Vous êtes sur la page 1sur 13

Competitive Reactions to Market Entry: Explaining Interfirm Differences Author(s): Hubert Gatignon, Erin Anderson and Kristiaan Helsen

Source: Journal of Marketing Research, Vol. 26, No. 1 (Feb., 1989), pp. 44-55 Published by: American Marketing Association Stable URL: http://www.jstor.org/stable/3172668 . Accessed: 09/10/2013 16:17
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

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.

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


This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions


Competitivereactionsare recognizedas a drivingforce influencing marketing decisions.Theauthorsseek to explainhow establishedcompetitors in an oligopoly react to a significant new entryin theirmarket.It has been suggestedthat at least some establishedcompetitors will react to a marketentry positivelyand at least some competitors will react negativelyor not at all. Boththeoryand evidencesuggest that not all firmswill react to an entry in the same way. The authorsposit that interfirm differencesin competitive reactionsto entrycan be predictedby obmix variable. Comserving,for each competitor,the elasticityof each marketing mix"weapons" and retreatwith petitorswill retaliatewiththeireffectivemarketing their ineffectivemarketinginstruments. These predictions are tested by estimating the parameters of an econometric modelof demandresponsefunctions and reaction withdata from the marketfor an over-the-counter functions gynecologicalproduct and from the airline industry.Results,replicatedin the two markets,are substanwith predictions. tially consistent

Competitive Reactions
Explaining Interfirm





In a Journal of Marketing Research editorial, Weitz (1985, p. 229) notes: The effectivenessof marketing usuallydeprograms pends on the reactionof both customersand comtheoriesand research petitors.However,marketing have emphasized issues related to customer response and have directedless attentionto competito competitive tive response.This lack of attention becauseit is difficultto imagine effects is surprising a marketing decision that is not affected by competitiveactivity. Weitz goes on to call for empirical research to identify patterns of competitive response under a variety of conditions.
*Hubert Gatignon and Erin Anderson are Associate Professors and Kristiaan Helsen is a doctoral candidate, The Wharton School, University of Pennsylvania. The research was supported by the Center of Marketing Strategy Research and by The Wharton School Summer Salary Support Program. The authors thank three anonymous JMR reviewers for their comments, as well as Jerry Wind for assistance in obtaining part of the data for the study. They also thank an anonymous manager at a major pharmaceutical house for considerable market research and industry background information.

Our research is a step in that direction. We examine competitive response to an event that can profoundly affect a firm or even an entire industry: the entry of a new competitor (Baumol 1982). How will established firms react to this potentially momentous event? Scherer (1980, p. 244) outlines the possibilities as either accommodation (cutting back to "make room for the newcomer") or retaliation (fighting back to "make life as difficult as possible for the interloper") and calls for economists to build "realistic theories" about which reaction to expect. The strategy literature amplifies, pointing out that different competitors not only choose different reactions but differ in how they employ specific "instrumentsof warfare" (Kotler and Singh 1981). For example, firm A may react by increasing advertising, firm B may cut price, and firm C may alter none of its marketing mix decisions (Hanssens 1980; Lambin, Naert, and Bultez 1975). In sum, each competitor decides, for each marketing instrument, whether to respond to an entrant by counterattacking (raising expenditures, a positive reaction), retreating (reducing expenditures, a negative reaction), or not responding (a zero reaction). Predicting these reactions is an importantcomponent of strategic marketing (Oxenfeldt and Moore 1978; Porter 1979; Rothschild 1979). Predicting competitive reactions in detail is a
44 Journal of Marketing Research Vol. XXVI (February 1989), 44-55

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions



complex, subjectiveendeavorand often is based on dubious personificationsof each competitor(Day 1984; Miles 1980). Thoughbroad guidelineshave been suggested (in particular,in the industrialorganizationliterature),they often yield conflicting predictions.Furmodelswhose come fromnormative ther,theseguidelines to actualsituationshas not been assessed. applicability The purposeof our study is to develop and test an of when andwhy some finnrms (or brandmanexplanation in in the case of firms) a given maragers multiproduct ket reactnegativelyto a marketentry, othersreactpositively, and others do not react at all. We survey the literatures and array strategicand industrial organization their argumentsinto reasons for expecting negative or on the positive reactions.We concentratein particular across of in reaction variations competitors explanation We then suggest a continand marketinginstruments. based on the principlethat competitors gency approach fight only with their "best weapons"and avoid doing andSingh1981). battlewiththeir"poor (Kotler weapons" mix variWe proposethatgood weaponsare marketing mix variablesthat induce a small change are marketing Elasin the firm's sales or marketshare performance. in the marwhich measures the change ticity, percentage ket's responseto a branddue to a percentage changein variablesof thatbrand,is thought one of the marketing into be an indicatorof the strengthof that marketing strument. Our fundamental propositionis that each established thatthreatens its position an entrant to reacts competitor with variables on relatively high by raisingexpenditures on variables with andby loweringexpenditures elasticities relativelylow elasticities. Here, we considermarketresponsein termsof shiftsin marketshare.We offer a test of these ideas using an econometricmodel of competitive rivalryin the marketfor an over-the-counter gynecological productand we replicatethe resultsin the airline industry.
COMPETITIVEREACTIONS: CONFLICTING THEORIESAND CONTINGENCYPROPOSITIONS Overview ables that the firm uses well-those for which the market response is relatively elastic. Conversely, poor weapons

The theoreticalliterature bearingon competitivereaction comes largely from industrialorganization(IO)

and strategy, whereas much of the empirical research comes from marketing. An oft-noted feature of the IO literature is its emphasis on structure (number and size of competitors), from which behavior ("conduct") is inferred (Scherer 1980; Weitz 1985). This literature is covered thoroughly by Scherer (1980). In contrast, the strategy literature is concerned directly with the actions of each competitor (Miles 1980). These literatures are a rich source of rationale for hypotheses. Perhaps because of the difficulty of obtaining comprehensive data on competitor behavior, empirical re-

