mksc%2E2014%2E0867

This article was downloaded by: [142.103.180.205] On: 19 June 2015, At: 13:59Publisher: 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

Untangling Searchable and Experiential Quality Responsesto CounterfeitsYi Qian, Qiang Gong, Yuxin Chen

To cite this article:Yi Qian, Qiang Gong, Yuxin Chen (2014) Untangling Searchable and Experiential Quality Responses to Counterfeits. MarketingScience

Published online in Articles in Advance 22 Sep 2014

. http://dx.doi.org/10.1287/mksc.2014.0867

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 articles 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 2014, INFORMS

Please scroll down for articleit 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

Articles in Advance, pp. 117ISSN 0732-2399 (print) ISSN 1526-548X (online) http://dx.doi.org/10.1287/mksc.2014.0867

2014 INFORMS

Untangling Searchable and Experiential QualityResponses to Counterfeits

Yi QianSauder School of Business, University of British Columbia, Vancouver, British Columbia V6T 1Z2 Canada; and

National Bureau of Economic Research, Cambridge, Massachusetts 02138, [email protected]

Qiang GongSouthwestern University of Finance and Economics, Chengdu, Sichuan, China, [email protected]

Yuxin ChenNYU Shanghai, Pudong, Shanghai, China, [email protected]

In this paper, we untangle the searchable and experiential dimensions of quality responses to entry bycounterfeiters in emerging markets with weak intellectual property rights. Our theoretical framework analyzesmarket equilibria under competition from counterfeiting as well as under monopoly branding. A key theoreticalprediction is that emerging markets can be self-corrective with respect to counterfeiting issues in the followingsense: First, counterfeiters can earn positive profits by pooling with authentic brands only when consumers havegood faith in the market (i.e., they believe there is low probability that any product is a counterfeit). Whenthe proportion of counterfeits in the market exceeds a cutoff value, brands invest in self-differentiation fromthe competitive-fringe counterfeiters. Second, to attain a separating equilibrium with counterfeiters, brandedincumbents upgrade the searchable quality (e.g., appearance) of their products more and improve the experientialquality (e.g., functionality) less compared with monopoly equilibrium. However, in the pooling equilibrium withsporadic counterfeits, authentic firms instead may invest in experiential quality to attract more of the expertconsumers who are well versed in quality. This prediction uncovers the nature of product differentiation in thesearchable dimension and helps with analyzing real-world innovation strategies employed by authentic firms inresponse to entries by counterfeit entities. In addition, welfare analysis hints at a nonlinear relationship betweensocial welfare and intellectual property enforcement.

Keywords : counterfeit; emerging markets; searchable quality; experiential quality; signaling; two-dimensionalvertical differentiation

History : Received: January 17, 2013; accepted: January 25, 2014; Kannan Srinivasan and K. Sudhir served as theeditors-in-chief and Anthony Dukes served as associate editor for this article. Published online in Articles inAdvance.

1. IntroductionAlong with rapid growth and industrialization, emerg-ing markets are often characterized by underdevelop-ment in legal infrastructure. Emerging markets growfast despite these limitations. Last year, Chinese luxuryconsumption accounted for USD 12.6 billion, and by2015, China is expected to top luxury consumptionin the world. This study sheds light on some of theself-correcting mechanics of a market in a setting wherebrands face competition from counterfeiters underweak intellectual property rights (IPR), as is commonin many emerging markets. By providing counterfeitproducts with an appearance that is similar to thatof a brand, counterfeiters can trick consumers intopurchasing their products. Even nondeceptive coun-terfeits are desired by a large segment of customersin the emerging market. In a 2004 study on globalcounterfeiting, the World Customs Organization esti-mated that as much as 512 billion euros of traded world

merchandise may have been counterfeits (BusinessWeek2005). The sheer volume of counterfeit merchandisearound the world underscores the importance of study-ing how markets function in the face of counterfeiting,a topic that no multinational corporation operatingin emerging markets can ignore. In particular, howshould brands design various quality dimensions inresponse to counterfeiting?

To some extent, the literature has explored thedemand for counterfeits (Bloch et al. 1993, Wee et al.1995, Commuri 2009, Wilcox et al. 2009, Han et al. 2010),with price, social aspiration, and attitudes toward largebranded companies cited as main factors in drivingcounterfeit demand. On the supply side, several studieshave examined piracy effects (Anand and Galetovic2004, Conner and Rumelt 1991, Sinha et al. 2010, Verniket al. 2011); legal responsibilities (Olsen and Granzin1992); counterfeit impacts in the international tradesetting (Grossman and Shapiro 1988a, b); and firms

1

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.

Qian, Gong, and Chen: Searchable and Experiential Quality Responses to Counterfeits2 Marketing Science, Articles in Advance, pp. 117, 2014 INFORMS

internal organizations in complementing weak IPRenforcement (Zhao 2006).

Qian (2008) focuses on documenting the generalimpact of counterfeiting, whereas this paper providesa deeper theoretical examination of that impact bydisentangling the attributes of quality. The theoreticalmodel in Qian (2014) provides the microfoundations ofconsumer deception by a counterfeiter and an authenticfirms tactics (distribution choices and self-enforcement)to mitigate it. Neither of those papers dissects the extentand nature of innovations. This paper, in contrast, isfocused on the differential impact on two dimensions ofquality. That counterfeiters usually mimic an authenticproducts design but offer inferior functional qualityhas important implications for authentic producersincentives for innovation and the nature of innovations.Our theoretical framework helps unravel these com-plexities with intuitive closed-form solutions. AlthoughGrossman and Shapiro (1989) primarily define a pres-tige effect as a function of the total sales of the brand(and its counterfeits) and one-dimensional quality, wedecompose quality to a finer level. In addition, weendogenize quality choices and analyze the modelunder producers flexible cost structures. The find-ings provide practical guidance on brand protectionstrategies under different market conditions, wherecounterfeits may vary by their production costs andpervasiveness.

A key novelty in our theoretical model is that weextend Nelsons (1970) constructs on searchable andexperiential goods to quality dimensions within a good.Although Nelson categorizes goods into those thatare searchable and experiential, we observe that mostproducts possess a combination of searchable andexperiential attributes. For many products, includingauthentic products subjected to counterfeit imitations,some quality traits may be observable before purchase(e.g., brand logo, stitching, appearance) and othersunobservable (e.g., durability, breathability) at the timeof purchase. Counterfeits, by definition, share somesearchable (observable) quality traits with the authenticbrand, whereas the deception revolves around theexperiential (unobserved) quality. Untangling thesetwo quality dimensions yields insight into the uniqueimpact counterfeiting can have on branded productscompared with a generic entry. This study contributesto the literature on counterfeits and product qualitydifferentiation.

We build upon a vertical differentiation model(Gabszewicz and Thisse 1979, Shaked and Sutton 1983)with two dimensions of endogenous quality to analyzethe quality and pricing responses to counterfeit entry.Since counterfeiters attempt to copy authentic productsand usually produce inferior quality, the competitionis actually more vertical in nature. We incorporatetwo quality dimensions, as defined in the previous

paragraph, both vertical. This novel model capturesthe special feature of counterfeits in unbundling search-able quality (e.g., brand name and appearance) andexperiential quality (e.g., functionality) and examinesthe brands strategies in these two quality dimensionsin response to counterfeiting.

