1. No category

Why Do Corporations Give to Charity?
Author(s): Peter Navarro
Source: The Journal of Business, Vol. 61, No. 1 (Jan., 1988), pp. 65-93
Published by: The University of Chicago Press
Stable URL: http://www.jstor.org/stable/2352980
Accessed: 24/08/2008 18:08
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ucpress.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.
JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.
http://www.jstor.org
Peter Navarro
University of San Diego
Why Do Corporations Give
to Charity?*
Americancorporationscontributeover $3 billion
to charityeach year. These contributionsfinance
a broad range of activities, includinghealth and
welfare services, educationand research, culture
and the arts, and various civic activities.1
Why do corporations contribute? Previous
studies have attempted to answer this question
within the context of profit maximization and
managerialdiscretion.2However, at the theoretical level, none of these studies develops a formal
model that satisfactorily portrays the contributions process.3 At the same time, the empirical
efforts have been hamperedby the lack of firmspecific data. Perhapsas a result of this dataconstraint, previous empirical studies have relied
* I am indebted to Richard Carson, Jeffrey Dubin,
Malcolm Harris, Joseph Kalt, and especially Richard E.
Caves for many helpfulcommentsand criticisms.The paper
also benefittedgreatly from the comments of two excellent
refereesandfrompresentationsat the IndustrialOrganization
Seminars of Harvard University and the University of
California, San Diego. All errors and omissions are, of
course, my own.
1. Over 70% of corporate contributionsgo to education
and research (41%)and health and welfare (31%).The remainderis split fairly equally between arts and culture and
civic activities. See Troy (1984).
2. Johnson (1966), Schwartz (1968), Nelson (1970),
Whitehead(1976), Levy and Shatto (1978), Maddox (1981),
McElroyand Siegfried(1984, 1985),and Clotfelter(1985).
3. The most rigorousattemptis that of Maddox(1981).
(Journal of Business, 1988,vol. 61, no. 1)
? 1988by The Universityof Chicago.All rightsreserved.
0021-9398188/6101-0001$01.50
65
This paper explores
whether corporatecontributionsshould be tax
deductiblewithin the
more generalcontext
of an examination of
the profitand utility
maximizationmotives
drivingcontributions.
The theoreticalsection
develops a formal
structuralmodel of the
contributionsprocess,
illustratescomparative
statics, and derives a
set of empiricallytestable hypotheses. Using
a new source of firm
data, the empiricalresults indicatethat profit
maximization is an im-
portantmotive driving
contributions.This
findingsupportsthe
currenttax-deductible
status of contributions
(up to a seldom encounteredceiling) and
favors a reformthat allows firmsto treat contributionsas ordinary
business expenses.
66
Journal of Business
primarilyon IRS industrydata to focus their attentionon estimationof
the income and tax rate elasticities of contributions.4While this approach generates informationuseful for predictingfuture contribution
levels and measuring the sensitivity of giving to tax policy and the
business cycle, it has left unanswered the broader question of why
individualfirms contribute.
The answer to this question does, however, have importantimplications for federal tax policy as well as for the broaderideologicalstruggle over the legality and social desirability of contributionswithin
which the tax policy debate has occurred. On one side of the debate,
conservatives have arguedthat legalizing corporatecontributionshas
sanctioneda flagrantabuse of shareholderpropertyrights at the same
time that granting tax-deductibility status to contributions has encouraged an inefficient use of corporate and, by extension, societal
resources.5 Implicit in this critique of contributionsis, of course, the
assumptionthat firmsgive to charityfor reasons other thanprofitmaximization, such as to satisfy the goals of shirkingmanagersratherthan
those of shareholders.For if this assumptiondid not hold, both shareholders and the productiveforce of society would benefit.
On the other side of the debate, liberals have arguedthat corporations have a social responsibility to contribute that transcends any
obligation to shareholders.6Concerned more with social welfare and
income redistributionthan shareholderrights or efficiency losses, this
camp was first instrumentalin legalizing contributionsand, over the
years, it has vigorously supportedthe maintenanceand extension of
the tax-deductibilitystatus of contributions.7Like their conservative
counterparts,liberals also implicitlyassume that contributionsserve a
functionother thanprofitmaximization,but in this case it is an eleemosynary one.
Profitmaximizationis, however, a testable assumption,and it is the
purpose of this paper to make that test within the context of a more
general examination of corporate contributionmotives. The paper's
4. Maddox(1981)overcamethe lack of firmdataby conductingher own survey(with
John Siegfried)of 166firms.The most recent study, that of Clotfelter(1985),is devoted
almost entirelyto estimatingthese elasticitiesfrom industrydata (ch. 5).
5. Until as late as 1953,corporatecontributionsthatdid not yield a "directbenefit"to
firms(e.g., increasedprofits)were per se illegal. See, e.g., OldMissionPortlandCement
Co. v. Helvering,293 U.S. 289 (1934).The directbenefitdoctrinewas overturnedby a
New Jersey SupremeCourt decision and upheldby the U.S. SupremeCourtin A. P.
SmithManufacturing
Co. v. Barlowet al., 13 N.J. 145,98 A.2d 581 (1953).The Revenue
Act of 1936,? 23 (q), 74th Cong., 2d Sess., ch. 690, p. 1661,firstpermittedcorporations
to deduct contributions(presentlymadefor directbenefit)up to, but not exceeding, 5%
of this net income. A classic statementof the conservativeposition may be found in
Friedman(1962),pp. 132-36.
6. See, e.g., Andrews (1971), p. 120.
7. For a good history of the liberalizationof corporatecontributiontax policy, see
Fremont-Smith(1972).
67
Corporate Charity
majortheoreticalcontributionin Section I is to providea formalstructural model of the contributions process that treats the paradigms
of profit maximizationand managerialdiscretion as complementary
ratherthan competing. On the basis of a revised specificationof Oliver
Williamson'sclassic managerialdiscretionmodel (1963, 1964),the theory also arguesfor a reversalof the impliedpositive relationof tax rate
changes to tax-deductiblepreferredexpenditures(e.g., corporatecontributions).Section II presents an empiricalmodel that is based on the
theory, and Section III presents the results of estimatingthat model.
The implicationsof these results for tax policy and the legal status of
contributionsare examined in a concludingsection.
I.
The Theoretical Model
A.
Profit Maximization
To consider the profit motive, it is useful to distinguishbetween revenue enhancement,cost reduction, and tax considerations.
1. Demand considerations. On the revenue side, profit-maximizing
managersmay use corporatecontributionstacticallyas partof an overall advertisingstrategydesignedto promotea productdirectlyor, alternatively, to promote the firm's image.8One purpose of such advertising, which provides an additionalgoodwill attributefavored by some
customers, is to enhance revenues by shifting, or decreasingthe elasticity of, the demandcurve for the firm's product. To formalizethese
demand side effects, assume that managerschoose a level of product
output, Q, and contributions,G, so as to maximize
,a = PQ(P,G) - C[Q(P,G)] - G,
(1)
where arequals profits; Q is a function of G and the price, P; and
productioncost, C, is a function of Q. The fifst-orderconditions may
be written as
dP = dC
ac
p +P+QaQ
(2)
and
p
aQ
~aG
-
ac aQ +1.
aQ aG
(3)
8. An example of contributionsas productpromotionis that of the Brown and Williamson Tobacco Company's sponsorshipof the Newport Jazz Festival, renamedthe
Kool Jazz Festival after one of its brands.Well-knownexamples of image advertising
include Texaco's sponsorship of the MetropolitanOpera broadcasts and AT & T's
underwritingof the public affairsprogram,the MacNeil-Lehrer News Hour.
Journal of Business
68
In (2), we have the standardmarginalconditionfor the output decision while (3) similarlyindicates that the firmwill increase G up to the
point where the marginalrevenue of giving equals the marginalcost, in
this case its price (by assumption, equal to 1) plus the change in costs
broughtaboutby a change in Q. Defining qGand nrpas the elasticitiesof
demand in giving and in price, respectively, (3) can be arrangedto
obtain the familiar Dorfman-Steinerrule for advertising adapted to
contributions:9
G =
S
G..
(4)
11P
Thus, the firm'sgiving-to-salesratio (GIS) shouldrise with the elasticity of demandwith respect to contributionadvertisingand fall with the
price elasticity of demand.
Cost considerations. On the cost side, profit-maximizingmanagers may use corporatecontributionsto reduce labor, capital, operating, or regulatoryand governmentalcosts as well as to sow the seedsand later reap the benefits-of product or technologicalinnovation.
