Breaking Free of a Stereotype: Should a Domestic Brand Pretend …cdn.persiangig.com/preview/ZwyMhAN7xN/zhang2015.pdf · 2019. 4. 12. · of brand origin is largely driven by consumers

This article was downloaded by: [130.133.8.114] On: 09 May 2015, At: 22:32Publisher: Institute for Operations Research and the Management Sciences (INFORMS)INFORMS is located in Maryland, USA

Marketing Science

Publication details, including instructions for authors and subscription information:http://pubsonline.informs.org

Breaking Free of a Stereotype: Should a Domestic BrandPretend to Be a Foreign One?Kaifu Zhang

To cite this article:Kaifu Zhang (2015) Breaking Free of a Stereotype: Should a Domestic Brand Pretend to Be a Foreign One?. Marketing Science

Published online in Articles in Advance 21 Apr 2015

. http://dx.doi.org/10.1287/mksc.2014.0895

Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions

This article may be used only for the purposes of research, teaching, and/or private study. Commercial useor systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisherapproval, unless otherwise noted. For more information, contact [email protected].

The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitnessfor a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, orinclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, orsupport of claims made of that product, publication, or service.

Copyright © 2015, INFORMS

Please scroll down for article—it is on subsequent pages

INFORMS is the largest professional society in the world for professionals in the fields of operations research, managementscience, and analytics.For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org

http://pubsonline.informs.org

http://dx.doi.org/10.1287/mksc.2014.0895

http://pubsonline.informs.org/page/terms-and-conditions

http://www.informs.org

Articles in Advance, pp. 1–16ISSN 0732-2399 (print) � ISSN 1526-548X (online) http://dx.doi.org/10.1287/mksc.2014.0895

© 2015 INFORMS

Breaking Free of a Stereotype: Should a DomesticBrand Pretend to Be a Foreign One?

Kaifu ZhangTepper School of Business, Carnegie Mellon University, Pittsburgh, Pennsylvania,

[email protected]

Consumers in many emerging markets exhibit a pronounced preference for western and global brands, whiledomestic brands are often associated with a cheap but low quality image. Frustrated with the negative

country-of-origin (COO) stereotype imposed on them, many domestic brands from emerging markets follow avariety of approaches to disguise their COO and pretend to be foreign. This paper studies the strategic aspectsof this phenomenon. We consider an experience good market where consumers learn about each brand’s quality,using its COO as a prior. Each firm can invest in product quality or COO dissociation; the former generatesbetter quality signals while the latter simply masks the firm’s COO identity. The analysis reveals a few maininsights. Sharing a reputation leads to a common good problem where firms free-ride on each other’s qualityinvestments. As a high quality firm dissociates itself from the stereotype, it ceases to contribute to the COOimage but also prevents its low quality peers from free-riding on it. This incurs a negative direct effect buta positive strategic effect on the group image. Consequently, a country’s COO image may actually improvewhen more high quality brands shun their identities and pretend to be foreign. In equilibrium, COO imageimproves monotonically as firms become more efficient in providing quality. However, the prevalence of COOdissociation may first increase then decrease. We discuss the implications of these results for emerging marketbrands as well as policy makers.

Keywords : game theory; branding; country-of-origin effectHistory : Received: January 9, 2013; accepted: September 5, 2014; K. Sudhir served as the special section

editor-in-chief and Yuxin Chen served as associate editor for this article. Published online in Articles inAdvance.

1. IntroductionThe purchasing prowess of emerging market con-sumers has long been under the spotlight as com-panies seek the next growth opportunities. McKinseyand Company (2012b) estimates that by 2025, half ofthe world’s population will join the consuming classand the total annual consumption in emerging mar-kets will grow to $30 trillion. While these rapidlygrowing markets pose enormous opportunities, manyemerging market consumers have exhibited a pro-nounced preference for foreign and multinationalbrands, creating a real disadvantage for aspiring localbrands. For example, a 2013 survey (World Brand Lab2013) revealed that in China, the top 10 brands thatenjoy the most brand loyalty were all foreign. By con-trast, in Japan, the top 10 brands were all domes-tic. Across a large number of product categories,McKinsey and Company (2012a) found a clear prefer-ence for foreign brands by Chinese consumers; evenmore worrisome for the domestic firms, this prefer-ence for foreignness is more pronounced among theyoung and newly affluent consumer cohorts. Indeed,despite the large differences among emerging mar-kets, this preference for western and global brands

seems to be a robust feature of consumer mentality inmany developing countries (Credit Suisse 2014).

In general, favorable or unfavorable country stereo-types stem from the performance history of a coun-try’s products (Maheswaran et al. 2009). Admittedly,the negative country-of-origin (COO) images are oftenaccurate reflections of emerging market firms’ over-all lack of technical sophistication. Frustrated withthis negative association, however, many emergingmarket firms have made various attempts to de-emphasize or disguise their COO identity. While afirm cannot outright lie about its COO, it can manip-ulate consumer perceptions about its COO. As shownin a number of studies (Samiee et al. 2005, AndersonAnalytics 2007), consumer recognition of a brand’sCOO is typically imperfect and sometimes malleable.Samiee et al. (2005) demonstrate that the recognitionof brand origin is largely driven by consumers asso-ciation of brand names with languages that suggestcountry origins. Similarly, Leclerc et al. (1994) showthat American brands who choose French soundingnames can successfully affect the perceived hedonismof their products. In a survey study, Anderson Ana-lytics (2007) demonstrated that consumers have very

1

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.

Zhang: Should a Domestic Brand Pretend to Be a Foreign One?2 Marketing Science, Articles in Advance, pp. 1–16, © 2015 INFORMS

limited knowledge about real-world brands’ COO,including some very well known brands.

Indeed, choosing a foreign sounding brand namehas become a popular practice among emergingmarket firms who are concerned about their COO.In some cases, firms may deliberately choose brandnames that trigger more favorable COO associations.For example, Giordano and Bossini, despite its Italian-sounding name, is a successful Hong Kong basedclothing brand with outlets across Asia, the MiddleEast, and North Africa. Rossini, a Chinese watch-maker who prides itself as the top domestic brand ina market dominated by foreign competition, ironicallyhas an Italian sounding brand name.1 Other exam-ples include garment brand “Metersbonwe” based inSouth China, and outdoor brands “Kailas” and “KingCamp.”2 While the latter set of brand names may notnecessarily suggest a particular COO, they are defi-nitely foreign sounding to Chinese ears. Similar exam-ples abound in India, with brands such as La Opala,Da Milano, Franco Leone, and Munich Polo sellinghigh quality, premium products to the domestic mar-ket, but under foreign sounding names (The EconomicTimes 2012).

In addition to the brand name, a plethora of othercues can shape COO perception. Store display, culturesymbols in the brand logo, and the accent and eth-nicity of ad models all have the potential to influenceCOO recognition. For example, Canada’s CompetitionBureau points out that misleading COO claims can bemade through the use of pictures and symbols, e.g.,the maple leaf.3 Walch, a Chinese firm producing anti-septic liquid, uses a medieval-European style helmetas its brand logo in addition to its German soundingbrand name. Another Chinese outdoor brand, Can-Torp, chooses to only display its English name onmany of its storefronts. Its logo features the silhouetteof a camel that is strikingly similar to the Americancigarette brand except that the camel has two humpsrather than one. The Chinese baby formula brand“Scient” used a stars-and-stripes pattern in its pack-aging.4 A host of other Chinese brands use Europeanstyle coats-of-arms as their brand logos. The EconomicTimes (2012) reported that Munich Polo prefer usingwestern models in its ads to reinforce the perceptionamong Indian consumers that it has a German origin.

Interestingly, the above mentioned examples are allsuccessful domestic brands that produce reasonably

1 See http://www.rossini.com.cn/en/index.aspx for a history of thecompany.2 In China’s case, once a foreign sounding name is chosen, the com-pany usually then transliterates the name into Chinese characters.3 See http://www.canadianadvertisinglaw.com/?p=3573.4 This company eventually went too far in claiming its Americanorigin and was sued by regulators for deceptive advertising.

high quality products. While it is easy to understandtheir desire to be less Chinese/Indian and more for-eign, the strategic implications of COO dissociationare less clear. Building on the rich literature on COOeffect, umbrella branding, and statistical discrimina-tion theory, this paper studies the strategic aspects ofthe COO stereotype. We ask (1) how does a sharedreputation affect firms’ incentives to provide quality?(2) What governs firms’ incentives to invest in COOdissociation to break free of the stereotype? (3) As acountry’s high quality firms mask their COO iden-tity, what are the implications for the country’s COOimage? (4) As a group, what are the paths the firmscan follow to improve the collective image? We con-sider a group of firms who share the same COO.Because quality signals are noisy, a brand’s COOserves as a prior when consumers learn about eachfirm’s quality, in a Bayesian fashion. The COO stereo-type is determined endogenously and corresponds tothe group’s aggregate past performance. Each firmcan invest in its product quality or disguise its COOidentity. Quality investment improves the chance thata firm’s product will generate a positive quality sig-nal. COO dissociation, on the other hand, is a costlyaction that merely shapes consumer perception abouta firm’s COO. When a firm invests in COO dissoci-ation, the consumers are less likely to perceive thisfirm to be from the focal country. Therefore, they areless likely to apply the COO prior to judge this firm’squality.

Our first set of results confirm the main insightsfrom the collective reputation literature (Tirole 1996,Winfree and McCluskey 2005). In particular, sharedreputation creates a common good problem whereeach firm has the incentive to free-ride on others’quality investment. This leads to insufficient qualityinvestment compared to the socially efficient level.When consumers have moderate tendencies to stereo-type, shared reputation hurts the more efficient firms(i.e., ex-post, high quality firms) but the less efficientfirms (i.e., ex-post, low quality firms) benefit from thefree-riding opportunity. However, when consumers’stereotyping tendency is strong (i.e., individual qual-ity signals are extremely scarce or noisy), the possi-bility of free-riding hurts high and low quality firmsalike. This happens because free-riding hurts incen-tives so much that even the most inefficient firmswould prefer the no-stereotyping benchmark.

In a second step, we study the case wherein firmscan invest in COO dissociation to break away fromthe stereotype. Typically, the most efficient firms, i.e.,those who suffer the most from the free-riding prob-lem, also have the strongest incentives to break awayfrom the stereotype. As a high quality firm dissociatesitself from its COO, it prevents the rest of the groupfrom free-riding on its quality investment. This incurs

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.

