the theory and practice of corporate nance: evidence from...
TRANSCRIPT
Journal of Financial Economics 60 (2001) 187}243
The theory and practice of corporate "nance:evidence from the "eld�
John R. Graham�, Campbell R. Harvey����*�Fuqua School of Business, Duke University, Durham, NC 27708, USA�National Bureau of Economic Research, Cambridge, MA 02912, USA
Received 2 August 1999; received in revised form 10 December 1999
Abstract
We survey 392 CFOs about the cost of capital, capital budgeting, and capital structure.Large "rms rely heavily on present value techniques and the capital asset pricing model,while small "rms are relatively likely to use the payback criterion. A surprising number of"rms use "rm risk rather than project risk in evaluating new investments. Firms areconcerned about "nancial #exibility and credit ratings when issuing debt, and earnings
�We thank Franklin Allen for his detailed comments on the survey instrument and the overallproject. We appreciate the input of Chris Allen, J.B. Heaton, Craig Lewis, Cli! Smith, Jeremy Stein,Robert Taggart, and Sheridan Titman on the survey questions and design. We received expertsurvey advice from Lisa Abendroth, John Lynch, and Greg Stewart. We thank Carol Bass, FrankRyan, and Fuqua MBA students for help in gathering the data, and Kathy Benton, Steve Fink, AnneHiggs, Ken Rona, and Ge Zhang for computer assistance. The paper has bene"ted from commentsmade by an anonymous referee, the editor (Bill Schwert), as well as Michael Bradley, Alon Brav,Susan Chaplinsky, Magnus Dahlquist, Gene Fama, Paul Gompers, Ravi Jagannathan, Tim Opler,Todd Pulvino, Nathalie Rossiensky, Rick Ruback, David Smith, ReneH Stulz, and seminar partici-pants at the Harvard Business School/Journal of Financial Economics Conference on the interplaybetween theoretical, empirical, and "eld research in "nance, the 2000 Utah Winter FinanceConference, the University of Wisconsin and the 2001 American Finance Association Meetings.Finally, we thank the executives who took the time to "ll out the survey. This research is partiallysponsored by the Financial Executives Institute (FEI). The opinions expressed in the paper do notnecessarily represent the views of FEI. Graham acknowledges "nancial support from the Alfred P.Sloan Research Foundation. Some supplementary research results are available athttp://www.duke.edu/&charvey/Research/indexr.htm.
* Corresponding author. Fuqua School of Business, Duke University, Durham, NC 27708, USA.Tel.: #1-919-660-7768; fax: #1-919-660-7971.
E-mail address: [email protected] (C.R. Harvey).
0304-405X/00/$ - see front matter � 2001 Published by Elsevier Science S.A.PII: S 0 3 0 4 - 4 0 5 X ( 0 1 ) 0 0 0 4 4 - 7
� See, for example, Lintner (1956), Gitman and Forrester (1977), Moore and Reichert (1983),Stanley and Block (1984), Baker et al. (1985), Pinegar and Wilbricht (1989), Wansley et al. (1989),Sangster (1993), Donaldson (1994), Epps and Mitchem (1994), Poterba and Summers (1995),Billingsley and Smith (1996), Shao and Shao (1996), Bodnar et al. (1998), Bruner et al. (1998) andBlock (1999).
per share dilution and recent stock price appreciation when issuing equity. We "nd somesupport for the pecking-order and trade-o! capital structure hypotheses but littleevidence that executives are concerned about asset substitution, asymmetric information,transactions costs, free cash #ows, or personal taxes. � 2001 Published by ElsevierScience S.A.
JEL classixcation: G31; G32; G12
Keywords: Capital structure; Cost of capital; Cost of equity; Capital budgeting; Discountrates; Project valuation; Survey
1. Introduction
In this paper, we conduct a comprehensive survey that describes the currentpractice of corporate "nance. Perhaps the best-known "eld study in this area isJohn Lintner's (1956) path-breaking analysis of dividend policy. The results ofthat study are still quoted today and have deeply a!ected the way that dividendpolicy research is conducted. In many respects, our goals are similar to Lin-tner's. We hope that researchers will use our results to develop new theories} and potentially modify or abandon existing views. We also hope that practi-tioners will learn from our analysis by noting how other "rms operate and byidentifying areas where academic recommendations have not been fully imple-mented.
Our survey di!ers from previous surveys in a number of dimensions.� First,the scope of our survey is broad. We examine capital budgeting, cost of capital,and capital structure. This allows us to link responses across areas. For example,we investigate whether "rms that consider "nancial #exibility to be a capitalstructure priority are also likely to value real options in capital budgetingdecisions. We explore each category in depth, asking more than 100 totalquestions.
Second, we sample a large cross-section of approximately 4,440 "rms. In total,392 chief "nancial o$cers responded to the survey, for a response rate of 9%.The next largest survey that we know of is Moore and Reichert (1983) who study298 large "rms. We investigate for possible nonresponse bias and conclude thatour sample is representative of the population.
188 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Third, we analyze the responses conditional on "rm characteristics. Weexamine the relation between the executives' responses and "rm size, P/E ratio,leverage, credit rating, dividend policy, industry, management ownership, CEOage, CEO tenure, and the education of the CEO. By testing whether responsesdi!er across these characteristics, we shed light on the implications of variouscorporate "nance theories related to "rm size, risk, investment opportunities,transaction costs, informational asymmetry, and managerial incentives. Thisanalysis allows for a deeper investigation of corporate "nance theories. Forexample, we go beyond asking whether "rms follow a "nancial pecking order(Myers and Majluf, 1984). We investigate whether the "rms that most stronglysupport the implications of the pecking-order theory are also the "rms mosta!ected by informational asymmetries, as suggested by the theory.
Survey-based analysis complements other research based on large samplesand clinical studies. Large sample studies are the most common type of empiri-cal analysis, and have several advantages over other approaches. Most large-sample studies o!er, among other things, statistical power and cross-sectionalvariation. However, large-sample studies often have weaknesses related tovariable speci"cation and the inability to ask qualitative questions. Clinicalstudies are less common but o!er excellent detail and are unlikely to `averageawaya unique aspects of corporate behavior. However, clinical studies use smallsamples and their results are often sample-speci"c.
The survey approach o!ers a balance between large sample analyses andclinical studies. Our survey analysis is based on a moderately large sample anda broad cross-section of "rms. At the same time, we are able to ask very speci"cand qualitative questions. The survey approach is not without potential prob-lems, however. Surveys measure beliefs and not necessarily actions. Surveyanalysis faces the risk that the respondents are not representative of the popula-tion of "rms, or that the survey questions are misunderstood. Overall, surveyanalysis is seldom used in corporate "nancial research, so we feel that our paperprovides unique information to aid our understanding of how "rms operate.
The results of our survey are both reassuring and surprising. On one hand,most "rms use present value techniques to evaluate new projects. On the otherhand, a large number of "rms use company-wide discount rates to evaluate theseprojects rather than a project-speci"c discount rate. Interestingly, the surveyindicates that "rm size signi"cantly a!ects the practice of corporate "nance. Forexample, large "rms are signi"cantly more likely to use net present valuetechniques and the capital asset pricing model for project evaluation than aresmall "rms, while small "rms are more likely to use the payback criterion.A majority of large "rms have a tight or somewhat tight target debt ratio, incontrast to only one-third of small "rms.
Executives rely heavily on practical, informal rules when choosing capitalstructure. The most important factors a!ecting debt policy are "nancial #exibil-ity and a good credit rating. When issuing equity, respondents are concerned
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 189
about earnings per share dilution and recent stock price appreciation. We "ndvery little evidence that executives are concerned about asset substitution,asymmetric information, transactions costs, free cash #ows, or personal taxes.We acknowledge but do not investigate the possibility that these deeper implica-tions are, for example, impounded into prices and credit ratings, and so execu-tives react to them indirectly.
The paper is organized as follows. In the second section, we present the surveydesign, the sampling methodology, and discuss some caveats of survey research.In the third section we study capital budgeting. We analyze the cost of capital inthe fourth section. In the "fth section we examine capital structure. We o!ersome concluding remarks in the "nal section.
2. Methodology
2.1. Design
Our survey focuses on three areas: capital budgeting, cost of capital, andcapital structure. Based on a careful review of the existing literature, we de-veloped a draft survey that was circulated to a group of prominent academics forfeedback. We incorporated their suggestions and revised the survey. We thensought the advice of marketing research experts on the survey design andexecution. We made changes to the format of the questions and overall surveydesign with the goal of minimizing biases induced by the questionnaire andmaximizing the response rate.
The survey project is a joint e!ort with the Financial Executives Institute(FEI). FEI has approximately 14,000 members that hold policy-making posi-tions as CFOs, treasurers, and controllers at 8,000 companies throughout theU.S. and Canada. Every quarter, Duke University and the FEI poll these"nancial o$cers with a one-page survey on important topical issues (Graham,1999b). The usual response rate for the quarterly survey is 8}10%.
Using the penultimate version of the survey, we conducted beta tests at bothFEI and Duke University. This involved having graduating MBA students and"nancial executives "ll out the survey, note the required time, and providefeedback. Our beta testers took, on average, 17 minutes to complete the survey.Based on this and other feedback, we made "nal changes to the wording onsome questions. The "nal version of the survey contained 15 questions, mostwith subparts, and was three pages long. One section collected demographicinformation about the sample "rms.
The survey instrument appears on the Internet at the addresshttp://www.duke.edu/&charvey/Research/indexr.htm. We sent out two di!er-ent versions with questions 11}14 and questions 1}4 interchanged. We wereconcerned that the respondents might "ll in the "rst page or two of the survey
190 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
but leave the last page blank. If this were the case, we would expect to seea higher proportion of respondents answering the questions that appear at thebeginning of either version of the survey. We "nd no evidence that the responserate di!ers depending on whether the questions are at beginning or the end ofthe survey.
2.2. Delivery and response
We used two mechanisms to deliver the survey. We sent a mailing from DukeUniversity on February 10, 1999 to each CFO in the 1998 Fortune 500 list.Independently, the FEI faxed out 4,440 surveys to their member "rms onFebruary 16, 1999. Three hundred thirteen of the Fortune 500 CFOs belong tothe FEI, so these "rms received both a fax and a mailed version. We requestedthat the surveys be returned by February 23, 1999. To encourage the executivesto respond, we o!ered an advanced copy of the results to interested parties.
We employed a team of 10 MBA students to follow up on the mailing to theFortune 500 "rms with a phone call and possible faxing of a second copy of thesurvey. On February 23, FEI refaxed the survey to the 4,440 FEI corporationsand we remailed the survey to the Fortune 500 "rms, with a new due date ofFebruary 26, 1999. This second stage was planned in advance and designed tomaximize the response rate.
The executives returned their completed surveys by fax to a third-party datavendor. Using a third party ensures that the survey responses are anonymous.We feel that anonymity is important to obtain frank answers to some of thequestions. Although we do not know the identity of the survey respondents, wedo know a number of "rm-speci"c characteristics, as discussed below.
Three hundred ninety-two completed surveys were returned, for a responserate of nearly 9%. Given the length (three pages) and depth (over 100 questions)of our survey, this response rate compares favorably to the response rate for thequarterly FEI-Duke survey.The rate is also comparable to other recent aca-demic surveys. For example, Trahan and Gitman (1995) obtain a 12% responserate in a survey mailed to 700 CFOs. The response rate is higher (34%) in Block(1999), but he targets Chartered Financial Analysts - not senior o$cers ofparticular "rms.
2.3. Summary statistics and data issues
Fig. 1 presents summary information about the "rms in our sample. Thecompanies range from very small (26% of the sample "rms have sales of lessthan $100 million) to very large (42% have sales of at least $1 billion) (see Fig.1A). In subsequent analysis, we refer to "rms with revenues greater than $1billion as `largea. Forty percent of the "rms are manufacturers (Fig. 1C). Thenonmanufacturing "rms are evenly spread across other industries, including
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 191
Fig. 1. Sample characterstics based on the survey respponses of 392 CFOs.
"nancial (15%), transportation and energy (13%), retail and wholesale sales(11%), and high-tech (9%). In the appendix, we show that the responding "rmsare representative of the corporate population for size, industry, and othercharacteristics.
The median price}earnings ratio is 15. Sixty percent of the respondents haveprice}earnings ratios of 15 or greater (Fig. 1D). We refer to these "rms as growth
192 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Fig. 1. (continued).
"rms when we analyze how investment opportunities a!ect corporate behavior.We refer to the remaining 40% of the respondents as nongrowth "rms.
The distribution of debt levels is fairly uniform (Fig. 1E). Approximatelyone-third of the sample "rms have debt-to-asset ratios below 20%, another thirdhave debt ratios between 20% and 40%, and the remaining "rms have debtratios greater than 40%. We refer to "rms with debt ratios greater than 30% as
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 193
highly levered. The creditworthiness of the sample is also dispersed (Fig. 1F).Twenty percent of the companies have credit ratings of AA or AAA, 32% havean A credit rating, and 27% have a BBB rating. The remaining 21% havespeculative debt with ratings of BB or lower.
Though our survey respondents are CFOs, we ask a number of questionsabout the characteristics of the chief executive o$cers. We assume that theCFOs act as agents for the CEOs. Nearly half of the CEOs for the responding"rms are between 50 and 59 years old (Fig. 1I). Another 23% are over age 59,a group we refer to as `mature.a Twenty-eight percent of the CEOs are betweenthe ages of 40 and 49. The survey reveals that executives change jobs frequently.Nearly 40% of the CEOs have been in their jobs less than four years, andanother 26% have been in their jobs between four and nine years (Fig. 1J). Wede"ne the 34% who have been in their jobs longer than nine years as having`long tenurea. Forty-one percent of the CEOs have an undergraduate degree astheir highest level of educational attainment (Fig. 1K). Another 38% have anMBA and 8% have a non-MBA masters degree. Finally, the top three executivesown at least 5% of the common stock of their "rm in 44% of the sample. TheseCEO characteristics allow us to examine whether managerial incentives orentrenchment a!ect the survey responses. We also study whether having anMBA a!ects the choices made by corporate executives.
Fig. 1M shows that 36% of the sample "rms seriously considered issuingcommon equity, 20% considered issuing convertible debt, and 31% thoughtabout issuing debt in foreign markets. Among responding "rms, 64% calculatethe cost of equity, 63% have publicly traded common stock, 53% issue divi-dends, and 7% are regulated utilities (Fig. 1N). If issuing dividends is anindication of a reduced informational disadvantage for investors relative tomanagers (Sharpe and Nguyen, 1995), the dividend issuance dichotomy allowsus to examine whether the data support corporate theories based on informa-tional asymmetry.
Table 1 presents correlations for the demographic variables. Not surprisingly,small companies have lower credit ratings, a higher proportion of managementownership, a lower incidence of paying dividends, a higher chance of beingprivately owned, and a lower proportion of foreign revenue. Growth "rms arelikely to be small, have lower credit ratings, and have a higher degree ofmanagement ownership. Firms that do not pay dividends have low creditratings.
Below, we perform univariate analyses on the survey responses conditional oneach separate "rm characteristic. However, because size is correlated witha number of di!erent factors, we perform a robustness check for the nonsizecharacteristics. We split the sample into large "rms versus small "rms. On eachsize subsample, we repeat the analysis of the responses conditional on "rmcharacteristics other than size. We generally only report the "ndings withrespect to nonsize characteristics if they hold on the full sample and the two size
194 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 195
subsamples. We also perform a separate robustness check relative to publicversus private "rms and only report the characteristic-based results if they holdfor the full and public samples. The tables contain the full set of results, includingthose that do not pass these robustness checks.
All in all, the variation in executive and "rm characteristics permits a richdescription of the practice of corporate "nance, and allows us to infer whethercorporate actions are consistent with academic theories. We show in the appen-dix that our sample is representative of the population from which it was drawn,fairly representative of Compustat "rms, and not adversely a!ected by non-response bias.
3. Capital budgeting methods
3.1. Design
This section studies how "rms evaluate projects. Previous surveys mainlyfocus on large "rms and suggest that internal rate of return (IRR) is the primarymethod for evaluation. For example, Gitman and Forrester (1977), in theirsurvey of 103 large "rms, "nd that only 9.8% of "rms use net present value astheir primary method and 53.6% report IRR as primary method. Stanley andBlock (1984) "nd that 65% of respondents report IRR as their primary capitalbudgeting technique. Moore and Reichert (1983) survey 298 Fortune 500 "rmsand "nd that 86% use some type of discounted cash #ow analysis. Bierman(1993) "nds that 73 of 74 Fortune 100 "rms use some type of discounted cash#ow analysis. These results are similar to the "ndings in Trahan and Gitman(1995), who survey 84 Fortune 500 and Forbes 200 best small companies, andBruner et al. (1998), who interview 27 highly regarded corporations. (Seehttp://www.duke.edu/&charvey/Research/indexr.htm for a review of the capitalbudgeting literature.)
Our survey di!ers from previous work in several ways. The most obviousdi!erence is that previous work almost exclusively focuses on the largest "rms.Second, given that our sample is larger than all previous surveys, we are able tocontrol for many di!erent "rm characteristics. Finally, we go beyond NPVversus IRR analysis and ask whether "rms use the following evaluation tech-niques: adjusted present value (see Brealey and Myers, 1996), payback period,discounted payback period, pro"tability index, and accounting rate of return.We also inquire whether "rms bypass discounting techniques and simply useearnings multiples. A price-earnings approach can be thought of as measuringthe number of years it takes for the stock price to be paid for by earnings, andtherefore can be interpreted as a version of the payback method. We are alsointerested in whether "rms use other types of analyses that are taught in manyMBA programs, such as simulation analysis and value at risk (VaR). Finally, we
196 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Fig. 2. Survey evidence on the popularity of di!erent capital budgeting methods. We report thepercentage of CFOs who always or almost always use a particular technique. IRR representsinternal rate of return, NPV is net present value, P/E is the price-to-earnings ratio, and APV isadjusted present value. The survey is based on the responses of 392 CFOs.
are interested in the importance of real options in project evaluation (see Myers,1977).
3.2. Results
Respondents are asked to score how frequently they use the di!erent capitalbudgeting techniques on a scale of 0 to 4 (0 meaning `nevera, 4 meaning`alwaysa). In many respects, the results di!er from previous surveys, perhapsbecause we have a more diverse sample. An important caveat here, and through-out the survey, is that the responses represent beliefs. We have no way ofverifying that the beliefs coincide with actions.
Most respondents select net present value and internal rate of return as theirmost frequently used capital budgeting techniques (see Table 2); 74.9% of CFOsalways or almost always (responses of 4 and 3) use net present value (rating of3.08); and 75.7% always or almost always use internal rate of return (rating of3.09). The hurdle rate is also popular. These results are summarized in Fig. 2.
The most interesting results come from examining the responses conditionalon "rm and executive characteristics. Large "rms are signi"cantly more likely touse NPV than small "rms (rating of 3.42 versus 2.83). There is no di!erence intechniques used by growth and nongrowth "rms. Highly levered "rms aresigni"cantly more likely to use NPV and IRR than are "rms with small debt
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 197
Tab
le2
Surv
eyre
spon
ses
toth
equ
esti
on:
how
freq
uen
tly
doe
syo
ur"
rmus
eth
efo
llow
ing
tech
niq
ues
wh
end
ecid
ing
wh
ich
pro
ject
so
rac
qui
siti
ons
topu
rsue
?�
%al
way
sSi
zeP
/EL
ever
age
Inve
stm
ent
grad
eP
ayd
ivid
ends
Ind
ustr
yM
anag
emen
to
wn
or
alm
ost
alw
ays
Mea
nSm
all
Lar
geG
row
thN
on-
GL
ow
Hig
hY
esN
oY
esN
oM
anu
.O
ther
sL
ow
Hig
h
(b)
Inte
rnal
rate
ofre
turn
75.6
13.