We analyze the market equilibrium under asymmet-ric information (specific to deceptive counterfeiting)and compare equilibrium solutions with the monopolyequilibrium. We take into account asymmetric informa-tion by building on the literature of quality uncertainty.Price is the conventional signal for product quality,but Nelson (1974) points out the importance of adver-tisement as a form of a nonprice signal for quality.Milgrom and Roberts (1986) argue that prices are bet-ter signals for quality than nonprice signals (notablyadvertisements) unless we assume repeated purchases.Moorthy and Srinivasan (1995) propose a money-backguarantee as another effective quality signal. Metrickand Zeckhauser (1999) use a simplified vertical dif-ferentiation framework to model competition underasymmetric information. Despite the sophistication ofthe previous literature, these models only consider amonopolistic market and assume exogenous quality lev-els. A few exceptions are Moorthy (1988), who examinesproduct and price competition, and Purohit (1994), whomodels a Cournot game with endogenous quality. How-ever, their models are confined to a single dimensionof quality. There exists a small literature on nonpricesignaling in nonmonopolistic markets. Hertzendorf andOvergaard (2001) and Fluet and Garella (2002) considerduopolies facing partially uninformed consumers. Theyshow that adjustments along nonprice investments(e.g., advertising and packaging) can be used to signalunderlying quality.

Unlike these traditional models that consider one pre-determined and often fixed quality for each producer,we introduce two quality variables for the producers tooptimize over. We incorporate asymmetric informationbetween deceptive counterfeiters and consumers byassuming that a fraction of consumers cannot differen-tiate counterfeits from authentic products at the sameprice and searchable quality levels.1 We then analyzethe quality decisions by producers. This endogenizationof qualities helps explain detailed quality dynamics inthe face of competition.

The key finding among a rich set of results is thatemerging markets can be self-corrective with respect tocounterfeit issues in the following sense: First, coun-terfeiters can earn positive profits by pooling withauthentic brands only when consumers have a priorbelief that there is low probability that any product isa counterfeit. When the proportion of counterfeits in

1 The asymmetric information due to status signaling is not dealtwith here, but it is in Qian (2014).

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.

Qian, Gong, and Chen: Searchable and Experiential Quality Responses to CounterfeitsMarketing Science, Articles in Advance, pp. 117, 2014 INFORMS 3

the market exceeds a cutoff value, brands invest in self-differentiation and profits dissipate for the competitive-fringe counterfeiters. Second, in the pooling equilibriumwith sporadic counterfeits, the authentic brand hasthe incentive to innovate experiential quality to attractmore expert consumers. As counterfeits become morerampant, the authentic producer invests in searchablequality instead of experiential quality to differentiate itsproduct from counterfeits in a separating equilibrium.

Although there is no consensus on how entry affectsinnovation in the economics literature (Purohit 1994,Aghion et al. 2005), marketers often note product dif-ferentiation as a practical consequence. Establishingsuch results in a closed-form solution is novel. Ourfindings yield additional insights into the nature andextent of quality differentiation. Our model aids inunderstanding the observations in practice that brandschoose to upgrade different quality dimensions underdifferent circumstances and, as a result, that they oftenraise prices after entry by counterfeiters (Barnett 2005).In addition, we highlight the strategic nature of search-able quality differentiation and analytically reveal thetwo uses of searchable quality upgrades in the face ofcounterfeiting: namely, widening the searchable qualitygap to alleviate competition and disentangling asym-metric information due to counterfeiting. Our studyprovides one of the first analyses on the theoreticalfinding that searchable quality often provides signalvalue for experiential quality and on the conditionsunder which entry leads to innovation.

It is worth noting that such innovation responses cangenerate mixed welfare implications across productcategories. In industries where the marginal benefitsof searchable quality are high, such as fashion prod-ucts, searchable quality innovations lead to welfaregains. However, in other industries where there islittle marginal utility associated with searchable quality,such as pharmaceuticals, such innovation responses arerather wasteful. Our predictions shed new light on thedebate in the marketing and economics literature onthe relationship between competition and innovation.

We buttress our main theoretical predictions by busi-ness cases across industries and empirical settings inQian (2014). Using a panel data set on the Chinesefootwear industry that includes detailed quality dimen-sions of a representative sample of brands and theircorresponding counterfeits, we reveal through analysesthat authentic branded firms invest handsomely in var-ious searchable quality dimensions (shoe style, surface,and side materials, etc.), but not in shoe bottoms andfunctional quality, where differences, if any, are minor.Such practices are starting to be adopted by variouscompanies, ranging from Microsofts software encryp-tions to the most recent conspicuous CD packagingand from unique imprints on pharmaceutical pills toholograms on Fendi scarfs (Passariello 2006). All these

findings align with our theoretical predictions. Theweakened regulation of counterfeit shoe manufacturersin China is a nice setting for the theorys empiricaltest. Therefore, this research can provide insights intomarketing in emerging markets.

We organize the rest of the paper as follows. We firstdevelop the theoretical model that incorporates twodimensions of quality under competition with asym-metric information. We then proceed to validate thekey theoretical predictions. Finally, we conclude bysummarizing the managerial insights and contributions.

2. Theoretical FrameworkIn this section, we first set up the game. We thensolve the benchmark case with a monopoly brandbefore probing into the competitive market equilibria.In particular, we derive the pooling and separatingequilibria properties in cases where counterfeitershave either a cost disadvantage or advantage over theauthentic producer. We refine the equilibria with theintuitive and divinity (D1) criteria (Cho and Kreps 1987,Banks and Sobel 1987). Finally, we end this sectionwith welfare analyses and discussions.

2.1. Model SetupExtending the traditional vertical differentiation frame-work to incorporate two dimensions of quality, wecharacterize a good with an experiential quality (e.g.,functionality) as index hi and a searchable quality(e.g., appearance) as index si, where i indexes good iand each company produces only one good. Experi-ential quality includes dimensions of quality that areunobserved and can only be inferred. In the data forshoe companies, h includes cushioning effects, friction,and other shoe functionalities. Searchable quality, onthe other hand, includes dimensions of quality thatare searchable or observable at the time of purchase.For instance, s includes the brand name, shoe surfacematerials, and other appearance characteristics. Eachdimension of quality could take on either a high orlow value, and we normalize the low value to be 1without loss of generality: 811H9.2

Experiential quality can be inferred based on pastexperiences and signals such as product price andsearchable quality. An authentic firm has been in themarket for a long time, so there are lots of reviews,advertisements, expert magazines, etc. There may bemany other ways to convey signals that we may notcapture in the model because we focus on the interac-tion with counterfeits. In our theoretical framework,consumers expect experiential quality ha based on priorexperience and beliefs. As a preview of later sections,

2 The key results are robust in the setting with continuous quality;please see Qian et al. (2013). This discrete quality setting makespredictions more salient and reveals additional insights.

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


in a separating equilibrium, the expected experientialquality for the authentic product is the authentic ha.In a pooling equilibrium, the expert consumers canstill correctly infer the true ha. However, the confusedconsumers form the expectation around the authenticha equaling the weighted average of the authentic andcounterfeit experiential qualities.