Zhang: Should a Domestic Brand Pretend to Be a Foreign One?Marketing Science, Articles in Advance, pp. 1–16, © 2015 INFORMS 3

a negative direct effect on the COO image. However,facing diminished free-riding opportunities, the lessefficient firms now have stronger incentives to investin quality. This translates into a positive strategiceffect on the COO image. Overall, when consumers’tendency to rely on the COO stereotype is moder-ate, the direct effect dominates the strategic effect.When consumers have strong tendencies to rely onthe COO stereotype, the strategic effect comes to dom-inate. This leads to an interesting scenario in whichthe COO image improves as more high quality firmsdissociate from their COO identity.

Finally, we investigate the firms’ joint decisions onquality investment and COO dissociation. We dis-cuss comparative static results focusing on two keyparameters, i.e., consumers’ tendency to stereotypeand firms’ cost of quality (i.e., technical efficiency).As expected, the equilibrium COO image increasesmonotonically as firms gain greater technical effi-ciency and as consumers rely less on the COO stereo-type. Firms’ incentives to dissociate from the COO, onthe other hand, diminish as consumers rely less on theCOO stereotype. Interestingly, when firms make col-lective improvement in technical efficiency, the impacton COO dissociation is dubious. On one hand, a firmtypically finds COO dissociation more attractive asit makes progress in product quality. On the otherhand, when firms collectively improve quality, theCOO image improves, which makes it less imperativefor each firm to break away from the stereotype. As aresult, the prevalence of COO dissociation may firstincrease then decrease when firms make progress intechnical efficiency.

We organize the rest of the paper as follows. In §2,we discuss the innovation of the present paper in rela-tion to the literature. Section 3 presents the model set-up. The main results are presented in §4. Section 5provides summarizing remarks and discusses futureavenues of research.

2. Literature ReviewOur work is related to several literature streams. First,it is related to the rich literature on the COO effect.Maheswaran et al. (2009) and Dinnie (2004) provideexcellent surveys of this literature. Early studies inthe COO literature established that COO does influ-ence consumer evaluation of a product, by servingas an information cue (Bilkey and Nes 1982, Han1989). We model COO effect following the informa-tional approach. We do not consider the situationswherein factors external to the products (e.g., hostil-ity between countries) may influence consumer choice(Maheswaran and Chen 2006, Hong and Kang 2006).

More recently, a number of studies have shown thatconsumers’ ability to recognize a brand’s COO is in

fact quite limited and often malleable (Samiee et al.2005, Liefeld 2004). Sometimes, consumers cannot tellthe COO of a brand. When they do reach a judgment,it is strongly influenced by the association betweenbrand names with languages that are suggestive ofCOOs (Samiee et al. 2005). Leclerc et al. (1994) foundthat a French sounding name increases the perceivedhedonism of a product made in the United States.Overall, these studies point to the importance of lin-guistic cues in influencing consumer perception ofa brand’s COO. This is consistent with the popularpractice of adopting a foreign sounding name to maskone’s COO. We develop a general model of COO dis-sociation that includes, but is not limited to, the caseof foreign naming.

Second, our work is related to the marketingliterature on umbrella branding (Wernerfelt 1988,Montgomery and Wernerfelt 1992, Erdem 1998,Erdem and Sun 2002, Anand and Shachar 2004,Hakenes and Peitz 2008, Cabral 2009) and brandextension (Loken and John 1993, John et al. 1998,Simonin and Ruth 1998). Similar to the papers in thisliterature, our model revolves around the notion ofreputation spillover. There is a key difference betweenour model and umbrella branding. When a multi-product firm adopts an umbrella branding strategy,it can coordinate the quality or pricing decisionsfor all of the products to maximize the total profit.By contrast, independent firms who share a com-mon reputation will make decisions without internal-izing other firms’ payoffs. This leads to interestingreputational externalities. These are the focus of thisstudy. We allow self-interested firms to unilaterallydissociate themselves from the group identity. To ourknowledge, this element is not present in the umbrellabranding literature.

Finally, our paper is closely related to the eco-nomic literature on statistical discrimination theory(see Fang and Moro 2010 for a review). In hisseminal paper on this topic, Phelps (1972) postu-lates that when a principal observes a noisy signalabout an agent’s performance, her group identity(e.g., race, gender) contains useful information. ThePhelps model corresponds to a Bayesian frameworkwhere an agent’s group identity serves as a prior.The same basic idea is also discussed by Arrow(1973), and Lundberg and Startz (1983) considers acase where human capital is endogenous but groupprior is exogenous. The recent literature on collec-tive and transferable reputation (Tirole 1996, Levin2009, Tadelis 1999) build on the statistical discrimina-tion paradigm, but focus on the possibility that groupreputation is self-fulfilling via a contractual process.We model COO stereotyping following the statisti-cal discrimination paradigm. Consumers are intuitiveBayesian learners who use a brand’s COO as the

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


prior. We consider endogenous quality choice as wellas the endogenous formation of the group prior. Mostimportant, we allow for the possibility of COO dis-sociation. COO dissociation influences the perceivedgroup identity of a focal firm. This departs from theabove mentioned papers in statistical discriminationliterature, all of which (to our knowledge) assumethat the agents’ group identity is exogenous and fixed.

3. ModelConsider N firms who share a common COO, g.While the early literature on COO effect considerscountry image as a single construct, more recentworks have pointed out that COO image is often cat-egory specific (Roth and Romeo 1992). Toys, electron-ics, and footwear from China may suffer from a poorCOO image, but silk and tea do not. The firms weconsider should come from the category(-ies) that,broadly speaking, share the same COO image.5

We consider a simple two-stage game. In the firststage, the firms decide how much to invest in prod-uct quality, a decision denoted as qi, and whetherto engage in COO dissociation, a decision denotedby bi ∈ 80119, where bi = 1 implies that firm i dis-guises its COO identity. Both quality investment andCOO dissociation are costly. In the second stage, con-sumers form the COO stereotype, learn about eachproduct’s quality, and payoffs are realized. We firstexplain our formulation of COO stereotyping assum-ing that a subset of firms engage in COO dissociation.

COO Stereotyping. We model consumer stereotyp-ing following the statistical discrimination paradigm(Phelps 1972, Fang and Moro 2010). In this frame-work, a firm’s group identity serves as a prior whenconsumers learn about the firm’s underlying quality.Consider a subgame where firms choose quality levelsq11 0 0 0 1 qN and a subset B engage in COO dissociation.Each firm i produces a Bernoulli product that gen-erates a positive experience with probability qi anda negative experience with probability 1 − qi. Con-sumers learn about each firm’s quality in a Bayesianfashion.

Consumers have two pieces of information aboutbrand i. They observe i’s COO and a number of qual-ity signals generated by product i. Consumers holdprior belief Beta4Sg1Mg − Sg5 about country g firms’quality which, as we will explain shortly, is derivedfrom past experience with country g’s products ingeneral. They obtain Ms 0–1 signals about each firmi’s performance out of which Ssi signals are posi-tive. As such, consumers’ posterior belief about qi forany i y B is E4qi � Sg1Mg1 Ssi1Ms5. The posterior also

5 We thank an anonymous reviewer for this comment.

Table 1 Definition of Key Parameters

Parameter Definition A function of

Ms The number of quality signals each brandgenerates.

Exogenous

Ssi The number of positive signals brand i

generates.qi

Mg The precision of the COO prior, when COOdissociation decisions are Eb.

Eb

Sg The positivity of the prior. Eq, Eb

� The weight consumers put on the COO prior. Eb

Mgb The “baseline” precision of the COO prior,when no brand engages in COOdissociation.

Exogenous

�b The “baseline” weight consumers put on theCOO prior, when no brand engages in COOdissociation.

Exogenous

follows a beta distribution. When Ssi out of the Ms

signals are positive, the posterior mean is

E4qi �Sg1Mg1Ssi1Ms5=Mg

Mg +Ms

Sg

Mg

+Ms

Mg+Ms

SsiMs

0 (1)

Given the multitude of parameters in the abovemodel, we summarize them in Table 1. Equation (1)reflects the exact idea put forth in Phelps (1972).In essence, consumers’ posterior belief about firm i’squality is a weighted average of the group (prior)mean and the sample mean. As such, a firm’s COOidentity contains useful information when qualitysignals are limited and noisy. We can define � =

Mg/4Mg +Ms5 as the extent to which consumers relyon the group stereotype to judge a brand’s quality.When �→ 1, consumers acquire very little data abouteach individual brand and completely revert to thegroup prior to judge each firm’s quality. When �→ 0,consumers completely rely on the quality signals toinfer each brand’s true quality. Consequently, theCOO stereotype does not affect their judgment. Priorliterature has uncovered many factors that affect theextent to which consumers rely on the COO stereo-type to evaluate brands, such as brand familiarity(Han 1989), category efficiency, and attribute strength(Maheswaran 1994). As we will explain shortly, wetreat Ms as a model parameter but do not makestatements about the behavioral determinants of thisparameter. For example, a small Ms may imply a lackof actual experience or difficulty with recalling anypast experiences (Keller 1993).

Stereotype Formation. We now explain the deter-minants of Mg and Sg in the game. In the Phelps(1972) formulation, the precision and mean of thegroup prior (Mg and Sg in our model) are treated asexogenous parameters. By contrast, we assume thatMg and Sg are endogenously determined, a processwhich we term stereotype formation.