092.
873.
41**
*3.
363.
362.
853.
36**
*3.
523.
353.
432.
68**
*3.
192.
94**
3.34
2.85
***
(a)
Net
pre
sent
valu
e74
.93
3.08
2.83
3.42
***
3.30
3.27
2.84
3.39
***
3.47
3.38
3.35
2.76
***
3.23
2.82
***
3.35
2.77
***
(f)P
ayba
ckpe
riod
56.7
42.
532.
722.
25**
*2.
552.
412.
582.
462.
482.
362.
462.
632.
682.
33**
*2.
392.
70**
(c)
Hu
rdle
rate
56.9
42.
482.
132.
95**
*2.
782.
872.
272.
63**
3.01
2.92
2.84
2.06
***
2.60
2.29
**2.
702.
12**
*(j)
Sen
siti
vity
anal
ysis
(e.g
.,`g
ooda
vs.
`fai
ravs
.`ba
da)
51.5
42.
312.
132.
56**
*2.
352.
412.
102.
56**
*2.
602.
622.
422.
17**
2.35
2.24
2.37
2.18
(d)
Ear
nin
gsm
ulti
ple
app
roac
h38
.92
1.89
1.79
2.01
*1.
972.
111.
672.
12**
*1.
902.
22*
1.88
1.88
1.85
2.00
1.85
2.04
(g)
Dis
cou
nted
payb
ack
per
iod
29.4
51.
561.
581.
551.
521.
671.
491.
641.
841.
49*
1.54
1.62
1.61
1.50
1.49
1.76
*(l)
We
inco
rpor
ate
the`r
eal
opt
ionsa
of
apr
ojec
tw
hen
eval
uati
ng
it26
.59
1.47
1.40
1.57
1.31
1.55
1.50
1.41
1.34
1.61
1.37
1.52
1.49
1.45
1.40
1.52
(i)A
cco
unti
ngra
teof
retu
rn(o
rb
ook
rate
ofre
turn
on
asse
ts)
20.2
91.
341.
411.
251.
431.
191.
341.
321.
221.
211.
401.
271.
361.
341.
301.
44
(k)
Val
ue-a
t-ri
skor
oth
ersi
mu
lati
onan
alys
is13
.66
0.95
0.76
1.22
***
0.84
0.86
0.78
1.10
***
1.09
1.04
1.04
0.82
**0.
950.
920.
950.
86(e
)A
dju
sted
pre
sent
valu
e10
.78
0.85
0.93
0.72
*0.
970.
69**
0.87
0.80
0.80
0.79
0.80
0.91
0.78
0.92
0.79
0.99
*(h
)P
ro"
tabi
lity
inde
x11
.87
0.83
0.88
0.75
0.73
0.81
0.74
0.96
*0.
660.
670.
810.
830.
900.
760.
810.
98
198 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
%al
way
sC
EO
age
CE
Ote
nure
CE
OM
BA
Reg
ula
ted
Tar
get
deb
tra
tio
Pu
blic
corp
.F
orei
gnsa
les
For
tun
e50
0m
ailin
go
ral
mos
tal
way
sM
ean
'59
Yng
erL
ong
Shor
tY
esN
oY
esN
oN
oY
esY
esN
oY
esN
oN
oY
es
(b)
Inte
rnal
rate
ofre
turn
75.6
13.
093.
213.
062.
973.
16*
3.17
3.03
3.76
3.04
***
3.03
3.18
3.27
2.77
***
3.31
3.01
**3.
003.
57**
*
(a)
Net
pre
sent
valu
e74
.93
3.08
3.08
3.09
2.90
3.17
**3.
173.
00*
3.50
3.07
**2.
993.
23**
3.24
2.78
***
3.38
2.95
***
2.97
3.60
***
(f)P
ayba
ckpe
riod
56.7
42.
532.
832.
43**
*2.
802.
37**
*2.
482.
552.
052.
56**
2.65
2.43
*2.
452.
67*
2.62
2.49
2.57
2.35
(c)
Hu
rdle
rate
56.9
42.
482.
882.
38**
*2.
392.
512.
572.
423.
182.
42**
2.33
2.64
**2.
702.
10**
*2.
562.
432.
303.
28**
*(j)
Sen
siti
vity
anal
ysis
(e.g
.,`g
ooda
vs.`fa
ira
vs.`
bada)
51.5
42.
312.
202.
362.
202.
372.
412.
253.
142.
26**
*2.
242.
432.
372.
182.
362.
282.
222.
76**
*
(d)
Ear
nin
gsm
ulti
ple
app
roac
h38
.92
1.89
2.25
1.79
**1.
931.
861.
981.
861.
621.
901.
851.
962.
081.
56**
*1.
981.
841.
832.
15*
(g)
Dis
cou
nted
payb
ack
per
iod
29.4
51.
561.
941.
48**
*1.
721.
46*
1.68
1.49
1.52
1.60
1.57
1.61
1.56
1.60
1.62
1.53
1.51
1.84
*(l)
We
inco
rpor
ate
the`r
eal
opt
ionsa
of
apr
ojec
tw
hen
eval
uati
ng
it26
.59
1.47
1.68
1.40
*1.
561.
361.
491.
390.
951.
48*
1.44
1.46
1.40
1.59
1.53
1.43
1.44
1.57
(i)A
cco
unti
ngra
teof
retu
rn(o
rb
ook
rate
of
retu
rnon
asse
ts)
20.2
91.
341.
491.
331.
391.
341.
421.
291.
761.
30*
1.30
1.39
1.31
1.43
1.27
1.38
1.36
1.26
(k)
Val
ue-a
t-ri
skor
oth
ersi
mu
lati
onan
alys
is13
.66
0.95
1.07
0.90
0.92
0.93
0.99
0.88
1.76
0.89
*0.
771.
12**
*0.
891.
010.
900.
960.
861.
36**
*(e
)A
dju
sted
pre
sent
valu
e10
.78
0.85
1.18
0.75
***
0.88
0.80
0.74
0.91
*0.
670.
860.
880.
810.
830.
900.
740.
890.
860.
80(h
)P
ro"
tabi
lity
inde
x11
.87
0.83
0.87
0.83
0.95
0.77
*0.
830.
850.
570.
850.
750.
99**
0.76
1.00
**0.
810.
830.
850.
75
�Res
pon
den
tsar
eas
ked
tora
teon
asc
ale
of0
(nev
er)t
o4
(alw
ays)
.We
rep
ort
the
over
allm
ean
asw
ella
sth
e%
ofr
espo
nden
tsth
atan
swer
ed3
(alm
ost
alw
ays)
or4
(alw
ays)
.***
,**,
*d
eno
tes
asi
gni"
can
tdi!
eren
ceat
the
1%,
5%,a
nd
10%
leve
l,re
spec
tive
ly.
All
tab
leco
lum
ns
are
de"
ned
inT
able
1.
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 199
ratios. This is not just an artifact of "rm size. In unreported analysis, we "nda signi"cant di!erence between high- and low-leverage small "rms as well ashigh- and low-leverage large "rms. Interestingly, highly levered "rms are alsomore likely to use sensitivity and simulation analysis. Perhaps because ofregulatory requirements, utilities are more likely to use IRR and NPV andperform sensitivity and simulation analyses. We also "nd that CEOs with MBAsare more likely than non-MBA CEOs to use net present value, but the di!erenceis only signi"cant at the 10% level.
Firms that pay dividends are signi"cantly more likely to use NPV and IRRthan are "rms that do not pay dividends. This result is also robust to our analysis bysize. Public companies are signi"cantly more likely to use NPV and IRR than areprivate corporations. As the correlation analysis indicates in Table 1, many of theseattributes are correlated. For example, private corporations are also smaller "rms.
Other than NPV and IRR, the payback period is the most frequently usedcapital budgeting technique (rating of 2.53). This is surprising because "nancialtextbooks have lamented the shortcomings of the payback criterion for decades.(Payback ignores the time value of money and cash #ows beyond the cuto! date;the cuto! is usually arbitrary.) Small "rms use the payback period (rating of2.72) almost as frequently as they use NPV or IRR. In untabulated analysis, we"nd that among small "rms, CEOs without MBAs are more likely to use thepayback criterion. The payback is most popular among mature CEOs (rating of2.83). In separate examinations of small and large "rms, we "nd that matureCEOs use payback signi"cantly more often than younger CEOs. Payback isalso frequently used by CEOs with long tenure (rating of 2.80). Few "rms use thediscounted payback (rating of 1.56), a method that eliminates one of the paybackcriterion's de"ciencies by accounting for the time value of money.
It is sometimes argued that the payback approach is rational for severelycapital constrained "rms: if an investment project does not pay positive cash#ows early on, the "rm will cease operations and therefore not receive positivecash #ows that occur in the distant future, or else will not have the resources topursue other investments during the next few years (Weston and Brigham, 1981,p. 405). We do not "nd any evidence to support this claim because we "nd norelation between the use of payback and leverage, credit ratings, or dividendpolicy. Our "nding that payback is used by older, longer-tenure CEOs withoutMBAs instead suggests that lack of sophistication is a driving factor behind thepopularity of the payback criterion.
McDonald (1998) notes that rules of thumb such as payback and hurdle ratescan approximate optimal decision rules that account for the option-like featuresof many investments, especially in the evaluation of very uncertain investments.If small "rms have more volatile projects than do large "rms, this could explainwhy small "rms use these ad hoc decision rules. It is even possible that small"rms use these rules not because they realize that they approximate the optimalrule but simply because the rules have worked in the past.
200 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
A number of "rms use the earnings multiple approach for project evaluation.There is weak evidence that large "rms are more likely to employ this approachthan are small "rms. We "nd that a "rm is signi"cantly more likely to useearnings multiples if it is highly levered. The in#uence of leverage on theearnings multiple approach is also robust across size (i.e., highly levered "rms,whether they are large or small, frequently use earnings multiples).
In summary, compared to previous research, our results suggest increasedprominence of net present value as an evaluation technique. In addition, thelikelihood of using speci"c evaluation techniques is linked to "rm size, "rmleverage, and CEO characteristics. In particular, small "rms are signi"cantlyless likely to use net present value. They are also less likely to use supple-mentary sensitivity and VaR analyses. The next section takes this analysis onestep further by detailing the speci"c methods "rms use to obtain the cost ofcapital, the most important risk factors, and a speci"c capital budgetingscenario.
4. Cost of capital
4.1. Methodology
Our "rst task is to determine how "rms calculate the cost of equity capital.We explore whether "rms use the capital asset pricing model (CAPM), a multi-beta CAPM (with extra risk factors in addition to the market beta), averagehistorical returns, or a dividend discount model. The results in Table 3 andsummarized in Fig. 3 indicate that the CAPM is by far the most popular methodof estimating the cost of equity capital: 73.5% of respondents always or almostalways use the CAPM (rating of 2.92; see also Fig. 1H). The second and thirdmost popular methods are average stock returns and a multibeta CAPM,respectively. Few "rms back the cost of equity out from a dividend discountmodel (rating of 0.91). This sharply contrasts with the "ndings of Gitmanand Mercurio (1982) who survey 177 Fortune 1000 "rms and "nd thatonly 29.9% of respondents use the CAPM `in some fashiona but "nd that31.2% of the participants in their survey use a version of the dividend discountmodel to establish their cost of capital. More recently, Bruner et al. (1998)"nd that 85% of their 27 best-practice "rms use the CAPM or a modi"edCAPM. While the CAPM is popular, we show later that it is not clear thatthe model is applied properly in practice. Of course, even if it is applied pro-perly, it is not clear that the CAPM is a very good model (see Fama and French,1992).
The cross-sectional analysis is particularly illuminating. Large "rms are muchmore likely to use the CAPM than are small "rms (rating of 3.27 versus 2.49,respectively). Smaller "rms are more inclined to use a cost of equity capital that
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 201
Tab
le3
Surv
eyre
spon
ses
toth
equ
esti
on:
does
your"
rmes
tim
ate
the
cost
ofeq
uity
cap
ital
?(I
f`n
oa,p
leas
ego
ton
ext
ques
tio
n).
If`y
esa,
how
doyo
ude
term
ine
you
r"
rm's
cost
ofeq
uity
cap
ital
?�
%al
way
sSi
zeP
/EL
ever
age
Inve
stm
ent
grad
eP
ayd
ivid
ends
Ind
ustr
yM
anag
emen
to
wn
ersh
ipo
ral
mos
tal
way
sM
ean
Smal
lL
arge
Gro
wth
No
n-G
Lo
wH
igh
Yes
No
Yes
No
Man
u.
Oth
ers
Lo
wH
igh
(b)
usi
ng
the
cap
ital
asse
tpr
icin
gm
odel
(CA
PM
,th
e`b
etaa
appr
oach
)73
.49
2.92
2.49
3.27
***
3.19
3.03
2.57
3.23
***
3.13
3.34
3.00
2.76
3.02
2.87
3.26
2.36
***
(a)w
ith
aver
age
hist
oric
alre
turn
so
nco
mm
on
sto
ck39
.41
1.72
1.80
1.65
1.65
1.78
1.80
1.56
1.67
1.48
1.77
1.63
1.60
1.84
1.66
1.87
(c)
usin
gth
eC
AP
Mbu
tin
clu
din
gso
me
extr
a`r
isk
fact
orsa
34.2
91.
561.
391.
70*
1.62
1.48
1.57
1.45
1.71
1.76
1.51
1.54
1.69
1.49
1.59
1.44
(f)ba
ckou
tfr
om
dis
coun
ted
divi
den
d/e
arn
ings
mo
del,
e.g.
,Pri
ce"
Div
./(co
sto
fca
p.}
gro
wth
)
15.7
40.
910.
960.
870.
901.
020.
721.
05**
0.92
0.98
0.90
0.95
0.98
0.80
0.97
1.10
(d)
wha
teve
rou
rin
vest
ors
tell
usth
eyre
quir
e13
.93
0.86
1.22
0.54
***
0.76
0.44
**0.
920.
880.
480.
79*
0.70
1.12
**0.
800.
970.
651.
23**
*(e
)by
regu
lato
ryd
ecis
ions
7.04
0.44
0.37
0.50
0.56
0.32
*0.
480.
360.
510.
440.
540.
24**
0.44
0.44
0.51
0.41
CE
Oag
eC
EO
tenu
reC
EO
MB
AR
egu
late
dT
arge
tde
bt
rati
oP
ubl
icco
rp.
For
eign
sale
sF
ortu
ne
500
mai
ling
'59
Yng
erL
ong
Shor
tY
esN
oY
esN
oN
oY
esY
esN
oY
esN
oN
oY
es
(b)
usi
ng
the
cap
ital
asse
tpr
icin
gm
odel
(CA
PM
,th
e`b
etaa
app
roac
h)2.
852.
932.
832.
963.
082.
77*
3.00
2.87
2.83
3.03
3.13
2.13
***
3.23
2.75
**2.
783.
46**
*
(a)
wit
hav
erag
eh
isto
rica
lre
turn
son
com
mo
nst
ock
2.43
1.54
***
1.70
1.73
1.53
1.90
*1.
601.
701.
641.
801.
651.
911.
621.
781.
801.
38*
(c)u
sin
gth
eC
AP
Mbu
tin
clud
ing
som
eex
tra`r
isk
fact
orsa
1.91
1.48
*1.
661.
491.
621.
482.
171.
41**
1.53
1.49
1.56
1.53
1.57
1.52
1.38
2.17
***
(f)ba
ckou
tfr
om
dis
coun
ted
div
iden
d/ea
rnin
gsm
odel
,e.
g.,P
rice
"D
iv./(
cost
of
cap.}
gro
wth
)1.
210.
82**
1.05
0.83
0.78
1.02
*1.
200.
880.
930.
920.
990.
68*
0.81
0.97
0.90
0.95
(d)
wha
teve
rou
rin
vest
ors
tell
usth
eyre
quir
e0.
760.
871.
020.
790.
720.
99*
0.69
0.87
0.94
0.81
0.67
1.53
***
0.65
0.97
**0.
960.
46**
(e)
byre
gula
tory
dec
isio
ns0.
320.
470.
390.
430.
410.
472.
190.
28**
*0.
490.
430.
490.
27*
0.20
0.55
***
0.37
0.71
**
�Res
pon
den
tsar
eas
ked
tora
teo
na
scal
eo
f0
(nev
er)
to4
(alw
ays)
.We
rep
ort
the
over
all
mea
nas
wel
las
the
%o
fre
spon
den
tsth
atan
swer
ed3
(alm
ost
alw
ays)
and
4(a
lway
s).*
**,*
*,*
den
ote
sa
sign
i"ca
nt
di!
eren
ceat
the
1%,
5%an
d10
%le
vel,
resp
ecti
vely
.All
tabl
eco
lum
nsar
ed
e"n
edin
Tab
le1.
202 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Fig. 3. Survey evidence on the popularity of di!erent methods of calculat the cost of equity capital.We report the percentage of CFOs who always or almost always use a particular technique. CAPMrepresents the capital asset pricing model. The survey is based on the responses of 392 CFOs.
is determined by `what investors tell us they requirea. CEOs with MBAs aremore likely to use the single-factor CAPM or the CAPM with extra risk factorsthan are non-MBA CEOs, but the di!erence is only signi"cant for the single-factor CAPM.
We also "nd that "rms with low leverage or small management ownership aresigni"cantly more likely to use the CAPM. We "nd signi"cant di!erences forprivate versus public "rms (public more likely to use the CAPM). This is perhapsexpected given that the beta of the private "rm could only be calculated viaanalysis of comparable publicly traded "rms. Finally, we "nd that "rms withhigh foreign sales are more likely to use the CAPM.
Given the sharp di!erence between large and small "rms, it is important tocheck whether some of these control e!ects just proxy for size. It is, indeed, thecase that foreign sales proxy for size. Table 1 shows that that there is a signi"-cant correlation between percent of foreign sales and size. When we analyze theuse of the CAPM by foreign sales controlling for size, we "nd no signi"cantdi!erences. However, this is not true for some of the other control variables.There is a signi"cant di!erence in use of the CAPM across leverage that isrobust to size. The public/private e!ect is also robust to size. Finally, thedi!erence in the use of the CAPM based on management ownership holds forsmall "rms but not for large "rms. That is, among small "rms, CAPM use isinversely related to managerial ownership. There is no signi"cant relation forlarger "rms.
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 203
4.2. Specixc risk factors
Table 4 investigates sources of risk other than market risk, and how they aretreated in project evaluation. The list of risk factors includes the fundamentalfactors in Fama and French (1992), and momentum as de"ned in Jegadeesh andTitman (1993), as well as the macroeconomic factors in Chen et al. (1986) andFerson and Harvey (1991).
The format of Table 4 is di!erent from the others. We ask whether, inresponse to these risk factors, the "rm modi"es its discount rate, cash #ows,both, or neither. We report the percentage of respondents for each category. Inthe cross-tabulations across each of the demographic factors, we test whetherthe `neithera category is signi"cantly di!erent conditional on "rm character-istics.
Overall, the most important additional risk factors are interest rate risk,exchange rate risk, business cycle risk, and in#ation risk (see Fig. 4). For thecalculation of discount rates, the most important factors are interest rate risk,size, in#ation risk, and foreign exchange rate risk. For the calculation of cash#ows, many "rms incorporate the e!ects of commodity prices, GDP growth,in#ation, and foreign exchange risk.
Interestingly, few "rms adjust either discount rates or cash #ows for book-to-market, distress, or momentum risks. Only 13.1% of respondents consider thebook-to-market ratio in either the cash #ow or discount rate calculations.Momentum is only considered important by 11.1% of the respondents.