There is a continuum of consumers. Each consumerderives a utility from consuming a unit of the authenticproduct U =V 4ha1 sa15Pa = 4ha + sa5Pa if oneunit with quality 4ha1 sa5 is consumed by a consumerwith taste unif60117 at price Pi, and U = 0 if theconsumer does not buy any unit. All consumers preferhigh quality, given the same price. We therefore cap-ture consumer heterogeneity in taste by : a higher indicates more willingness to purchase a given quality.The cumulative distribution function F 45 is thereforethe fraction of consumers with a taste parameter lessthan . We start with the assumption that the tastes forthe two dimensions of quality are collinear becausehigh-income consumers are more able to afford bothdimensions of quality or because true lovers of theproduct are more able to appreciate its finer attributes.3

This assumption is similar to the one commonly madein the price discrimination literature: that individualsdemand curves do not cross. Relaxing this assumptiondoes not change the predictions qualitatively but onlycomplicates calculations.4

At first, one branded producer leads in the mar-ket, and a set of counterfeiters could potentially seekopportunities to enter. To offer a product of qualityqi = 4hi1 si5, the producer i faces a combination ofmarginal and fixed costs. The sunk cost for innovat-ing high quality H on either dimension of qualityis I . We assume that the production cost is linearand separable: Ca4s1h5= cs+ ch. The counterfeitersfree-ride on the authentic producers research anddevelopment and hence incur no sunk costs. They arelimited by technology and have no capacity to innovate.They can only produce the low quality by themselves.However, they can imitate the high quality by theauthentic producer as followers. That is, whenever thebrand sets either dimension of quality to the baselinelevel 1, the counterfeiters quality in that dimensionwill be 1 as well. Whenever the brand innovates eitherquality to H , counterfeiters have the liberty to choose

3 Analogously, Grossman and Shapiro (1988) model the utility of astatus good as being derived from its quality and prestige, whichpositively correlates with quality and negatively relates to totalsales. Our model differs from theirs in three important ways: (1)We disentangle quality into two dimensions. (2) We endogenizequality and price choices. (3) We incorporate potential asymmetry ininformation about quality due to counterfeiting. It is a commonphenomenon in emerging markets that consumers are fooled intopurchasing counterfeits.4 Results are available upon request.

between 1 and H for that quality dimension. As theyimitate the authentic quality qa = 4ha1 sa5, their qualityis qc = 4hc1 sc5, which brings the consumers utility levelto Uc = 4 hc + sc5Pc, where 0< 1. They usuallycannot perfectly reproduce the authentic experientialquality or bring about the same associated utility toconsumers ( < 1). We do incorporate the best-case sce-nario of = 1, whereby the counterfeiters bring aboutthe same experiential quality utility as the authenticfirm. The counterfeiters face perfect competition amongthemselves. Without loss of generality, we assume thatthese counterfeiters are symmetric in their marginalcosts: Cc4s1h5= 4c +c5s + 4c +c5h, where c > 0implies that counterfeiters have higher production coststhan does the authentic firm, and c 0 implies theopposite.5

We assume based on the practical intuition that thecost differential is moderate and the sunk investmentcosts are larger than the marginal cost differential.6 Thegeneral solution without this simplifying assumptionis outlined in Online Appendix A.16 (available assupplemental material at http://dx.doi.org/10.1287/mksc.2014.0867). The assumption helps to stream-line calculations without sacrificing intuitions andpredictions.

We model a game with a dominant authentic brandedfirm and a set of competitive-fringe imitators or coun-terfeiters that is closest to the actual interactions amongconsumers, authentic producers, and counterfeiters.The sequence of events is the following:

1. An authentic producer chooses her quality, bothexperiential and searchable 4ha1 sa5, with the corre-sponding costs as specified before. She also sets theoptimal price Pa to correspond to this quality level.

2. Based on the brands price and quality decisions,counterfeiters decide whether to enter the market and,if so, whether to enter without deception (i.e., revealthe true product source and quality), with deception(i.e., fool consumers into thinking they are real), orboth.

3. Counterfeiters pick their searchable quality scand the corresponding price Pc for a product withexperiential quality hc.

4. Each consumer purchases at most one unit of aproduct. They can observe searchable quality at thetime of purchase but can infer experiential quality onlybased on past experiences, word of mouth, and signalssuch as price and searchable quality.

5 Generally speaking, a product has multiple aspects for both search-able and experiential qualities. Hence, the cost for the authentic firmcan be expressed in general as Ca4si1hj 5= c4

Ni=1 si +

Mi=1 hj 5, and the

cost for the counterfeits is Cc4si1hj 5=N

i=14c+csi 5si +M

i=14c+chj 5hj .6 The production-cost difference between authentic and counterfeitproducts c and the innovation cost of each quality I 4c1 I5 have tosatisfy certain simplifying conditions for closed-form solutions.

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


To solve this game, we first solve for benchmarkprices and profits when the authentic firm is amonopoly. We then introduce asymmetric informationbetween counterfeiters and consumers, a unique featureassociated with counterfeiters compared to generalentrants. We examine the pooling equilibrium underthe constraint that the brands quality and price choicesneed to be such that counterfeiters have incentivesto pool. We solve for separating equilibrium prices,qualities, and profits when some consumers can befooled, under the constraint that counterfeiters have noincentive to match price and quality with the authenticfirm, which is necessary to guarantee separation. Sincethe authentic firm moves first, its actions will determinewhether we are in a pooling or separating equilibrium.We derive the perfect Bayesian equilibria solutionsunder both pooling and separating conditions.

2.2. Monopoly BenchmarkLet us start with the benchmark scenario where themarket has just a monopoly branded producer. Con-sumers will purchase the authentic product of quality4ha1 sa5 and price Pa if they derive positive utility from it.That is, 4ha + sa5Pa 0, implying that Pa/4ha + sa5.The market demand for the authentic product is then1 Pa/4ha + sa5. The authentic producer maximizesprofit:

maxh1 sa1Pa

{PaDa4sa1ha1Pa5 4c sa + c ha5Da4sa1ha1Pa5

I 6m4sa5+m4ha57}1 (1)

where m415= 0, m4H5= 1.Proposition 1. The optimal quality choices for the

branded monopoly are either qa = 41115 or 4H1H5. There isa threshold Im = 44H 1541 c525/2 such that it is profitablefor the authentic producer to innovate to qa = 4H1H5 whenI < Im and adopt qa = 41115 otherwise.Proof. Please see Online Appendix A.1.

The authentic firm has the same innovation costfor improving each dimension of quality here. Whenthere exists no counterfeit, consumers can observe thesearchable quality and they can infer the authenticexperiential quality; that is, consumers can get the trueinformation on both dimensions of quality. Thus, theauthentic firm is indifferent about improving eitherdimension of quality.