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


It has been shown in the literature that a plethoraof factors can drive the stereotype formation process,such as past experience, word of mouth, media andexpert opinion, etc. See Maheswaran et al. (2009) fora discussion on some of these processes. For exam-ple, the media can make a country’s COO image moresalient by extensively covering the country’s productin a trade dispute or product recall. Jointly, these fac-tors will determine the shape and precision of theprior belief in a complex process. We remain relativelyagnostic on the exact psychological process throughwhich consumers obtain their prior beliefs about thegroup. However, we focus on the case in which aCOO image is an accurate reflection of a country’soverall product quality.6 We make the following twoassumptions about COO stereotype formation

Mg =N −

∑

i biN

Mgb1 (2)

Sg

Mg

=

∑

i41 − bi5qi∑

i41 − bi50 (3)

Equations (2) and (3) introduce additional parame-ters and require some explanation. Let us start fromEquation (2). Recall that bi stands for firm i’s decisionto dissociate from group i. Therefore, N −

∑

i bi cor-responds to the number of firms who do not masktheir COO. Equation (2) essentially states that the pre-cision of the COO stereotype decreases as more firmsbreak away from group g. This reflects the reality thatconsumers have greater opportunities to interact withbrands from country g when there are more brandsfrom that country. Consequently, they are likely tohold a more precise belief about country g’s COO.Therefore, when a subset of brands break away fromCOO g, Mg will decrease accordingly. For the purposeof our analysis, we specify a linear function, Mg =

4N −∑

i bi54Mgb/N5, where Mgb is the baseline pre-cision of the COO prior when no firms break awayfrom the stereotype. When no firm engages in COOdissociation Mg =Mgb. When all firms engage in COOdissociation Mg = 0.

Thus, Mgb is an exogenous parameter in the modelwhile Mg is endogenously determined. Consequently,� as defined above, is also endogenously determined.Because we would like to present our equilibriumresults in terms of �, which has a clear conceptualmeaning, we introduce another exogenous parame-ter �b. We define �b = Mgb/4Mgb + Ms5 as a parame-ter that measures the consumers’ baseline tendency to

6 Indeed, our use of stereotype does not imply any cognitive bias orcarry any negative connotation. As is often the case, producers fromthe same country may share similar cost structures, technical back-ground or even similar work ethics. As such, their common COOdoes contain useful information about each firm’s quality. We thankthe reviewing team for helping us clarify these assumptions.

apply the COO stereotype, when no firm breaks away.The connection between �b and � will be explained inthe appendix.

We next explain the determination of Sg , which isdescribed by Equation (3). Recall that qi is the qualitychoice of firm i. Equation (3) states that given qi, themean of the group prior is a weighted average of theactual quality levels of the firms who do not breakaway from the COO identity. Put differently, the COOis unbiased in the sense that the COO mean equalsthe expected quality of a brand with which consumerswill interact. A brand does not enter into the calcula-tion of country g’s COO image if it disguises its COOidentity (bi = 1).

In our formulation, the COO prior is defined bytwo parameters, i.e., the mean and the precision. Notethat the COO prior distributions cannot be rankedin familiar orders such as stochastic dominance.7 Forexample, in certain product categories, country Amay have a less positive COO image than country B,but having less direct (e.g., product trial) and indi-rect (e.g., media coverage) experiences with productsmade in country A, consumers’ prior belief aboutcountry A’s COO image may also be less precise. Assuch, given the same amount of data, they may putless weight on the COO prior when judging the qual-ity of a country A brand.

Note that in this model, we assume that each brandgenerates the same number of signals, denoted as Ms .In reality, each brand can generate a different numberof signals Msi. When Msi is different for each firm i,consumers may apply a different �i when judgingeach brand’s quality. Similarly, Equation (3) shouldnow be Sg/Mg =

∑

i1 bi=0 Msiqi/∑

i1 bi=0 Msi, reflectingthe fact that consumers have different opportunitiesto interact with different brands; therefore, the qual-ity of the brands carry different weights in the COOformation process.

When a firm increases Msi, it decreases the weightof the COO prior with which consumers learn aboutits quality. Put differently, by endogenizing Msi, wecould study the firms’ incentives to promote theirbrand awareness (i.e., generate more quality signals)as a way to counter the effect of a negative stereotype.This is a different mechanism from disguising one’sCOO identity. We discuss this in detail in §5 as animportant direction for future research. Table 1 sum-marizes the definitions of the key parameters. Notethat the parameters Ms , Mgb, �gb are the only exoge-nous parameters in the model. We will present theequilibrium results in terms of these parameters.

COO Dissociation. In the first stage, each firmchooses whether to engage in COO dissociation. COO

7 We thank one of the reviewers for pointing out this fact.

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


dissociation creates uncertainty about a brand’s COO.When firm i is successful in dissociating from itsCOO, it is perceived to be completely unrelated tocountry g. Consequently, consumers do not judgeits brand by country g’s COO stereotype, nor dothey incorporate firm i’s quality in the stereotypeformation process for country g. Formally, insteadof applying the group stereotype Beta4Sg1Mg − Sg5to judge firm i’s quality, consumers use an alterna-tive, exogenous prior Beta4S01M0 − S05 when updat-ing their beliefs about brand i’s quality. For example,Beta4S01M0 − S05 could stand for the COO image ofItaly when Chinese firms choose an Italian soundingname. For simplicity, we focus on the case whereina firm can always dissociate itself from the group aslong as it incurs the cost.8 Specifically, denote bi ∈

80119 as firm i’s decision to engage in COO dissoci-ation. Consumer prior is therefore S4bi5= 41 − bi5Sg +

biS0 and M4bi5= 41 − bi5Mg + biM0.As discussed in §1, firms who suffer from an infe-

rior COO image follow a variety of strategies to masktheir COO. The variables S0 and M0 may take differ-ent values depending on how exactly a firm disguisesits COO. The variables We highlight one special caseof Beta4S01M0 − S05 that we will use as a continuousexample in our analysis. Consider the special case inwhich M0 is very small, such that Beta4S01M0 − S05 isan uninformative prior. Roughly speaking, this couldbe the case where an emerging market brand i dis-guises its COO by choosing a neutral brand name,which is not associated with any particular foreignculture. As consumers become uncertain about thisbrand’s COO, they do not apply the g stereotype tojudge its quality. Moreover, since no specific COO ishinted at by the name, consumers cannot apply anyspecific COO stereotype to help them judge brandi’s quality. Consequently, they judge brand i’s qual-ity by putting more weight on its own quality sig-nals, Ssi and Ms . In effect, � approximates 0 as firm iinvests in COO dissociation.9 Therefore, COO dissoci-ation may not only induce a more positive image butalso endogenously change the extent to which con-sumers rely on a brand’s own performance data tojudge its quality.

More broadly, we impose the following parame-ter restrictions for the purpose of the current anal-ysis: S0/M0 > Sg/Mg and M0 < Mg . Together, these

8 We started by studying a general case where COO dissociation isonly partially effective. With a certain probability, a consumer canstill tell the firm i’s COO even if it invests in COO dissociation. Theanalytical expressions become more complicated, but the intuitionand results remain the same. We choose to present the simplestcase to best illustrate the intuition.9 Note however, as long as S0 > 0 and M0 > 0, � is never exactly 0.To enable �= 0, we need to apply the Haldane prior Beta40105. Theconceptual debate about improper priors is beyond the scope ofthis paper. We do use the Haldane prior for the current analysis.

assumptions state that COO dissociation leads toa more favorable perception but also creates moreuncertainty about a firm’s COO. These assumptionsare reasonable in the context of emerging marketbrands pretending to be from developed countries.As emerging market consumers typically have a bet-ter knowledge of domestic brands than global brands,their prior belief on group g tends to be more pre-cise than their prior on group 0. Hence, Mg > M0.This is particularly true in less developed parts ofemerging markets. In the third and fourth tier cities ofChina, for example, consumers have great aspirationsfor global brands but their knowledge of a particularCOO is very limited. The other assumption, S0/M0 >Sg/Mg , simply ensures that the brands do not wantto be associated with COOs that have worse imagescompared with their home country.10 We assume thatfirm i incurs cost cb when it masks its COO. Theparameter cb may correspond to the cost of hiringa brand consulting agency, the cost of redesigningone’s website, creating an ad with a foreign accent, orany other costs associated with managing a disguisedCOO identity.

The above parameter restrictions allow us to focuson the context wherein domestic brands from emerg-ing markets pretend to be foreign brands. Technically,the model is tractable when the above inequalities arereversed. In the appendix, we replicate the analysisunder the assumption M0 > Mg . This is reminiscentof a case in which emerging market brands enter adeveloped market.

As a technical point, as Equations (2) and (3) sug-gest, since Mg and Sg endogenously depend on firmchoices Eq and Eb, the inequality M0 <Mg may flip whenmore firms break away from the stereotype. Put dif-ferently, emerging market consumers become uncer-tain about the quality image of their own COO, astoo many firms now hide their COO. We believe thisis not a very relevant case. Although COO dissocia-tion may marginally lower the precision about coun-try g’s COO image, it is quite unlikely that this willreverse the M0 <Mg inequality. In the main analysis,we assume that M0 < Mg is satisfied in equilibriumand in all relevant subgames.

As COO dissociation masks a firm’s COO identity,consumers do not incorporate this firm in the stereo-type formation process. As such, they update theprior about group g based on all of the firms who donot mask their COO identity. The updated stereotypeis determined as in Equations (2) and (3). In otherwords, we consider a simple model in which con-sumer belief about firm i’s COO is malleable. A veryimportant caveat applies to this modeling choice. Our

10 We are grateful to the review team who helped us clarify theseassumptions.

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


formulation of COO dissociation by nature assumesbounded rational consumers. If consumers are per-fectly rational, they should be able to infer the firms’tendencies to engage in COO dissociation. In this case,knowing that what appears to be a foreign brandcan in fact be a domestic one, consumers may applythe g prior (with a certain probability) to judge abrand’s quality even if the brand appears to be inthe 0 group. This may reduce firms’ incentives toinvest in COO dissociation. We believe that boundedrationality should be a reasonable assumption. This isparticularly likely to be true for emerging market con-sumers, whose sophistication is limited in the earlydevelopment stages of these markets.

Quality Investment. COO dissociation serves todisassociate a brand from the (negative) stereotype.However, it does not have any real effect on a firm’squality and does not generate more positive sig-nals Ssi. The likelihood of observing a positive sig-nal depends entirely on a firm’s quality level, qi. Theprobability that consumers observe Ssi success out ofMs signals is F 4Ssi �Ms1 qi5=

(

Ms

Ssi

)

qSsii 41−qi5

Ms−Ssi . Firm i

incurs quadratic cost cqiq2i to achieve quality qi.