Small and large "rms have di!erent priorities when adjusting for risk. Forlarge "rms, the most important risk factors (in addition to market risk) areforeign exchange risk, business cycle risk, commodity price risk, and interest raterisk. This closely corresponds to the set of factors detailed in Ferson and Harvey(1993) in their large-sample study of multibeta international asset pricing mod-els. Ferson and Harvey "nd that the most important additional factor is foreignexchange risk. Table 4 shows that foreign exchange risk is by far the mostimportant nonmarket risk factor for large "rms (61.7% of the large "rms adjustfor foreign exchange risk; the next closest is 51.4% adjusting for business cyclerisk).
The ordering is di!erent for small "rms. Small "rms are more a!ected byinterest rate risk than they are by foreign exchange risk. This asymmetry in riskexposure is consistent with the analysis of Jagannathan and Wang (1996) andJagannathan et al. (1998). They argue that small "rms are more likely to beexposed to labor income risk and, as a result, we should expect to "nd these"rms relying on a di!erent set of risk factors, and using the CAPM lessfrequently, when estimating their cost of capital.
As might be expected, "rms with considerable foreign sales are sensitive tounexpected exchange rate #uctuations. Fourteen percent of "rms with substan-tial foreign exposure adjust discount rates for foreign exchange risk, 22% adjust
204 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Fig. 4. Survey evidence on types of multibeta risk that are important for adjusting cash #ows ordiscount rates. We report the percentage of CFOs who always or almost always adjust fora particular type of risk. The survey is based on the responses of 392 CFOs.
cash #ows, and 32% adjust both. These "gures represent the highest incidence of`adjusting somethinga for any type of non-market risk, for any demographic.
There are some interesting observations for the other control variables.Highly levered "rms are more likely to consider business cycle risk important;surprisingly, however, indebtedness does not a!ect whether "rms adjust forinterest rate risk, term structure risk, or distress risk. Growth "rms are muchmore sensitive to foreign exchange risk than are nongrowth "rms. (Table 4 onlyreports the results for four control variables; A full version of Table 4 is availableon the Internet at http://www.duke.edu/&charvey/Research/indexr.htm.)
4.3. Project versus xrm risk
Finally, we explore how the cost of equity models are used. In particular, weconsider an example of how a "rm evaluates a new project in an overseasmarket. We are most interested in whether corporations consider the company-wide risk or the project risk in evaluating the project.
Table 5 contains some surprising results. Remarkably, most "rms would usea single company-wide discount rate to evaluate the project; 58.8% of therespondents would always or almost always use the company-wide discountrate, even though the hypothetical project would most likely have di!erent riskcharacteristics. However, 51% of the "rms said they would always or almost
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 205
Tab
le4
Surv
eyre
spon
ses
toth
eq
uest
ion
:wh
enva
luin
ga
pro
ject
,do
you
adju
stei
ther
the
dis
coun
trat
eor
cash#
ow
sfo
rth
efo
llow
ing
risk
fact
ors?
(Che
ckth
em
osta
ppr
opri
ate
box
for
each
fact
or)
.Per
cent
age
ofre
spo
nden
tsch
oosi
ng
each
cate
gory
isre
por
ted�
Ove
rall
Siz
eP
/E
Dis
coun
tra
teC
ash
Flo
wB
oth
Nei
ther
Dis
coun
tra
teC
ash
Flo
wB
oth
Nei
ther
Dis
c.ra
teC
ash#
ow
Bo
thN
eith
erSm
all
Lar
geSm
all
Lar
geSm
all
Lar
geSm
all
Lar
geG
row
thN
on-
GG
row
thN
on-
GG
row
thN
on-
GG
row
thN
on-
G
(b)
Inte
rest
rate
risk
(ch
ange
inge
nera
lle
vel
ofin
tere
stra
tes)
15.3
08.
7824
.65
51.2
717
.33
12.6
77.
4310
.67
29.7
017
.33
45.5
459
.33*
*13
.39
7.06
7.09
16.4
722
.83
18.8
256
.69
57.6
5
(f)F
orei
gnex
chan
geri
sk10
.80
15.3
418
.75
55.1
17.
4315
.44
9.90
22.8
215
.35
23.4
967
.33
38.2
6***
10.2
418
.75
14.9
622
.50
22.8
323
.75
51.9
735
.00*
*(d
)G
DP
or
bus
ines
scy
cle
risk
6.84
18.8
018
.80
55.5
66.
936.
7612
.87
27.0
319
.80
17.5
760
.40
48.6
5**
6.98
7.41
24.0
318
.52
22.4
814
.81
46.5
159
.26*
(a)
Ris
ko
fun
expe
cted
in#
atio
n11
.90
14.4
511
.90
61.7
613
.43
9.93
9.95
20.5
314
.93
7.95
61.6
961
.59
12.4
09.
6414
.73
16.8
710
.08
12.0
562
.79
61.4
5(h
)S
ize
(sm
all"
rms
bein
gri
skie
r)14
.57
6.00
13.4
366
.00
14.4
314
.67
7.46
4.00
16.9
28.
6761
.19
71.3
3**
14.8
415
.66
7.03
3.61
17.1
99.
6460
.94
68.6
7
(e)
Co
mm
odit
ypr
ice
risk
2.86
18.8
610
.86
67.4
32.
493.
3812
.94
27.0
39.
4512
.84
75.1
256
.76*
**3.
124.
9420
.31
24.6
912
.50
7.41
64.0
662
.96
(c)
Ter
mst
ruct
ure
risk
(cha
nge
inth
elo
ng-
term
vs.
shor
t-te
rmin
tere
stra
te)
8.57
3.71
12.5
775
.14
10.4
56.
082.
994.
7314
.93
9.46
71.6
479
.73*
7.03
6.10
3.12
6.10
10.9
417
.07
78.9
170
.73
(g)
Dis
tres
sri
sk(p
rob
abili
tyof
ban
kru
ptcy
)7.
416.
274.
8481
.48
5.94
9.40
4.95
8.05
6.93
2.01
82.1
879
.87
6.98
15.8
56.
986.
106.
98n/
a79
.07
76.8
3
(i)`M
arke
t-to
-bo
oka
rati
o(r
atio
ofm
ark
etva
lue
of
"rm
tobo
okva
lue
of
asse
ts)
3.98
1.99
7.10
86.9
34.
463.
361.
492.
688.
914.
7085
.15
89.2
62.
388.
433.
171.
205.
566.
0288
.89
84.3
4
(j)M
omen
tum
(rec
ent
stoc
kpr
ice
perf
orm
ance
).3.
432.
864.
8688
.86
3.98
2.70
2.99
2.70
6.47
2.70
86.5
791
.89
3.15
4.94
2.36
4.94
4.72
1.23
89.7
688
.89
Lev
erag
eF
ore
ign
sale
s
Dis
coun
tra
teC
ash
Flo
wB
oth
Nei
ther
Dis
coun
tra
teC
ash
Flo
wB
oth
Nei
ther
Lo
wH
igh
Lo
wH
igh
Lo
wH
igh
Lo
wH
igh
Yes
No
Yes
No
Yes
No
Yes
No
(b)
Inte
rest
rate
risk
(ch
ange
inge
nera
lle
vel
ofin
tere
stra
tes)
14.2
918
.12
10.7
16.
5224
.40
23.1
950
.60
52.1
713
.54
15.9
48.
338.
7619
.79
26.2
958
.33
49.0
0
(f)F
orei
gnex
chan
geri
sk12
.88
7.09
12.8
818
.44
17.1
821
.99
57.0
652
.48
13.8
39.
5222
.34
12.3
031
.91
13.4
931
.91
64.6
8***
(d)
GD
Po
rb
usin
ess
cycl
eri
sk6.
834.
9613
.66
28.3
716
.15
24.8
263
.35
41.8
4***
6.45
7.14
26.8
815
.87
16.1
319
.44
50.5
457
.54
(a)
Ris
ko
fun
expe
cted
in#
atio
n13
.94
10.7
110
.91
16.4
38.
4813
.57
66.6
759
.29
7.29
13.5
519
.79
12.7
513
.54
11.5
559
.38
62.1
5(h
)Siz
e(s
mal
l"rm
sbe
ing
risk
ier)
10.3
715
.60
6.71
5.67
17.6
89.
9365
.24
68.0
912
.77
15.0
27.
455.
5311
.70
14.2
368
.09
64.4
3(e
)C
om
mod
ity
pric
eri
sk1.
244.
3214
.29
26.6
212
.42
8.63
72.0
560
.43*
*3.
232.
7926
.88
15.1
410
.75
10.7
659
.14
71.3
1**
(c)
Ter
mst
ruct
ure
risk
(cha
nge
inth
elo
ng-
term
vs.
shor
t-te
rmin
tere
stra
te)
6.17
11.4
36.
172.
1410
.49
15.7
177
.16
70.7
16.
459.
524.
303.
5713
.98
12.3
075
.27
74.6
0
(g)
Dis
tres
sri
sk(p
rob
abili
tyof
ban
kru
ptcy
)4.
828.
456.
636.
344.
824.
2383
.73
80.9
99.
386.
757.
295.
952.
085.
9581
.25
80.9
5
(i)`M
arke
t-to
-bo
oka
rati
o(r
atio
ofm
ark
etva
lue
of
"rm
tobo
okva
lue
of
asse
ts)
3.61
4.32
3.61
0.72
6.63
7.19
86.1
487
.77
4.26
3.95
5.32
0.79
5.32
7.91
85.1
187
.35
(j)M
omen
tum
(rec
ent
stoc
kpr
ice
perf
orm
ance
)3.
683.
552.
453.
554.
914.
2688
.96
88.6
54.
263.
193.
192.
794.
265.
1888
.30
88.8
4
�Per
cent
age
ofre
spo
nden
tsch
oosi
ng
each
cate
gory
isre
por
ted.
Th
ep
erce
nta
ges
for
dis
coun
tra
te,c
ash#
ow,b
oth
and
neit
her
shou
ldsu
mto
100.
***,
**,*
den
ote
sa
sign
i"ca
ntdi!
eren
ceat
the
1%,5
%,a
nd
10%
leve
l,re
spec
tive
ly.
All
tab
leco
lum
ns
are
de"
ned
inT
able
1.
Tab
le5
Surv
eyre
spon
ses
toth
equ
esti
on:
Ho
wfr
equ
entl
yw
ould
your
com
pan
yus
eth
efo
llow
ing
dis
coun
tra
tes
wh
enev
alu
atin
ga
new
pro
ject
inan
over
seas
mar
ket?
To
eval
uat
eth
ispr
ojec
tw
ew
oul
dus
e�
%al
way
sSi
zeP
/EL
ever
age
Inve
stm
ent
grad
eP
ayd
ivid
ends
Ind
ustr
yM
anag
emen
to
wn
ersh
ipo
ral
mos
tal
way
sM
ean
Smal
lL
arge
Gro
wth
No
n-G
Lo
wH
igh
Yes
No
Yes
No
Man
u.
Oth
ers
Lo
wH
igh
(a)
The
dis
coun
tra
tefo
ro
uren
tire
com
pan
y58
.79
2.50
2.50
2.50
2.76
2.37
**2.
452.
582.
412.
83**
2.46
2.53
2.56
2.32
*2.
612.
41(d
)A
risk
-mat
ched
disc
ount
rate
for
this
part
icul
arp
roje
ct(c
onsi
der
ing
both
coun
try
and
indu
stry
)50
.95
2.09
1.86
2.36
***
2.20
2.26
1.99
2.30
**2.
432.
252.
311.
82**
*2.
222.
012.
222.
01
(b)
Th
edi
scou
ntra
tefo
rth
eo
vers
eas
mar
ket
(co
untr
yd
isco
unt
rate
)34
.52
1.65
1.49
1.82
**1.
841.
691.
541.
81*
1.82
2.01
1.75
1.52
*1.
861.
42**
*1.
701.
52
(c)
Ad
ivis
iona
ld
isco
unt
rate
(ifth
epr
ojec
tlin
eof
busi
nes
sm
atch
esa
dom
esti
cd
ivis
ion)
15.6
10.
950.
821.
09**
1.12
1.04
0.88
1.08
*1.
171.
051.
050.
84*
1.01
0.90
0.96
1.08
(e)
Ad
i!er
ent
dis
coun
tra
tefo
rea
chco
mpo
nen
tca
sh#
ow
that
has
adi!
eren
tri
skch
arac
teri
stic
(e.g
.de
pre
ciat
ion
vs.o
pera
ting
cash#
ows)
9.87
0.66
0.68
0.64
0.49
0.85
***
0.61
0.68
0.75
0.58
0.68
0.64
0.68
0.65
0.56
0.85
**
%al
way
sC
EO
age
CE
Ote
nure
CE
OM
BA
Reg
ula
ted
Tar
get
deb
tra
tio
Pu
blic
corp
.F
orei
gnsa
les
For
tun
e50
0m
ailin
go
ral
mos
tal
way
sM
ean
'59
Yng
erL
ong
Shor
tY
esN
oY
esN
oN
oY
esY
esN
oY
esN
oN
oY
es
(a)
The
dis
coun
tra
tefo
ro
uren
tire
com
pan
y58
.79
2.50
2.54
2.49
2.18
2.64
***
2.49
2.51
2.00
2.52
*2.
392.
64*
2.55
2.42
2.87
2.33
***
2.57
2.20
**(d
)A
risk
-mat
ched
disc
ount
rate
for
this
part
icul
arp
roje
ct(c
onsi
der
ing
both
coun
try
and
indu
stry
)50
.95
2.09
2.31
2.02
*2.
112.
062.
201.
992.
552.
03*
1.90
2.25
**2.
241.
79**
*2.
212.
021.
972.
61**
*
(b)
Th
edi
scou
ntra
tefo
rth
eo
vers
eas
mar
ket
(co
untr
yd
isco
unt
rate
)34
.52
1.65
1.80
1.61
1.49
1.73
*1.
771.
601.
501.
661.
701.
581.
781.
41**
1.81
1.58
1.58
1.92
*
(c)
Ad
ivis
iona
ld
isco
unt
rate
(ifth
epr
ojec
tlin
eof
busi
nes
sm
atch
esa
dom
esti
cd
ivis
ion)
15.6
10.
951.
180.
87**
0.99
0.92
0.88
0.98
1.27
0.89
*0.
911.
011.
080.
66**
*0.
940.
930.
891.
17*
(e)
Ad
i!er
ent
dis
coun
tra
tefo
rea
chco
mpo
nen
tca
sh#
ow
that
has
adi!
eren
tri
skch
arac
teri
stic
(e.g
.de
pre
ciat
ion
vs.o
pera
ting
cash#
ows)
9.87
0.66
0.72
0.62
0.55
0.68
0.59
0.67
0.38
0.67
0.67
0.57
0.61
0.79
*0.
630.
680.
710.
46*
�Res
pon
den
tsar
eas
ked
tora
teo
na
scal
eo
f0
(nev
er)
to4
(alw
ays)
.We
rep
ort
the
over
all
mea
nas
wel
las
the
%o
fre
spon
den
tsth
atan
swer
ed3
(alm
ost
alw
ays)
and
4(a
lway
s).*
**,*
*,*
den
ote
sa
sign
i"ca
nt
di!
eren
ceat
the
1%,
5%,a
nd
10%
leve
l,re
spec
tive
ly.
All
tab
leco
lum
ns
are
de"
ned
inT
able
1.
208 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
always use a risk-matched discount rate to evaluate this project. These resultsare related to Bierman (1993) who "nds that 93% of the Fortune 100 industrial"rms use the company-wide weighted average cost of capital for discounting,72% use the rate applicable to the project based on the risk or the nature of theproject, and 35% use a rate based on the division's risk.
The reliance of many "rms on a company-wide discount rate might makesense if these same "rms adjust cash #ows for foreign exchange risk whenconsidering risk factors (i.e., in Table 4). However in untabulated results, we "ndthe opposite: "rms that do not adjust cash #ows for foreign exchange risk arealso relatively less likely (compared to "rms that adjust for foreign exchangerisk) to use a risk-matched discount rate when evaluating an overseas project.
Large "rms are signi"cantly more likely to use the risk-matched discount ratethan are small "rms (rating of 2.34 versus 1.86). This is also con"rmed in ouranalysis of Fortune 500 "rms, which are much more likely to use the risk-matched discount rate than the "rm-wide discount rate to evaluate the foreignproject (rating of 2.61 versus 1.97). Very few "rms use a di!erent discount rate toseparately value di!erent cash #ows within the same project (rating of 0.66), asBrealey and Myers (1996) suggest they should for cash #ows such as depreciation.
The analysis across "rm characteristics reveals some interesting patterns.Growth "rms are more likely to use a company-wide discount rate to evaluateprojects. Surprisingly, "rms with foreign exposure are signi"cantly more likelyto use the company-wide discount rate to value an overseas project. Publiccorporations are more likely to use a risk-matched discount rate than areprivate corporations; however, this result is not robust to controlling for size.CEOs with short tenures are more likely to use a company-wide discount rate(signi"cant at the 5% level for both large and small "rms).
5. Capital structure
Our survey has separate questions about debt, equity, debt maturity, convert-ible debt, foreign debt, target debt ratios, credit ratings, and actual debt ratios.Instead of stepping through the responses security by security, this sectiondistills the most important "ndings from the capital structure questions andpresents the results grouped by theoretical hypothesis or concept. These group-ings are neither mutually exclusive nor all-encompassing; they are intendedprimarily to organize the exposition.
5.1. Trade-ow theory of capital structure choice
5.1.1. Target debt ratios and the costs and benexts of debtOne of the longest-standing questions about capital structure is whether "rms
have target debt ratios. The trade-o! theory says that "rms have optimal
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 209
Fig. 5. Survey evidence on some of the factors that a!ect the decision to issue debt. The survey isbased on the responses of 392 CFOs.
debt}equity ratios, which they determine by trading o! the bene"ts of debt withthe costs (e.g., Scott, 1976). In traditional trade-o! models, the chief bene"t ofdebt is the tax advantage of interest deductibility (Modigliani and Miller, 1963).The primary costs are those associated with "nancial distress and the personaltax expense bondholders incur when they receive interest income (Miller, 1977).In this section we discuss the traditional factors in the trade-o! theory, namelydistress costs and tax costs and bene"ts. Many additional factors (e.g., informa-tional asymmetry, agency costs) can be modeled in a trade-o! framework. Wediscuss these alternative costs and bene"ts in separate sections below.
Table 6 and Fig. 5 show the factors that determine the appropriate amount ofdebt for the "rm. The CFOs tell us that the corporate tax advantage of debt ismoderately important in capital structure decisions: Row a of Table 6 showsthat the mean response is 2.07 on a scale from 0 to 4 (0 meaning not important, 4meaning very important). The tax advantage is most important for large,regulated, and dividend-paying "rms } companies that probably have highcorporate tax rates and therefore large tax incentives to use debt. Desai (1998)shows that "rms issue foreign debt in response to relative tax incentives, so weinvestigate whether "rms issue debt when foreign tax treatment is favorable. We"nd that favorable foreign tax treatment relative to the U.S. is fairly important(overall rating of 2.26 in Table 7). Big "rms (2.41) with large foreign exposure(2.50) are relatively likely to indicate that foreign tax treatment is an important
210 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
factor. This could indicate that "rms need a certain level of sophistication andexposure to perform international tax planning.
In contrast, we "nd very little evidence that "rms directly consider personaltaxes when deciding on debt policy (rating of 0.68 in Table 6) or equity policy(rating of 0.82 in Table 8, the least popular equity issuance factor). Therefore, itseems unlikely that "rms target investors in certain tax clienteles (although wecan not rule out the possibility that investors choose to invest in "rms based onpayout policy, or that executives respond to personal tax considerations to theextent that they are re#ected in market prices, see Graham, 1999a).
