best practices in estimating the cost of capital - an update
DESCRIPTION
aTRANSCRIPT
"Best Practices" in Estimating the Costof Capital: An Update
W. Todd Brotherson, Kenneth M. Eades, Robert S. Harris, and Robert C. Higgins
Theories on cost of capital have been around for decades.Unfortunately for practice, the academic discussions typicallystop at a high level of generality, leaving important questionsfor application unanswered. Recent upheavals in financialmarkets have only made the practitioner s task more difficultThis paper updates our earlier work on the state of the artin cost of capital estimation to identify current best practicesthat emerge. Unlike many broadly distributed multiplechoice or fill-in-the-blank surveys, our findings are basedon conversations with practitioners at highly regardedcorporations and leading financial advisors. We also reporton advice from best-selling textbooks and trade books.We find close alignment among all these groups on use ofcommon theoretical frameworks to estimate the cost ofcapital and on many aspects of estimation. We find largevariation, however, for the joint choices of the risk-free rateof retum, beta and the equity market risk premium, as wellas for the adjustment of capital costs for specific investmentrisk. When compared to our 1998 publication, we find thatpractice has changed .wme since the late 1990s but thereis still no consensus on important practical issues. Thepaper ends with a synthesis of messages from best practicecompanies and financial advisors and our conclusions.
The authors thank Bob Bruner for collaboration on prior research. Hisduties as a dean, however, prevent his joining this effort. We thank MSCIfor sharing Barra data. The research would not have been possible withoutthe cooperation of the companies and financial advisors surveyed. Thesecontributions notwithstanding, any errors remain the authors '.
W. Todd Brotherson is Assistant professor at Southern Virginia Universityin Buena Vista, VA. Kenneth M. Eades is a Professor at the DardenGraduate School of Business Administration at the University of Virginiain Charlottesville, VA. Robert S. Harris is a Professor at the DardenGraduate School of Business Administration at the University of Virginia inCharlottesville, VA. Robert C. Higgins is Professor Emeritus at the FosterSchool of Business at the University of Washington in Seattle, WA.
•"Cost of capital is so critical to things we do, and CAPMhas so many holes in it—and the books don't tell you whichnumbers to use... so at the end of the day, you wonder a bitif you've got a solid number. Am I fooling myself with thiswell-disciplined, quantifiable number?"
- A corporate survey participant
Over the years, theoretical developments in financeconverged into compelling recommendations about the costof capital to a corporation. By the early 1990s, a consensushad emerged prompting descriptions such as "traditional...textbook...appropriate," "theoretically correct," "a usefialmle of thumb" and a "good vehicle." In prior work withBob Bmner, we reached out to highly regarded firms andfinancial advisors to see how they dealt with the many issuesof implementation.' Fifteen years have passed since our firststudy. We revisit the issues and see what now constitutesbest practice and what has changed in both academicrecommendations and in practice.
We present evidence on how some of the most financiallysophisticated companies and financial advisors estimatecapital costs. This evidence is valuable in several respects.First, it identifies the most important ambiguities in theapplication of cost of capital theory, setting the stage forproductive debate and research on their resolution. Second, ithelps interested companies to benchmark their cost of capitalestimation practices against best-practice peers. Third, theevidence sheds light on the accuracy with which capital costscan be reasonably estimated, enabling executives to use theestimates more wisely in their decision-making. Fourth, it
' To provide a self-contained article we draw directly on portions of ourearlier work, Bruner, Eades, Harris, and Higgins (1998), which alsodiscusses surveys from earlier years.
15
16
enables teachers to answer the inevitable question, "But howdo companies really estimate their cost of capital?"
The paper is part of a lengthy tradition of surveys ofindustry practice. For instance. Bums and Walker (2009)examine a large set of surveys conducted over the lastquarter century into how US companies make capitalbudgeting decisions. They find that estimating the weightedaverage cost of capital is the primary approach to selectinghurdle rates. More recently, Jacobs and Shivdasani (2012)report on a large-scale survey of how financial practitionersimplement cost of capital estimation. Our approach differsfrom most papers in several important respects. Typicallystudies are based on written, closed-end surveys sentelectronically to a large sample of firms, often covering awide array of topics, and commonly using multiple choiceor fill-in-the-blank questions. Such an approach typicallyyields low response rates and provides limited opportunifyto explore subtleties of the topic. For instance, Jacobs andShivdasani (2012) provide useful insights based on theAssociation for Finance Professionals (AFP) cost of capitalsurvey. While the survey had 309 respondents, AFP (2011,page 18) reports this was a response rate of about 7% basedon its membership companies. In contrast, we report theresult of personal telephone interviews with practitionersfrom a carefully chosen group of leading corporations andfinancial advisors. Another important difference is that manyexisting papers focus on how well accepted modem financialtechniques are among practitioners, while we are interestedin those areas of cost of capital estimation where financetheory is silent or ambiguous and practitioners are left totheir own devices.
The following section gives a brief overview of theweighted-average cost of capital. The research approachand sample selection are discussed in Section II. Section IIIreports the general survey results. Key points of disparity arereviewed in Section IV. Section V discusses further surveyresults on risk adjustment to a baseline cost of capital, andSection VI highlights some institutional and market forcesaffecting cost of capital estimation. Section VII offersconclusions and implications for the financial practitioner.
I. The Weighted-Average Cost of Capital
A key insight from finance theory is that any use of capitalimposes an opportunity cost on investors; namely, funds arediverted from eaming a retum on the next best equal-riskinvestment. Since investors have access to a host of financialmarket opportunities, corporate uses of capital must bebenchmarked against these capital market altematives.The cost of capital provides this benchmark. Unless a firmcan earn in excess of its cost of capital on an average-riskinvestment, it will not create economic profit or value forinvestors.
JOURNAL OF APPLIED FINANCE - No. 1, 2013
A standard means of expressing a company's cost ofcapital is the weighted-average of the cost of individualsources of capital employed. In symbols, a company'sweighted-average cost of capital (or WACC) is:
W K ) ,equity equity^
(1)
where:K = component cost of capital.W = weight of each component as percent of total capital,t = marginal corporate tax rate.
For simplicify, this formula includes only two sources ofcapital; it can be easily expanded to include other sourcesas well.
Finance theory offers several important observationswhen estimating a company's WACC. First, the capital costsappearing in the equation should be current costs reflectingcurrent financial market conditions, not historical, sunk costs.In essence, the costs should equal the investors' anticipatedintemal rate of retum on future cash flows associated witheach form of capital. Second, the weights appearing in theequation should be market weights, not historical weightsbased on often arbitrary, out-of-date book values. Third,the cost of debt should be after corporate tax, reflecting thebenefits of the tax deductibility of interest.
Despite the guidance provided by finance theory, use ofthe weighted-average expression to estimate a company'scost of capital still confronts the practitioner with a numberof difficult choices.- As our survey results demonstrate, themost nettlesome component of WACC estimation is the costof equify capital; for unlike readily available yields in bondmarkets, no observable counterpart exists for equities. Thisforces practitioners to rely on more abstract and indirectmethods to estimate the cost of equity capital.
II. Sample Selection
This paper describes the results of conversations withleading practitioners. Believing that the complexify of thesubject does not lend itself to a written questionnaire, wewanted to solicit an explanation of each firm's approachtold in the practitioner's own words. Though our telephoneinterviews were guided by a series of questions, theconversations were sufficiently open-ended to reveal manysubtle differences in practice.
Since our focus is on the gaps between theory andapplication rather than on average or fypical practice, we
- Even at the theoretical level, the use of standard net present value (NPV)decision rules (with, for instance, WACC as a discount rate) does notcapture the option value of being able to delay an irreversible investmentexpenditure. As a result, a firm may find it better to delay an investmenteven if the current NPV is positive. Our survey does not explore the waysfirms deal with this issue; rather we focus on measuring capital costs.
BROTHERSON ET AL. - "BEST PRACTICES" IN ESTIMATING THE COST OF CAPITAL: A N UPDATE
, , Table I. Three Survey SamplesFull titles of textbooks and trade books are listed in the references at the end of this paper.
