an SEO effectively amounts to a limiting case of our model with no
inertia, because all shares must be sold off in the market, and none can simply be placed in
the hands of inertial investors. Thus the SEO is associated with a more negative price
impact than the stock-for-stock merger, all else equal, and hence tends to be less attractive.
So we would expect the firm to finance the merger with a stock swap, but to finance the
greenfield investment with cash. This is a prediction that fits closely with the empirical
results of Fama and French (2005).
Pushing the logic further, our model also suggests that the relative preference for
doing the merger as a stock swap as opposed to financing the greenfield investment
with an SEO is stronger when the difference of opinion H among acquiring-firm investors
is greater, more target shareholders are asleep, and the scale of the two investments
is larger.
Although we do not test any of these hypotheses formally, we can offer a bit of
suggestive evidence. We have done some further comparisons of merger and SEO activity
that speak to scale effects. Based on a sample of 6,526 SEOs and 2,040 stock-for-stock
mergers of public companies over the period 1990–2003, we find that mergers raise
substantially more total equity financing than SEOs, $2,559 billion versus $968 billion.
Also, the largest transactions, both in absolute terms and relative to issuer size, are more
likely to be mergers than SEOs. For example, about 53% of stock-for-stock mergers are
for more than $100 million, while only 33% of SEOs are. Alternatively, 23% of mergers
are for more than 50% of the issuer’s market value, while only 10% of SEOs are. (These
numbers are based on an analysis of transactions in the Thomson Financial mergers and
acquisitions and seasoned equity offerings databases.)
14
A firm that considers its stock to be overvalued, and that wants to exploit this overvaluation, can be thought
of as needing to finance some of its investments with equity issues in this sense. Our model then explains why
stock-for-stock mergers are a more attractive option for such an overvalued firm than SEO-financed capital
expenditures, even when the assets ultimately acquired would be similar in either case. This contrasts with Shleifer
and Vishny (2003) who argue that stock-for-stock mergers are motivated by overvaluation but do not explain why
they are any better an outlet for overvalued acquirers than SEO-financed capital expenditures.
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298296
Again, we stress that these stylized facts do not represent a decisive test. They are also
consistent with other interpretations, such as mergers having an advantage over greenfield
investment when it comes to big projects in which time-to-build considerations are likely to
be important. Nevertheless, they do fit nicely with the model’s implication that the relative
appeal of a stock-for-stock merger is greatest when the firm’s growth plan is such that it
needs to raise a very large amount of equity financing.
5. Conclusions
Most people are reluctant to make active decisions. Instead, they tend to follow the path
of least resistance, accepting defaults. This has important implications for corporate
finance. Raising equity in an SEO requires investors to actively buy the shares of the
issuing firm. The default option is to buy no shares. By contrast, raising equity in a stock-
swap merger works well to the extent that target investors do not actively opt out of
holding the shares of the acquiring firm. The default option is to accept the shares of the
acquiring firm as consideration for shares in the target. We find that this sort of inertia is a
pervasive aspect of investor behavior. Individuals accept the default roughly 80% of the
time, and institutions accept it a third of the time.
In a classical stock market with horizontal demand curves, inertial behavior of this kind
would be irrelevant for prices. However, combining inertia with a downward-sloping
demand curve makes the price impact of a stock-swap merger relatively small, and the
terms for existing shareholders better as a result. We test this idea within the sample of
stock-swap mergers, using institutional ownership to proxy for low investor inertia.
Consistent with the theory, acquirer announcement returns are more negative when inertia
is low. The broader conclusion for corporate finance is that, when firms face downward-
sloping demand curves, as numerous studies suggest, stock-swap mergers could play a
particularly important role in supporting a strategy of rapid, equity-financed growth.
References
Akerlof, G.A., 1991. Procrastination and obedience. American Economic Review 81, 1–19.
Andrade, G., Mitchell, M., Stafford, E., 2001. New evidence and perspectives on mergers. Journal of Economic
Perspectives 15, 103–120.
Bagwell, L., 1992. Dutch auction repurchases: an analysis of shareholder heterogeneity. Journal of Finance 47,
71–106.
Barber, B., Odean, T., 2000. Trading is hazardous to your wealth: the common stock investment performance of
individual investors. Journal of Finance 55, 773–806.
Barberis, N., Shleifer, A., Vishny, R.W., 1998. A model of investor sentiment. Journal of Financial Economics 49,
307–343.
Bergman, N., Jenter, D., 2006. Employee sentiment and stock option compensation. Journal of Financial
Economics, in press, doi:10.1016/j.jfineco.2006.01.008.
Chen, J., Hong, H., Stein, J.C., 2002. Breadth of ownership and stock returns. Journal of Financial Economics 66,
171–205.
Choi, J.J., Laibson, D., Madrian, B., Metrick, A., 2002. Defined contribution pensions: plan rules, participant
choices, and the path of least resistance. In: Poterba, J. (Ed.), Tax Policy and the Economy. MIT Press,
Cambridge, pp. 67–114.
Choi, J.J., Laibson, D., Madrian, B., Metrick, A., 2004. For better or for worse: default effects and 401(k) savings
behavior. In: Wise, D. (Ed.), Perspectives in the Economics of Aging. University of Chicago Press, Chicago,
IL, pp. 81–121.
dx.doi.org/10.1016/j.jfineco.2006.01.008
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298 297
Constantinides, G.M., 1986. Capital market equilibrium with transactions costs. Journal of Political Economy 94,
842–862.
Daniel, K.D., Hirshleifer, D., Subrahmanyam, A., 1998. Investor psychology and security market under- and
over-reactions. Journal of Finance 53, 1839–1885.
Dierkens, N., 1991. Information asymmetry and equity issues. Journal of Financial and Quantitative Analysis 26,
181–199.
Fama, E.F., French, K.R., 1997. Industry costs of equity. Journal of Financial Economics 43, 153–193.
Fama, E.F., French, K.R., 2005. Financing decisions: who issues stock? Journal of Financial Economics 76,
549–582.
Greenwood, R., 2005. Short- and long-term demand curves for stocks: theory and evidence on the dynamics of
arbitrage. Journal of Financial Economics 75, 607–649.
Harris, L., Gurel, E., 1986. Price and volume effects associated with changes in the S&P 500 list: new evidence for
the existence of price pressures. Journal of Finance 41, 815–882.
Hodrick, L., 1999. Does stock price elasticity affect corporate financial decisions? Journal of Financial Economics
52, 225–256.
Hong, H., Scheinkman, J., Xiong, W., 2006. Asset float and speculative bubbles. Journal of Finance 61,
1073–1117.
Hong, H., Stein, J.C., 1999. A unified theory of underreaction, momentum trading and overreaction in asset
markets. Journal of Finance 54, 2143–2184.
Kahneman, D., Knetsch, J.L., Thaler, R.H., 1990. Experimental tests of the endowment effect and the Coase
theorem. Journal of Political Economy 98, 1325–1348.
Kahneman, D., Knetsch, J.L., Thaler, R.H., 1991. The endowment effect, loss aversion, and status quo bias.
Journal of Economic Perspectives 5, 193–206.
Kaul, A., Mehrotra, V., Morck, R., 2000. Demand curves for stocks do slope down: new evidence from an index
weights adjustment. Journal of Finance 55, 893–912.
Lang, L., Stulz, R., Walkling, R., 1989. Managerial performance, Tobin’s q, and the gains from successful tender
offers. Journal of Financial Economics 24, 137–154.
Madrian, B., Shea, D., 2001. The power of suggestion: inertia in 401(k) participation and savings behavior.
Quarterly Journal of Economics 116, 1149–1187.
Martinez-Jerez de, A.F., 2004. Shareholder value impact of accounting choice in business combinations: the role
of corporate governance. Unpublished working
paper, Harvard
University, Cambridge, MA.
Matvos, G., Ostrovsky, M., 2006. Cross-ownership and returns to shareholders in mergers. Unpublished working
paper, Harvard University, Cambridge, MA.
Miller, E., 1977. Risk, uncertainty and divergence of opinion. Journal of Finance 32, 1151–1168.
Mitchell, M., Pulvino, T., Stafford, E., 2004. Price pressure around mergers. Journal of Finance 59, 31–63.
Moeller, S.B., Schlingemann, F.P., Stulz, R., 2004. Do acquirers with more uncertain growth prospects gain less
from acquisitions? Unpublished working paper, Ohio State University, Columbus, OH.
Morck, R., Shleifer, A., Vishny, R.W., 1990. Do managerial objectives drive bad acquisitions? Journal of Finance
45, 31–48.
Myers, S.C., Majluf, N.S., 1984. Corporate financing and investment decisions when firms have information that
investors do not have. Journal of Financial Economics 13, 187–221.
O’Donoghue, T., Rabin, M., 1999. Procrastination in preparing for retirement. In: Aaron, H. (Ed.), Behavioral
Dimensions of Retirement Economics. Brookings Institution, Washington, DC, pp. 125–156.
Oyer, P., 2004. Why do firms use incentives that have no incentive effects? Journal of Finance 59,
1619–1649.
Scheinkman, J., Xiong, W., 2003. Overconfidence and speculative bubbles. Journal of Political Economy 111,
1183–1219.
Scherbina, A., 2004. Analyst disagreement, forecast bias and stock returns. Unpublished working paper, Harvard
University, Cambridge, MA.
Schwert, G.W., 2000. Hostility in takeovers: in the eyes of the beholder? Journal of Finance 55, 2599–2640.
Shefrin, H., Statman, M., 1985. The disposition to sell winners too early and ride losers too long: theory and
evidence. Journal of Finance 40, 777–790.
Shleifer, A., 1986. Do demand curves for stocks slope down? Journal of Finance 41, 579–590.
Shleifer, A., Vishny, R.W., 2003. Stock market driven acquisitions. Journal of Financial Economics 70,
295–311.
Stein, J.C., 1996. Rational capital budgeting in an irrational world. Journal of Business 69, 429–455.