When we ask "rms directly about whether potential costs of distress a!ecttheir debt decisions, we "nd they are not very important (rating of 1.24 inTable 6), although they are relatively important among speculative-grade "rms.However, "rms are very concerned about their credit ratings (rating of 2.46, thesecond most important debt factor), which can be viewed as an indication ofconcern about distress. Among utilities and "rms that have rated debt, creditratings are a very important determinant of debt policy. Credit ratings are alsoimportant for large "rms (3.14) that are in the Fortune 500 (3.31). Finally, CFOsare also concerned about earnings volatility when making debt decisions (ratingof 2.32), which is consistent with the trade-o! theory's prediction that "rmsreduce debt usage when the probability of bankruptcy is high (Castanias, 1983).
We ask directly whether "rms have an optimal or `targeta debt}equity ratio.Nineteen percent of the "rms do not have a target debt ratio or target range (seeFig. 1G). Another 37% have a #exible target, and 34% have a somewhat tighttarget or range. The remaining 10% have a strict target debt ratio (see Fig. 6).These overall numbers provide mixed support for the notion that companiestrade o! costs and bene"ts to derive an optimal debt ratio. However, un-tabulated analysis shows that large "rms are more likely to have target debtratios: 55% of large "rms have at least somewhat strict target ratios, comparedto 36% of small "rms. Targets that are tight or somewhat strict are morecommon among investment-grade (64%) than speculative "rms (41%), andamong regulated (67%) than unregulated "rms (43%). Targets are important ifthe CEO has short tenure or is young, and when the top three o$cers own lessthan 5% of the "rm.
Finally, the CFOs tell us that their companies issue equity to maintaina target debt}equity ratio (rating of 2.26; Row e of Table 8), especially if their"rm is highly levered (2.68), "rm ownership is widely dispersed (2.64), or theCEO is young (2.41). Overall, the survey evidence provides moderate support forthe trade-o! theory.
5.1.2. Deviations from target debt ratiosActual debt ratios vary across "rms and through time. Such variability might
occur if debt intensity is measured relative to the market value of equity, and yet"rms do not rebalance their debt lock-step with changes in equity prices. Our
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 211
Tab
le6
Surv
eyre
spon
ses
toth
equ
esti
on:
Wha
tfa
cto
rsa!
ect
how
you
cho
ose
the
appr
opri
ate
amou
nto
fde
bt
for
you
r"
rm?�
%im
port
ant
Size
P/E
Lev
erag
eIn
vest
men
tgr
ade
Pay
div
iden
dsIn
dus
try
Man
agem
ent
ow
ner
ship
or
very
impo
rtan
tM
ean
Smal
lL
arge
Gro
wth
No
n-G
Lo
wH
igh
Yes
No
Yes
No
Man
u.
Oth
ers
Lo
wH
igh
(g)
Fin
anci
al#
exib
ility
(we
rest
rict
deb
tso
we
hav
een
oug
hin
tern
alfu
nds
avai
lab
leto
pur
sue
new
proj
ects
wh
enth
eyco
me
alon
g)
59.3
82.
592.
542.
652.
612.
752.
612.
602.
712.
592.
732.
40**
*2.
672.
522.
682.
41**
(d)
Our
cred
itra
tin
g(a
sas
sign
edb
yra
ting
agen
cies
)57
.10
2.46
1.92
3.14
***
2.89
2.81
2.29
2.64
**3.
363.
11**
2.76
2.04
***
2.52
2.39
2.81
1.99
***
(h)
Th
evo
lati
lity
of
our
earn
ings
and
cash#
ows
48.0
82.
322.
292.
362.
412.
252.
252.
322.
112.
44**
2.33
2.28
2.35
2.31
2.32
2.41
(a)
The
tax
adva
ntag
eo
fin
tere
stde
duc
tib
ility
44.8
52.
071.
772.
44**
*2.
362.
271.
992.
26**
2.32
2.54
2.35
1.65
***
2.30
1.79
***
2.27
1.89
***
(e)
The
tran
sact
ion
sco
sts
and
fees
for
issu
ing
deb
t33
.52
1.95
2.07
1.81
**1.
981.
801.
941.
871.
852.
061.
912.
021.
891.
951.
882.
02
(c)
The
deb
tle
vels
ofo
ther"
rms
ino
urin
dus
try
23.4
01.
491.
291.
77**
*1.
721.
521.
361.
70**
*1.
801.
711.
631.
34**
1.38
1.66
**1.
571.
37*
(b)
Th
epo
tent
ial
cost
sof
ban
kru
ptcy
,n
ear-
ban
kru
ptcy
,or"
nan
cial
dis
tres
s21
.35
1.24
1.36
1.10
**1.
291.
02*
1.16
1.37
**0.
991.
40**
1.27
1.21
1.31
1.22
1.30
1.33
(i)W
elim
itd
ebt
soo
urcu
stom
ers/
supp
liers
are
not
wor
ried
abo
uto
ur"
rmgo
ing
out
ofb
usin
ess
18.7
21.
241.
201.
301.
431.
00**
*1.
341.
201.
231.
141.
191.
301.
211.
40*
1.17
1.45
**
(n)
We
rest
rict
our
bor
row
ing
soth
atp
ro"
tsfr
om
new
/fu
ture
proj
ects
can
be
capt
ure
dfu
llyb
ysh
areh
old
ers
and
do
not
hav
eto
be
paid
out
asin
tere
stto
deb
thol
der
s
12.5
71.
011.
160.
80**
*1.
090.
69**
*1.
180.
83**
*0.
770.
850.
951.
061.
080.
970.
781.
30**
*
(j)W
etr
yto
have
enou
ghd
ebt
that
we
are
not
anat
trac
tive
tak
eove
rta
rget
4.75
0.73
0.57
0.91
***
0.95
0.86
0.62
0.90
***
0.84
0.96
0.76
0.66
0.83
0.66
*0.
850.
74
(f)T
hep
erso
nal
tax
cost
our
inve
sto
rsfa
cew
hen
they
rece
ive
inte
rest
inco
me
4.79
0.68
0.59
0.72
*0.
530.
80**
0.68
0.63
0.87
0.51
***
0.71
0.55
*0.
650.
630.
650.
72
(k)
Ifw
eis
sue
deb
tou
rco
mp
etit
ors
kno
wth
atw
ear
eve
ryu
nlik
ely
tore
duce
our
out
put
2.25
0.40
0.41
0.37
0.48
0.32
*0.
330.
47**
0.38
0.51
0.38
0.41
0.46
0.36
0.37
0.52
**
(m)
To
ensu
reth
atu
pper
man
agem
ent
wor
ksha
rdan
de$
cien
tly,
we
issu
esu$
cien
td
ebt
tom
ake
sure
that
ala
rge
port
ion
of
our
cash#
owis
com
mit
ted
toin
tere
stp
aym
ents
1.69
0.33
0.33
0.32
0.32
0.28
0.22
0.49
***
0.28
0.38
0.32
0.34
0.40
0.26
**0.
330.
35
(l)A
hig
hd
ebt
rati
oh
elps
usba
rgai
nfo
rco
nce
ssio
ns
from
our
empl
oye
es0.
000.
160.
160.
150.
180.
130.
130.
19*
0.14
0.17
0.13
0.19
*0.
180.
150.
170.
18
212 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
%im
port
ant
CE
Oag
eC
EO
tenu
reC
EO
MB
AR
egu
late
dT
arge
tde
bt
rati
oP
ubl
icco
rp.
For
eign
sale
sF
ortu
ne
500
mai
ling
or
very
impo
rtan
tM
ean
'59
Yng
erL
ong
Shor
tY
esN
oY
esN
oN
oY
esY
esN
oY
esN
oN
oY
es
(g)
Fin
anci
al#
exib
ility
(we
rest
rict
deb
tso
we
hav
een
oug
hin
tern
alfu
nds
avai
lab
leto
pur
sue
new
proj
ects
wh
enth
eyco
me
alon
g)
59.3
82.
592.
542.
592.
682.
522.
512.
642.
762.
572.
632.
542.
682.
40**
2.91
2.45
***
2.60
2.55
(d)
Our
cred
itra
tin
g(a
sas
sign
edb
yra
ting
agen
cies
)57
.10
2.46
2.52
2.44
2.28
2.56
**2.
372.
503.
592.
32**
*2.
192.
73**
*2.
861.
68**
*2.
772.
30**
*2.
263.
31**
*
(h)
Th
evo
lati
lity
of
our
earn
ings
and
cash#
ows
48.0
82.
322.
382.
332.
402.
292.
222.
40*
2.27
2.31
2.34
2.26
2.34
2.31
2.43
2.27
2.32
2.30
(a)
The
tax
adva
ntag
eo
fin
tere
stde
duc
tib
ility
44.8
52.
072.
152.
051.
922.
14*
2.11
2.07
2.64
1.98
**2.
032.
132.
241.
76**
*2.
451.
91**
*1.
972.
53**
*
(e)
The
tran
sact
ion
sco
sts
and
fees
for
issu
ing
deb
t33
.52
1.95
1.95
1.98
2.22
1.83
***
2.03
1.97
1.71
1.95
2.02
1.89
1.92
2.03
1.98
1.94
2.00
1.70
**
(c)
The
deb
tle
vels
ofo
ther"
rms
ino
urin
dus
try
23.4
01.
491.
431.
521.
461.
531.
611.
452.
321.
40**
*1.
371.
60**
1.63
1.27
***
1.41
1.51
1.41
1.86
***
(b)
Th
epo
tent
ial
cost
sof
ban
kru
ptcy
,n
ear-
ban
kru
ptcy
,or"
nan
cial
dis
tres
s21
.35
1.24
1.12
1.29
1.37
1.20
1.24
1.25
1.38
1.25
1.32
1.19
1.15
1.42
**1.
291.
221.
271.
08
(i)W
elim
itd
ebt
soo
urcu
stom
ers/
supp
liers
are
not
wor
ried
abo
uto
ur"
rmgo
ing
out
ofb
usin
ess
18.7
21.
241.
321.
231.
391.
17**
1.23
1.25
1.33
1.23
1.27
1.24
1.27
1.16
1.20
1.26
1.30
0.98
**
(n)
We
rest
rict
our
bor
row
ing
soth
atp
ro"
tsfr
om
new
/fu
ture
proj
ects
can
be
capt
ure
dfu
llyb
ysh
areh
old
ers
and
do
not
hav
eto
be
paid
out
asin
tere
stto
deb
tho
lder
s
12.5
71.
010.
991.
001.
050.
971.
040.
980.
861.
021.
030.
990.
951.
101.
011.
001.
120.
48**
*
(j)W
etr
yto
have
enou
ghd
ebt
that
we
are
not
anat
trac
tive
take
ove
rta
rget
4.75
0.73
0.82
0.70
0.78
0.70
0.76
0.73
0.71
0.71
0.71
0.77
0.94
0.34
***
0.93
0.64
***
0.70
0.88
*
(f)T
hep
erso
nal
tax
cost
our
inve
sto
rsfa
cew
hen
they
rece
ive
inte
rest
inco
me
4.79
0.68
0.56
0.68
0.67
0.63
0.65
0.65
0.67
0.62
0.73
0.58
*0.
650.
640.
780.
61*
0.67
0.72
(k)
Ifw
eis
sue
deb
tou
rco
mp
etit
ors
kno
wth
atw
ear
eve
ryu
nlik
ely
tore
duce
our
out
put
2.25
0.40
0.45
0.39
0.48
0.34
**0.
370.
420.
380.
380.
440.
360.
430.
350.
420.
390.
400.
36
(m)
To
ensu
reth
atu
pper
man
agem
ent
wor
ksha
rdan
de$
cien
tly,
we
issu
esu$
cien
td
ebt
tom
ake
sure
that
ala
rge
port
ion
of
our
cash#
owis
com
mit
ted
toin
tere
stp
aym
ents
1.69
0.33
0.38
0.32
0.42
0.28
**0.
300.
360.
140.
34*
0.34
0.34
0.31
0.36
0.27
0.35
0.37
0.17
**
(l)A
hig
hd
ebt
rati
oh
elps
usba
rgai
nfo
rco
nce
ssio
ns
from
our
empl
oye
es0.
000.
160.
140.
160.
160.
150.
160.
160.
140.
160.
160.
180.
170.
150.
160.
160.
170.
14
�Res
pon
den
tsar
eas
ked
tora
teo
na
scal
eof
0(n
otim
port
ant)
to4
(ver
yim
por
tant
).W
ere
por
tth
eo
vera
llm
ean
asw
ell
asth
e%
of
resp
ond
ents
that
answ
ered
3an
d4
(ver
yim
port
ant)
.***
,**
,*d
eno
tes
asi
gni"
can
tdi!
eren
ceat
the
1%,
5%,a
nd
10%
leve
l,re
spec
tive
ly.
All
tab
leco
lum
ns
are
de"
ned
inT
able
1.
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 213
Tab
le7
Surv
eyre
spon
ses
toth
equ
esti
on:
has
you
r"
rmse
riou
sly
cons
ider
edis
suin
gd
ebt
info
reig
nco
untr
ies?
If`y
es:,
wh
atfa
ctor
sa!
ect
you
r"
rm's
deci
sion
sab
out
issu
ing
fore
ign
deb
t?�
%im
port
ant
Size
P/E
Lev
erag
eIn
vest
men
tgr
ade
Pay
div
iden
dsIn
dus
try
Man
agem
ent
ow
ner
ship
or
very
impo
rtan
tM
ean
Smal
lL
arge
Gro
wth
No
n-G
Lo
wH
igh
Yes
No
Yes
No
Man
u.
Oth
ers
Lo
wH
igh
(c)
Pro
vidi
ng
a`n
atur
alhe
dgea
(e.g
.,if
the
fore
ign
curr
ency
deva
lues
,w
ear
eno
tob
ligat
edto
pay
inte
rest
inU
S$)
85.8
43.
153.
063.
222.
983.
293.
203.
323.
063.
233.
123.
363.
322.
94*
3.00
3.28
(b)
Kee
pin
gth
e`s
our
ceof
fund
sacl
ose
toth
e`u
seo
ffu
ndsa
63.3
92.
673.
092.
52**
2.73
2.35
*2.
702.
792.
382.
702.
573.
12**
2.92
2.23
***
2.55
2.74
(a)
Fav
orab
leta
xtr
eatm
ent
rela
tive
toth
eU
.S(e
.g.,
di!
eren
tco
rpor
ate
tax
rate
s)52
.25
2.26
1.94
2.41
**2.
272.
292.
262.
392.
372.
402.
292.
082.
362.
132.
162.
33
(e)
For
eign
inte
rest
rate
sm
ayb
elo
wer
than
dom
esti
cin
tere
stra
tes
44.2
52.
192.
332.
112.
272.
032.
222.
132.
202.
482.
082.
402.
222.
102.
042.
54**
(d)
Fo
reig
nre
gula
tion
sre
quir
eu
sto
issu
ed
ebt
abro
ad5.
500.
630.
600.
640.
750.
29**
0.55
0.72
0.65
0.57
0.63
0.73
0.64
0.66
0.59
0.61
%im
port
ant
CE
Oag
eC
EO
tenu
reC
EO
MB
AR
egu
late
dT
arge
tde
bt
rati
oP
ubl
icco
rp.
For
eign
sale
sF
ortu
ne
500
mai
lo
rve
ryim
port
ant
Mea
n'
59Y
nger
Lo
ngSh
ort
Yes
No
Yes
No
No
Yes
Yes
No
Yes
No
No
Yes
(c)
Pro
vidi
ng
a`n
atur
alhe
dgea
(e.g
.,if
the
fore
ign
curr
ency
deva
lues
,w
ear
eno
tob
ligat
edto
pay
inte
rest
inU
S$)
85.8
43.
153.
303.
133.
393.
133.
333.
063.
333.
143.
303.
173.
212.
953.
342.
92**
3.22
3.00
(b)
Kee
pin
gth
e`s
our
ceof
fund
sacl
ose
toth
e`u
seo
ffu
ndsa
63.3
92.
672.
572.
712.
742.
672.
772.
663.
332.
66*
2.78
2.64
2.65
2.95
2.72
2.65
2.85
2.30
**
(a)
Fav
orab
leta
xtr
eatm
ent
rela
tive
toth
eU
.S(e
.g.,
di!
eren
tco
rpor
ate
tax
rate
s)52
.25
2.26
2.13
2.30
2.00
2.39
*2.
422.
04*
2.11
2.22
2.44
2.12
2.37
1.67
**2.
501.
94**
2.34
2.11
(e)
For
eign
inte
rest
rate
sm
ayb
elo
wer
than
dom
esti
cin
tere
stra
tes
44.2
52.
192.
302.
162.
262.
172.
222.
141.
672.
142.
401.
93**
2.18
2.26
2.25
2.08
2.28
2.03
(d)
Fo
reig
nre
gula
tion
sre
quir
eu
sto
issu
ed
ebt
abro
ad5.
500.
630.
770.
570.
500.
690.
600.
581.
110.
57*
0.57
0.64
0.61
0.56
0.59
0.64
0.64
0.62
�Res
pon
den
tsar
eas
ked
tora
teo
na
scal
eo
f0(n
oti
mpo
rtan
t)to
4(v
ery
imp
orta
nt).
We
rep
ortt
he
ove
rall
mea
nas
wel
las
the
%of
resp
ond
ents
that
answ
ered
3an
d4
(ver
yim
por
tant
).**
*,**
,*de
not
esa
sign
i"ca
nt
di!
eren
ceat
the
1%,
5%,a
nd
10%
leve
l,re
spec
tive
ly.
All
tab
leco
lum
ns
are
de"
ned
inT
able
1.
214 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
evidence supports this hypothesis: the mean response of 1.08 indicates that "rmsdo not rebalance in response to market equity movements (Row g in Table 9).Further, among "rms targeting their debt ratio, few "rms (rating of 0.99) statethat changes in the price of equity a!ect their debt policy. Similarly, in theirlarge-sample study of Compustat "rms, Opler and Titman (1998) "nd that "rmsissue equity after stock price increases, which they note is inconsistent with "rmstargeting debt ratios because it moves them further from any such target.
Fisher et al. (1989) propose an alternative explanation of why debt ratios varyover time, even if "rms have a target. If there are "xed transactions costs toissuing or retiring debt, a "rm only rebalances when its debt ratio crosses anupper or lower hurdle. We "nd moderate evidence that "rms consider transac-tions costs when making debt issuance decisions (rating of 1.95 in Row e ofTable 6), especially among small "rms (2.07) in which the CEO has been in o$cefor at least ten years (2.22). Many papers (e.g., Titman and Wessels, 1988)interpret the "nding that small "rms use relatively little debt as evidence thattransaction costs discourage debt usage among small "rms; as far as we know,our analysis is the most direct examination of this hypothesis to date. However,when we ask whether they delay issuing debt (rating of 1.06 in Table 9) orretiring debt (1.04) because of transactions costs, which is a more direct test ofthe Fisher et al. (1989) hypothesis, the support for the transactions cost hypothe-sis is weak.
5.2. Asymmetric information explanations of capital structure
5.2.1. Pecking-order model of xnancing hierarchyThe pecking-order model of "nancing choice assumes that "rms do not target
a speci"c debt ratio, but instead use external "nancing only when internal fundsare insu$cient. External funds are less desirable because informational asym-metries between management and investors imply that external funds areundervalued in relation to the degree of asymmetry (Myers and Majluf, 1984;Myers, 1984). Therefore, if "rms use external funds, they prefer to use debt,convertible securities, and, as a last resort, equity.