17
Company Sample Advisor Sample Textbook/Trade BookSample
AmerisourceBergenCaterpillarChevronCoca ColaCostco WholesaleIBMIntemational PaperIntuitJohnson ControlsPepsiCoQualcommSyscoTargetTexas InstrumentsUnion PacificUnited TechnologiesUPSW.W. GraingerWalt Disney
Bank of America MerrillLynchBarclays CapitalCredit SuisseDeutsche Bank AGEvercore PartnersGoldman Sachs & Co.Greenhill & Co, LLCJP MorganLazardMorgan StanleyUBS
TextbooksBrigham and Ehrhardt (2013)Ross, Westerfield and Jaffe (2013)Brealey, Myers, and Allen (2011)Higgins (2012)
Trade booksKoller, Goedhart, and Wessels (2005)Ibbotson(2012)
aimed to sample practitioners who were leaders in the field.We began by searching for a sample of corporations (ratherthan investors or financial advisors) in the belief that theyhad ample motivation to compute WACC carefully and toresolve many of the estimation issues themselves. Severalpublications offer lists of firms that are well-regarded infinance; of these, we chose Fortune's 2012 listing of MostAdmired Companies.^ Working with the Hay Group, Fortunecreates what it terms "the definitive report card on corporatereputations." Hay provided us with a listing of companiesranked by the criterion "wise use of assets" within industry.To create our sample we only used companies ranked first orsecond in their industry. We could not obtain raw scores thatwould allow comparisons across industries.
The 2012 Fortune rankings are based on a survey of 698companies, each of which is among the largest in its industry.For each of 58 industry lists. Hay asks executives, directors,and analysts to rate companies in their own industry on a setof criteria. Starting with the top two ranked firms in eachindustry, we eliminated companies headquartered outside
' For instance, Institutional Investor publishes lists of firms with the bestChief Financial Officers (CFOs), or with special competencies in certainareas. We elected not to use such lists because special competencies mightnot indicate a generally excellent finance department, nor might a stellarCFO. The source used by Bruner et al. (1998) is no longer published. Ourapproach, however, mirrors that earlier work.
North America (eight excluded)."* We also eliminated theone firm classified as a regulated utility (on the groundsthat regulatory mandates create unique issues for capitalbudgeting and cost of capital estimation) and the sevenfirms in financial services (inclusive of insurance, banking,securities and real estate). Forty-seven companies satisfiedour screens. Of these, 19 firms agreed to be interviewed andare included in the sample given in Table I. Despite multipleconcerted attempts we made to contact appropriate personnelat each company, our response rate is lower than Bruner,Eades, Harris, and Higgins (1998) but still much higher thantypical cost of capital surveys. We suspect that increases inthe number of surveys and in the demands on executives'time infiuence response rates now versus the late 1990s.
We approached corporate officers first with an emailexplaining our research. Our request was to interviewthe individual in charge of estimating the firm's WACC.We then arranged phone conversations. We promised ourinterviewees that, in preparing a report on our findings, wewould not identify the practices of any particular companyby name—we have respected this promise in the presentation
'' This screen was also used in Bruner et al. (1998). Our reasons forexcluding these firms were the increased difficulty of obtaining interviews,and possible difficulties in obtaining capital market information (such asbetas and equity market premiums).
18
that follows. ; 'In the interest of assessing the practices of the broader
community of finance practitioners, we surveyed two othersamples: •-' .:••-
• Financial advisors. Using a "league table" of merger andacquisition advisors from Thomson's Securities DataCommission (SDC) Mergers and Acquisitions database,we drew a sample of the most active advisors based onaggregate deal volume in M&A in the US for 2011. Ofthe top twelve advisors, one firm chose not to participatein the survey, giving us a sample of eleven.
We applied approximately the same set of questionsto representatives of these firms' M&A departments.'Financial advisors face a variety of different pressuresregarding cost of capital. When an advisor is represent-ing the sell side of an M&A deal, the client wants a highvaluation but the reverse may be tme when an advisoris acting on the buy side. In addition, banks may be en-gaged by either side of the deal to provide a FaimessOpinion about the transaction. We wondered whether thepressures of these various roles might result in financialadvisors using assumptions and methodologies that re-sult in different cost of capital estimates than those madeby operating companies. This proved not to be the case.
• Textbooks and Trade books. In parallel with our priorstudy, we focus on a handful of widely-used books. Froma leading textbook publisher we obtained names of thefour best-selling, graduate-level textbooks in corporatefinance in 2011. In addition, we consulted two populartrade books that discuss estimation of the cost of capitalin detail.
III. Survey FindingsTable II summarizes responses to our questions and shows
that the estimation approaches are broadly similar across thethree samples in several dimensions:
• Discounted Cash Flow (DCF) is the dominant investmentevaluation technique.
• WACC is the dominant discount rate used in DCF analy-ses.
• Weights are based on market not book value mixes of debtand equity.''
' Specific questions differ, reflecting the facts that financial advisorsinfrequently deal with capital budgeting matters and that corporate financialofficers infrequently value companies.
' The choice between target and actual proportions is not a simple one.Because debt and equity costs depend on the proportions of each employed,it might appear that the actual proportions must be used. However, if thefirm's target weights are publicly known and if investors expect the firmto move to these weights, then observed costs of debt and equity mayanticipate the target capital structure.
JOURNAL OF APPLIED FINANCE - NO. 1, 2013
• The after-tax cost of debt is predominantly based on mar-ginal pretax costs, and marginal tax rates.^
• The Capital Asset Pricing Model (CAPM) is the domi-nant model for estimating the cost of equity. Despiteshortcomings of the CAPM, our companies and financialadvisors adopt this approach. In fact, across both compa-nies and financial advisors, only one respondent did notuse the CAPM.«
These practices parallel many of the findings fromour earlier survey. First, the "best practice" firms showconsiderable alignment on many elements of practice.Second, they base their practice on financial economicmodels rather than on mies of thumb or arbitrary decisionmies. Third, the financial fi-ameworks offered by leadingtexts and trade books are fiindamentally unchanged fromour earlier survey.
On the other hand, disagreements exist within and amonggroups on matters of application, especially when it comesto using the CAPM to estimate the cost of equity. TheCAPM states that the required retum (K) on any asset canbe expressed as:
K=R+ß{R-R), (2)
where:R^ = interest rate available on a risk-free asset.R^^ = return required to attract investors to hold thebroad market portfolio of risky assets.ß = the relative risk of the particular asset.
According to CAPM then, the cost of equity, K , fora company depends on three components: retums on risk-free assets (Rj.), the stock's equity "beta" which measuresrisk of the company's stock relative to other risky assets(ß = 1.0 is average risk), and the market risk premium(R, - Rf) necessary to entice investors to hold risky assetsgenerally versus risk-free instmments. In theory, each ofthese components must be a forward-looking estimate. Oursurvey results show substantial disagreements, especially interms of estimating the market risk premium.
' In practice, complexities of tax laws such as tax-loss carry forwardsand carry backs, investment tax credits, state taxes, and intemational taxtreatments complicate the situation. Graham and Mills (2008) estimateeffective marginal tax rates. While most companies in their sample hit thestatutory marginal rate, the average effective marginal tax rate tumed outto be about 5% lower
"See, for instance, Brealey, Myers, and Allen (2011) pgs. 196-199 fora high-level review of some of the empirical evidence and the textbookauthors' rationale for continued use of the CAPM.
BROTHERSON ET AL. - "BEST PRACTICES" IN ESTIMATING THE COST OF CAPITAL: A N UPDATE19
S<l>
O
(A
gSoS-ou
ide
BOu
cd
C/Í
cd 3<D EË 60
T3Ccd
1 ^Z ECd
^ ^ 3>% (Ucd íCi=^'3
T3 1)B O.cd Í "
^ cd
cd c/5
00 ^
3oB
2
O
u•33 »1
Cd / ^
o Q
T3 g
II••B ' 5 i
•i ^cd i_i
IB
•2 S
1°
o .2cao B
3 E -So o E
o g-t u Cd
£f
o o
o
(N
'S
I
a
>. pa
o^ 3
f- (U
O-o
2 »i ^o > ^
cS 00 o
'^ S ^
S "S tí _^ Ê? ^ 'S'S p u 2
rit
so3
o_c'§>cd
^
'C
'SEo
ield
'S
3U
•S ''C
03
E2-a[3IÄ
ideb
t
s:Z
^ 11cd c Ciû
s- I I-a 'S "Î
'?> s o s00 .s
ï l'bO bo . ^
i l s
yiel
d
(N (N (N
o z^ S?i n -—'
11Is
Ë ëo §..S 8
IS?
cd
3 -ao Bà ==
^ 00T3 .ËS'c-i3 0.^> U O'S g' Ec ö .x:
Ë -s ¿o s i
<C dû [X(
<D ' S'S B
III I-,c,. BC- 2
II!