ARTICLE IN PRESS
M. Baker et al. / Journal of Financial Economics 84 (2007) 266–298298
Thaler, R.H., 1980. Toward a positive theory of consumer choice. Journal of Economic Behavior and
Organization 1, 39–60.
Wurgler, J., Zhuravskaya, E., 2002. Does arbitrage flatten demand curves for stocks? Journal of Business 75,
583–608.
Zhang, D., 2004. Why do IPO underwriters allocate extra shares when they expect to buy them back? Journal of
Financial and Quantitative Analysis 39, 571–594.
- Corporate financing decisions when investors take the path of least resistance
Introduction
The model
Investor beliefs
Stock-for-stock merger when all target shareholders are awake
Stock-for-stock merger when some target shareholders are asleep
Empirical implications
Merger arbitrage
Capital gains taxes
Empirical analysis
Data
Measuring investor inertia
Individual investor inertia
Institutional investor inertia
The determinants of inertia
Acquirer returns and volume around merger announcements
Institutional ownership and acquirer announcement returns
Institutional ownership and acquirer announcement volume
Demand-curve and relative-size interactions
Further implications
Financing choice
Equity financing in mergers and SEOs
Conclusions
References
Journal of Financtal Economics 32 (1991) 263-192. North-Holland
The investment opportunity set and
corporate financing, dividend, and
compensation policies*
Clifford W. Smith, Jr. and Ross L. Watt
s
Lirirersi~~~ yf‘ Rocl~vrer. Rodwsrrr. !V Y 146_‘7. LJ,4
Received August 1990. final version received August 1992
We examine explanations for corporate financing-. dividend-. and compensation-policy choices. We
document robust empirical relations among corporate policy decisions and various firm character-
istics. Our evidence suggests contracring theories are more important in explaining cross-sectiona
l
variation in observed financial. dividend. and compensation policies than either tax-based or
signaling theories.
1. Introduction
To date. there has been little empirical analysis of the cross-sectional structure
of corporate financing, dividend, and compensation policies. Although much
effort has been devoted to developing the theory of these basic corporate
policies, empirical support for the models is largely anecdotal. Our primary
objective in this paper is to examine whether there are robust empirical relations
among corporate policy decisions and various firm characteristics. We believe
a more balanced interaction between theory and testing in corporate finance
will produce richer models and more powerful econometric methods of data
analysis.
A model of the cross-sectional variation in corporate policies requires speci-
fication of the exogenous variables that drive policy selection. Many potential
variables vary over time, but not across firms. For example, all firms have access
to the same contracting technology (e.g., sinking funds, dividend covenants,
Correspontlem~ to: Clifford W. Smith. Jr.. William E. Simon Graduate School of Business .Adminis-
tration. University of Rochester. Rochester. NY 11677. US;\.
*This work has been partially supported by the Simon School’s Bradley Policy Research Center.
We thank Amy P. Sweeney for computational assistance. Previous drafts of this work uere titled:
‘The Structure of Executive Compensation and the Control of Management’.
0304405X 92 SOj.00 ( 1992-Elserier Science Publishers B.V. .All rights reserved
executive stock options, cancelable leases). And. given well-functioning labor.
capital. and product markets, all firms have access to any potential stockholder,
bondholder. manager. lessor, or customer: they all have access to individuals
with different risk preferences or personal tax rates. Thus. neither personal ta
x
provisions, risk preferences, nor the contracting technology appear able to
explain observed cross-sectional variation in corporate policies.
In making investment and employment decisions. however, firms invest in
specialized physical and human capital. These firm-specific investments result in
variation in firms’ investment opportunity sets (i.e., their prospective investment
opportunities and associated payoff distributions). Corporate taxes and regula-
tion also vary across firms. Thus, the investment opportunity set, regulation,
and corporate tax provisions offer the potential to explain cross-sectional policy
variation. These are the explanatory variables we use in our analysis.
Of course, aspects of these variables are endogenous. For example, regulation
and tax policy are determined within the political process, and we observe
innovation in both the real investment activities of firms as well as the contracts
they employ. Our statistical analysis, however. requires only that these factors
be predetermined, not that they be ccmpletely exogenous.
Using industry-level data from 1965 to 1985 we find that measures of the
firm’s investment opportunity set (such as the availability of growth options and
firm size) are related to its financing. dividend. and executive-compensation
policies. In particular, we document that firms with more growth options (i.e.,
greater access to positive net present value projects) have lower leverage,’ lower
dividend yields [see also Rozeff (1982)]. higher executive compensation, and
greater use of stock-option plans. We also find that regulated firms have higher
leverage, higher dividend yields, lower executive compensation, and less fre-
quent use of both stock-option and bonus plans. Finally, we find that larger
firms have higher dividend yields and higher levels of executive compensation
[see also Fox (1986) and Murphy (1985)].
These relations imply associations among the corporate policies themselves.
Our evidence indicates positive associations between leverage and dividend
yield and between compensation and the use of both bonus and stock-option
plans. Negative associations are documented between leverage and compensa-
tion, bonus and stock-option plans, as well as between dividend yield and both
bonus and stock-option plans.
Our empirical analysis includes a broader range of investment-policy charac-
teristics than previous studies.’ and focuses on the partial effects of each
exogenous variable (i.e., holding the other variables constant). We also relate
firm characteristics not just to a single corporate policy choice but to financing,
‘See also Ferri and Jones I 1979). Castamas (1983). BradIs). Jarrsll. and Kim (198-I). Long and
Malitz (1985). and Titman and Wessels (1988).
LSee footnote I: also Rozeff (19921.
dividend. and compensation policies. In this way, we help control for potential
sources of spurious correlation that can be troublesome if a single corporate
policy is examined in isolation.
Other research confirms the empirical relations documented in this paper.
Gaver and Gaver (1993) use firm-level data and measure growth options by the
frequency of a stock’s inclusion in growth-stock mutual funds. Holthausen and
Larcker (1991) use firm-level data supplemented by confidential firm-level
compensation data. Kale (1991) uses firm-level data on compensation plans to
investigate the variation in the board of directors’ authority to award stock or
stock options to management.
In section 2 we describe our data and the instrumental variables used to
measure corporate financing, dividend, and compensation policies as well as our
independent variables. We also discuss our empirical methods. In section 3 vve
predict the empirical relations between these policies and investment-opportun-
ity-set, size, regulation, and tax variables, and present evidence on the estimated
relations. In section 4 we examine the implications of our analysis for the
relations among financing. dividend, and compensation policies. We present our
conclusions in section 5. The appendix contains sensitivity analysis and provides
evidence on the robustness of our results.
2. Data and empirical methods
Investigating the empirical relations among the investment opportunity
set. regulation. and firm size on the one hand, and firms’ financing. dividend.
and executive-compensation policies on the other, requires a wide range of
data. some of which (especially compensation data) can be difficult to obtain.
Data on executive compensation and the use of formal incentive plans are
available by industry in the Conference Board surveys of executive compensa-
tion. We use the Conference Board survey data for every fourth year from
1965 through 1985 as reported in Fox (1966, 1970, 1974, 1978. 1982. 1986).
Because this compensation data is available at an industry level only. we
estimate investment-opportunity-set. financing-policy, and dividend-policy
variables for each of Fox’s industry definitions for each year in the study
using annual firm data for a sample of C’ornpnst~t firms chosen to match
the firm-size attribute Fox reports (which is typically industry sales). We
then generate industry-level data by averaging data on individual firms sorted
by industry. The use of industry-level data should reduce measurement error in
the variables if Fox’s classification of industries using SIC codes effectively
groups firms by the nature of their investment opportunity set. It should also
maintain dispersion among the variables. We describe how we assemble our
data and match the compensation data with other data in section A.1 of the
appendix.
In this section. we describe measures used in the empirical analysis. Relatively
accurate financing, dividend. compensation, regulation, and firm-size measures
are available. but measures of the investment opportunity set involve substantial
measurement error. We attempt to address this problem by using several
alternate measures, as well as by using an instrumental-variables approach and
testing the specification of the relations among the measures.
Firmnciny policy. A firm’s financing policy is represented by its equity-
to-vnlue t-do (E,‘V). The equity-to-value ratio for industry i in year T is
calculated using four years of data:
(1
)
T = 1965, 1969, 1973, 1977, 1981, 1985,
where Ni, is the number of sample firms in industry i with data available in year
t, and kj, is the proxy for the market value of firm j at the end of year C. C$ is
equal to the market value of firmj’s equity at the end of year t (Ejt) plus the book
value of its assets at the end of year f (Ai,) minus the book value of its equity at
the end of year r.
Dicihd policy. A firm’s dividend policy is represented by its dicidenrf Jield
or dividend-to-price ratio (D/P). The dividend yield for industry i in year T is
calculated as
T = 1965, 1969, 1973, 1977, 1981, 1985.
where Nit is the number of sample firms in industry i with data available in year
t. D, is dividends per share for firm j in year t, and Pjl is firm j’s share price at the
end of year r.
Compensarion. We use the CEO’s s&r?; as a surrogate for management
compensation. Since this surrogate ignores compensation under incentive plans,
it measures compensation with error. Ignoring incentive compensation prob-
ably reduces the likelihood of observing any relation between the investment
opportunity set and compensation. however, since it reduces the variation in the
measured level of compensation. We adjust the median CEO salary for each
industry-year using the GNP deflator. The log of the resulting median CEO real
salary is used to measure compensation.
L:sr qf‘incenri~e plnns. The variables for the use of incentive plans are the
percentage of firms in each industry with bonus plans and the percentage with
stock-option plans. (Fox also reports the use of stock appreciation rights. but
only for years after 1977.) Data on the combination of plans (for example. the
percentage of firms in each industry with at least one incentive plan) are not
available in Fox.
7 ’ E.yogenous wriddes _._.
Inresrment opporrunity ser. The primary variable used in this study as
a proxy for the investment opportunity set is the ratio of book value of assets to
firm value (A’Q The book value of assets (Aj,) is used as a surrogate for assets in
place. We predict that the higher A ‘r. the higher the ratio of assets in place to
firm value. and the lower the ratio of the value of investment opportunities to
firm value. Since book value is historical cost less depreciation, it contains
potentially significant measurement error for firms with long-lived assets.