Myers and Majluf (1984) assume that "rms seek to maintain "nancial slack toavoid the need for external funds. Therefore, if we "nd that "rms value "nancial#exibility, this is generally consistent with the pecking-order theory. However,#exibility is also important for reasons unrelated to the pecking-order model(e.g., Opler et al., 1999), so "nding that CFOs value "nancial #exibility is notsu$cient to prove that the pecking-order model is the true description of capitalstructure choice.
We ask several questions related to the pecking-order model. We ask if "rmsissue securities when internal funds are not su$cient to fund their activities, andseparately ask if equity is used when debt, convertibles, or other sources of"nancing are not available. We also inquire whether executives consider equity
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 215
Tab
le8
Surv
eyre
spon
ses
toth
equ
esti
on:
has
you
r"
rmse
riou
sly
cons
ider
edis
suin
gco
mm
onst
ock
?If`y
esa,
wha
tfa
cto
rsa!
ect
your"
rm's
dec
isio
nsab
out
issu
ing
com
mon
sto
ck?�
%im
port
ant
Size
P/E
Lev
erag
eIn
vest
men
tgr
ade
Pay
div
iden
dsIn
dus
try
Man
agem
ent
ow
ner
ship
or
very
impo
rtan
tM
ean
Smal
lL
arge
Gro
wth
No
n-G
Lo
wH
igh
Yes
No
Yes
No
Man
u.
Oth
ers
Lo
wH
igh
(m)
Ear
nin
gs-p
er-s
hare
dilu
tion
68.5
52.
842.
653.
12**
3.17
3.03
2.81
2.93
3.00
3.18
3.06
2.63
**3.
032.
60**
3.07
2.63
**(k
)T
he
amo
unt
byw
hich
our
stoc
kis
und
erva
lued
or
ove
rval
ued
by
the
mar
ket
66.9
42.
692.
672.
712.
942.
652.
502.
93**
2.58
3.08
**2.
702.
662.
762.
502.
932.
47**
(a)
Ifo
urst
ock
pri
ceh
asre
cent
lyri
sen
,th
epr
ice
atw
hic
hw
eca
nse
llis`h
igha
62.6
02.
532.
572.
472.
572.
612.
452.
672.
422.
92*
2.35
2.69
*2.
792.
26**
2.62
2.45
(c)
Pro
vidi
ng
shar
esto
empl
oye
ebo
nus/
stoc
kop
tion
plan
s53
.28
2.34
2.22
2.50
2.20
2.38
2.66
2.00
***
2.77
1.97
**2.
462.
172.
162.
472.
342.
30
(e)
Mai
ntai
nin
ga
targ
etde
bt-
to-e
qui
tyra
tio
51.5
92.
262.
042.
58**
2.56
2.03
**1.
862.
68**
*2.
442.
582.
681.
85**
*2.
481.
91**
2.64
1.84
***
(j)D
iluti
ngth
eho
ldin
gso
fce
rtai
nsh
areh
old
ers
50.4
12.
142.
301.
90*
1.94
2.23
2.20
2.09
1.46
2.24
**1.
972.
311.
952.
202.
002.
38*
(b)
Sto
ckis
our`l
east
risk
yaso
urce
offu
nds
30.5
81.
761.
931.
52*
2.07
1.37
***
1.80
1.71
1.44
1.68
1.56
1.97
*1.
761.
691.
621.
91(g
)W
heth
ero
urre
cen
tpr
o"ts
hav
ebe
ensu$
cien
tto
fund
our
acti
viti
es30
.40
1.76
1.91
1.54
*1.
931.
39**
1.71
1.79
1.52
1.82
1.67
1.76
1.84
1.69
1.60
1.88
(f)U
sing
asi
mila
ram
oun
tof
equi
tyas
isu
sed
byot
her"
rms
inou
rin
dust
ry22
.95
1.45
1.33
1.63
*1.
701.
00**
*1.
351.
571.
561.
431.
741.
09**
*1.
361.
381.
591.
32
(h)
Issu
ing
sto
ckgi
ves
inve
stor
sa
bet
ter
impr
essi
ono
fou
r"
rm's
pros
pect
sth
anis
suin
gde
bt
21.4
91.
311.
521.
00**
1.48
0.89
***
1.22
1.37
0.92
1.43
**1.
101.
46*
1.14
1.50
*1.
181.
51*
(l)In
abili
tyto
obta
infu
nds
usi
ng
deb
t,co
nver
tibl
es,
orot
her
sou
rces
15.5
71.
151.
360.
84**
1.00
0.79
1.09
1.20
0.68
1.45
***
1.03
1.19
1.03
1.22
1.16
1.21
(d)
Com
mo
nst
ock
iso
urch
eap
est
sou
rce
of
fun
ds14
.05
1.10
1.35
0.73
***
1.02
0.97
1.26
0.96
0.68
0.68
0.93
1.28
*0.
981.
170.
861.
36**
(i)T
heca
pit
alga
ins
tax
rate
sfa
ced
by
our
inve
stor
s(r
elat
ive
tota
xra
tes
on
div
iden
ds)
5.00
0.82
0.78
0.88
0.88
0.79
0.98
0.63
**0.
800.
920.
800.
770.
750.
920.
810.
88
216 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
%im
port
ant
CE
Oag
eC
EO
tenu
reC
EO
MB
AR
egu
late
dT
arge
tde
bt
rati
o
Pu
blic
corp
.F
orei
gnsa
les
For
tun
e50
0
mai
ling
or
very
impo
rtan
tM
ean
'59
Yng
erL
ong
Shor
tY
esN
oY
esN
oN
oY
esY
esN
oY
esN
oN
oY
es
(m)
Ear
nin
gs-p
er-s
hare
dilu
tion
68.5
52.
843.
042.
812.
643.
00*
2.62
2.95
*3.
642.
72**
*2.
692.
973.
181.
48**
*2.
892.
802.
733.
29**
(k)
Th
eam
oun
tby
whi
chou
rst
ock
isu
nder
valu
ed
or
ove
rval
ued
by
the
mar
ket
66.9
42.
692.
522.
742.
862.
602.
732.
672.
432.
692.
692.
662.
901.
78**
*2.
962.
58*
2.74
2.43
(a)
Ifo
urst
ock
pri
ceh
asre
cent
lyri
sen
,th
epr
ice
atw
hic
hw
eca
nse
llis`h
igha
62.6
02.
532.
542.
552.
512.
562.
452.
562.
642.
502.
472.
572.
701.
83**
*2.
362.
592.
462.
79
(c)
Pro
vidi
ng
shar
esto
empl
oye
ebo
nus/
stoc
k
opti
onpl
ans
53.2
82.
342.
652.
23*
2.44
2.29
2.13
2.42
2.15
2.31
2.28
2.38
2.24
2.72
**2.
502.
292.
242.
74**
(e)
Mai
ntai
nin
ga
targ
etde
bt-
to-e
qui
tyra
tio
51.5
92.
261.
722.
41**
2.12
2.38
1.79
2.46
***
3.14
2.11
***
1.71
2.68
***
2.40
1.73
**2.
212.
242.
242.
38
(j)D
iluti
ngth
eho
ldin
gso
fce
rtai
nsh
areh
old
ers
50.4
12.
142.
322.
132.
272.
142.
162.
192.
002.
162.
242.
022.
251.
68**
1.93
2.20
2.25
1.65
**
(b)
Sto
ckis
our`l
east
risk
yaso
urce
offu
nds
30.5
81.
761.
711.
741.
721.
731.
531.
831.
691.
751.
791.
731.
791.
621.
821.
751.
901.
17**
(g)
Whe
ther
our
rece
nt
pro"
tsh
ave
been
su$
cien
t
tofu
ndou
rac
tivi
ties
30.4
01.
761.
361.
86**
1.84
1.73
1.42
1.91
**1.
691.
701.
751.
771.
731.
801.
551.
801.
881.
22**
(f)U
sing
asi
mila
ram
oun
tof
equi
tyas
isu
sed
by
oth
er"
rms
inou
rin
dust
ry
22.9
51.
451.
121.
52*
1.41
1.47
1.13
1.58
**2.
151.
30**
1.46
1.37
1.43
1.54
1.11
1.54
*1.
481.
30
(h)
Issu
ing
sto
ckgi
ves
inve
stor
sa
bet
ter
impr
essi
on
of
our"
rm's
pros
pect
sth
anis
suin
gde
bt
21.4
91.
310.
921.
39**
1.32
1.30
1.11
1.41
1.23
1.28
1.24
1.36
1.29
1.33
1.21
1.35
1.41
0.91
**
(l)In
abili
tyto
obta
infu
nds
usi
ng
deb
t,co
nver
tibl
es,
orot
her
sou
rces
15.5
71.
150.
791.
26*
1.32
1.10
0.76
1.35
***
1.38
1.09
1.22
1.10
1.06
1.42
0.72
1.29
**1.
200.
91
(d)
Com
mo
nst
ock
iso
urch
eap
est
sou
rce
of
fun
ds14
.05
1.10
0.88
1.12
1.00
1.12
1.16
1.05
0.69
1.15
1.32
0.92
**1.
011.
46*
1.11
1.11
1.23
0.52
***
(i)T
heca
pit
alga
ins
tax
rate
sfa
ced
by
our
inve
stor
s(r
elat
ive
tota
xra
tes
on
div
iden
ds)
5.00
0.82
0.79
0.80
0.95
0.72
0.57
0.92
**0.
380.
81*
0.84
0.76
0.84
0.71
0.93
0.78
0.81
0.83
�Res
pon
den
tsar
eas
ked
tora
teo
na
scal
eo
f0(n
oti
mpo
rtan
t)to
4(v
ery
imp
orta
nt).
We
rep
ortt
he
ove
rall
mea
nas
wel
las
the
%of
resp
ond
ents
that
answ
ered
3an
d4
(ver
yim
por
tant
).**
*,**
,*de
not
esa
sign
i"ca
nt
di!
eren
ceat
the
1%,
5%,a
nd
10%
leve
l,re
spec
tive
ly.
All
tab
leco
lum
ns
are
de"
ned
inT
able
1.
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 217
Fig. 6. Survey evidence on whether "rms have optimal or target debt}equity ratios. The survey isbased on the responses of 392 CFOs.
� Pinegar and Wilbricht (1989) survey 176 unregulated, non"nancial Fortune 500 "rms. Like us,they "nd that #exibility is the most important factor a!ecting "nancing decisions, and thatbankruptcy costs and personal tax considerations are among the least important. Our analysis,examining a broader cross-section of theoretical hypotheses and using information on "rm andexecutive characteristics, shows that the relative importance of these factors is robust to a moregeneral survey design.
undervaluation when deciding which security to use, and whether "nancial#exibility is important.
The most important item a!ecting corporate debt decisions is management'sdesire for `"nancial #exibility,a with a mean rating of 2.59 (Table 6).� Four "rmswrite in explicitly that they remain #exible in the sense of minimizing interestobligations, so that they do not need to shrink their business in case of aneconomic downturn. In untabulated analysis, we "nd that "rms that value"nancial #exibility are more likely to value real options in project evaluation,but the di!erence is not signi"cant. Fifty-nine percent of the respondents saythat #exibility is important (rating of 3) or very important (rating of 4). This"nding is interesting because Graham (2000) shows that "rms use their "nancial#exibility (i.e., preserve debt capacity) to make future expansions and acquisi-tions, but they appear to retain a lot of unused #exibility even after expanding.However, the importance of #exibility in the survey responses is not related toinformational asymmetry (size or dividend payout) or growth options in themanner suggested by the pecking-order theory. In fact, #exibility is statistically
218 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
more important for dividend-paying "rms, opposite the theoretical prediction (ifdividend-paying "rms have relatively little informational asymmetry). There-fore, a deeper investigation indicates that the desire for "nancial #exibility is notdriven by the factors behind the pecking-order theory.
Having insu$cient internal funds is a moderately important in#uence on thedecision to issue debt (rating of 2.13, Row a in Table 9). This behavior isgenerally consistent with the pecking-order model. More small "rms (rating of2.30) than large "rms (1.88) indicate that they use debt in the face of insu$cientinternal funds, which is consistent with the pecking-order if small "rms su!erfrom larger asymmetric-information-related equity undervaluation. However,there is only modest evidence that "rms issue equity because recent pro"ts havebeen insu$cient to fund activities (1.76 in Table 8), and even less indicating that"rms issue equity after their ability to obtain funds from debt or convertibles isdiminished (rating of 1.15 in Table 10).
Firms are reluctant to issue common stock when they perceive that it isundervalued (rating of 2.69, the most important equity issuance factor inTable 8). In a separate survey conducted one month after ours, when the DowJones 30 was approaching a new record of 10,000, Graham (1999b) "nds thatmore than two-thirds of FEI executives feel that their common equity isundervalued by the market and that only 3% of CFOs think their stock isovervalued, suggesting that the preference for pecking-order-like behaviormight be driven by managerial optimism (Heaton, 2000). Taken together, these"ndings indicate that a large percentage of companies are hesitant to issuecommon equity because they feel their stock is undervalued. Many "rms issueconvertible debt instead: equity undervaluation is the second most popularfactor a!ecting convertible debt policy (rating of 2.34 in Table 10), a responseparticularly popular among growth "rms (2.72).
Finding that "rms avoid equity when they perceive that it is undervalued isgenerally consistent with the pecking order. However, when we examine morecarefully how equity undervaluation a!ects "nancing decisions, the support forthe pecking-order model wanes. In debt decisions, large (rating of 1.76 in Rowd of Table 9), dividend-paying (1.65) "rms are relatively more likely to say thatequity undervaluation a!ects their debt policy (versus ratings of 1.37 for bothsmall and nondividend-paying "rms). In equity decisions, the relative import-ance of stock valuation on equity issuance is not related to informationalasymmetry as indicated by small size and nondividend-paying status, though itis more important for "rms with low executive ownership. In general, these"ndings are not consistent with the pecking-order idea that informationallyinduced equity undervaluation causes "rms to avoid equity "nancing. Helwegeand Liang (1996, p. 457) also "nd that `asymmetric information variables haveno power to predict the relative use of public bonds over equity.a
In sum, the importance of "nancial #exibility and equity undervaluation tosecurity issuance decisions is generally consistent with the pecking-order model
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 219
Tab
le9
Surv
eyre
spon
ses
toth
equ
esti
on:
wh
atot
her
fact
ors
a!ec
tyo
ur"
rm's
deb
tp
olic
y?�
%im
port
ant
Size
P/E
Lev
erag
eIn
vest
men
tgr
ade
Pay
div
iden
dsIn
dus
try
Man
agem
ent
ow
ner
ship
or
very
impo
rtan
tM
ean
Smal
lL
arge
Gro
wth
No
n-G
Lo
wH
igh
Yes
No
Yes
No
Man
u.
Oth
ers
Lo
wH
igh
(c)
We
issu
ed
ebt
whe
nin
tere
stra
tes
are
part
icul
arly
low
46.3
52.
222.
072.
40**
2.35
2.42
2.13
2.29
2.40
2.43
2.37
1.98
***
2.25
2.16
2.39
2.02
***
(a)
We
issu
ed
ebt
whe
nou
rre
cent
pro"
ts(in
tern
alfu
nds
)ar
en
otsu$
cien
tto
fun
do
urac
tivi
ties
46.7
82.
132.
301.
88**
*2.
091.
862.
102.
121.
812.
28**
2.09
2.16
2.24
1.94
**2.
142.
13
(d)
We
use
deb
tw
hen
our
equi
tyis
unde
rval
ued
by
the
mar
ket
30.7
91.
561.
371.
76**
*2.
141.
851.
521.
721.
562.
17**
*1.
651.
37*
1.67
1.47
1.83
1.49
**
(g)
Ch
ange
sin
the
pric
eof
our
com
mo
nst
ock
16.3
81.
080.
911.
25**
*1.
451.
380.
961.
27**
1.05
1.52
***
1.14
0.95
1.14
1.01
1.25
1.07
(e)
We
dela
yis
suin
gd
ebt
bec
ause
oftr
ansa
ctio
ns
cost
san
dfe
es10
.17
1.06
1.25
0.83
***
1.06
0.87
1.09
1.00
0.90
0.92
0.97
1.20
**1.
061.
070.
921.
22**
(f)W
ede
lay
reti
rin
gde
bt
beca
use
of
reca
pit
aliz
atio
nco
sts
and
fees
12.4
31.
041.
041.
051.
161.
040.
911.
18**
1.10
1.30
1.13
0.93
*1.
190.
86**
*1.
051.
02
(b)
Usi
ng
deb
tgi
ves
inve
sto
rsa
bett
erim
pre
ssio
no
fou
r"
rm's
pros
pect
sth
anis
suin
gco
mm
on
stoc
k9.
830.
960.
851.
05*
1.19
1.14
0.91
1.09
1.00
1.39
**1.
000.
841.
010.
871.
070.
95
(h)W
eis
sue
deb
tw
hen
we
hav
eac
cum
ula
ted
sub
stan
tial
pro"
ts1.
140.
530.
500.
550.
610.
550.
460.
540.
570.
600.
550.
500.
580.
450.
610.
52
%im
port
ant
CE
Oag
eC
EO
tenu
reC
EO
MB
AR
egu
late
dT
arge
tde
bt
rati
oP
ubl
icco
rp.
For
eign
sale
sF
ortu
ne
500
mai
ling
or
very
impo
rtan
tM
ean
'59
Yng
erL
ong
Shor
tY
esN
oY
esN
oN
oY
esY
esN
oY
esN
oN
oY
es
(c)
We
issu
ed
ebt
whe
nin
tere
stra
tes
are
part
icul
arly
low
46.3
52.
222.
132.
262.
242.
212.
362.
152.
192.
202.
302.
122.
391.
90**
*2.
382.
152.
192.
35
(a)
We
issu
ed
ebt
whe
nou
rre
cent
pro"
ts(in
tern
alfu
nds
)ar
en
otsu$
cien
tto
fun
do
urac
tivi
ties
46.7
82.
132.
242.
092.
352.
00**
2.09
2.18
2.00
2.14
2.21
2.00
2.01
2.33
**1.
932.
182.
211.
75**
(d)
We
use
deb
tw
hen
our
equi
tyis
unde
rval
ued
by
the
mar
ket
30.7
91.
561.
511.
571.
441.
601.
501.
581.
861.
501.
631.
462.
100.
54**
*1.
891.
41**
*1.
541.
67
(g)
Ch
ange
sin
the
pric
eof
our
com
mo
nst
ock
16.3
81.
080.
951.
111.
051.
061.
041.
081.
101.
041.
160.
991.
480.
31**
*1.
151.
021.
081.
10(e
)W
ede
lay
issu
ing
deb
tb
ecau
seof
tran
sact
ion
sco
sts
and
fees
10.1
71.
060.
971.
091.
270.
95**
*1.
131.
060.
761.
101.
130.
991.
031.
151.
111.
051.
170.
57**
*
(f)W
ede
lay
reti
rin
gde
bt
beca
use
of
reca
pit
aliz
atio
nco
sts
and
fees
12.4
31.
041.
081.
011.
200.
93**
1.10
0.98
1.05
1.06
1.07
0.99
1.14
0.87
**1.
220.
971.
070.
89
(b)
Usi
ng
deb
tgi
ves
inve
sto
rsa
bett
erim
pre
ssio
no
fou
r"
rm's
pros
pect
sth
anis
suin
gco
mm
on
stoc
k9.
830.
961.
100.
900.
940.
950.
791.
04**
1.10
0.91
1.01
0.91
1.18
0.51
***
1.00
0.92
0.92
1.14
(h)W
eis
sue
deb
tw
hen
we
hav
eac
cum
ula
ted
sub
stan
tial
pro"
ts1.
140.
530.
510.
530.
610.
