Contractual Arrangements as Signaling Devices: Evidence from Franchising
Author(s): Francine Lafontaine
Source: Journal of Law, Economics, & Organization, Vol. 9, No. 2 (Oct., 1993), pp. 256-289
Published by: Oxford University Press
Stable URL: http://www.jstor.org/stable/764853
Accessed: 18/01/2009 18:00
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=oup.
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 [email protected].
Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to Journal of Law,
Economics, & Organization.
http://www.jstor.org
256
JLEOV9 N2
ContractualArrangementsas SignalingDevices:
Evidencefrom Franchising
Francine Lafontaine
Universityof Michigan
1. Introduction
In the presence of asymmetricinformation,the notion that firms may use a
variety of instrumentsto provide relevant informationabout themselves or
theirproductis now well establishedin industrialorganization.For example,
pricing, advertising,and warrantydecisions have all been suggestedas means
by which a firm can signal the quality of its product.1Recently, this type of
reasoninghas also been appliedto the areaof contractualarrangements.More
Financialsupportfrom the NationalScience Foundation(NSF) via grantSES-8908646, from
the GraduateSchool of IndustrialAdministration(GSIA) at Carnegie Mellon University, and
from the Universityof Michigan School of Business Administrationis gratefullyacknowledged.
As requiredby NSF, the data set used in this paperhas been depositedwith the Inter-University
Consortiumfor Political and Social Research(ICPSR).I would like to thankStan Baiman, Steve
Donald, Nancy Gallini, Pauline Ippolito, Nancy Lutz, Scott Masten, Marius Schwartz, Subrata
Sen, KathrynShaw, MargaretSlade, Ted Snyder, KannanSrinivasan,ValerieSuslow, andparticularly Sugato Bhattacharyyafor helpful comments and discussions, as well as RobertPicardfor
his assistance. I also thankthe editor, Alan Schwartz, for his guidance. Finally, I am gratefulto
workshop participantsat the AntitrustDivision of the Departmentof Justice, the FederalTrade
Commission Bureau of Economics, the MassachusettsInstitute of Technology Departmentof
Economics, the University of PittsburghDepartmentof Economics, the Universityof Michigan
GraduateSchool of Business Administrationand the Yale School of Management, for useful
comments. All errorsare, of course, mine. An earlierversion of this paper,entitled"An Empirical Look at FranchisingContractsas Signaling Devices," was circulatedas GSIA workingpaper
1990-19.
1. See, notably, Cooper and Ross (1984), Kihlstrom and Riordan (1984), and Wolinsky
(1983); the possibility that firms may use multiple signals is examined in Matthewsand Moore
(1987), Milgrom and Roberts (1986), and Wilson (1985), among others.
? 1993 by Oxford University Press. All rights reserved. 8756-6222/93/$5.00
as SignalDevices 257
Contractual
Arrangements
specifically, Gallini and Wright(1990) and Beggs (1992) arguethatthe use of
royalties on sales in licensing agreementsmight be explainedby the need for
the licensor to signal the value of the offered technology. A similar point is
made relative to verticalrelationshipsin general, and franchisingcontractsin
particular,in Tirole (1988). Finally, Gallini and Lutz (1992) argue that in the
context of franchising,not only can the use of royalty rates act as a signal to
potential franchisees, but so can the proportionof outlets franchisorschoose
to operatedirectly ratherthan franchise. Theirs is the firstmodel of franchising to address directly the issue of dual distribution,that is, of franchisors'
tendency to operate some stores directly and to franchise the others. Other
theoriesof franchisingmustrely on outlet heterogeneityand limitationson the
variety of contractsthat can be used to give rise to dual distribution.2
This article's contributionis to provide an empirical assessment of this
proposition that signaling may explain franchisors'choices of royalty rates
and franchisefees, as well as their tendency to operateoutlets directly.Franchisors' observed choices of royalty rates, franchisefees, and proportionof
company-ownedoutlets at the time they begin franchisingare relatedempirically to a measureof their types. This measureis calculatedbased on the rate
of growth in the numberof outlets of each franchisedchain in the years that
follow the franchisor'sinvolvementin franchising.
This analysis of whetheror not franchisecontractsserve a role as signaling
devices should be useful in at least two respects. First, it guides us in the
process of identifying the theories that can best explain franchising. This is
importantgiven that franchisinghas become a very prevalenttype of organizational arrangementin the U.S. economy. At present, the U.S. Department
of Commerce(U.S. DOC) estimatesthatmore thanone-thirdof all retailsales
occur through outlets of franchisedcompanies (U.S. DOC, 1988). Second,
because it provides an empiricaltest of the importanceor relevanceof signaling in a particularcontext, this article allows us to betterevaluatethe empirical importanceand/or relevance of signaling theory in general. Given that
signaling theoryhas been and continuesto be appliedto a numberof different
issues, it is importantthatwe assess its empiricalrelevance, andthatwe begin
to determinethe circumstancesunder which signaling may be more or less
importantas an explanationfor a particularphenomenon. Yet to date, very
little empirical work has been done in this area in industrialorganization.3
2. See Lafontaine(1992a) and Shepard(1993) for more on this.
3. A numberof empiricalanalyses of signaling models have been carriedout in finance and
accounting. These are concernednotablywith dividendpolicy or debt/equityratios as signals of
firms'quality, or with common stock repurchasesat a premiumor the underpricingof new issues
as signals of the same. (See, for example, Balvers, McDonald, and Miller, 1988; Downes and
Heinkel, 1982; Vermaelen, 1981; and the references therein). In general, the results in this
literatureare supportiveof signaling models. However, in the marketingliterature,the empirical
evidence has not been very supportiveof either the positive price-quality relationshipimplied by
signaling models (see Curryand Riesz, 1988; Gerstner,1985; Wiener, 1985; and the references
therein)or the warranty-qualityrelationshipsuggested by such models (see Garvin, 1983;Gemer
and Bryant, 1981; and Priest, 1981). One exception is Wiener (1985), who finds that product
warrantiesdo signal productreliability.
258 The Journalof Law.Economics.& Organization,
V9 N2
The present article focuses on the signaling hypothesis, but one finds a
numberof other explanationsfor franchisingin the literature.These include
models based on risk sharing (notably, Stiglitz, 1974, although he is concerned with sharecroppingratherthan franchising);on franchiseemoral hazard combined with risk aversion (e.g., Stiglitz, 1974 and Mathewson and
Winter, 1985), on two-sided moral hazard (Bhattacharyyaand Lafontaine,
1992; Lal, 1990; Rubin, 1978); and on screening (e.g., Hallagan, 1978, also
on sharecropping);as well as the traditionalexplanationfor franchisingbased
on capital marketimperfections(e.g., Caves and Murphy,1976). The empirical implications of these theories have been tested elsewhere (see, notably,
Brickley and Dark, 1987; Lafontaine, 1992a; Martin, 1988; and Norton,
1988). Unfortunately,the frameworkneeded to assess the implicationsof the
signaling model does not readily lend itself to drawing conclusions about
othertheories. As a result, the main empiricalevidence presentedhere cannot
really be used to infer much aboutalternativetheories. However,the existence
of these alternativeexplanationsimplies that other factors, such as riskiness
and incentive considerations,are likely to play a role in the design and use of
franchisecontracts.Thus these must be controlledfor in the empiricalanalysis. Consequently,some of the results obtained herein-in particular,those
that relate to the control variables-will bear on alternativemodels of franchising to some extent. These relationshipswith alternativetheories, and with
the existing empirical literature,are noted when appropriate.
The article is organizedas follows. Section 2 provides a discussion of the
theoreticalframework,and of the empiricalimplicationsof signaling models
for franchisingcontracts.Section 3 sets out the empiricalmodel, and includes
a brief descriptionof the data and of the primarydependentand explanatory
variables. Section 4 elaborateson some of the data issues, providing more
detailed information on variables and measurement,and on the empirical
methodology.Empiricalresults, and their interpretation,are found in Section
5. Section 6 concludes.
2. Theory and Empirical Implications
A franchiseagreementis a contractualarrangementbetween two independent
firms, whereby the franchisee pays the franchisorfor the right to sell the
franchisor'sproduct and/or the right to use the franchisor'strademarkat a
given place and for a certainperiod of time. In business formatfranchising,
franchise contracts typically involve the payment, by the franchisee to the
franchisor,of a proportionof the franchisee'ssales in royalties.4This propor4. The U.S. Departmentof Commerceclassifies franchisesaccordingto the main component
of the transaction:"ProductandTradeName Franchising"(also called 'TraditionalFranchising")
is characterizedby dealers who "concentrateon one company'sproductline and to some extent
identify theirbusinesswith thatcompany"(U.S. DOC 1988: 1). This type of franchisingis limited
to car dealerships, soft-drinkbottlers, and gasoline service stations. In "Business FormatFranchising" the relationship between franchisor and franchisee "includes not only the product,
service, and trademark,but the entire business format itself-a marketingstrategy and plan,
operating manuals and standards, quality control, and continuing two way communication"
Contractual
as SignalDevices 259
Arrangements
tion is usually constant over all sales levels.5 In addition, franchisorsalmost
always requirea lump-sum franchisefee that is paid up front and only once
for the durationof the contract-which, according to U.S. Departmentof
Commerce (1988) is about 15 years on average.
Since franchisinginvolves the transferfrom the franchisorto the franchisee
of a tradename and whole way of doing business, which are basically intangible assets, it has been arguedthatit is difficultfor the buyerto assess the value
of these assets a priori. This is especially true for potentialfranchiseesinterested in new franchisesystems, thatis, in systems offeredby franchisorswith
no established reputation.This is emphasized in the trade literatureby the
pervasive calls for franchisees to "investigatebefore they invest."
In the theoreticalliterature,authorshave analyzedthis informationaladvantage of the franchisorand developed an explanationfor franchisingand franchise contractsthat relies on franchisors'need to signal theirtype to potential
franchisees. This signaling explanationhas been used to explain franchisors'
choices of contractterms, namely their royalty rates and franchisefees [see
Beggs (1992) and Gallini and Wright(1990) on licensing contracts;Desai and
Srinivasan(1990), Gallini and Lutz (1992), and Tirole (1988) on franchising],
as well as their tendency to operateoutlets directly (Gallini and Lutz, 1992).
This section discusses the empiricalimplicationsof the above models in the
context of franchising.The interestedreaderis referredto Appendix A for a
description of a representativesignaling model as it has been applied to
franchising, as well as a more systematic derivation of the testable implications.
The main insight from the signaling literaturein the context of linear
payment rules such as those used in franchisingis that "good"franchisorsthat is, those with a high-value trade name-who have not yet established
their reputationcan signal their type by offering a contractthat makes their
revenues highly dependentupon the performanceof the outlet. The argument
is very similar to that found in Leland and Pyle (1977), where entrepreneurs
can convince potentialinvestorsof the value of theirprojectby investingmore
in it themselves. In otherwords, the contracta high-typefranchisor(one with
a high-value tradename) should offer to differentiateor separatehimself from
a low-type franchisor (one with a low-value trade name) would involve a
higher royalty rate and lower franchisefee-to the point where the low type
would find it unprofitableto imitate (see Beggs, 1992; Gallini and Wright,
1990; and Tirole, 1988). The reason why the low-type franchisorfinds it
unprofitableto imitatethis contractis thatthe impositionof royaltyratesleads
to distortionsin franchisees'price and quantitydecisions due to double marginalization. It is assumed that the reduction in revenues arising from in(1988: 3). Examples include restaurants,business and employment services, and real estate
agencies. In the remainderof this articlethe word "franchising"is used to mean business format
franchising.
5. In Lafontaine(1992b), 93 of the 117respondentssaid that their royalty rate was constant
over all sales levels.