2.3. Entry by CounterfeitersThere are two types of counterfeits: nondeceptiveand deceptive. Nondeceptive counterfeiters revealtheir products as different from the authentic ones byoffering a lower price and searchable quality than theauthentic products. The deceptive counterfeiters intendto fool consumers by fully imitating authentic priceand searchable quality. Asymmetric information is

important in the context of counterfeiting because manyarticles and news stories reveal how consumers areconned into buying counterfeits. In a Chinese nationalsurvey with 30 retailers and 200 consumers (Qian 2014),some retailers were fooled by the counterfeiters whoclaimed to be the sales force of a branded companyand who offered huge discounts to fulfill their year-endsales quotas. The majority of consumers at the mallintercepts claim that they cannot tell counterfeits apart,and they usually rely on price or store signals to inferthe quality and authenticity of the product.

2.3.1. Modeling Deception. Although the opti-mized quality and prices are public information(ha1 sa1Pa1hc1 sc1Pc), we assume that a fraction, , ofconsumers may not be able to tell counterfeits apartfrom their authentic counterparts at the same price andsearchable quality (henceforth appearance), at leastnot until after the purchase. This setup is intuitivebecause authentic producers tend to provide detailedinformation about their products in order to buildreputation and brand recognition. Counterfeits, onthe other hand, mostly try to mimic the appearanceof authentic products and misrepresent attributes toextract short-term windfalls (Qian 2014).

The other 1 fraction of consumers are expertsin the product (e.g., based on the source or channelof the product) and know the exact quality of theproduct they are purchasing, even if they, too, do notdirectly observe the experiential quality. They haveaccurate expectations on the authentic experientialquality based on past experiences, word of mouth,and rational inferences. They may purchase counter-feits at a lower price depending on their individualwillingness to trade off quality for price (similar tothe case with complete information). Let 4i5 denoteuninformed consumers prior beliefs on the probabilityof any product being each type (counterfeit or authen-tic) =i 4i= a1 c5, where subscripts a and c denoteauthentic and counterfeit products, respectively. Wedenote 4c5= and 4a5= 1 . They are drawnuniformly from all consumers in the valuation distri-bution.7 They can calculate P sa , s

sa, s

pa , and P

pa , where

superscripts s and p denote separating and poolingequilibria, respectively. However, they cannot recognizethe experiential quality in practice. After observing thesellers price and searchable quality, consumers updatetheir beliefs about the sellers type. Let 4i p1 s5= 1denote uninformed consumers updated beliefs aboutseller is type being exactly =i.

We handle the technical issue of specifying the out-of-equilibrium beliefs following the prior literature.There are two main methods. One of them is to assume

7 Relaxing this assumption and drawing more heavily from thelow-valuation consumers would not qualitatively change the results.

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


a particular set of beliefs following a deviation (e.g.,McAfee and Schwartz 1994). This approach is some-times criticized for considering the specific set of chosenbeliefs to be arbitrary. The second method is to startwith an unconstrained set of out-of-equilibrium beliefsand then narrow it down using existing refinement.This approach imposes some structure on the out-of-equilibrium beliefs by suggesting that the morereasonable belief would be the one that is consistentwith the refinement. The intuitive criterion (Cho andKreps 1987) and the divinity criterion (Banks and Sobel1987) are the two existing refinements most commonlyused in signaling models (e.g., Simester 1995, Desaiand Srinivasan 1995, Zhao 2000, Feltovich et al. 2002,Harbaugh and To 2005, Mayzlin and Shin 2011).

We apply these popular criteria to refine theset of perfect Bayesian equilibria. The intuition isbriefly described here and proved in detail in OnlineAppendix A.14. Consider the set of best responsesassociated with a particular out-of-equilibrium belief.Suppose that the counterfeit type benefits from thedeviation under a set of best responses bigger than theauthentic types. Furthermore, this is the case for allpossible beliefs. The refinement criteria then requirethat the consumer not believe the deviating type isan authentic producer. More generally, suppose thatin deviation 4p1 s5 6= 4P 1 s5, type producer is morelikely to yield higher profits in equilibrium under abigger set of best responses from the consumer than istype producer. D1 then requires that the consumernot believe the product could be from type .

The D1 criterion is stronger than the intuitive crite-rion in that it does not require that the authentic typenot benefit from the deviation under any possible belief.Instead, it requires that the set of consumers bestresponses, which are based on the consumers beliefs,be strictly smaller than that of the counterfeit type.We require that a potential equilibrium be supported byout-of-equilibrium beliefs that survive the refinement.We detail the intuitive criterion and D1 criterion withtheir applications in our setting (both separating andpooling equilibria) in Online Appendix A.14.

The authentic producer maximizes profit as in Equa-tion (1). The counterfeiters can choose searchable orexperiential quality of either 1 or H . We denote thecounterfeiters searchable and experiential qualities as4sc1hc5. Since the nondeceptive counterfeiters are underperfect competition with each other, they set the priceat the marginal cost level: Pc = 4c+c5sc + 4c+c5hc =4c+c54sc + 15.Lemma 1. The counterfeiters will always keep the expe-

riential quality low at hc = 1.Proof. Please see Online Appendix A.2.

The intuition is simple: since there is only one deal-ing between consumers and the counterfeit products,

counterfeiters have no incentive to invest in experientialquality that consumers do not observe.

We find that the authentic firm can adopt differentstrategies depending on the production costs to coun-terfeiters. We consider the following two scenarios: (1)when counterfeiters have higher production costs thandoes the authentic brand, i.e., c > 0, and (2) whencounterfeiters enjoy lower production costs, i.e., c 0.In the following sections, we analyze the two typesof equilibria under each of these two scenarios: thepooling equilibrium and the separating equilibrium.

2.3.2. Counterfeiters Have a Higher Marginal Cost.Sometimes counterfeiters marginal costs can be higherthan the brands. For instance, the elegant appearanceof authentic branded shoes requires superior materials,fine labor, and pattern-press technology during theproduction process. Without economies of scale andlegitimate accesses to technology and knowhow, it isoften more costly for counterfeiters to achieve the exactappearance of the authentic product. Notably, as illegalentities, counterfeiters do not have an import licenseto acquire fancy materials (Qian 2008), and brandedcompanies abide by strict allocation of raw materials totheir subcontracted manufacturers to prevent them fromcounterfeiting (Zhao 2006). Counterfeiters usually useinferior materials to mimic the authentic appearance,and it naturally becomes harder to do so when theappearance becomes more sophisticated. Because theauthentic producer enjoys economies of scale anddesigns an appearance level that is hard to imitate, wealso consider the case where c > 0.

The results derived in this section hold in the generalsetting of Cc4si1hj5=

Ni=14c+csi 5si +

Mi=14c+chj 5hj ,

whenever maxNi=1 chj > 0, that is, whenever there exists

one aspect of either dimension of quality that requiresa higher marginal cost from the counterfeiters.Pooling equilibrium. Our theoretical framework yields

insights into the pooling equilibrium where counterfeit-ers opt into the same price and searchable quality levelsas authentic producers. In this section, we analyze themodel to draw out these implications.

Because only 1 fraction of consumers is expert indifferentiating authentic products from counterfeits, theother fraction of consumers will have probability ofbeing conned into purchasing counterfeits.8 There arenow three types of products: authentic branded prod-ucts; nondeceptive counterfeits, which self-differentiatefrom the authentic products with a lower quality andprice; and counterfeits that are indistinguishable from

8 Here, is the availability of counterfeits and is a function ofgovernment law enforcement. Bearing in mind that counterfeitingcarries the risk of being caught and punished, we assume that therecan be at most N counterfeiters in the market, covering proportionof the market.