In the labormarket,for instance, workersmay be willingto work for
lower wages in communities that provide better recreational,educational, cultural,and health-relatedfacilities. If the costs to the firmsof
financingsuch facilities are more than offset by the wage reductions,
profitsare increased.'0At the same time, a firm'scharitablesupportof
education can increase the long-run supply of labor (e.g., chemical
engineers) and thereby lower the wage the firm must offer. Similar
considerations can lower capital and other operating costs. For example, a firmoperatingin a communitywith a lower crimerate will pay
less for insuranceand securitybecause of the reducedrisks of theft and
vandalism.
In the political arena, corporate contributionsmay likewise be a
useful tactic in an overalllobbyingstrategydesignedto create or maintain a favorablebusiness climate. Withinthis context, the cultivationof
a good public image can help insulatethe firmfrom unfavorabletax or
regulatorypolicies (e.g., stringentenvironmentalregulations),or it can
help the firm gain access to favorable laws or regulations, such as a
relaxation of zoning restrictions or a tax abatement. The charitable
supportof basic or appliedresearch likewise can lead to new product
or technological developments that lower costs.
To formalizethese cost side effects, the model in equation(1) can be
expanded such that cost is a function not only of Q but also a set of
communityor environmentalattributes,E. (Below, E is treatedwithin
2.
9. Dorfmanand Steiner (1954).
10. See, e.g., Clotfelter(1985), p. 188.
Corporate Charity
69
the context of the labor market, but the discussion can be generalized
to capital, operating,and regulatorycosts or technologicaleffects.) In
this formulation, costs fall as E rises. That is, C = C[Q(P,G); E], and
aC/aE < 0.
Assume that employees take into considerationas a condition for
work not only the pecuniary compensation that employers offer but
also an index of environmentalattributes, E, that characterizetheir
prospectiveresidentialcommunity.Underthis assumption,an employee's potential consumption bundle consists of an array of marketed
goods, such as autos, food, clothing, housing, entertainment,and medical care, purchasedwith pecuniarycompensation.It also includes an
arrayof nonmarketedgoods-"quasi-fringe benefits" such as clean
air, safety from crime, and access to adequatehousing, medical care,
and culturalopportunitiesthat can only be "purchased"by residingin
the community. In any given community, these nonmarketedgoods
can be summarizedby E.
To the extent that an employee puts a positive value on E, he or she
should be willing to accept lower compensationto work in a community that is characterizedby a higherlevel of E. This "laborattraction"
dynamicis analogousto CharlesTiebout's (1956)well-knownobservation that citizens "vote with theirfeet," choosing communitiesand tax
burdensthat match their preferences for a level of governmentgoods
and services. In this case, workers choose communities that match
their preferencesfor environmentalattributes(tradedoff againstpecuniary compensation).
In relatingE to G, a productionfunctionfor environmentalattributes
can be posited as
E = E(G, F, V),
(5)
where G represents the financing of contributionsby the individual
firm,F representsfinancingby other firmsand privateindividuals,and
V representsfinancingby taxpayers. In this formulation,F implies the
free rider problem while V indicates the possible substitutionof private-sectorcontributionsfor public-sectorsocial welfareexpenditures.
While measures of F and V will be incorporatedinto the empirical
model, for simplicity(andwithoutloss of generality)they are treatedas
exogenous here, and the firm's objective function can be rewritten:
H = PQ(P,G) - C[Q(P,G); E(G,F0,V0)] - G.
(6)
From manipulationof the first-orderconditions, cost side effects can
be incorporatedinto this modifiedDorfman-Steinerrule:
G _G
CP<l
S
(7)
+ CEEG)
Since aC/dE < 0 and aElaG > 0 by assumption, (7) implies that firms
70
Journal of Business
will choose a larger giving-to-sales ratio when cost side motives are
added to demand side motives (by comparing[4] and [7]). l
3. Tax considerations. To complete the discussion of profit motives, the effect of a proportionaltax rate, t, on the level of G can be
examined. For simplicity, let PQ(P,G) equal total revenues, R, and
maximize:
=(1-
t)(R - C-G).
(8)
From the first-orderconditions, it can easily be shown that the imposition of a proportionaltax will have no effect on the level of giving;
that is, under profit maximization,aG/at = 0. 12
B.
Utility Maximization and Managerial Discretion
The traditionalmanagerialdiscretion model (Williamson1963, 1964)
rests on the assumptionthat inadequateshareholderpolicing, born of
the separation of ownership and control in the modern corporation,
provides utility-maximizingmanagers with the opportunityto shirk
their responsibilityto maximize the value of the firm. Instead, managers divert "discretionaryprofits"-that surplus above some minimumprofitdemandedby shareholders-to financethe consumptionof
various preferredexpenditures:unnecessarilyluxurious office suites,
excess staff, lavish expense accounts, salaries above those necessary
to retain managers,and the like. Such consumptionenhances managerial utility by fulfillingthe drive for status, power, security, and prestige (Williamson 1964, pp. 29-32). The traditionalmodel further assumes (1) that the extent of managerialshirkingrises with the degree of
separationof ownership and control and a collateralreductionin the
ability of shareholdersto overcome organizationalobstacles stemming
from a diffusion of interests, imperfect information,and transactions
costs (Monsen, Chiu, and Cooley 1968;Nyman and Silberstan 1978);
and (2) that there are imperfectcapital markets.
This section assumptioncan be relaxed, however, if managerialdiscretion is viewed within the full generalityof the principal-agentproblem (Jensen and Meckling 1976; Shavell 1979). In particular,wellinformed owners bargainingto Pareto optimalitywith managersmay
11. For the statementto be true, it mustbe that 1 + CEEG > 0. In general,this will be
true if price exceeds cost in the productmarket,a conditionseen by rewritingthe firstorder conditions (P - CQ)QG = 1 + CEEG. Thus, the result should hold in all but
perfectlycompetitivemarkets.
12. It is importantto note that this, and all otherresults regardingthe tax rate in this
paper,hingeon the assumptionof the full deductibilityof contributions.This assumption
mirrorsthe present tax code up to a ceiling that virtuallyall corporationstoday never
reach. In addition,Clotfelterhas shown that in a multiperiodframeworkin which the
benefits of contributionsare treated as a depreciatingasset, tax policy will affect the
timing,but not the level, of contributions(1985, pp. 189-90).
71
CorporateCharity
allow the managers to trade off preferred expenditure consumption
against pecuniary compensation. (Such a possibility is of empirical
interest below because expectations about the sign of the relationbetween executive compensationand contributionswill be dependenton
the presence or absence of such bargaining.)
With this background, as previous studies have noted, corporate
contributionsmay fit into the managerialdiscretion framework.'3In
particular,the level of contributionsabove that necessary to maximize
profitscan be viewed as a preferredexpendituredesigned to boost the
manager'sprestige in the communityor, more altruistically,simply as
a way to generate the warm glow of the "performanceof office for the
benefit of society" (Williamson1964, p. 31). In this characterization,
then, the profit motive may be regarded as being nested within the
utility motive. That is, some level of giving may be consistent with
profitmaximization,but we can allow for the possibilitythat managers
may choose to contribute at a level above the profit optimumto enhance their utility.
To formalizethis, assume that managerschoose a level of giving, G,
a level of other preferred expenditures, X, and output, Q, so as to
maximizeutility, U, subject to a shareholder-capitalmarketconstraint
that nests the profitmotive within the utility motive underthe assumptions of nonsatiationand the positive marginalutilities of G andX. This
constraint may be written such that discretionaryprofits, HD, must
equal reportedafter-taxprofits, HR, minus the minimumlevel of earnings, HO,necessary to preventeither a changein managementby shareholders or a corporate takeover by outside buyers in the marketfor
corporatecontrol.14Thus, managersmaximize
U[G,X]; G,X,Q,
subject to
HD = HR
0,
(9)
where
HR =
(1
-
t)[R(Q,G)
-
C[Q(P,G); E(G,F0,V0)] - G
-
X].
Note that (9) differs in a small but significantway from the classic
managerialdiscretion model (Williamson1963, 1964;Clotfelter 1985).
In that model, HD is commonly included as an additionalargumentin
13. The list includes Williamson(1964), Jensen and Meckling(1976), and Clotfelter
(1985).
14. In general, we expect that Howill be related to corporatecontributionsin the
followingway. The negativeeffect of contributionson reportedprofitsmust not reduce
the value of stockholders' equity by more than the cost of organizinga successful
stockholderchallengeto managementor by morethanthe cost of organizinga buyoutby
buyers in the marketfor corporatecontrol, whicheveris smaller.
72
Journal of Business
the managerialutility function (i.e., U = U[G,X,HID]).However, HDis
excluded from (9) on the groundsthat it is merely a means to finance
preferredexpenditureconsumptionratherthan a utility-generatingend
in itself.'5 Furthermore,rational behavior coupled with nonsatiation
requires that HD will always be driven to zero; that is, any additional
dollar of HD will always be expended on additionalunits of G or X.