20JOURNAL OF APPLIED FINANCE - NO. 1, 2013
I
ÏV)O
I(O
20)
O
.Q
.«0
O(A
IU
2oS-oo
O
••s;
o
a
io
ies
asur
iTr
e;
<u
Lon
g-t
S?o
les:
íes
unsí
Tre;
¿D
3
a"gs.o.S
"cdo
hist
orii
nst
Tre
í
O
oc1)
Mat
ch
o?en
tst
m
.S
that
of
r*^ ^-<
43
ra
siI °t
2
H
M O.Ë ""o •g V 'o >
1
Ic
oB
iso
Bo a
a.
•Ö
ao
X IaOo,
oo^t—(N
O
•c>
,•—«
O
so;
î
BROTHERSON ET AL. - "BEST PRACTICES" IN ESTIMATING THE COST OF CAPITAL: AN UPDATE
w3(0a>
DC>
3(0
iî
Oo
CQa>•a
I(A
O
o
I(U
2o¡2
•o<
oa.oü
(A
O
8X(U
"cd
O
'r-|
Ient
ion
turn
s
" 2oo
o
hmet
i
cd
•anao8
o
eom
e
60N ^
'—1
'era
ges;
. ^
cd
s; 3
3'
i>tj
c3tri(^
som
e il
g
'S_uen
othe
i(Ä
3O
C
o'Stu
cd
I
o_o•aid
y, -gO o
c« 'S
p aX X ^m tx) tjs
o a•a 3S a
I 1. 2 Ö0
Î 5 I•p r j S-i
T3
1 I00 .tí
<U _ L , t u >! -tS
e b tí cd tuo ä o ^ c«o Ë .2 là o
m < Zm
cC <u
'o a
I i
os
60
g
a
est
i
atítu
-occdcd
sset
>>
cß
'S
tar
the
260
g C > .-S
UU
O =0
a a
D P 5
00
60 .
S •-atí
c§•
o.Moa>
J3ocd
Q, -^3 5
tu - MTT - ^ (11
E " Cd 'MO 00 ' 3" C 60 O
a '-5 tii
I « -S £
-oc o" 2 5X
111 11o a D. t- a
c
1toO
t :
M
O
S
3
g
ccd
.11
ea
•otícd
tß' ^
'Stu
•o"
O-
^ cen . 2^ .22
(D;3r/1 is:
(U
OC61
Bcd
1-^ OsOs
o 1
E l^ ^tícdO
So>• Z
^ 6b.3_ „
5 o
o 2 b c> a " gtu o ï: M'S > o g
li i i" "cd s 360 .-s a -5S g. -g •?> S 3 á
21
a
2 S .§tí =« -t;o tu .^
tí V ^
cd.tí
'S
.>! 3 ^(U c« f j
" l a
s, «ON o ^
^ votN —
cdtíO ! 8
O "g
ocd *> ^
a g I
m T)^ (U cß ^
o o
3 O,
t*H t j
il
22 JOURNAL OF APPLIED FINANCE - No. 1, 2013
tn
ST3(0
20)a>
CD
o
VIS
•o
"5'0cmc
S o> zr - • *(N vo
nse
o&2o
EoB-ou
too
a
uBOeneo
> ,.ti
til
—2o
ga2ena>o
cS
1eoeutn3
S£eu
g
Cti
2BOc/)
rea
cet
èôE
2Beo
•S
eus:2J
S'5.eao00ut.
1—1
3
EBeu
.C
c/î
bO "bô
we:
we:
m
ii
O cd
z B
ao-y
•o
icd
sIBIO
BeoB
^eo
O
3O
used
eaoe/î
eaeu
tn.2<u
. Ee4.H
O
3•DBeoeu
timat
eneu3O• * ^
eo
eu.aeu00Benx:o
S
capi
otoOO
eu
Iea
>. 22 =
oZ
.P -S "
3 eo— Eea eO
o aÍÍ So «Ü .5eu V-
x; eu• .g
'S•S S S
eLT e : ^ .
tieo 2
fc eu'—' D.
% ^cd O
>•5t= 0
o oo
'S
oZ T3
O
•a
fi .ë B
III'3 — "3bû cd "Q
ë g Ien 'r; 3
:E ea _
èa 00o eu
ns.
equa
tioD
Cl-
îmen
dre
corr
<
1 usi
ng
e
IU3U
JIU
jrec
o
0
tí
lati
ng 01
cuss
ion
iang
u10
dis
and
tritip
les
i
^-1
2eu
eu3
val
rate
s fo
iou
nt 1
îg d
ise
iffe
rir
T3
0
•S
teri
m c
and
in
/alu
e a
îinal
\
gB
c/î
flo
ng t
heva
ryi
îmen
d
II003J
<
0
•t th
e ri
s
1
ate
toou
nt I
0
3ne(
17'
t
low
s. (
•lCJ
C>
ion
bva
luat
usse
s
(/
B
•g
E
-ëo
a. MH
B X!« IO S« &
o
OBeu>eo
1 >o H
I ëCd Q
•S- ^ -Gs :S eu3 eu &
2 ï;s.^ ^ -S? NOo^ o^ o^ o^t~- 00 —I ONfN ^ os
«
g s we M 3
• rH t/î —-
"g OJ g
os!1c " ,0 _t/î OJ
o -2 eO
g Cd C
8 «
Ia
uu
i 1« oe« B
Sa 2D Q-^ s?
I «g CI;o o
•—• encd a>
g ^
eu
eu o3 a
•3 "> eu
.II
g
0000 • «
en ' '
Î%.ä "o•o .5§ eo
3 3 -O e/1
euen3
you
0T3
2ea0 ,euc/î
B0
vis
.aeOeu
Ifor
ea
D.ea0
0
0Ü
E2
ffe
BROrHERSON ET AL. - " B E S T PRACTICES" IN ESTIMATING THE CoST OF CAPITAL: A N UPDATE 23
3
Ë0)
0)
Ü
re
Oo
COo•an
!£o¡2
•o
oo.oo
tncotf)
a
tíoc«go
tí
.a
_>
ren
diff
e
iIR
•S -aM tu" tí
13 to
(NOO
oj ;
for
C
ISC
OU
tu
iffe
•ata
CA CA
r-i
(U
ira
i
S00
otu
iffe
"OCO
<D
O
ON
títa(U
3
you
o• a
ions
3
val
1...
you
hfl
yne
CA
alue
" ^
o
ods
Ití
ffer
e
00
'.S00ü
ties
(
3
port
a .o
gic
tu
Stra
ons
• ^
§•inB2
unti
SCO
•3tu
low
.2 3KS T3 o o
ca
¿3 Cd
^ CA
^ tu 'tí
•a -o g.
O o
tí
i I 8o •^ 13"
a,CJ
s
ii
I ^
'í .S
24
A. The Risk-free Rate of Retum
As originally derived, the CAPM is a single periodmodel, so the question of which interest rate best representsthe risk-free rate never arises. In a multi-period worldtypically characterized by upward-sloping yield curves, thepractitioner must choose. The difference between realizedretums on short-term US Treasury-bills and long-termT-bonds has averaged about 150 basis points over the long-run; so choice of a risk-free rate can have a material effect onthe cost of equity and WACC
Treasury bill yields are more consistent with the CAPMas originally derived and reflect risk-free retums in thesense that T-bill investors avoid material loss in value frominterest rate movements. However, long-term bond yieldsmore closely reflect the default-free holding period retumsavailable on long-lived investments and thus more closelymirror the types of investments made by companies.
Our survey results reveal a strong preference on the part ofpractitioners for long-term bond yields. As shown in Table II(Question 9), all the corporations and financial advisors useTreasury bond yields for maturities of 10 years or greater,with the 10-year rate being the most popular choice. Manycorporations said they matched the term of the risk-freerate to the tenor of the investment. In contrast, a third of thesample books suggested subtracting a term premium fromlong-term rates to approximate a shorter term yield. Half ofthe books recommended long-term rates but were not preciseon the choice of maturity.