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


the authentic product in price and appearance (search-able quality). The latter two types of products are bothproduced by the set of competitive-fringe counterfeiters.Their production capacity is not satiated in either theperfectly competitive low-end market with residualdemand or the pooling market with fixed probability of being purchased by the uninformed consumers.These counterfeiters thus produce both types of prod-ucts in a pooling equilibrium. They set 4sa1Pa5 fortheir counterfeit products in a pooling equilibrium and4sc1Pc5 for the nondeceptive counterfeits.

If the authentic firm does not improve searchablequality (sa = 1), the nondeceptive counterfeiters canonly choose sc = 1 because that is the only availablelevel they can imitate. If the authentic firm improvesthe searchable quality to H , the counterfeiters canchoose either sc = 1 or sc = H . Given the technicalassumption that c is not too big,9 since the counter-feiters compete perfectly and set the price equal toits marginal cost, consumers will strictly prefer thecounterfeiters with sc =H . We therefore have sc = sa inequilibrium.

Expert consumers will always choose authenticproducts over counterfeits in this pooling equilib-rium. The expert consumers who purchase the authen-tic product must derive nonnegative utility from it,V 4ha1 sa1 5Pa 0, and that utility must exceed theirutility derived from a unit of the lower quality product,V 4ha1 sa1 5Pa V 411 sc1 5Pc. These two conditionsimply that the indifferent customer who is equallyhappy with an authentic or counterfeit product musthave taste = 4Pa Pc5/4ha 5 for the nondeceptivecounterfeits market to persist. The demand of informedconsumers for the authentic product is

Di4Pa5= 1 = 1 Pa Pcha

0

The novices who purchase any product with appear-ance sa at price Pa must have a positive expected utility,satisfying 41 5V 4ha1 sa1 5+ V 411 sa1 5Pa 0 and41 5V 4ha1 sa1 5+ V 411 sa1 5 Pa V 4hc1 sc1 5 Pc.This implies that the customer who is indifferentbetween purchasing the authentic and counterfeit prod-ucts has taste

4154ha+ sa5+4hc+ sc5Pa= hc+ scPc = 24Pa2c2c5

4154H5 0

It follows that the demand of uninformed consumersfor the pooled authentic and deceptive counterfeits is

Du4Pa5= 1 = 1 Pa Pc

41 54ha hc50

9 In particular, when c 4c41 55/4sc 15, sc = sa.

The total demand is Da4Pa5= 415Di4Pa5+ 41 5 Du4Pa5, and the authentic firm will maximize its profitas follows:

pa = maxPa1 sa1ha

{4Pa csa cha5Da4Pa5 I 6m4sa5+m4ha57

}0

Proposition 2. In a pooling equilibrium, the authenticfirm will not increase searchable quality. In addition, thereexists a threshold I ph as follows:

Ip

h =641 54H 5+ 2c c4H 1572

44H 5

641 541 5+ 2c72

441 5 0

When I I ph , the authentic producer will only improve theexperiential quality, and when I > I ph , the authentic producerwill improve neither quality.

Proof. Please see Online Appendix A.3.

This proposition illustrates that when counterfeitersare able to pool with the authentic firm on price andsearchable quality, the authentic producer will not havethe incentive to improve its searchable quality sincecounterfeiters are capable of completely replicating.However, when innovation costs are low, the authenticfirm will improve experiential quality to capture moreof the expert consumers who know all quality levels.

Corollary 1. (1) When informed consumers increase,that is, decreases, the authentic firm will have a strongerincentive to improve experiential quality. (2) When counter-feits become more rampant in the market, that is, increases,the authentic firm will have less of an incentive to improveexperiential quality.


The intuition of part (1) directly follows the dis-cussion underlying the above proposition since thehigher experiential quality is especially tailored to theinformed consumers. Part (2) hints at the fact that in apooling equilibrium, some of the authentic innovationsalso benefit counterfeiters because of the pooling natureof the uninformed consumers. Therefore, as counterfeitsbecome more rampant in the market, the authenticfirm has less incentive to innovate.Separating equilibrium. If the authentic producer and

counterfeiters choose different prices and searchablequalities from each other, then all consumers can tellthe products apart based on the model setup. To dif-ferentiate their products from counterfeits, authenticproducers should set the price at Pa 4c+c54sc +hc5=4c + c54sa + hc5the incentive compatibility (IC)constraintto preempt the deceptive counterfeiters.The brand competes with nondeceptive counterfeitswith quality 41115 and price Pc = 24c+c5.

We again denote as the threshold level of consumertaste for qualities, at which a consumer is indifferentbetween consuming a branded product or a counterfeit,

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


and consumers with utility level will buy a brandproduct if and only if . We have

ha+ saPa= hc+ scPc =Pa24c+c5ha+sa1

0

The demand for the authentic products is

D4Pa5= 1 = 1 Pa 24c+c5ha + sa 1

0

The authentic firm will maximize its profit as follows:

maxPa1ha1 sa

a4sa1ha1Pa5= maxPa1ha1 sa

{4Pa Ca5Da

I 6m4sa5+m4ha57}

s.t. Pa 4c+c54sa + 150 (IC constraint)We then arrive at the following proposition.

Proposition 3. When c > 0, the authentic producercan separate its product from counterfeits by improvingsearchable quality. Because it is unprofitable for counterfeit-ers to mimic the high searchable quality when the authenticproducer sets a low price, the market forms a separatingequilibrium. The authentic firm will not improve its expe-riential quality. In addition, there exists a threshold I ss asfollows:

I ss =c[4H 15 4c+c5

(H 2 1H 2

)]0

The authentic producer will improve searchable quality iffI < I ss .


In the case where counterfeiting involves a highermarginal cost than the authentic cost and the IC con-straint favors the separating equilibrium, the brandwill choose a higher searchable quality to differentiateitself from counterfeits. The brand is more likely tooffer the baseline level of experiential quality since ahigher level would not help the brand to differentiateitself from counterfeits in the eyes of all consumers.The counterfeiter will choose the lowest quality levelin both dimensions. This proposition highlights one ofthe key findings: counterfeiting can stimulate the brandto offer a higher searchable quality but a stagnantexperiential quality.

Because of its nature, the innovation induced bycounterfeits may be more valuable in certain industries,where searchable quality brings utility in its ownright (e.g., fashion), than in others where consumerutility is primarily derived from experiential quality(e.g., pharma). We provide empirical support for thisprediction based on various cases and a unique paneldata set on the Chinese footwear industry in the nextsection.

2.3.3. Counterfeiters with Lower Production Costs.We next consider the case where counterfeiters canimitate the branded product at a lower cost than that

incurred by the authentic producer (c 0). Counter-feiters will always keep the same searchable quality fortheir product as that of the branded product, whichmeans sa = sc. However, when it comes to experientialquality, Lemma 1 shows that counterfeiters will alwayskeep it at hc = 1. There will always be some incentivefor the authentic producer to improve the experientialquality of the product if the innovation cost is lowenough.