The inclusionversus exclusion of HD fromthe utilityfunctionleads to
quite differentimplicationsfor tax policy. If HD is included,an increase
in the tax rate generates two effects. First, there is a negative income
effect, because after-tax profits and therefore managerialincome to
finance G and X are reduced. Second, there is a positive substitution
effect because the price of HD, which is taxable, changesrelativeto the
other argumentsin the utility function, G and X, which are tax deductible. 16
As will be demonstratedbelow, in the classic expense preference
model, this positive substitutioneffect offsets a negativeincome effect
and leads to the implicationaG/at > 0; that is, an increase in the tax
rate stimulatesan increase in preferredexpenditureconsumption,such
as contributions.If, however, HD is excluded from the utilityfunction,
there is only a negativeincome effect, and aG/at< 0. Thus, an increase
in the tax rate would reduce preferredexpenditurecontributions.
To illustratethis result and, more generally,the theoreticalimplications of (9), the following Lagrangeancan be solved:
L = U[G,X]
+ X [H`0- (1 - t)[R(Q,G) - C[Q(P,G); E(G,F0,V0)] - G -X]
(10)
where X is a Lagrangeanmultiplier.The first-orderconditions are
adL=HIl0- (1 - t)(R - Cax
aL
aG
aR
aC aQ
aQ aE
G-X)=O?
(11)
aC aE
aE aG
(12)
ax
-=
U2+(1
- t) = 0,
(13)
15. It may be arguedthat some formof profitsdoes in fact belongin the utilityfunction
if managersderive satisfactionor statusfromthem. This may be a validargument,but it
is not the one set forthin defense of [ID'Sexclusionin the utilityfunction.Nor wouldthe
same implicationsfollow, since it is not [D that is indicated,but some measureof total
profits.
16. If unspent,HD is subjectto the corporatetax while G andX arefullytax deductible.
Thus, a changein the tax ratechangesthe cost of HD relativeto G andX, a pointcrucialto
understandingthe tax implicationsof HD's placementin the utilityfunction.
CorporateCharity
73
and
-(1
-
aQ
-) O (
c
_
aQ
)
TQ_
0,
(14)
where U1 and U2 representthe first partialsof the utilityfunctionwith
respect to the first and second arguments,respectively.
Since Xmeasuresthe change in utility for a changein the constraint,
and since the constraintis binding,Xis less than0. At the same
aLlaHlo,
time, (1 - t) is positive by assumption.Hence, (14) can be rewrittenas
aR
dR
aQ
_ac
=
C
V
aQ ,'15
(15)
which is equivalent to (2). Accordingly, utility-maximizingmanagers
choose the firm's output in the same way as profit-maximizingmanagers. By a similarargument,(12) can be rewrittenas
U1 = X(1 - t)(
aG
aC aE
aE aG
ac aQ
aQ aG
iI/
(16)
This indicates that utility-maximizingmanagerswill want to choose a
level of G that is higher than strict profit maximizationrequires. To
prove this, recall from (3) that profit-maximizingmanagersequate the
marginalrevenue of G to marginalcost. However, since U1and (1 - t)
are positive and X is negative, equation (16) implies
aR < ac aQ + aC aE +
aG
aE +G
aQ aG
(17)
Thus, utility-maximizingmanagerschoose G such that marginalrevenue is less than marginalcost. Under diminishingmarginalutility of
giving (i.e., U1 < 0), G must thereforebe higherunderutilitymaximization. While this result provides little empiricalleverage, it is bf interest for its potential social welfare implications;specifically, in the absence of adequate contributionpolicing and imperfect competition,
utility-maximizingmanagerswith a taste for contributionswill runtheir
firms inefficiently.
To examine comparativestatics, equations (11)-(14) are first totally
differentiated:
(1 - t)dG + (1
-(1 - t)ZldX + [Ull
-
XZ2]dG + U12 dX
-
t) dX = -Z6dt - dflo,
XZ3dQ= XZldt,
(1 - t)dX + U21dG + U22dX = Xdt,and
- XZ4dG - XZ5 dQ = 0,
(18)
Journal of Business
74
where
Zl=
-
__
aG
aQ aG
aC aE__ I < 0
-
aiz aG
by (17), or
G
-1G
-p[l + CEEG(F, VA)]
=S
by (6) and (7);
Z
__2R
(-
2R
aQ
a2Q aC
OGQ
aG
aG2 aQ
+
LG2
a2E\1
aE2 kaG!J'
a2E aC _ a2C
_
0G2 aE
Z3 = (1-
Z4Z== (1
Z5 = (
2R - a2C aQ~
d-G]
[
)
aQ-
0ra
- t)[a2
2R
dQ2
a2C aQ 1
a2R
a9Q
+
Q2 aG 1'
aQaG
aG
J-a2R_a2c
t)Q2
a2C dQ 2
aQ2 k aG v
a
1
aQ2J'
and
Z6 = (R - C-
G-X).
After rendering(18) in matrixform, Cramer'srule can be appliedto
solve for aGIaHlo
(holding t constant):
aG
= -XZ5[-(U12 + U22Z1)(1 -
t)]
(19)
This is the "income effect" because it measuresthe response of giving
to a change in the constraint.Intuitively,we would expect this income
effect to be negative because, as H`0rises, the funds availablefor preferred expenditurebehavior fall, ceteris paribus. On inspection, both
the denominatorand - XZ5are negative by the definitionof a negative
definite matrix. Thus, for aG/alloto be negative, the term in brackets
must be negative. In the bracketedterm, U22is negative and (1 - t) is
positive by assumption while Z1 is negative by (17). Hence, the
numeratorwill be positive if U12is greaterthan or equal to zero. The
rate of change of U1 for a change in X is U12,and vice versa; it will
always be zero in the case of an additive utility function. Likewise,
for a multiplicativeutility function, it should generally be positive.'7
17. On this point, see Williamson(1964, p. 44).
75
CorporateCharity
Therefore, the numeratorof (19) is generallypositive, and the income
effect is negative.
Similarly,we can solve for aG/at(holdingHOconstant):
-XZ54-(U12 + U22Z1)(1 - t)Z6]
aG
at ri0=Hi
(20)
A11
By similarreasoningand because Z6 is positive, this derivativeis likewise negative. That is, giving falls as the tax rate increases. Substituting (19) into (20) illustrateswhy this is so; namely, an increase in the
tax rate results in a pure negative income effect:
aG
at |n
-
-
HO=HO
= 6aGl
arll t
< 0.
(21)
The results of (19)-(21) can be contrastedwith those obtainedfrom
the classic expense preferencemodel. IncludingHD in the utility function, managersmaximize
U[G,X,HD],
subject to
HD =
(1 -
t)(R - C-G-X)
0.
-Ho
(22)
The classic approachto solving this problemis to treatthe constraint
as an equality and to regardit as redundantbecause it takes the same
form as the last argumentin the utility function.'8 The problem then
reduces to one of straightforwardmaximizationof the utility function.19Using a proceduresimilarto that above then yields the following
and aG/at:20
expressions for aGIaHlo
aG
an t|
duo t= t
=
{U3Z5[Ul3U22 -
UO3U22(1 -
t) + U22U33(1 - t)ZI
(23)
-
U12U23
U23(1 -
t)ZI
+ U12U23(1
-
t)]}/IJI,
18. This assumes that second-orderconditionsare satisfiedand that cornersolutions
are disallowed(Williamson1963,p. 1036).
19. This approachsuggests that perhapsthe implicationsof including[D in the utility
functionwere not fully recognized.Indeed,the treatmentof the constraintas an equality
is one of the best argumentsfor exclusion of HD in the utilityfunction,because it clearly
implies that HD equals 0, as does the total utility derivedfrom its consumption.
20. In particular,the first-orderconditionsaretotallydifferentiated,andCramer'srule
is appliedto solve for aG/alloand aG/at.For the complete derivationand an extended
discussion, the interestedreadermay refer to Williamson(1964, pp. 66-71).
Journal. of Business
76
and
= {U3Z54U13U22- U13U23(1 - t) + U22U33(1 - t)Z
aG i
at no=Ho
-
U2U
-
-
U3(
OZ
U3Z23(1 - t)Z
+ {U3Z5[U3U22ZI -
U3U12-
+ U12U33(1
U3U13(1 -
-
0]Z6}/IJI
(24)
+ U3U12
0)]}1JI
where IJI
is the relevant determinant.