The ratio is calculated for each industry i in year T using four years of data:
T = 1965, 1969, 1973, 1977, 1981, 1985,
where .Vi, is the number of sample firms in industry i with data available in year
f. Sensitivity analysis using other investment-opportunity-set measures is re-
ported in section A.2 of the appendix.
Reyuhion. We use dummy variables for regulation. We consider the insur-
ance, gas and electric utility, and banking industries as regulated and the other
thirteen industries as unregulated. In our base-case regressions, we use a single
intercept dummy variable for regulation. though the effect of regulation on
policy choices probably varies across the three industries. (See the section A.2 of
appendix for an examination of the sensitivity of the results to different dummy
variables for each industry.)
Firm size. Strictly speaking. firm size is an endogenous variabie.that de-
pends on economies of scale in both production and organization of the firm.
Size is thus a function of the investment opportunity set. Yet given our limited
knowledge of the determinants of size. we include size itself as an exogenous
variable. We measure firm size by the log of the Cotnpustar sample’s median real
sales for each industry year (i.e., 1965. 1969. 1973. 1977, 1981. or 1985) for the
unregulated industries. We use the GNP deflator to restate nominal sales from
Cotnpustm so that we measure size in constant dollars. For the regulated
industries, we measure size by median real premium income in the insurance
268 C. W’. Smith. Jr. und R.L. Nutts. Fimtnctng. rliridtwct. und compunsutton policirs
industry, median real operating income in the utilities industry, and median real
worldwide deposits in the banking industry. The use of a different size measure
in the regulated industries introduces noise in the size measure, but it does not
appear to introduce any bias (although we investigate this possibility in section
A.2 of the appendix).
Accounring return. We add the accounting return as an additional indepen-
dent variable in the compensation regression. The mean accounting return for
industry i in year t is calculated using annual accounting data:
Rj, = (Ofj, + IN Tj,),!vjl_ 1 (
where for sample firmj in year t Rj, is the accounting return, 01, is the operating
income, and IlvTj, is interest expense. We obtain the mean return by averaging
R, over the sample firms in the industry and over four years (the contempo-
raneous and three previous years). We include this variable because CEO
compensation varies with performance [see Murphy (1985)].
2.3. Empirical methods
We pool cross-section and time-series observations and regress the various
policy variables (financing, dividend, compensation, and incentive plans) on
measures representing all three exogenous variables (investment opportunity
set, regulation, and size). We also regress the policy variables on exogenous
variables separately each year. Since a simultaneous system of equations under-
lies the data, our estimated parameters are thus reduced forms. not structural
parameters.
The regressions are over industry-years. The two regressions with dependent
variables obtained from Compustnt (E/V and D/P) are estimated over 93 obser-
vations (insurance and banking are unavailable for 1965). The three compensa-
tion regressions are estimated over 9 1 observations (construction is unavailable
for 1965, 1969, and 1973).
Speci’cation tests. Two diagnostic tests are used in all regressions: the White
(1980) specification test and a test for nonlinearities. The White test indicates
whether the regression errors are heteroskedastic or if the errors and explana-
tory variables are (nonlinearly) dependent. Although White’s specification test is
valid asymptotically, its accuracy in small samples is more conjectural. For our
regressions, the White chi-square test of first- and second-moment specification
generally shows that the null hypothesis of no misspecification can be rejected at
the 0.0001 level. The White procedure also generates a variance-covariance
matrix of coefficient estimators that converges to the true variance-covariance
matrix in large samples. This gives the opportunity to produce test statistics
that have the right size. The White asymptotic standard error for our estimated
coefficients is typically lower than that from an ordinary-least-squares
regression. so the significance of the coefficients generally increases if the White
standard error is used to calculate the r-statistic.
To test for nonlinearities, we sort the residuals by the values of each continu-
ous explanatory variable and calculate a Durbin-Watson statistic. Nonlineari-
ties for an explanatory variable show up as correlated errors. (We assume that,
except for the influence of nonlinearities, regression errors are cross-sectionally
independent.) Firms tend to follow the same financing. dividend. and compensa-
tion policies over time. When we run cross-sectional regressions separately for
each year, however, none of the Durbin-Watson statistics is significant. Hence,
there is no evidence of significant cross-sectional dependence.
3. Theory and evidence
We discuss each policy variable in turn, first developing predictions about the
relation between the policy and the exogenous variables from contracting,
tax-based, and signaling theories. We then examine evidence from our regres-
sion results. Table 1 summarizes the contracting-hypothesis predictions and
reports the estimated regressions for the various policy variables based on
observations pooled over time and across industries.
If there are multiple partial effects in the estimated coefficients, we are unable
to separate them without additional structure. For example, contracting ar-
guments imply that firms with more growth options should have lower debt in
their capital structure. whereas signaling and tax effects imply higher debt. If the
estimated relation between growth options and leverage is significantly negative,
we conclude that the contracting effect is significant. whereas if the estimated
relation is positive. we conclude that the combination of signaling and tax effects
is significant. Since we estimate only the net effect. we cannot separately identify
the significance of less important partial effects. In most cases in which two
explanations lead to predictions for the sign of the relation between a variable
and a policy (e.g.. the proportion of assets in place and dividend yield), the
predictions are for opposite signs, so we can reject one of them.
Also, we do not specify any interdependencies among policies. For example,
we predict that firms with more growth options (fewer assets in place) use stock
options more frequently because management is more difficult to monitor in
such firms. We do not allow for the possibility that the management of firms
using stock options increases leverage to increase the value of the options by
raising equity volatility. To sort out these partial effects we would have to
develop and estimate a structural model specifying the nature of the interdepen-
dence. Titman and Wessels (1988) follow such an approach and impose a com-
plex structure on the estimated relations amon, 0 variables. If the structure they
use is correct, the power of their estimates is increased, but if their structure is
incorrect, they impose bias. Given our current knowledge of these empirical
9
-l
1
.3
-I
1
7
.0
3
0
I
1
.7
7
1
3
.1
I
O
.I
I)
0
.5
I
0
7
8
7
.4
7
O
.?
O
_
2
.5
1
_
~
0
. I
’)
3
.0
3
_
–
0
.2
s
1
2
.3
3
+
0
.0
I
6
.
I
4
t
-~
0
.1
7
_
9
.W
_
–
0
.1
3
–
X
.7
9
–
0
.4
h
I7
.3
4
_
~
0
.0
3
—
3
.4
3
~
,
il
.0
I
1
.2
0
+
0
.
I7
0
.4
7
+
o
.o
fl
2
.1
3
+
0
.0
3
2
. I-
l
+
I.
5
0
2
.x
0
0
.7
’)
1
1
4
.5
3
0
.W
)
I
0
.0
I
0
.0
2
0
.4
x
0
.7
8
so
.z
o
0
.0
0
0
I
3
x
3
0
0
.0
0
0
I
IO
i.
9
3
O
(H
W
)I
relations. we believe progress is better served by documenting robust empirical
relations between policy parameters and exogenous variables before attempting
to subdivide the relations into component effects.
3. I. Fintmcing policy.
Conrractiny hypotheses. There are several contracting arguments relating to
financing policy. Myers (1977) describes the firm’s potential investment oppor-
tunities as call options whose values depend on the likelihood that management
will exercise them. If the firm has risky debt outstanding. situations arise in
which exercising the option to undertake a positive net present value project
potentially reduces share value because debtholders have a senior claim on the
project’s cash flows. Unless this conflict between the shareholders and debthol-
ders is controlled. the probability that these real investment options will be
exercised is reduced, thereby reducing firm value. One way to control this
underinvestment problem and its associated value loss is to finance growth
options with equity rather than debt. Hence Myers predicts that the larger the
proportion of firm value represented by growth options (i.e., the lower the assets
in place), the lower the firm’s leverage. and the higher its equity-to-value ratio.
Regulation also controls incentive problems between stockholders and fixed
claimholders by reducing discretion over the firm’s projects. Hence regulated
firms are predicted to have lower equity-to-value ratios. Jensen (1986) suggests
that firms with more free cash flow choose higher levels of debt in their capital
structure as a credible precommitment to pay out the excess cash. If firms with
more growth options have less free cash flokv. this analysis also predicts a nega-
tive relation between assets in place and equity-to-value.
Tax hypotheses. Progressivity in the tas structure implies that greater vola-
tility in taxable income raises the firm’s expected tax liabilities [Smith and Stulz
(198.5)]. If firms with more growth options have more volatile cash flows, they
have incentives to reduce the amount of debt in their capital structure over the
range of progressivity. Hence this tax effect implies a negative relation between
the proportion of assets in place and the equity-to-value ratio.
Other tax-code provisions, however, potentially affect financing policy differ-
ently. DeAngelo and Masulis (1980) argue that firms that generate substantial
noninterest tax shields. such as investment tax credits, have a comparative
disadvantage in using interest tax shields and thus should have less debt in their
capital structures. If capital-intensive firms are more likely to generate invest-
ment tax credits than firms whose value derives largely from growth options,
such firms should have lower equity-to-value ratios.
Siynaliny hypotheses. A substantial literature examines the impact of in-
formation asymmetries on financing policy. but most of it does not attempt to
develop implications for cross-sectional variation in leverage. Analyses such as
that of Myers and Majluf (1984) focus on explaining stock-price reactions to
announcements of security offers. Cross-sectional implications for financing
policy are not apparent.
The information asymmetry models that do have potential implications for
cross-sectional variation in firms’ policy choices are signaling models. For
example. Ross (1977) develops a signaling model that examines the relation
between leverage and firm quality. holdin g any information disparity fixed.
Issuing debt in his model is a signal of high quality because the firm exposes
itself to the costs of financial distress. Therefore, high-quality firms choose
higher leverage.