Because the yield curve is ordinarily relatively flatbeyond ten years, the choice of which particular long-term yield to use often is not a critical one. However, at thetime of our survey. Treasury markets did not display these"normal" conditions in the wake of the financial crisis andexpansionary monetary policy. In the year we conducted oursurvey (2012), the spread between 10- and 30-year Treasuryyields averaged 112 basis points.'" While the text and tradebooks do not directly address the question of how to dealwith such markets, it is clear that some practitioners arelooking for ways to "normalize" what they see as unusualcircumstances in the govemment bond markets. For
'This was estimated as the difference in arithmetic mean retums on long-term govemment honds and US Treasury hills over the years 1900 to 2008,using data from E. Dimson and P. Marsh as cited in Brealey, Myers, andAllen (2011) in Tahle 7.1 at page 158.
'" Over the period 1977-2012 the difference hetween the yields on 10- and30-year Treasury honds averaged 29 hasis points. However, for the periodafter the financial crisis this spread has heen much higher in the midst ofexpansive Federal Reserve policy. For the period 2008-2012 the differenceaveraged 103 hasis points. During 2012 when we conducted our surveyit averaged 112 basis points. As we write in mid-January 2013, the yieldson 10-, 20-, and 30-year Treasury honds are 1.89%, 2.66%, and 3.06%respectively. Calculations use Federal Reserve H-15 interest rate data.
^ : JOURNAL OF APPLIED F I N A N C E - N o . 1, 2013
instance, 21% of the corporations and 36% of the financialadvisors resort to some historical average of interest ratesrather than the spot rate in the markets. Such an averagingpractice is at odds with finance theory in which investors seethe current market rate as the relevant opportunity. We retumto this issue later in the paper.
B. Beta Estimates
Finance theory calls for a forward-looking beta, onereflecting investors' uncertainty about the future cash flowsto equity. Because forward-looking betas are unobservable,practitioners are forced to rely on proxies of various kinds.Often this involves using beta estimates derived fromhistorical data.
The usual methodology is to estimate beta as the slopecoefficient of the market model of retums:
R. = a.+ß.{RJ, (3)
where:R.^ = return on stock I in time period (e.g., day, week,month) t.R^^^ = return on the market portfolio in period t.a. = regression constant for stock i.ß. = beta for stock i.
In addition to relying on historical data, use of thisequation to estimate beta requires a number of practicalcompromises, each of which can materially affect the results.For instance, increasing the number of time periods used inthe estimation may improve the statistical reliability of theestimate but risks including stale, irrelevant information.Similarly, shortening the observation period from monthlyto weekly, or even daily, increases the size of the sample butmay yield observations that are not normally distributed andmay introduce unwanted random noise. A third compromiseinvolves choice of the market index. Theory dictates thatR_ is the retum on the "market portfolio," an unobservableportfolio consisting of all risky assets, including humancapital and other non-traded assets, in proportion to theirimportance in world wealth. Beta providers use a varietyof stock market indices as proxies for the market portfolioon the argument that stock markets trade claims on asufficiently wide array of assets to be adequate surrogatesfor the unobsei-vable market portfolio.
Another approach is to "predict" beta based on underlyingcharacteristics of a company. According to Barra, the"predicted beta, the beta Barra derives from its risk model,is a forecast of a stock's sensitivity to the market. It is alsoknown as ftindamental beta because it is derived fromfijndamental risk factors.... such as size, yield, and volatility— plus industry exposure. Because we re-estimate these
BROTHERSON ET AL. - " B E S T PRACTICES" IN ESTIMATING THE COST OF CAPITAL: A N UPDATE 25
Table III. Compromises Underlying Beta Estimates and Their Effect on Estimated Betas ofSample Companies
Bloomberg* Value Line* Barra
Number of observationsTime intervalMarket index proxy
Sample mean betaSample median beta
102weekly over 2 yrs.
S&P 500
0.960.98
260weekly over 5 yrs.NYSE composite
0.930.90
statistical models usingcompany characteristics
0.910.96
* With the Bloomberg service it is possible to estimate a beta over many differing time periods, market indices, and smoothed or unadjusted.The figures presented here represent the baseline or default-estimation approach used if one does not specify other approaches. Value Linestates that "the Beta coefficient is derived from a regression analysis of the relationship between weekly percentage changes in the price ofa stock and weekly percentage changes in the NYSE Index over a period of five years. In the case of shorter price histories, a smaller timeperiod is used, but two years is the minimum. The betas are adjusted for their long-term tendency to converge toward 1.00."
AmerisourceBergen
Caterpillar
Chevron
Coca-Cola
Costco Wholesale
IBM
Intemational Paper
Intuit
Johnson Controls
PepsiCo
Qualcomm
Sysco
Target
Texas Instruments
Union Pacific
United Technologies
UPS
W.W. Grainger
Walt Disney
Mean
Median
Standard Deviation
Table
2-year raw
0.60
1.48
1.11
0.58
0.64
0.85
1.37
0.92
1.56
0.34
1.10
0.62
0.54
1.13
1.14
1.17
0.85
0.98
1.20
0.96
0.98
0.34
IV. Betas for' Corporate SurveyBloomberg Betas
5-year raw
0.61
1.45
1.01
0.55
0.73
0.83
L57
0.71
1.41
0.46
0.85
0.79
1.07
0.92
1.08
0.99
0.92
0.91
1.10
0.94
0.92
0.30
2-yearadjusted
0.74
1.32
1.07
0.72
0.76
0.90
1.25
0.95
1.38
0.56
1.07
0.75
0.69
1.09
1.09
1.12
0.90
0.98
1.13
0.97
0.98
0.23
RespondentsValue Line
5-year Betaadjusted
0.74
1.30
1.01
0.70
0.82
0.89
1.38
0.80
1.28
0.64
0.90
0.86
1.04
0.95
1.05
0.99
0.95
0.94
1.06
0.96
0.95
0.20
0.70
1.35
0.95
0.60
0.75
0.85
1.40
0.90
1.30
0.60
0.85
0.70
0.90
0.90
1.15
1.00
0.85
0.95
1.00
0.93
0.90
0.23
Barra
Beta
0.64
1.47
0.99
0.53
0.64
0.73
1.37
0.77
1.40
0.54
1.10
0.60
0.82
0.97
0.96
0.96
0.84
1.05
0.96
0.91
0.96
0.28
Value Line betas are as of January 11, 2013.Bloomberg betas are as of Jan 22, 2013. The adjusted beta is calculated as 2/3 times the raw beta plus 1/3 times 1.0.Barra betas are from January 2013. Source: Barra. The Barra data contained herein are the property of Barra, Inc. Barra,information providers make no warranties with respect to any such data. The Barra data contained herein is used undernot be further used, distributed or disseminated without the express written consent of Barra.
Range
iVIax-Min
0.14
0.18
0.16
0.19
0.18
0.17
0.32
0.24
0.29
0.30
0.26
0.26
0.53
0.23
0.19
0.21
0.11
0.14
0.24
0.23
0.21
0.09
its affiliates andlicense and may
26 JOURNAL OF APPLIED FINANCE - NO. 1, 2013
Table V. Choice of Beta
We asked our sample companies and advisors "what do you use as your beta estimate?" A sampling of responses shows the choice is notalways a simple one.
"We use Bloomberg's default calculation. We take special care to support beta chosen if doing an acquisition or major investment.""We use Bloomberg (both historical and historical adjusted) plus Barra and calibrate based on comparing the results.""We use our beta and check against those of competitors. We unlever the peers and then relever to our capital structure. The results comepretty close to our own beta, so that's confirming.""We use the median beta of comparable companies, based on analysis of size and business. We have a third party identify comparablecompanies and betas."We use Barra and Bloomberg and triangulate based on those numbers. The problem comes up for getting a pure play for a division or acompany that hasn't been public for a long time. We then have to pick comparable companies.""Each line of business has its own peer group of companies and cost of capital. Using five-year betas from Bloomberg, we find an averageunlevered beta for the peer group and then relever to our corporate-wide capital structure."
risk factors daily, the predicted beta reflects changes in thecompany's underlying risk structure in a timely manner.""Table III shows the compromises underlying the betaestimates of three prominent providers (Bloomberg, ValueLine and Barra) and their combined effect on the betaestimates of our sample companies. The mean beta of oursample companies is similar from all providers, 0.96 fromBloomberg, 0.93 according to Value Line, and 0.91 fromBarra. On the other hand, the averages mask differences forindividual companies. Table IV provides a complete list ofsample betas by provider.