is relatively searchor competitive research sparse(Scherer 1980; Weitz 1985). Empiricalmarketingresearch on on the competitiverivalryor reactionshas concentrated difficult issue of how to measurethe extent of rivalry. Apartfrom marketsharemodels, the first econometric modelsexplicitlyincorporating did so by incompetition efforts as independent troducingcompetitivemarketing variables (Bass andParsons1969;Beckwith1972;Clarke 1973; Schultz1971). ThenWildt(1974) explicitlymodeled rivalryin a systemof equations wherethe marketing decisionswere endogenous.Further in the developments of reactionsand the effects of competitive measurement reactionon optimaldecisionshavebeen offeredby Lambin, Naert, and Bultez (1975), Metwally (1978), and Hanssens(1980). This work has advancedthe study of competitionto the point that a state of rivalrycan be observed and measured(Hanssens 1980), therebyovercominga very difficult barrierto empiricalstudy of the impactof competitivereactions. We focus on how establishedfirms in an oligopoly of a new competitor in theirmarket. reactto the entrance The purposeof our studyis to investigatewhy thereare differencesin the directionof response(positiveor negative) among firms in a given market. This research by the empirical findingthatfirms questionis motivated do not all reactin the samedirection.Forexample,some firms mightreactto a competitor's increasein advertistheirown or, conversely, ing expenditures by increasing firmsmay not reactat all, them. Further, by decreasing reactionelasticities(Gatishowingzero or insignificant in reactionto the same Such differences gnon 1984). event-a marketentry-are due partiallyto variations in the perceptionsof the relevantorganizational members and in the enactment processesby which individual perceptionsbecome corporateviewpoints (Day 1984; Miles 1980;Oxenfeldt andMoore1978;Rothschild 1979). Differencesin reactionalso may dependon the agendas and values of influentialindividualswithin the established firmsin an industry (Cyertand March1963;Williamson 1965). Unfortunately, these factorsare difficult to study systematically; theireffect may be to introduce "noise" into the relationshipbetween competitive actions and reactions. However, anotherfactor-differences in firm abilities-operates in a more systematic way and is moreamenableto theorydevelopment.Firm abilitiesare the focus of our study. Writersin strategypointout that some firms are simply very good at some functions (e.g., advertising) and very poor at others (e.g., distribution). Several scholars, notably Porter (1979) and Day (1984), build their approach on the identification of each competitor's "vul-

nerabilities" and strengths Such (competitive advantage).

knowledge is useful not only in deciding where a firm should attack, but also in predicting where it will be attacked and what weapons each competitor is likely to use. The prevailing (and reasonable) assumption is that firms will use only the weapons they wield well. In the following sections we first discuss reasons to expect neg-

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions



ative or little reaction,then reasons to expect positive reaction.

Negative Reaction (Withdrawal)

We suggestthatestablishedcompetitors may respond to entryby cuttingback resourcesdevoted to a market for two reasons:inabilityto formulatean effective reis the optimal sponseand the possibilitythatwithdrawal response. (profit-maximizing) to battlepresumesone knows how to fight back. Devisis not difficult for some kinds of ing a counterattack recompetitive action, but for others the appropriate (Schmalensee 1976). As Scherer (1980, sponseis unclear but counteracting a clever advertising gambitis far from an entrant(likely to be somethingof easy." Countering an unknownquantity)is also far from easy. Hence, in the face of uncertainty abouthow to respond,some firms may respondvery little, if at all. Firmsmay even reduce resource commitments, either permanentlyor while waiting to determinehow the marketwill react to the entrant(Cooperand Schendel 1976; Kotler and Singh 1981). A firm also may cut back becauseit is certainthat it cannot counter effectively (Day 1984; Oxenfeldt and Moore 1978). As Rothschild (1979, p. 23) states, "Ifthe will have less key skills aren'tavailable,the competition thanoptimalresults." to the questionof reactionto entry is offered approach by the DEFENDERmodel (Hauserand Shugan 1983). This normativeanalytical model predicts the optimal (profit-maximizing) responseto entrygiven a set of assumptionsabout the market.Under a range of conditions, some degree of negativereactionproves optimal (profit maximizing)for the three variablesconsidered: (cut expenadvertising(cut expenditures),distribution ditures),and pricing(raise price). One reasonfor these resultsis that, underHauserand Shugan'sassumptions, the entrantusually decreasesevery other competitor's profit, makingthe marketless worthwhile.Hence, expenditures generatelower returnsafterentryand established competitors will have an interestin not commitAn inherent ting the samelevel of resources. assumption, does not expandas the result however,is thatthe market of the productintroduction.
Positive Reactions (Retaliation) Firms may respond to entry by positive reactions (increasing marketing expenditures to combat an entrant) for two reasons: objectives other than profit and underspending. Objectives other than profit. The DEFENDER model's recommendation to cut back in response to market entry follows from two premises: firms behave rationally and they seek to maximize profit. An extensive literature in organizationtheory establishes the possibility that firms Negative reaction as a profit-maximizing response. One p. 388) puts it, ". . . any fool can match a price cut, Inability to respond effectively. Committing resources

behaveirrationally (e.g. Cohen,March,andOlson 1972). It is also possible that, for a given brandor strategic businessunit (SBU), the objectiveis not profit maximizationbut is insteadmarket of dominance,maintenance a niche, or some othergoal not relateddirectlyto brand or SBU profit.Suchobjectivescan leadto a "fightback" even at the expenseof profit. reaction) response(positive modelsusuallyassumethat Underspending. Analytical at a level of marketing such all firmsoperate expenditures thatthe elasticityof each marketing variableis less than one (decreasingreturnsto scale). This is also an asof the DEFENDER model(Hauser andShugan sumption is violated(increas1983). However,if this assumption ing returnsto scale), negative reactionswill not maximize profit because a firm is already underspending. Hence, furthercutbacksperpetuatethe underspending error.Firms that are underspending may have positive rather thannegativereactionsbecausepositivereactions may be the profit-maximizing responsefor them. A firm might underspend for several reasons, even though doing so is normativelyincorrect.First, man(at least initially)of the elasticity agers may be ignorant of a marketing mix variable(Chakravarti, Mitchell,and Staelin 1981); they may not be aware they are underspending.Second, to find where they are on the sales responsecurve, firmswould need to experiment,which they typicallyhesitateto do (Ackoff andEmshoff 1975; Little 1966;Pekelman andTse 1980). Third,firmsmight not be able to increasetheireffort substantially to reach the zone of optimalreturns (e.g., becauseof lackof budor lack of access to channelsof disget for advertising tribution).Fourth,as Hauserand Shugan(1983) note, some industries are "sleepy"(not very competitive). The entrant may "wakeup" the industry,galvanizing as they competitorsto use marketingmix instruments shouldbe used. In essence, entryheightenscompetition, which forces a firm to reexamineits practicesand correct errors, such as spendingtoo little on a marketing mix variable.This "errorcorrection" is observedas a positive reactionto marketentry.
A Contingency Approach to Competitive Reaction

If positive, zero, and negativereactionscan vary by both competitorand marketinginstrument,how can a given competitor'sreaction be predicted?Though we cannothope to providea completeanswer, we suggest that the directionof competitivereactionis contingent
upon the elasticity of a firm's marketing mix variables. We propose that firms react positively (increase expenditures) with their best weapons-the marketing instruments that have relatively high elasticity. Conversely, firms react negatively (decrease expenditures) with relatively inelastic marketing instruments. If firms react negatively when elasticity is low and positively when elasticity is high, it follows that there is a "turning point"-the level of elasticity where no reaction would occur. In practice, given the uncertainty of the assessment of elasticity, there is a zone of elasticity within