Yet in this signaling literature, quality is not defined in terms of observable
variables. To derive testable implications, we assume that with no information
disparity there is no incentive to signal, and that the greater the information
disparity. the greater the derived demand for signaling. We also assume that if
the costs of signaling vary, they are less sensitive than the benefits of signaling to
variation in the size of the information disparity. With these assumptions, the
signaling analysis implies that if firms with more growth options face greater
information disparities. they should be high-debt (i.e., low equity-to-value) firms.
Also, if regulated firms have lower information disparities, they should be
low-debt firms.
Size hypotheses. If costs of financial distress limit leverage, the greater
diversification (and consequent lower return variance) of larger firms enables
them to have higher leverage than smaller firms. We therefore predict a negative
relations between the equity-to-value ratio and firm size.
Regression results. In table 1, the coefficients of asset/value, regulation, and
firm size from the regression are all reliably negative and the regression itself is
significant at the 0.001 level. The Durbin-Watson statistics when the obser-
vations are ranked on the basis of A,‘V and log of real sales suggest there is no
significant departure from linearity for those independent variables. Hence, the
evidence from the regression is consistent with E,‘V being a reliably negative
function of assets in place in the investment opportunity set and a reliably
negative function of regulation. Both negative functions are consistent with our
contracting arguments and inconsistent with the signaling hypothesis. The
negative relation between EiV and assets in place is also consistent with the
progressive-tax effect, but inconsistent with the effects of investment tax credits.
That E I’ is a negative function of firm size is consistent with costs of financial
distress limiting leverage.
3.1. Dick/end policy.
Cormvcriny hypotheses. The firm’s cash-flow identity links investment and
dividend policy: the greater the amount of investment during the period, the
smaller the dividend or the more the new equity issued. Jensen (1986) argues that
firms with more growth opportunities have lovver free cash flow and pay lower
dividends. Hence. there should be a positive relation between the proportion of
assets in place and dividend yield. Tvvo contracting arguments reinforce this
predicted relation. First, Rozeff (1982) and Easterbrook (1951) argue that the
new-issue market lowers agency costs by providing effective monitoring. Firms
with fewer growth options would go to the new-issue market less frequently and
forego this benefit if they pay fewer dividends. Second. dividend covenants that
specify a maximum on payouts effectively impose a minimum investment
requirement [see Smith and Warner (1979) and Kalay I 1982)], thereby reducing
the underinvestment problem. The more binding the dividend constraint, how-
ever. the more likely it is that managers will be forced to undertake negative net
present value investments (although this cost is eliminated if firms can invest in
financial assets that offer normal returns). Firms with more profitable invest-
ment options can tolerate more restrictions on dividends before the expected
benefits of controlling payout are offset by the expected cost of forced negative
net present value investments. Hence firms with more growth options (i.e., lower
assets in place) are expected to pay lovver dividends.
Smith (1986) argues that the regulatory process gives managers of regulated
utilities an incentive to pay higher dividends in order to force the utility to raise
funds more frequently in the capital market. New issues provide evidence on the
firm’s cost of capital that is useful in the regulatory process. Without such evidence,
the utility commission faces fewer constraints in reducing the firm’s rate of return.
Higher dividends thus discipline the regulators as vvell as the firm’s managers. In
addition, some regulatory authorities still set required returns by using a form of the
dividend-growth model. whereby higher dividend payments raise allowed rates of
return. Therefore. we expect that regulated firms pay higher dividends.
T(I.Y h~porhrsr.s. In the dividend literature we find no tax analysis that can
explain cross-section variation in dividends. An important reason for this lack of
explanatory power is the endogeneity of personal shareholder tax rates. For
esample. Litzenberger and Ramasvvamy ( 1982) argue that since dividends are
taxed at higher effective personal rates than capital gains, higher-dividend firms
generate higher expected personal tax liabilities and thus require higher ex-
pected before-tax returns. But since all firms have access to potential sharehol-
ders of the various tax brackets. the shareholder tax rate is endogenous. Thus,
the Litzenberger and Ramaswamy analysis has no implications for cross-sec-
tional corporate dividend policy choice.
Siyrl‘l/irl~/ Ii \~porllrs~.s. Bhattacharya (1979) develops a signaling model in
vvhich he argues that high-quality firms pay high dividends. Again. if the signal
increases with the information disparity betvveen managers and investors, firms
with greater information disparities (typically unregulated firms with more
growth options) should pay higher dividends.
Si:r I~~.pothrsis. To make the regression comparable to the other base policy
regressions. the log of real safes is included in this regression although we have
no reason to espect firm size to affect dividend policy.
Rryrrssion resulrs. The regression evidence in table 1 indicates that the
estimated coefficients of regulation and assets, value are both reliably positive.
This evidence is consistent with the contracting predictions and inconsistent
with the signaling predictions. The size coefficient also is reliably positive in the
regression. The overall regression is significant and again there is no evidence of
nonlinearity for .-I L’ or the log of real sales.
3.3. Cort~pensutiotl
Cotltroctitly hypotheses. We hypothesize that the marginal product of in-
vestment decision makers is greater than the marginal product of supervisors
and good decision makers are less numerous than good supervisors.3 Therefore.
the larger the proportion of firm value represented by growth options, the
greater the manager’s compensation.
Regulation restricts the manager’s investment discretion and reduces the
marginal product of the decision maker, so regulation should reduce the level of
compensation. Some regulatory authorities appear to regulate compensation
policy directly. placing limits on payments to executives. Such limit should
increase perquisite consumption. This presents a potential problem in our
empirical analysis. since consumption of perquisites is difficult to measure. If
managers of regulated firms consume more perquisites than managers of un-
regulated firms. our evidence overstates the difference in compensation between
the two groups. We also include the accounting return in this regression because
we expect that CEO compensation varies with performance [see Murphy
(19X5)].
Size It~~pothrs~.s. In general. the larger the firm. the larger the stock of real
resources that can be affected by a given managerial decision. Managers of
larger firms thus have a higher value added. so we expect higher compensation
for executives of larger firms.
Rryrmion results. Coefficients on all four independent variables are signi-
ficant in the compensation regression in table 1. The coefficients of Ai V and
regulation are negative and the coefficients of the log of real sales and the
accounting return are positive. The regression is significant at the 0.0001 level,
and there is no evidence of nonlinearity for A/V, the log of real sales, or
accounting return. The growth options. regulation, and firm-size results are
JWe expect that this elTect will be reinforced by managers’ compensating differenttals for risk.
Given risk-a\erre manayers with firm-specific human capital who cannot completeI! diversify their
compensation risk. the higher the firm’s risk. the higher the risk of the manager’s compensation. and
the higher the managers’ equilibrium compensation. We expect that. as an empirical proposition, the
larger the proportion of firm value represented by growth options. the greater the tirm’s risk. Also.
we argue below that the lamer the proportion of firm value represented b! grouth options, the more
likely it is that the manape:s compensation is tied to firm Lalue and the greater the variance of the
manager’s compensation. We believe. ho\ve\er. that these compensating ditferenttals for risk are
secondar). See sectton ,A.? of the appendix.
consistent with our contracting predictions. The accounting-return result is
consistent with previous evidence that CEO compensation varies with firm
performance. Although increased perquisite consumption by CEOs of regulated
firms potentially explains the difference in compensation between regulated and
unregulated firms, perquisite consumption cannot explain our reported financ-
ing-policy or dividend-policy coefficients for regulated firms.
3.1. Iflcmtice comprnsution
Contracting hypotheses. The typical problem analyzed in the principal-
agent literature is that of a risk-neutral principal attempting to induce a risk-
averse agent to take the action the principal would take.’ If the principal can
observe the agent’s actions. the optimal contract pays the agent a fixed wage and
penalizes him for taking suboptimal actions: that contract imposes all the risk
on the risk-neutral principal. If the principal cannot observe the agent’s actions.
the optimal contract gives the agent a share in the outcome of his actions. That
contract provides an incentive to expend effort to achieve the principal’s objec-
tive, thus justifying the increased compensation of the agent for bearing the
additional risk.
When we apply this principal-agent analysis to large firms, shareholders are
considered risk-neutral because they can diversify firm-specific risk. If managers
cannot effectively diversify the risk of their compensation payments, they are
risk-averse in their actions. We suggest that managers’ actions are less readily
observable if the firm has more investment opportunities. It is difficult for
shareholders or outside board members who do not have the manager’s specific
knowledge to observe all the investments from which the manager chooses. In
general. the larger the proportion of firm value represented by growth options,
the more likely that the firm ties compensation to the effect of the manager’s
actions on firm value.
This linkage does not by itself imply the use of formal incentive plans. The
manager’s salary could be informally renegotiated periodically on the basis of
previous performance. But the effectiveness of future salary renegotiation de-
pends on expected future employment (e.g.. a 64year-old manager facing
retirement at 65 would be little motivated by an annual salary-renegotiation
scheme). It also depends on the degree to which the renegotiation promise is
bonded. Informal salary renegotiation is less effective if there is higher manage-
ment turnover and thus less reason to expect future managers to honor unwrit-
ten. informal contracts. These problems encourage the use of explicit incentive
plans that tie the manager’s compensation to a performance measure that
reflects the effects of the manager’s actions on firm value (e.g., stock price or
accounting earnings). Hence the larger the proportion of firm value represented
“For a suney of this literature see \lacDonald I 1984).
276 C. I+.. Smrrh. Jr. atrd R. L. N’uirs. Finuncing. dirldmd. and ~vnrprnsurron policies
by intangible investment opportunities, the more likely the firm is to have
a formal incentive compensation plan. We thus expect a negative relation
between the proportion of assets in place and the use of stock-option plans.
More growth options are likely to make accounting numbers poorer mea-
sures of performance. For example, Rao (1989) provides evidence that most of
a start-up firm’s value is represented by investment opportunities. The impact of
managers’ actions on those opportunities is not accurately measured by ac-
counting numbers. This effect should reduce the use of accounting-based incen-
tive plans and offset the incentive of firms with more growth options to use
incentive compensation plans. Thus. the relation between the proportion of
assets in place and the use of formal accounting-based incentive plans is
ambiguous.
