Вы находитесь на странице: 1из 16
Theoretical Modeling in Marketing Author(s): K. Sridhar Moorthy Source: The Journal of Marketin g, Vol.

Theoretical Modeling in Marketing Author(s): K. Sridhar Moorthy Source: The Journal of Marketing, Vol. 57, No. 2 (Apr., 1993), pp. 92-106 Published by: American Marketing Association

Accessed: 23/08/2010 18:39

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=ama.

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 service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

information about JSTOR, please contact support@jstor.org. American Marketing Association is collaborating with JSTOR

American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Marketing.


K. Sridhar Moorthy

Theoretical Modeling



Over the last 10 years or so, theoretical modeling has rapidly become an

marketing. To many people,

theoretical modeling. The author argues that even though theoretical modeling is quantitative, it is closer

to behavioral marketing in purpose and methodology than to quantitative decision

Whereas behavioral

periments. Using

support modeling.

marketing involves empirical experiments, theoretical modeling involves logical ex-

important style of research in

an attempt at explaining

however, this style is still a mystery. This article is

this framework, the author addresses such issues as the internal and external validity

of model-based theories. The

of theoretical models, the purpose of theoretical modeling, and the testing

agency theory explanation

of salesforce compensation is

used as a case study.

N essentiallynew style of researchhas sprungup in marketingrecently: mathematicaltheoretical

modeling.' Scarcely an issue of Marketing Science

style. Some examples Staelin (1983), Moor-

passes withoutan articlein this arethe articles by McGuireand

thy (1984), Basu et al. (1985), Mahajan and Muller (1986), Hess and Gerstner (1987), Hauser (1988),

Wilson and Norton (1989), and Rao (1990). Lately,


and the term "theoretical modeling" is used. A theoretical model need not be mathematical (cf. the verbal and graphic models in "behavioral

marketing": Bettman 1979; Puto 1987; Sujan 1985; Wright 1975) and

For ex-

purpose. ample, most mathematical models in marketing are really measure- ment models-models set up to estimate demand functions (Hans- sens, Parsons, and Shultz 1990).

a mathematical model need not have a theoretical

'In the title of the article and elsewhere, "mathematical"is

K.Sridhar Moorthy is AssociateProfessorof Marketing, WilliamE.Si-


ester.Thearticlewaswrittenwhiletheauthorwas visiting theAnder-

sonGraduateSchoolof Management atUCLAin


Business Administration,University of Roch-

May 1990. Preliminary

presented at theAmerican Marketing Association's1989

DoctoralConsortiumat Harvard University, the 1989SummerMarket-


University ofToronto.Theauthorthanksthe

ings fortheircommentsand encouragement. Heis

to JoaoAssuncao,RajivLal,GaryLilien,Rick Staelin,andthe

mousJMreviewersfortheirdetailedcommentsona previous version




Chicago, theWharton School,andthe

participants atthesemeet-


92 / Journalof Marketing,April 1993

theoretical modeling seems

to have invadedthe Jour-

nal of Marketing Research as well (e.g.,

Hauser and

Wererfelt 1989; Lal 1990; Wilson, Weiss, and John


nonparticipant, the popularity and growth

of theoretical modeling may seem like an oddity, a

passing fad. The methodseems to violateall the norms of good research.The articlesare (generally) all the-

are unrealistic.Man-

difficult to find. To make


less lemmas, propositions,theorems,proofs.

gitimate to ask: What is all this in aid of? How does


How can we apply these models? How can we test


from the verbal

ture and the quantitative models in the decision sup- port system literature?


in the "behavioral"litera-

models?How does quantitativetheorizing

To the

ory, no data. The assumptions

agerial implications are

matters worse, the readermust wade


It is le-

methodology work? Why is it useful to marketing?