260
The Joural of Law. Economics, & Organization, V9 N2
creases in the royalty rates are greaterfor low-type franchisorsthan they are
for high types (this is the usual sortingcondition;see AppendixA). Hence, in
equilibrium, the high type will be able to separate from the low type by
choosing a high enough royalty rate and a correspondinglylow enough franchise fee. The low type, unableto profitablyimitatethis contract,will opt for
a contractwith no royalties, that is, a contractinducingno distortionsdownstream. In equilibrium,the royalty rate for the high type will be greaterthan
the royalty rate of the low type, which is zero. At the same time, it can be
shown that the franchisefee of the high type will necessarilybe smaller than
the franchisefee of the low type (see Appendix A).
Using T to denote the franchisor'stype (or, equivalently,the value of the
trade name) and extending the argumentto a continuum of types or trade
name values, the testable implications from the signaling model are as
follows:
Al. For franchisorswith no established reputation,holding other factors
constant, the royalty rate, r, increases with T.
A2. For franchisorswith no established reputation,holding other factors
constant, the franchise fee, F, is decreasing in T.
Note that as a result of these two relationships,we should also find that the
franchise fee F and the royalty rate r are inversely related. This will be the
case here irrespective of whether or not the franchisee's participationconstraintis binding. In most other models of franchising,a negative correlation
between the two fees would arise only because franchiseesare kept at their
reservationlevel of utility-that is, because franchisee's participationconstraintsare binding.
The focus of the above discussion was on franchisors'capacity to signal
throughthe terms of their franchisecontracts.The argumentunderlyingthis
result is thatfranchisorscan signal theirtype by makingtheirrevenuesdepend
on the value of the franchise. As noted in Gallini and Lutz (1992), besides
levying royalties based on outlets' sales, franchisorscan make their revenues
depend on the overall value of their franchiseby operatingoutlets directly.
And most franchisorsdo own and operatea numberof their outlets.6 Hence
Gallini and Lutz (1992) suggest that signaling might explain not only the use
of royalties in franchising,but also why it is that franchisorsoperatesome of
their units, while franchisingthe others.
Three types of separatingequilibriamay occur in this new model. First, it
6. See, for example, Lafontaine(1992a):431 of her 548 franchisorsoperatedoutlets directly.
The average proportionof company-operatedstores across the 548 firms was 17.25 percent. In
fact, dual distributionis so much a partof franchisingthat most of the empiricalwork to date in
this area has focused on explaining the franchise versus company-owndecision. See, notably,
Brickley and Dark (1987), Martin(1988), Minkler (1990), and Norton (1988).
Contractual Arrangements as Signal Devices
261
could be that franchisorswould decide not to use company ownership as a
signal, in which case all signaling would again be achievedthroughthe terms
of their franchise contact. Hence the testable implicationswould still be Al
and A2. Alternatively,signalingmay be done completely throughfranchisors'
choice of contract mix, and not at all through the terms of the franchise
contract. In that case, in equilibrium,good franchisorswill signal their type
by relying heavily on company ownership, while bad franchisors,unable to
imitate, will choose not to operate any outlets directly. Extending this to a
continuumof types, we have:
Bl. For firmswith no establishedreputation,holding otherthings constant,
the proportionof company-ownedoutlets, a, increases with the franchisor's
type or trade name value T.
Finally, both company-ownedoutlets and royalty rates may be used as
signals. But even if firms use both of these to signal their type, there is still
only one basic signal. In both cases, high-valuefranchisorsdifferentiatethemselves from lower-valuechains by makingtheirrevenuesdependon the value
of their franchise. And the extent to which their revenues do depend on
downstreamsales-that is, franchisors'stakes in downstreamoperationscan be estimated through a combinationof their royalty rates and their reliance on company ownership. If the contract mix and/or royalty rates are
used as signaling devices, we should find that this stake increases with the
franchisor'stype, or that:
Cl. For franchisorswith no established reputation,holding other things
constant, the franchisor'sstake in downstreamoperations,which reflectsboth
royalty rate and tendency to operateoutlets directly,is increasingin the type
or trade name value T.
In addition, if both signals are used, given a franchisor'stype, the royalty
rate and the use of company ownershipshould be negatively correlated.The
two signals are basically substitutes, so that as one signal is used more to
achieve the desired level of dependency of the franchisor'srevenues on the
outlets' performance, the other should be used less. Which combinationof
royalties and company-ownedoutlets is actually used will depend on the
relative cost and efficiency of each signal. Hence we have:
C2. For a given type or trade name value T, the proportionof companyowned outlets, a, and the royalty rate r, are negatively correlated.
Finally, if the franchisee'sparticipationconstraintis binding, the franchise
fee and the royalty rate should again be negatively correlated,but the franchise fee need not be decreasing, and the loyalty rateincreasing,in the type of
the franchisorwhen both signals are used simultaneously.
262 TheJournal
of Lav.Economics,
&Organization,
V9N2
3. Empirical Specification
The data used in this article are obtainedprimarilyfrom EntrepreneurMagazine's yearly "Franchise500" surveys (Entrepreneur,various years).7 These
surveys cover an average of about 1,000 franchisorseach year, providing
informationon royalty rates, advertisingfees, franchisefees, and the number
of company-ownedand franchisedoutlets in the chain.8The surveysalso state
the year in which the franchisorsbegan their operations, the year in which
they became involved in franchising,the type of businessthey are involved in,
and the amount of capital requiredto open an outlet. The earliest usable
survey covers data for 1980, and the last survey included in this research
contains the 1989 data.9
A numberof issues arise when tryingto adaptthese datato test the hypotheses developed above. First, the signaling model addressesthe issue of how
new franchisorscan convey informationaboutthemselvesthroughtheirinitial
contractmix and contractterms. As a result, one needs to constructa sample
of franchisorsfor which dataon the initial choices of contractterms, r and F,
and of contractmix, a, are available.Hence only firmsthatstartedfranchising
on or after the first of the Entrepreneursurvey years (that is, 1980) can be
included in the sample.10This eliminates from the sample well-established
franchisors such as McDonald's and Burger King, both of which started
franchisingaround1955.
Second, establishing empirical differences in behaviorbetween high- and
low-value franchisorsrequires that one measures their type or trade name
value T. Since no data on sales or profits, at the outlet or the chain level, are
availablefor these firms, this articlerelies on a measureof type based on the
observed growth in the numberof outlets of the franchisedchain in the years
that follow its involvement in franchising. This, of course, is based on the
notion thata growing chain is a profitablechain. In the contextof franchising,
such an assumption is probably quite reasonable: Given the geographical
natureof the types of businesses that are usually franchised,total sales and
profits for chains tend to be highly correlatedwith the numberof outlets."
However, measuringa firm'stype based on its ex post success requiresthatwe
observe the firm a numberof years after it startsfranchising,so that we can
7. Other sources used here include Bond (1989) and the Departmentof Commerce'sFranchise OpportunitiesHandbook(U.S. DOC, 1985, 1986, 1988).
8. Due in large part to entry and exit, these are not the same 1,000 franchisorseach year.
9. The first survey was actuallydone in 1980, andcontainedthe 1979 data. Unfortunately,in
that survey, the advertisingfee was not reported.
10. Because the calendaryears do not provide appropriatecutoff points, I include those firms
that startedfranchisingin year t - 1 if they appearin the year t survey but did not appearin the
t - I survey. Thus firms that startedfranchisingin 1979 can be found in the sample despite the
fact thatthe firstsurvey was for 1980. The fees and contractmix reportedin the year t survey are
interpretedas initial contractofferingsand contractmixes in such cases. The five-yeardifference
requiredto calculate growth implies that these firms must be observed again in the yeart + 5
survey or later, since their initial data are in fact for year t.
11. Also, measuresof "success"based on outlets are used quite frequentlyin the tradeliterature on franchising. See for example Venture's"The Franchise 100" (Venture,Variousyears).
as SignalDevices 263
Contractual
Arrangements
assess its growth. Given 10 years of data, I chose 5 years as the minimum
elapsed time over which to measure growth. Combined with the constraint
from above that we observe their first year in franchising, this implies that
only firms that startedfranchisingbetween 1980 and 1984 can be part of the
sample. A total of 168 franchisorswere found to fulfill these conditionsacross
the 10 surveys. Missing or inconsistentdataled to the rejectionof 43 of these,
resulting in a final sample of 125 franchisors.
HypothesesAl, A2, B1, and C1 imply thatfranchisors'choices of contract
terms and company ownershipshould be explainedby their qualitylevels, or
types. Hence the firm'stype, measuredby its ex post growth, will be the main
explanatoryvariableof interesthere. However, severalothervariablesshould
be controlled for in the empirical model, for a numberof differentreasons.
First, input sales at a markupcould be a substitutefor royalties on sales, and
hence for company operation of individual outlets in the present context.12
But only if franchiseescan be requiredto buy the inputs as part of the franchise agreementwill the franchisorbe able to chargeprices above the competitive level.13This would imply that franchisorsmay have an additionalsignal
at their disposal, namely the percentage of their franchisees' inputs they
supply contractually,combined with the markupthey take on the sale of such
inputs. This new instrumentwould warrantempiricalanalysis on its own, but
because only aggregatedata on input sales are available, ratherthan treatthis
variable as an additionalsignal, I include input sales here as a control variable. Second, othervariables, such as the numberof years the franchisorwas
in business before deciding to startfranchising(a measureof reputation),or
the sector in which the firm operates, and the year in which the firm began
franchising, can convey informationabout the franchisor'stype. These may
reduce the need for the firmto signal throughthe dependentvariables.Third,
the existing empirical literatureon franchising implies that factors such as
risk, franchisee moral hazard, franchisormoral hazard, and capital requirements could influencefranchisors'choices of contracttermsand contractmix.
Hence they should be controlled for in the empirical model.
Thus hypotheses Al, A2, B1, and C1 are characterizedempirically as
follows:
n
YZ=
+ Pi Ti + I
k=2
PkZh + E
j = 1,...4
(l)
where Y} is franchisori's royalty rate, Y? is the franchisefee, Y7 stands for
franchisori's proportionof company-ownedoutlets, and Y4 is the stake the
firm has in downstreamoperations.All of these are measuredat the time the
firm becomes involved in franchising.The main explanatoryvariablein this
12. See, for example, Caves and Murphy(1976) on the equivalence between input markups
and royalty rates.
13. I thank the editor, Alan Schwartz, for pointing this out.
264
V9 N2
The Journalof Lav,.Economics.& Organization.
model, Ti, is the measure of ex post growth that capturesthe type or trade
name value of franchisori. The Zk'sare the controlvariables, and the e's are
independentidentically distributederrorterms.
4. Data, Variable Specification, and Estimation Methods
This section provides more informationon the various measurementissues
involved in estimating the above model, and the variables used in the analyses. The reader is referredto Appendix B for more details on the data.