If regulation restricts the investment opportunity set and makes observation
of the manager’s actions easier, regulated firms are less likely to use formal
incentive plans.
Tas hypotheses. In addition to these contracting arguments, taxes are po-
tentially important in determining the use of incentive compensation plans.
Miller and Scholes (1982) show that incentive compensation plans frequently
contain a deferral aspect that is attractive only if the executive’s effective tax rate
is higher than that of the corporation (as happened during the period
1965-1985). This hypothesis also has implications for the compensation policy
of banks and insurance companies. Firms in these industries are allowed to
receive tax-exempt income from municipal bonds while deducting interest paid
on CDs or indemnity payments to policyholders. Thus, banks and insurance
companies face lower effective tax rates and should use incentive compensation
provisions more frequently. Yet these industries use incentive compensation less
frequently.
Size hypothrses. Given fixed costs and scale economies in the administration
of incentive compensation plans, such plans should be observed more often in
large firms. Eaton and Rosen (1983) argue that the problems of monitoring
management increase with firm size. Also, Christie, Joye, and Watts (1989) offer
span-of-control arguments and present evidence that larger firms are more likely
to decentralize, implying that large firms employ incentive contracts more
frequently [see Smith and Watts (1982) and Sloan (1993)]. We thus expect a
positive association between firm size and the use of incentive compensation plans.
Regression results. Consistent with contracting predictions, the r-statistic in
table I from the bonus-plan regression shows that the coefficient on the regula-
tion dummy is reliably negative. The coefficient of A;‘V is positive and significant
in the regression, which is consistent with the hypothesis that accounting
numbers are less useful as performance measures for firms with growth oppor-
tunities. The coefficient of the log of real sales also is positive and significant.
All three coefficients (A,‘C’, regulation. and the log of real sales) have the signs
predicted by contracting arguments and are significant in the stock-option
regression. The regression is significant at the 0.0001 level, and there is no
evidence of nonlinearities for A V or the log of real sales. Generally. the results
suggest that the existence of a stock-option plan is a reliably negative function of
both regulation and A C*.
3.5. Sensiticitj~ ancal~~sis
In general, the regression results for financing policy. dividend policy, com-
pensation. use of bonus plan. and use of stock-option plan are consistent with
the contracting predictions. In contrast. the results for the two policies for which
the signaling hypothesis has predictions (financial and dividend policy) are
inconsistent with those predictions. Taxes could explain the NC’ coefficient in
the financial-policy regression via an association between A/V and cash flow
variance. but contrary to tax implications, banks and insurance firms have
higher equity-to-value ratios and less frequently use incentive compensation
plans. Overall, the evidence is more consistent with the contracting hypothesis
than with the signaling or tax hypotheses.
In the appendix we examine the robustness of our results. In particular, we
investigate alternate investment-opportunity-set variables, sensitivity to prob-
lems with the regulatory subsample. the time-series stationarity of the
relations.
and positive dependence in corporate policies. The evidence in the appendix
indicates that the results in table 1 are robust to alternate measures of invest-
ment opportunities. When we estimate the policy regressions without the regu-
lated industries. the general tenor of the results is unchanged. When we allow the
effect of regulation to vary across regulated industries. our basic results for the
assettvalue, size, and accounting-return variables do not change. The estimated
coefficients are relatively stationary over time, especially the financing policy,
dividend policy, and stock-option-plan relations. Finally, positive dependence
in corporate policies is a problem for the regulation variable only. Even then. the
regulation coefficients are still significant when cross-sectional variation and not
time-series variation is used to estimate the corporate policy relations.
Gaver and Gaver (1993) provide an independent test of the robustness of the
results. Using individual-firm data rather than industry data and different
investment-opportunity-set variables. they replicate the results presented in this
paper.
4. Relations among policies
If contracting theories are more important than signaling or taxes in explain-
ing cross-sectional variation in corporate policy choices, we should observe
predictable relations among policies. For example, we expect a negative relation
between E; bP and D, P because the larger the proportion of firm value
represented by growth options, the higher the firm’s equity-to-value ratio and
the lower its dividend yield. Table 2 contains predictions on these policy
relations.
Under the contracting argument, firms with more growth options have less
debt because of the more severe incentive problems associated with debt [Myers
(1977)] and because they have less use for debt as a creditable commitment to
distribute excess cash flow. These firms have less incentive to use dividends to
subject themselves to the discipline of the new-issue market when their invest-
ments create demand for new capital. Signaling models reinforce the prediction
of a positive association between leverage and dividend yield, since high-quality
firms should choose both high leverage and high dividends. Regulatory restric-
tions on investment reduce incentive problems associated with debt and so
encourage regulated firms to have higher leverage. Regulated firms also have
incentives to pay higher dividends and thus discipline regulators through their
more intensive use of capital markets.
For the full sample, the estimated correlation between the ratio of equity to
value and dividend yield is negative, as predicted, and is significant. (The table
2 results are generally unaffected by using the Spearman rank-order correla-
tion.) When the regulated industries are excluded, the absolute value of the
estimated correlation between E,‘C’ and dividend yield falls from 0.49 to 0.33,
suggesting that regulation reinforces the negative relation between E/ C’and D,sP.
When we estimate the E/V and D,P regressions excluding the regulated indus-
tries, the t-statistics for assets,‘value are still highly significant, but the t-statistic
for firm size is insignificant in the leverage regression. This evidence suggests
growth options are responsible for much of the correlation between leverage and
dividend yield in the unregulated sample.
We predict a positive relation between E; V and compensation. Contracting
arguments suggest managers of firms with more growth options are paid more
because of their greater marginal product. In table 2 the relation between E,‘C’
and compensation is reliably positive for both the full and unregulated samples.
Regulation should reinforce this positive relation because it reduces compensa-
tion by reducing the manager’s marginal product. In fact, the absolute correla-
tion between E ‘C’and compensation in table 2 is less for the unregulated sample
than for the full sample (0.50 versus 0.70). The table 1 regressions for financing
and compensation both indicate significant effects of size. Since the estimated
size coefficients show the opposite signs. however. size effects would imply
a negative correlation between E,‘V and compensation, not the observed posi-
tive correlation.
We predict E ‘If to be positively related to the use of stock-option plans. Firms
with more growth options are more likely to use stock-option plans because the
Table 2
Unconditional relations among financtng-policy. dividend-policy. and compensation-policy vari-
ables for 16 industries and 13 unregulated industries. 1963-1985.
Compensation policy
f iwIcin(/ polic!
Equit) value ratto
Predicted sign
;\I1 industries
Cnregulated Industries
DirtJrftd policy,
Dtvidend yield
Predicted sign
All industries
Unregulated industries
Cor~rprrrscztim policy
Log of real salark
Predicted sign
,411 industries
Unreylated industries
Di\ idend policy Use of incentive plans
Log of real
Di\ idend yield salary Bonus Stock-option
7 –
– 0.19” Of70” 0.61’ Ot73”
– 0.33 0.50h – 0.05 O.jY
‘1 – –
– 0.19 – b.68’ – 0.64”
0.32 0.03 0.1 I
.1
b.W Ot70”
0.01 0.5gb
“Significant at the I”0 Iebel (one-tailed test).
*StSnifcant at the IO”0 Ie\cl tone-tailed test).
‘Significant at the IO~O Ietel (two-tailed test).
“SiSniticant at the IO” U Ic\el ttuo-tatled test).
manager’s actions are less likely to be observable. Regulation is expected to
reduce the use of incentive plans and so reinforce the expected positive relation
between E,‘C7and stock options. In table 2 the correlations between E,:Vand the
use of stock-option plans are reliably positive for both the full and regulated
samples. Consistent with the reinforcement effect of regulation, the association is
stronger for the full sample than for the unregulated sample. The estimated
correlation between E I/ and bonus plans is close to zero for the unregulated
sample. but for the full sample it is reliably positive, consistent with the
implications of the effect of regulation.
On the basis of contracting arguments. we predict dividend yield to be
negatively associated with both compensation and the use of stock-option plans.
since we expect firms with more growth options to have lower dividend yields
and higher compensation. and to use stock-option plans more often. Regulation
should reinforce these predictions, since we expect it to increase dividend yield,
reduce compensation. and reduce the use of incentive plans. For the full sample,
the estimated correlations in table 2 have the expected sign. The correlation
between dividend yield and stock-option plans is reliably negative, but the
correlation between dividend yield and compensation is insignificant. U’hen we
exclude regulated firms the estimated correlations are insignificant and have the
wrong sign, suggesting that the estimated associations for the full sample are due
to the regulatory effect. Similarly. the use of bonus plans is significantly nega-
tively correlated with dividend yield for the full sample but not for the un-
regulated sample, which is again consistent with the effect of regulation on the
use of incentive plans.
4.3. Correkitions among c~on~pmsiition cmiuhles
Finally, we predict that compensation is positively correlated with the use of
stock-option plans. We expect firms with more growth options to have higher
compensation and to use stock-based incentive plans more often. Regulation
should reinforce this expected relation. As predicted. the estimated correlation
in table 2 is reliably positive in both the full and unregulated samples, but less so
for the unregulated sample. The estimated correlation between compensation
and the use of bonus plans is also reliably positive for the full sample. but not for
the unregulated sample, again suggesting regulation is important in reducing the
use of bonus plans.
The predicted and observed correlations among the policies in table 2 are
generally consistent. When the regulated industries are included, each correla-
tion has the predicted sign and regulation has the predicted directional effect on
the correlation. We believe these unconditional correlation results are important
in interpreting the results of many empirical studies. For example, Ang and
Peterson (1981) examine tradeoffs between leasing and debt. They find that firms
that issue more debt tend to engage in more leasing. Smith and Wakeman (1985)
argue that this result should not be surprising; although leasing and debt are
substitutes for a given firm, when investment opportunity sets provide high debt
capacity they also tend to provide more profitable leasing opportunities. To
measure the extent of substitutability between leasing and debt, differences in
investment opportunities must be controlled. Lambert, Lanen, and Larcker
(1989) find evidence that the initial adoption of executive stock option plans is
associated with dividend reductions. Kole (1991) notes that this association
could reflect changes in firms’ investment opportunity sets rather than simply
a compensation-induced change in dividend policy. Finally. Nance, Smith, and
Smithson (1991) find no significant relation between leverage and hedging,
which they interpret as an inability to separate two effects that work through
leverage: (1) given investment opportunities. more leverage should produce
stronger incentives to hedge: and (1) firms with more leverage have fewer growth
options and lower incentives to hedge.