Proposition 4. (1) When c 0, the authentic firmwill never improve its searchable quality. (2) When 0< < 1,there exists a threshold I ph as follows:

Ip

h =641 54H 5+ 2c c4H 1572

44H 5

641 541 5+ 2c72

441 5 0

When I < I ph , it is profitable for the authentic producerto upgrade experiential quality. When I > I ph , the authen-tic producer will improve neither experiential quality norsearchable quality.


When counterfeiters have a marginal cost advantagein producing the searchable quality compared with theauthentic producer, the authentic producer will nothave the incentive to improve its searchable qualityjust to be copied by the counterfeiters.

Corollary 2. (1) When the proportion of informedconsumers increases, that is, decreases, the authenticfirm will have a stronger incentive to improve experientialquality. (2) When counterfeits become more rampant inthe market, that is, increases, the authentic firm willhave less incentive to improve experiential quality. That is,6a411H5a411157/ < 0, 6a411H5a411157/< 0, I ph/ < 0, I

p

h/ < 0. In addition, 6Pa411H5Pa411157/ < 0, 6Pa411H5 Pa411157/ < 0.Proof. Please see Online Appendix A.7.

The comparative statics reveal that when the propor-tion of informed consumers increases, i.e., decreases,the profits of the authentic firm will increase, asdepicted in Figures 1 and 2. However, the increasein a41115 is faster than that in a411H5 with respectto . The price of the authentic product will increase,whereas Pa411H5 increases faster than Pa41115. Whencounterfeits become more rampant in the market, thatis, increases, the price of the authentic productwill decrease. Authentic price Pa41115 decreases fasterthan Pa411H5. This is very intuitive since experientialquality is known by the informed consumers only.The more informed consumers there are in the market,the larger the demand for the authentic product. Themore pervasive counterfeits are, the more free-riding

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


Figure 1 Authentic Profits Under Pooling and Separating Equilibria as Functions of s= 1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Profit

Pooling Separating

Figure 2 Authentic Profits Under Pooling and Separating Equilibria as Functions of s= 1

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Profit

Pooling Separating

of the authentic innovations by counterfeiters (espe-cially in confusing the uninformed consumers), theless incentive the authentic producer would have forinnovation.

Finally, we explore the extreme case when = 1,which implies that counterfeiters can imitate the authen-tic producer completely at a lower marginal cost.

Corollary 3. When c 0 and = 1, there exists athreshold I ph :

Ip

h =1 cH 1+ 2c2

4H 1

such that when I > Iph , the authentic producer will improveneither its experiential nor searchable quality. Since qai =11 does not break even, the authentic producer will exitthe market.


This corollary illustrates the lemon problem perAkerlof (1970). When deceptive counterfeiters enjoya cost advantage over the authentic brand, the low-quality counterfeits can drive higher-quality authenticproducts out of the market. The only way for theauthentic producer to survive in the market is byimproving its experiential quality in order to capturethe informed consumers who value higher experientialquality, provided that such innovations will not incurtoo high of a sunk cost.

In general, when the authentic producer adopts thepooling strategy, the experiential quality is alwaysimproved to capture the high-valuation informed con-sumers, in order for the authentic brand to survive inthe market. In addition, the authentic firm cannot usethe separating strategy with c 0.10 That is,Corollary 4. When c 0, the authentic producer

cannot be separated from counterfeiters by improving itssearchable quality. Hence, no separating equilibrium isachieved.


2.4. Self-Correcting MechanismIn analyzing and comparing an authentic firms incen-tives for the pooling and separating strategies, wediscover a self-correction mechanism in the emergingmarket: when counterfeits become too rampant in themarket, the authentic brand self-differentiates fromcounterfeits, switching from a pooling strategy to aseparating strategy.

When the marginal cost advantage of the authenticbrand over counterfeits is not big enough, a poolingstrategy is adopted and the experiential quality is

10 The corresponding condition in a more general form ismaxNi=1 c

hj 0.

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


improved first. The maximized profit and the optimalprice can be written as

pa 411H5=641 54H 5+ 2c c4H 1572

44H 5 I1

P pa =41 54H 5+ c4H + 35+ 2c

20

As the searchable cost advantage of the authenticfirm increases, it becomes profitable for the authenticfirm to strive for a separating equilibrium wherebyit signals authentic quality against counterfeits. Thebrand would improve its searchable quality rather thanits experiential quality. The maximized profit of theseparating equilibrium and its optimal price are

sa 4H115=c4H + 15[

1 4c+c54H 15H

] I1

P sa = 4c+c54H + 150We contrast the maximized profits under different

conditions and obtain the following: When pa 411H5 sa 4H115, the authentic producer will adopt a poolingstrategy. When pa 411H5 < sa 4H115, the authenticproducer will adopt a separating strategy by improvingsearchable quality.

Proposition 5. There exist thresholds and , suchthat when either or , the authentic producerchooses the pooling strategy.

= 1 (

4c4H + 154H 5[

1 4c+c54H 15H

] 2c+ c4H 15

) 1H

and

= 1 (

4c4H + 154H 5[

1 4c+c54H 15H

] 2c+ c4H 15

) 1H 0


The set of predictions here weaves into a story ofthe self-corrective mechanism of emerging markets. Wedetail the predictions and the corresponding intuitions.Proposition 5 demonstrates that the authentic producerhas more incentive to improve experiential quality ina pooling equilibrium to capture the high-valuationexpert consumers. When there are lots of uninformedconsumers in the market, the authentic producer prefersthe signaling strategy, thus improving the brandssearchable quality in order to inform the consumersof its authenticity and superior quality. In emergingmarkets, where infrastructures are highly imperfect,

signaling becomes even more crucial. Consistent withour empirical findings, the authentic producer prefersto improve searchable quality instead of experientialquality.

Proposition 6 below shows that asymmetric informa-tion aggravates competition and promotes the need todifferentiate. In the pooling equilibrium, if attainablegiven the incentive compatibility constraints of thebrand and counterfeiters, the counterfeiters split themarket with the brand and earn a positive profit. Thisis more desirable for counterfeiters than the separatingequilibrium, where counterfeiters as competitive fringescompete away the economic rents from the residualdemand.

Proposition 6. (1) When the fraction of uninformedconsumers 45 increases, the authentic firm will have strongerincentives to adopt the signaling strategy. (2) When the pro-portion of counterfeits in the market 45 increases, theauthentic firm will have stronger incentives to adopt thesignaling strategy and to separate from the counterfeits.Formally, 6pea 411H5 sea 4H1157/ < 0, 6pea 411H5 sea 4H1157/ < 0, 4I

pe

h I se5/ < 0, 4I peh I se5/ < 0.In addition, 6P pea 411H5P sea 4H1157/ < 0, 6P pea 411H5P sea 4H1157/ < 0.


When there are more uninformed consumers andmore counterfeits in the market, the probability thatuninformed consumers will purchase counterfeitsincreases. When asymmetric information becomes toosevere, the brand has incentives to self-differentiatefrom the counterfeiters to alleviate the erosion in profits.In addition, under more severe asymmetric information,the authentic producer can charge a higher price whenimproving searchable quality than experiential quality.

Private incentives to compensate for public enforce-ment surge when counterfeits become very pervasiveand when too many consumers cannot tell counter-feits from authentic products. The following corollaryestablishes that the authentic producer will upgradesearchable quality to differentiate its product fromcounterfeits when the proportion of uninformed con-sumers is above a threshold level, given any relativelylarge proportion of counterfeits in the market.