Under the standardassumptions of the classic model, the income
effect represented by (23) is, like its counterpartin (19), negative.2'
Collectingterms in Z6 in turnallows us to rewrite(24)as the sum of this
negativeincome effect and a positive substitutioneffect where the term
to the left of the plus sign is the negative income effect and the termto
the right is the positive substitutioneffect:22
aG
aTdTr0ii
+ {U3Z51U3U22Z1
+
-
U3U23(
=
-
aG
Z6-tt
6aH0
t= i
t)Z + U3U12
~~~~~~~~IJ
-
U3U13(1
-
t)]}
(25)
In the classic model, this substitutioneffect is normally assumed to
dominatethe income effect so that aG/at> 0.
The conflictingresults of (21), where aG/at< 0, and (25), where aG/at
> 0, have two important empirical implications and one important
implicationfor tax policy. The first empiricalimplicationis that these
results provide an empirical means of discriminatingbetween profit
maximizationand managerialdiscretion motives. Specifically, if the
parameterestimate associated with the federaltax rate in the empirical
model below is not significantlydifferentfrom zero, the hypothesis of
managerialdiscretion can be rejected. The second empiricalimplication is that a negative sign on the parameterestimate associated with
the tax rate would support exclusion of HD in the managerialutility
function, whereas a positive sign would support the classic expense
preference model (i.e., inclusion of HD).
Empiricalresolution of this issue, in turn, has an importanttax policy implication.If the sign is negative/positive,an increase in the tax
rate will lead utility-maximizingmanagersto decrease/increasethe corporate contributionsof their firms.
21. The assumptionsinclude U1> 0, Uji< 0, and U1j= Ofor i $j. In addition,both IJI
and U3Z5 are negative by the definitionof a negativedefinitematrix.The most critical
assumptionhere is that the Us = 0, which impliesthat the utilityfunctionis additivein
its components.For a discussion, see Williamson(1964, pp. 70-71).
22. That the substitutioneffect is positive follows from the same set of assumptions
identifiedin n. 21 above.
CorporateChariy
77
II. The EmpiricalModel
The theoretical conditions established above suggest that the optimal
level of contributions, G*, will be functionally related to the set of
theoretical variables
G
= fhG,
IP, CEEG,
F, V, Ho, X, t),
(26)
where 9G and n9 reflect demandside advertisingmotives, CEEG measures cost side labor attractionmotives, F indicatesthe free riderproblem, V measuresthe substitutabilityof private-sectorcontributionsfor
public-sector welfare expenditure, Horeflects the shareholder-capital
marketconstraint,X represents other preferredexpenditures,and t is
the federal tax rate. G* will be related positively to IG and CEEG and
negatively to F, and V by equations (6) and (7). Likewise, G* will be
relatednegativelyto Hoby (19) and, if expense preferencebehavioris in
evidence, negatively to t by (21) or positively to t by (25).
The following empiricalmodel is designed to test these theoretical
relations:
GIS = Bo + B1ADVER + B2PCM + B3LINTENSE + B4FREERIDE,
+ B5GOVT + B6MANAGE + B7DERATIO
(27)
+ B8DIVCHANGE + B9MANSAL,
and
+ BOFEDTXR + B11TITHE + Ai,
where [ is an errorterm, sales, S, is a normalizingvariablesuggested
by equation (7), and G is derived from an importantnew source of
previously unanalyzed, firm-specificcontributionsdata, namely, the
American Council for the Arts (ACA) Guide to Corporate Giving.23
The ACA is the leadingprivate nationalorganizationthat serves the
arts. The ACA guide has been published to date in three separate
volumes-1978, 1981, and 1983. Each volume reportsthe results of a
written survey (with telephone follow-up) sent to a large sample of
corporations, with primaryfocus on the 1,000 largest industrialsas
listed in the Fortune Double 500 Directory. The sample used herein
consists of a pooled cross section of 249 firmsdrawnfrom these three
volumes of the ACA guide. The firms are all listed on the Compustat
data base and representthe sample of firmsfrom the guide for which a
complete set of regressors could be obtained. The observations span
23. Other possible normalizingvariablesfor contributionsinclude assets and value
added. However, the use of sales in the denominatorof the dependentvariableis well
supportedby the theoretical model as well as the empiricalresults: when the log of
contributionswas regressedon the log of sales and the other variablesin the empirical
model, the impliedrestrictionthatB = 1for the parameterestimateassociatedwith sales
could not be rejected. (The point estimate was 1.08.)
78
Journal of Business
the years 1976-82, with no firm representedin more than one year.
Immediately below, the regressors in (27) and their relation to the
theoretical variables in (26) are explained while a discussion of two
samplingproblems associated with the ACA guide data is postponed
until the next section.
A.
The Demand Side
1. The giving elasticity (NOG).The firm's advertisingexpenses (normalized by sales), ADVER, measures the firm'spropensityto rely on
advertisingas a marketingtool. As a measureof IG, it may be viewed
as an instrumentfor the underlyingattributesof the firm's product(s)
that determine this propensity.24These firm-specificattributes(e.g.,
product heterogeneity and durability)that cause some firmsto advertise more than others should also cause the same firms to contribute
more than others. Thus, GISshouldrise with ADVER.25Implicitin this
formulationis that contributionsand advertisingexpendituresmay be
jointly determined.Hence, we also shall want to test for the endogeneity of ADVER below.
2.
The price elasticity of demand (u9p). Following Dorfman and
Steiner (1954),Schmalensee(1972),and others, the inverse of the price
elasticity of demand, lp, can be easily shown to equal the firm's
"price-cost margin," that is, output price minus marginalcost as a
proportionof output price:26
1
P - (aClaQ)
(28)
Thus, GIS should rise with PCM.
However, statistical confirmationof this relation is not sufficient
evidence of goodwill advertisingmotives. The problemis one of multiple inference: a positive relation between giving and the price-cost
marginis also consistent with a "rule of thumb" method of determining contributionlevels, a method given some weight in survey stud24. For a discussion of the various attributesthat determinea firm's propensityto
advertise, see Caves and Williamson(1985).
25. There is a collateralissue here that has arisen in the literatureas to whethera
positive/negativesign of the parameterestimateassociatedwith the advertisingvariable
may be interpretedas evidence that advertisingexpenses and contributionsare complements/substitutes.While Whiteheadarguesfor this interpretation(1976,p. 65), Maddox
correctlypoints out that the sign can shod no light on the matter(1981,pp. 35-36). The
reasonis that the concepts of gross substitutabilityand complementarityhave to do with
the cross-priceelasticity, which the parameterestimate sheds no light on.
26. The result, which is also the usual conditionfor monopolypricing,may be easily
derivedfromeq. (3). In the empiricalmodel,PCMis representedby the ratioof revenues
less costs to revenues, which follows from multiplyingthe rightside of (28) by Q/Q.
79
CorporateCharity
ies.27 Accordingly,we must comparethe findingsbelow regardingPCM
to those for the other measure of advertisingmotives (i.e., ADVER).
B.
The Cost Side
1. Labor motives (CEEG). Because many of the benefits of corporate contributions are likely to accrue to labor rather than capital,
one importantfirm attributebearing on the size of aC/aEand the labor attraction motive should be labor intensity. In particular, GIS
should rise as the percentage of a firm's costs attributableto labor,
LINTENSE, rises. However, any given firm's contributionsmay be
targeted to areas outside the corporate headquartersand plant communities (e.g., nationally). Because giving outside these employeeinhabitedcommunitiesis likely to have little to do with labor motives,,
our measure of labor intensity can be refinedby multiplyingit by the
percentage of contributionsgiven in the headquartersand plant communities.
2.
The free rider problem (F).
The possibility that employees from
other firms operatingin the donor firm's communitymay "free ride"
off the donor firm's contribution-financingof community benefits is
incorporatedin the variable FREERIDE. This variablemeasures the
numberof other firms (with more than 20 employees) operatingin the
firm's SMSA; the expected sign is negative.
3. Governmentspending (V). The distributionof observations in
the contributionsdata sample across the years 1976-82 provides a test
of the substitutabilityof private sector contributionsfor federalbudget
outlays on social welfare needs and, more generally, of the effect of
governmentspending on contributions.Duringthat period, aggregate
statistics indicate that corporatecontributionschanged a great deal in
response to the changingrole of the federal government.Specifically,
contributionsrose significantlyin 1981and 1982as the Reaganadministration dramaticallycut the federal budget.28One possible political
interpretationof this rise portraysit merely as a response to the wellpublicized "jawboning"of the corporatesector by the Reaganadministration.However, from an economic standpoint,it is also true that a
decrease in federal spendingon welfare needs shouldincreaseboth the
firm's marginalefficiency of corporate giving as well as the marginal
advertising benefits gained from contributions. Thus, from both the
cost and demand sides, the observed substitutabilityof private for
public sector dollars is not surprising.To incorporatethis factor into
27. See, e.g., Maddox(1981, ch. 3).
28. See the Councilon Foundations'CorporatePhilanthropy(1982),p. 140.
80
Journal of Business
the model, an indicatorvariable,GOVT, is set equalto 1 if an observation is from 1981 or 1982, and zero otherwise.29The expected sign is
positive.