5. Conclusions
Although evidence on the relations between growth options and leverage and
dividend policies had been provided previously, this paper is the first to present
evidence on the relations between growth options and compensation policy,
between regulation and leverage, dividend. and compensation policies, and
among the policies themselves. Documentation of these empirical relations is an
important step in focusing the profession’s attention on the explanation of
empirically important phenomena. Refinement of the relations examined here
and examination of additional relations should provide guideposts for the
development of richer theory. Our evidence suggests contracting theories are
more important in explaining cross-sectional variation in observed financial.
compensation, and dividend policies than either tax-based or signaling theories.
Although we believe our results, as well as those of Gaver and Gaver (1993)
Holthausen and Larcker (1992). and Kole (1992) are suggestive, much work
remains. There are potentially important limitations of this initial analysis and
thus several ways in which the power of our tests might be increased. First, our
exogenous variables are at least partially endogenous. A model that more
effectively separates exogenous from endogenous components of the investment
opportunity set would increase the power of our tests. Second, we do not have
measures of the specific tax status of companies in our industries. More detailed
data vvould enable more powerful tests of tax-based hypotheses. Third, other
corporate policies can be examined: for example, leasing. hedging, and account-
ing policies also should be driven by the firm’s investment opportunity set.
Appendix
A.I. .CIatching the Fo.u atd Cottlplrstrrt ciuta
Table A.1 gives Fox’s industry groupings by SIC code for each year of the
analysis. Some of the definitions are not constant across years. For the un-
regulated industries in table A.1 (all except insurance, utilities. and banking), we
begin with all Cotnpustat firms that fall into one of Fox’s industries in a survey
year. For each industry in each year, we sort the firms by sales. We find the firm
with sales closest to the median industry sales reported by Fox for that year. If.
for example. that firm is number 53 in our sorted list of firms in that year, we
keep twice 53 or 106 firms in the Cotnpusrat sample for that industry-year; the
107th firm and firms further down the list are dropped. (The median size of
Tuble A. f
Industry detinittons (SIC codes\ for 16 tndustrtcs from Fox by year for 1965. 1969. 1973. 1977. 1981.
and 1985.
Industry
–
I. Insurance
1. Gas s( electrtc utilities
3. Banking
1. Manufacturing machinery
5. Electrical machmery
6. Paper
7. Stone. clay &i glass
8. Food
9. Textile mill products & apparel
IO. Primary metals
I I. Construction
12. Retail trade
13. Consumer chemicals
I-l. Fabricated metals
15. Transportation equipment
16. Industrial chemicals (petroleum)
Years” Dehnition (SIC codes)
._~ __.._~..-.- _~
69-85 6312-6332
65-85 -191 l-4932
69-85 6012-6026
65 35OG3599
69-85 3510-3580
65-8 I 360&3699
85 3600-3699 & 3800-3899
65-8 I 26O(f2699
85 26OG2799
65-85 32OG-3299
65-85 2oOG2099
65-73 & 85 2200-2399
77-8 I 3200-2299
65-85 3300-3399
77-85 1600-1799
65-85 521 l-5999
65-71 2800-2899
81-85 283O-‘848
65 3100-3199
69-85 3JlG3199
65 3700-3799
69-85 371 l-3791
65-77 2900-2999
81-85 2810-2820 & 285&2890
“If a year is not included in the ranges specified for an industry. data are not abailabie for firms in
that Industry in that year and the industry-year is not included in the empirical work. Insurance and
banking are not included in 1965 and construction is not included before 1977.
Compusrar firms in a given unregulated industry is alvvays smaller than the
median Fox firm, so we always drop smaller rather than larger Corqwstnr firms
in forming the Compusrc~ samples for those industries.) In calculating each
variable for Fox’s industries we use every firm in the Comprrsrat industry-year
subsample that has available data. Hence, the number of firms used for a given
industry can vary across years and for different variables, although for most
variables (research and development being the major exception), the number is
the same.
For the regulated industries (insurance, gas and electrical utilities, and com-
mercial banking), Fox reports size attributes other than sales (premium income.
total current operating revenue, and deposits, respectively). We use those
attributes instead of sales in forming the Cornpust~zt sample for the regulated
industries. Insurance firms’ financial data for 1965 are not available on Compu-
stcrf at Rochester. and the insurance industry is included in the empirical work
only from 1969 on (see table A.1). Premium income is not available on Compu-
srclt and has to be collected from Mootlls’ Barrk and Finunce .Ifanual. which
reduces the Compusr~~r sample substantially and results in almost all available
Co~pu.src~ insurance firms being larger than the insurance firms in Fox’s
samples. We nonetheless include every Contpusr~~r insurance firm with data
available in a given year in order to have an adequate sample.
The procedures used to obtain tht Cor?~p~srar sample for unregulated indus-
tries are also applied to obtain the Compllsiat sample for utility firms in 1969.
1973. 1977. 1981, and 1985. In 1965, however, only 10 gas and electric utility
firms are available on Compustczt at Rochester. We include all 20 in the 1965
Cmprrstar utility sample. although because Fox’s utility sample contains large
NYSE-listed utilities, our 1965 C’ompustat utility sample tends to be smaller
than Fox’s 1965 sample.
Fox uses deposits as the size statistic for the banking industry. The Compuscar
file includes two deposit numbers: worldwide and domestic deposits. Since Fox’s
deposit definition is unclear. we use worldwide deposits because it yields the
larger number of observations. This time the Compustnt banks are larger than
Fox’s banks (possibly because Fox uses domestic deposits), but to give us
a sample of acceptable size, we include all Con~pmtar banks with financial data
available for at least one of the five years beginning with 1969. (Bank data are
not available on our Con~pusrr~t files for 1965 – see table A.1.)
Besides the median. Fox also reports the distribution of the size measures
across five to seven size intervals. We calculate a chi-square statistic to compare
the Fox sample and Co~npustczt sample size distributions. The two samples are
significantly different at the 0.10 level in 10 of the 11 industry-years for which
data are available for the insurance and banking industries and for the utilities
industry for 1965. We expect these results, however, since in those industry-years
we are unable to select the Cornpustat samples to match the Fox samples by firm
size. For the 13 unregulated industries and the utilities industry for years other
than 1965 (when we could select the Conzpr~at sample by firm size). the two
sample distributions are significantly different at the 0.10 level in only 15 out of
the 80 industry-year observations, and at the 0.01 level in only IO of the 80
industry-years. (There are 80 industry-years available instead of 83 – 13 indus-
tries for six years and one industry for five years – because Fox does not report
compensation data for the construction industry until 1977.)
Although the limitations in matching the two samples introduce noise into the
estimated relations that involve compensation-policy variables. estimated rela-
tions involving only financial variables are not affected. When the compensation
variables are used as dependent variables in regressions that have Comprrstnr
variables as independent variables. the effect of the noise on the other variables’
Table .A.2
Distributions of number of tirms in industry-years for Fox and Compustat samples across 16
industries and six years.”
Number of tirms in industry-year
Quantlles Fox Compustat
Maximum 213 208
0.75 84 94.5
0.50 56 55.5
0.25 31 36
Minimum 21 IO
“Total number of industry-years is 91 rather than 96 because data are not available in Comprtstclr
for the insurance and banking industries in 1965 and Fox does not report compensation data for the
construction industry in 1965. 1969. and 1973.
coefficients should be to bias their t-statistics toward zero. (In two of the three
regulated industries. however, the mean firm size for the Cornpustat sample is
larger, and this could bias the estimated coefficient of the regulation variable
toward its predicted sign-see section A.2 of the appendix.)
Table A.2 gives the distributions of the number of firms available for calculat-
ing mean variables in industry-years for the Fox and Compustat samples. The
total number of industry-years in both samples is 91 (75 unregulated and 16
regulated). The median industry-year in the Fox sample includes 56 firms and
the median Cornpustar industry-year includes 55.5 firms. A Wilcoxon-Xlann-
Whitney test [see Siegel and Castellan (1988, p. I’S)] does not reject the
hypothesis that the two distributions are the same at conventional critical
probability levels (1 = 0.003).
A 2. Sensiticit~ ana/j~.sis
Alternate itwestment opportunity set measures. As a specification check, we
use other investment-opportunity-set measures: the ratio of depreciation to firm
value (DEP: V), the ratio of research and development to firm value (R&D V), the
variance of the rate of return on the firm’ (VAR), the earningsi’price ratio (X/f’),
and the ratio of capital expenditures to firm value (CZ-lP/v). We average the
ratios across firms and years as in eq. (3) and calculate the variance for each firm
over four years and then average across firms. There is considerable correlation
among these alternate measures. however. To deal with this multicollinearity,
we focus on the base case that includes only the ratio of book-to-market values,
‘The rate of return on the tirm (r,) is defined as
T, = lr,.,(&_ ,) f I.V7-,) I’,_,
C. Ct.. Smirh. Jr. unri R. L. Wuri.r. Finuncmg. riirichd. und con~prnsurion policks 285
XC’, reported in section 3. Here. we reestimate the base-case regression substitu-
ting each of these variables in turn for A/C_.
When VAR is substituted for A/V, the results are very similar to those in
table 1. All regressions are significant. though the dividend-policy and financial-
policy regressions are less significant than those using A,‘V. All estimated
coefficients have the same sign and significance as the equivalent coefficients in
table I.