Over half of the corporations in our sample (Table II,Question 10) cite Bloomberg as the source for their betaestimates, and some of the 37% that say they calculatetheir own may use Bloomberg data and programs. 26% ofthe companies cite some other published source and 26%explicitly compare a number of beta sources before makinga final choice. Among financial advisors, there is strongreliance on fundamental betas with 89% of the advisorsusing Barra as a source. Many advisors (44%) also useBloomberg. About a third of both companies and financialadvisors mentioned levering and unlevering betas eventhough we did not ask them for this information. And inresponse to a question about using data from other firms(Table II, Question 12), the majority of companies and alladvisors take advantage of data on comparable companies toinform their estimates of beta and capital costs.'-
Within these broad categories, the comments in TableV indicate that a number of survey participants use morepragmatic approaches which combine published betaestimates or adjust published estimates in various heuristicways.
" Barra (2007) on page 18. The actual quote from 2007 mentions monthlyupdating. In obtaining Barra data, we leamed that updates are now dailyso have substituted that in the quotes since we did not find an updatedreference.
'- Conroy and Harris (2011) compare the impacts of various approaches tousing comparable company data to estimate the cost of capital for firms inthe S&P 500.
C. Equity Market Risk Premium
This topic prompted the greatest variety of responsesamong survey participants. Finance theory says the equitymarket risk premium should equal the excess retumexpected by investors on the market portfolio relative toriskless assets. How one measures expected future retumson the market portfolio and on riskless assets is a problemleft to practitioners. Because expected future retums areunobservable, past surveys of practice have routinelyrevealed a wide array of choices for the market risk premium.For instance, Femandez, Aguirreamalloa, and Corres (2011)survey professors, analysts and companies on what theyuse as a US market risk premium. Of those who reported areference to justify their choice, the single most mentionedsource was Ibbotson/Momingstar, but even among thoseciting this reference, there was a wide dispersion of marketrisk premium estimates used. Carleton and Lakonishok(1985) demonstrate empirically some of the problems withsuch historical premiums when they are disaggregated fordifferent time periods or groups of firms. Dimson, Marsh,and Staunton (201 la, 201 Ib) discuss evidence from an arrayof markets around the globe.
How do our best practice companies cope? Amongfinancial advisors, 73% extrapolate historical retums intothe future on the presumption that past experience heavilyconditions future expectations. Among companies, 43%cite historical data and another 16% use various sourcesinclusive of historical data. Unlike the results of our earlierstudy (1998) in which historical retums were used by allcompanies and advisors, we found a number of respondents(18% of financial advisors and 32% of companies) usingforward-looking estimates of the market risk premium. Theadvisors cited versions of the dividend discount model. Thecompanies used a variety of methods including Bloomberg'sversion of the dividend discount model."
" In our conversations, company respondents sometimes noted they usedthe Bloomberg estimate of the market risk premium but did not havedetailed knowledge of the calculation. As described by Bloomberg, their
BROTHERSON ET AL. - " B E S T PRACTICES" IN ESTIMATING THE COST OF CAPITAL: A N UPDATE 27
Table VI. Historical Averages to Estimate the Equity Market Risk Premium, (R^ - R,)Figures are the annual average risk premium calculated over the period 1926-2011 based on data drawn from Table II-I, Ibbotson (2012),where the Rm was drawn from the "Large Company Stocks" series, and Rf drawn from the "Long-Term Govemment Bonds" and "USTreasury Bills" series.
Relative toT-Bill Returns
Relative toT-Bond Returns
Arithmetic MeanGeometric Mean
8.2%6.0%
5.7%3.9%
Even when historical retums are used to estimate themarket risk premium, a host of differences emerge includingwhat data to use and what method to use for averaging. Forinstance, a leading textbook cites US historical data backto 1900 from Dimson and Staunton (as cited by Brealey,Myers, and Allen (2011), p.l58) while 73% of our financialadvisors cite Ibbotson data which traces US history back to1926. Among companies, only 32% explicitly cite Ibbotsonas their main reference for data and 11% cite other historicalsources.'''
Even using the same data, another chief difference wasin their use of arithmetic versus geometric averages. Thearithmetic mean retum is the simple average of past retums.Assuming the distribution of retums is stable over time andthat periodic retums are independent of one another, thearithmetic retum is the best estimator of expected retum. Thegeometric mean retum is the intemal rate of retum between asingle outlay and one or more future receipts. It measures thecompound rate of retum investors eamed over past periods.It accurately portrays historical investment experience.Unless retums are the same each time period, the geometricaverage will always be less than the arithmetic average andthe gap widens as retums become more volatile."
Based on Ibbotson data (2012) from 1926 to 2011, TableVI illustrates the possible range of equity market riskpremiums depending on use of the geometric as opposedto the arithmetic mean equity retum and on use of realizedretums on T-bills as opposed to T-bonds. Even widervariations in market risk premiums can arise when onechanges the historical period for averaging or decides to use
estimates of country risk premiums are based on projections of futuredividends in a multistage dividend discount model. At the time of oursurvey, the US country risk premium from Bloomberg was in the rangeof 8-9%, well above averages from historical retums. See Harris andMarston (2001, 2013) for discussion of changing risk premiums over time.
''' With only minor exceptions, our respondents used US data rather thanrely on global indices for both their estimates of beta and the marketrisk premium. Dimson, Marsh, and Staunton (2011a) discuss historicalestimates of the market risk premium using data from many countries.
'5 See Brealey, Myers, and Allen (2011), pages 158-159 for the argumentto support use of an arithmetic mean. For large samples of returns, thegeometric average can be approximated as the arithmetic average minusone half the variance of realized retums.
data from outside the US For instance, Dimson, Marsh, andStaunton (2011a) provide estimates of historical equity riskpremiums for a number of different countries since 1900.
Since our respondents all used longer-term Treasuries astheir risk-free rate, the right-most column of Table VI mostclosely fits that choice. Even when respondents explicitlyreferenced the arithmetic or geometric mean of historicalretums, many rounded the figure or used other data to adjusttheir final choice. The net result is a wide array of choicesfor the market risk premium. For respondents who provideda numerical figure (Table II, Question 11), the average forcompanies was 6.49%, very close to the average of 6.6%from financial advisors. These averages mask considerablevariation in both groups. We had responses as low as 4%and as high as 9%. The 4% value is in line with the Ibbotson(2012) historical figures using the geometric mean spreadbetween stocks and long-term govemment bonds. The upperend of 9 percent comes from forward-looking estimatesdone in 2012 when US financial markets reflected a very lowinterest rate environment." We add a word of caution in howto interpret some of the differences we found in the marketrisk premium since the ultimate cost of capital calculationdepends on the joint choice of a risk premium, the risk-freerate and beta. We retum to this issue when we illustratepotential differences in the cost of capital.
As shown in Table VII, comments in our interviewsexemplify the diversity among survey participants. Thisvariety of practice displays the challenge of applicationsince theory calls for a forward-looking risk premium, onethat refiects current market sentiment and may change withmarket conditions. What is clear is that there is substantialvariation as practitioners try to operationalize the theoreticalcall for a market risk premium. And, as is clear in someof the respondent comments, volatility in markets hasmade the challenge even harder. Compared to our earlierstudy (1998) in which respondents almost always appliedhistorical averages, current practice shows a wider variationin approach and considerable judgment. This situationpoints the way for valuable future research on the marketrisk premium.
" Harris and Marston (2001, 2013) discuss evidence that the market riskpremium is inversely related to interest rates.
28 , JOURNAL OF APPLIED FINANCE-No. 1,2013
Table VII. Choice of the Market Risk Premium
"What do you use as your market risk premium?" A sampling of responses from our best practice companies shows the choice can be acomplicated one.