C.
Managerial Discretion
1.
The shareholder-capital market constraint (HO). While the best
measureof Howould be the returnon equityrequiredto keep shareholders from removing managementor to keep raidersfrom organizinga
takeover, as a practicalmatter,it is not possible to determinethis rate
directly. Three indirect measures of Hoare, however, proposed.
First, in previous studies, Monsen, Chiu, and Cooley (1968),
Chevalier(1969), and Nyman and Silberstan(1978)have examinedthe
extent of separation of ownership and control (and by implication,
the size of llo)within the context of a two-regimeclassificatorysystem
that separates firms into "managerially controlled" and "ownercontrolled" on the basis of factors such as stockholderdiffusion,composition of the board of directors, and other informationpertinentto
control.30In developingone measurefor Ho,this classificatoryapproach
is adopted here: an indicatorvariable, MANAGE, is set equal to 1 if
the firm is manageriallycontrolled, and zero otherwise. This determination is made on the basis of the percent of stock held by the largest
investor (with families treated as a unit), the numberof managerson
the boardof directors, and other relevantevidence of control found in
10K reports, proxy statements, and annualreports.31If contributions
are, in fact, a discretionarypreferredexpenditure,the coefficient on
MANAGE should be positive.
A second measureof Hois suggestedby the literatureon agency costs
and optimal capital structure. In particular,it is well known that the
existence of tax subsidieson interestpaymentsprovidesfirmmanagers
with an opportunityto increase the value of the firm by increasing
leverage.32Because, however, the probabilityof bankruptcyrises with
the degree of leverage (and attendant fixed payment obligations),
29. Indicatorvariableswere also tested for individualyears and indicatedthat the
groupingfor GOVTwas appropriate.
30. In this regard,it has been observedthat the holdingof a majorityof sharesby one
individual(or group)is not necessaryfor effective corporatecontrol.Rather,controlcan
be achieved when the dominantstock interest is well below 50%(Monsen, Chiu, and
Cooley 1968,p. 437).
31. For a good discussion of the requirementsof corporatecontrol, see Herman
(1982).
32. Modiglianiand Miller(1963).In searchof an optimalcapitalstructure,subsequent
studies have focused on limits to debt financing,such as bankruptcycosts, e.g., Kraus
and Litzenberger(1973);creditrationing,e.g., Jaffee (1971);differentialpersonaltaxes,
e.g., Miller (1977);growth opportunities,e.g., Myers (1977);imperfector incomplete
capitalmarkets,e.g., Robichekand Myers (1966);and the uncertaintyof the tax benefit
stream, e.g., Brennanand Schwartz(1978).
Corporate Charity
81
Donaldson (1963) first argued that managersmay, in the interests of
job security and stabilizationof personal wealth, seek to avoid high
leveraging. Jensen and Meckling (1976) subsequentlyinterpretedthis
suboptimaluse of leverage as an agency cost imposed on shareholders
by utility-maximizingmanagers.This line of argumentsuggeststhatthe
firm'sdebt-equityratio, DERATIO,may be used as a second measure
of Hlo.33The expected sign is negative;that is, GISshouldfall as leverage increases.
A finalindirectmeasureof Horelates to the firm'sdividendpolicy. In
general, we expect that an increase/decreasein the annual dividend
per-sharewill be associated with a loosening/tighteningof the shareholder constraint.Accordingly,GIS shouldrise with an increase in the
dividend from the previous year, DIVCHANGE, and fall with a decrease in the dividend;in other words, the expected sign is positive.
2. Other preferred expenditures (X). The direct measurement of
other preferred expenditures, X, is difficult because some fraction
of most (if not all) preferred expenditures are legitimate, profitmaximizing activities. Thus, it is difficult to determine the relevant
utility-maximizing"residual," for example, how many squarefeet of
an office, how many staff people, how much of an expense account,
and so on, represent non-profit-maximizingbehavior. At the same
time, managerswho indulgein preferredexpenditurebehaviorhave no
incentive to collect data and/or report such behavior. Indeed, they
have every incentive to hide it, both from shareholdersand empiricists
tryingto detect it. Not surprisingly,then, the most commonmeasureof
X in the literatureis the one preferredexpenditurefor which excellent
data is kept, namely, executive compensation.34This measure is
adopted here with the caveat that it neither measures the residual
salarycomponentindicativeof discretionarybehaviornor capturesthe
full range of preferredexpenditureconsumption.Nonetheless, contributions are expected to vary systematicallyin manageriallycontrolled
firmswith the remunerationof the firm'schief executive officer, SALARY, if contributionsare a preferredexpenditure.To isolate and measure this effect, an interactiveterm, MANSAL, is createdby multiplying MANAGE times SALARY. The expected sign on MANSAL is
negative if managersbargainwith well-informedowners to trade off
33. DERATIO,as it is calculatedon the Compustatdatabase, equals [(totalassets total equity + preferredstock)/(totalequity - preferredstock)] * 100. In preliminary
analyses, the equity-to-salesratio (EIS) was also used interchangeablywith DERATIO
as a measureof capitalstructureandperformedvery similarlyto DERATIO.Because of
this similarityand because inclusion of EIS created collinearityproblems,only the results for DERATIOare reported.
34. Those adoptingthis measureincludeWilliamson(1963, 1964),Nelson (1970),and
Whitehead(1976).
82
Journal of Business
salary for contributionconsumption. Otherwise, the expected sign is
positive.
3.
The tax rate (t).
By equations (21) and (25), GIS should either fall
or rise, respectively, as the firm's average federal corporatetax rate,
FEDTXR, rises if utility maximizationmotives are present, and, by
equation(8), GIS.should be insensitive to FEDTXRif only profitmaximization motives exist.
D.
Other Factors
In additionto the variablesabove suggested by the theoreticalmodel,
one additionalexogeneous variable, TITHE, has been includedin the
empirical model to reflect a unique and potentially powerful positive
influenceon contributionlevels across firms.In particular,in a number
of majorU.S. cities, strong organizedpressures to contributeare exerted on the firmand its managersin the form of "tithingclubs." Cities
that have such clubs include Baltimore,Birmingham,Denver, Duluth,
Jacksonville, Kansas City (Missouri), Louisville, Minneapolis, Norfolk, Oakland,Pine Bluff (Arkansas),Phoenix, Rochester, San Francisco, Seattle, Tuscaloosa, and St. Cloud.35Table 1 presents the average contributions-to-sales ratio (GIS) for firms in tithing versus
nontithingcities as well as by U.S. Census Bureauregions. Note that
the average ratio is twice as high for firms in cities with tithing clubs
versus cities without tithing clubs.
Firm membershipin tithingclubs entails pledginga certainpercentage of income to charity (typically, 2%-5%). If a firm fails to join a
tithing club in its city, it risks negative publicity and an attendant
negative reactionfrom its customers. At the same time, firmmanagers
risk diminutionof status and prestige amongtheirpeers and withinthe
broadercommunity.To capturethese pressures, an indicatorvariable
TITHE is set equal to one if the firm is headquarteredin a city with
tithing clubs, and zero otherwise. The expected sign is positive.
E.
Summary
Table 2 summarizesthe regressors, states the expected parameterestimate signs suggestedby the developmentabove, and indicatesthe data
sources (which are explained in more detail in the Appendix).
III.
A.
Statistical Considerations and Estimation
Sampling Problems
The ACA guide contributionssurvey data is characterizedby two sampling problems that call for the use of more specialized estimation
techniques than ordinaryleast squares (OLS).
35. Vivian Stark, Dayton Hudson Corporation,personal communication,August
1985.
CorporateCharity
TABLE 1
83
Average Giving-to-Sales Ratio by Firms in Tithing Versus Nontithing
Cities
Giving-to-Sales
Number of
Ratio (GIS)
Firms
Firms in cities:
Tithing
Nontithing
.001258
.000632
25
224
Total
...
249
Regions:
New England
MiddleAtlantic
South Atlantic
East North Central
East South Central
West North Central
West South Central
Mountain
Pacific
Total
.000777
.000770
.000721
.000550
.000754
.001089
.000308
.000495
.000744
...
34
59
17
78
4
20
14
1
22
249
First, firmssurveyed by the ACA that did not contribute,or contributed very little, generallydid not respondto the survey request.36This
implies a truncated distributionof observations in the data set and,
collaterally, both biased and inconsistent estimatorsif OLS is used.37
The technique of truncatedregressionwas used to address this problem. We found, however, that this techniqueyielded virtuallyidentical
parametersestimates as OLS. This suggests that truncationis not a
problem.38Accordingly, we only report OLS estimates below for
which heteroscedastic-consistent (White 1980) covariance matrices
may be more readily computed.