46*
0.45
0.58
0.71
0.52
0.56
0.50
0.56
0.47
0.57
0.51
0.52
0.55
�Res
pon
den
tsar
eas
ked
tora
teo
na
scal
eo
f0(n
oti
mpo
rtan
t)to
4(v
ery
imp
orta
nt).
We
rep
ortt
he
ove
rall
mea
nas
wel
las
the
%of
resp
ond
ents
that
answ
ered
3an
d4
(ver
yim
por
tant
).**
*,**
,*de
not
esa
sign
i"ca
nt
di!
eren
ceat
the
1%,
5%,a
nd
10%
leve
l,re
spec
tive
ly.
All
tab
leco
lum
ns
are
de"
ned
inT
able
1.
220 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Tab
le10
Surv
eyre
spon
ses
toth
equ
esti
on:
has
you
r"
rmse
riou
sly
cons
ider
edis
suin
gco
nver
tibl
ed
ebt?
If`y
esa,
wha
tfa
cto
rsa!
ect
your"
rm's
dec
isio
nsab
out
issu
ing
conv
erti
ble
deb
t?�
%im
port
ant
Size
P/E
Lev
erag
eIn
vest
men
tgr
ade
Pay
div
iden
dsIn
dus
try
Man
agem
ent
ow
ner
ship
or
very
impo
rtan
tM
ean
Smal
lL
arge
Gro
wth
No
n-G
Lo
wH
igh
Yes
No
Yes
No
Man
u.
Oth
ers
Lo
wH
igh
(a)
Co
nver
tibl
esar
ean
inex
pen
sive
way
tois
sue
`del
ayeda
com
mo
nst
ock
58.1
12.
492.
542.
432.
672.
502.
382.
602.
732.
422.
592.
432.
402.
572.
422.
52
(f)O
ur
stoc
kis
curr
entl
yu
nder
valu
ed50
.68
2.34
2.26
2.44
2.72
2.19
*2.
212.
522.
402.
642.
252.
462.
412.
432.
282.
42(g
)A
bili
tyto`c
alla
or
forc
eco
nver
sio
no
fco
nve
rtib
lede
bt
if/w
hen
we
need
to47
.95
2.29
2.28
2.29
2.58
2.56
2.32
2.20
2.21
2.65
2.42
2.17
2.26
2.33
2.08
2.52
*
(e)
Avo
idin
gsh
ort-
term
equ
ity
dilu
tion
45.8
32.
182.
032.
352.
452.
192.
152.
282.
472.
382.
441.
97*
2.23
2.14
2.05
2.33
(h)
To
attr
act
inve
sto
rsun
sure
abou
tth
eri
skin
ess
of
our
com
pan
y43
.84
2.07
2.35
1.73
**1.
881.
882.
022.
101.
361.
88*
1.83
2.31
*2.
002.
131.
822.
47**
(c)
Co
nver
tibl
esar
ele
ssex
pen
sive
than
stra
ight
deb
t41
.67
1.85
2.08
1.58
*1.
562.
31**
1.80
1.83
1.43
1.80
1.57
2.14
**1.
582.
10*
1.71
2.00
(d)
Oth
er"
rms
ino
urin
dus
try
succ
essf
ully
use
conv
erti
bles
12.5
01.
101.
121.
061.
220.
69*
1.29
0.83
**0.
931.
250.
861.
21*
0.92
1.30
*1.
051.
06
(b)
Pro
tect
ing
bond
hold
ers
agai
nst
unf
avo
rab
leac
tio
nsby
man
ager
so
rst
ock
hold
ers
1.41
0.62
0.61
0.64
0.72
0.31
**0.
570.
660.
430.
640.
540.
710.
580.
720.
610.
67
%im
port
ant
CE
Oag
eC
EO
tenu
reC
EO
MB
AR
egu
late
dT
arge
tde
bt
rati
oP
ubl
icco
rp.
For
eign
sale
sF
ortu
ne
500
mai
ling
or
very
impo
rtan
tM
ean
'59
Yng
erL
ong
Shor
tY
esN
oY
esN
oN
oY
esY
esN
oY
esN
oN
oY
es
(a)
Co
nver
tibl
esar
ean
inex
pen
sive
way
tois
sue
`del
ayeda
com
mo
nst
ock
58.1
12.
492.
792.
462.
742.
422.
612.
472.
782.
512.
362.
682.
542.
272.
522.
412.
512.
41
(f)O
ur
stoc
kis
curr
entl
yu
nder
valu
ed50
.68
2.34
2.00
2.45
2.28
2.42
1.87
2.57
**2.
782.
272.
302.
322.
451.
93*
2.48
2.25
2.30
2.47
(g)
Ab
ility
to`c
alla
or
forc
eco
nver
sio
no
fco
nve
rtib
lede
bt
if/w
hen
we
need
to47
.95
2.29
2.64
2.21
2.42
2.22
1.91
2.39
*2.
252.
282.
232.
372.
292.
272.
482.
202.
282.
31
(e)
Avo
idin
gsh
ort-
term
equ
ity
dilu
tion
45.8
32.
182.
002.
252.
282.
162.
002.
243.
112.
10**
2.05
2.37
2.21
2.07
2.24
2.12
2.05
2.59
*(h
)T
oat
trac
tin
vest
ors
unsu
reab
out
the
risk
ines
so
fou
rco
mp
any
43.8
42.
072.
292.
002.
002.
081.
572.
33**
*1.
882.
122.
321.
63**
1.77
3.07
***
2.00
2.10
2.16
1.75
(c)
Co
nver
tibl
esar
ele
ssex
pen
sive
than
stra
ight
deb
t41
.67
1.85
2.50
1.70
**1.
941.
762.
041.
781.
381.
932.
071.
44**
1.81
2.00
1.81
1.86
2.02
1.25
**(d
)O
ther"
rms
ino
urin
dus
try
succ
essf
ully
use
conv
erti
bles
12.5
01.
101.
001.
110.
721.
25**
0.57
1.33
***
1.50
0.95
*1.
330.
78**
1.09
1.00
1.33
1.00
1.18
0.80
(b)
Pro
tect
ing
bond
hold
ers
agai
nst
unf
avo
rab
leac
tio
nsby
man
ager
so
rst
ock
hold
ers
1.41
0.62
1.08
0.53
***
0.61
0.66
0.48
0.73
*0.
620.
590.
600.
670.
610.
670.
620.
620.
640.
56
�Res
pon
den
tsar
eas
ked
tora
teo
na
scal
eo
f0(n
oti
mpo
rtan
t)to
4(v
ery
imp
orta
nt).
We
rep
ortt
he
ove
rall
mea
nas
wel
las
the
%of
resp
ond
ents
that
answ
ered
3an
d4
(ver
yim
por
tant
).**
*,**
,*de
not
esa
sign
i"ca
nt
di!
eren
ceat
the
1%,
5%,a
nd
10%
leve
l,re
spec
tive
ly.
All
tab
leco
lum
ns
are
de"
ned
inT
able
1.
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 221
of "nancing hierarchy. However, asymmetric information does not appear tocause the importance of these factors, as it should if the pecking-order model isthe true model of capital structure choice.
5.2.2. Recent increase in price of common stockWe investigate whether "rms issue stock during a `window of opportunitya
that arises because their stock price has recently increased, as argued byLoughran and Ritter (1995). Lucas and McDonald (1990) put an informationalasymmetry spin on the desire to issue equity after stock price increases: Ifa "rm's stock price is undervalued due to informational asymmetry, it delaysissuing until after an informational release (of good news) and the ensuingincrease in stock price.
Recent stock price performance is the third most popular factor a!ectingequity-issuance decisions (rating of 2.53 in Table 8), in support of the `window ofopportunitya. Consistent with Lucas and McDonald (1990), the window ofopportunity is most important for "rms su!ering from informational asymmet-ries (i.e., not paying dividends).
5.2.3. Signaling private information with debt and equityRoss (1977) and Leland and Pyle (1977) argue that "rms use capital structure
to signal their quality or future prospects. However, very few "rms indicate thattheir debt policy is a!ected by factors consistent with signaling (rating of 0.96 inTable 9). In addition to small absolute importance, companies more likely tosu!er from informational asymmetries, such as small, private (0.51) "rms, arerelatively unlikely to use debt to signal future prospects (see Row b in Table 9).We also "nd little evidence that "rms issue equity to give the market a positiveimpression of their prospects (rating of 1.31 in Table 8). Sending a positive signalvia equity issuance is relatively more popular among speculative, nondividend-paying "rms.
5.2.4. Private information and convertible stock issuanceBrennan and Kraus (1987) and Brennan and Schwartz (1988) argue that the
call or conversion feature makes convertible debt relatively insensitive to asym-metric information (between management and investors) about the risk of the"rm. We "nd moderate support for this argument. Firms use convertible debt toattract investors unsure about the riskiness of the company (rating of 2.07 inTable 10). This response is relatively more popular among "rms for whichoutside investors are likely to know less than management about "rm risk, i.e.,small "rms (2.35) with large managerial ownership (2.47).
Stein (1992) argues that if "rms privately know that their stock is under-valued, they prefer to avoid issuing equity. At the same time, they want tominimize the distress costs that come with debt issuance. Convertible debt is`delayeda common stock that has lower distress costs than debt and smaller
222 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
undervaluation than equity. We "nd strong evidence consistent with Stein'sargument that convertibles are `back-door equity.a Among "rms that issueconvertible debt, the most popular factor is that convertibles are an inexpensiveway to issue delayed common stock (rating of 2.49 in Table 10).
5.2.5. Anticipating improvement in credit ratingsHaving private information about credit quality can a!ect a "rm's optimal
debt maturity. In theory, if "rms privately know they are high quality but arecurrently assigned a low credit rating, they issue short-term debt because theyexpect their rating to improve (Flannery, 1986; Kale and Noe, 1990). In practice,the evidence that "rms time their credit worthiness is weak. The mean responseis only 0.85 (Row e, Table 11) that companies borrow short-term because theyexpect their credit rating to improve. This response receives more support fromcompanies with speculative grade debt (1.18) and those that do not pay divi-dends (0.99). Though not of large absolute magnitude, this last answer isconsistent with "rms timing their credit ratings when they are subject to largeinformational asymmetries.
5.2.6. Timing market interest ratesAlthough relatively few executives time changes in their credit ratings (some-
thing about which they might reasonably have private information), we "ndsurprising indications that they try to time the market in other ways. We inquirewhether executives attempt to time interest rates by issuing debt when they feelthat market interest rates are particularly low. The rating of 2.22 in Table6 provides moderately strong evidence that "rms try to time the market in thissense. Market timing is especially important for large "rms (2.40), which impliesthat companies are more likely to time interest rates when they have a large orsophisticated treasury department.
We also "nd evidence that "rms issue short-term debt in an e!ort to timemarket interest rates. CFOs borrow short-term when they feel that short ratesare low relative to long rates (1.89 in Table 11) or when they expect long-termrates to decline (1.78). Finally, we check if "rms use foreign debt because foreigninterest rates are lower than domestic rates. There is moderate evidence thatrelatively low foreign interest rates a!ect the decision to issue abroad (rating of2.19). Though insigni"cant, small (2.33) growth (2.27) "rms are more likely tomake this claim. If covered interest rate parity holds, it is not clear to us why"rms pursue this strategy.
5.3. Agency costs
5.3.1. Conyicts between bondholders and equityholdersMyers (1977) argues that investment decisions can be a!ected by the presence
of long-term debt in a "rm's capital structure. Shareholders might `underinvesta
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 223
Tab
le11
Surv
eyre
spon
ses
toth
equ
esti
on:
wh
atfa
ctor
sa!
ect
you
r"
rm's
cho
ice
betw
een
sho
rt-
and
long
-ter
md
ebt?
�
%im
port
ant
Size
P/E
Lev
erag
eIn
vest
men
tgr
ade
Pay
div
iden
dsIn
dus
try
Man
agem
ent
ow
ner
ship
or
very
impo
rtan
tM
ean
Smal
lL
arge
Gro
wth
No
n-G
Lo
wH
igh
Yes
No
Yes
No
Man
u.
Oth
ers
Lo
wH
igh
(b)
Mat
chin
gth
em
atu
rity
of
our
deb
tw
ith
the
life
of
our
asse
ts63
.25
2.60
2.69
2.46
**2.
702.
46*
2.57
2.63
2.60
2.45
2.53
2.67
2.51
2.72
*2.
542.
62
(g)
We
issu
elo
ng-t
erm
deb
tto
min
imiz
eth
eri
skof
hav
ing
tore"
nan
cein`b
adti
mesa
48.8
32.
152.
052.
29*
2.31
2.03
*1.
952.
55**
*2.
262.
51*
2.22
2.05
2.39
1.79
***
2.18
2.10
(a)
We
issu
esh
ort
-ter
mw
hen
shor
t-te
rmin
tere
stra
tes
are
low
com
par
edto
long
-ter
mra
tes
35.9
41.
891.
792.
01**
1.97
2.11
1.82
1.93
2.22
2.05
2.00
1.74
**2.
031.
77**
1.95
1.67
**
(c)
We
issu
esh
ort
-ter
mw
hen
we
are
wai
tin
gfo
rlo
ng-
term
mar
ket
inte
rest
rate
sto
dec
line
28.7
01.
781.
661.
93**
2.01
1.82
1.67
1.90
**2.
002.
021.
911.
61**
*1.
901.
65**
1.82
1.67
(d)
We
bor
row
sho
rt-t
erm
soth
atre
turn
sfr
omne
wp
roje
cts
can
beca
ptu
red
mor
efu
llyb
ysh
areh
old
ers,
rath
erth
anco
mm
itti
ng
top
aylo
ng-
term
pro"
tsas
inte
rest
tode
bth
old
ers
9.48
0.94
1.03
0.80
**0.
870.
891.
010.
85*
0.84
0.77
0.98
0.87
1.05
0.81
**0.
890.
97
(e)
We
expe
cto
urcr
edit
rati
ng
toim
prov
e,so
we
borr
owsh
ort-
term
unti
lit
does
8.99
0.85
0.86
0.84
0.87
0.68
*0.
790.
99*
0.66
1.18
***
0.73
0.99
**0.
890.
850.
890.
87
(f)B
orro
win
gsh
ort-
term
red
uces
the
chan
ceth
atou
r"
rmw
illw
ant
tota
ke
onri
sky
pro
ject
s4.
020.
530.
620.
40**
*0.
540.
32**
0.56
0.49
0.36
0.56
**0.
470.
59*
0.53
0.51
0.40
0.70
***
224 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
%im
port
ant
CE
Oag
eC
EO
tenu
reC
EO
MB
AR
egu
late
dT
arge
tde
bt
rati
oP
ubl
icco
rp.
For
eign
sale
sF
ortu
ne
500
mai
ling
or
very
impo
rtan
tM
ean
'59
Yng
erL
ong
Shor
tY
esN
oY
esN
oN
oY
esY
esN
oY
esN
oN
oY
es
(b)
Mat
chin
gth
em
atu
rity
of
our
deb
tw
ith
the
life
of
our
asse
ts63
.25
2.60
2.28
2.69
***
2.69
2.53
2.59
2.64
2.81
2.60
2.53
2.66
2.47
2.85
***
2.33
2.69
***
2.65
2.39
*
(g)
We
issu
elo
ng-t
erm
deb
tto
min
imiz
eth
eri
skof
hav
ing
tore"
nan
cein`b
adti
mesa
48.8
32.
152.
092.
202.
252.
122.
202.
152.
482.
152.
002.
36**
*2.
232.
02*
2.40
2.06
**2.
112.
31
(a)
We
issu
esh
ort
-ter
mw
hen
shor
t-te
rmin
tere
stra
tes
are
low
com
par
edto
long
-ter
mra
tes
35.9
41.
891.
781.
931.
871.
901.
981.
871.
951.
861.
931.
852.
001.
72**
2.11
1.80
**1.
862.
03
(c)
We
issu
esh
ort
-ter
mw
hen
we
are
wai
tin
gfo
rlo
ng-
term
mar
ket
inte
rest
rate
sto
dec
line
28.7
01.
781.
681.
801.
791.
781.
741.
792.
401.
71**
*1.
721.
871.
931.
50**
*2.
001.
69**
1.74
1.94
(d)
We
bor
row
sho
rt-t
erm
soth
atre
turn
sfr
omn
ewp
roje
cts
can
beca
ptu
red
mor
efu
llyb
ysh
areh
old
ers,
rath
erth
anco
mm
itti
ngto
pay
long
-ter
mp
ro"
tsas
inte
rest
tod
ebth
old
ers
9.48
0.94
0.86
0.95
0.98
0.90
0.99
0.89
0.90
0.93
0.96
0.90
0.87
1.07
**0.
950.
930.
990.
70**
(e)
We
expe
cto
urcr
edit
rati
ng
toim
prov
e,so
we
borr
owsh
ort-
term
unti
lit
does
8.99
0.85
0.79
0.87
0.89
0.82
0.84
0.87
0.90
0.85
0.98
0.65
***
0.88
0.82
0.89
0.85
0.89
0.70
*
(f)B
orro
win
gsh
ort-
term
red
uces
the
chan
ceth
atou
r"
rmw
illw
ant
tota
ke
onri
sky
pro
ject
s4.
020.
530.
510.
530.
660.
44**
*0.
450.
560.
430.
540.
550.
510.
460.
67**
0.44
0.57
*0.
590.
29**
*
�Res
pon
den
tsar
eas
ked
tora
teo
na
scal
eo
f0(n
oti
mpo
rtan
t)to
4(v
ery
imp
orta
nt).
We
rep
ortt
he
ove
rall
mea
nas
wel
las
the
%of
resp
ond
ents
that
answ
ered
3an
d4
(ver
yim
por
tant
).**
*,**
,*de
not
esa
sign
i"ca
nt
di!
eren
ceat
the
1%,
5%,a
nd
10%
leve
l,re
spec
tive
ly.
All
tab
leco
lum
ns
are
de"
ned
inT
able
1.
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 225
and pass up positive NPV projects if they perceive that the pro"ts will be used topay o! existing debtholders. This cost is most acute among growth "rms. Myers(1977) argues that "rms can limit total debt, or use short-term debt, to minimizeunderinvestment costs. Froot, Scharfstein, and Stein (1993) argue that "rms canhedge or otherwise maintain "nancial #exibility to avoid these costs of underin-vestment.
We ask "rms if their choice between short- and long-term debt, or theiroverall debt policy, is related to their desire to pay long-term pro"ts to share-holders, not debtholders. The absolute number of "rms indicating that theirdebt policy is a!ected by underinvestment concerns is small (rating of 1.01 inTable 6). However, more growth (1.09) than nongrowth "rms (0.69) are likely toindicate that underinvestment problems are a concern, which is consistent withthe theory. We "nd little support for the idea that short-term debt is used toalleviate the underinvestment problem. The mean response is only 0.94 (Rowd in Table 11) that short-term borrowing is used to allow returns from newprojects to be captured by long-term shareholders, and there is no statisticaldi!erence in the response between growth and nongrowth "rms.
Overall, support for the underinvestment argument is weak. This is interestingbecause it contrasts with the "nding in many large sample studies that debtusage is inversely related to variables measuring growth options (i.e., market-to-book ratios), which those studies interpret as evidence that underinvestmentcosts a!ects debt policy (e.g., Graham, 1996).
Stockholders capture investment returns above those required to service debtpayments and other liabilities, and at the same time have limited liability whenreturns are insu$cient to fully pay debtholders. Therefore, stockholders preferhigh-risk projects, in con#ict with bondholder preferences. Leland and Toft(1996) argue that using short-term debt reduces this agency con#ict (see alsoBarnea et al., 1980).
In contrast to this hypothesis, however, we "nd little evidence that executivesissue short-term debt to minimize asset substitution problems. The mean re-sponse is only 0.53 (Table 11) that executives feel that short-term borrowingreduces the chance that shareholders will want to take on risky projects.