This articleis an


to one

attempt at answering these ques-


of research. (For a more



tions in an informal way. It is not meantto be a philo-

sophical discussionof research methodology, butrather

a user's

formal treatment, see Cook and

principal aim is to relatethe-

oretical modeling to the other research paradigms in

that the methodbecomes accessible to

marketing, so

1991; Suppe 1977.) The

Journal of Marketing

Vol. 57 (April 1993), 92-106

a broad array of marketing academiciansand practi-

tioners. At the same time, the perspective provided here should be useful to theoreticalmodelers as they appraise their own and others' works. The mainthemeof the articleis thatmathematical

theories are built

tation, muchlike the empiricalexperimentation in be- havioral marketing research (e.g., Bettman 1979; Puto

1987; Sujan 1985; Wright 1975). Theoreticalmodels

are the "treatments"in this

are thus

in the decision support and measurementliteraturein



a process of logical experimen-

logical experiment.They

quantitative models


differentfrom the

Bultez and Naert 1988; Little 1970,

Lodish 1971; Silk andUrban 1978). Even

though both types of models use mathematics, in pur-

pose and methodologythey arefar apart.By the same

token, even

from behavioral

ical level, in purpose and methodologythey are very


though theoretical modeling is different

1975, 1979;


researchin its mathemat-

In the next section, this view of theoreticalmod-


lengthy case study


pensationpractices, which has received a lot of atten-

tion lately (Basu et al. 1985). This example is used throughout the article. Among the topics discussedare the internaland external validity of theoreticalmod-

eling experiments, the

ing, the relative importance of

theoretical modeling, the

realism of modeling assumptions, and the testing

model-basedtheories. Finally, theoretical modeling is

relatedto the other two

in mar-

validitygiven the purpose of

purpose of theoreticalmodel-


is developed further.Then a

of theoretical modeling is

presented. The example


the "agencytheory"explanation of salesforcecom-





keting: behavioral marketing and decision support


Overview of Theoretical Modeling

Theoretical modeling begins with the

need to under-

standsome marketingphenomenon. For example, may wantto understand why storeshave sales, or why

some manufacturersare

tributionand othersare not. The researcherthen con-

a explained take place.

A model is specified by a seriesof assumptions. Some

assumptions are purelymathematical; their purpose


empirical content. They

can describe such

they care about, the

exogenous conditions under which they make deci-

sions, what their decisions

are about, and so on. (In

many of them there are, what

things as who the actors are, how

substantive, with verifiable

tractable.Other assumptions are

model-in whichtheactionsto be


verticallyintegrated into dis-


or she calls


structs an environment-which

make the analysis

marketingmodels, the actors usually are manufactur- ing firms, channel intermediaries, or consumers.)Only

the substantive assumptionsparticipate in the expla- nation being offered. For example, in Hauser's (1988) model of product and price competition, two (or three) manufacturers are deciding on the configuration of their products.

Each is assumedto offer

attributeshave to be set for each

attributesare related by the productiontechnology. Consumers' preferences are additiveand linearin the

two attributes. Every feasible product configuration

has the same constant The firms choose their

Then, after committing to a

multaneously chooses its price.

These assumptions,clearly, do not describe real- world markets.At best they define an artificialworld with some connections to the real world. Thus the

concept of a model in


tems and behavioral marketing research. In decision

support modeling, a model is a "mathematicalde-

scription of how something works" (Little 1979, ital-

ics added); in theoretical modeling, a model is simply

a setting in which a question is investigated, a "lab- oratory."2Hence, whereasdecision support models-

because they are descriptions of how things work-

emphasizerealism, theoreticalmodels-because

are laboratories-are

necessarily reticalmodel is also differentfroma behavioralmodel.

The latteris a verbalor

graphicdescription searcher's theory. For example, Puto (1987) describes



ioralresearchersreferto theirmeasurementmodel (e.g.,

a regression equation) as the model, even though a model describing their theory also exists. Once a theoreticalmodel has been built, the re-

searcher analyzes its logical implications for the nomenon being explained. Then another model,

stantivelydifferent fromthe first, is built-very likely

by anotherresearcher-and its implications are ana-

lyzed. The process

model, if necessary, until all ramificationsof the ex-

planation being proposed have been examined.

comparing the implications of one model with those of another, and tracing the differences to the model

design, we hope to understandthe cause-effect rela-

tionshipsgoverning the phenomenon in question.


as though a logical experiment were beingrun, the variousmodels as the treatmentsand the


The key differencefrom empiricalexperiments is that

only one product.Only two


andthe two

production cost for each firm. productsfirst, simultaneously.