4.1 Dependent Variables
In order to estimate the empirical model described above, one must obtain
measuresof the contractterms, namelythe royaltyratesand franchisefees, of
the contractmix, and of the stake franchisorshave in downstreamoperations
at the time they become involved in franchising. The first two are obtained
directly from the survey data. The last two dependentvariablespresentmore
of a measurementproblem.That is because, by definition,when firmsarejust
becoming involved in franchising, they are 100 percent company-owned.In
otherwords, if the contractmix (or the stake in downstreamoperations)is to
signal the franchisor'stype in any way, the initial numberof company-owned
outlets must be the variablethat serves as the signal, giving franchiseessome
idea about what the proportionof company-operatedoutlets mightbe over the
next few years. In this article, I use first the logarithm of the number of
company-ownedoutlets (denoted by d) as a measure of the contractmix.14
Second, I use a measureof what the proportionof company-ownedoutlets is
expected to be after t years in franchising, given the number of franchises
typical franchisorsopen up duringtheir first t years in franchising.This last
measure is denoted by a,.15
To implementa test for the case where firmsuse both r and a to signal their
type, given a measuresuch as a,, one can estimatethe stakethe franchisorhas
in downstreamoperationsby combiningroyalties and companyownershipas
follows:
SSt
at * AvSalesc + (1 - a)AvSalesf
r/100l
a, ?AvSalesc + (1 a,)AvSalesf
(2)
whereAvSalescandAvSalesfreferrespectivelyto averagesales per companyowned and per franchisedoutlet, at the time the firm becomes involved in
franchising. This formula gives the share of the chain's revenues that is
14. Because some firmshave no company-ownedoutlets when they startfranchising,I add 0.5
to the numberof outlets before takingthe logarithm.Also, resultswere equivalentwhetherI used
the numberof company-ownedoutlets, or its logarithm.
15. In other words, I use a, = co/(co + f,) as one way to capturethe theoreticalownership
proportion,wheref, is the numberof franchisedunits typically opened within t years in franchising. This measureassumes that the numberof company-operatedoutlets remainsconstantat the
initial level of co. This is well supportedby the data. The median numberof franchisedunits
opened during the first 3 (5) years in franchising,f3 (f5), was 20 (32).
Contractual
as SignalDevices 265
Arrangements
at timet. Unfortunately,
dataon averdirectlyby the franchisor
appropriated
outlet
unavailable
are
for
sales
franchisors.
individual
data
Aggregate
age
per
andfranchisedoutletsareobtainedon a secon sales percompany-operated
toralbasisfromthe U.S. Department
of Commerce(1988).16(See Appendix
B for moredetails.)
Finally,one mayassumethatfirmscontinueto signaltheirtypesfor a few
yearsbeyondthetimeat whichtheyactuallybecomeinvolvedin franchising.
In thatcase, the actualproportionof company-owned
outlets,a,, and the
actualS, [estimatedfrom(2) givena, ] threeor fouryearsafterthe firmgets
involvedin franchisingmay also be used to measurea andS, respectively.
Specifically,a, andS, aremeasuredtwoyearsbeforetheendof thefive-yearthetwoobservations
on eachfirm.(SeeAppendixB for
plusperiodseparating
moredetails).
4.2 ExplanatoryVariables
As noted earlier,the measureof type used in this articleis basedon the
observedrateof growthof eachchainin theyearsthatfollowits involvement
in franchising.Thegrowthrateis measuredas thedifferencein thelogarithm
of thenumberof outletsbetweentwotimeperiodsatleastfiveyearsapart(but
as close as possible to five years apart)dividedby the numberof years
between the two observations(which varies from 5 to 8 years).17Given that
we knowfromJovanovic(1982),as well as previousempiricalwork,thata
firm'sgrowthrate tendsto declineas it becomesmoremature,the actual
measureusedhereto capturethetypeof thefirmis thedeviationof thefirm's
rateof growthfromits expectedrateof growthgivenits age. In otherwords,
the actualmeasureof type is the errortermfroma regressionof the growth
rate on the numberof years that a firmwas in businessbeforeit began
franchising.18
Onedrawbackassociatedwithusinga measurebasedon thegrowthratein
thenumberof outletsto capturetheintrinsicvalueof thefranchiseis thatlowfranchisors
qualityor even fraudulent
maysucceedin gettinga largenumber
of franchiseesto invest in their system initially.As a result,their initial
growthin numberof outletswouldbe verylarge,despitetheirbeingof low
quality.To alleviatethisproblem,I measuregrowthovera periodof at least
16. AvSales*and AvSalesf must be differentiatedin this formula because average sales per
company-ownedoutlets tend to be larger than those of franchisedunits. See Franchising in the
Economy, 1986-88 (U.S. DOC, 1988: Table 19).
17. Results were the same when the growth rate was measuredas the proportionalchange in
the number of outlets. Results were also unchanged when the growth rate in the number of
franchises only was considered. Since I was interestedin measuringthe change in profitability,
growth in the total numberof outlets seemed a more appropriatemeasure.
18. The reciprocalfunctionalform was used, in this regression, as it provedmost satisfactory.
Consistent with Jovanovic's model, and with most of the existing evidence, franchisors'growth
rates were significantlyinversely relatedto theirage. An attemptwas also madeto controlfor the
number of outlets a firm had as it became involved in franchising, but the coefficient for this
variable was never significant. The qualitative results presented here remain the same if the
growth rate is used directly as the measureof type.
266 The Journalof Law.Economics,& Organization,
V9 N2
five years; this relatively long period should minimize the effect that early
phenomenal growth of low-quality franchisedchains might have on the results. In addition, I examine how sensitive results are to initial growth by
contrastingthose obtained over the whole five-year-plusrange with results
obtainedwhen the measureof growth is limited to the last two of the five or
more years. These two ways of measuringgrowth are referredto as Growthl
and Growth2in the remainderof this paper. After controllingfor the firm's
expectedgrowthrategiven its age, theybecome Typel andType2,respectively.
Having describedin some detail the main explanatoryvariableused in the
empiricalanalyses below, I now turnto a discussion of the controlvariables.
First, as noted earlier, it is necessary to controlfor input sales in considering
the relationshipbetween the signals and firmtypes, given that inputsales at a
markupcould serve as an alternativesignal. Unfortunately,the data do not
allow a distinctionbetween those sales from franchisorto franchiseethat are
contractuallyrequired (and hence on which the franchisorcould charge a
markup)and those that are voluntaryon the part of the franchisee. Second,
factorsthat allow the firmto differentiateitself from otherfranchisorsmust be
included in the regressions. For example, the number of years in business
prior to startingfranchisingmay be related to or give informationabout the
success a firm will achieve in franchising:In a sense, the firm has already
establisheda reputationfor itself in the market.In that case, a new franchisor
need only distinguish itself from other new franchisorsthat have been in
business for the same length of time. Thus one should controlfor the number
of years a firm was in business priorto franchisingin the empiricalspecification. Similarly, franchisorsmay need only to distinguish themselves from
other franchisorsgetting involved in franchisingat the same time they are.
Because there are only a few possible years in which firmsin this sample may
have started franchising, from 1979 to 1984, dummy variables are used to
control for this effect (with 1979 as the base case). Furthermore,individual
franchisorsmay only compete with, and hence need only separatethemselves
from, other firms involved in the same type of business. This implies that
sectoraldummy variablesshould also be used as controlvariables. Appendix
B identifies which sectors are differentiatedin this way.
Finally,controllingfor factorssuggestedby alternativetheoriesof franchising as potential forces influencing the firms' choices of contractterms and
contractmix requiresthatone quantifiesfactorssuch as risk, franchiseemoral
hazard, franchisormoralhazard,and capitalrequirements.This task presents
many challenges. First is the fact that it is not clear what the best measures
are, and second are the constraintsimposed by the availabledata. However,
since most of these factors are likely to be more homogeneouswithin sectors
than across them, sectoral dummy variableswill be useful here as well. The
dummy variablesrepresentingthe year the firmbecame involved in franchising should also be useful in this respect. In addition, informationconcerning
the sectoral rates of discontinuationof individualoutlets for the appropriate
years is providedin Franchising in the Economy (U.S. DOC, variousyears).
These can be used as a measure of the degree of risk faced by potential
Contractual Arrangements as Signal Devices
267
franchisees. Finally, the Entrepreneursurveys directly include an estimate of
the amount of capital requiredto open an outlet in each franchisedchain.19
Having found proxies for risk and for capitalrequirements,we are left with
the measurementof both franchiseeand franchisormoral hazard.In the case
of franchisee moral hazard, measures of geographicaldispersion have been
used in the literatureto capturedifferences in the cost of monitoringoutlet
performance(e.g., Brickley and Dark, 1987; Lafontaine,1992a).These measures are based on the notion that a higher cost of monitoringmakes franchisee moralhazardmore problematic.The numberof states in which a given
chain has establishedoutlets at the time of its initialfranchisecontractoffering
is the measure of geographicaldispersionused here. It is availablefrom the
U.S. Departmentof Commerce'sFranchise OpportunitiesHandbook (U.S.
DOC, various years) for 66 of the 125 firms in the sample.
Finally, for franchisormoral hazard,one measureof the importanceof the
franchisor'srole in the business thatis availablefor the same 66 firmsfromthe
same source is the amount of training (in days) that is provided to new
franchisees.This is assumedto capturethe amountof know-howthatis being
transferred.Also, given that the tradename is the main franchisorinput, and
that it is likely to become more valuable the longer a firm has been in
business, the numberof years in business can be used to capturethe importance of the franchisor'sinput. In this context, the effect of this variableon the
royalty rates and on the reliance on company-operatedoutlets would be the
opposite of those suggested by the signaling theory:Under franchisormoral
hazard, the more valuable the trade name is (that is, the more importantthe
franchisor'srole in the franchise)the more incentives one would want to give
to the franchisor. These greater incentives can be achieved either through
higher royalties or throughmore company ownership.
4.3 EstimationMethods
All of the dependentvariablesof interestin this paperare boundedbelow at 0,
and some are also bounded above at 100 (percent). Given that these boundaries representa binding constraintfor at least some of the observations, a
Tobit estimatoris used. In other words, the estimated model for the royalty
rate, the proportionof company-ownedstores, and the stake of the franchisor
in downstreamoperationis
n
Y; = So +
fT,i >+
/AZki +e
if 100 > RHS > 0,
(3)
if RHS ' 0,
(4)
k=2
and
Y = 0
19. Franchisorsoften give a rangeof values for this variable.I use the averagevalue, assuming
that it representsan estimate of the amount of capital necessary to open an averageoutlet.
&Organization,
268 TheJouralof Law,Economics.
V9N2
and
if RHS : 100,
Y' = 100
(5)
for j = 1, 3, and 4. For the franchisefee, which is not boundedabove, the
model is
n
y2 = PI2
fZT+
T+ +
k Z,
if RHS > 0
(6)
otherwise.
(7)
k=2
and
2= 0
Assuming thatthe errortermsin the above equationsare normallydistributed,
these equationscan be estimatedunder maximumlikelihood. In those cases
where the upperbound never binds, such as for the royaltyrate and for some
of the measures of both the contractmix and the stake of the franchisorin
downstreamoperations,the two-tail Tobitestimatordefinedby equations(3)
to (5) reduces to the single-tail version used for the franchise fee, as in
equations (6) and (7).20
Since the data are cross-sectional, equationswere tested for heteroscedasTests were also carriedout to deterticity, and correctedwhen appropriate.21
mine whether multiplicativesectoral dummies should be included in the regressions. In general, these dummies were not found to be significant and
hence they were excluded from the analyses.22Finally, the requirementthat
firms be observed at two points in time at least five years apartimplies that
only relatively successful firms-that is, firms that survive at least five
years-are included in this sample. This in turn raises the possibility of a
selection problem. This is not addressed in the estimations for a number
of reasons. First, as noted in Appendix A, it is necessary that the "bad"
franchisorsbe profitablefor separationto occur in this model. Hence there
is a theoretical reason for excluding firms that fail from the sample under
study here.23 Second, it is not possible, using the available survey data, to
20. Note that results obtained under ordinaryleast squares (OLS) were very similar to the
Tobit estimates.
21. See Maddala(1983) for a descriptionof the methodologyused here. The functionalform
used for a was a = exp(a + fBType).I am grateful to Stephen Donald for suggesting this
functionalform. The test for heteroscedasticitythen reducesto a test of the significanceof /. The
value of the heteroscedasticitycoefficient P and its asymptotic t-ratio are reported for each
equationat the bottom of the tables. Correctionswere carriedout wheneverthe #3coefficientwas
found to be significantat a level of .10 or better.