Green (1984) argues that convertible debt can circumvent the asset substitu-tion problem that arises when "rms accept projects that are riskier thanbondholders would prefer. However, we "nd little evidence that "rms useconvertibles to protect bondholders against unfavorable actions by managers orstockholders (rating of 0.62 in Table 10).
5.3.2. Conyicts between managers and equityholdersJensen (1986) and others argue that when a "rm has ample free cash #ow, its
managers can squander the cash by consuming perquisites or making ine$cientinvestment decisions. We inquire whether "rms use debt to commit to pay outfree cash #ows and thereby discipline management into working e$ciently
226 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
along the lines suggested by Jensen. We "nd very little evidence that "rmsdiscipline managers in this way (mean rating of 0.33, the second lowest ratingamong all factors a!ecting debt policy in Table 6). It is important to note,however, that 1) managers might be unwilling to admit to using debt in thismanner, or 2) perhaps a low rating on this question re#ects an unwillingness of"rms to adopt Jensen's solution more than a weakness in Jensen's argument.
5.4. Product market and industry factors
Bradley et al. (1984) "nd that debt ratios di!er markedly across industries.One explanation for this pattern is that the product market environment ornature of competition varies across industries in a way that a!ects optimal debtpolicy. For example, Titman (1984) suggests that customers avoid purchasinga "rm's products if they think that the "rm might go out of business (andtherefore not stand behind its products), especially if the products are unique;consequently, "rms that produce unique products might avoid using debt.Brander and Lewis (1986) model another way that production and "nancingdecisions can be intertwined. They hypothesize that, by using substantial debt,a "rm can provide a credible threat to rivals that it will not reduce production.
We "nd little evidence that product market factors a!ect debt decisions.Executives assign a mean rating of 1.24 to the proposition that debt should belimited so that a "rm's customers or suppliers do not become concerned that the"rm might go out of business (Table 6). Moreover, high-tech "rms (which weassume produce unique products) are less likely than other "rms to limit debtfor this reason, contrary to Titman's prediction. We do "nd that, in comparisonto nongrowth "rms (1.00), relatively many growth "rms (1.43) claim that cus-tomers might not purchase their products if they are worried that debt usagemight cause the "rm to go out of business. This is consistent with Titman'stheory if growth "rms produce unique products. Finally, there is no evidencesupporting the Brander and Lewis hypothesis that debt provides a credibleproduction threat (rating of 0.40).
Though we do not "nd much evidence that product market factors driveindustry di!erences in debt ratios, we ask executives whether their capitalstructure decisions are a!ected by the "nancing policy of other "rms in theirindustries. This is important because some papers de"ne a "rm's target debtratio as the industry-wide ratio (e.g., Opler and Titman, 1998; Gilson, 1997).
We "nd only modest evidence that managers are concerned about the debtlevels of their competitors (rating of 1.49 in Table 6). Recall, however, that creditratings are important to debt decisions and note that industry debt ratios are animportant input for bond ratings. Rival debt ratios are relatively important forregulated companies (2.32), Fortune 500 "rms (1.86), public "rms (rating of 1.63versus 1.27 for private "rms), and "rms that target their debt ratio (1.60).Moreover, equity issuance decisions are not in#uenced greatly by the equity
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 227
policies of other "rms in a given industry (rating of 1.45 in Table 8). Finally, we"nd even less evidence that "rms use convertibles because other "rms in theirindustry do so (1.10 in Table 10).
5.5. Control contests
Capital structure can be used to in#uence, or can be a!ected by, corporatecontrol contests and managerial share ownership (e.g., Harris and Raviv, 1988;Stulz, 1988). We "nd moderate evidence that "rms issue equity to dilute thestock holdings of certain shareholders (rating of 2.14 in Table 8). This tactic ispopular among speculative-grade companies (2.24); however, it is not related tothe number of shares held by managers. We also ask if "rms use debt to reducethe likelihood that the "rm will become a takeover target. We "nd little supportfor this hypothesis (rating of 0.73 in Table 6).
5.6. Risk management
Capital structure can be used to manage risk. GeH czy et al. (1997, p. 1331) notethat `foreign denominated debt can act as a natural hedge of foreign revenuesaand displace the need to hedge with currency derivatives. We ask whether "rmsuse foreign debt because it acts as a natural hedge, and separately how impor-tant it is to keep the source close to the use of funds. Among the 31% ofrespondents who seriously considered issuing foreign debt, the most popularreason they did so is to provide a natural hedge against foreign currencydevaluation (mean rating of 3.15 in Table 7). Providing a natural hedge is mostimportant for public "rms (3.21) with large foreign exposure (3.34). The secondmost important factor a!ecting the use of foreign debt is keeping the sourceclose to the use of funds (rating of 2.67), especially for small (3.09), manufacturing"rms (2.92).
Risk-management practices can also explain why "rms match the maturity ofassets and liabilities. If asset and liability duration are not aligned, interest rate#uctuations can a!ect the amount of funds available for investment and day-to-day operations. We ask "rms how they choose debt maturity. The most popularexplanation of how "rms choose between short- and long-term debt is that theymatch debt maturity with asset life (rating of 2.60 in Table 11). Maturity-matching is most important for small (2.69), private (2.85) "rms.
5.7. Practical, cash management considerations
Liquidity and cash management a!ect corporate "nancial decisions, often inways that are not as `deepa as the factors driving academic models. Forexample, many companies issue long-term so that they do not have to re"nancein `bad timesa (rating of 2.15 in Table 11). This is especially important for highly
228 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
levered (2.55), manufacturing (2.37) "rms. The CFOs also say that equity is oftenissued simply to provide shares to bonus/option plans (2.34 in Table 8), parti-cularly among investment-grade "rms (2.77) with a young CEO (2.65).
The hand-written responses indicate that practical considerations a!ect thematurity structure of borrowing (see B.7 on the Internet site, Appendix B). Four"rms explicitly say that they tie their scheduled principal repayments to theirprojected ability to repay. Another six diversify debt maturity to limit themagnitude of their re"nancing activity in any given year. Other "rms borrow forthe length of time they think they will need funds, or borrow short-term untilsu$cient debt has accumulated to justify borrowing long-term.
5.8. Other factors awecting capital structure
5.8.1. DebtWe ask if having debt allows "rms to bargain for concessions from employees
(Chang, 1992; Hanka, 1998). We "nd no indication that this is the case (meanrating of 0.16 in Table 6, the lowest rating for any question on the survey). Nota single respondent said that debt is important or very important as a bargain-ing device (rating of 3 or 4). We also check if "rms issue debt after recentlyaccumulating substantial pro"ts (Opler and Titman, 1998). The executives donot recognize this as an important factor a!ecting debt policy (rating 0.53 inTable 9).
Fourteen "rms write that they choose debt to minimize their weightedaverage cost of capital (see B.5 on the Internet site, Appendix B). Ten write,essentially, that they borrow to fund projects or growth, but only as needed. Fiveindicate that bond or bank covenants a!ect their debt policy.
5.8.2. Common stockWe investigate whether concern about earnings dilution a!ects equity
issuance decisions. The textbook view is that earnings are not diluted if a "rmearns the required return on the new equity. Conversely, if funds are obtained byissuing debt, the number of shares remains constant and so EPS can increase.However, the equity is levered and therefore more risky, so Modigliani andMiller's `conservation of valuea tells us that the stock price will not increase dueto higher EPS. Nonetheless, Brealey and Myers (1996) indicate that there isa common belief among executives that share issuance dilutes earnings per share(on p. 396, Brealey and Myers call this view a `fallacya). To investigate this issue,we ask if earnings per share concerns a!ect decisions about issuing commonstock.
Among the 38% of "rms that seriously considered issuing common equityduring the sample period, earnings dilution is the most important factor a!ect-ing their decision (mean rating of 2.84 in Table 8 and a mean rating of 3.18among public "rms). The popularity of this response is intriguing (see Fig. 7). It
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 229
Fig. 7. Survey evidence on some of the factors that a!ect the decision to issue common stock. Thesurvey is based on the responses of 392 CFOs.
either indicates that executives focus more than they should on earnings dilution(if the standard textbook view is correct), or that the standard textbook treat-ment misses an important aspect of earnings dilution. EPS dilution is a bigconcern among regulated companies (3.60), even though in many cases theregulatory process ensures that utilities earn their required cost of capital,implying that EPS dilution should not a!ect share price. Concern about EPSdilution is strong among large (3.12), dividend-paying "rms (3.06). EPS dilutionis less important when the CEO has an MBA (2.62) than when he or she does not(2.95), perhaps because the executive has read Brealey and Myers!
We inquire whether common stock is a "rm's least risky or cheapest source offunds. Williamson (1988) argues that equity is a cheap source of funds withwhich to "nance low-speci"city assets. A modest number of the executives statethat they use equity because it is the least risky source of funds (rating of 1.76 inTable 8). The idea that equity has low risk is more popular among "rms with thecharacteristics of a new or start-up "rm: small (1.93) with growth options (2.07).The idea that common stock is the cheapest source of funds is less popular(rating of 1.10), although "rms with start-up characteristics are more likely tohave this belief. Unreported analysis indicates that there is a positive correlationbetween believing that equity is the cheapest and that it is the least risky sourceof funds.
Nine companies indicate that they issue common stock because it is the`preferred currencya for making acquisitions, especially for the pooling methodof accounting (see B.9 on the Internet site, Appendix B). Two "rms write that
230 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Fig. 8. Survey evidence on the factors that a!ect the decision to issue convertible debt. The survey isbased on the responses of 392 CFOs.
they issue stock because it is the natural form of "nancing for them in theircurrent stage of corporate development.
5.8.3. Convertible debtWe ask the executives whether the ability to call or force conversion is an
important feature a!ecting convertible debt policy. Among the one-in-"ve "rmsthat seriously considered issuing convertible debt, there is moderate evidencethat executives like convertibles because of the ability to call or force conversion(rating of 2.29 in Table 10). Though not a direct test, the popularity of thecall/conversion feature is consistent with Mayers' (1998) hypothesis that con-vertible debt allows funding of pro"table future projects but attenuates overin-vestment incentives. The factors used in decisions to issue convertible debt arepresented in Fig. 8.
Billingsley et al. (1985) document that convertibles cost on average 50 basispoints less than straight debt. However, relatively few CFOs indicate that theyuse convertible debt because it is less expensive than straight debt (rating of1.85). Companies run by mature executives are more likely to issue convertiblesbecause they are less costly than straight debt (2.50).
Billingsley and Smith (1996) also "nd that convertibles are favored as delayedequity and because management feels that common equity is undervalued.Contrary to our results, Billingsley and Smith "nd fairly strong evidence that"rms are in#uenced by the convertible use of other "rms in their industry. Alsoin contrast to our results, they "nd that the most important factor a!ecting theuse of convertibles is the lower cash costs/coupon rate versus straight debt. Onedi!erence between our study and theirs is that they request a response relative to
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 231
a speci"c o!ering among "rms that actually issue convertible debt. We condi-tion only on whether a "rm seriously considered issuing convertibles.
5.8.4. Foreign debtGrinblatt and Titman (1998) note that capital markets have become increas-
ingly global in recent decades and that U.S. "rms frequently raise funds over-seas. We indicate above that "rms issue foreign debt in response to taxincentives, to keep the source close to the use of funds, and in an attempt to takeadvantage of low foreign interest rates. Five "rms write that they borrowoverseas to broaden their sources of "nancing (see B.8 on the Internet site,Appendix B). Few "rms indicate that foreign regulations require them to issueabroad (rating of 0.61 in Table 7).
6. Conclusions
Our survey of the practice of corporate "nance is both reassuring andpuzzling. For example, it is reassuring that NPV is dramatically more importantnow as a project evaluation method than, as indicated in past surveys, it was 10or 20 years ago. The CAPM is also widely used. However, it is surprising thatmore than half of the respondents would use their "rm's overall discount rate toevaluate a project in an overseas market, even though the project likely hasdi!erent risk attributes than the overall "rm. This indicates that practitionersmight not apply the CAPM or NPV rule correctly. It is also interesting thatCFOs pay very little attention to risk factors based on momentum and book-to-market value.
We identify fundamental di!erences between small and large "rms. Ourresearch suggests that small "rms are less sophisticated when it comes toevaluating risky projects. Small "rms are signi"cantly less likely to use the NPVcriterion or the capital asset pricing model and its variants. Perhaps these andour other "ndings about the e!ect of "rm size will help academics understandthe pervasive relation between size and corporate practices. Further, the factthat the practice of corporate "nance di!ers based on "rm size could be anunderlying cause of size-related asset pricing anomalies.
In our analysis of capital structure, we "nd that informal criteria such as"nancial #exibility and credit ratings are the most important debt policy factors.Other informal criteria such as EPS dilution and recent stock price appreciationare the most important factors in#uencing equity issuance. The degree of stockundervaluation is also important to equity issuance, and we know from othersurveys that most executives feel their stock is undervalued.
We "nd moderate support that "rms follow the trade-o! theory and targettheir debt ratio. Other results, such as the importance of equity undervaluationand "nancial #exibility, are generally consistent with the pecking-order view.However, the evidence in favor of these theories does not hold up as well under
232 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
closer scrutiny (e.g., the evidence is generally not consistent with informationalasymmetry causing pecking-order-like behavior), and is weaker still for moresubtle theories. We "nd mixed or little evidence that signaling, transactionscosts, underinvestment costs, asset substitution, bargaining with employees, freecash #ow considerations, and product market concerns a!ect capital structurechoice. Table 12 summarizes our capital structure "ndings.
In summary, executives use the mainline techniques that business schoolshave taught for years, NPV and CAPM, to value projects and to estimate thecost of equity. Interestingly, "nancial executives are much less likely to followthe academically proscribed factors and theories when determining capitalstructure. This last "nding raises possibilities that require additional thoughtand research. Perhaps the relatively weak support for many capital structuretheories indicates that it is time to critically reevaluate the assumptions andimplications of these mainline theories. Alternatively, perhaps the theories arevalid descriptions of what "rms should do}but corporations ignore the theoret-ical advice. One explanation for this last possibility is that business schoolsmight be better at teaching capital budgeting and the cost of capital than atteaching capital structure. Moreover, perhaps NPV and the CAPM are morewidely understood than capital structure theories because they make moreprecise predictions and have been accepted as mainstream views for longer.Additional research is needed to investigate these issues.
Table 12Summary of the relation between survey evidence and capital structure theories
A capital structure theory or concept is listed in the "rst column, followed by the related surveyevidence in the right column. � (�) indicates that the evidence drawn from the unconditionalresponses to a survey question supports (does not support) the idea in the "rst column. An indented� (�) indicates whether the survey evidence supports (does not support) the idea conditional on "rmcharacteristics or other detailed analysis. The conditional (i.e., indented) evidence usually quali"esthe unconditional result it lies directly below.
Theory or concept Survey evidence
Trade-ow theory of choosing optimal debt policy �corporate interest deductions moderatelyimportant.
Trade-o! bene"ts and costs of debt (Scott, 1976). �foreign tax treatment moderately important.Often tax bene"ts are traded o! with expecteddistress costs or personal tax costs (Miller, 1977).
�cash #ow volatility important.
�expected distress/bankruptcy costs notimportant.�maintaining "nancial #exibility important(expected distress costs low).
�unrelated to whether "rm has target debtratio.�personal taxes not important to debt orequity decision.
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 233
Table 12 (continued)
Theory or concept Survey evidence
Firms have target debt ratios �44% have strict or somewhat stricttarget/range.
A static version of the trade-o! theory implies that"rms have an optimal, target debt ratio.
�64% of investment-grade "rms havesomewhat strict target/range.�target D/E moderately important for equityissuance decision.�37% have #exible and 19% have notarget/range.�issue equity after stock price increase.�changes in stock price not important to debtdecision.�execs say same-industry debt ratios are notimportant.
�there are industry patterns in reporteddebt ratios.
The ewect of transactions costs on debt ratios: �transactions costs a!ect debt policy.Transactions costs can a!ect the cost of externalfunds.
�more important for small "rms.
Firms avoid or delay issuing or retiring securitybecause of issuance/recapitalization cost(Fisher et al., 1989)
�absolute importance is small fortransactions costs delaying debt issue.�transactions costs relatively important forsmall, no-dividend "rms.
�transactions costs do not cause "rms todelay debt retirement.
Pecking-order theory of xnancing hierarchy: �"rms value "nancial #exibility.Financial securities can be undervalued due toinformational asymmetry between managers andinvestors. Firms should use securities in reverseorder of asymmetry: use internal funds "rst, debtsecond, convertible security third, equity last
�desire for #exibility is unrelated to degreeof informational asymmetry (size) or growthstatus.
�#exibility less important for no-dividend"rms.�issue debt when internal funds areinsu$cient.
�more important for small "rms.To avoid need for external funds, "rms may preferto store excess cash (Myers and Majluf, 1984).
�no relation to growth or dividendstatus.
�issue equity when internal fundsinsu$cient.�relatively important for small "rms.
�equity issuance decision a!ected by equityundervaluation.
�no relation to size, dividend status, orexecutive ownership.
234 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Table 12 (continued)
Theory or concept Survey evidence
�equity issuance decision una!ected byability to obtain funds from debt, convertibles,or other sources.�debt issuance una!ected by equityvaluation.
�even less important for small, growth, no-dividend "rms.
Stock price: Recent increase in stock price presentsaawindow of opportunitya to issue equity(Loughran and Ritter, 1995). If stock undervalueddue to informational asymmetry, issue afterinformation release and ensuing stock priceincrease (Lucas and McDonald, 1990)
�issue equity when stock price has risen
�recent price increase most important for"rms that do not pay dividends (signi"cant)and small "rms (not signi"cant).
Credit ratings: "rms issue short-term if they expecttheir credit rating to improve (Flannery, 1986).
�In general, rating is very important to debtdecision.�short-term debt not used to time ratingimprovement.
Interest rates: do absolute coupon rates or relativerates between long- and short-term debt a!ectwhen debt is issued?
�issue debt when interest rates low.
�short-term debt used only moderately totime the level of interest rates or because ofyield curve slope.
Underinvestment: "rm may pass up NPV'0project because pro"ts #ow to existingbondholders. Can attenuate by limiting debt orusing short-term debt.
�low absolute importance of limiting the useof debt, or borrowing short-term, to avoidunderinvestment.
Most severe for growth "rms (Myers, 1977). �growth status has no e!ect on relative useof short-term debt.
�growth status a!ects relative importanceof limiting total debt.
Asset substitution: shareholders take on riskyprojects to expropriate wealth from bondholders(Jensen and Meckling, 1976). Using convertibledebt (Green, 1984) or short-term debt (Myers,1977) attenuates asset substitution, relative tousing long-term debt.
�neither convertible debt nor short-term debtis used to protect bondholders from the"rm/shareholders taking on risky orunfavorable projects.
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 235
Table 12 (continued)
Theory or concept Survey evidence
Free cash yow can lead to overinvestment orinezciency:
�debt is not used with intent of commitingfree cash #ows.
Fixed commitments like debt payments commitfree cash so management works hard ande$ciently (Jensen, 1986).
Product market and industry inyuences:
Debt policy credibly signals production decisions(Brander and Lewis, 1986).
�debt policy is not used to signal productionintentions.
Sensitive-product "rms use less debt so customersand suppliers do not worry about "rm enteringdistress (Titman, 1984).
�absolute importance of this explanation islow.
�not important for high-tech "rms.�relatively important for growth "rms.
Debt ratios are industry-speci"c (Bradley et al.,1984).
�"rms report that the debt, equity, andconvertibles usage of same-industry "rms doesnot a!ect "nancing decisions.
�empirical debt ratios di!er systematicallyacross industries.