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions



which no reactionwould be expected. The precedingdiscussionleads to the followingpropositions. effortelasticity for a given PI: A firmwithlow marketing its effortfor reactsto an entryby decreasing variable thatvariable. the variable of a marketing increases, P2:As the elasticity firmreacts its effortby lesser to anentry by decreasing At some point,the firm'sreaction turns increments. to entryby increasing positive,thatis, the firmreacts its effortforthatvariable. effortelasticity fora given P3: A firmwithhighmarketing its effortfor variable reactsto an entryby increasing thatvariable. P, proposesa zone of negativereaction,P2a zone of no reaction,and P3 a zone of positive reaction.

are tested in two industries.Though Ourpropositions this test may not generalizeacrossall industries,its replication offers evidence about the existence of the hyto a parthatis not idiosyncratic phenomenon pothesized thedataforeachmarket/ We firstdescribe ticular industry. followed studied.Then, becausethe procedures industry in the two cases are identical, we discuss each step of the processandthe resultsof each step for bothdatasets in parallel.
Data Over-the-counter gynecological product. Our first da-

taset is fromthe marketfor a women's healthcareprodover the counter.The product, uct thatcan be purchased which can be used once only, was formerlyavailable only througha visit to a doctor'soffice. Once approved as an over-theby the Food and Drug Administration counterproduct,the first brandwas launchedin the late 1970s and was an immediatesuccess. By 1982 the inreadustrywas large and competitive.For proprietary sons the natureof the productcannot be describedin greaterdetail. We refer to the productas "OTC-gyn" (over-the-counter gynecologicalproduct). 1982to June dataareavailablefromJanuary Quarterly 1986, a periodof intense competitioncenteringaround a few nationalbrands,includingtwo majorentries. In fourestablished brands ouranalysis,we consider forwhich data are availablethroughthe period and two entrants.
The four established brands (brands A, B, C, and D) were the "major players" in this market during this period, with a combined market share of more than 90% before the entries of two new brands, which we study. Three of the four brands use advertising as their principal marketing mix instrument. Because the market changes quickly, test marketing and rollouts are rare. Brands typically go from concept test to national launch, which allows pinpointing of entry periods. We use Nielsen data for the dependent variable, market share in units. The predictor variables are repre-

sentedby proprietary dataon all majorcomadvertising petitors.Prices were relativelystable duringthe period investigatedand, with the exceptionof brandC, there were few price differencesamong the brandsavailable duringthatperiod.' BrandC is a low pricedbrandthat did virtuallyno advertising.Given that media expendituresareon a quarterly the market basis, we transformed sharedatafromtheiroriginalbimonthlybasis to a quarlinear smoothingtechterly basis by a straightforward nique.2 The OTC-gynindustry consistsof threetypes of products, which have differentbenefits and drawbacksand is effective coexist in the market.The "firstgeneration" but is relatively Onebrand easy to misuseunintentionally. in quarter3 of 1985, is of entry (brandE), introduced this type. "Second-generation" in products,introduced the early 1980s, are less sensitive to mishandling.One establishedsecond-generation brand, A, was extended with the additionof a variantwe label "brand A-plus," in the first quarter of 1984. "Third-generaintroduced tion" productsare exceptionallyeasy to use and have a researchindicates Qualitative very differentappearance. the third-generation productsappealto manyconsumers who believe they carrya moremedically"correct" connotation.BrandF, whose entryis includedin our data, is the firstbrand of this formandwas launched in quarter 4 of 1984. Airlinemarket.In this industry, are large, competitors the ratio of fixed to variablecosts is high (which, accordingto Porter1979, would make positive reactions line is flexible (the product morelikely), andthe product is a flight andthe flight scheduleis not difficultto vary). The relativeflexibility of the numberof flights should increasethe likelihoodof observingany change in this decision variableover time. Further,the environment hostile and uncertainafter deregubecame particularly lation (Robertson,Ward, and Caldwell 1982). Indeed, the marketingliteraturehas shown the occurrenceof competitivereactionsin this industry,some firmsreacting positively and others negatively (Gatignon 1984, Hanssens 1980; Wildt 1974). Empirically,the industry is relativelytractable becausethe impactof a new entry in the marketis fast, as are competitors'reactions,so

'Thoughprice data were not availableto verify this information, which was providedby the managerof one of the brands,it is conof the product,a high-involvement, occasional gruentwith the nature purchasethat is not expensive in absolute terms. In addition, the indicates goodnessof fit of the empiricalmodels reportedhereafter thatthe variables includedin the modelexplainmostof the variances in marketshare, suggestingthatrelativelylittle would be left to exvariablessuch as price. plain by additional 2Though higherdegreesof freedomwould be availableby disagof the independata, errorin measurement gregatingthe advertising dent variablesintroduces a bias in the estimatedmodel parameters, whichis not the case for measurement errorin the dependent variable it is Therefore, (only inefficiencyresultsbecauseof heteroelasticity). to sacrificedegreesof freedomfor unbiasedness of coefpreferable ficient estimates(Judgeet al. 1985).

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions



effects should not be confusedby lags and inertia. As mix variables,the there are few significantmarketing "noise"createdby the effects of multiplechangesin the is avoided (Gatignon1984; Schultz 1971; marketplace Schultzand Hanssens1977). The market (Los Angeles-Phoenix travel) was selected becauseit met the criterion thatan entryoccurred when the marketbecamemuchmore afterderegulation, to the 24 months The time seriescorresponds competitive. to December 1979 1980) following deregula(January tion. This series is long enough to estimateelasticities with reasonableprecisionwithoutjeopardizingthe staconditions. bility of otherenvironmental Thisperiod to the timeof observation when corresponds demonstrated the of Gatignon(1984) importance comin the last of airline industry deregulation petition phase (Graham,Kaplan,and Sibley 1981). The time series is observedbefore deregulation destabilizedthis industry. Therearethreecompetitors, all with similar market shares and in a and none dominant (.342, .393, .265), position. 3 is a nationalcarrier,for whichthis route's Competitor revenues constitutea small portionof total sales, and competitors1 and 2 are largeregionalairlines.The new entrant 4), whichis also a largeregionalcar(competitor on thisroutein April 1980. During started rier, operating thistimeperiod,the volumeof passengers flyingthe route varied greatlybut without any majortrend. The entry did not significantlyaffect primary demand. The dependent variableused in the studyis the share of the number of passengers on directflights on the Los route. Past research(Gatignon1984; Angeles-Phoenix Hanssens1980;Schultz1971;Schultz andHanssens 1977) has shownthatthe numberof flights is the most significant predictorof the numberof passengerscarriedby an airline.Advertising in the city pairalso expenditures to a lesser to predicting either contribute, extent, though the numberof passengersor marketshareof an airline on a given route. In addition,after deregulation, price became a significantexplanatory variableas well. Gatignon (1984) definedprice as the "averagelowest fare (one-way) weightedby the durationof that lower fare wereonly duringthe month,"wherethe faresconsidered those withoutrestrictions on types of individualsor on we used as incapacity.Followingthis priorliterature, variables the share of the dependent city-pairadvertising (obtainedfrom the Media Records Green expenditures BookandfromBroadcast Advertisers Reports,Inc.), the
relative number of flights, and the relative price. These data were obtained from records of the Federal Aviation Bureau (FAB). Though no significant pattern or trend could be observed in the number of flights or advertising series, a clear trend of increasing prices occurred in this market during the period. However, the relative price measure shows variability over the period. Model Specification The same procedure was used for both the OTC-gyn product data and the airline route data, which were mod-