Second, the ACA survey results are characterizedby differential
response rates across four distinctstratagroupedby size. In particular,
of the Fortune 1000firms (by sales) respondingto the survey, the top
200 responded at a 62%rate, firms 201-300 at a 47%rate, firms301500 at a 29%rate, and firms501-1000 at a 19%rate.39As a result, the
sample employed in the estimation is not random across strata and
36. American Council for the Arts, Guide to Corporate Giving (1983), 3:563.
37. Amemiya(1985, p. 367).
38. This conclusion is furtheredbolsteredby the observationthat only a very small
percentage of firms in the Fortune 1000 (roughly5%) do not contribute(Councilon
Foundations,CorporatePhilanthropy[1982],p. 104). For discussion of the relationof
OLS to truncationestimators,see Green (1983).
39. AmericanCouncilfor the Arts, Guideto CorporateGiving(1983),3:563.Response
rate clearly falls by size. One majorreason, cited by the ACA guide (3:563)is fewer
resourcesto accommodatethe survey requestin smallerfirms.Thereis nothingin this
patternto suggest, however, that firmswithineach stratathat contributedo not respond
in a nonrandomway.
Journal of Business
L.
cd
Z
4,-
-cd
od
0
0
0
u
u
u
u
k)
u
0
c>?
C)
0
u u
tw
cd
V)
O
+J
4)
0
4.4
4-4
cd
..*
-4
O (:) O I
64
+
+
w
(R)
>
cd
C.
>
0
-4
4-4
-4
r.
4-4
0
4)
N
-4
- E -4M.
i-4
00
64
0
A4-4
9 =
cld
cd
>
*j
=3 0
6.,
Cd
4-4
r-4 r-4
w
-4
-4
C's
cd
W
64
C) +j
*j
j -.4
11-1=O
+
4
cld to
42'
-4
W
c
0
4-4
-4
cd
4-
4-4
r-4
O
0
U
4
-4
x
.0
C's
SW
0
cd
El
cid
W
>
(L)
c)
Z
0
42'
4-i
'O
.;'4 -.4
0
fA
0 O
- 64
-4
.,::$
,.j
r. 4-4
Cd 0
0
0
00
;..
;
i.-k 4)
-1::
;>
1.4
o
03
=1
v
C) O
M .0
U -4
c
-O
C's
-
17:1 eq
C's
00
>
64
14
cd
0
0
4.4
4)
;.4 0 4- 4)
x
O'..4-i O U0
!O
W 4) =I (I
4--
CIS
03
El 64
0
0
*.A64
00
0
1010.4
00
4-4
c
IZ
0
L:L. 0 0
cd C)
-Mc
CZ
El 0
.4 (=)
04
1-4
0
c
0 OX
X-
u
-6,
0C)4
El
0
m
Ei
z
Cd
>
-4
CZ
a
4C.
Corporate Charity
TABLE 3
85
Results of the OLS and WLS Models
Regressors
ADVER
PCM
LINTENSE
FREERIDE
GOVT
DERATIO
DIVCHANGE
MANAGE
MANSAL
FEDTXR
TITHE
INTERCEPT
Adjusted R2
OLS
Restricted
OLS
WLS
.025802
(2.11)
3.48380
(3.03)
.009165
(2.84)
-.000037
(2.29)
.211598
(1.84)
-.000927
(1.67)
.374644
(1.76)
.020972
(.12)
-.000037
(.13)
-.354570
(1.01)
.569101
(3.53)
-3.48043
(16.77)
.557086
(3.58)
-3.53427
(19.08)
.034293
(2.75)
.484420
(.40)
.008963
(2.51)
-.000031
(1.52)
.187853
(1.60)
-.001276
(1.79)
.525538
(1.72)
.155938
(1.03)
-.000523
(1.91)
-.620574
(1.68)
.642394
(4.29)
-3.21343
(12.73)
.223
.229
.164
.025527
(2.08)
3.31368
(2.90)
.008849
(2.75)
-.000036
(2.28)
.223757
(2.01)
-.000928
(1.66)
.396373
(1.83)
...
...
...
NOTE.-Numbersin parenthesesare t-statistics.
therefore not representativeof the desired population.While there is
some controversy in the literatureon this point, Klein (1953) has argued that it is necessary to weight stratifiednonrandomsamples to
avoid biased estimates.40Following Klein, observationsin each strata
were weightedby the inverse of the strata'sresponse rate so thatfirms
in the strataswith the lower response rates received greaterweight.41
As is evident from table 3, such weightingdoes not affect the qualitative results regardingprofit maximizationmotives but there is an important difference (discussed below) regardingutility maximization
motives.
B.
The Results
Table 3 reports the results of estimating equation (27) for the unweighted (OLS) and weighted least squares (WLS) samples. The
dependentvariableis the log of GIS. To furthertest utility maximiza40. For a strongopposingview on the need to weight stratifiedrandomsamples, see
DuMoucheland Duncan (1983).
41. This techniqueis used underthe assumptionthat, of the firmsthatcontributed,the
response was randomwithin strata. See discussion, n. 39 above.
86
Journal of Business
tion motives, the table also presents a restrictedversion of the unrestricted OLS model in which three managerialdiscretion variables
with insignificantparameterestimates-MANAGE, MANSAL, and
FEDTXR-are droppedfrom the model for the purpose of constructing an F-test. All t-statistics are based on heteroscedastic consistent
covariance matrices suggested by White (1980).
The explanatorypower of the models, as measuredby adjustedR2s,
is moderatelyhigh for cross-sectionaldata. These R2s rangefrom .223
and .229 for the unrestrictedand restrictedOLS models, respectively,
to .164 for the unrestrictedWLS model. This latterresult suggests, not
surprisingly,that there is slightly more variabilityin the contributions
behavior (if not in the motives) of the relatively smallerfirms, which
are given greaterweight in the WLS estimations.42
A comparison of parameter estimates across the models likewise
indicates considerable stability with respect to both size and significance levels for variables measuringprofit maximizationmotives, but
there is less stability among several of the variablesmeasuringutility
maximizationmotives (MANAGE, MANSAL, and FEDTXR). Closer
inspection of these estimates suggests that both demandand cost side
profit motives influence the corporate contributionsdecision, but the
evidence is less clear cut in supportof managerialdiscretionmotives.
Withrespect to demandside considerations,the parameterestimates
associated with PCM are correctly signed and, in the OLS models,
statistically significantabove the .01 level, suggestingthe presence of
advertisingmotives by the modifiedDorfman-Steinerrule. This conclusion is further bolstered by the results for ADVER, which yield
correctly signed parameterestimates significantabove the .05 level
across models. At the same time, a Hausmantest allows us to weakly
reject the hypothesis of the endogeneity of ADVER, a findingconsistent with its use in the regression as an instrumentfor underlying
unobserved structuralcharacteristics.43
With respect to cost side considerations, the estimated parameters
for LINTENSE are positive and significantat the .02 level or above,
providingstrongsupportfor the labormarkethypothesis. The parameter estimates associated with FREERIDElikewise are correctlysigned
and significant above the .05 level in the OLS models and weakly
significantin the WLS model, suggesting the presence of free rider
42. The similarityof the parameterestimates for the OLS and WLS models suggest
similarmotives.
43. The Hausmantest is based on Hausman(1978). The test involves estimatinga
separateequationwith ADVER as the dependentvariable.As suggestedby Albionand
Farris(1981, p. 130), the regressorsincludeda consumernondurableindicatorvariable
and measures of market size, regionalityof markets, marketconcentration,and frequency of purchasealongwith contributions.The predictionof ADVER,ADVERHAT,
was used as an additionalregressorin eq. (23). The parameterestimateassociatedwith
ADVERHATwas very weakly significant(above the .20 level) and suggestedonly very
weak endogeneity.
Corporate Charity
87
problems.In a similarpattern,the parameterestimatesassociated with
GOVT are all correctly signed but exhibit slightly largersize and significancelevels in the OLS models than in the WLS model. One possible interpretationof these differences for FREERIDE and GOVT in
the unweighted versus weighted (inversely by size) samples is that
largerfirmsare both more prone to free riderproblemsand, being more
visible, more likely to take up the slack and substituteincreasedcontributions for falling federal budget dollars.
Turningto utility-maximizingmotives, the parameterestimatesassociated with both DERATIO and DIVCHANGE are correctly signed
and consistently significantabove the .10 level across models. This
suggests, following Donaldson (1963), that more highly levered firms
systematicallycontributeat lower levels, while an increase in the dividend is also likely to indicate an increase in contributionsthrougha
loosening of the shareholdersconstraint. The remainingevidence is,
however, less supportiveof the managerialdiscretiontheory.