product, each firm si-

theoretical modeling is different

the concept of a model in decision supportsys-


unrealistic.A theo-

of the re-

"proposedconceptual model of the buying deci-

framingprocess"graphically. Sometimesbehav-



continueswith a thirdand a fourth





"dependent variables."

being explained as the

2Subsequently a distinction is made between a supermodel and a

reserved for the supermodel.

model and the

For the present,

laboratoryinterpretation is

this distinction is not necessary.



in Marketing/ 93

in empiricalexperiments the subjectsproduce the ef- fects, whereasherethe researcher produces the effects by logical argument. Theoretical modeling also re- sembles meta-analysis in some ways (Assmus, Far- ley, and Lehmann 1984). Both involve "post-model"

analyses, but meta-analysis is used

to discover the

patterns in empirical results across a number of situ-

ations, whereas theoretical modeling's purpose is to

construct cause-effect explanations of marketing phe-

nomena. Figure 1 summarizesthis view of theoretical modeling.


Modeling in Action:


A Case

Firms compensate theirsalesforcesin a variety of ways,

for example, salaries,

contests, and free vacation

to ask is: Why is theresuch a variety of

schemes and what

component serve in a


et al. (1985) have used agency theory to explain these

features of salesforce compensation plans. Agency theory originates from economics, where it was developed to address situations in which a

"principal" must use an "agent" to carry out certain


actions. The principal cannot observe the

pose we consider just salariesandcommissions.

compensationpackage? Sup-

commissions, quotas, sales

trips. A natural question


function does each compensation


tions costlessly, so the question arises:Whatkind of

contract should the principal offer the agent


so that

agent is motivatedto act in the principal's inter-



of Theoretical


Marketingphenomenon to be explained

ModelofmarketingenvironmentPropositionsP aboutphenomenon

Model 2 of

marketingenvironment = PropositionsP2 about


Modeln of

marketingenvironment = Propositions P2 about phenomenon

Model n of marketing environment

Propositions P. about phenomenon

Develop theory by relatingpropositions to models

94 /

Journalof Marketing,April 1993

est?3Notable contributorsto the theory include Wil- son (1969), Spence and Zeckhauser (1971), Ross (1973), Mirrlees (1976), Harris and Raviv (1979), Holmstrom (1979, 1982), Shavell (1979), Grossman and Hart (1983), Nalebuff and Stiglitz (1983), and Holmstromand Milgrom (1987, 1990).

Stated verbally, the agency theory explanation of

fluctuating, uncertainin-


as the

by salespeople. If all

the salesperson's income came as salary, however,

salaries and commissions is as follows. Salespeople,

like most human beings, are risk averse. They prefer

a stable, known income to a

come, even if the latteris the same on

former. Salaries, by definition, lend stability and pre-

dictability to a compensationpackage, so they areused

to reducethe income risk borne


he or she would have no incentiveto workhard given thatthe firm cannot observe how hardeach salesper- son works. Commissions are used to motivate sales-

people to work hard in situationswhere their effort cannot be observed. Thus, the use of salaries and commissions in

compensationpackages represents a tradeoff between reducing the income risk borne by salespeople and providing themthe incentivesto work hard.


is to constructa "su-

permodel"specifying the overallenvironmentin which the explanation will be constructed. Subsequently, we

specify submodelsof this supermodel and derive the

this understanding. The first step

Let us see how

theoretical modeling helps us

logical implications of these submodels. This proce- dureis analogous to a behavioralresearcherfirst spec-

ifying the


context will be, how

be manipulated, what the

overall boundariesof his or her experi-


many (and which) levels of the

then actuallyrunning the

variableswill be used-and



The following assumptions describeour supermodel.

* Assumption






the firm,

is designing a compensation package for salespeople



* Assumption 2: The compensationpackage consistsof a

salary and/or

the salesperson.