The regressions when CAPf’Cp or DEP/V is substituted for A,‘V are very
similar to each other and to those in table 1. All regressions are significant at the
0.0001 level, though the financial-policy regressions are less significant than
those using A,‘V. The one change in estimated coefficient sign for both CAP: V
and DEP;V regressions is for the coefficient of log of real sales in the financial-
policy regression. That coefficient is positive in the CAP; V and DEP,I V finan-
cial-policy regressions and is significantly positive in the CAP,: V regression. The
only other different result for the CAP/V and DEP; V regressions is that the
coefficient of log of real sales in the bonus regression is insignificant.
When the earningsprice ratio (X/P) is substituted for A,‘V, the results are
similar to those in table 1. All regressions are significant, though the dividend-
and financial-policy regressions are less significant than those using A/V. The
difference in results from those in table I is that the coefficient of the investment-
opportunity-set variable in the compensation, bonus, and stock-option regres-
sions becomes insignificant. The two of those coefficients with the signs
predicted by the contracting hypotheses (compensation and stock-option) have
significance levels of 0.28 and 0.11, respectively. Overall, X/P is a less effective
measure of the effect of the investment opportunity set on compensation policy
than A/V.
When R&D/V is used instead of A,‘V, the coefficient in the bonus regression
changes sign and is significant. There is one other change in sign. The coefficient
of log of real sales in the financial-policy regression is insignificantly different
from zero, but positive, contrary to prediction. Four coefficients that are
significant in table 1 are insignificant in the R&D/V regressions: the coefficients
of the investment opportunity set in the dividend and financial-policy regres-
sions and the coefficients of log of real sales in the bonus and stock-option
regressions. Overall these results suggest that R&D/,2’ is more associated with
compensation policy than with financial and dividend policy. All the regressions
using R&D, V are significant at the 0.0001 level except the financial-policy
regression. which is significant at the 0.03 level.
These results show that the results in table 1 are generally robust to alternate
specification of the investment-opportunity-set variable. The strongest results in
table I are for the financial-policy, dividend-policy. compensation, and stock-
option regressions. In those regressions, all estimated coefficients have the sign
predicted by the contracting hypotheses and are significant. When the alternate
investment-opportunity-set variables are substituted in those four regressions,
the estimated coefficients of the investment-opportunity-set variable have the
predicted signs in all five alternate specifications for each of the four regressions
and are significant in four of the five specifications for each of the four regres-
sions. The X’P (compensation and stock-option regressions) and R&D:‘V (finan-
cial and dividend-policy regressions) specifications each provide two insignitic-
ant coefficients.
Instrwt7entril lariclhles. To examine further the robustness of the results to
alternate specifications of the investment-opportunity-set variable, we use an
instrumental-variable approach. A,;V is regressed on the other investment-
opportunity-set variables and the predicted values from that regression are
substituted for Xc’ in the base-case regressions. As might be expected from the
lack of sensitivity to the substitution of the individual alternatives to A/V. the
results in table 1 do not change in any substantive way under the instrumental-
variable approach. None of the coefficients change sign or significance.
1 Y-oblrtm with rite regrtlatory subsample. There are several potential prob-
lems with the use of the regulated industries in our empirical analysis. One is
that while sales are used as the firm size measure for the unregulated industries,
other measures are used for regulated firms. A stock measure (deposits) is used in
banking although a flow measure (sales) is used in other industries. The effect of
this use is unclear, since it is like using accounts receivable for sales and the effect
will depend on how frequently deposits turn over. Bank deposits are higher than
any other industry’s size measure. Total current operating revenue (utilities’ size
measure) and premium income (insurance-industry size measure) are analogous
to sales and are less likely to involve bias or noise than the measure for banking.
To assess the effects of the different size measures we reestimate the five policy
regressions excluding all three regulated industries and dropping the regulation
dummy variable.
Table A.3 reports the results. All the regressions remain significant, though
the significance level drops. Excluding the regulated industries causes size to
become insignificant in the financing-policy and stock-option regressions and
reduces the significance of AiV in the compensation regression. The A/V
coefficient retains the sign predicted by the hypotheses and remains significant
in both regressions involving Compusrar-based dependent variables (financial
and dividend policies) as well as in the compensation and stock-option regres-
sions. The firm-size coefficient retains its predicted sign and remains significant
in the compensation and bonus regressions. Thus the general tenor of our results
remains.
Another problem is identified earlier in the paper. The banking, gas and
electric utilities. and insurance industries are subject to different regulations, so
the effect of regulation on firms’ policies is likely to differ across these industries.
To assess the effect of this variation. we substitute separate intercept dummy
variables for each regulated industry for the single regulatory dummy variable in
the policy regressions.
0
.
I!
,
0
.7
L
J
1
l.X
’)
7
.0
I
A
ss
c,
s,
v
;d
u
t
_
–
a
.5
7
–
1
3
.6
4
+
o
.o
‘t
1
2
.4
9
_
~
~
0
.
t 0
–
I.4
0
‘, ;,.
30
2.
4’
)
0
20
_
3.
12
_
~
O
.(
W
)
–
0
.5
3
‘,
&
a
s.
2
3
i w
7
9
.9
8
t 0
.0
4
1
.5
1
f O
.O
I
0
.0
2
0
.7
I
9
3
.3
2
o
.t
x
u
I
0
.6
X
x3
.x
4
o
.,x
xI
I
+
I2
0
0
.5
’)
3
6
.3
0
X
V
O
.o
c
N
) I
0
.0
7
0
.
I2
5
.‘
)
I
0
0
0
s
A,
I’
2.
I
’J
Ln
S
Ik
!S
t .
0
x
/I,
I’
1
.8
2
I.
1
1
SI
IL
!S
t .
0
x
:I
;
I
2
.7
1
1
.1
1
S
k
!?
,
t .
‘)
L
When we make this substitution in the base case. the coefficients of all I5
dummy variables have the sign predicted by the contracting hypotheses and all
but the coefficient of the insurance dummy variable in the dividend price
regression are significant. But, the individual-industry dummy coefficients are
significantly different across the three industries. All else being equal, the banks
have a significantly lower equity-to-value ratio than the insurance companies.
which in turn have a significantly lower ratio than the utilities. Utilities’
dividend yields are significantly higher than those of the insurance companies or
banks, and banks pay a significantly lower salary than the insurance companies
or utilities. Finally, utilities are significantly less likely to have a bonus plan and
a stock-option plan than either banks or insurance firms. Despite these differing
industry effects, the substitution of separate industry dummy variables does not
change the tenor of the results for the coefficients of assets/value. firm size, and
accounting return.
It is possible that regulation affects the slope coefficients as well as the
intercept. Although we expect this effect to work against confirming the con-
tracting predictions, we introduce both regulated-industry intercept dummies
and multiplicative dummies for the explanatory variables to check that the
significant results for the investment-opportunity-set and firm-size variables in
the base cases are not due to misspecification of the regulatory effect. This
procedure also allows us to check for the problem (mentioned in section A.l)
that the insurance and banking firms in the Co~~pwstar sample are much larger
than those in Fox’s sample. This difference affects the compensation and
incentive-plan regressions. Since it appears smaller firms pay less compensation
and are less likely to have incentive plans, the insurance and banking industries
will show lower compensation and use of incentive plans (which come from
Fox’s sample) than their firm sizes (which come from Compustar) suggest.
A possible result is a negative sign for the regulation coefficient and a lower
slope coefficient for firm size.
Specific regulated-industry intercept and slope dummy variables do not
change the tenor of the results for the assetsvalue, firm-size, and accounting-
return variables. Few of the 33 slope dummy coefficients are significant. None of
the predictions for the firm-size coeficients based on the sample matching
problems are confirmed. The primary effect is to reduce the significance of the
intercept dummy coefficients.
Time-series stationarity. The regressions assume that relations between the
exogenous variables and the endogenous policy variables are stationary over time.
These relations may have changed over time. For example, regulated firms have
increased their use of bonus plans significantly in more recent years. This could be
due to more relaxed regulation. The effect of nonstationarity is to increase the
estimated standard errors of the coefficients and reduce the estimated coetlicients’
significance. But the changes in the relations can provide insight into the relations
themselves, and for that reason we examine their stationarity.
The first hypothesis we test is that all coefficients (including the intercept) are
constant across the full period (19651985). The sums of squared errors from
separate cross-sectional regressions for each year are compared with sums of
squared errors from the regressions that require all coefficients to be constant
over time. The hypothesis is not rejected at standard levels, except for the bonus
and compensation regressions, where it is rejected at the 0.01 and 0.05 levels.
respectively.
To provide more information on nonstationarities. we test the stationarity of
each of the five coefficients (coefficients of regulation, A./V, log of real sales.
accounting return and intercept) separately. Regressions are run restricting the
particular coefficient whose stationarity is being tested to be constant and
allowing all other coefficients and the intercept to vary over time (by use of year
dummies). The sums of squared errors from those regressions are compared with
the sums of squared errors from separate regressions for each year. The hypo-
thesis that the coefficient is stationary over time cannot be rejected at any
reasonable probability level for any coefficient in any of the five regressions.
The rejection of stationarity of all coefficients for some regressions and the
failure to reject stationarity for any single coefficient suggest that the rejections
in the first test are due to joint effects. To test that hypothesis, we run regressions
restricting the intercept and all but one coefficient to be constant over time and
compare the regressions’ sums of squared errors with the sums of squared errors
of the individual-year regressions. The hypothesis that other coefficients are
stationary can be rejected only for the bonus regression and the compensation
regression when the regulation coefficient is allowed to vary (i.e., A/V and
log-of-sales coefficients are constant). Thus it appears the rejection of the
stationarity of all coefficients for the bonus and compensation regressions is due
to the joint nonstationarity of the A/V and log-of-sales coefficients.
Generally, the estimated relations are stationary over time. The financial-
policy. dividend-policy, and stock-option regressions show no significant non-
stationarity. These equations also have the most explanatory power (see table 1)
and are among the most robust to other specification checks. Also. one of the
two regressions showing some evidence of nonstationarity is the bonus regres-
sion. where the sign of the predicted relation with XV is ambiguous.