"We take an average of arithmetic mean and geometric mean for equity risk premium.""5.5% - 6.5%. Refiects a judgmental synthesis of various academic views and financial market perspectives.""Estimate a forward-looking risk premium. Use a methodology that incorporates S&P P/E ratio. It is a more volatile metric than using ahistorical risk premium. Use historicals as a sanity check.""6.5-7%. We ask several banks how we compare to comparable companies, including comparable size and industry.""Use the long-term (1926-2011) Ibbotson's S&P 500 premium over the risk-fi-ee rate."Comments from financial advisors also were revealing. While some simply responded that they use a published historical average, otherspresented a more complex picture."We used to apply the geometric mean, and then we switched to arithmetic. Now we use 4.6 - 6.6%, so we'll show for both ends ofthatdistribution. Based mainly on Ibbotson, but may add to that for emerging market companies.""Forward-looking estimate using dividend discount model...bank's proprietary model, which is forward-looking and uses S&P pricelevel, plus projections and payout to get an implied cost of equity"
IV. The Impact of Various Assumptions forUsing CAPM
To illustrate the effect of these various practices onestimated capital costs, we mechanically selected the twosample companies with the largest and smallest range ofbeta estimates in Table IV. We estimated the hypotheticalcost of equity and WACC for Target Corporation, whichhas the widest range in estimated betas, and for UPS whichhas the smallest range. Our estimates are "hypothetical" inthat we do not adopt any information supplied to us by thecompanies and financial advisors but rather apply a rangeof approaches based on publicly available information asof early 2013. Table VIII gives Target's estimated costs ofequity and WACCs under various combinations of risk-free rate, beta, and market risk premiums. Three clustersof possible practice are illustrated, each in turn using betasas provided by Bloomberg, Value Line, and Barra. The firstapproach, adopted by a number of our respondents, uses a10-year T-bond yield and a risk premium of 6.5% (roughlythe average response from both companies and financialadvisors). The second approach also uses a 6.5% riskpremium but moves to a 30-year rate to proxy the long-terminterest rate. The third method uses the ten-year Treasuryrate but a risk premium of 9%, consistent with what someadopters of a forward-looking risk premium applied. Werepeated these general procedures for UPS
The resulting ranges of estimated WACCs for the twofirms are as follows:
Target
UPS
MaximumWACC
8.58%
9.17%
iVIinimumWACC
4.58%
6.62%
Difference inBasis Points
400
255
The range from minimum to maximum is considerable forboth firms, and the economic impact potentially stunning. Toillustrate this, the present value of a level perpetual annual
stream of $10 million would range between $117 millionand $218 million for Target, and between $109 million and$151 million for UPS.
Given the positive yield curve in early 2013, the variationsin our illustration are explained by choices for all threeelements in applying the CAPM: the risk-free rate, beta, andthe equity market premium assumption. Moreover, we notethat use of a 10-year Treasury rate in these circumstancesleads to quite low cost of capital estimates if one sticks totraditional risk premium estimates often found in textbookswhich are in the range of 6%. From talking to our respondents,we sense that many are struggling with how to deal withcurrent market conditions that do not fit typical historicalnorms, a topic we discuss in more detail in Section VI.
V. Risk Adjustments to WACC
Finance theory is clear that the discount rate should riseand fall in concert with an investment's risk and that a firm'sWACC is an appropriate discount rate only for average-riskinvestments by the firm. High-risk, new ventures shouldface higher discount rates, while replacement and repairinvestments should face lower ones. Attracting capitalrequires that prospective retum increase with risk. Mostpractitioners accept this reasoning but face two problemswhen applying it. First, it is often not clear precisely how mucha given investment's risk differs from average, and second,even when this amount is known, it is still not obvious howlarge an increment should be added to, or subtracted from,the firm's WACC to determine the appropriate discount rate.
We probed the extent to which respondents alterdiscount rates to refiect risk differences in questions aboutvariations in project risk, strategic investments, terminalvalues, multidivisional companies, and synergies (Table II,Questions 13 and 17-20). Responses indicate that the greatpreponderance of financial advisors and text authors strongly
BROTHERSON ET AL. - " B E S T PRACTICES" IN ESTIMATING THE COST OF CAPITAL: A N UPDATE 29
Table VIII. Variations in Cost of Capitai (WACC) Estimates for Target CorporationUsing Different Metiiods of Implementing the Capital Asset Pricing iViodel
Data are as of January 22, 2013. The CAPM is used to estimate the cost of equity. Interest rates are from Bloomberg and the FederalReserve. The cost of debt is assumed to be 4.21% (average of Aaa and Baa yields). We approximated the capital structure and tax rateusing the market value of equity and the most recent financial statements, yielding debt as 32% of capital and a 34% tax rate.
1. Ten-year Treasury rate plus risk premium of 6.5%(Most common choice of proxy for risk-free rate and average risk premium used by respondents)Rf=1.92%, yield to maturity on 10-year US Treasury bondR -R =6.50%, average value for respondents
"" Cost of Equity Cost of CapitalBeta Service K WACCBloomberg (default), ß=.54 5.42% 4.58%Bloomberg (5-year, adj.), ß= 1.04 8.67% 6.81%Value Line, ß=.9O 7.76% 6.18%Barra, ß=.82 7.24% 5.83%
2. Thirty-year Treasury rate plus risk premium of 6.5%(Longer-term maturity choice favored by some analysts)R^=3.03%, yield to maturity on 30-year US Treasury bond
0, average value for respondents
Beta ServiceBloomberg (default), ß=.54Bloomberg (five year, adj.), ß=1.04Value Line, ß=.9OBarra, ß=.82
Cost of Equity
—£6.5V/09.79%8.88%8.36%
Cost of CapitalWACC5.34%7.57%6.94%6.59%
3. Ten-year Treasury rate plus risk premium of 9.0%(Consistent with some applications of a forward-looking risk premium estimate)R=L92%, yield to maturity on 10-year US Treasury bondR -R =9.0%, upper-end of risk premium estimates
Cost of Equity Cost of CapitalBeta Service K WACCBloomberg (default), ß=.54 6.71% 5.51%Bloomberg (five year, adj.), ß=l.04 11.27% 8.58%Value Line, ß=.9O 10.01% 7.72%Barra, ß=.82 9.29% 7.23%
favor varying the discount rate to capture differences in risk(Table II, Questions 13, 19, and 20). Corporations, on theother hand, are more evenly split, with a sizeable minorityelecting not to adjust discount rates for risk differencesamong individual projects (Table II, Questions 13 and 17).Comparing these results with our earlier study, it is worthnoting that while only about half of corporate respondentsadjust discount rates for risk, this figure is more thandouble the percentage reported in 1998. Despite continuinghesitance, companies are apparently becoming morecomfortable with explicit risk adjustments to discount rates.
A closer look at specific responses suggests thatrespondents' enthusiasm for risk-adjusting discount ratesdepends on the quality of the data available. Text authorslive in a largely data-free world and thus have no qualmsrecommending risk adjustments whenever appropriate.Financial advisors are a bit more constrained. They regularlyconfront real-world data, but their mission is often to valuecompanies or divisions where extensive market information
is available about rates and prices. Correspondingly, virtuallyall advisors questioned value multi-division businesses byparts when the divisions differed materially in size and risk,and over 90% are prepared to use separate division WACCsto reflect risk differences. Similarly, 82% of advisors valuemerger synergies and strategic opportunities separately fromoperating cash flows, and 73% are prepared to use differentdiscount rates when necessary on the various cash flows.
There is a long history of empirical research on howshareholder retums vary across firm size, leading someacademics to suggest that a small cap premium shouldbe added to the calculated cost of capital for such firms."Our study focuses on large public companies, so it is not
'• Banz (1981) and Renganum (1981) first identified that firms with smallermarket capitalization had earned higher average retums than stocksgenerally and higher than those predicted even if one adjusts for risk usingthe CAPM. Fama and French (1992) identify size as a critical factor inexplaining differences in returns. Pratt and Grabowski (2010) discuss sizeadjustments and the use of Ibbotson data on size groupings.
30 JOURNAL OF APPLIED FINANCE - NO. 1, 2013
Table IX. Adjusting Discount Rates for Risk
When asked whether they adjusted discount rates for project risk, companies provided a wide range of responses.