22. These results are available from the authorupon request.
23. In particular,one must be sure to exclude from the sample fly-by-nightoperators.The
reason is that these firms are not "playing the same game." Unlike the set of firms considered
here, they do not care about royalties, since they do not really intendto collect any. As a result,
Contractual
Arrangementsas Signal Devices 269
infer franchisorsurvival. In otherwords, the absence of a franchisorfrom the
survey in a given year cannot be interpretedas a sign of failure. Third, and
most important,even if the sample under study excludes or underrepresents
those firms at the bottom of the distributionof types, the theory would still
predict, and hence we should still observe, over the remainingset of firms, the
relations described in Section 2.
5. Empirical Results
Table 1 presents some descriptive statistics about initial royalty rates, franchise fees, number of company-operatedoutlets, and ex post growth rates.
The franchisorin this sample who requiredthe largestroyaltypaymentwhen
it first startedfranchisingwas Jazzercise, a fitness programcompany.24The
firm with the largest franchisefee was Qual Krom, in the automotiveservice
industry.The company that had the greatestnumberof company-ownedoutlets when it began franchisingwas Jack in the Box, a fast-food hamburger
chain, with 780 units. The company that experienced the greatest level of
growth in the presentsample, accordingto Growthl (and Typel), was Sparks
Tune-Up, while according to Growth2 (and Type2), it was Fantasy Coachworks Ltd., a chain of "Auto Boutiques."
While this is not shown in the table, it is interestingto note that royalty
rates and franchise fees did not tend to vary much over the five-year-plus
period between the two observationson each firm. The averageroyalty rate
declined from 7.04 to 6.65, and the averagefranchisefee went from 14.24 to
16.21 (thousandsof 1980 dollars). Similarly,the numberof company-owned
outlets went down slightly from an averageof 20.96 to 19.28. None of these
differences is statisticallysignificant.25These results are consistent with survey resultsreportedin Lafontaine(1992b).They are also supportedby Banerji
and Simon (1991), who find that the contract terms of their 27 franchisors
remainedrelatively stable duringthe five- to six-year period over which they
observed them.
Table 1 also gives descriptive statistics for the various measures of the
proportionof franchisedstores and of the stake the firm has in downstream
operations, that is, a = log(co), a, and S, when t = 3 and 5 years, respectively, as well as their actualproportionof company-ownedstores and actual
stake in downstreamoperationsthree or four years later, a, and S~, respectively.
Notice thatsince most firmsoperateunits directly,andhave positive royalty
they may well choose a relativelyhigh royaltyrateand a low franchisefee. The lattermight make
it easier to sell a large numberof franchisesin a very limited amountof time, while the former
makes the low franchise fee credible. In other words, these firms do not satisfy the sorting
condition: They incur no loss of revenues from increasingr.
24. Five years later this fee was down from 31.5 percent to 20 percent.
25. The statistical tests involve calculating the mean and standarderror of the individual
differences across firms. Confidence intervalsaroundthe mean differencesincluded zero for all
reasonableconfidence intervals, for all three variables.
270
The Joural of Lav, Economics,& Organization,
V9 N2
Table 1. DescriptiveStatistics for the Franchisorsin the Sample
Variable
Royaltyrate (%)
Franchisefee ($000)
Company-ownedunits
&
a3 (3 years)
as (5 years)
a, (3-4 years)
S3 (3 years)
Ss (5 years)
Sa (3-4 years)
Growthl(5 years +)
Growth2(last 2 years)
Typel (5 years +)
Type2 (last 2 years)
Years in business
Capitalneeded ($000)
Inputsales (%)
Days of training
Numberof states
N
Mean
a
125
125
125
125
125
125
125
125
125
125
125
125
125
125
125
125
125
66
66
7.04
14.24
20.96
1.11
15.93
11.99
23.96
29.56
24.20
37.95
0.23
0.13
0.00
0.00
6.50
96.53
13.11
13.13
5.48
3.87
11.11
101.65
1.38
19.72
18.17
27.02
21.33
19.49
30.22
0.19
0.20
0.17
0.20
9.60
171.50
12.27
11.64
8.15
Minimum
0.00
0.00
0.00
-0.69
0.00
0.00
0.00
0.00
0.00
0.00
-0.22
-0.44
-0.44
-0.54
0.00
1.00
0.18
2.50
1.00
Maximum
31.50
79.30
780.00
6.66
97.50
96.06
100.00
99.19
98.71
100.00
0.83
0.69
0.42
0.59
52.00
1087.5
34.59
62.50
50.00
rates, the data are not really consistent with the notion that firms use strictly
theircontractterms (firsttype of equilibriumdiscussed above) or strictlytheir
mix of company-ownedand franchisedoutlets (second type of equilibrium)to
signal theirtypes. Hence, takingthese dataat face value, we areleft only with
the possibility that franchisorsuse both signals to differentiatethemselves.
However, the data in Table 1 might still be consistent with the first type of
equilibriumif we assume that company ownershipis used for purposes that
have nothing to do with signaling. For example, one might argue that firms
operateunits directly to obtain informationabouttheir markets,or to provide
training grounds for franchisees.26Similarly, the data in Table 1 might be
consistent with the second type of equilibrium,where signaling is achieved
completely throughthe contractmix, if one assumedthat franchisorsrely on
royalty payments for reasons unrelatedto signaling.
5.1 Main Regression Results
Regression results for the various dependentvariablesover the sample of 125
franchisorsare found in Tables2, 3 and4. Table2 shows the resultspertaining
to the termsof the franchisecontract,r andF. Table3 containsresultsrelative
to the contractualmix, a, while Table4 shows resultspertainingto the stakeS
26. In Lafontaine (1992b), 44 of the 130 franchisorsresponding said that company-owned
outlets serve as some sort of marketinformation-gathering
device, and 84 said thatthey were used
as researchcenters and traininggrounds for new franchisees. (Note that multiple answers were
allowed for this question; the total numberof answers was 234, not 130.)
as SignalDevices 271
Contractual
Arrangements
Table 2. Tobit Regression Results for RoyaltyRates and FranchiseFees
Variable
Royalty
Type1
-0.16
(-0.09)
Franchise
Fee
0.07**
(2.06)
Capital required
Discontinuations
Inputsales
Firstfranchised in
1980
1981
1982
1983
1984
-0.20
(-1.04)
0.69*
(1.69)
0.15*
(1.91)
0.88
(0.92)
-1.16
(-1.14)
2.94***
(2.86)
1.29
(1.22)
-0.16
(-0.11)
Constant
5.13***
(3.21)
LimitObservations
Non-LimitObservations
Log-likelihood
Hetero.Coeff. (fB)
Franchise
Fee
-3.00
(-0.57)
Type2
Years in business
Royalty
0.03
(0.32)
1.82***
(3.19)
-2.62**
(-2.25)
-0.13
(-0.58)
6.73**
(2.23)
4.87
(1.61)
2.79
(0.96)
-1.87
(-0.63)
5.28
(1.30)
18.03***
(4.05)
-1.05
(-0.67)
0.07**
(2.06)
-0.21
(-1.07)
0.71*
(1.74)
0.15*
(1.94)
0.83
(0.87)
-1.14
(-1.13)
2.88***
(2.83)
1.26
(1.22)
-0.24
(-0.17)
5.14***
(3.24)
-0.65
(-0.14)
-0.04
(-0.34)
2.40***
(3.93)
-2.76**
(-2.21)
-0.07
(-0.30)
5.11*
(1.74)
4.42
(1.42)
3.32
(1.07)
-2.42
(-0.76)
7.85*
(1.85)
17.76***
(3.65)
4
6
4
6
121
-320.41
0.26
(0.60)
119
-446.72
-1.17**
(-2.55)
121
-320.12
-0.18
(-0.34)
119
-449.70
-0.46
(-1.22)
tests:"'p < .01;"p < .05;'p < .10
in parentheses.
Resultsof two-tailed
Note:Asymptotic
t-statistics
the firm has in downstreamoperations.27In all three tables, results obtained
when growth is measured over the whole five or more years (Typel) are
presented first, followed by those obtained when growth is measured over
only the last two of the same five-years-plusperiod (Type2).28
The predictionsfrom the signaling models-that the effect of franchisors'
types on the royaltyrates, or on theirproportionsof franchisedstores, or on the
27. In terms of the growth variable, all equations were estimated under both a linear and a
quadraticform. The two sets of results were equivalent. Also, since they were used as control
variablesand do not providemuch insight to the problemat hand, the coefficientsobtainedfor the
six sectoral dummy variables are not reportedin the tables.
28. Note that the initial royalty rates and franchisefees are not significantlynegatively correlated in these data. This is consistent with results found in Baneji and Simon (1991) and
Lafontaine(1992a), and it suggests that there might be rents left downstream.See also Mathewson and Winter (1985) and Kaufmannand Lafontaine(1994) on this issue.
Table 3. TobitRegression Resultsfor the Use of Company-OwnedOutlets
Variable
Type1
a
a3
-0.88
(-1.46)
-11.34
(-1.22)
-
aa
a.
CX3
-43.58***
(-3.23)
Type2
Years in business
Capitalrequired
Discontinuations
Inputsales
Firstfranchisedin
1980
1981
1982
1983
1984
Constant
LowerLimitObservations
Upper LimitObservations
Non-LimitObservations
Log-likelihood
Hetero.coeff. (f)
(0.04***
(3.56)
0.12*
(1.96)
0.10
(0.71)
0.05*
(1.86)
-0.30
(-0.92)
-0.76**
(-2.20)
0.02
(0.05)
0.48
(1.49)
0.15
(0.32)
0.56
(1.06)
20
0
105
-175.37
-1.13**
(-2.37)
0.68***
(3.94)
1.79**
(2.05)
1.55
(0.75)
0.71*
(1.83)
-4.67
(-0.93)
-10.78**
(-2.08)
-0.74
(-0.15)
8.89*
(1.84)
4.12
(0.59)
5.86
(0.74)
20
0
105
-460.02
-1.44***
(-2.92)
1.08***
(4.58)
1.57
(1.39)
3.16
(1.20)
0.58
(1.18)
-5.85
(-0.85)
-4.92
(-0.68)
-1.71
(-0.27)
10.12
(1.59)
6.64
(0.74)
-0.99
(-0.10)
24
3
98
-467.83
-1.86***
(-4.03)
Note:Asymptotic
inparentheses.
t-statistics
Resultsof two-tailed
tests:"*^p< .01;"p < .05;*p< .10.