The sales

on the revenues manager designs



the package

or re-

jects the compensationpackage offered.If he or she re-

she will worksomewhereelse and get ex-

jects, he or

and commitsto it. The

salesperson then accepts


* Assumption 3: Each salesperson'sutility fromincomeI

and selling effort W is given by U(I,W) = V(I) -


3Strictlyspeaking, agency

theory addresses situations in which the

the principal, that is, the prin-


agent's actions cannot be verified by

cipal cannot prove in a court of law whether or not the

out the desired actions. Observability of the agent's actions is nec-

essary for verifiability,

to use "observable" in place of "verifiable."

but not sufficient. We will, however, continue


V is an increasing, twice-continuously differentiable, concave function of I. Salespeople decide how hard to work by maximizing their expected utility.


package to maximize the firm's expected net profits, ticipating the salesforce's reaction. The net profits are

nI, where 'F denotes the gross profits of

given by Tr -


* Assumption 4: The manager designs the

the firm and n denotes the number of salespeople.

* Assumption 5: The gross profits of the firm are a func- tion of W, the work put in by each salesperson, and E, a random variable representing the uncertainty in the revenues generated. e is independently and identically

distributed across salespeople. Neither the


the salesperson observes the resolution of this uncer- tainty. Both can, however, observe the revenues ob- tained. As the salesperson works harder, he or she shifts the distribution of rr such that higher gross profit out- comes are more likely.

* Assumption 6: Assumptions 1 through 5 are known to




sales manager and the salespeople and both know

These assumptions have substantive and mathe- matical components. The distinction between the two is that the former are verifiable empirically (in prin- ciple), whereas the latter are not.4 It is the substantive assumptions that define the marketing environment:

* Each salesperson's output is independent of


salesperson's output(think about a situationin which each

salesperson is selling a unique product).


* Only salaries and commissions are available as compen- sation elements.

on a

take-it-or-leave-it basis (i.e.,

there is no room for ne-

gotiating compensation after the revenues have been re- alized).

* The manager is risk neutral; he or she is indifferent be-

dollars for sure or getting a gamble with

* The manager proposes the compensation package

tween getting x

the same expected value.

* Salespeople, however, can be risk neutral (as just

fined) or risk averse (i.e.,

preference to the gamble)-both

ted by the (weak) concavity assumption.


prefer x dollars for sure in

possibilities are admit-

* Salespeople dislike putting in effort, and their dislike is independent of the amount of money they make.

* Salespeople cannot completely control the revenues they produce.

The mathematical content of assumptions 1 through

6 resides in assumptions 3 and 4. V is assumed to be twice-continuously differentiable, which means that the salesperson's utility is a sufficiently smooth func- tion of income I. (It has no kinks or discontinuities,

its slope has no kinks or discontinuities,

of its slope has no discontinuities.) This

and the slope assumption

enables the researcher to use calculus as the primary

4The standards of evidence in the

essary to meet this standard. For

most marketing researchers empirical test, an as-

sumption such as: ceteris paribus, a salesperson's utility increases with

would willingly accept, even without a formal

of verifiability are the generally accepted standards


Formal empirical tests may not be nec-


monetary income.

analytical tool. The "maximization assumptions"-

manager maximizing expected profits, salesperson

mathematical and

maximizing expected utility-have

substantive content; they are difficult to verify em-

pirically, but we can find situations in which the "stakes" are high enough for optimizing behavior to be a reasonable assumption.

Running the Experiment

Let us now construct a series of submodels (hereafter, simply "models") from this supermodel by special- izing assumptions 1 through 6,5 and state their logical implications for the optimal salesperson compensation

contract. (The derivations of these implications are in the articles cited previously.) The models are the treatments in the experiment defined by the super- model; various aspects of the optimal compensation

scheme (e.g.,

the firm's expected profits) are the dependent vari-

ables. See Figure 1.

salary, commissions,

expected income,

Model 1 (salespeople are risk-neutral and their ef-

fort is observable).