For example, the parameterestimates associated with FEDTXR are
both negatively signed, suggestingthe presence of utility maximization
motives (and exclusion of rD in the managerialutility function), but
only the estimate in the WLS model is even moderately significant
(above the .10 level).
A similarpatternfor the OLS versus WLS models prevails for the
parameterestimates associated with MANSAL and MANAGE. The
parameterestimates associated with MANSAL are both negatively
signed, suggestinga managerialbargainingtradeoffbetween salaryand
contributions,but only the estimatein the WLS model is above the .10
significancelevel. Even less robust, for MANAGE, the parameterestimates are correctly signed but insignificantin the OLS model and only
very weakly significantin the WLS model.
Given this pattern of evidence for MANAGE, MANSAL, and
FEDTXR, it would be imprudentto make any strongclaims that utility
maximizationmotives are an importantfactor in the contributiondecision. However, given the apparentdifferencesin the results for these
variablesin the OLS and WLS models, it is at least possible to conjecture that managerialdiscretion behavior regardingcontributionsperhaps may be more prevalent in smaller firms. Such a conjectureworthy as a topic for future research-is at least indirectlysupported
by the results of two F-tests testing thejoint significanceof MANAGE,
MANSAL, and FEDTXR in the OLS and WLS models.44On the basis
of these tests, we cannot reject the hypothesis (above the .01 level) that
the parameterestimates associated with MANAGE, MANSAL, and
FEDTXR are jointly zero in the OLS model, but this hypothesis is
rejected in the WLS model, at least above the .10 level.
44. This test is describedin Pindyckand Rubinfeld(1981, pp. 117-19).
Journal of Business
88
Finally, the positive signs and high significancelevels of TITHE's
parameterestimates indicatethat the presence of tithingclub pressures
is a strong predictorof contributions.
In summary,the level of charitablecontributionsappearsto rise with
ADVER, PCM, LINTENSE, GOVT, DIVCHANGE,and TITHEand
fall with FREERIDE and DERATIO, while MANAGE, MANSAL,
and FEDTXRappearto weakly influencecontributions,but only in the
WLS model.
C. Robustness of Results
To assess the robustness of the results above to several common data
and estimation problems, the regression model was subjectedto several additionaltests.
First, the conditionindex methodof Belsey, Kuh, and Welsch (1980)
was used to examine the data sample for ill-conditioning,with no
significantproblems indicated. Second, an examinationof the partialregression leverage plots and studentized residuals indicated a small
numberof significantoutliers in the data sample. The Tukey biweight
method of iteratively weighted least squareswas used to address this
problem.The methodsystematicallydownweightsthe influenceof outliers by assigning small weights to observations with large residual
value (Mosteller and Tukey 1977, pp. 356-65). The results, not reportedbelow because of space constraints,yielded very similarparameter estimates and significancelevels, thereforeindicatingstrongresistance of the results to the influenceof outliers.
Finally, Box-Cox estimationof the regressionmodel was employed
to test the appropriatenessof the logarithmicfunctional form used
above. This procedure (Box and Cox 1964)estimates the appropriate
power transformationcoefficient, X,for the dependentvariable,with X
= 1 implyinga linearform and X = 0 implyinga logarithmicform. The
Box-Cox results (also not reportedfor space constraints)once again
yielded parameter estimates and significance levels very similar to
those above and led to clear rejectionof a linear specificationfor GIS.
D.
Comparison of Results to Previous Studies
Before comparing the empirical results above to those of previous
studies, it is useful to remindthe readerthat, with the exception of the
Maddox study, previous studies have been restrictedto industrydata
while this study employs firmdata.45At the same time, this is a crosssectional study while roughly half of the previously reported results
involve time-series data.
45. Clotfelterprovidesan in-depthreview of the previousstudies(1985,pp. 193-207).
The studies of McElroy and Siegfried (1984, 1985) are not discussed in this section
because they are largely derivativeof Maddox's doctoral dissertation,with Siegfried
advising(Maddoxis McElroy's maidenname).
Corporate Charity
89
Schwartz, Whitehead,Levy and Shatto, and Maddoxall finda positive and significantrelation between charitablecontributionsand advertising, lending supportto the conclusion here that contributionsare
a form of advertising.However, Bennett and Johnsonfound advertising significantin only two of their 16 reported regressions and only
when a specific variable (i.e., unionization)was in the model.
Nelson findsa positive and significantrelationbetween contributions
and labor intensity, as does this study, but Whitehead estimates a
similarmodel that yields an insignificantrelation. In contrast, Whitehead finds an insignificantrelation between contributionsand executive compensationmotives, but Nelson reports a significantrelation.
Clotfelter,Levy and Shatto, Nelson, and Schwartzall finda positive
and significantrelationbetween contributionsand the tax rate in their
time-seriesanalyses. This findingappearsto contradictboth the theory
presented here and the reported results. However, Clotfelter has
shown that in a dynamic context (e.g., time series), profit-maximizing
firmswill change the timingof, as opposed to the total level of, contributions in response to an anticipatedchange over time in the tax rate,
with relative contributionlevels higher in periods of higher taxation
(e.g., an excess profitstax in wartime).46This findingis consistentwith
the theory presented here, the findings of this cross-sectional study,
and the findingsof previous time-series studies.
Maddox reports results similarto this study for FREERIDEwhile
every previous study includes some measureof income in theirregressions whose parameterestimates are regularlyfound to be positive and
highly significant.These findingson income supportthe results found
for PCM, which is highly correlatedwith income.
Levy and Shatto as well as Nelson find a positive and significant
relation between contributionsand total dividends while Whitehead
finds a weakly significantpositive relation between contributionsand
the change in dividends.
In summary, this comparison of results yields little evidence that
sharplycontradictsthe hypotheses tested in this study.47
IV.
A.
Summary and Policy Implications
Summary
This paper has examined the motives behind corporatecontributions
within the context of profit maximizationand managerialdiscretion.
46. See n. 12 above.
47. The hypothesisthatcontributionsmay be a formof nonpricecompetitionhas been
tested and rejectedin the literatureby Johnson(1966), Whitehead(1976),and Bennett
and Johnson(1980).This hypothesishas likewise been rejectedin preliminaryanalyses
alongwith the hypothesisthat contributionssystematicallydifferbetweenown-firmand
foundationvehicles. Because the variablestesting these hypotheseswere unrelatedto
the theoreticalmodel, they were not includedin the empiricalmodel above.
Journal of Business
90
The theoretical model has illustrated that the profit motive may be
nested within the managerialdiscretion motive. On the basis of a revised specificationof the classic managerialdiscretionmodel, the theory also has argued for a reversal of the hitherto postulatedpositive
relationof changes in the federal tax rate to changes in tax deductible
preferredexpenditures.
An empirical test of the model that is based on an importantnew
source of firm data indicates that profit maximizationis an important
motive drivingcontributions.The followinghypotheses are supported:
(1) corporatecontributionsrepresenta form of advertising,(2) contribution-financed environmental attributes represent a quasi-fringe
benefit to firmemployees, and (3) the free riderproblemreduces contributionlevels. The substitutabilityof private sector contributionsfor
public sector welfare expendituresis also supported.
In contrast, the evidence that contributionsare a utility-maximizing
managerialpreferredexpenditureis less clear-cut.The empiricalanalysis supportsthe hypotheses that contributionsare (1) negativelyrelated
to the amount of debt versus equity in the firm's capital structureand
(2) positively related to increases in dividends. However, we found no
relation in the unweighted sample between contributions,on the one
hand, and the federal tax rate, executive compensation,and the degree
of managerialcontrol, on the other hand, and only a weak relation
when the samplewas weighted inversely by size. Finally, the presence
of tithing club pressures is a strong predictorof higherfirm contributions.
B.
Policy Implications and Future Research
From the revenue side of the budget, the empiricalresult that profit
maximizationis an importantcontributionsmotive supportsthe legality of contributionsas well as the currentfull deductibilityof contributions (up to a seldom encountered ceiling). The evidence would also
favor a reform that removed that ceiling and allowed firms to treat
contributionsas ordinarybusiness expenses.
From the expenditureside, the albeit weaker evidence that privatesector contributionsmay substitutefor public-sectorwelfare expenditures appears to lend some supportto the argumentthat federal intervention is unnecessaryto meet the nation'swelfareneeds. But ultimate
resolution of this private-versus public-sectordebate must hinge on a
worthy goal of future research, namely, estimating the elasticity of
substitutionbetween contributionsand governmentexpenditures. At
the same time, future studies may find it worthwhileto study further
the effect of firm size (and, collaterally, mergers) on contributions
within the context of managerialdiscretionas well as to examine why
tithingclubs that appearto significantlyaffect contributionlevels exist
in some cities, such as Minneapolis,and not in others.