Corporate control:Capital structure can be used to a!ect thelikelihood of success for a takeover bid/controlcontest. Managers may issue debt to increase theire!ective ownership (Harris andRaviv, 1988; Stulz, 1988).
�equity issued to dilute holdings of particularshareholders.�dilution strategy unrelated to managerialshare ownership.
�takeover threat does not a!ect debt decisions.
Risk management: "nance foreign operations withforeign debt as a means of hedging FX risk.
�foreign debt is frequently viewed as a naturalhedge.
Maturity-matching: match maturity between assetsand liabilities.
�important to choice between short- andlong-term debt.
Cash management: match cash out#ows to cashin#ows.
�long-term debt reduces the need to re"nancein bad times.
�spread out required principal repaymentsor link principal repayment to expected abilityto repay.
Employee stock/bonus plans: shares of stock neededto implement employee compensation plans.
�when funding employee plans, "rms avoidissuing shares, which would dilute theholdings of existing shareholders.
Bargaining with employees: high debt allowse!ective bargaining with employees (Chang, 1992).
�debt policy is not used as bargaining device
Earnings per share dilution �most important factor a!ecting equityissuance decision.
236 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Appendix. Nonresponse bias and other issues related to survey data
We perform several experiments to investigate whether nonresponse biasmight a!ect our results. The "rst experiment, suggested by Wallace and Mellor(1988), compares the responses for "rms that returned the survey on time (i.e., byFebruary 23) to those that did not return the survey until February 24, 1999 orlater. The "rms that did not respond on time can be thought of as a sample fromthe nonresponse group, in the sense that they did not return the survey until wepestered them further. We "rst test, for each question, whether the meanresponse for the early respondents di!ers from the mean for the late respon-dents. There are 88 questions not related to "rm characteristics. The meananswers for the early and late respondents are statistically di!erent for only eight(13) of these 88 questions at a 5% (10%) level.
Because the answers are correlated across di!erent questions, we also performmultivariate �� tests comparing the early and late responses. We calculatemultivariate test statistics for each set of subquestions, grouped by the mainquestion. (That is, one �� is calculated for the 12 subquestions related to the "rstquestion on the survey, another �� for the six subquestions related to the secondsurvey question, etc.) Out of the 10 multivariate �� s comparing the means forthe early and late responses, none (two) are signi"cantly di!erent at a 5% (10%)level. Following the order of the tables as they appear in the text, the multivari-ate analysis of variance p-values for each of the ten questions are 0.209, 0.063,0.085, 0.892, 0.124, 0.705, 0.335, 0.922, 0.259, and 0.282. A low p-value indicatessigni"cant di!erences between the early and late responses. Finally, a singlemultivariate �� across all 88 subquestions does not detect signi"cant di!erencesbetween the early and late responses ( p-value of 0.254). The rationale of Wallaceand Mellor suggests that because the responses for these two groups of "rms aresimilar, non-response bias is not a major problem.
The second set of experiments, suggested by Moore and Reichert (1983),investigates possible non-response bias by comparing characteristics of respon-ding "rms to characteristics for the population at large. If the characteristicsbetween the two groups match, then the sample can be thought of as represent-ing the population. This task is somewhat challenging because we have onlylimited information about the FEI population of "rms. (Given that mostFortune 500 "rms are also in the FEI population, we focus on FEI character-istics. We ignore any di!erences in population characteristics that may beattributable to the 187 "rms that are in the Fortune 500 but not in FEI.) Wehave reliable information on three characteristics for the population of "rmsthat belong to FEI: general industry classi"cation, public versus private owner-ship, and number of employees.
We "rst use �� goodness-of-"t analysis to determine whether the responsesrepresent the industry groupings in roughly the same proportion as that foundin the FEI population. Sixty-three percent of FEI members are from heavy
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 237
manufacturing industries (manufacturing, energy, and transportation), as are62% of the respondents. These percentages are not signi"cantly di!erent at the5% level. Therefore, the heavy manufacturing versus non-manufacturing break-down that we use in the tables is representative of the FEI population. We alsoexamine public versus private ownership. Sixty percent of FEI "rms are publiclyowned, as are 64% of the sample "rms. Again, these numbers are not statisticallydi!erent, suggesting that our numbers represent the FEI population, and alsothat our public versus private analysis is appropriate.
Although we do not have reliable information about the dividend policies,P/E ratios, sales revenue, or debt ratios for the FEI population, our analysisrelies heavily on these variables, so we perform Monte Carlo simulations todetermine the representativeness of our sample. Speci"cally, we take a randomsample of 392 "rms from the Compustat database, stratifying on the number ofemployees in FEI "rms. That is, we sample from Compustat so that 15.4% of thedraws are from "rms with at least 20,000 employees, 24.7% are from "rms withbetween 5,000 and 19,999 employees, etc., because these are the percentages forthe FEI population. We then calculate the mean debt ratio, sales revenue, andP/E ratio (ignoring "rms with negative earnings), and the percentage of "rmsthat pay dividends for the randomly drawn "rms. We repeat this process 1,000times to determine an empirical distribution of mean values for each variable.We then compare the mean values for our sample to the empirical distribution.If, for example, the mean debt ratio for the responding "rms is larger than 950 ofthe mean debt ratios in the Monte Carlo simulation, we would conclude thatthere is statistical evidence that respondent "rms are more highly levered thanare "rms in the overall population.
The sample values for sales revenue and debt ratios fall comfortably nearthe middle of the empirical distributions, indicating that the sample is repre-sentative for these two characteristics. The mean P/E ratio of 17 for the sampleis statistically smaller than the mean for the Compustat sample (overall meanof approximately 20). Fifty-four percent of the sample "rms pay dividends,compared to approximately 45% in the strati"ed Compustat sample.Although the sample and population di!er statistically for these last two traits,the economic di!erences are small enough to indicate that our sample isrepresentative of the population from which it is drawn. There are at leastthree reasons why our Monte Carlo experiment might indicate statisticaldi!erences, even if our sample "rms are actually representative of the FEIpopulation: (1) there are systematic di!erences between the Compustat andFEI populations not controlled for with the strati"cation based on numberof employees, (2) the strati"cation is based on FEI "rms only, although thesurvey `oversamplesa Fortune 500 "rms, and (3) we deleted "rms withnegative P/E ratios in the Monte Carlo simulations, although survey res-pondents might have entered a P/E ratio of zero or something else if they hadnegative earnings.
238 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Finally, given that much corporate "nance research analyzes Compustat"rms, we repeat the Monte Carlo experiment without stratifying by number ofemployees. That is, we randomly draw 392 "rms (1,000 times) from Compustatwithout conditioning on the number of employees. This experiment tells uswhether our sample "rms adequately represent Compustat "rms, to provide anindication of how directly our survey results can be compared to Compustat-based research. The mean debt ratio, sales revenue, and P/E ratios are notstatistically di!erent from the means in the Compustat data; however, thepercentage of "rms paying dividends is smaller than for the overall Compustatsample. Aside from dividend payout, the "rms that responded to our survey aresimilar to Compustat "rms.
If one accepts that nonresponse bias is small, there are still concerns aboutsurvey data. For one thing, the respondents might not answer truthfully. Giventhat the survey is anonymous, we feel this problem is minimal. Moreover, ourassessment from the phone conversations is that the executives would not takethe time to "ll out a survey if their intent was to be untruthful.
Another potential problem with survey data is that the questions, no matterhow carefully crafted, either might not be properly understood or might notelicit the appropriate information. For example, Stigler (1966) asks managers iftheir "rms maximize pro"ts. The general response is that, no, they take care oftheir employees, are responsible corporate citizens, etc. However, when Stiglerasks whether the "rms could increase pro"ts by increasing or decreasing prices,the answer is again no. Observations such as these can be used to argue thatthere is some sort of `economic Darwinisma, in which the "rms that survivemust be doing the proper things, even if unintentionally. Or, as Milton Fried-man (1953) notes, a good pool player has the skill to knock the billiards ballsinto one another just right, even if he or she can not solve a di!erential equation.Finally, Cli! Smith tells about a chef who, after tasting the un"nished product,always knew exactly which ingredient to add to perfect the day's recipe, butcould never write down the proper list of ingredients after the meal wascomplete. These examples suggest that managers might use the proper tech-niques, or at least take the correct actions, even if their answers to a survey donot indicate so. If other "rms copy the actions of successful "rms, then it ispossible that many "rms take appropriate actions without thinking within thebox of an academic model.
This set of critiques is impossible to completely refute. We have attempted tobe very careful when designing the questions on the survey. We also feel that bycontrasting the answers conditional on "rm characteristics, we should be able todetect patterns in the responses that shed light on the importance of di!erenttheories, even if the questions are not perfect in every dimension. Ultimately,however, the analysis we perform and conclusions we reach must be interpretedkeeping in mind that our data are from a survey. Having said this, we feelthat these data are representative and provide much unique information that
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 239
complements what we can learn from traditional large-sample analysis andclinical studies.
References
Baker, H.K., Farrelly, G.E., Edelman, R.B., 1985. A survey of management views on dividend policy.Financial Management 14, 78}84.
Barnea, A., Haugen, R., Senbet, L., 1980. A rationale for debt maturity structure and call provisionsin the agency theoretic framework. Journal of Finance 35, 1223}1234.
Bierman, H.J., 1993. Capital budgeting in 1992: a survey. Financial Management 22, 24.Billingsley, R.S., Lamy, R.E., Marr, M.W., Thompson, G.R., 1985. Explaining yield savings on new
convertible bond issues. Quarterly Journal of Business and Economics, 92}104.Billingsley, R.S., Smith, D.M., 1996. Why do "rms issue convertible debt? Financial Management 25,
93}99.Block, S.B., 1999. A study of "nancial analysts: practice and theory. Financial Analysts Journal,
86}95.Bodnar, G.M., Hayt, G.S., Marston, R.C., 1998. Wharton survey of "nancial risk management by
US non-"nancial "rms. Financial Management 27, 70}91.Bradley, M., Jarrell, G.A., Kim, E.H., 1984. On the existence of an optimal capital structure. Journal
of Finance 39, 899}917.Brander, J.A., Lewis, T.R., 1986. Oligopoly and "nancial structure: the limited liability e!ect.
American Economic Review 76, 956}970.Brealey, R.A., Myers, S.C., 1996. Principles of Corporate Finance, 5th Edition. McGraw-Hill, New
York.Brennan, M.J., Kraus, A., 1987. E$cient "nancing under asymmetric information. Journal of
Finance 42, 1225}1243.Brennan, M.J., Schwartz, E.S., 1988. The case for convertibles. Journal of Applied Corporate
Finance 1, 55}64.Bruner, R.F., Eades, K.M., Harris, R., Higgins, R.C., 1998. Best practices in estimating the cost of
capital: survey and synthesis. Financial Practice and Education 8, 13}28.Castanias, R., 1983. Bankruptcy risk and optimal capital structure. Journal of Finance 38, 1617}1635.Chang, C., 1992. Capital structure as an optimal contract between employees and investors. Journal
of Finance 47, 1141}1158.Chen, N.-f., Roll, R., Ross, S.A., 1986. Economic forces and the stock market. Journal of Business 59,
383}404.Desai, M.A., 1998. A multinational perspective on capital structure choice and internal capital
markets. Unpublished Working Paper, Harvard Business School.Donaldson, G., 1994. Corporate Restructuring: Managing the Change Process from Within. Har-
vard University Press, Boston.Epps, R.W., Mitchem, C.E., 1994. A comparison of capital budgeting techniques with those used in
Japan and Korea. Advances in International Accounting 7, 205}214.Fama, E.F., French, K.R., 1992. The cross-section of expected stock returns. Journal of Finance 47,
427}465.Ferson, W.E., Harvey, C.R., 1991. The variation of economic risk premiums. Journal of Political
Economy 99, 285}315.Ferson, W.E., Harvey, C.R., 1993. The risk and predictability of international equity returns. Review
of Financial Studies 6, 527}566.Fisher, E.O., Heinkel, R., Zechner, J., 1989. Dynamic capital structure choice: theory and tests.
Journal of Finance 44, 19}40.
240 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Flannery, M.J., 1986. Asymmetric information and risky debt maturity choice. Journal of Finance41, 19}37.
Friedman, M., 1953. Essays in Positive Economics. University of Chicago Press, Chicago.Froot, K.A., Scharfstein, D.S., Stein, J.C., 1993. Risk management: coordinating corporate invest-
ment and "nancing policies. Journal of Finance 48, 1629}1658.GeH czy, C., Minton, B.A., Schrand, C., 1997. Why "rms use currency derivatives. Journal of Finance
52, 1323}1354.Gilson, S.C., 1997. Transactions costs and capital structure choice: evidence from "nancially
distressed "rms. Journal of Finance 52, 111}133.Gitman, L.J., Forrester Jr., J.R., 1977. A survey of capital budgeting techniques used by major U.S."rms. Financial Management 6, 66}71.
Gitman, L.J., Mercurio, V., 1982. Cost of capital techniques used by major U. S. "rms: survey andanalysis of Fortune's 1000. Financial Management 14, 21}29.
Graham, J.R., 1996. Debt and the marginal tax rate. Journal of Financial Economics 41,41}73.
Graham, J.R., 1999a. Do personal taxes a!ect corporate "nancing decisions? Journal of PublicEconomics 73, 147}185.
Graham, J.R., 1999b. Quarter 2, 1999 FEI Survey. http://www.duke.edu/&jgraham.Graham, J.R., 2000. How big are the tax bene"ts of debt? Journal of Finance 55, 1901}1941.Green, R., 1984. Investment incentives debt and warrants. Journal of Financial Economics 13,
115}136.Grinblatt, M., Titman, S., 1998. Financial Markets and Corporate Strategy. Irwin McGraw-Hill,
Boston.Hanka, G., 1998. Debt and the terms of employment. Journal of Financial Economics 48,
245}282.Harris, M., Raviv, A., 1988. Corporate control contests and capital structure. Journal of Financial
Economics 20, 55}86.Heaton, J.B., 2000. Managerial optimism and corporate "nance. Unpublished Working Paper,
University of Chicago.Helwege, J., Liang, N., 1996. Is there a pecking order? Evidence from a panel of IPO "rms. Journal of
Financial Economics 40, 429}458.Jagannathan, R., Kubota, K., Takehara, H., 1998. Relationship between labor income risk and
average return: empirical evidence from the Japanese stock market. Journal of Business, 71,319}348.
Jagannathan, R., Wang, Z., 1996. The conditional CAPM and the cross-section of expected returns.Journal of Finance, 51, 3}53.
Jegadeesh, N., Titman, S., 1993. Returns to buying winners and selling losers: implications for stockmarket e$ciency. Journal of Finance 48, 65}91.
Jensen, M.C., 1986. Agency costs of free cash #ow, corporate "nance, and takeovers. AmericanEconomic Review 76, 323}339.
Jensen, M.C., Meckling, W., 1976. Theory of the "rm: managerial behavior, agency costs andownership structure. Journal of Financial Economics 3, 305}360.
Kale, J.R., Noe, T.H., 1990. Risky debt maturity choice in a sequential game equilibrium. Journal ofFinancial Research 13, 155}165.
Leland, H.E., Pyle, D.H., 1977. Informational asymmetries, "nancial structure, and "nancial inter-mediation. Journal of Finance 32, 371}387.
Leland, H.E., Toft, K.B., 1996. Optimal capital structure, endogenous bankruptcy, and the termstructure of credit spreads. Journal of Finance 51, 987}1019.
Lintner, J., 1956. Distribution of incomes of corporations among dividends, retained earnings, andtaxes. American Economic Review 97}113.
Loughran, T., Ritter, J.R., 1995. The new issues puzzle. Journal of Finance 50, 23}52.
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 241
Lucas, D.J., McDonald, R.L., 1990. Equity issues and stock price dynamics. Journal of Finance 45,1019}1043.
Mayers, D., 1998. Why "rms issue convertible bonds: the matching of "nancial and real investmentoptions. Journal of Financial Economics 47, 83}102.
McDonald, R.L., 1998. Real options and rules of thumb in capital budgeting. In: Brennan, M.J.,Trigeorgis, L. (Eds.), Innovation, Infrastructure, and Strategic Options. Oxford University Press,London.
Miller, M.H., 1977. Debt and taxes. Journal of Finance 32, 261}275.Modigliani, F., Miller, M.H., 1963. Corporate income taxes and the cost of capital: a correction.
American Economic Review 53, 433}443.Moore, J.S., Reichert, A.K., 1983. An analysis of the "nancial management techniques currently
employed by large. U.S. corporations. Journal of Business Finance and Accounting 10,623}645.
Myers, S.C., 1977. Determinants of corporate borrowing. Journal of Financial Economics 5,147}175.
Myers, S.C., 1984. The capital structure puzzle. Journal of Finance 39, 575}592.Myers, S.C., Majluf, N., 1984. Corporate "nancing and investment decisions when "rms have
information that investors do not have. Journal of Financial Economics 13, 187}224.Opler, T.C., Pinkowitz, L., Stulz, R., Williamson, R., 1999. The determinants and implications of
corporate cash holdings. Journal of Financial Economics 52, 3}46.Opler, T.C., Titman, S., 1998. The debt}equity choice, Unpublished Working Paper, Ohio State
University.Pinegar, J.M., Wilbricht, L., 1989. What managers think of capital structure theory: a survey.
Financial Management 18, 82}91.Poterba, J., Summers, L., 1995. A CEO survey of U.S. companies' time horizon and hurdle rates,
Sloan Management Review, Fall, 43}53.Ross, S.A., 1977. The determination of "nancial structure: the incentive signaling approach. Bell
Journal of Economics 8, 1}32.Sangster, A., 1993. Capital investment appraisal techniques: a survey of current usage. Journal of
Business Finance and Accounting 20, 307}332.Scott, J.H., 1976. A theory of optimal capital structure. Bell Journal of Economics 7 (1),
33}54.Shao, L.P., Shao, A.T., 1996. Risk analysis and capital budgeting techniques of U. S. multinational
enterprises. Managerial Finance 22, 41}57.Sharpe, S.A., Nguyen, H.H., 1995. Capital market imperfection and the incentive to lease. Journal of
Financial Economics 39, 271}294.Stanley, M.T., Block, S.B., 1984. A survey of multinational capital budgeting. The Financial Review
19, 36}54.Stein, J.C., 1992. Convertible bonds as backdoor equity "nancing. Journal of Financial Economics
32, 3}21.Stigler, G.J., 1966. The Theory of Price, 3rd Edition. Macmillan, New York.Stulz, R., 1988. Managerial control of voting rights: "nancing policies and the market for corporate
control. Journal of Financial Economics 20, 25}54.Titman, S., 1984. The e!ect of capital structure on a "rm's liquidation decision. Journal of Financial
Economics 13, 137}151.Titman, S., Wessels, R., 1988. The determinants of capital structure choice. Journal of Finance 43,
1}19.Trahan, E.A., Gitman, L.J., 1995. Bridging the theory}practice gap in corporate "nance: a survey of
chief "nancial o$cers. Quarterly Review of Economics and Finance 35, 73}87.Wallace, R., Mellor, C., 1988. Nonresponse bias in mail accounting surveys: a pedagogical note.
British Accounting Review 20, 131}139.
242 J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243
Wansley, J.W., Lane, W.R., Sarkar, S., 1989. Management's view on share repurchase and tendero!er premiums. Financial Management 18, 97}110.
Weston, J.F., Brigham, E., 1981. Managerial Finance, 7th Edition. Holt, Rinehart & Winston,Hinsdale, IL.
Williamson, O.E., 1988. Corporate "nance and corporate governance. Journal of Finance 43,567}591.
J.R. Graham, C.R. Harvey / Journal of Financial Economics 60 (2001) 187}243 243