eled separately.Our data analysis proceededin three stages. In stage 1, we estimated,for eachbrand,a model of marketshareas a functionof the marketing mix variables (e.g., advertisingshare, relativeprice or number of flights, lagged marketshare, and dummyvariables the two entrantsand one productline exrepresenting for the airtensionfor OTC-gynbrandsand one entrant line route).This stageyieldedestimates of marketing mix elasticities,as well as the impactof each entry. In stage mix decisions(advertising 2, we modeledthe marketing for the OTC-gynmarketand numberof flights for the airlineindustry) of each brandas a functionof competitive activity. In stage 3, we modified the marketing variabledecision functions,introducing a constraint on the coefficientsthatreflectsthe impactof an entrant on decisionsaboutadvertising or number of flights. spending This constrainttakes the form of a process function, whereinthe degreeof reactionis expressedas a function of the brand'sadvertising elasticity(OTC-gyn)or of the number-of-flights elasticity (airlines). The coefficients of the processfunctionform the test of our hypotheses. were specified, as is commonly done in econometric models in marketing(Beckwith 1972; Lambin 1976; ParsonsandSchultz 1976), as a functionof lagged market shareand shareof marketing mix expenditures. For the airlinemodel, quarterly variables inwere dummy troducedto capturethe seasonalityof the market.The lagged dependentvariable representsthe dynamics of marketshare. It is a commonlyused formulation that tendsto contribute to the robustness of the marketshare model (Naertand Leeflang 1978). The functionalform is linearin the logarithms,so the coefficientsare interpretableas elasticities. Competitiveeffects were capturedby usingeach brand'sshareof industry advertising (OTC-gyn)or numberof flights and relativeprice (airline). In addition, we introduceddummy variablesto the impactof each entry.3 represent In the OTC-gynmarket,therewere two new entries, brandF and brandE, as well as the productline extension of brandA (A-plus). Each brandequationis therefore of the form:

Market share equations. The market share equations

(1) m,(t)

ee',0m(t 1)O'a'ai(t) H k=1 etk+a)eUs()


3Thenew entry may influenceconsumers'responsesto the marketing mix variables,therebychangingelasticitiesand cross-elasticities, but there is little theoryas to why they would change and in which direction.With enoughobservations,Chow's (1960) test can determinewhetherthe slopes are stable pre- and post-entry.In our in both datasets, study, becauseof the small sampleof observations we tested the stabilityof the parameters when the observations after an entry occurredwere added (Maddala1977) for each entry. The tests failedto rejectthe stability of the parameters pre-andpost-entry, except in the case of brandC. However,given thatthe branddid not decisionscould not be modeledin advertise,this brand'sadvertising the rest of the analysis.Consequently, the effect of new competition is modeledparsimoniously as a "main"effect in the relevantbrand modelsof marketshare.

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions



where: mi(t)= marketshareof brandi at time t, ai(t) = advertisingshareof brandi at time t, Dk(t) = dummyvariablefor entry of brandA-plus (k
= 1), brand F (k = 2), and brand E (k = 3),

Table 1
RESULTSa Independent variables Intercept Laggedshare share Advertising BrandA plus
Brand F entry Brand E entry

and 3's = responsefunctionparameters, = disturbance term. u,(t) variables thatthe important Overall,the fits confirmed were included(R2 = .94, .97, .74, and .92 respectively for brandsA, B, C, and D, thoughthey shouldbe evaluated with caution given the relativelyfew degrees of freedom).The OLS estimationresultsindicatedthe importanceof the lagged dependentvariablefor most of the brandmodels. However, variationsacrossbrandsin were observed. the significanceof the otherparameters The lack of significancecould be due to the inefficiency of the OLS estimationif the errortermsare contempor(Beckwith 1972; Reibsteinand Gaaneouslycorrelated sharemodelswere the market tignon1984).Consequently, reestimated simultaneously as seemingly unrelated beto take into accountpossiblecorrelations regressions terms. Given that the OLS estitween the disturbance mates are not efficient, a cutoff t-valueof 1.2 was used for includingeach variablein the SUR model specification accordingto a procedureused by Schultz and the lagged Hanssens (1984). Further, (1977) andGatignon to marketsharecoefficientfor brandB was constrained estimateof 1.033 (though 1 because the unconstrained not statistically differentfrom 1) could lead to inconsisdid not signiftent marketshareresults. This constraint coefficients. of the other affect any icantly in Table 1. Advertising The results are summarized shareis relatedsignificantlyto the marketshareof the two largest brands,A and D. These coefficients (.068 and .136 for A and D, respectively)are in the typical range of advertisingelasticities (Lambin 1976). It is noteworthythat the advertisingof brandD is approxiof brandA, matelytwice as effective as the advertising differencesin firm abilities.In adindicatingsubstantial dition, the marketleader, brandA, is the only brand negativelyaffectedby the two entriesof new competitors (-.102 and -.078 for brands F and E, respectively). However, the brandA-plus productline extension hada positiveeffect on brandA's share(.086). This additionalshare was taken from brandC (-.133) and brandB (-.134). The positive effect of the brandE entry on brand C's market share (.139) may be due to a difference in their positioning. Though both are secondgeneration formulations, brand E was launched as a relatively expensive, heavily advertised brand. Brand C's positioning as a similar formulation but as a low priced national brand may have become clearer to the consumer as a result of the "splashy" entry of a physically similar brand. In other words, brand E may have increased awareness and acceptance of second-generation national brands; hence consumers may have been more willing to notice brand C's price advantage and to "experiment"

BrandA 2.108 (7.36) .369


BrandB -.068 (1.27)

1.0b -

BrandC 1.570 (3.66) .385


BrandD .398 (1.20) .725


.068 (4.44) .086 (5.02)


-.004 (.44) -.134 (1.84)


NA -.133 (3.17)

.136 (3.59)

-.078 .139

(3.17) (3.46) 16 16 16 Numberof observa- 16 tions .74 .92 .94 .97 R2 (basedon OLS) witht-valuesless than1.2 arein parentheses. Variables "t-statistics in OLS estimation areomittedand indicated by a dash;NA indicates for the brand. no advertising to 1.0. constrained bCoefficient

nationalbrand(albeitan with another second-generation unadvertised name). Finally, the lagged marketsharecoefficientsof brand B's share(not significantlydifferentfrom 1.0 and conto thatvalue), combinedwith the negativeconstrained stantterm(-.068), reflectthe smoothlydecliningshare trendof the brand.BrandA andbrandC have relatively low values for lagged share (.369 and .385, respecthatthesetwo brands aremorevolatile tively), indicating and need advertisingto supporttheir share. Brand D benefitsfrom a relativelystrongstability(with a lagged sharecoefficientof .725), in additionto the strongimpact of advertising. By the same procedure,the marketshare model for each airlineis represented by equation2.