((-
1
-17
(-
ContractualArrangementsas Signal Devices 273
Table 4. TobitRegression Results for the Stake of the Franchisor
in DownstreamOperations
Variable
Typel
S3
Sa
-19.65**
(-2.19)
-50.96***
(-3.85)
S3
-9.15
Type2
(-1.19)
Years in business
Capital required
Discontinuations
Inputsales
Firstfranchised in
1980
1981
1982
1983
1984
Constant
LowerLimitObservations
Upper LimitObservations
Non-LimitObservations
Log-likelihood
Hetero.Coeff. (8)
0.66***
(3.73)
2.13**
(2.31)
2.31
(1.14)
0.28
(0.74)
-4.70
(-0.93)
-13.66**
(-2.63)
-1.29
(-0.26)
9.43*
(1.87)
-2.88
(-0.41)
20.14**
(2.57)
3
0
122
-520.97
-1.13***
(-2.83)
1.09***
(4.24)
1.56
(1.21)
2.50
(0.89)
0.39
(0.72)
-5.41
(-0.75)
-9.08
(-1.20)
-3.00
(-0.43)
13.08*
(1.85)
-0.05
(-0.05)
19.36*
(1.81)
4
3
118
-549.99
-1.37**
(-3.21)
0.70***
(3.93)
2.09**
(1.99)
2.97
(1.44)
0.43
(1.08)
-4.34
(-0.86)
-14.98***
(-2.91)
-2.27
(-0.44)
8.73
(1.62)
-3.82
(-0.51)
16.92**
(2.15)
3
0
122
-524.41
-0.67*
(-1.78)
Sa
-16.25
(-1.27)
1.28***
(4.47)
0.67
(0.42)
1.24
(0.37)
0.63
(1.01)
-5.64
(-0.72)
-5.47
(-0.66)
-10.00
(-1.22)
6.08
(0.72)
-0.90
(-0.08)
23.60*
(1.87)
4
3
118
-564.33
-0.03
(-0.06)
Note:Asymptotict-statisticsin parentheses.Resultsof two-tailedtests: "*p < .01; *p < .05; *p < .10.
stakes they have in downstreamoperationsshouldbe positive-are clearlynot
borne out in these tables. In Table 2, one finds basically no relationship
between the fees and the two measures of type. In Table 3, a relates negatively, not positively, to the measures of type, although in some cases the
effect is not significantlydifferentfrom zero. Results in Table4 also indicatea
negative ratherthan a positive relationshipbetween the franchisor'stype and
the stake the franchisorinitially has in downstreamoperations.Finally, Table
5 shows the partial correlations(rank correlationsin parentheses)between
firms' royalty rates and their reliance on company-ownership.The first column refers to the correlationsholding Typel constant, while the second column shows the same correlationswhen Type2 is held fixed. According to
hypothesis C2, conditional on the type of the firm, these correlationcoeffi-
274 The Journalof Law,Economics,& Organization,
V9 N2
Table 5. Correlationand Rank-Correlation
Coefficients
p(r,&|T)
p(r,a31T)
p(r,aalT)
Typel
Type2
0.19**
(0.17)*
0.23***
(0.18)**
0.04
0.19**
(0.16)*
0.23***
(0.17)*
0.04
(0.05)
(0.06)
Note:Rankcorrelationcoefficientsare shown in
parentheses.The significancetests are performedas
follows:
(i) Undernormality,and underHo: p = 0, we have
(rVr
/)/(Vl - r2) .tn-2,wherer is the
ordinarycorrelationcoefficient,and n is the
numberof observations(here,n = 125).
-. N(0,1).whereR is
(ii) Forn > 30, z = RVn
the rankcorrelationcoefficient.Note thatthis test
does not requireany distributional
assumption.
Resultsof two-tailedtests: "*p < .01;"p < .05; p < .10.
cients should be negative. But in fact the coefficients are positive, indicating
that franchisorsdo not tradeoff royalty rates and the companyownershipof
outlets.29Overallthen, one really finds no supportin these datafor the notion
thatfranchisorsuse the termsof theirfranchisecontractsand/ortheircontract
mix as a way to signal their quality to potentialfranchisees.
Unfortunately,these negativeresultscannoteasily be reinterpretedin a way
that would supportsome alternativetheory of franchising.The reasonfor this
is that no other theory has implicationsconcerningthe effect of a firm'stype,
or ex post growth, on initial choices of royalty rates, franchise fees, and
contractmix. However, some of the results relative to control variablesbear
on these alternativetheories. We now turn to these results.
5.2 The Effects of ControlVariables
The measures of franchisee and franchisormoral hazard, respectively the
numberof states in which a franchisorhas establishedoutlets and the number
of days of trainingfor new franchisees, are availablefor only 66 of the 125
franchisors.For this reason, these variablesarenot includedin the regressions
presentedin Tables2, 3, and 4. To determinehow sensitive the results might
be to the exclusion of these variables, the regressionswere reestimatedwith
these two variablesfor the subsampleof 66 firmsfor which this information
was available. The results are found in AppendixC in TablesCl, C2 and C3.
They show thatthese two variableshave a significanteffect only on the actual
proportionof company-ownedoutlets and the actual stake the firm has in
29. When the calculations were done holding fixed not only the type, but also the other
variables used in the regressions, the correlationcoefficients were still positive or nil.
Contractual
as SignalDevices 275
Arrangements
downstreamoperationsthree or four years after it becomes involved in franchising. And the effects that are significantare consistentwith doubled-sided
moral hazard. For all other dependentvariables, a joint test that the coefficients are zero could not be rejected. Furthermore,there is no case where the
inclusion of these variablesaffects the resultsrelativeto the measuresof type.
In thatsense, Tables2, 3, and 4 providea meaningfulset of resultsrelative to
the signaling hypothesis despite the absence of these control variablesin the
regressions.
In terms of the input sales variable, one finds that it either has no effect or
has a positive effect on royaltyrates, franchisefees, and the use of companyownership.None of these is consistentwith the notionthatthese sales areused
as an alternativesource of revenues, or an alternativesignal, by franchisors.30
As noted earlier,one problemwith this variableis thatit containsno information as to what proportionof the sales from franchisorsto franchiseesis based
on a contractualrequirement,and what proportionis voluntaryon the part of
the franchisee.Given antitrustrestrictionson tying, andespecially the Court's
decision in Siegel et al. v. ChickenDelight, Inc., 448 F.2d 43 (9th Cir. 1971),
it is likely that a large proportionof the observed sales from franchisorsto
franchisees are voluntaryratherthan compulsory.31The fact that franchisors
cannot charge a price above the competitive price for voluntarysales could
explain the results obtainedhere for this variable.32.33
30. In Lafontaine(1992a), input sales, measuredin thousandsof dollars per franchise, were
found to have a negative effect on royaltyratesand franchisefees (thoughthe observedeffects on
the use of company ownershipwere similar to those obtainedhere). In the presentarticle, input
sales are measuredas a proportionof averagesales per franchisedstore. Lafontaine(1992a)uses a
variable defined as one minus this proportionas a measure of franchisee'sjurisdiction in the
outlet. The effects of this variableon the franchisefee and the royaltyrate areconsistentwith the
results obtainedin the presentpaper.
31. For a discussion of tying and franchising,see for example Hunt and Nevin (1975), Klein
and Saft (1985) and Justis and Judd (1989:147-150). There is some evidence that the antitrust
treatmentof tying in the U.S. has affected franchisors'propensityto requirefranchiseesto buy
inputs from them. In particular,one finds that while U.S. franchisorstypically resortto quality
standardsand lists of approvedsuppliers, franchisorsin the U.K., where the approachtowards
vertical restrictionssuch as tying has been more "open-minded"(George, 1990:144),use tying
much more often and to a muchgreaterextent. (See Dnes (1992) for a numberof franchisingcase
studies and Mendelsohn(1990) for a descriptionof franchisingin the U.K.) Interestingly,in their
survey of franchisees,Hunt and Nevin (1975) found thatabout53 percentof franchiseesbelieved
the prices they paid for the inputs they bought from their franchisorswere at or below market
prices. The other 47 percent thoughtthese prices were above marketprices.
32. I thank the editor, Alan Schwartz, for making this point. Consistentwith this argument,
Sen (1993) finds that franchisorswho requirefranchiseesto buy certain inputs from them (captured by a dummy variable)have lower fees than those who do not.
33. There might also be an efficiency argumentto be made for franchisorsopting for royalty
rates over input mark-ups:if the technology downstreamis not one of fixed proportions,franchisees will substituteaway from inputsthey must buy at a relativelyhigherprice. Royalty rates,
on the other hand, will not lead to such distortions in franchisees' input buying behavior.
However, the observationthat U.K. franchisorsrely much more on input sale requirementsthan
U.S. franchisorsdo points toward antitrustrestrictionsratherthan efficiency as the main reason
for U.S. franchisors'limited use of input sales requirements.
of Lavw.
&Organization,
V9N2
276 TheJournal
Economics,
In a signalingframework,the numberof years in businesspriorto franchising could providepotentialfranchiseeswith informationaboutthe type of the
firm, with the franchisorthat has been in business for a longer period being
more likely to be a high-value franchisor.In that case, high-type franchisors
would not need to signal their type through royalties or company-owned
outlets. This would imply that the number of years in business prior to
franchisingwould have a negative effect on royalty rates and on the use of
company-ownedoutlets, and a positive one on the franchise fee. But in all
tables, one typically finds a positive and significantrelationshipbetween the
numberof years in business priorto franchisingand the use of royaltiesor the
use of company ownership. Hence results relative to this variable are also
inconsistent with the signaling explanationfor franchising.34However, they
are supportiveof the interpretationof years in business as a measure of the
value of the trade name in the context of franchisormoral hazard.
As for the othercontrolvariables, one finds firstthatthe risk variablehas a
positive effect (though generally not an effect that is significantly different
from 0) on the royalty rates and on the use of company ownership, and a
negative effect on the franchise fees. These results, which to some extent
supportthe notion that franchisorsprovide insuranceto their franchisees, are
consistent with results from Baneji and Simon (1991), but opposite to those
found in Lafontaine(1992a)and Martin(1988). Why this is so remainsunclear
at this point. Second, as in Lafontaine(1992a), one finds thata higheramount
of capital required implies both a higher franchise fee and more company
ownership.While the first effect is consistentwith the notion that franchisors
use franchisingas a source of capital, the second effect is not. Brickley and
Dark (1987) argue that this second effect can be explained by the increased
contracting costs related to higher capital requirements:When franchisees
must invest heavily in firm-specific assets, they face more risk and hence
require a higher level of compensation. This would make franchising less
attractivefor firms whose concepts call for large capital expenditures.However, one way to reducethe risk to franchiseesis to ask for a smallerfranchise
fee. Yet as we just saw, franchisorsincreasethe franchisefee when the capital
requiredto open an outlet goes up.
5.3 Conclusions
The resultspresentedabove do not supportthe notion thatfranchisorsuse the
terms of their franchise contracts(royalty rates and franchisefees) or their
capacity to operate outlets directly, or a combination of the two, to give
34. These resultsare the oppositeof those reportedfor this variablein Lafontaine(1992a).The
difference is attributableto the fact that the franchisorsincluded herein are all just beginningto
franchise.As a result, in the presentarticle, the numberof years in businesscapturesthe effect of
the "%time not franchising"variableused in Lafontaine(1992a), thatis, some notionof the value
of the business formatfranchise. In other words, the longer the firmshave been in business, the
more time they have had to develop their business format, leading them potentially to ask for
higher royalty payments. Also, these "older" firms are more likely to have opened a larger
numberof outlets, outlets that by definitionare company-owned.
Contractual Arrangements as Signal Devices
277
informationto potential franchisees about the value of their franchise. Not
only do the results related to the measures of type used here lead to this
conclusion, but so do two otherempiricalpatterns:(i) the fact thatthe termsof
the contractsdo not vary significantlyover the five or more years separating
the two observationson each firm (as noted in Gallini and Lutz (1992), in a
signaling model, one would expect the royalty rates to decrease and the
franchisefees to increaseover time); and (ii) the fact thatthe numberof years
in business priorto franchisinghas a positive effect on royaltyratesand on the
use of company-ownedoutlets, while undersignaling these effects should be
negative.
The existing empiricalliteratureon franchisingfurthersupportsthis "negative" conclusion vis-a-vis the signaling hypothesis in that it suggests that
incentives issues and monitoring costs are a central issue in franchising.