Conditional on observed quality signals, we assumethat firm i’s payoff is simply its perceived qual-ity level, E4qi � S4bi51M4bi51 Ssi1Ms5, minus the costof quality investment and COO dissociation. Takingexpectation over the quality signals Ssi, Ms , we obtainfirm i’s expected payoff in the first stage

çi4qi1 bi3 Eq−i1 Eb−i5

=

∫

E4qi � S4bi51M4bi51 Ssi1Ms5 dF 4Ssi �Ms1 qi5

− cqiq2i − cbbi1 (4)

where E4qi � S4bi51M4bi51 Ssi1Ms5 is defined in Equa-tion (1), S4bi5, M4bi5, and F 4Ssi � Ms1 qi5 are defined asabove. In the first stage, firm i chooses qi and bi tomaximize the expected payoff. Finally, we introduceheterogeneity into firms’ cost of quality invest-ment, cqi. This leads to dispersion in equilibrium qual-ity levels and allows us to compare high and lowquality firms’ incentives to mask their COO. In ouranalysis, we do not impose specific functional formrestrictions on the distribution of cqi. We refer tothe inverse of each firm’s cost of quality, 1/cqi, asfirm i’s technical efficiency. As such, in this model,COO stereotype stems from the fact that firms fromthe same country have similar cost structures.

To summarize, we consider a model in which con-sumers use the group average as a prior for each indi-vidual firm’s quality. They learn about each brand’squality based on the group prior as well as fromthis particular brand’s quality signals. A firm caninvest in quality to generate more favorable quality

signals; they can also invest in COO dissociation thatmasks its COO identity. We are interested in under-standing firms’ equilibrium decisions and, in particu-lar, the implication for the COO stereotype, ä4 Eq1 Eb5=

Sg4 Eq1 Eb5/Mg4Eb5.

4. AnalysisIn this section, we attempt to present a set of propo-sitions that are as general as possible. We do notimpose further restrictions on the distribution of cqiand Beta4S01M0 − S05 in our lemmas and proposi-tions. To sharpen the intuitions, however, we con-sider a running example that we discuss after everyproposition. Specifically, we consider a case wherein1/42cqi5 is equidistantly distributed on 6�−�1�+�7.11

In addition, we assume that COO dissociation cre-ates maximal uncertainty about a firm’s country oforigin; consumers apply a diffuse prior when a firmmasks its COO.12 There are two key parameters inthis running example, i.e., �b and �.13 The parame-ter � captures the median firm’s technical efficiency.As � increases, the firms can achieve higher prod-uct quality at a lower cost. �b captures the extent towhich consumers rely on the COO stereotype to judgea firm’s quality. As explained above, �b is the base-line stereotyping tendency when no firm breaks awayfrom the stereotype.

Shared reputation may create a common goodproblem where firms do not fully internalize the ben-efits of quality investment. Put differently, it is exactlybecause of the consumers’ tendencies to stereotypethat each firm finds it profitable to free-ride on otherfirms’ efforts. This lack of incentives to improve qual-ity, in turn, leads to a negative COO stereotype.We first provide a benchmark result that formalizesthis intuition. Consider a scenario in which none ofthe firms engage in COO dissociation. Proposition 1describes firms’ equilibrium quality choices.

Proposition 1. Rank the firms according to 1/42cqi5,such that i = 1 is the most efficient firm. Consider a casewherein no firm engages in COO dissociation. The equi-librium quality choice for firm i is: q∗

i = min841 − �b +

�b/N541/42cqi55119. Firms underinvest by a larger marginwhen �b is higher and N is larger.

Proposition 1 echoes the insight from the publicgood literature, i.e., when firms enjoy a shared reputa-tion they suffer from a free-riding problem that leads

11 Under quadratic cost, the equilibrium quality choice is propor-tional to 1/cqi. This cost distribution yields equilibrium qualitychoices that have nice analytical properties.12 We use the Haldane prior S0 =M0 = 0.13 For simplicity, we fix Ms and present most of our results in termsof �b . Thus there is a one-to-one correspondence between Mgb and�b in our analysis.

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Figure 1 (Color online) Shared Reputation May Hurt Both High andLow Quality Firms

0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.00.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

1/cqi

Πi

�b = 0.6

�b = 0.2

�b = 0

to underinvestment in quality. When consumers judgethe brands based on their shared COO, a high qual-ity firm incurs positive spillover on the other firms.As such, each firm cannot fully internalize the ben-efit of quality investment, and the equilibrium qual-ity choices are lower compared with the socially effi-cient level, 1/�2cqi�.14 As expected, the extent of free-riding is most accentuated when group size is largeand when consumers have a strong tendency to relyon the COO stereotype to judge individual brands’quality.

Consumers’ tendency to engage in COO stereotyp-ing clearly hurts the high quality firms. However,the public good problem may create so much ineffi-ciency that even the lowest quality firms can becomeworse off when free-riding opportunity increases. Fig-ure 1 illustrates a case of 20 firms and 1/�2ci� isequidistantly distributed on �0�2�1�. The horizontalaxis represents 1/ci where the vertical axis representsthe firm’s equilibrium payoff, which corresponds to�i = �b�

∑

k=1��N q∗k/N �+ �1−�b�q∗i − cqiq

∗2i .

Figure 1 compares firm payoffs under three dif-ferent levels of stereotyping. The firms are indexedby their technical efficiency 1/cqi along the horizon-tal axis. Consumers put a larger weight on the COOprior when �b is larger. As can be seen, when con-sumers’ tendency to apply the COO stereotype is mild(�b = 0�2), low quality firms become better off com-pared with the no-stereotyping benchmark (�b = 0),benefiting from the possibility of free-riding. Highquality firms, as expected, become worse off. When�b = 0�6, however, the public good problem dimin-ishes firms’ quality investment incentives to such an

14 Note that since we interpret “quality” as the probability withwhich a product will generate a positive experience, q∗i is boundedbetween �0�1�. Our key insights are not driven by this assumption.

extent that both high quality and low quality firms areworse off, as measured by the difference between the�b = 0 curve and the �b = 0�6 curve.

When COO dissociation is not too costly, it rep-resents a clear benefit for high quality firms whoattempt to do so.15 When a firm succeeds in disasso-ciating itself from the group, its perceived quality isno longer weighted down by the group prior. Inter-estingly, as high quality firms dissociate themselvesfrom their COO, the impact on the COO image isambiguous. Proposition 2 considers a case in whichan exogenous number of firms, Nb, disguise theirCOO identity. It states wherein high quality firmsbreak away from the group, it may hurt or benefit thegroup stereotype.

Proposition 2. Rank the firms according to 1/cqi,such that i = 1 is the most efficient firm. Consider a casein which the top Nb firms engage in COO dissociation.Each firm chooses its quality level conditional on the COOdissociation decisions. Define q∗i �Nb� as the optimal qualitychoice of firm i when the top Nb firms break away. The equi-librium COO stereotype �� q� �b� can be re-parameterizedas �� q∗�Nb��Nb�, which has the following properties:

• �� q∗�Nb��Nb + 1�/�� q∗�Nb��Nb�≤ 1.• ∀ i > Nb + 1, q∗i �Nb + 1�/q∗i �Nb�≥ 1.

Put differently, high quality firms’ COO dissociation deci-sion has a negative direct effect and a positive strategiceffect on the group stereotype.

Proposition 2 states that high quality firms’ COOdissociation decision has two effects on the equilib-rium group stereotype. First, keeping quality choicesconstant, group stereotype deteriorates as high qual-ity firms mask their COO. When a high quality firmhides its COO identity, it ceases to contribute tothe group reputation, leading to a negative directeffect on the COO stereotype. Interestingly, however,COO dissociation has a positive strategic effect on thegroup image. The intuition is as follows: As a branddisconnects itself from the group, it prevents the otherbrands from free-riding on itself. When more highquality firms break away from the group, the remain-ing firms have stronger incentives to improve theirown qualities. As such, q∗i �Nb + 1� ≥ q∗i �Nb� for ∀ i >Nb + 1. Since �� q∗�Nb�/�q

∗i > 0, it follows that COO

dissociation has a positive strategic effect on the COOstereotype.

The overall effect of COO dissociation on equilib-rium stereotype depends on the relative magnitudesof the direct effect and the strategic effect. Broadlyspeaking, the strategic effect becomes less prominent

15 In Proposition 4, we show that high quality firms indeed havestronger incentives to invest in COO dissociation when S0/M0 >Sg/Mg and M0 <Mg .

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


when the free-riding problem is less severe. Specifi-cally, when �b is low, the strategic effect is dominatedby the direct effect. Therefore, group image deterio-rates as more firms break away from the group. When�b is high, however, the strategic effect dominates thedirect effect. The group image can in fact improve asmore high quality firms disassociate themselves fromthe group. Proposition 3 presents this comparativestatic result. To better sharpen the intuitions, Corol-lary 1 outlines the exact conditions in our runningexample.

Proposition 3. Regardless of the distribution of cqi,given N , Nb, the strategic effect is (weakly) increasingin �b; the direct effect is (weakly) decreasing in �b.

Corollary 1. Consider the case wherein 1/42cqi5 isequidistantly distributed on 6�−�1�+�7, and the qualitychoice of each firm has an interior solution. The strate-gic effect is q∗

i 4Nb + 15/q∗i 4Nb5= �b/4N −�b4Nb + 155+ 1.

The direct effect is ä4 Eq∗4Nb51Nb + 15/ä4 Eq∗4Nb51Nb5 =

1 − �/4N� − Nb�5. Overall, when �b > �/�, the COOstereotype improves as more firms break away from thestereotype.16

The intuition for Proposition 3 is as follows: Thestrategic effect refers to the fact that as more firmsbreak away from the COO stereotype, free-ridingis alleviated and quality investment increases. Thestrategic effect is increasing in �b and is null (e.g.,q∗i 4Nb + 15/q∗

i 4Nb5 = 1) when �b = 0. This is because,when �b = 0, consumers attach very little weight tothe COO prior and full weight to each firm’s indi-vidual quality signals. As such, free-riding is not aproblem from the beginning and each firm choosesthe efficient quality investment level, regardless ofhow many firms break away from their COO iden-tity. Thus, the strategic effect is minimal since qual-ity decisions do not depend on Nb. When �b is large,free-riding becomes an increasingly serious prob-lem, and the strategic effect becomes proportionallylarger. Corollary 1 presents the special case wherein1/42cqi5 is uniformly distributed. The strategic effectis q∗

i 4Nb+15/q∗i 4Nb5=�b/4N −�b4Nb+155+1, which is

clearly increasing in �b.In contrast, the direct effect simply depends on

the arithmetic difference between the cqi of a firmwho breaks away and that of the rest who staywith their COO identity. As we will show inthe appendix, when the quality investment games

16 This result applies to the case wherein interior solutions exist forthe quality investment games. Relaxing this assumption will notmake the game intractable, but the boundary condition can only bedefined implicitly. In fact we do not impose this restriction in themain analysis.

have interior solutions, the direct effect is sim-ply 44N − Nb5

∑

i=Nb+210001N 41/42cqi555/44N − Nb − 15 ·∑

i=Nb+110001N 41/42cqi555, and is invariant in �b. As Corol-lary 1 states, this ratio corresponds to 1−�/4N�−Nb�5when 1/42cqi5 is uniformly distributed. Put together,we can derive the conditions under which COOstereotype improves as more firms break away fromit. In the running example, this condition comes downto �b > �/�. The boundary is �/�, which increases asfirms have more heterogeneous costs. This is intuitivesince the direct effect is more pronounced when qual-ity variance is large.17 Figure 2 illustrates the insightfor five firms with 1/42cqi5 equidistantly distributed on6005117.