esD,4 ' ee,.?m,(t 1)i''f;(t)s'2a,a(t)l2,pi(t)l"

+ (t) Sef.a,2t) + i,70(t POWs4(t)e,

where: mi(t)= marketshareof airlinei at time t, shareof airlinei at time ai(t) = advertising f(t) = share of numberof flights of airline i at time t, p,(t) = priceof airlinei relativeto average price on the route at time t, DAt)= dummy variablefor entry of airline 4, dummyvariables, Q2(t),Q3(t),Q4(t)= quarterly = function and response parameters, 1's term. u,(t) = disturbance

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions



As is similarto the fit obtainedin the OTC-gynmarvariablesexplainrespectively98.5, ket, the independent in marketsharefor the 97.7, and98.2% of the variances three airlines. The seemingly unrelated regressionestiin Table2 (variables for which mationresultsarereported the OLS coefficient estimateshad a t-value inferiorto 1.2 are eliminated). In this market, as typically of flights found,the number is the strongestdeterminant of marketshare, with share elasticitiesof .905, .907, and .880 for each airline, respectively. Advertisingshareis not significant,but the relativeprice of airline3 has a significantimpacton its marketshare(with an elasticityof -. 185). Thoughthe lagged sharevariablewas eliminatedafterthe OLS resultsshowedinsignificance, the market sharesof airlines seasonalcomponent.Both air1 and2 have a significant lines 1 and 3 are affected significantlyby the entrant. The nationalairline,3, was hardest hit (-.187), whereas the regionalairlineswere affectedless (-.066 for airline 1). In fact, airline2's marketshare increasedafter the else being constant.Its sales (i.e., the entry, everything numberof passengers),however, were affected negatively by the entry. To this point we have estimatedthe effectivenessof each competitor's mix variables. We also have marketing establishedthat, in the OTC-gynmarket,the two new competitorsand the productline extension had a discernibleimpacton the leader'sshareof the market.The impact was similarin the airlinemarket.We now test our propositions via the estimationof decision models, which representthe advertisingand number-of-flights levels (for the OTC-gyn and airline markets, respec-

before and after entively) chosen by each competitor tries. to be modeledis the advertising expenses in the OTCIn market. airline the market, gyn only the numberof flights is consideredbecause this marketingdecision variableis significantly moreimportant thanpriceor advertising, as discussedbefore. The specificationof decision models is madedifficultby the complexityof the to incorporate. The decision equationmust phenomena interfirm coordination of mix variablesand incorporate reactionsto competitors, both with lags in reactionsand anticipationof competitors'actions. Hanssens (1980) discusses the modeling of these issues and proposes a methodbased on time series analysisto assess empirimodelspecification.However,withthis cally the correct methodone mustassumethe availability of a substantial numberof observations to considerall cross (leads and lags) correlations. The methodwe used is similarin spirit, though the limitedsampleperioddid not enableus to use time series analysis. Instead,we consideredall leads and lags in a a set of variables stepwisemanner,each step introducing enteredin a stepwise regression(forward).In the first mix of one brandwas specifiedas a step, the marketing functionof all otherbrands'marketing mix withlags and leads of one and two periods.Only variableswith t-statistics greaterthan 1.2 were retainedfor the second step as describedbefore. In that secby the same procedure ond step, the competitiveentry dummyvariableswere introduced.For the OTC-gyn market,where multiple entriesoccurred,only the currentvalue of the dummy for the entry of brandE was specified, whereasup to two-periodleads and lags were specified and entered stepwisefor the entryof brandF. BrandA-plus was not included witha dummyvariable, as it is not a new brand. The reactionsto brandA's total advertising are already modeled in step one of the procedure.In a final step, the variablesthathad been enteredwere forcedinto the model andthe leads andlags for the last entry(brandE) in a forward were investigated stepwiseregression.This procedurewas followed for each brandthat advertised (i.e., A, B, andD). The homogeneity(pre/post entries) for the coefficientsof the decision equationswas estabof the parameters lishedby testingthe stability when new observations were added(Maddala1977). All the tests in both datasets. provedinsignificant This procedureled to the following specificationof
the model for each brand in the OTC-gyn market.
(3) A1(t) = ea'oA2(t - 2)aA4(t - 2)''2e1'3D'+2)eelf
A2(t) = eQ2A l(t 1)a2IA4(t)a22eoa23D2(-2)eaUD2430)e2()

Marketing decision equations. The decision variable

Table 2
Independent variables Airline 1 Airline 2 Airline 3

Lagged share Advertising share

.428 (.94)

.371 (1.29)

1.290 (2.76)

Shareof number
of flights Relative price

(21.52) -.035

(1.10) .907
(10.57) -

(13.55) -.185

Quarter 2 Quarter 3 Quarter 4

(.40) -.066
(2.44) .045 (2.04)

(2.86) -.073 (2.26) -.061 (1.98) -.041 (1.53)

(2.38) -.187

Numberof observations 23 23 23 .977 .982 .985 R2(basedon OLS) in are Variables with t-values less than 1.2 "t-statistics parentheses. in OLS estimationare omittedand indicatedby a dash.

.016 (.68)

A4(t) =

eJe 43192(


where: A,(t) = advertising expenditures at time t for brand A (i = 1), brand B (i = 2), and brand D (i = 4),

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions


Dk(t) = value of entry dummy at period t for brand F


(k = 2) and brand E (k = 3) (value is 0 before ei(t) = disturbance term.