Brickley and Dark (1987), Lafontaine (1992a), Minkler (1991), and Norton
(1988) all have found results suggesting thatthe termsof the contract,as well
as the decision to franchise or operate directly, are influenced by factors
related to franchisee and/or franchisorincentives.35
All this is not to say thatfranchisorsdo not have an informationaladvantage
over potential franchisees when it comes to the value of their franchise or
trade name, for they clearly do. Gallini and Lutz (1992) documentthis fact
very well. But there are two main problemswith the signaling explanationin
the context of long-termrelationshipssuch as the ones involved here.
First, signaling via the termsof the franchisecontractimplies inefficiencies
and a lack of flexibilityfor the durationof this contract.Hence it is likely to be
very costly. Put differently,if royalty rates and franchise fees were used to
provide informationto franchisees about the fundamentalvalue of the franchise, and given thatfranchisorsget to choose the length of theircontract,one
would not expect the contracts to be set for such long periods. Similarly,
signaling through the contract mix can be very costly. In Gallini and Lutz
(1992), it involves a suboptimaleffortlevel being put into the developmentof
a particularlocation, the effects of which are irreversible.Withoutsuch high
costs, however, using the contract mix as a signal would not be credible:
Firms could modify their contractmix in a significantway once a numberof
contracts were signed. The main point here is that there likely exist more
efficient ways for franchisorsto reveal informationabout their quality level.
For example, membershipin the InternationalFranchiseAssociation is conditional on franchisorsmeeting certain criteriafor admission and on their respecting a particularcode of ethics. If the admissioncriteriafor this associa35. Gallini and Lutz (1992) presentsome evidence supportingthe signaling hypothesis. They
find that the proportionof company-owned stores decreases in newly established franchised
chains as the number of years since they began franchising increases-as it should if firms
initially used company-ownershipas a signal of their quality, and subsequentlyreduced their
reliance on this signal as informationabout their value became known. But as some of the
discussion in Section 4 suggests, these patternsshould be interpretedwith care in light of the fact
that new franchisors necessarily start out as 100 percent company-ownedchains. Given the
composition of their samples, this problem is likely to explain also, at least in part, the results
obtained in Lafontaine(1992a) and Martin(1988).
278 TheJournal
&Organization,
of Lavw,
V9N2
Economics,
tion were "discriminating"enough, then simply belonging to this association
might be taken as a credible way to convey information.36
Second, and more importantly,the kind of informationthe franchisee is
seeking when consideringbuying a franchiseis not limited to how good the
franchisoris now, and/or how valuable the trade name is now, but includes
informationabout how good they are likely to be in the future. To the extent
that the future worth of the outlet will depend on the franchisor'sfuture
behavior, the franchiseewill be looking for guaranteesin the contractabout
this future behavior. In this context, royalties on sales, and possibly the
company ownership of outlets, are used not as signals of an exogenously
given and fixed franchisorvalue, but rather as incentive mechanisms that
reassurethe franchiseeas to the futurebehaviorof the franchisor,and therefore as to the future value of the franchise.37
More generally,the "negative"resultobtainedhereinis interestingnot only
in what it tells us about franchising but also in what it reveals about the
circumstancesunder which signaling, and informationasymmetries,may or
may not provide appropriateexplanationsfor particularphenomena.Specifically, the results in this articlesuggest that even if there is evidence of asymmetric information,signaling cannot easily be used to explain the terms by
which long-termrelationshipsaregoverned,becausethe longertherelationship
is (a) the more likely it is that alternativecheaperinformationaldevices will
exist, and, more importantly,(b) the morelikely it is thatthe requiredinformation is not exogenously given and fixed but ratheris a function of the future
behavior of the informed party.38Hence one must look elsewhere, toward
monitoringcosts and otherorganizationalefficiency arguments,to explainthe
natureand use of various types of long-termcontractualarrangements.
Appendix A: Signaling Models of Franchising
This appendixdiscusses more precisely the signaling explanationin the context of franchising.The firstpartis concernedwith the morebasic model from
Desai and Srinivasan(1990), Gallini and Wright (1990), and Tirole (1988),
where all signaling is achieved throughthe terms of the franchisecontract,
namely royalty rate and franchise fee only. The second model, based on
Gallini and Lutz (1992), allows franchisorsto signal their types throughtheir
contractmix as well as the terms of their contracts. Since pooling equilibria
can be ruled out by an appropriatechoice of refinement,and, more importantly, since only separatingequilibria imply differences in firms' behavior
that can be tested empirically,only separatingequilibriaare discussed here.
36. Fromthe financeliterature,one finds evidence thatsuch alternativesexist and thatthey are
used: Balvers and colleagues (1988) find that firms' underpricingof new issues (the standard
signal) is significantlyreduced by an appropriatechoice of investmentbankerand auditor.The
reputationsof the bankerandauditorconvey informationaboutthe value of the stock such thatthe
issuing firm does not need to underpricethe offer as much.
37. See Rubin (1978), Lal (1990) and Bhattacharyyaand Lafontaine(1992) for this type of
argument, and Lafontaine(1992a) for empiricalevidence supportingit.
38. See Ippolito(1990) for a similarargumentconcerningsignaling in the contextof consumer
durables.
as SignalDevices 279
Contractual
Arrangements
A.1 SignalingThroughthe Termsof the FranchiseContract
Considera risk-neutralfranchisorwith no establishedreputation,who obtainsa
tradename of high (h) or low (1)value from a common knowledgeprobability
distribution, where the probabilityof getting a low-value trade name is 0.
Inversedemandat the retaillevel is given by p = p(q,T), whereT again stands
for the type of the franchisor,or, equivalently,for the value of the tradename.
Alternatively,demand may be written as ph(q) and pl(q), with ph(q) 2 p'(q)
for all q > 0.
A competitive marketfor potentialfranchisees,who are also risk neutral,is
assumedto exist. The franchisecontractc = (r,F) stipulatesa royalty rate on
sales, r E [0,1), and an up-frontfee, F - 0, such thatthe franchisor'srevenues
from a franchisedoutlet over the durationof the contractare given by
R(r,F) = r - pi(qi(c))q'(c) + F,
i = h,l,
where qi(c) is the franchisee's choice of output over the duration of the
contract, given that the contractis c and that the franchisoris of type i. The
franchisees are assumed to learn their franchisor'stype after signing their
contract but before choosing their action. It is assumed that the contract
cannot be renegotiated after the franchisee has observed the franchisor's
type.39 The franchisee'sprofits before payment of the franchisefee are
i(qi) = (1 - r)pi(qi(c))qi(c)- K,
i = h,l,
where K stands for the setup costs of the franchisee,which are assumedto be
the same whether the franchisor'stype is high or low. For notational simplicity, marginalcosts at the outlet level are assumedto be 0. The value of t7
is assumed strictly concave in qi.
Without asymmetric information, the contract that would maximize the
franchisor'sprofits would have each type extractingall of the franchisee's
profitsthroughthe fixed fee F. Any r > 0 would lead to a suboptimalquantity
being chosen by the franchiseesince the royalty rate increases marginalcost
(or reduces marginalrevenue)downstream.40With a fixed-fee contract,franchisees would be asked to pay a transfer price equal to the franchisor's
marginal cost, and hence they would choose the optimal qi. The franchise
contractswould be given by ch = (O,1Hh) and cl = (0,17') for the high- and
low-value franchisors,where 1hand 17 are the first-bestprofits of the outlets, before paymentof the fixed fee, when the tradenames arehigh- and lowvalue, respectively.
39. It must also be thatthe value of the tradename, althoughobservedex post by both parties,
is not verifiableby a thirdparty.Otherwise, the optimalex ante contractwould simply have to be
made contingent on this value to achieve the first best.
40. This is the standarddouble-marginalizationargument. It might be argued that in an
oligopolistic setting, positive royalty rates are beneficial to franchisorssince they reduce output
and hence bring firms closer to the collusive outcome. However, after franchisorshave signed
theircontractswith franchisees,they extracttheirprofitsfrom the sales levels of theirfranchisees.
As a result, they have no incentive to increase profits at the expense of lower sales.
280 The Journalof Lav/,Economics,& Organization,
V9 N2
With asymmetricinformation,in a separatingequilibrium,both types are
revealed. Given this, the low type's optimal contractis cl by definition. But
the franchisorwith the good trade name cannot offer ch since the franchisor
with a low-value tradename would also choose ch, and separationwould fail.
The "good"franchisormust offer a contractc = (r > ,F < 17h)thatsatisfies
the franchisee's participationconstraintand is less profitablethan cl for the
low-value franchisor.A necessaryconditionfor a set of separatingcontractsto
exist in this context is thatd(Rr(r,T))/dT> 0, whereRr is the derivativeof the
franchisor'srevenuesR with respect to the royalty rate r.
In order for the contractchosen by the high type to be unappealingto the
low type, it must be that
r p'(q'(c))q'(c) + F - max [0,171].
(Al)
Notice that the low-value franchisor'srevenues under c are necessarily nonnegative, since the franchisefee F is assumedto be nonnegativeand sales are
by definition nonnegative. Consequently,nH must begreater than 0 for a
separatingequilibriumto exist.41Hence, (Al) may be rewrittenas
r *pl + F <
0
(A2)
n,
where pl = pq(q'(c))and q = ql(e). If this condition is met, then the pair of
contracts(c,cl) signals perfectly.The franchisee,knowing that the firmoffering c has a high-value tradename, assigns an ex post probabilityg(c) = 1 to
the franchisorbeing a good type and acceptsthe contractif F < (1 - r)hh K. After eliminatingdominatedstrategies,42the unique separatingequilibrium contractfor the high-value franchisor,c* = (r*,F*), solves
max [r ph(qh(c))qh(c)+ F],
r,F
subject to
r p'(ql(c))
q(c) + F =
7l'
(A3)
K - F - 0.
(A4)
and
(1
-
r)ph(qh(c))
-
qh(c)
-
Notice thatfrom (A3) and the fact thatroyaltypaymentsare nonnegative,the
franchise fee F* in c* can never be greaterthan 71.43
41. In other words, low-value franchisesmust still be profitable,otherwise separationwould
never occur. As a result, an empiricaltest of this model cannot be based on a definitionof types
that is derived from a notion of survival.
42. See Tirole (1988: chap. 11)for an introductionto the refinementsliterature,and Cho and
Kreps (1987) and the references thereinfor more details.
43. Also, the franchisorwith type h must preferc* to c'. But that is implied by (A3) and the
fact that r.ph(qh(c))qh(c) > r-pl(q'(c))q1(c).
ContractualArrangementsas Signal Devices 281
The solution to the above programgives two schedules, one for r and one
for F, as functions of the franchisor'stype T: For r, we find that rh > rl = 0;
for F, thatFh < F = 171.Extendingto a continuumof types, the schedule for
r would be increasingin T, with the absolutelowest choosing r = 0, while the
schedule for F would be decreasingin T.
A.2 SignalingThroughthe ContractTermsand the ContractMix
The above discussion focused on a single franchisordesigning a contractfor a
single franchisee. In reality, franchisorshave a numberof outlets and while
they tend to franchisemost of them, they also generallyoperatesome of them
directly.I assume that in company-ownedoutlets the franchisorkeeps control
over q (and thereforep), and a hired manager, who is paid a fixed wage,
supervises the outlet.44Note that with a competitivemarketfor such managers, the fixed wage to be paid to each managerwill be independentof T.
It is necessary to impose a cost relatedto companyownership,for example
a higher capital cost, to prevent firmsfrom operatingall of their units in this
context.45This is because the problemsrelatedto asymmetricinformationare
assumednot to occur in a verticallyintegratedfirm.46Since signalinginvolves
costly distortions, without some sort of cost attachedto company operation,
high-value firms in this model would find it in theirbest interestto operateall
of their outlets directly (while low-value franchisorswould be indifferent
between franchisingunder cl and operatingdirectly).