This is the simplest model to ana-

lyze. Given the salesperson's risk neutrality, it is im- materialwhether the compensation package is all salary,

all commissions, or any combination, as long as all options yield the same expected income to the sales- person. Furthermore, given that the salesforce's work

is observable,

sation package such that if a salesperson does not work

as hard as the manager would like, that person will be penalized severely. So the salesforce will work as hard as the manager would like and each member of

the salesforce

utility Uo, the utility they would have gotten from the alternative job. The firm's expected profits will be as

high as they can be.

the manager will design the compen-

will get an expected income yielding

Model 2 (salespeople are risk averse and their ef-

Borch (1962) has shown that un-

der these circumstances an all-salary plan (with pen-

alties as in model 1) is optimal whereas an all-

commissions plan is not. The reason is that with all


so for any effort level his or her expected utility will be lower than it would be if he or she were given the

same expected income in salary. Therefore, the man- ager who wants the salesforce to put out effort W and get an expected utility Uo will have to pay them more compensation on average with commissions than with salary. With the optimal all-salary plan, however, the model 1 results are replicated.

fort is observable).

the salesperson's income will fluctuate,

5Foreach model, the defining special assumptions are in parenthe- ses.

Theoretical Modeling in Marketing/ 95

Model 3 (salespeople are risk neutral and their

effort is not observable). Now, a pure commissions



model 1

scheme, however, cannotbe the same as in model 1.

Now the commission rate will be such as to give

salesperson all of the firm's gross profits from the


by asking the salesperson to pay a lump-sum amount

equal to the firm's net profits

scheme is optimal for the firm and the

will work as hardas he or she did under

(Harris and Raviv 1979). The commission


he or she sells. The firm will make its money

undermodel 1. Essen-

tially the manager is selling the product to the sales-

person for a lump-sumprice. Both firm and salesper- son will be as well off as they were in model 1.

Model 4 (salespeople are risk averse and their ef- fort is not observable). This is the most complicated case. To analyzeit, Holmstrom (1979) makestwo ad- ditional substantive assumptions: the distributionof

gross profits

tio property and (2) is "convex" (Grossman and Hart

The distributionof sales satisfiesthe monotone


to reflect high effort on the part of the

Convexity means (loosely)

is more likely

likelihoodratio property if an observationof

(1) satisfies the monotonelikelihood ra-



salesperson thanlow effort.

that the

higher with "average"

bination of


Holmstromthen shows that any additional


profits if and only

(1985) show that the

probability of observing high gross profits is

effort than with a 50-50 com-

and low effort. (The gamma distri-


by Basu et al. 1985 has these properties.)


expected if it adds information.Basu et al.

optimal compensationpackage


effort will increase the firm's

must involve both salariesand commissions. Further-

more, the salesperson she did undermodel 2.

will not work as hardas he or

In addition, the following


resultsobtain with a gamma distribution6for rrand a

specific power functionfor

the utility functionof the

salesperson (V(I) = I8/8 with 8 =


Interpreting the Results

The four submodelscan be seen as a 2 x 2 full-fac-

torial experimentaldesign with two factors and two

levels of

plications of model 1 versus model

versus model 4, we see that the salesperson's risk

preference-whether he or she is riskneutralor not-

has a "maineffect"on the

Withrisk neutrality, salariesarenot needed; with risk

aversion, salaries are needed. Similarly, comparing model 1 with model 3 and model 2 with model 4, we

see that the observability of the salesperson's effort has a main effect on the optimalcompensationplan. If the salesperson's work is observable, commissions

are not needed; teractioneffects.

able "how hard the

interactioneffect between risk aversionand observa-

bility: lack of observability resultsin less work if the salesperson is risk averse, but with risk neutrality the observability has no effect on how hardthe salesper- son works.

What about the model4? They, too,

how certain independent variables affect certain de-

comparative-statics results from seem to indicate something about

each factor (Figure 2). Comparing the im-

2 and model 3


otherwise they are. Thereare also in-

For example, for the dependent vari-

there is an


pendent variables.All of them are ceteris paribus re- effect of

from the

alternative job andhis or herworkeffectiveness-and

all of them hold regardless of the values at which we

fix the

all of this

conductedwithin model

analysis 4, it is as though

sults-for example, when considering the uncertainty, we fix the salesperson'sutility

other independent variables. So, even though



several models with model 4's de-


Experimental Design

for Theoretical

Modeling of

Salesforce Compensationa

1. The

to the

greater the responsiveness of grossprofit variance


theless thesalesforce works,

expected income, the less the firm's ex-


the greater the proportion of salary

the less its

pected profits,

to expected income.7

Salesperson's attitudetoward risk





2. As the salesperson's work effectiveness increases, the

expected profits and the harder the

All salary, all commissions, or

any mixture;penalty for shirking; salesperson puts our

All salary; penalty for


salesforce works.