Dependence in corporate policirs. Positive dependence in corporate policies
would cause the standard errors of the coefficients in the pooled cross-section
and time-series regressions to be understated and the t-statistics overstated. To
assess the effect of this dependence on the significance of the estimated coeffi-
cients in table 1 we estimate the variance-covariance matrices for given policies
within each industry and use those estimated matrices in a generalized-least-
squares estimation [see Froot (1989)].
The results for the generalized-least-squares estimation are similar to those in
table 1. As in table I. each slope coefficient is significant. The significance levels
for the coefficients of ‘-4 ‘C’. log of real sales, and accounting return are of a similar
290 C. I+‘. Smirk. Jr. und R.L. M;Irrs. Fiminuncin~g, diridund. and comprnsu~ron policies
magnitude, but the levels for the regulation coefficients are less significant; for
example, the c-statistic for the regulation coefficient in the equity/value regres-
sion is – 4.04 versus – 12.33 in table 1. This latter result is to be expected since
industries are classified as regulated or unregulated for the entire estimation
period.
Further insight into the significance of the coefficients in table 1 is obtained by
regressing the mean policy variable on the mean independent variables for the
entire 1965-1985 period. This regression examines cross-sectional variation and
ignores time-series variation. The regulation coefficients are less significant in
these regressions just as in the generalized-least-squares estimation. However,
the other slope coefficients are also less significant in this specification than in
the generalized-least-squares and pooled regressions (table 1); for example, the
t-statistic for the A/V coefficient in the equity/value regression is – 3.49 versus
– 10.29 and – 12.47 in the generalized-least-squares and pooled regressions,
respectively. This suggests that a substantial part of the table 1 explanatory
power of A:‘V, log of real sales, and accounting return comes from their
(nondependent) times-series variation.
The previous inference is confirmed by a fixed-effects analysis [see Hsiao
(1986, ch. 3)]. The time-series mean for the industry is deducted from each
variable and the policy regressions estimated with the transformed variables.
The effect of this transformation is to eliminate the individual industry effect
that is constant across time and estimate the independent variable coefficients
solely on the basis of the within-industry (time-series) variation. Because the
regulatory variables are constant across time, their estimated coefficients in the
fixed-effect analysis are zero. The significance of the coefficients of log of real
sales and accounting return is similar to that in the pooled regression (table 1)
and in the generalized-least-squares estimation. However, the significance of the
A/V coefficient increases in all policy regressions; for example, the t-statistic for
the AIVcoefficient in the equity/value regression in table 1 is – 12.47 and in the
fixed-effects analysis it is – 16.13.
Our analysis therefore indicates positive dependence in corporate policies in
the pooled cross-section and time-series regressions reported in table 1 is
a problem for the estimation of the regulation coefficients only. Even so, the
regulation coefficients remain significant when they are estimated using cross-
sectional and not time-series variation.
On compensation diflerentials for risk. An alternate interpretation of our
evidence is that risk considerations. rather than investment-opportunity-set
characteristics. drive our results. Managerial risk aversion would imply that
managers of high-risk projects receive higher compensation. The predicted
impact on the use of incentive compensation plans is less clear; it ultimately
should depend on the proportion of controllable and uncontrollable risk. But if
more total risk tends to be associated with more controllable risk, the use of
incentive compensation plans should increase. The simple bankruptcy-cost
theory of capital structure su,, Ooests that more volatile firms should use less debt.
We can find no analysis, however. to link volatility and dividend policy.
To attempt to distinguish between the investment-opportunity-set and vola-
tility hypotheses, we add IPAR to our benchmark regression. In all five regres-
sions. the size and significance of the benchmark regression coefficients remain.
but L..-lR is insigniticant. Although insignificant, the signs of the I-AR coetli-
cients are the same as those reported in table A.?. where 4c’was omitted. From
this evidence. we conclude that our results are not driven simply by volatility.
However, we recognize that this test faces potential problems: (1) A lo and VAR
are correlated. Using them in the same regression makes it difficult to separately
identify their effects because of multicollinearity. (1) We know that A I-is a noisy
instrumental variable for the investment opportunity set. We also know that
FAR is a noisy instrumental variable for risk. Thus the regression results reflect
the correlations among the true variables. but also correlations among the error
components.
References
Ang. J. and P. Peterson. 1984. The leasing puzzle. Journal of Finance 39. 1053-1065.
Bhattacharya. S.. 1979. Imperfect information. dividend policy. and ‘the bird-m-the-hand’ fallac!.
Bell Journal of Economics IO. 259-270.
Bradley. ht.. G.A. Jarrell. and E.H. Kim. 1984. On the existence of an optimal capital structure:
Theory and evidence, Journal of Finance 39. 357478.
Castanias. R.. 1983. Bankruptcy risk and optimal capital structure, Journal of Finance 3s. 1617-1625.
Christie. ,A.. M. Joye. and R.L. Watts, 1989. Organization of the firm: Theory and e\:dcnce (Simon
Graduate School of Business Administration. Cniversity of Rochester, Rochester. NY).
DeAnSrIo. H. and R. Masulis. 1980. Optimal capital structure under corporate and personal
taxation. Journal of Financial Economics 8. 3-29.
Easterbrook. F.H., 1981. Two agency-cost explanattons of dividends, American Economic Rsbisu
74. 650-659.
Eaton. J. and H.S. Rosen. 1983. Agency, delayed compensation and the structure of executibs
remuneration. Journal of Finance 38: l189-1505.
Ferri. h1.G. and W.H. Jones. 1979. Determinants of financial structure: A new msthodoloeical
approach. Journal of Finance 34, 631-643.
Fox. H.. 1986. Top executive compensation (Conference Board, New York. NY,.
Froot. K..A., 1989. Consistent covariance matrix estimation with cross-sectional dependence and
heteroskedasticity in tinancial data. Journal of Financial and Quantitative Anal>sis 2-I. 333-356.
Gaver. J.J. and K.M. Gaver. 1993. Additional evidence on the association between the investment
opportunity set and corporate financing. dividend and compensatton policies. Journal of .Ac-
countin: and Economics 16. 125-160.
Holthausen. R.W. and D.F. Larcker. 1991. Financial performance and organizarlon structure
(Wharton School. Cnibersitj of Pennsylvania. Phtladelphia. PA).
Hsiao. C.. 1986. Analysis of panel data (Cambridge Lnibersity Press. Cambridget.
Jensen. \l.C.. 1986. Agency costs of free cash Hou. corporate finance. and takeo\:rs. .American
Economic Rebieu 76, 323-329.
Kalay. A.. 1982. Stockholder-bondholder confiict and dividend constraints, Journal of Financial
Economics IO. 2 I l-233.
Kole. S.R.. 1991. .A cross-sectional investigation of managerial compensation from an ex ante
perspsctice (Simon Graduate School of Business Administration. Lniverstt! oi Rochester.
Rochester. NY I.
‘2 9 C. It. Snrlrh. Jr. und R.L. Nir~[s. Financmg. dicidend. und compmsuriun policies
Lambert. R.A.. W.N. Lanen. and D.F. Larcker. 1989, Executive stock option plans and corporate
dividend policy. Journal of Financial and Quantitative Analysts 24. 409175.
Lttzenberger. R.H. and I<. Ramaswamy, 1981. The effect of dividends on common stock prices: Tar
effects or information effects’?. Journal of Finance 37. 429443.
Long. M. and E. Malitz. 1985. The investment-financing nexus: Some empirical evidence. Midland
Corporate Finance Journal 3. 53-59.
MacDonald. G.. 1981. Sew directions in the economic theory of agency. Canadian Journal of
Economics 17. II5440.
Miller. M. and MM. &holes. 1982. Executive compensatmn. taxes and incentives. in: W.F. Sharpe and
C.M. Cootner. eds.. Financial economics: Essays in honor of Paul Cootner (Prentice-Hall.
Engleuood Cliffs. NJ).
Murphy. K.J.. 1985, Corporate performance and managerial remuneration. Journal of Accounting
and Economics 7, 1141.
Myers. S.. 1977. Determinants of corporate borrowing. Journal of Financial Economics 5. 117-175.
Myers. S. and N.S. Majluf. 1981, Corporate financing and investment decisions when firms have
information that investors do not have. Journal of Financial Economics 13. 187-221.
Nance. D.R.. C.W. Smith. Jr.. and C.W. Smithson. 1992, The determinants of corporate hedging,
Journal of Finance, forthcoming.
Rao. G.. 1989. The relation between stock returns and earnings: A study of newly-public firms,
Unpublished thesis (Simon Graduate School of Business Administration. University of Roches-
ter. Rochester. NY).
Ross. S.. 1977. The determination of financial structure: The incentive-signaling approach. Bell
Journal of Economics 8. 1310.
Rozetf. MS.. 1982. Growth. beta and agency costs as determinants of dividend payout ratios,
Journal of Financial Research 5. 249-259.
Siegel. S. and N.J. Castellan. Jr.. 1988, Nonparametric statistics for the behavioral sciences. 2nd ed.,
J.D. Anker. ed. (McGraw-Hill. New York, NY).
Sloan. R.G.. 1993. Accounting earnings and top executive compensation. Journal of Accounting and
Economics 16. 55-100.
Smith. C.W.. Jr.. 1986. Investment banking and the capital acquisition process. Journal of Financial
Economics 15. 3-29.
Smith. C.W.. Jr. and R. Stulz. 1985. The determinants of firm’s hedging policies, Journal of Financial
and Quantitative Analysis 20, 391405.
Smith, C.W.. Jr. and L.Ecl. Wakeman. 1985. Determinants of corporate leasing policy. Journal of
Finance 10. 895-908.
Smith. C.W.. Jr. and J. Warner. 1979, On financial contracting: An analysis of bond covenants,
Journal of Financial Economics 7. 117-161.
Smith. C.W.. Jr. and R.L. Watts. 1981. Incentive and tax etlects of U.S. executive compensation
plans. Australian Journal of Management 7. 139-157.
Titman. S. and R. Wessels. 1988. The determinants ofcapital structure choice. Journal of Finance 33,
I-19.
White. H.. 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for
heteroskedasticity. Econometrica 48. 817-838.