Assume the total numberof units in the chain, n, is fixed, and define ai as
the proportion of company-operatedstores in a given franchise chain. A
separatingequilibriumin this case could be characterizedby dh > 0 and a
contractc = (r,F) for franchisorswith a high-qualitytradename, and a' = 0
and c = cl for the others. Low-qualityfranchisorswould not want to mimic
this offer as long as, for a given dh,
dh ' (H71- Z1) + (1 - dh)(r * plI q + F)
I71,
where n7- > 0 as before, andZ', i = h,l, representsthe increasedcapitalcost
associated with company operation for the high- and low-value franchisor,
respectively. To ensure that a separatingequilibriumwould exist if signaling
was achieved strictly throughthe contractmix, I assume that it is less costly
for the high type to increasea than it is for the low type. Combinedwith the
sorting condition imposed above, this also guarantees the existence of a
44. What is necessary is that the franchisorbe able to dictate the quantityto be producedin
company-operatedoutlets, and that the manager'srevenues not depend on the "quality"of the
franchise. One way to guaranteethatmanagersdo not care aboutthe qualityof the franchiseis to
offer them a fixed wage.
45. Gallini and Lutz (1992) do not need to introducea highercapitalcost in theirmodel since
they argue that the cost of company ownershipis the inefficiency introducedby managerswho
cannot appropriateincreasesin the resale value of theiroutlet that are due to their good management. As a result, they pick a lower level of effort than franchisees do. But since I do not
incorporatethe choices of effort level in the following discussion, it is simplerto thinkin termsof
a capital cost.
46. This assumptionis quite customary;see, for example, Crocker(1983).
282 TheJournal
of Law,Economics.
&Organization,
V9N2
separatingequilibriumwhen a combinationof r and a is used to signal the
franchisor'stype. The contractoffered by the high-qualityfranchisorwould
again have to satisfy the franchisee'sparticipationconstraint,that is,
(1 - r)h.
qh
- K - F.
For a given ah, the set of separatingequilibriaimpliedby these conditionscan
be reduced to a singleton through the elimination of dominatedstrategies.
This unique separating equilibrium contract cah max{ah(Inh - Zh) + (1 - ah) [r . ph(qh(c))
r,F
(rah,Fah)
is the solution to
qh(c) + F]},
subject to
ah(7l
- Zl) + (1 -
ah)
[r *p(ql(c))
q(c) + F] = IIn
(A5)
and
(1 - r)ph(qh(c))qh(c) - K
'
F.
(A6)
The separatingequilibriummight take one of two forms in this case. First,
there exists an a such that c = ch when ah E [d,l]. In other words, all
signaling can be achieved throughthe contractmix ratherthan the contract
terms. In this case, we have that ah = a and a' = 0, so that ah > a'.
Extendingto a continuumof types again, if signaling is a considerationin the
choice of contractmix, and all signaling is achieved throughthe choice of
contractmix, then we should find that a is increasingin T.
The second type of equilibriumoccurs for values of ah such that0 < ah <
a. In this case, r > 0 becomes necessaryagain to achieve separation.In other
words, high-valuefranchisor'suse a combinationof r and a to signal his type.
With this kind of equilibrium,the signal is the franchisor'sstake (S) in the
whole chain. If the contract mix and/or royalty rates are used as signaling
devices, we should find that this stake increases with T. In addition, for a
given value of T, r and a should be negatively correlated.
Appendix B: Data Sources and Variable Definitions
As noted above, the data used in this paper are obtained primarely from
EntrepreneurMagazine's yearly "Franchise500" surveys. These surveys give
informationon, among otherthings, royaltyrates, advertisingfees, and franchise fees for all franchisorssurveyed each year.47Because the royalty rates
and franchise fees used in the empirical analyses should represent all the
47. At a point in time, the royalty rate and franchisefee demandedby a franchisortend to be
the same for all potentialfranchisees.This is not clear from the Entrepreneur'ssurveys:Some of
the franchisorssay that their fees vary, while others give ranges. Lafontaine(1992a) analyzes
these statementsand concludes thatthe fees arerelativelyconstantacross franchiseesat a point in
time. In this article, cases where fees were said to vary were eliminatedfrom the data since one
cannot assign fees to these firms. However, averages were used for those that gave a range.
Contractual
as SignalDevices 283
Arrangements
variable and fixed payments from the franchisee to the franchisor,respectively, advertisingfees given as a percentageof sales are addedto the stated
royaltyrate to generatethe notion of royaltyrateused in this article. When the
royalty payments and/or the advertisingfees are given as fixed amountsper
time period, they are discountedand addedto the franchisefee.48 Finally, all
franchisefees are in 1980 U.S. thousandsof dollars.49Royalty ratesand other
proportionsare in percentages.
In terms of the proportionof company-ownedstores and the stake the
franchisorhas in downstreamoperations, both aa and Sa are measuredtwo
years before the end of the five-year-plusperiod separatingthe two observations on each firm. This is because in each survey year t, the number of
franchisedand company-ownedoutlets is given for years t, t - 1, and t - 2.
Since I observe each firmin two surveys at least five years apart,I can use the
data for t - 2 in the second observationto constructaa andSa. By so doing, I
eliminate the need to observe the firm in another survey between the two
original ones, which would reduce my sample size. The fact that the number
of years elapsed between the beginningof franchisingand the time I observe
aa andSa varies a little across firms(the numberof years is eitherthreeor four
for 120 firmsout of 125) is not a problemin the regressionssince I controlfor
the year in which the firm began franchising.Note that the royaltyrate in the
final survey is used in combinationwith a, to generateS,
Finally, some of the data used in the analyses are obtained on a sectoral
basis from the U.S. Dept. of Commerce'sFranchisingin the Economy(U.S.
DOC, 1988). This includesdataon inputsales from franchisorsto franchisees,
which are measuredas the total amountof such sales dividedby total sales of
franchised outlets in the sector. This source also provides discontinuation
rates, and the average sales level of franchisedand company-ownedstores.
The 14 business formatfranchisingsectors definedby the DOC are as follows:
AutomotiveProductsand Services;Business Aids and Services;Construction,
Home Improvement,Maintenance,and CarpetCleaning;ConvenienceStores;
EducationalProductsand Services; Restaurants;Hotels, Motels, and Campgrounds;Laundryand DrycleaningServices; Recreation,Entertainment,and
Travel; Rental Services (Auto-Trucks);Rental Services (Equipment);Nonfood Retailing; Food Retailing, Non-convenience;Miscellaneous.
In the case of two sectors, Business Services and Restaurants,the average
sales data are also broken down by subsectors. For Business Services, there
are six subsectors: Accounting and Collection Services, Employment Services, Printing and Copying, Tax Preparation,Real Estate, and Miscella48. In reality, fixed up-frontfees and ongoing fixed paymentsare different:The obligationto
pay the latterdisappearsif the outlet goes out of business duringthe contractperiod. Since failure
rates are low in franchising [according to the U.S. Dept. of Commerce (1988), the numberof
discontinuedoutlets in 1986 was 7,934 out of 246,664 outlets, for a discontinuationrate of 3.2
percent per year], and ongoing fixed payments are also uncommon (4 cases out of 125 in the
present sample), I disregardthis difference. The discount rate used was 10 percent (fees are in
nominal dollars). Given the small numberof cases where ongoing fixed paymentsoccur, results
are not sensitive to these data manipulations.
49. Some of the franchisorsin the sample are Canadian.Theirfranchisefees were transformed
to $U.S.
284 The Journalof Law,Economics.& Organization,
V9 N2
neous. For Restaurants,there are eight subsectors: Chicken; Hamburgers,
Franks, Roast Beef; Pizza; Mexican; Seafood; Pancakes and Waffles;Steak
and Full Menu; and Sandwiches and Others.
Finally, because there were often only a few firmsin the sample in each of
the sectors, the dummy variables identify only six separatesectors, namely:
Automotive Services, Business Services, Construction and Maintenance,
Restaurants,Non-food Retailing, and Non-convenienceFood Retailing. The
other firms form a miscellaneous sector.
Appendix C: Tobit Results
for the Subsample of 66 Franchisors
Table C1. Tobit Results for RoyaltyRates and FranchiseFees
Variable
Royalty
Type1
-1.51
(-0.67)
Franchise
Fee
Capital required
Discontinuations
Inputsales
Firstfranchised in
1980
1981
1982
1983
1984
Days of training
Numberof states
Constant
LimitObservations
Non-LimitObservations
Log-likelihood
Hetero.Coeff. (8)
0.04
(0.53)
0.84
(0.90)
0.74
(1.54)
0.14*
(1.76)
-0.54
(-0.39)
-1.81
(-1.03)
1.89
(1.56)
-0.87
(-0.70)
-0.59
(-0.43)
-0.01
(-0.16)
-0.01
(-0.24)
6.45***
(3.33)
2
64
-157.75
-2.54**
(-2.03)
Franchise
Fee
-0.90
(-0.16)
Type2
Years in business
Royalty
-0.19
(-1.43)
10.21**
(6.14)
-3.62***
(-2.86)
-0.18
(-0.86)
-1.43
(-0.50)
6.89
(1.66)
-2.43
(-0.79)
-5.00*
(-1.65)
4.31
(1.14)
0.04
(0.44)
-0.04
(-0.29)
19.05***
(3.83)
1
65
-218.86
-0.78
(-1.04)
-0.12
(-0.06)
0.12**
(2.08)
-0.50
(-0.86)
0.97*
(1.87)
0.17**
(1.96)
0.77
(0.65)
-2.10
(-1.22)
2.60**
(2.01)
0.62
(0.52)
0.05
(0.03)
0.01
(0.40)
-0.01
(-0.27)
4.25**
(2.11)
2
64
-160.64
-0.17
(-0.20)
-5.58
(-1.13)
-0.20
(-1.46)
10.16**
(6.18)
-3.41**
(-2.70)
-0.18
(-0.89)
-1.49
(-0.53)
6.61
(1.61)
-2.13
(-0.70)
-4.50
(-1.55)
4.39
(1.20)
0.04
(0.49)
0.01
(0.04)
18.34***
(3.70)
1
65
-218.23
1.10
(1.39)
Note:Asymptotict-statisticsin parentheses.Resultsof two-tailedtests: "*p < .01; "p < .05; *p < .10.
Table C2. TobitResults for the Use of Company-OwnedOutlets
Variable
Type1
a3
-0.93
(-0.93)
-12.83
(-0.94)
a
-51.28***
(-3.51)
Type2
Years in business
Capital required
Discontinuations
Inputsales
Firstfranchisedin
1980
1981
1982
1983
1984
-0.
0.07***
(2.95)
0.21
(0.86)
0.15
(0.68)
0.09**
(2.53)
-0.92*
(-1.82)
-0.41
(-0.57)
-0.30
(-0.58)
0.25
(0.48)
-0.45
(-0.70)
1.17***
(3.57)
3.60
(1.11)
2.86
(0.97)
1.18**
(2.38)
-13.16*
(-1.93)
-7.30
(-0.76)
-4.93
(-0.69)
4.10
(0.59)
-5.55
(-0.63)
1.16***
(3.27)
-2.22
(-0.55)
2.51
(0.79)
0.57
(1.19)
(-1.
0
(2
0
(0
0
(0
0
(2
6.64
(0.80)
-0.13
(-0.01)
4.69
(0.60)
23.65***
(3.65)
25.86***
(2.99)
-0.
(-1.
-0.
(-0.
-0.
(-0.
0
(0
-0.
(-0.