Corollary 5. For any given > , there exists > such that when 601 5, the authentic producer prefers apooling strategy and improves experiential quality. When 6117, the authentic producer chooses a signaling strategythrough innovations in searchable quality, as depicted inFigure 2.

= 1

(

4c4H + 154H 5[

1 4c+c54H 15H

] 2c+ c4H 15

) 14H 5 0

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.



Analogous to the previous corollary, the followingresult illustrates that the authentic producer will chooseseparating strategies to differentiate its product fromcounterfeits when the probability of randomly picking acounterfeit from all of the products is above a thresholdlevel, given the existence of any relatively large fractionof uninformed consumers in the market.

Corollary 6. For any given > , there exists > such that when 601 5, the authentic producer prefers thepooling strategy with innovation in experiential quality.When 6117, the authentic producer chooses a signalingstrategy with innovation in searchable quality, as depictedin Figure 1.

= 1

(

4c4H + 154H 5[

1 4c+c54H 15H

] 2c+ c4H 15

) 14H 5 0


When there are few uninformed consumers orthere is a small probability of purchasing counter-feits unknowingly, the authentic producer will optfor the pooling equilibrium because the small loss ofdemand to the counterfeits does not justify the cost forupgrading searchable quality in order to separate fromcounterfeits.

Those counterfeiters who imitate the brand withthe same searchable quality and price in the poolingequilibrium gain a total profit as follows:

c = 6Pa4c+c54sa+hc57[

1 PaPc4154hahc5

]= 464154H5+c4H152c7641+25

4H5c4H15+2c75 144154H5 0

When = 11c = 41 54H 5/4 6c4H 152c72/4441 54H 55, there is a 40115 such thatthe counterfeiter will choose to maximize profits.It is clear that this part of the counterfeiters profitscomes from cheating uninformed consumers. When thefraction of uninformed consumers or the proportionof counterfeits in the market increases, the profit tocounterfeiters increases. As the positive rents attractmore and more counterfeit entries, the authentic pro-ducer will eventually find it more profitable to strivefor a separating equilibrium. The fringe market iscompetitive, so the profit to counterfeiters dissipates tozero.

In sum, this section uncovers the self-correctionproperty of the emerging market: when counterfeit-ing becomes too pervasive and confuses too manyconsumers in the market, the authentic producer is

motivated to upgrade searchable quality and forcecounterfeiters into a separating equilibrium.

In refining the perfect Bayesian equilibria (PBE) ofthis game, we have the following:

Proposition 7. The derived pooling and separatingequilibria survive the intuitive criterion and D1. There is aunique PBE for any parameter values under the intuitiveand D1 criteria. In addition, c = 1 4p= P sa1 s = ssa5 isconsistent with the intuitive criterion and D1.


2.5. Welfare ImplicationsIn this section, we calculate the welfare impacts ofcounterfeiting. Notably, we are interested in derivingthe welfare implications of the different market struc-tures. While the detailed derivations of social welfare inOnline Appendix A.15 do not lend themselves to sim-ple solutions for comparisons, we conduct simulationexercises to show the important comparative statics.Social welfare declines if the asymmetric informationbecomes more severe in the market; i.e., and arebigger.

In the monopoly case, given qa = 41115, we havePMa = c+ 1. Given qa = 4H1H5, we have PMa =H41 + c5.The consumer surplus and social welfare is as follows:

CSM 4sa1ha5= 164sa +ha5 PMa 7 d1

SWM 4sa1ha5=CS+a4sa1ha50We show in the next proposition that innovation

increases consumer surplus and social welfare.

Proposition 8. (1) When I < Im = 4H 1541 c52/4,the authentic producer will innovate and choose qualityqa = 4H1H5, CSM 4H1H5=H41c52/4, and SWM 4H1H5=3H41 c52/4 2I . When I Im = 4H 1541 c52/4,the authentic producer will not implement any improve-ment; i.e., qa = 41115. We get CSM 41115 = 41 c52/4and SWM 41115= 341 c52/4. (2) In addition, we haveCSM 4H1H5 > CSM 41115 and SWM 4H1H5 > SWM 41115.


We next turn to analyzing welfare when counter-feiters enter the market. We separately discuss twoscenarios, depending on the extent of asymmetricinformation in the market (the values of and ).

(a) When and are not too big, the authen-tic firm will choose the pooling strategy. In equilib-rium, social welfare is the sum of consumer surplus,authentic profits, and counterfeit profits. The consumersurplus for the informed consumers is CSpi 4sa1ha5= 164sa + ha5 Pa7 d+

64sc + hc5 Pc7 d, and the

consumer surplus for the uninformed consumersis CSpu4sa1ha5 = 41 5

164sa +ha5Pa7 d+

164sa+

hc5 Pa7 d+

64sc +hc5 Pc7 d7. Total social welfare

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


is SW p4sa1ha5= 41 5CSpi +CSpu4sa1ha5+pa 4sa1ha5+

pc 4sa1hc5.Since the solution in the general situation is compli-

cated, we first consider some possible results when theproduction cost is the same and the number of unin-formed consumers or the proportion of counterfeits issmall enough; i.e., 0 or 00

When I < I ph = 4H 1541 2c5/4 + c24H 152/444H 55, the authentic firm has the incentive toimprove its experiential quality

CSp411H5= 2c2

1 + +4H + c4H 1552

84H 5+ 1 +

2 c4H + 35

2

SW p411H5= 2c2

1 + +34H 5

8+ 36c4H 157

2

84H 5 c42H + 55

4+ 1 +

2 Ih0

However, when I > I ph = 4H 1541 2c5/4 + c24H 152/444H 55, the authentic firm has no incentive toimprove its quality

CSp41115= 2c2

1 + +41 52841 5 +

1 + 2

2c1

SW p41115= 2c2

1 + +7 +

8 2c0

It is easy to check CSp411H5 > CSp41115 andSW p411H5 > SW p41115. This implies that qualityimprovement will increase both consumer surplus andsocial welfare. We discuss some special cases below.

We can show that I ph < Im (see Online Appendix A.15).

Hence, if counterfeits enter the market and I < I ph , theentry encourages market competition but not innova-tion. Social welfare increases.

When I 4I ph1 Im5, the authentic firm will not improvesearchable quality. That is, the entry of counterfeitersdepresses social welfare. Because I > Im, there is noquality improvement in both cases, but the entry ofcounterfeiters adds to the market competitivenessand increases social welfare. It is easy to check thatwhen c 0, we have I pb 0. The above resultshint at the uncertainty associated with the impacts ofcounterfeiting on authentic quality and social welfare.

(b) As and increase, the profits of the authenticfirm decrease. The authentic firm prefers the signal-ing strategy. Social welfare is the sum of consumersurplus and the profits of the authentic firm. That is,CSs4sa1ha5=

164sa +ha5Pa7 d+

64sc +hc5Pc7 d,

and SW s4sa1ha5= CSs +a4sa1ha5. Figures 3 and 4 showthese comparative statics.