In Figure 2, we plot ä4 Eq∗4Nb51Nb5 and q∗i as func-

tions of Nb. Each dot represents the equilibrium qual-ity choice q∗

i of a corresponding firm. The solid curverepresents equilibrium stereotype ä4 Eq∗4Nb51Nb5 =

Sg4 Eq∗4Nb51Nb5/Mg4Nb5 as a function of Nb. To illustratethe idea of breaking away, the figures omit the top Nb

firms who mask their COO identities. Thus, for eachNb, the figures show q∗

i for i =Nb+11 0 0 0 1N . The leftpanel corresponds to the case �b = 0015 wherein theconsumers do not have strong tendencies to rely onthe COO stereotype. As can be seen, when firm i = 1disassociates itself from the group, the new stereo-type is based on firms i = 21 0 0 0 15 who are less effi-cient compared with firm 1. This implies a negativedirect effect. However, firms i = 21 0 0 0 15 now havestronger incentives to invest in quality, as they can nolonger free-ride on i = 1’s quality investment. Thus,q∗i increases for each of the remaining firms, as indi-

cated by the upward tilting arrows in Figure 2. How-ever, under �b = 0015, the strategic effect is not strongenough to compensate for the direct effect. Conse-quently, group stereotype deteriorates as more firmsengage in COO dissociation. When �b = 005, however,the strategic effect dominates. Collectively, the firmsimprove their quality to such an extent that COOstereotype improves although the highest quality firmsbreak away from the group.

Suppose a central planner has the option of ban-ning firms from disguising their COO. Is this a goodstrategy to follow? While we treat Nb as an exogenousparameter in Propositions 2 and 3, the main insightsalso hold in an equilibrium framework. Proposition 3and Corollary 1 imply that, somewhat counterintu-itively, banning the high quality firms from engagingin COO dissociation can in fact hurt the COO image.

17 Note also that Proposition 3 speaks to the effect of breakingaway on equilibrium stereotype, not on equilibrium profit. When�b is small, for example, the strategic effect is limited, and thestereotype decreases as more firms break away. Yet profits do notdecrease as much since consumers put relatively little weight onCOO prior when they make a quality judgment.

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Figure 2 (Color online) “Breaking Away” May Hurt or Improve COO Image

0 1 20.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

0.90

Nb Nb

q i q i

�b = 0.15, � = 0.75, � = 0.25

0 1 2

0.35

0.40

0.45

0.50

0.55

0.60

0.65

�b = 0.5, � = 0.75, � = 0.25

Firm 2

Firm 3

Firm 3

Firm 4

Firm 5

Firm 1 Firm 2

Firm 2

Firm 3

Firm 3

Firm 3Firm 2

Firm 3

Firm 1

Propositions 2 and 3 shed light on the marginalimpact as high quality brands break away from thegroup stereotype. We next endogenize both COOdissociation and quality decisions in an equilibriumframework. To best illustrate the intuition, we firstpresent two lemmas that discuss how firms’ incen-tives to engage in COO dissociation vary with the keyparameters. Specifically, we define i�qi� �q−i� �b−i� =

�i�qi�1� �q−i� �b−i� − �i�qi�0� �q−i� �b−i� as firm i’s incen-tive to engage in COO dissociation, conditional onits quality choice qi and every other firm’s decisions.Lemma 1 states the first result.18

Lemma 1. Fix Ms such that �b corresponds to Mgb one-to-one. ∀ qi >�� q� �b�, i�qi� �q−i� �b−i� is larger under larger�b. Put differently, firms have stronger incentives to investin COO dissociation when consumers have a stronger ten-dency to apply the COO stereotype.

In general, holding �q and �b−i constant, firm i ismore likely to engage in COO dissociation when con-sumers have stronger tendencies to revert to the COOstereotype to judge a brand’s quality. Consequently,high quality firms have stronger incentives to breakaway from the stereotype. This result is not very sur-prising. In a second lemma, we examine how equilib-rium quality levels �q affect firm i’s incentive to engagein COO dissociation. Does COO dissociation becomemore prevalent as firms gain greater efficiency in

18 Note that in the most general case, i cannot be uniquely deter-mined by �b . However, since �b , bounded between 0 and 1, is con-ceptually clear for presentation purposes, we present the results interms of �b whenever possible by fixing Ms , such that �b corre-sponds to Mgb one-to-one.

improving product quality? Lemma 2 states the basicinsight.

Lemma 2. Given �b� �q, ∀ qi >�� q� �b�,• i�qi� �q−i� �b−i� is increasing in qi as long as N and

N−Nb are not too small, such that M0 < Mg��N −

Nb − 1�/�N −Nb��N �1−�b�/�N �1 − �b� + �b��. Other-wise, i�qi� �q−i� �b−i� is decreasing in qi.

• i�qi� �q−i� �b−i� is decreasing in qk, k �= i.As such, when cqi decreases for every i, the equilibrium

number of firms who engage in COO dissociation mayincrease or decrease.

Lemma 2 states that, as expected, a firm’s incen-tive to break away from the stereotype is increasingin its own quality as long as N −Nb is not too small.Note that we assumed M0 <Mg in the baseline modeland that �b ≤ 1 by construction. Thus the conditionin Lemma 2 is satisfied as long as ��N − Nb − 1�/�N − Nb��N �1 − �b�/�N �1 − �b�� + �b� is close to 1,which is the case for large enough N −Nb.

However, as other firms improve their qualities,firm i’s incentive to engage in COO dissociationdecreases. The intuition is as follows: As COO disso-ciation separates a firm from the rest of the group,its benefit is largest when most other firms fromthe same country have low quality. As firms k �= iattain higher quality, the COO stereotype improvesand the need for breaking away is diminished. Thisinsight has interesting implications for the prevalenceof COO dissociation when firms collectively makeprogress in technical efficiency.

To illustrate the idea, consider our running examplewherein 1/�2cqi� is equidistantly distributed on ��−�� + ��. When � is fixed, the parameter �, i.e., the

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Figure 3 (Color online) Prevalence of COO Dissociation as FirmsImprove in Quality

0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.80

1

2

3

4

5

6

7

8N = 20, � = 0.15, cb = 0.08

Nb

�b = 0.5

�b = 0.45

�

inverse of the median 2cqi, measures the firms’ capa-bility to produce high quality products. As emergingmarket firms collectively catch up in technological ormanagerial skills, cqi will decrease across the board.This collective progress can be approximated with anincrease in �, while holding cb as a constant. Figure 3illustrates the impact of higher � on the prevalenceof COO dissociation. It depicts equilibrium N ∗

b as afunction of �. As � increases, firms collectively gainhigher technical efficiency. Interestingly, the equilib-rium number of firms who engage in COO dissocia-tion follows an inverted-U shaped curve. When � isvery low, most firms are only capable of producinglow quality products. Their profits are low regardlessof whether they break away from the stereotype, mak-ing costly COO dissociation an unprofitable propo-sition. As � improves, more firms become capableof achieving high quality and can afford the cost ofbreaking away from the group stereotype. This effectenables more firms to engage in COO dissociation.However, when � further increases, the stereotypeeventually improves to such a point that the gainfrom breaking away starts to decrease. This pointsto an interesting, nonmonotonic relationship betweenfirms’ collective technical efficiency (�) and the preva-lence of COO dissociation (N ∗

b ). As expected, N ∗

b issmaller at every � for smaller �b, reflecting the insightfrom Lemma 1.

Note that the exact relationship between N ∗

b and cqi’sdepends on the distributional assumptions about cqi.We replicated the analysis under a number of othercost distributions as well by relaxing the assumptionthat q is bounded on �0�1�. While N ∗

b does not alwaysfollow an inverted-U shape curve, it is a rather robustresult that N ∗

b can increase or decrease as cqi’s decrease.Lemmas 1 and 2 shed some light on the strate-

gic rationale behind COO dissociation. Together withPropositions 2 and 3, they suggest an interesting inter-action between COO dissociation and quality invest-ment. Building on these insights, we next investigatethe equilibrium outcome in the first stage, when firmschoose qi and bi jointly. Proposition 4 states the result.

Figure 4 (Color online) Equilibrium Quality Choice and COODissociation

0.1 0.5 0.90.15

1.00

2.00

0.150.15

0.15

0.3

0.3

0.3

0.5

0.50.5

0.65

0.650.6

5

�b

N = 20, � = 0.15, cb = 0.08

BA

�Proposition 4. Rank the firms according to 1/cqi,

such that i= 1 has the lowest cost. Define �b as before and�′

b = M0/�Ms +M0�. In a pure strategy equilibrium, thetop N ∗

b firms invest in COO dissociation: b∗i = 1 iff i≤N ∗

b .The equilibrium quality choices are: q∗i = min�1� �1 −

�N −Nb − 1��b/�N − �bNb��1/�2cqi�� for i > N ∗

b ; q∗i =min�1� �1 − �′

b��1/�2cqi�� for i ≤ N ∗

b . An implicit defini-tion of N ∗

b is given in the appendix.