the firm's


shirking; salesperson puts out desired effort

desired effort



All commissions; salesperson

Specific mixture of

3. As the expected utility from the alternative job

6Most of these

expected profits,

in- Observability of

salesperson effort

creases, the less the salesforce works, the more its ex-

pected income, the less the firm's

the greater the proportion of salary to expected income.



salary and

comparative-statics resultsare replicated for a bi-

(Basu et al. 1985).

puts out desired effort commissions; salesperson



the nature of the

nomialdistributionas well

7The varianceof gross profits for the gamma distributionis given

by g2(t)/q,

effort, t, and q > 0 is a

the smallerthe

q, to the salesperson's effort.

where g(t) is some

the more

increasing functionof the salesperson's

parameter of the gamma distribution. Thus,

responsive is the varianceof grossprofits

aEntriesin the cells

optimal compen-

sation scheme and the

ious treatments.

salesperson's effort level under var-

96 / Journalof Marketing,April 1993

fining characteristics (salesperson's work not ob-

served, salesperson risk averse) are being analyzed for their main effects with respect to certain independent variables. For example, comparative-statics result 2 says that the salesperson's work effectiveness has a main effect on the firm's profits and how hard the




each with a differentlevel of

tiveness. Hence, comparative-staticsanalysis is es- sentially an efficient way to run an experiment when

the "causes" being manipulated are continuousvari-

ables.8 Model 4

(sub)models are the ones defined

nationsof levels of the independent variableson which

the comparative statics is run.

ceived quality in a product class in which there is a strong marketcorrelationbetween price and objective quality (women's blazers). They used three levels of

familiarity as a covariate, four price levels, and two objective quality levels. In other words, they per-



as they did, they were unable to determinewhether


class in

market price and objectivequality. The pointis, how-

ever, that Rao and Monroe had to limit their experi-


relationshipsthey found would apply in a product

fairlycomplex experiment with at least three Nevertheless, by constructing the experiment

and we could have discoveredthis

approximately and much more labori-

a "large" numberof model 4's,

salesperson work effec-


supermodel now and the

by various combi-


which there is a weak correlationbetween

somewhere. Their experiment is already much


perceivedquality(Olson 1977).

assumption is empiricallytrue,


more complex than previous studies of the relation-

ship between price

In the salesforce compensation context, our su- permodel assumes thata salesperson'sproductivity is

independent of other salespeople'sproductivity.Though

in some situationsthis


or the other of the

salespeople work as a team in selling to an account

or (2) even though different

pendently, their productivity is


and Models

many othersit is false. It is often false becauseone


conditions holds: (1)

As is apparent from the example just considered, a

interpreting the impli- supermodel it would be

difficult to compare one model with another.Never-

theless, the choice of a

and clarity.

For example, in physics, the general field theory that



general theory


comprising the following models: electromag-

is still being developed can be thought

issue is the tradeoff between generality

supermodel is aframework for

cations of models. Withouta

salespeople work inde- affected by the same

state of the

is the

key to explaining why salespeople in a team are all

compensated alike and why sales contestsareused (as

economy). Such dependenceamongsalespeople

underlying environmentalfactors (e.g.,

supermodel is not easy. One

of as a


we learnedfrom Holmstrom 1982 and Nalebuff and

Stiglitz 1983). By restricting our supermodel as we

did, we


us isolate and understandthe effects of

of salesperson's actionsandriskaversionon the choice

between salary and commissions.

complication in specifying

is thatthe supermodel is constantlychanging. It evolves

as our

understanding evolves. Each successive study

theory, quantummechanics,

of relativity. Each of these models, however, can be

thought of as a supermodel in its own right. For ex-

ample, the generaltheory of relativity is a supermodel

for the special theory of relativity. (Similarly, model

4 is a supermodel for its comparative-staticsresults.)