entryand 1 afterentry), ae's= reactionfunctioncoefficients, and

of deGiven this model specificationwith simultaneity model we decision estimated the cisions, advertising by Table3 reports the results.Brand leastsquares. three-stage A's andbrand B showsthe largestreactions to bothbrand D's advertising (reactionelasticitiesare 5.639 and4.21, in relation to brand with a lag of one quarter respectively) to D advertising. BrandA reactsrelatively strongly brand D's advertising(.329), whereas it avoids face-to-face with B (the smallbut negativereactionelascompetition occurwith a two-quarter ticity is -.028). Both reactions lag. Brand D seems to make its advertisingdecisions of its competitors. independently These reactioncoefficients (elasticities) demonstrate the asymmetryof competitivebehavior. The estimates of the coefficientsindicatethat brandsA and B reacted exto F's entryby increasingtheir level of advertising for A and 5.767 for The B). (.794 negative penditures coefficientfor the effect of F's entryon the advertising level of D (-1.329) may reflecta decisionto reallocate resourcesacrossbrandswithin the manufacturer's portfolio. In contrast,in reactionto brandE's entry, D increased its advertisingexpenditures(.845) whereas B its expenditures decreased (- 11.78) andA did not react. The purposeof our study is to explain the diversityof to note beforeproceedthese reactions.It is interesting A anticipated F's entryandbrand ing, however,thatbrand D anticipated E's entry (both lead two quarters).4 They before the both increasedtheir advertising expenditures entries. Table 1 also shows that D's marketshare was not affectedby E's entry, whereasF's entryhad a neg-

ative effect on A's share. These outcomes can be explainedpartlyby the competitiveresponseanticipation and by the fact that D's advertisingelasticity is much largerthanA's. Forthe airlinedata,the flightdecisionmodelresulting from the analysisis shown in equation4. (4) FI(t) =
ea'oP2(t)a"P3(t)a2A2(t)a'3A2(t - 1)O.4 SA3(t).lSF3(t).6eal7DI(t)ell(t)
F2(t)= F3(t) =
)ef2() eaIoP2(t)a2,P3(t)a"Fi(t)a3ea27D(t


Pi(t) = Fi(t) = A,(t) = D,(t) = price of airline i at time t (i = 1,2,3), number of flights of airline i at time t, advertising expenditures at time t, dummy variable for entry of airline 4,

a's = reactionfunctioncoefficients, and

Ei(t)= disturbance term.

The three-stage least squareestimationresultsare given in Table 4. The interpretation of the resultsis straightforward.For competitivebehaviorin general, airline 1 seems to competedirectlywith airline2, which has the highestmarketshare.Whenairline2 decreasesits price or increasesits advertising,airline 1 (with the second largestshare)reactsby increasingits numberof flights behaviorin relationto airline 3. In responseto airline 3's decreasein price, increasein advertising, or increase in flights, airline1 decreasesits number of flights (.825, behav-.178, and -.464), suggestinga "cooperative" ior with airline3 againstthe leader. Airline 2 avoids competingin reactionto advertising changesby its competitors.However, it has a tendency
(ao of airline2, thoughthe coefficientis not statistically significant. Further,airline 2 offers more flights when its rivals raise theirprices (.392 and .380). = .145) to react to increases in the number of flights (-.890 and .150). Airline 1 has the opposite competitive

of brand A managers thatbrand (postanalysis) 4Interviews confirmed was boostedbefore entryin F's entry was expectedand advertising of the new competition. anticipation

Table 3
variables Brand A Brand B Brand D

Constant BrandA advertising

Brand B advertising




-.028 (1 = 2)

(3.39) (3.60)

5.639 (1 = 1)


BrandD advertising BrandF entrydummy BrandE entrydummy

(1.72) .329 (1 = 2) (1.78) .794 (L = 2) (4.62)

4.21 (1.788) 5.767 (1 = 2) (1.87) -11.78


-1.329 (5.05) .845 (L = 2)


in parentheses L = n meansa lead of n periodsand I = n means a lag of n periods. are t-statistics; "Numbers

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions



Table 4
Independent variables Constant Airline 1 8.640 Airline 2 1.69 Airline 3 9.061

of airline thereactions to theentry and represents 4),5 a + IlkNN2 enables us to express the following 0ik = Y-0 hypotheses.
< 0 Hi: Y~ok H2:Yk > 0

1 Price,airline
Price, airline 2

(6.85) -.890


(5.67) -.425 (2.08)

an inelastic marketinstrument (when ,a is small) and is adverinstrument large). In this case, the marketing and of for market the number the OTC-gyn flights tising for the airlinemarket. In otherwords,if elasticityis low, the interceptterm dominates.We expect to be y y negative, indicatingnegativereactionwith a low elas(H,). As elasticity ticityinstrument grows,the Ylk-,2 term (positiveby H2)dominates,indicatingpositivereaction. These two hypothesesalso containthe idea thatthereis a mediumrangeof elasticitywhere no reactionscan be expected (P2). Therefore,the two researchhypotheses stated completelycoverthe threetheoretical propositions before. deResults. For the OTC-gynmarket,the advertising cision model (3) was reestimated with the constraint on the coefficients as hypothesized.The process function was not appliedto the brandF entrydummycoefficient for brandD (c43) becauseboth brandsare marketed by the same company. Consequently,the coefficient does but instead repnot have a competitiveinterpretation, of company resentsthe reallocation resources to multiple the coefficient (043) Therefore,constraining products.6 to a processrepresenting a competitiverationalewould be invalid. The processfunctionconstraints are applied across equations.The estimationwas carriedout by usLSQ, which appliesto our case ing the TSP procedure across of simultaneous equationswith linearconstraints equations. of the Table 5 shows the estimatesof the parameters decisionfunctionsfor the OTC-gynmarket, advertising includingestimatesof the coefficientsof the linearconstraint,which constituteour hypothesistest. The signs
5The processfunctiondoes not containan errorterm;the EGLS estimatethat would result has unknownproperties given the small is only moreefcross-sectional sample,becausethe EGLSestimator A specificationwithoutan errorterm in the ficient asymptotically. test of the hypothesis, process functionactuallyprovidesa stronger is asymptotically less efficient. becausethe estimator andextent of ourpropositions 6Therationale appliesto thedirection of competitive reaction,andnot the speedof reaction.It is clearfrom Table 3 that A's reactionto F's entry and D's reactionto E's entry thanB's reaction.Brand werefaster(evenbeforetheentriesoccurred) D's coefficientrepresenting changes in its advertisingexpenditures withthe entry,as it reflects dueto brand F's entryis contemporaneous bothbrands. of the companymarketing the portfolio Though planning of reaction is an important dimension the speedof reaction (Heil 1987), its analysisis beyond the scope of our article. Therefore,the constraints by the processfunctionwere appliedsimilarlyto represented of the speed of the reaction. all reactions due to an entry,regardless The assumption is thatthe explanation by the constraints represented whetherthe reactionis fast or slow. does not distinguish

H, expresses the idea that firms retreat (oak< 0) with

pi2 is

attack (ask > 0) with an elastic instrument (when

3 Price,airline 2 airline Advertising,

Advertising, airline 2

(5.39) .825 (5.33) .150 (5.03)


(2.83) .380 (2.78)


.073 (1.62)

(lagged) Advertising,airline3

(1.88) -.178


of flights, Number
airline 1 airline3
Number of flights, Entry dummy

(4.08) -.464


(.077) -.310

-.225 .239 -.459

in parentheses "Numbers are t-statistics.



in responseto a changein the numbutby "cooperating" ber of flights (-.310). Airline3 respondsonly to airline 2's changein advertising expenses (.073). The resultspertaining to new entry show that airline 2, with the highest number-of-flights elasticity, reacted to the entrant by increasingits numberof flights (.239), This analysisof competitiveresponsesshows thatreactionsin generaland reactionsto a new entrydifferby An explanation for thesedifferences was tested competitor. empiricallyin the two marketsstudied.
Hypotheses but the other two competitors decreased theirs (-.225 and -.459).