Proposition 4 states that, regardless of the distributionof cqi, firms’ equilibrium COO dissociation decisionscan be characterized by a threshold N ∗

b . The N ∗

b mostefficient firms will break away from the group. Theother firms do not disguise their COO. The firms whobreak away from the stereotype have greater incen-tives to invest in quality. While this is a general result,the exact conditions for N ∗

b depend on the distributionof cqi. To clarify the intuitions, let us consider our run-ning example in which Beta�S0�M0 − S0� correspondsto the flat prior, and 1/�2cqi�’s are equidistantly dis-tributed on ��−��+��. Figure 4 describes the equi-librium outcomes in the parameter space ��b��; � isfixed to 0.15.

Figure 4 depicts N ∗

b with varying shades on agrayscale. A darker area corresponds to equilibriawith higher N ∗

b , i.e., COO dissociation is more preva-lent. The lighter area represents equilibria in whichfewer firms mask their COO identity.19 A set of “iso-stereotype” curves are superimposed on the graph,which we explain in the next paragraph. A fewinsights are evident from Figure 4. First, note that theequilibrium pattern of N ∗

b is consistent with the intu-ition laid out in Lemmas 1 and 2. Consistent with

19 We discretized the range of N ∗

b into a five-level grayscale.

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


the intuition in Lemma 1, COO dissociation is mostprevalent when consumers have strong tendencies tostereotype (i.e., �b is high). As �b increases, N ∗

b mono-tonically increases. Consistent with Lemma 2, as firmscollectively make progress in quality (i.e., � is higher),the prevalence of COO dissociation may first increasethen decrease. In Figure 4, for any given �b, N ∗

b fol-lows an inverted-U shaped relationship in �. Overall,COO dissociation is most prevalent for large �b andintermediate �. None of the firms practice COO dis-sociation (N ∗

b = 0) for sufficiently small �b, and eithersufficiently small or large �.

A second insight concerns the equilibrium COOimage, ä∗4 Eq1 Eb5, which corresponds to the mean of theCOO prior.20 Figure 4 presents four “iso-stereotype”curves in the parameter space. Each iso-stereotypecurve corresponds to a set of 4�b1�5 values underwhich the equilibrium COO image, ä∗4 Eq1 Eb5= Sg/Mg ,are identical. The labels on top of each iso-stereotypecurve denote the value of ä∗4 Eq1 Eb5 under that set ofparameters.

Examining these iso-stereotype curves, several in-teresting insights emerge. First, broadly speaking,equilibrium COO image is more favorable for lower�b or higher �.21 Put differently, the country’s qualityimage improves as firms gain greater technical effi-ciency (� ↑) or as consumers rely less on the COOstereotype to judge the brands (�b ↓). Second, an iso-stereotype curve typically spans different equilibriawith varying levels of N ∗

b . For example, consider theiso-stereotype curve with S∗

g/M∗g = 003. At point A,

4�b1�5 ≈ 40049100535, four firms engage in COO dis-sociation and the remaining firms achieve an aver-age quality of S∗

g/M∗g = 003. At point B, 4�b1�5 ≈

40058100585, seven firms engage in COO dissociation,and the firms collectively achieve a COO image ofS∗g/M

∗g = 003 as well. Relatively speaking, point B rep-

resents a case in which firms’ collective technical effi-ciency is high, but consumers have strong tendenciesto stereotype. As a consequence, a larger number ofhigh quality firms break away from the COO iden-tity. In comparison, point A represents a case whereinfirms’ technical efficiency is slightly lower, but con-sumers rely less on the COO stereotype to judge eachbrand’s quality. In equilibrium, firms do not havestrong incentives to disguise their COO and they

20 COO stereotype is the main focus of this paper, but we can alsoinvestigate equilibrium firm surplus, by calculating the total pay-offs accrued to all of the firms. This is also a meaningful measuresince it takes into account both the firms who break away from thegroup and the firms who remain in the group. It turns out thatin our example, the “iso-welfare” curves follow largely the samepattern. Firm welfare is increasing in � and decreasing in �b .21 Some anomalies may occur since “breaking free” is a discretedecision. This is technically possible but conceptually not veryinteresting.

achieve the same level of average quality. In general,the iso-stereotype curves tend to be upward sloping,as depicted in Figure 4.

5. DiscussionThis paper studies a curious phenomenon that iswidely observed in many emerging markets. Manyemerging market consumers hold pronounced COOpreferences for western and global brands. In re-sponse, local brands from emerging markets pursue avariety of measures to pretend to be foreign. We studythe strategic implications of this phenomenon, focus-ing on the antecedents and consequences of breakingfree of a stereotype.

Our analysis illustrates the public good problemwhen consumers apply the COO stereotype to judgethe firms’ quality. Firms cannot fully internalize thebenefit of quality investment, which leads to insuf-ficient product qualities that fulfill a negative COOstereotype. As such, high quality firms have incen-tives to dissociate from their COO identity. Interest-ingly, as high quality firms break away, the free-ridingproblem is alleviated, leading to a higher qualityinvestment by the rest of the firms. Overall, high qual-ity firms’ COO dissociation may hurt or benefit theequilibrium COO image. We also show that as firmscollectively make progress in quality, the prevalenceof COO dissociation can first increase then decrease.

We highlight one potential direction for futureresearch. The current analysis does not address thescenario in which firms can endogenously investin Msi and generate more quality signals on theirown. Recall that in our formulation of the COOeffect, the group stereotype affects the individualbrands by serving as a prior. When consumers donot make inference from Msi, roughly put, the weightshe puts on the prior is determined by the followingparameter:

�i =Mg

Mg +Msi

0

A high quality firm certainly has incentives todecrease �i such that it is less affected by the groupstereotype. This can be achieved by a decrease in Mg

or an increase in Msi. In this paper, we focused onthe possibility that firm i can decrease Mg by disguis-ing its COO, for example, by choosing a brand namethat does not evoke a particular COO association.The other approach is to increase Msi, which will alsodecrease �i. Put differently, a firm can break free of astereotype by building its brand. A high quality firm ican, for example, invest in marketing campaigns thatfocus on the uniqueness of its individual brand. Assuch, consumers realize that firm i is different fromother brands from country g.22

22 We thank one anonymous reviewer for this suggestion.

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


We believe that this represents a very differentbut equally interesting research question. While COOdissociation is a fairly unique and interesting phe-nomenon in many emerging markets, brand build-ing definitely has more general appeal. We discusssome tentative conjectures on what would happen ina model with endogenous Msi. First, when consumersdo not make inferences based on Msi, it is easy to seethat a firm’s incentive to increase Msi is quite similarto its incentive to decrease Mg , since both actions areequivalent in the sense that they decrease �i, which isthe only parameter that affects the profit function.

However, when consumers are fully Bayesian, as afirm increases Msi, its impact on the posterior beliefis quite different. When firms are heterogeneous interms of Msi, Bayesian consumers should infer theproduct qualities based on three pieces of informa-tion, i.e., the quality signals each firm generates,the firm’s country of origin identity, and the firm’sbrand awareness Msi. Put differently, the consumersno longer apply a single, overarching stereotype to allfirms from group g. When we allow two levels of Msi

in the model, for instance, there should exist a sepa-rating equilibrium in which only firms whose qualityis above a certain level invest in their brand aware-ness. Consumers may apply two different stereotypesto brands with a high level of awareness and brandswith a low level of awareness. Free-riding occurswithin each subgroup, i.e., among the firms who havethe same Msi, respectively. A formal analysis of thismodel should be left to future analysis.

The duality of brands and COO stereotypes andthe important question of brand building has gen-eral appeal for firms from developed markets andemerging markets alike. For example, for Italian shoesand French wines a positive stereotype is establishedamong the firms. In this case, the firms conceivablyhave no incentives to disguise their COO, but brandbuilding is still attractive especially for the firms whoenjoy the highest quality. We believe that furtheranalyses along these lines can be a very interestingresearch direction.

AcknowledgmentsThe author is grateful to Jiwoong Shin, K. Sudhir, YuxinChen, Liang Guo, Durairaj Maheswaran, Zsolt Katona,Miklos Sarvary, participants at the Marketing ScienceEmerging Markets conference at Wharton, seminar partic-ipants at Cheung Kong Graduate School of Business, Uni-versity of British Columbia, London Business School, HongKong University of Science and Technology, and FudanUniversity for their helpful comments. This research is sup-ported by research funding from the Cheung Kong Gradu-ate School of Business.

Appendix A. M0 >Mg

In this section, we reverse a central assumption in thebaseline model that, M0 < Mg . Recall that Mg stands for

the precision of the COO stereotype of the focal country,while M0 stands for the precision of the alternative COOprior. We believe that this is a reasonable assumption sinceemerging market consumers conceivably have more preciseknowledge about the local firms. Their COO knowledgeabout foreign firms are more positive but less precise.

In this section, we explore what would happen whenM0 >Mg . This may correspond to a case in which emerg-ing market firms enter a developed market.23 As we will showshortly, reversing the assumption M0 < Mg changes someof the key insights in the baseline model while preserv-ing the others. We will put the differences in perspective.Throughout the analysis, we keep the other model elementsunchanged. As is clear, when firms do not invest in COOdissociation, the equilibrium quality choices are identical tothat in the baseline model. Lemma A.1 discusses the firms’incentives to invest in COO dissociation.

Lemma A.1. When M0 >Mg and S0/M0 > Sg4 Eq1 Eb5/Mg4Eb5,the firms with higher cqi have stronger incentives to invest inCOO dissociation.

Lemma A.1 states that, as expected, when M0 >Mg , thehigh cost (i.e., low quality) firms have stronger incentivesto engage in COO dissociation. This happens because COOdissociation not only increases the mean of the prior butalso lowers the weight consumers put on firm i’s individualquality signals. While high quality and low quality firmsequally benefit from a higher prior mean, it is the low qual-ity firms who benefit the most from a lower weight on thequality signals. Moreover, a firm lowers its quality invest-ment when it disguises its COO. Therefore, given any distri-bution of cqi, an equilibrium of the game is characterized byNb such that the Nb least efficient firms engage in COO dis-sociation. Proposition A.1 echoes Proposition 2 in the base-line model, and states the impact of COO dissociation onequilibrium stereotype.

Proposition A.1. Rank the firms according to 1/cqi, suchthat i = 1 is the most efficient firm. Consider a case in whichthe bottom Nb firms engage in COO dissociation. The equi-librium COO stereotype ä4 Eq∗1 Eb∗5 can be re-parameterized asä4 Eq∗4Nb51Nb5, which has the following properties:

• ä4 Eq∗4Nb51Nb + 15/ä4 Eq∗4Nb51Nb5≥ 1.• ∀ i < N −Nb − 1, q∗

i 4Nb + 15/q∗i 4Nb5≥ 1.