Table C2. (Continued)
Variable
Days of training
Numberof states
Constant
LowerLimitObservations
Upper LimitObservations
Non-LimitObservations
Log-likelihood
Hetero.Coeff.(,)
a
-0.00
(-0.34)
-0.00
(-0.06)
-0.01
(-0.02)
11
0
55
-97.77
-0.55
(-0.82)
aa3
-0.08
(-0.42)
0.09
(0.30)
-2.50
(-0.22)
11
0
55
-238.27
-0.90
(-1.31)
0.39*
(1.77)
-0.86*
(-1.86)
-3.88
(-0.32)
12
1
53
-236.78
-3.24***
(-3.62)
Note:Asymptotict-statisticsin parentheses.Resultsof two-tailedtests: "*p < .01; **p< .05; 'p < .10.
-0.
(-0.
0
(0
-0
(-0
11
0
55
-97
-0
(-1
Contractual Arrangements as Signal Devices
287
Table C3. TobitResults for the Stake of the Franchisorin DownstreamOperations
Variable
Typel
S3
-12.53
(-0.83)
Sa
Capital required
Discontinuations
Inputsales
Firstfranchised in
1980
1981
1982
1983
1984
Days of training
Numberof states
Constant
LowerLimitObservations
Upper LimitObservations
Non-LimitObservations
Log-likelihood
Hetero.Coeff. (8)
0.95***
(2.63)
3.16
(0.86)
4.97
(1.54)
0.74
(1.37)
-14.89*
(-1.95)
-19.52*
(-1.79)
-8.82
(-1.09)
0.82
(0.10)
-13.82
(-1.40)
0.03
(0.15)
0.26
(0.77)
10.87
(0.88)
Sa
-10.42
(-0.80)
0.94***
(2.61)
3.32
(0.90)
5.32
(1.63)
0.76
(1.40)
-11.39
(-0.81)
1.40***
(3.38)
-5.33
(-0.92)
6.44*
(1.70)
0.46
(1.09)
-15.37**
(-2.02)
-20.41 *
(-1.87)
-9.17
(-1.13)
-0.06
(-0.01)
-15.03
(-1.54)
0.06
(0.30)
0.37
(1.08)
9.86
(0.79)
20.36**
(2.38)
-28.29**
(-2.18)
31.14**
(3.26)
31.26***
(4.72)
45.09***
(4.62)
1.14***
(4.59)
-0.13
(-0.59)
-12.38
(-0.94)
-61.79***
(-3.70)
Type2
Years in business
S3
0.99**
(2.59)
-1.72
(-0.39)
6.73**
(1.99)
0.32
(0.67)
13.35
(1.57)
-27.26**
(-2.03)
10.90
(1.33)
28.42***
(4.29)
26.21**
(2.78)
0.57**
(2.53)
-0.65
(-1.48)
4.42
(0.37)
1
1
1
0
65
-281.71
-0.88
(-1.38)
1
0
64
-282.67
-4.12***
(-3.67)
65
-281.73
-0.76
(-1.17)
1
1
64
-293.22
-5.39***
(-2.67)
Note: Asymptotic t-statistics in parentheses. Results of two-tailed tests: "'p < .01; "p < .05; *p < .10.
References
Balvers,RonaldJ., BillMcDonald,andRobertE. Miller.1988."Underpricing
of NewIssuesand
the Choiceof Auditoras a Signalof Investment
BankerReputation,"
63 Accounting
Review
605-22.
Vs. Ownership:
A Contracting
Baneji, Shumeet,andCarolSimon.1991."Franchising
Explanation,"Universityof Chicago,mimeo.
10International
Beggs,A. W. 1992."TheLicensingof PatentsunderAsymmetric
Information,"
Journal of Industrial Organization171-91.
MoralHazardandthe
Bhattacharyya,
Sugato,andFrancineLafontaine.1992. "Double-Sided
Natureof ShareContracts,"
Universityof Michigan,School of BusinessAdministration,
Divisionof Research,Working
PaperNo. 698.
of Law,Economics,
288 TheJournal
&Organization,
V9N2
Bond, RobertE. 1989. TheSourceBook of FranchiseOpportunities.Homewood, Ill.: RichardD.
Irwin Publishing Co.
Brickley,JamesA., and FrederickH. Dark. 1987. "TheChoice of OrganizationalForm:The Case
of Franchising,"18 Journal of Financial Economics 401-20.
Caves, RichardE., and William F. MurphyI. 1976. "Franchising:Firms, Markets,and Intangible Assets," 42 (April) SouthernEconomicJournal 572-86.
Cho, In-Koo, and David M. Kreps. 1987. "SignalingGames and Stable Equilibria,"102 Quarterly Journal of Economics 179-221.
Ross. 1984. "Prices, ProductQualitiesand AsymmetricInforCooper, Russell, and Thomas VW.
mation:The Competitive Case," 51 Review of Economic Studies 197-207.
Crocker,Keith C. 1983. "VerticalIntegrationand the StrategicUse of PrivateInformation,"14
Bell Journal of Economics 236-48.
Curry,David J., and PeterC. Riesz. 1988. "Pricesand Price/QualityRelationships:A Longitudinal Analysis," 52 Journal of Marketing36-51.
Desai, Preyas, and KannanSrinivasan.1990. "A ChannelManagementIssue:New Franchisingin
the Presence of Two-Sided InformationAsymmetry,"Carnegie Mellon University, mimeo.
Dixon, E. L., Ed. 1986, 1987. The FranchiseAnnual. St. Catherines,Ontario:Info Press Inc.
Dnes, Antony W. 1992. Franchising: A Case-study Approach. Aldershot, England: Avebury.
Downes, David H., and RobertHeinkel. 1982. "Signalingand the Valuationof UnseasonedNew
Issues," 37 Journal of Finance 1-10.
EntrepreneurMagazine. Various years. Annual Franchise 500 issues: February,1981;January
issues of 1982-1989.
Gallini, Nancy T., and Nancy A. Lutz. 1992. "Dual Distributionand Royalty Fees in Franchising," 8 Journal of Law, Economics, & Organization471-501.
, and Brian D. Wright. 1990. "TechnologyTransferUnderAsymmetricInformation,"21
RAND Journal of Economics 147-60.
Garvin, David. 1983. "Qualityon the Line," 61 HarvardBusiness Review 65-75.
George, Ken. 1990. "UK CompetitionPolicy: Issues and Institutions,"in A. Jacquemin,ed.,
CompetitionPolicy in Europe and North America. Chur, Switzerland:Harwood Academic
Publishers.
Gerner, Jennifer, and W. Keith Bryant. 1981. "ApplianceWarrantiesas a MarketSignal?" 15
Journal of ConsumerAffairs 75-86.
Gerstner, Eitan. 1985. "Do Higher Prices Signal Higher Quality?"22 Journal of Marketing
Research 209-15.
Hallagan, William. 1978. "Self-Selectionby ContractualChoice and the Theory of Sharecropping," 9 Bell Journal of Economics 344-54.
Hunt, Shelby D., and John R. Nevin. 1975. "TyingAgreementsin Franchising,"39 Journal of
Marketing20-26.
Ippolito, Pauline M. 1990. "Bondingand NonbondingSignals of ProductQuality,"63 Journalof
Business 41-60.
Jovanovic, Boyan. 1982. "Selection and the Evolutionof Industry,"50 Econometrica649-70.
Justis, Robert, and RichardJudd. 1989. Franchising. Cincinnati,Ohio: South WesternPub. Co.
Kaufmann,PatrickJ., and FrancineLafontaine.1994. "Costsof Control:The Source of Economic Rents for McDonald's Franchisees,"forthcomingin Journal of Law and Economics.
KihlstromRichardE., and Michael H. Riordan. 1984. "Advertisingas a Signal," 92 Journalof
Political Economy 427-50.
Klein, Benjamin, and Lester F. Saft. 1985. "The Law and Economics of Franchiselying Contracts,"28 Journal of Law and Economics 345-61.
Lafontaine, Francine. 1992a. "Agency Theory and Franchising:Some EmpiricalResults," 23
RAND Journal of Economics 263-83.
. 1992b. "How and Why Do FranchisorsDo What They Do: A Survey Report," in
P. Kaufmann,ed., Franchising:Passportfor Growth& Worldof Opportunity(Sixth Annual
Proceedingsof the Society of Franchising,InternationalCenterfor FranchiseStudies, University of Nebraska).
Lal, Rajiv. 1990. "ImprovingChannelCoordinationthroughFranchising,"9 MarketingScience
299-318.
ContractualArrangementsas Signal Devices 289
Leland, Hayne E., and David H. Pyle. 1977. "InformationalAsymmetries, FinancialStructure
and Financial Intermediation,"32 Journal of Finance 371-87.
Maddala,G. S. 1983. LimitedDependentand QualitativeVariablesin Econometrics.Cambridge:
CambridgeUniversity Press.
Martin, Robert E. 1988. "Franchisingand Risk Management,"78 AmericanEconomic Review
954-68.
Mathewson, G. Frank,and RalphA. Winter. 1985. "TheEconomics of FranchiseContracts,"28
(October)Journal of Law and Economics 503-26.
Matthews, Steven, and John Moore. 1987. "MonopolyProvision of Quality and Warranties:An
Explorationin the Theory of MultidimensionalScreening," 55 Econometrica441-67.
Mendelsohn, Martin. 1992. "UnitedKingdom,"in M. Mendelsohn,ed., Franchisingin Europe.
London: Cassell PublishersLtd.
Milgrom, Paul, and John Roberts. 1986. "Priceand AdvertisingSignals of ProductQuality,"94
Journal of Political Economy 796-821.
Minkler, Alanson P. 1990. "An Empirical Analysis of a Firm's Decision to Franchise,"34
Economics Letters 77-82.
Norton, Seth WV.1988. "An Empirical Look at Franchisingas an OrganizationalForm," 61
Journal of Business 197-217.
Priest, George L. 1981. "A Theory of the ConsumerProductWarranty,"90 YaleLaw Journal
1297-1352.
Rubin, Paul H. 1978. "TheTheory of the Firm and the Structureof the FranchiseContract,"21
Journal of Law and Economics 223-33.
Sen, KabirC. 1993. "The Use of Initial Fees and Royalties in Business FormatFranchising,"14
Managerial and Decision Economics 175-90.
Shepard, Andrea. 1993. "ContractualForm, Retail Price and Asset Characteristicsin Gasoline
Retailing," 24 RANDJournal of Economics 58-77.
Spence, A. Michael. 1973. "JobMarketingSignaling," 87 QuarterlyJournalof Economics 35574.
. 1974. Market Signaling: InformationalTransferin Hiring and Related ScreeningProcess. Cambridge,Mass.: HarvardUniversity Press.
Stiglitz, JosephE. 1974. "Incentivesand Risk-Sharingin Sharecropping,"41 Reviewof Economic
Studies 219-55.
Tirole, Jean. 1988. The Theoryof Industrial Organization.Cambridge,Mass.: MIT Press.
U.S. Departmentof Commerce. Variousyears. Franchise OpportunitiesHandbook(compiledby
AndrewKostecka).Editions:October1985, November 1986, January1988. Washington,D.C.:
U.S. GovernmentPrintingOffice.
. 1988. Franchisingin the Economy,1986-88 (preparedby AndrewKostecka).Also other
editions. Washington,D.C.: U.S. GovernmentPrintingOffice.
Venture. Variousyears. The Franchise 100.
Vermaelen, Theo. 1981. "Common Stock Repurchasesand Market Signalling," 9 Journal of
Financial Economics 139-83.
Wiener, JoshuaL. 1985. "AreWarrantiesAccurateSignals of ProductReliability?"12Journalof
ConsumerResearch 245-50.
Wilson, Robert. 1985. "Multi-dimensionalSignaling," 19 Economics Letters 17-21.
Wolinsky,Asher. 1983. "Prices as Signals of ProductQuality,"50 Review of Economic Studies
647-58.