of flights in response to a price cut of airline 1 (-.425),

Airline3 shows a somewhatless generalizable behavior. It competeswith airline 1 by increasingits number

Ourprincipal interestis the coefficientsof the dummy variablesin the reactionfunctionequations(3 and 4 rechangesin spectivelyfor each market),which represent firm i in advertising by responseto the new expenditures entryor entries.These coefficientsarethe a,L's andca4's in the OTC-gynmarketand the Ca,'s in the airlinemarto the propositions,the following ket. Corresponding formulated in termsof the model pacan be hypotheses rameters. function andk = 7 for the airline as a process market) to theentryof brand the reactions (where a,3represents the reactions to the entryof brand F, a,4 represents E,
Expressinga&i(with k = 3, 4 for the OTC-gyn market

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions



Table 5
Independent variables Brand A 1 Brand B 2 Brand D 3

Brand A advertising Brand B advertising

3.517 (2.69)
-.025 (1 = 2)

-63.373 (2.98)
5.211 (1 = 1)

5.968 (58.37)

(2.63) BrandD advertising BrandF entrydummy (1.25) .044 (1 = 2) (2.04) 5.234 (2.08) -3.575 + 61.48 p, (1.71) (2.05) -1.042 + 16.23 p, (1.62) (2.37) -1.513 (5.16)

BrandE entrydummy

Where(fromTable 1):

=0 022

= .067 = .137

I = n meansa lag of n periods. are t-statistics; in parentheses "Numbers

pected sign (Yo7= -23.08,

andmagnitude of the coefficientsof the linearconstraint on the new entry dummyvariables,boxed in Table 5, support Hi andH2. Coefficientsy arenegative(-3.575 and -1.042 respectivelyfor the entriesof brandF and brand E) andcoefficients Ylkarepositive(61.48 and 16.23 respectively). These findings indicatethat brandsfor which advertising has little impacton marketshare(3, is low) react to the entryby decreasing theiradvertising expenditures. This reactionis in accordwith the idea thatfinns do not fight with weaponsthat are not highly effective. However, as brandadvertising increases,finrmns elasticity(03,2) cut back by smallerincrements(at grows largeras the termcompensatesfor the negativeimpactof yo). Y1kP,2 Eventually,the reactionturnspositive. For brandF, the = .0581; for greaterelasticities, turningpoint is at P3,2 the reactionis positive. For brandE, reactionsturnpositive for firmswhose advertising than elasticityis greater .0642. These resultsapply to the brandsin this market acrosstwo entries.Thoughit does not involve manyentries and many brands,the model is estimatedsimultaneouslywith all the datain the sampleacrossbrandsand time. Consequently, modelparameters are estimated with relativelylarge statisticalpower. By the same procedure,the airlineanalysisprovides a replicationof the OTC-gynmarket.Table 6 indicates thatthe coefficientssupport our hypothesis.The y coefficients are both statistically significant with the exy17 = 25.65).

fight with weaponsthat are not highly effective. However, as flight elasticity (3,2) increases, finns cut back increments term (a, growslargeras the ya/I3,2 by smaller

Table 6
Independent variables Airline 1 Airline 2 Airline 3


1 Price,airline
Price, airline 2

5.896 (5.80)

1.318 (2.20)

7.950 (5.85) -.322 (1.86)

Price, airline3 Advertising,airline2

(6.43) .905

(3.86) .336

.138 (5.13) (1.90) -.150 (3.82) -.0001


airline 2 Advertising,
(lagged) Advertising,airline3 Numberof flights, airline 1
Number of flights,


.055 (1.42)

.168 (1.81)

-.100 (1.88) -.184 (1.21)


airline 3

-23.08 + 25.65k,1

Entrydummy Where(fromTable 2):

(5.79) (5.80)

This finding indicates that finnrms whose flight decisions affect their own sales very little (13,2 is low) react to the new entryby decreasing theirmarket effort (number of flights), in accordwith the idea thatfirmsdo not

.905 .907 032= .880

12 = 22 =

in parentheses are t-statistics. "Numbers

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions

54 compensates for the negative impact of Yk).The turning point from negative to positive reaction occurs for flight elasticity greater than .90; beyond that point, firms retaliate against the entrant by offering more flights. DISCUSSION AND CONCLUSION Our results offer some support of the hypotheses proposed. In two unrelated oligopolistic markets, we observe considerable variation in reaction to three entries. The effectiveness of the major marketing instrumentsadvertising expenditures and number of flights for OTCgyn and airlines, respectively-is shown to predict how competitors react to a new entry: that is, whether they increase or decrease their efforts. In accord with the literature on competitive behavior, we find that firms react with positively to entrants(by retaliationor counterattack) their effective weapons, where "effective" means having relatively large elasticity. Also consistent with the literature, firms cut back (withdraw) their inelastic marketing mix instruments. Undoubtedly many factors influence a competitors's reaction. Nonetheless, our results suggest that reactions can be better understood and predicted by observing one factor: the effectiveness of a current competitor's marketing mix instruments. The contribution of our study is to suggest when a positive reaction with a marketing instrumentwill be observed and when a negative reaction will be observed instead. The criterion, elasticity, is relatively easy to measure and conforms with extant theory in industrial organization and strategic marketing. The ideas expressed here are very general; they can be applied to various marketing mix instruments and to multiple scenarios. We consider response to the entry of a new product, but one could substitute, for example, the repositioning of a current brand. Repositioning, if successful, operates much as does an entry by changing the competition set. Other competitors may react by either increasing or decreasing their marketing effort, depending on the effectiveness of that effort. Our study has certain limitations. First, in both industries studied, elasticities were not significantly altered by the entrant. In other settings, an entrant may shift elasticities of competitors, thereby affecting their reactions and complicating the prediction of competitive response. Second, inferences that can be made from modeling one industry are necessarily limited to that industry. The model is merely a summarizing of the data evidence. It is the nature of scientific research that knowledge progresses by convergent studies and replications (Zaltman, Pinson, and Angelmar 1973). We provide such a replication by testing our hypothesis in two industries. Nevertheless, more replications are necessary for generalization. Methodologically, given the limited data, any significance at the usual confidence level provides a strong rejection of the null hypothesis of no effect. Nonetheless, the small sample size does limit the scope of the effects that can be observed and measured. A natural