These successive attempts at generalization have as their goal the explanation of more phenomena within

a common

external validity (discussed subsequently) and is

ier to test. However, for

effects captured in the less generaltheory, that theory

is better.

framework; the generaltheory has greater

were unableto explain these salesforcecom-

the restrictions helped


a supermodel

The other

eas- is based on a "big picture" (as it exists then), but it

also contributesto the big picture. In the

text, the

ity made possible the

in turn, the general field

roe (1988) study, the inclusionof familiarity as a co-

variate is

standing of



development of the special theory of relativ-

generaltheory of relativity,and,

theory. In the Rao andMon-

testimony to the evolution of our under-


behavior;familiarity does not



understanding the specific

Similar tradeoffs are faced by behavioral experi- menters. They mustdecidehow many effectsto "throw

in" to a given experiment. For example, Rao and

Monroe (1988) examined the relationships among

productfamiliarity,objective quality, price, and

per- appear as a constructin the studiesreviewed

(1977). In the salesforce compensationcontext, recent


that some


because output is

missionary on the ob-

servable output dimensions, salespeople may misal- locate their effort with respect to the unobservable

work). If compensation

dimensions cannot be measured (e.g.,

X"Essentially" because comparative statics usually requires strong

and Shannon

with y as the risk

by Holmstromand Milgrom(1990) suggests

salespeople may be


compensatedby salary

theireffortcannotbe observed-

multidimensionaland some of the

were based






convexity assumptions (Milgrom

we assume that

V(I) =


example, if

aversion parameter, a comparative-statics

respect to y does not give us the expected result

The optimal compensation plan turns out to be (I/y)log

B Tr)], where A and B are

sion is not well defined for

becomes imperative if we want to see what happens with


analysis of model 4 with

for risk neutrality.



constants independent of 7r, and this expres-

y equal to zero. Constructing model 3

risk neu-

Theoretical Modeling in Marketing/ 97

output dimensions. In effect, this researchidentifies another"cause"for salaries,but because this learning is new, we do not see an awarenessof this issue in

previous supermodels.9Supermodels

however, must explicitly

multidimensional output, depending on which cause

is the focus of study. In summary, a supermodel defines a manageable

experiment, with the word "manageable" left delib-


for the phenomenon in question and the submodel

analyses then help identify the implications of those

causes. Which

to leave out depends on our knowledge of potential

causes-knowledge ing improves-and

clarity. How this tradeoffis resolved depends on one's

purpose. If understanding of

a less generalsupermodelfocusing on those effects is

indicated; if

the relevant goals, a more general supermodel is ap-


from now on,

assume unidimensionalor

vague. It builds in a set of

potential causes

potential "causes"

to include and which

that changes as our understand- the tradeoffbetween generality and

specific effects is the goal,

external validity and empiricaltesting are

Internal Validity

of Theoretical


The internal validity of the salesforce


models were chosen carefully to form a factorialde-

sign of the two forces

described is very high. Because the


underlying the explanation and

logically (as shown by

the conclusions were derived

the fact that the results have stood the test of time),

there is essentially no question aboutthe cause-effect


is used because there is a

4 with the use of environmental assumptions that are more restrictivethanthose in models 1, 2, and 3. For

some results we assumed that the distributionof Tr

satisfied the

was convex, and for the

assumedin additionthatthis distributionwas

and the

function. One


emphasized, are they due to these additional, more restrictiveas-

due to the model differences that we

established. The qualifier "essentially"


in model

monotone likelihood ratio property and

comparative-statics resultswe


salesperson's utility function was a


might legitimately ask: Are our results

sumptions that we "sneakedin" to the

cause this question is really about the generality of

the results, however, it is

better to address it as an


analysis? Be-

validity issue (discussed

subsequently) rather

thanas an internal validity issue (Cook and Campbell

1979). To see this, observe thatwe

can get rid of any

following trick:


questions aboutinternal validityby the

any special assumptions

made under model 4

9In fact, in the first version of this article, there was no discussion of this issue.

98 / Journalof Marketing,April 1993

have been made up front, under the supermodel, as assumption 5.1! Thatwould have given us a less gen- eral supermodel, but internal validity would be se- cure.

Realism of TheoreticalModels

The need for internal validity in theoretical modeling necessarilyimplies thattheoreticalmodels will be un- realisticto some extent. The reasonis twofold. First, it is easierto infercause-effect rel</