Put differently, low quality firms’ COO dissociation effort has apositive direct effect and a positive strategic effect on the groupstereotype.

Proposition A.1 examines the impact of COO dissociationon equilibrium stereotype by comparing subgames with dif-fering levels of Nb . As expected, when low quality firmsare the first to break away from the stereotype, they incura positive direct effect on the group stereotype. Moreover,the strategic effect remains positive. This is because theextent of free-riding depends only on consumers’ tendencyto stereotype and the number of firms who share the groupimage. As such, when low quality firms break away fromthe group, high quality firms also gain additional incentivesto improve their qualities. Figure A.1 mirrors Figure 2 in the

23 We thank the review team for making this suggestion.

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Figure A.1 (Color online) When M0 >Mg

0 0.5 1.0 1.5 2.00.35

0.40

0.45

0.50

0.55

0.60

0.65

0.70

0.75

� = 0.5, � = 1.5, � = 0.5

0 0.5 1.0 1.5 2.00.50

0.55

0.60

0.65

0.70

0.75

0.80

0.85

0.90

0.95

Nb Nb

q i q i

� = 0.15, � = 1.5, � = 0.5

Firm 1

Firm 2

Firm 5

Firm 4

Firm 1

Firm 4

Firm 3

Firm 4

Firm 4

Firm 5

Firm 3

baseline model. Overall, COO stereotype always improvesas low quality firms break away from the stereotype. Underhigher �b , the increase (in terms of ratio) is faster due to astronger strategic effect.

More broadly, Propositions 2 and A.1 describe the scenar-ios in which high or low quality firms are the first to breakaway from the stereotype. In reality, many factors other thanM0 and Mg may determine whether the high or low qual-ity firms have stronger incentives to disguise their COO.For example, high quality firms may also have more mar-keting resources (e.g., better foreign language proficiency)to disguise their COO, for example, by choosing a foreignsounding name. Low quality firms may lack the knowledgeto do so. In this case, the high quality firms will be the firstto break away from the stereotype even if M0 > Mg . Theequilibrium of such a game will resemble what is describedin Proposition 2.

Appendix B. Proofs for Propositions

Connection Between � and �b . To help clarify the proof,we first explain the connections between these two param-eters. � measures the consumers’ tendency to stereotypewhen judging a firm’s quality. It is defined as Mg/�Mg+Ms�.Since Mg is in fact a function of �b, Mg = �N −

∑

i bi��Mgb/N�,� is determined endogenously in the game.

�b is an exogenous parameter in the model, defined as�b =Mgb/�Mgb +Ms�. Thus, we can derive � as a function of�b . Specifically, note that Mgb = �bMs/�1 − �b�. Thus, Mg =

�N −∑

i bi��bMs/��1−�b�N ��= �N −Nb��bMs/��1−�b�N ��.Plug this into � = Mg/�Mg + Ms�, we obtain that � =

�N −Nb��b/�N −Nb�b�.Clearly, when more firms in group g break away, Mg

decreases accordingly; so does �. While �b is a good mea-sure of consumers’ tendency to rely on the COO stereotype,the exact weight they put on the prior depends on the num-ber of firms who disassociate from the COO, Nb .

Proof of Proposition 1. When no firm invests in COOdissociation, we have �= �b . The equilibrium stereotype is∑N

i=1 qi/N . When each firm offers a Bernoulli product withsuccess probability qi, the expected payoff to firm i is

�i�qi�0� �q−i� �0�= �1−�b�qi +�b

∑Ni=1 qiN

− cqiq2i � (B1)

Since ��−cqiq2i �/�qi��qi=0 = 0, the optimal quality level

is greater than zero. The first order condition ��i�qi�0��q−i� �0�/�qi = 0 is met when qi = �1 − �b + �b/N��1/�2cqi��.Since qi ∈ �0�1� for a Bernoulli product, equilibrium qualitylevel is q∗i = min��1−�b+�b/N��1/�2cqi��1�. We have q∗i = 1when cqi is sufficiently low. �

Proof of Proposition 2. In the proof for Proposition 4,we show that given any �q and �b, high quality firms indeedhave stronger incentives to invest in COO dissociation. Herewe consider a subgame in which the top Nb firms breakaway from the stereotype. The firms that do not break awayfrom the stereotype optimize the following profit function:

�i�qi�0� �q−i�Nb�= �1−��qi +�

∑Ni=Nb+1 qi

N −Nb

− cqiq2i � (B2)

where �= �N −Nb��b/�N −Nb�b�. The optimal quality levelis therefore q∗i �Nb� = min�1� ��N − �b�N − 1��/�N − Nb�b�� ·�1/�2cqi��. Clearly for all i, q∗i �Nb� is (weakly) increasing inNb , proving the second part of Proposition 2.

To prove the first part of Proposition 2, note that�� q∗�Nb��N

′b� =

∑NN ′b+1 q

∗i �Nb�/�N − N ′

b�. In addition, q∗i �Nb�

is decreasing in i. Therefore,∑N

Nb+1 q∗i �Nb�/�N − Nb� ≥

∑NNb+2 q

∗i �Nb�/�N − Nb − 1�, with the inequality being strict

when interior solution exists in the quality investmentgames at least for some firms. This proves the first part ofProposition 2. �

Proof of Proposition 3. From the proof for Proposi-tion 2, it is clear that q∗i �Nb + 1�/q∗i �Nb� = 1 + �b/�N −

�Nb + 1��b� when q∗i �Nb + 1� < 1 and q∗i �Nb� < 1. This ratio

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


is increasing in �b . Therefore, the strategic effect is strictlyincreasing in �b when interior solutions exist for the qual-ity investment game. The same holds true when we con-sider corner solutions. When q∗

i 4Nb + 15 = 1 and q∗i 4Nb5 < 1,

q∗i 4Nb +15/q∗

i 4Nb5= 1/q∗i 4Nb5, which also increases in �b since

q∗i 4Nb5 is a decreasing function of �b . When q∗

i 4Nb + 15 =

q∗i 4Nb5= 1, the ratio is a constant.

From Proposition 2, the direct effect is ä4 Eq∗4Nb51Nb + 15/ä4 Eq∗4Nb51Nb5 = 44N − Nb5

∑NNb+2 q

∗i 4Nb55/44N − Nb − 15 ·

∑NNb+1 q

∗i 4Nb55 < 1. When for all j , q∗

j 4Nb5= 44N −�b4N − 155/4N −Nb�b5541/42cqi55, the direct effect is

ä4 Eq∗4Nb51Nb + 15

ä4 Eq∗4Nb51Nb5=

4N −Nb5∑

i=Nb+210001N 41/42cqi55

4N −Nb − 15∑

i=Nb+110001N 41/42cqi551

since the term 4N − �b4N − 155/4N − Nb�b5 cancels out.Hence, the direct effect is invariant in �b when interior solu-tions exist for the quality investment game. When thereexists k such that q∗

j 4Nb5 = 1 for j ≤ k but ∃ i, q∗i 4Nb5 < 1,

similar algebra shows that ä is decreasing in �b . When ∀ i,q∗i 4Nb5= 1, the direct effect is invariant in �b . �

Proof of Corollary 1. Corollary 1 follows directly fromProposition 3. When 1/42cqi5 is uniformly distributed on6� − �1� + �7,

∑

i=Nb+110001N 41/42cqi55/4N − Nb5 = � − � +

4N −Nb5�/N . It follows that ä4 Eq∗4Nb51Nb +15/ä4 Eq∗4Nb51Nb5= 1 − �/4N�−Nb�5.

To get the boundary condition �b > �/�, let us exam-ine the conditions under which ä4 Eq∗4Nb + 151Nb + 15/ä4 Eq∗4Nb51Nb5 > 1. When all firms choose interior qualitylevels, it turns out that

ä4 Eq∗4Nb + 151Nb + 15

ä4 Eq∗4Nb51Nb5=

ä4 Eq∗4Nb51Nb + 15

ä4 Eq∗4Nb51Nb5·q∗i 4Nb + 15q∗i 4Nb5

1

where i can take any value. Solving this inequality, weobtain the boundary condition �b > �/�. Note that theboundary condition is independent of Nb . Thus, the equi-librium COO stereotype is always a monotone functionof Nb . �

Proof of Lemmas 1 and 2. Consider the functionçi4qi113 Eq−i1 Eb−i5 − çi4qi103 Eq−i1 Eb−i5. From Equation (4), wehave

âi4qi1 Eq−i1 Eb−i5 = çi4qi113 Eq−i1 Eb−i5−çi4qi103 Eq−i1 Eb−i5

=

(

Ms

M0 +Ms

−Ms

Ms +Mg

)

qi

+

(

S0

M0 +Ms

−Sg4 Eq1 Eb5

Ms +Mg

)

− cb0 (B3)

Both lemmas follow naturally from Equation (B3). First,note that given Eq, Eb, ¡âi4qi1 Eq−i1 Eb−i5/¡Mg > 0. Therefore,âi4qi1 Eq−i1 Eb−i5 is increasing in �b when Ms is fixed, provingLemma 1. Next, note that Sg4 Eq1 Eb5 is (weakly) increasing inqk, k 6= i. Thus âi is decreasing in qk. Meanwhile,

¡âi4qi1 Eq−i1 Eb−i5

¡qi=

Ms

M0 +Ms

−Ms

Ms +Mg

−Mg

4N −Nb54Ms +Mg50

It can be shown that ¡âi4qi1 Eq−i1 Eb−i5/¡qi > 0 when the con-dition in Lemma 2 is satisfied. �

Proof of Proposition 4. The proof of Proposition 4 isdivided into two parts. First, we show that the equilibriumpattern can indeed be described by a threshold N ∗

b , suchthat all firms, ranked by 1/cqi, i ≤N ∗

b choose bi = 1 while allfirms i > N ∗

b choose bi = 0. Second, we prove that the equi-librium quality choices q∗

i are as described in Proposition 4and provide an implicit definition of N ∗

b .From Equation (B3), we obtain that for all j , k, qj > qk, âj >

âk when M0 < Mg , since Ms/4M0 + Ms5 − Ms/4Ms + Mg5 >0. Therefore, high quality firms have stronger incentives toengage in COO dissociation. In any subgame in which bj = 0but bk = 1, we have âk ≥ 0 but âj ≤ 0, a contradiction. Thesubgames in which the top N ∗

b firms, in terms of quality q∗i ,

break away from the stereotypes indeed correspond to anequilibrium pattern.