FEBRUARY 1989 JOURNAL OF MARKETING RESEARCH, progression of our research therefore would be to increase the number of explanatory variables in order to yield a wider explanation of competitive moves and countermoves. For example, environmental uncertainty may affect response, perhaps blunting reactions or even leading firms not to react. Another consideration may be the firm's impression of the entrant's capabilities. A firm may react strongly to an entrant judged to be capable and may not react to another entrantregarded as a minor threat. Further research also is needed to establish the generalizability of the ideas developed here, which are a step toward a more complete theory of competitive reactions, in particularas it discriminates between current competitors' reaction strategies. Nevertheless, our study does present an explanation, with empirical support in two industries,for observabledifferencesin firms'/brands' reaction patterns to entries of new competitors in the market. As such, the study enhances our understanding of a complex and crucial strategic issue-the pattern of competitive rivalry. REFERENCES ReAckoff, R. L. and J. R. Emshoff (1975), "Advertising search at Anheuser-Busch, Inc. (1963-68)," Sloan ManBass, FrankM. and LeonardParsons(1969), "Simultaneous EquationRegressionAnalysis of Sales and Advertising," Markets: An UprisBaumol,WilliamJ. (1982), "Contestable AmericanEcoing in the Theory of IndustryStructure," Beckwith,Neil E. (1972), "Multivariate Analysisof Sales ReJournal of sponses of CompetingBrandsto Advertising," AndrewMitchell,andRichard Staelin Chakravarti, Dipankar, (1981), "JudgmentBased Marketing Decision Models: (Fall), 13-23. Chow, G. C. (1960), "Tests of Equality Between Sets of in Two Linear Coefficients Econometrica, 28, Regressions," 591-605. Cross-ElasticiClarke,DarralG. (1973), "Sales-Advertising ties and AdvertisingCompetition," Journal of Marketing Research, 10 (August),250-61. andJohnP. Olson(1972), D., JamesG. March, Cohen,Michael "A Garbage Can Model of Organizational Choice,"AdminReCooper,ArnoldC. and Dan Schendel(1976), "Strategic sponses to TechnologicalThreats,"BusinessHorizons, 19 61-9. (February), M. andJamesM. March(1963), A Behavioral Cyert,Richard Theoryof the Firm. EnglewoodCliffs, NJ: Prentice-Hall, Inc.
Day, George S. (1984), Strategic Market Planning: The Pursuit of Competitive Advantage. St. Paul, MN: West Pubistrative Science Quarterly, 17 (March), 1-25. Problems and Possible Solutions," Journal of Marketing, 45 Marketing Research, 9 (May), 168-76. nomic Review, 72 (March), 1-15. Applied Economics, 1 (2), 103-24. agement Review, 10 (Winter), 1-15.

lishing. as a Moderator of the Gatignon,Hubert(1984), "Competition search, 21 (November),387-98. Graham,David R., Daniel P. Kaplan, and David S. Sibley in the Airline Indus(1981), "Efficiencyand Competition
Effect of Advertising on Sales," Journal of Marketing Re-

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions


try," working paper, Civil Aeronautics Board (September). Hall, William K. (1980), "Survival Strategies in a Hostile Environment," Harvard Business Review, 58 (September-October), 75-85. Hanssens, Dominique M. (1980), "Market Response, Competitive Behavior, and Time Series Analysis," Journal of Marketing Research, 17 (November), 470-85. Hauser, John R. and Steven M. Shugan (1983), "Defensive Marketing Strategies," Marketing Science, 2 (Fall), 31960. Heil, Oliver (1987), "Exploring and Predicting Competitive Reaction: A Marketing Signaling Approach," doctoral dissertation, University of Pennsylvania. Judge, George G., W. E. Griffiths, R. Carter Hill, Helmut Litkepohl, and Tsoung-Chao Lee (1985), The Theory and Practice of Econometrics. New York: John Wiley & Sons, Inc. Kotler, Philip and Ravi Singh (1981), "MarketingWarfare in the 1980s," Journal of Business Strategy, 10 (Winter), 3041. Lambin, Jean-Jacques (1976), Advertising, Competition and Market Conduct in Oligopoly Over Time. Amsterdam:NorthHolland Publishing Company. Philippe A. Naert, and Alain Bultez (1975), "Opti--, mal Marketing Behavior in Oligopoly," European Economic Review, 6, 105-28. Little, John D. C. (1966), "A Model of Adaptive Control of Promotional Spending," Operations Research, 14 (3), 1075. Maddala, G. S. (1977), Econometrics. New York: McGrawHill Book Company. Metwally, M. H. (1978), "Escalation Tendencies of Advertising," Oxford Bulletin of Statistics, 153-63. Miles, RobertH. (1980), Macro-OrganizationalBehavior. Santa Monica, CA: Goodyear Publishing. Naert, Philippe and Peter Leeflang (1978), Building Implementable Marketing Models. Boston: Martinus Nijhoff, Social Sciences Division. Oxenfeldt, Alfred R. and William L. Moore (1978), "Customer or Competitor: Which Guideline for Marketing?" Management Review, 67 (August), 43-8. Parsons, Leonard and Randall L. Schultz (1976), Marketing Models and Econometric Research. New York: North-Holland Publishing Company.

Pekelman, Dov and Edison Tse (1980), "Experimentationand Budgeting in Advertising: An Adaptive Control Approach," Operations Research, 28 (2), 321-47. Porter, Michael E. (1979), "How Competitive Forces Shape Strategy,"Harvard Business Review, 57 (March-April), 13745. Reibstein, David J. and Hubert Gatignon (1984), "Optimal ProductLine Pricing: The Influence of Elasticities and CrossElasticities," Journal of Marketing Research, 21 (August), 259-67. Robertson, Thomas S., Scott Ward, and William M. Caldwell IV (1982), "Deregulation: Surviving the Transition," Harvard Business Review, 60 (July-August), 20-5. Rothschild, William E. (1979), "Competitor Analysis: The Missing Link in Strategy," Management Review, 68 (July), 22-38. Scherer, F. H. (1980), Industrial Market Structure and Economic Performance. Chicago: Rand McNally College Publishing Co. Schmalensee, Richard (1976), "A Model of Promotional Competition in Oligopoly," Review of Economic Studies, 43, 493-507. Schultz, Randall L. (1971), "Market Measurement and Planning with a Simultaneous-Equation Model," Journal of Marketing Research, 8 (May), 153-64. and Dominique H. Hanssens (1977), "Logical Implications of Competitive Behavior: An Approach to Model Specification," paperpresentedat Joint National ORSA/TIMS Meeting, Atlanta. Weitz, Barton A. (1985), "Introduction to Special Issue on Competition in Marketing," Journal of Marketing Research, 22 (August), 229-36. Wildt, Albert (1974), "Multifirm Analysis of Competitive Decision Variables," Journal of Marketing Research, 11 (February), 50-62. Williamson, Oliver E. (1965), "A Dynamic Theory of Interfirm Behavior," Quarterly Journal of Economics (November), 579-607. Zaltman, Gerald, Christian R. A. Pinson, and Reinhard Angelmar (1973), Metatheory and Consumer Research. Hinsdale, IL: Dryden Press.
ReprintNo. JMR261103

This content downloaded from on Wed, 9 Oct 2013 16:17:08 PM All use subject to JSTOR Terms and Conditions