"Yes, but [the process] is fairly ad hoc. There's not a formal number.""We don't risk adjust, [but] we make sure we're on target by using sensitivity analysis and checking key assumptions.""For new deals, we add 20% to the cost of capital.""Yes, we make adjustments for different countries.""We make subjective adjustments [to our discount rate] for new technologies or products or customers, all of which carry more uncertainty.""We use a hurdle rate that is intentionally higher than our WACC. Recent WACC has been 9.64% and the hurdle rate is 15%, and then weadd qualitative considerations for the final decision.""Generally no further adjustments are made unless specifics are identified at the project level that would be very different risk. Forinstance, a contractual cash flow might be treated as a debt equivalent cash flow and given a lower discount rate."
surprising that firm responses do not reveal any such smallcap adjustments. In contrast, financial advisors work with awide spectrum of companies and are thus more likely to besensitive to the issue—as indeed they are. Among financialadvisors interviewed, 91% said they would at times increasethe discount rate when evaluating small companies. Ofthose who did make size adjustments, half mentioned usingIbbotson (2012) data which show differences in past retumsamong firms of different size. The adjustment process variedamong advisors, as the following illustrative quotes suggest."Adjustments are discretionary, but we tend to adjust forextreme size." "We have used a small cap premium, butwe don't have a set policy to make adjustments. It is fairlysubjective." "We apply a small cap premium only for micro-cap companies." "We use a small cap premium for $300million and below."
In important ways corporate executives face a morecomplex task than financial advisors or academics.They must routinely evaluate investments in intemalopportunities, and new products and technologies, for whichobjective, third party information is nonexistent. Moreover,they work in an administrative setting where decision rightsare widely dispersed, with headquarters defining proceduresand estimating discount rates, and various operating peoplethroughout the company analyzing different aspects of agiven project. As Table IX reveals, these complexities leadto a variety of creative approaches to dealing with risk.A number of respondents describe making discount rateadjustments to distinguish among divisional capital costs,intemational as opposed to domestic investments, andleased versus purchased assets. In other instances, however,respondents indicated they hold the discount rate constantand deal with risk in more qualitative ways, sometimes byaltering the project cash flows being discounted.
Why do corporations risk-adjust discount rates in somesettings and use different, often more ad hoc, approaches inothers? Our interpretation is that risk-adjusted discount ratesare more likely to be used when the analyst can establishrelatively objective financial market benchmarks for whatthe risk adjustment should be. At the division level, data
on comparable companies inform division cost of capitalestimates. Debt markets provide surrogates for the risksin leasing cash flows, and intemational financial marketsshed light on cross-country risk differences. When no suchbenchmarks exist, practitioners look to other more informalmethods for dealing with risk. In our view, then, practicaluse of risk-adjusted discount rates occurs when the analystcan find reliable market data indicating how comparable-risk cash flows are being valued by others.
The same pragmatic perspective was evident when weasked companies how frequently they re-estimated theircapital costs (Table II, Question 14). Even among firmsthat re-estimate costs frequently, there was reluctance toalter the underlying methodology employed or to revisethe way they use the number in decision making. Firmsalso appear sensitive to administrative costs, evidencingreluctance to make small adjustments but prepared to revisitthe numbers at any time in anticipation of major decisionsor in response to financial market upheavals. Benchmarkcompanies recognize a certain ambiguity in any cost numberand are willing to live with approximations. While the bondmarket reacts to minute basis point changes in interest rates,investments in real assets involve much less precision,due largely to greater uncertainty, decentralized decision-making, and time consuming decision processes. As noted inTable X, one respondent evidences an extreme tolerance forrough estimates in saying that the firm re-estimates capitalcosts every quarter, but has used 10% for a long time becauseit "seems to have been successful so far." Our interpretationis that the mixed responses to questions about risk adjustingand re-estimating discount rates reflect an often sophisticatedset of practical considerations. Chief among them are thesize of the risk differences among investments, the volumeand quality of information available from financial markets,and the realities of administrative costs and processes.When conditions warrant, practitioners routinely employrisk adjustments in project appraisal. Acquisitions, valuingdivisions and cross-border investments, and leasingdecisions were frequently cited examples. In contrast, whenconditions are not favorable, practitioners are more likely
BROTHERSON ET AL. - "BEST PRACTICES" IN ESTIMATING THE COST OF CAPITAL: A N UPDATE31
Table X. Re-estimating the Cost of Capital
"How frequently do you re-estimate your company's cost of capital? Here are responses from best practice companies.
"[We re-estimate] annually unless a fundamental event that is a game changer occurs, such as the banking crisis in 2008.""[We re-estimate] once a year around May since this is right before impairment testing.""We re-estimate twice a year and for special events such as a major acquisition.""We calculate the hurdle rate each year. We try to avoid any changes less than plus or minus 25 basis points.""Formally, we re-estimate our cost of capital every quarter, but generally we have used 10% for a long time, which seems to have been
successful so far."
Table XI. Judgments Related to Financial Market Conditions
Some of our best practice companies and advisors noted that their choice of a risk-free rate attempted to remove any unusual conditionsthey saw in current market yields. We asked "What do you use for a risk-free rate?" and heard the following:
"20-year US govemment bonds. If it seems to be unusually high or low, we may use a trailing average or other modified number.""Use a five-year rolling average in order to smooth out volatility associated with current/recent market trends.""Historically had used the spot rate yield to maturity on 10-year govemment bonds, but more recently have changed to thinking aboutfactors driving govemment interest rates, so we have moved away from the spot rate and toward the average yield over last 10 years."
to rely on cruder capital cost estimates and cope with riskdifferences by other means.
VI. Recent Institutional and MarketDevelopments
As discussed in the prior section, our interviews revealthat the practice of cost of capital estimation is shapedby forces that go beyond considerations found in usualacademic treatments of the topic. A feature that was morepronounced than in our prior study is the influence of a widearray of stakeholders. For instance, a number of companiesvoiced that any change in estimation methods would raisered flags with auditors looking for process consistency inkey items such as impairment estimates. Some advisorsmentioned similar concems, citing their work in venueswhere consistency and precedent were major considerations(e.g. faimess opinions, legal settings). Moreover, somecompanies noted that they "outsourced" substantial parts oftheir estimation to advisors or data providers. These itemsserve as a reminder that the art of cost of capital estimationand its use are part of a larger process of management—notsimply an application of finance theory.
The financial upheaval in 2008-2009 provided a naturaltest of respondents' commitment to existing cost ofcapital estimation methodologies and applications. Whenconfronted with a major extemal shock, did companies makewholesale changes or did they keep to existing practices?When we asked companies and advisors if financial marketconditions in 2008-2009 caused them to change the way theyestimate and use the cost of capital (Table II, Question 15),over three-fifths replied "No." In the main, then, there wasnot a wholesale change in methods. That said, a number ofrespondents noted discomfort with cost of capital estimationin recent years. Some singled out high volatility in markets.
Others pointed to the low interest rate environmentresulting from Federal Reserve policies to stimulate the USeconomy. Combining low interest rates and typical historicalrisk premiums created capital cost estimates that somepractitioners viewed as "too low." One company was sodistrustful of market signals that it placed an arbitrary eightpercent floor under any cost of capital estimate, noting that"since 2008, as rates have decreased so drastically, we don'tfeel that [the estimate] represents a long-term cost of capital.Now we don't report anything below 8% as a minimum [costof capital]."
Among the minority who did revise their estimationprocedures to cope with these market forces, one change wasto put more reliance on historical numbers when estimatinginterest rates as indicated in Table XI. This is in sharp contrastto both finance theory and what we found in our prior study.Such rejection of spot rates in favor of historical averagesor arbitrary numbers is inconsistent with the academic viewthat historical data do not accurately reflect current attitudesin competitive markets. The academic challenge today is tobetter articulate the extent to which the superiority of spotrates still applies when markets are highly volatile and whengovemments are aggressively attempting to lower ratesthrough such initiatives as quantitative easing.
Another change in estimation methods since our earlierstudy is reflected in the fact that more companies are usingforward-looking risk premiums as we reported earlier andillustrated in Table VIL Since the forward-looking premiumscited by our respondents were higher than historical riskpremiums, they mitigated or offset to some degree theimpact of low interest rates on estimated capital costs.'*
'* Pratt and Grabowski (2010) provide a discussion of estimating forwardlooking risk premiums. Harris and Marston (2001, 2013) discuss evidencethat the market risk premium is inversely related to interest rates.
32 JOURNAL OF APPLIED FINANCE - NO. 1, 2013
VII. Conclusions
Our research sought to identify the "best practice" incost of capital estimation through interviews with leadingcorporations and financial advisors. Given the huge annualexpenditure on capital projects and corporate acquisitionseach year, the wise selection of discount rates is of materialimportance to senior corporate managers.