Next, we show that the equilibrium ranking of q∗i is

indeed consistent with the equilibrium ranking of 1/cqi. Putdifferently, if a firm’s cost is lower, its equilibrium qual-ity choice must be higher. Suppose there exist j1 k cqj >cqk and q∗

j > q∗k . If bk = bj , the quality choices cannot be

simultaneously optimal since the two first order condi-tions cannot be satisfied simultaneously (corner solutionscannot hold either). Thus bj = 1 and bk = 0. No devia-tion conditions imply çk4q

∗k103 Eq−k1 Eb−k5 >çk4q

∗j 113 Eq−k1 Eb−k5

and çj4q∗k103 Eq−j1 Eb−j5 < çj4q

∗j 113 Eq−j1 Eb−j5. From these two

inequalities it is implied that cqk4q∗2j − q∗2

k 5 > cqj4q∗2j − q∗2

k 5.A contradiction is found.

Thus, in equilibrium, the more efficient firms alwayschoose higher quality; the higher quality firms always havestronger incentives to invest in COO dissociation. The equi-librium can thus be described with the parameter N ∗

b . Theequilibrium quality levels can be determined in a simi-lar fashion to Propositions 1 and 2. Next, we provide animplicit definition for N ∗

b .For all Nb , define Eb∗

−i4Nb5 and Eq∗−i4Nb5 as in the propo-

sitions. Rewrite âN ′ 4q′N ′1 Eq−N ′1 Eb−N ′ 5 as â4N ′1Nb5. Define

ê4N ′1Nb5 as follows:

ê4N ′1Nb5 = çN ′ 4q′′

N ′113 Eq∗−N ′ 4Nb51 Eb∗

−N ′ 4Nb55

−çN ′ 4q′

N ′103 Eq∗−N ′ 4Nb51 Eb∗

−N ′ 4Nb551 (B4)

where q′N ′ =min81144N −�b4N −155/4N −Nb�b5541/42cqN ′ 559

and q′′N ′ =min81141−�′

b541/42cqN ′ 559. It can be shown thatq′′N ′ is the optimal quality choice of firm N ′ when bN ′ =1.

From the insights above, ê4N ′1Nb5 is a decreasing func-tion in N ′. In words, Equation (B4) describes the payoffwhen a firm who chooses q′

i1bi =0 deviates by simul-taneously changing q and b, into q′′

i 1bi =1. In a purestrategy equilibrium, the number of firms who engage inCOO dissociation is given by N ∗

b , such that (1) ê4N ∗b 1N

∗b 5

>0 and (2) max8ê4N ∗b +11N ∗

b 51â4N∗b +11N ∗

b 59<0. Whenê4N −11N −15>0, all but one firm break away from thestereotype. When max8ê411051â411059<0, none of thefirms break away from the stereotype. Since ê and â aredecreasing functions, the equilibrium is the unique purestrategy equilibrium when it exists. �

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.


Proof of Lemma A.1 and Proposition A.1. It is easy tosee from Equation (B3) that when M0 >Mg , Ms/4M0 +Ms5− Ms/4Ms + Mg5 < 0 and that high quality firms havesmaller incentives to invest in COO dissociation. Further-more, given that Nb firms break away from the stereotype,the expression for q∗

i is identical to that in the baselinemodel. Thus, the strategic effect is the same as in Proposi-tion 3 and Corollary 1.

The direct effect is now

ä4 Eq∗4Nb51Nb + 15

ä4 Eq∗4Nb51Nb5=

4N −Nb5∑

i=110001N−Nb−1 41/42cqi55

4N −Nb − 15∑

i=110001N−Nb41/42cqi55

> 1

when interior solutions exist for the quality investmentgame. By similar logic as in the proof for Proposition 3,ä4 Eq∗4Nb51Nb +15/ä4 Eq∗4Nb51Nb5 > 1 remains true when somefirms obtain q∗

i = 1. �

ReferencesAnand BN, Shachar R (2004) Brands as beacons: A new source

of loyalty to multiproduct firms. J. Marketing Res. 41(2):135–150.

Anderson Analytics (2007) Brands and countries. Stamford, CT.http://www.andersonanalytics.com/reports/BrandAndCountries.pdf.

Arrow K (1973) The theory of discrimination. Ashenfelter O,Rees A, eds. Discrimination in Labor Markets (Princeton Univer-sity Press, Princeton, NJ).

Bilkey WJ, Nes E (1982) Country-of-origin effects on product eval-uations. J. Internat. Bus. Stud. 13(1):89–100.

Cabral L (2009) Umbrella branding with imperfect observabilityand moral hazard. Internat. J. Indust. Organ. 27(2):206–213.

Credit Suisse (2014) Emerging consumer survey 2014. ResearchInstitute, Zurich. https://publications.credit-suisse.com/tasks/render/file/?fileID=E78BA562-9486-225A-8E2196F57A74B6A4.

Dinnie K (2004) Country-of-origin 1965–2004: A literature review.J. Customer Behav. 3(2):165–213.

Erdem T (1998) An empirical analysis of umbrella branding. J. Mar-keting Res. 35(3):339–351.

Erdem T, Sun B (2002) An empirical investigation of the spillovereffects of advertising and sales promotions in umbrella brand-ing. J. Marketing Res. 39(4):408–420.

Fang H, Moro A (2010) Theories of statistical discrimination andaffirmative action: A survey. Technical report, National Bureauof Economic Research, Cambridge, MA.

Hakenes H, Peitz M (2008) Umbrella branding and the provisionof quality. Internat. J. Indust. Organ. 26(2):546–556.

Han CM (1989) Country image: Halo or summary construct. J. Mar-keting Res. 26(2):222–229.

Hong S-T, Kang DK (2006) Country-of-origin influences on prod-uct evaluations: The impact of animosity and perceptions ofindustriousness brutality on judgments of typical and atypicalproducts. J. Consumer Psych. 16(3):232–239.

John DR, Loken B, Joiner C (1998) The negative impact of exten-sions: Can flagship products be diluted? J. Marketing 62(1):19–32.

Keller KL (1993) Conceptualizing, measuring, and managingcustomer-based brand equity. J. Marketing 57:1–22.

Leclerc F, Schmitt BH, Dubé L (1994) Foreign branding and itseffects on product perceptions and attitudes. J. Marketing Res.31(2):263–270.

Levin J (2009) The dynamics of collective reputation. BE J. Theoret.Econom. 9(1):1–23.

Liefeld JP (2004) Consumer knowledge and use of country-of-origininformation at the point of purchase. J. Consumer Behav. 4(2):85–87.

Loken B, John DR (1993) Diluting brand beliefs: When do brandextensions have a negative impact? J. Marketing 57:71–84.

Lundberg SJ, Startz R (1983) Private discrimination and socialintervention in competitive labor market. Amer. Econom. Rev.73(3):340–347.

Maheswaran D (1994) Country of origin as a stereotype: Effects ofconsumer expertise and attribute strength on product evalua-tions. J. Consumer Res. 21(2):354–365.

Maheswaran D, Chen CY (2006) Nation equity: Incidental emo-tions in country-of-origin effects. J. Consumer Res. 33(3):370–376.

Maheswaran D, Chen CY (2009) Nation equity: Country-of-origineffects and globalization. Kotabe M, Kelsen K, eds. Handbook ofInternational Marketing (Sage, Thousand Oaks, CA), 91–113.

McKinsey and Company (2012a) From mass to mainstream: Keep-ing pace with China’s rapidly changing consumers. https://www.mckinsey.com/~/media/McKinseyOffices/Korea/PDFs/Chinese_Consumer_Report.ashx.

McKinsey and Company (2012b) Winning the $30 trillion decathlon:How to succeed in emerging markets. http://www.mckinsey.com/insights/strategy/winning_the_30_trillion_decathlon_going_for_gold_in_emerging_markets.

Montgomery CA, Wernerfelt B (1992) Risk reduction and umbrellabranding. J. Bus. 65(1):31–50.

Phelps ES (1972) The statistical theory of racism and sexism. Amer.Econom. Rev. 62(4):659–661.

Roth MS, Romeo JB (1992) Matching product category and coun-try image perceptions: A framework for managing country-of-origin effects. J. Internat. Bus. Stud. 23(3):477–497.

Samiee S, Shimp TA, Sharma S (2005) Brand origin recognitionaccuracy: Its antecedents and consumers’ cognitive limitations.J. Internat. Bus. Stud. 36(4):379–397.

Simonin BL, Ruth JA (1998) Is a company known by the companyit keeps? Assessing the spillover effects of brand alliances onconsumer brand attitudes. J. Marketing Res. 35(1):30–42.

Tadelis S (1999) What’s in a name? Reputation as a tradeable asset.Amer. Econom. Rev. 89(3):548–563.

The Economic Times (2012) Indian brands with foreign labels like“Munich polo” make name for themselves. http://articles.economictimes.indiatimes.com/2012-07-27/news/32889547_1_indian-brands-foreign-brand-da-milano.

Tirole J (1996) A theory of collective reputations (with applica-tions to the persistence of corruption and to firm quality). Rev.Econom. Stud. 63(1):1–22.

Wernerfelt B (1988) Umbrella branding as a signal of new productquality: An example of signalling by posting a bond. RAND J.Econom. 19(3):458–466.

Winfree JA, McCluskey JJ (2005) Collective reputation and quality.Amer. J. Agricultural Econom. 87(1):206–213.

World Brand Lab (2013) Asian brand loyalty survey. New York.www.worldbrandlab.com.

Dow

nloa

ded

from

info

rms.

org

by [

130.

133.

8.11

4] o

n 09

May

201

5, a

t 22:

32 .

For

pers

onal

use

onl

y, a

ll ri

ghts

res

erve

d.

Breaking Free of a Stereotype: Should a Domestic Brand Pretend …cdn.persiangig.com/preview/ZwyMhAN7xN/zhang2015.pdf · 2019. 4. 12. · of brand origin is largely driven by consumers

Documents