CorporateCharity
91
Appendix
Data Set Construction
Data on G, the years for GOVT, and informationon home and plantgivingto
construct LINTENSE were taken from the American Council for the Arts
Guide to CorporateGiving(1983, vols. 1-3). The variablesS, ADVER, PCM,
LINTENSE, DERATIO, DIVCHANGE, and FEDTXR were obtainedfrom
Standard& Poor's Compustatdatabase while informationto constructMANAGE, SALARY, and MANSAL was takenfrom Dun and Bradstreet'sMillion
Dollar Directory (1980), "1OK"reportsto the Securitiesand Exchange Commission, proxy statements, and Compustat.
Values for TITHEand FREERIDEwere assignedto each firmin the sample
on the basis of location of the firm's corporateheadquartersin a Standard
Metropolitan Statistical Area (SMSA) or major city if not in an SMSA.
FREERIDE was obtained from the Bureau of the Census 1977 Census of
Manufactures. Informationto construct TITHE was obtained from Vivian
Stark at the Dayton Hudson Corporation.
As a final data comment, two additionalsteps were taken to conserve degrees of freedom. First, industrydata were used to fill in missingobservations
for LINTENSE.48Second, mean values were substitutedfor missingobservations for FEDTXR.49
References
Albion, MarkS., and Farris, Paul W. 1981. TheAdvertisingControversy.Boston: Auburn.
Amemiya,Takeshi. 1985.AdvancedEconometrics.Cambridge,Mass: HarvardUniversity Press.
AmericanCouncilfor the Arts. 1983.Guideto CorporateGiving.Vols. 1-3. New York:
ACA.
Andrews, Kenneth R. 1971. The Concept of Corporate Strategy. Homewood, Ill.: Dow
Jones-Irwin.
Belsey, DavidA.; Kuh, Edwin;andWelsch, Roy E. 1980.RegressionDiagnostics. New
York: Wiley.
Bennett, JamesT., and Johnson,ManuelH. 1980.Corporatecontributions:Some additional considerations.Public Choice 35 (June):137-43.
Journalof theRoyal StatistiBox, G., and Cox, D. 1964.An analysisof transformations.
cal Society, ser. B, 26:211-43.
Brennan,M. J., and Schwartz,E. S. 1978. Corporateincome taxes, valuation,and the
problemof optimalcapital structure.Journalof Business 51 (January):116-29.
Caves, RichardE., and Williamson,Peter J. 1985. Whatis productdifferentiation,really? Journal of Industrial Economics 34, no. 2 (December): 113-32.
Chevalier,Jean-Marie.1969. The problemof control in large Americancorporations.
Antitrust Bulletin 14:163-80.
Clotfelter, Charles T. 1985. Federal Tax Policy and Charitable Giving. Chicago: Univer-
sity of ChicagoPress.
48. An indicatorvariable,LABDUM, was used in preliminaryanalysisto determine
whetherthe firmdata was significantlydifferentfrom the industrydata. It was set equal
to 1 if firmdataand0 if industrydata.No significantdifferencewas foundandLABDUM
was droppedfrom subsequentregressions.
49. On the validity of this procedure,see Pindyckand Rubinfeld(1981, pp. 245-52).
Use of means increased robustness of model as would be expected but did not
significantlychange estimates of the coefficients.
92
Journal of Business
Councilon Foundations.1982. CorporatePhilanthropy.Washington,D.C.: Councilon
Foundations.
Donaldson,G. 1963. Financialgoals: Managementvs. stockholders.HarvardBusiness
Review 41 (May-June): 116-29.
Dorfman,Robert, and Steiner, Peter 0. 1954.Optimaladvertisingand optimalquality.
American Economic Review 44 (December): 826-36.
DuMouchel,WilliamH., and Duncan, Greg J. 1983. Using sample survey weights in
multiple regression analysis of stratified samples. Journal of the American Statistical
Association 78 (September):535-43.
Dun and Bradstreet.1980.MillionDollar Directory.New York: Dun & Bradstreet.
Fremont-Smith, Marion R. 1972. Philanthropy and the Business Corporation. New
York: Russell Sage Foundation.
Friedman, Milton. 1962. Capitalism and Freedom. Chicago: University of Chicago
Press.
Green, W. H. 1983. Estimationof limiteddependentvariablemodels by ordinaryleast
squares. Journal of Econometrics 21 (February): 196-212.
Hausman,JerryA. 1978.Specificationtests in econometrics.Econometrica46 (November): 1251-71.
Herman, Edward S. 1982. Corporate Control, Corporate Power. New York: Cambridge
UniversityPress.
Jaffee, D. M. 1971. Credit Rationing and the Commercial Loan Market. New York:
Wiley.
Jensen, Michael C., and Meckling, WilliamH. 1976. Theory of the firm:Managerial
behavior, agency costs, and ownership structure.Journal of Financial Accounting
3:305-60.
Johnson, Orace E. 1966. Corporatephilanthropy:An analysis of corporatecontributions. Journal of Business 39 (October): 489-504.
Klein, L. R. 1953.A Textbookof Econometrics.Evanston, Ill.: Row, Peterson.
Kraus, A., and Litzenberger,R. H. 1973.A state-preferencemodel of optimalfinancial
leverage. Journalof Finance 28 (September):911-22.
Levy, FerdinandK., and Shatto, GloriaM. 1978.The evaluationof corporatecontributions. Public Choice 33 (March):19-28.
McElroy, KatherineMaddox, and Siegfried,John J. 1984. The effect of firm size and
mergerson corporatephilanthropy.In Betty Bock, Harvey J. Goldschmid,Ira M.
Millstein, and F. M. Scherer (eds.), The Impact of the Modern Corporation. New
York: ColumbiaUniversityPress.
McElroy, KatherineMaddox, and Siegfried,John J. 1985. The effect of firm size on
corporate philanthropy. Quarterly Review of Economics and Business 25, no. 2 (Sum-
mer): 18-26.
Maddox, KatherineE. 1981. CorporatePhilanthropy.Ph.D. dissertation,Vanderbilt
University.
Miller, M. H. 1977.Debt and taxes. Journalof Finance 32 (May):261-75.
Modigliani,F., and Miller, M. H. 1963. Taxes and the cost of capital:A correction.
American Economic Review 53 (June): 433-43.
Monsen, R. Joseph;Chiu,John S.; and Cooley, David E. 1968.The effect of separation
of ownershipand control on the performanceof the large firm. QuarterlyJournalof
Economics 82 (August):435-51.
Mosteller,Federic, and Tukey, JohnW. 1977.Data Analysis and Regression.Reading,
Mass.: Addison-Wesley.
Myers, Stewart C. 1977. Determinantsof corporateborrowing.Journal of Financial
Economics 5 (November):147-75.
Nelson, Ralph L. 1970. Economic Factors in the Growth of Corporation Giving. New
York: National Bureauof Economic Researchand the Russell Sage Foundation.
Nyman, Steve, and Silberstan,Aubrey. 1978. The ownershipand control of industry.
Oxford Economic Papers 30 (March): 74-101.
Pindyck,RobertS., and Rubinfeld,Daniel L. 1981.EconometricModels and Economic
Forecasts. 2d ed. New York: McGraw-Hill.
Robichek, A. A., and Myers, S. C. 1966. Problemsin the theory of optimal capital
structure. Journal of Financial and Quantitative Analysis 1 (March): 1-35.
Corporate Charity
93
Schmalensee, Richard. 1972. The Economics of Advertising. New York: American
Elsevier.
Schwartz,RobertA. 1968.Corporatephilanthropiccontributions.Journalof Finance23
(June):479-97.
Shavell, Steven. 1979.Risk sharingandincentivesin the principalandagentrelationship.
Bell Journal of Economics 10 (Spring): 55-73.
Tiebout, Charles M. 1956. A pure theory of local expenditures.Journal of Political
Economy64 (October):416-24.
Troy, Kathryn. 1984. Annual Survey of Corporate Contributions. New York: Confer-
ence Board.
White,Halbert.1980.A heteroskedasticity-consistent
covariancematrixestimatoranda
direct test for heteroskedasticity.Econometrica48 (May):817-38.
Whitehead,Paul. 1976.Some economic aspects of corporategiving. Ph.D. dissertation,
VirginiaPolytechnicUniversity.
Williamson,Oliver E. 1963. Managerialdiscretion and business behavior. American
Economic Review 53 (December): 1032-57.
Williamson, Oliver E. 1964. The Economics of Discretionary Behavior: Managerial Objectives in a Theory of the Firm. Englewood Cliffs, N.J.: Prentice-Hall.