Consistent with our 1998 study of the same topic, thisstirvey reveals broad acceptance of the WACC as the basisfor setting discount rates. In addition, the survey revealsgeneral alignment between the advice of popular textbooksand the practices of leading companies and corporateadvisors in many aspects of the estimation of WACC. Themain continuing area of notable disagreement is in thedetails of implementing the CAPM to estimate the cost ofequity. This paper outlines the varieties of practice in CAPMuse, the arguments in favor of different approaches, and thepractical implications of differing choices.
In summary, we believe that the following elementsrepresent "best current practice" in the estimation of WACC:
• Weights should be based on market-value mixes of debtand equity.
• The after-tax cost of debt should be estimated from mar-ginal pretax costs, combined with marginal tax rates.
• CAPM is currently the preferred model for estimatingthe cost of equity.
• Betas are drawn substantially from published sources.Where a number of statistical publishers disagree, bestpractice oñen involves judgment to estimate a beta.Moreover, practitioners often look to data on comparablecompanies to help benchmark an appropriate beta.
• Risk-free rate should match the tenor of the cash flowsbeing valued. For most capital projects and corporateacquisitions, the yield on the US govemment Treasurybond often or more years in maturity would be appropri-ate.
• Choice of an equity market risk premium is the subject ofconsiderable controversy both as to its value and methodof estimation. While the market risk premium averagesabout 6.5% across both our "best practice" companiesand financial advisors, the range of values starts from alow of around 4% and ends with a high of 9%.
• Monitoring for changes in WACC should be keyed tomajor investment opportunities or significant changes infinancial market rates, but should be done at least annual-ly. Actually flowing a change through a corporate systemof project appraisal and compensation targets must bedone gingerly and only when there are material changes.
• WACC should be risk-adjusted to reflect substantive dif-ferences among different businesses in a corporation. Forinstance, financial advisors generally find the corporateWACC to be inappropriate for valuing different parts ofa corporation. Given publicly traded companies in differ-ent businesses, such risk adjustment involves only mod-est revision in the WACC and CAPM approaches alreadyused. Corporations also cite the need to adjust capitalcosts across national boundaries. In situations wheremarket proxies for a particular type of risk class are notavailable, best practice involves finding other means toaccount for risk differences.
Best practice is largely consistent with finance theory.Despite broad agreement at the theoretical level, however,several problems in application can lead to wide divergencein estimated capital costs. Based on our results, we believethat two areas of practice cry out for further applied research.First, practitioners need additional tools for sharpening theirassessment of relative project and market risk. The variationin company specific beta estimates from different publishedsources can create substantial differences in capital costestimates. Moreover, the use of risk-adjusted discount ratesappears limited by lack of good market proxies for differentrisk profiles. We believe that appropriate use of comparable-risk, cross-industry or other risk categories deserves furtherexploration. Second, practitioners could benefit from furtherresearch on estimating the equity market risk premium.Current practice still relies primarily on averaging past dataoveroftenlengthy periods andyieldsawiderangeofestimates.Use of forward-looking valuation models to estimate theimplied market risk premium could be particularly helpfulto practitioners dealing with volatile markets. As the nextgeneration of theories sharpens our insights, we feel thatresearch attention to the implementation of existing theorycan make for real improvements in practice.
In fundamental ways, our conclusions echo those of ourstudy fifteen years ago. Our conversations with practitionersserve as a reminder that cost of capital estimation is part ofthe larger art of management—not simply an applicationof finance theory. There is an old saying that too often inbusiness we measure with a micrometer, mark with a pencil,and cut with an ax. Despite the many advances in financetheory, the particular "ax" available for estimating companycapital costs remains a blunt one. Best practice companiescan expect to estimate their weighted average cost of capitalwith an accuracy of no more than plus or minus 100 to150 basis points. This has important implications for howmanagers use the cost of capital in decision making. First,do not mistake capital budgeting for bond pricing. Despitethe tools available, effective capital appraisal continuesto require thorough knowledge of the business and wise
BROTHERSON ET AL. - "BEST PRACTICES" IN ESTIMATING THE COST OF CAPITAL: A N UPDATE 33
business judgment. Second, be careftil not to throw out thebaby with the bath water. Do not reject the cost of capital andattendant advances in financial management because your
finance people cannot provide a precise number. When inneed, even a blunt ax is better than nothing.*
References
Association for Finance Professionals, 2011, "Current Trends in Estimatingand Applying the Cost of Capital: Report of Survey Results," www.AFPonline.org, accessed March 2011.
Banz, R., 1981, "The Relationship between Retum and Market Value ofCommon Stock," Journal of Financial Economics 9 (No. 1), 3-18.
Barra, 2007, Barra Risk Model Handbook, MSCI Barra, www.Barra.com/support/library.
Brealey, R., S. Myers, and F. Allen, 2011, Principles of Corporate Finance,10th ed.. New York, NY, McGraw-Hill.
Brigham, E. and M. Ehrhardt, 2013, Financial Management: Theory andPractice, 14th ed.. Mason, OH, South-Western Publishing.
Bruner, R., K. Eades, R. Harris, and R. Higgins, 1998, "Best Practicesin Estimating the Cost of Capital: Survey and Synthesis," FinancialPractice and Education S (No. \), \3-2H.
Bums, R. and J. Walker, 2009, "Capital Budgeting Surveys: The Future is," Journal of Applied Finance 19 (No. 1-2), 78-90.
Carleton, WT. and J. Lakonishok, 1985, "Risk and Retum on Equity: TheUse and Misuse of Historical Estimates," Financial Analysts Journal41 (No. 1), 38-48.
Conroy, R. and R. Harris, 2011, "Estimating Capital Costs: PracticalImplementation of Theory's Insights" Capital Structure and FinancingDecisions: Theory and Practice (K. Baker and J. Martin, eds.). NewYork, NY John Wiley & Sons.
Dimson, E., P. Marsh, and M. Staunton, 2011a, "Equity Premia Around theWorld," London Business School Working Paper.
Dimson, E., P. Marsh, and M. Staunton, 2011b, The Dimson-Marsh-Staunton Global Investment Returns Database (the "DMS Database"),New York, NY Momingstar Inc.
Fernandez, P., J. Aguirreamalloa, and L. Corres, 2011, "US Market RiskPremium used in 2011 by Professors, Analysts and Companies: ASurvey with 5,731 Answers," IESE Business School Working Paper.
Graham, J. and L. Mills, 2008, "Using Tax Retum Data to SimulateCorporate Marginal Tax Rates," Journal of Accounting and Economics46 (No. 2-3), 366-380.
Harris, R. and F. Marston, 2001, "The Market Risk Premium: ExpectationalEstimates Using Analysts' Forecasts," Joumal of Applied Finance 11(No. 1), 6-16.
Harris, R. and F. Marston, 2013, "Changes in the Market Risk Premiumand the Cost of Capital: Implications for Practice," Journal of AppliedFinance 111 0^0. 1), 34-47.
Higgins, R.C., 2012, Analysis for Financial Management, 10th Ed., NewYork, NY McGraw-Hill.
Ibbotson, SBBI, 2012, Classic Yearbook, Market Results for Stocks, Bonds,Bills and Inflation 1926-2011, " Momingstar.
Jacobs, M.T. and A. Shivdasani, 2012, "Do You Know Your Cost ofCapital?" Harvard Business Review (July).
Koller, T., M. Goedhart, and D. Wessels, 2005, Valuation: Measuring andManaging the Value of Companies, 5th Ed., Hoboken, NJ, John Wiley& Sons, Inc.
Pratt, S. and R. Grabowski, 2010, Cost of Capital: Applications andExamples 4"' Ed., Hoboken, NJ, John Wiley and Sons.
Reinganum, M.R. 1981, "Misspecification of Capital Asset Pricing:Empirical Anomalies Based on Eamings' Yields and Market Values,"Journal of Financial Economics 9 (No. 1), 19-46.
Ross, S., R. WesterSeld, and J. Jaffe, 2013, Corporate Finance, 10th Ed.,New York, NY, McGraw-Hill.
Fama, E.F. and K.R. French, 1992, "The Cross-section of ExpectedRetums," Journal of Finance 47 (No. 2), 27-465.
Copyright of Journal of Applied Finance is the property of Financial Management Association and its content
may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express
written permission. However, users may print, download, or email articles for individual use.