Synthesizing all of the results, welfare is likely to benonlinear in government enforcement. Notably, whengovernment enforcement is so strong that there are

no counterfeits, the monopoly producer surplus isthe largest. When public enforcement is of mediumstrength, the brand will pool with a moderate numberof counterfeits, with consumer surplus compromisedfor the novices. When public enforcement is so weakthat counterfeits become rampant in the market, brandswill innovate to separate from counterfeits, resulting inhigher social welfare. Consumers enjoy more productvariety (innovated authentic products and low-qualitycounterfeits).

3. Empirical Validations3.1. Summary of CasesThe main theoretical predictions are buttressed by vari-ous business cases. As cited in the introduction, manyproducers have responded to counterfeit infringementsby upgrading searchable quality while limiting experi-ential quality improvements. To note a few examples,musicians and recording companies conspicuouslypackage their CDs in the face of rampant piracy. Promi-nent pharmaceutical companies like Abbott and Pfizeradopt unique color, shape, and RFID imprints on eachpill to differentiate their products from counterfeitdrugs even though the active ingredients remain thesame. Many fashion brands, such as Fendi, adopt avisible hologram on their products (Passariello 2006).

Using panel data on branded leather and sportshoes in China, one of the largest emerging markets,we provide additional empirical support for the keyproposition that authentic producers have incentives toimprove experiential quality upgrades while overin-vesting in searchable quality in the face of entry bycounterfeiters, compared to a monopoly equilibrium.Furthermore, the easier it is for counterfeiters to imitateauthentic quality, the less an authentic producer willupgrade experiential quality.

3.2. Data and Identification StrategyThe data used to validate our theoretical predictionsconsist of detailed information taken from companiesannual financial statements and other relevant companyrecords. In total, we tracked 31 branded companies(including multinational and national brands) andtheir corresponding counterfeits from 19932004. Ourdata set includes the average prices and costs for eachauthentic brand and each corresponding counterfeit ateach product quality level (e.g., high, medium, low).In addition, the data collected include the brandstotal domestic sales. All of the data were taken fromeach companys financial statements. The data setincludes both domestic brands and multinationalbrands operating in China and is supplemented by theChinese Industrial Census database, the eBay in Chinadata set, product catalog information, and interviews.Data on counterfeit entry, prices, and costs for eachbrand were obtained by contacting each companysbrand-protection offices. Also, government agencies

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


Figure 3 Social Welfare Under Pooling and Separating Equilibria as Functions of s= 1

0 0.1 0.2 0.3 0.4 0.5

0.6 0.7 0.8 0.9 1.0

0.70.80.91.01.11.21.31.41.5

Wel

fare

Pooling Separating

Figure 4 Social Welfare Under Pooling and Separating Equilibria as Functions of s= 1

0

0.70.80.91.01.11.21.3

Wel

fare

1.41.5

0.1 0.2 0.3 0.4 0.5 0.6

Pooling Separating

0.7 0.8 0.9 1.0

(specifically, the Quality and Technology SupervisionBureau) were used as alternative sources of data onauthentic and counterfeit firms. The Data Appendices,available online, provide further data checks as well assampling and survey methods.

An important part of our research was locating andanalyzing product-level data. The product-level dataused in our study were coded and compiled based oninformation found in the companies and stores annualcatalogs. These product catalogs assisted in better con-trolling for product quality and costs. We compiled adata set of the different characteristics for each typeof shoe listed in the catalogs, consisting of materialstructures, comfort levels, decorative patterns, supportand cushioning features, ventilation, etc. Recognizingthe importance of validating the data from firm reports,we ran Hedonic regressions of the unit productioncosts, as provided by the sampled companies, on thecorresponding material, machinery, and other char-acteristics of the shoes, as recorded in the catalogs.We conducted the analyses on the samples of leathershoes and sport shoes separately. Those characteristicstogether accounted for 90% of the products cost varia-tion. These results lend credibility to the company data.We detail the regression model and results in 3.4.

A key difficulty in empirically measuring entryeffects on prices and quality is that entry is oftenendogenous to these outcomes. Following Qian (2008),we exploit a natural experiment resulting from the

unexpected loosening of government enforcement inChina for footwear trademarks in 1995. The govern-ment reallocated enforcement resources away fromfootwear due to a series of accidents in several othersectors, notably gas explosions and food poisonings,leaving loopholes for massive counterfeit entry into theshoe markets soon after. In addition, this policy changemade it more likely for counterfeiters to enter a marketwhere authentic companies had a poor relationshipwith the local government than where authentic com-panies had strong ties to the government.11 For thatreason, monitoring differential relationships betweeneach branded company and its respective govern-ment was vital in order to identify the entry effectsof counterfeiters on different brands. We thereforeused the interaction between the policy change andthe differential relationships between each brandedcompany and the government to identify the entryeffects by counterfeiters for different brands, as inQian (2008). For brevity, we relegate discussions of

11 This finding was revealed in both qualitative interviews withmanagers and data analyses, as detailed in 3.4. The importanceof relationship is not confined to China. At an Imaging SuppliesCoalition Anti-Counterfeiting conference in 2008, Andrew Gardnerfrom Lexmark International gave a case study on an actual stingoperation to bust a counterfeiting ring in developing countries. Hearticulated how close collaboration between multiple vendors andthe local government was crucial to successfully executing thisoperation.

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


the institutional and identification details to OnlineAppendix B. Results are qualitatively similar whenusing this instrumental variable (IV) approach andwhen simply exploiting the regression discontinuity inthe natural policy experiment.

3.3. Shoe Quality CharacteristicsBased on interviews and analyses of authentic pricesand costs, Qian (2008) notes that authentic firms didnot engage in innovation when there was alreadysignificant competition from all the other authenticproducers. Innovation occurred and prices jumpedonly after these authentic brands were infringed uponby counterfeits (p. 1604). Using the data on shoecharacteristics that were gathered from product catalogs,we rank the quality in each observable or meaningfuldimension according to the cost of the materials usedfor that dimension. For instance, the variable for surfacematerials takes on a value of 1 for plastic leather1 0 0 0 14 for regular cow skin1 0 0 0 , and 14 for crocodile skin, inascending order of material costs. Similar proceduresare carried out to generate the variables for measuringthe quality of the shoe side and bottom materials.

For the less-quantifiable attributes, we mainly usedummy indicators. We construct the variable for shoeappearance (searchable quality) by summing up threedummy variablesfine, elegant, and patternseachtaking on a value of 1 if a pair of shoes is describedin the catalog as possessing the characteristics and0 otherwise. A simple sum is generated instead of aweighted sum to avoid biasing results according toprior beliefs on which attributes constitute the moreimportant quality components. We then construct avariable for functionality by summing up the followingindicator variables: versatility; cushioning (whethera pair of shoes has cushioning effects); absorption(whether it can absorb sweat); countering athletesfoot; softness; comfort; sturdiness; warmth; friction(for protection on slippery ground); and additionalfeatures for sport shoes such as durability, flexibility,and support.

We exhaust all of the attributes that are mentionedin the product catalogs. We also construct a variableindicating the technology applied to make the shoes,embodied in the equipment. Before counterfeit entry, allof the companies used domestic equipment. However,after entry, many of them imported Italian12 productionlines, pattern-pressing machines, and equipment tomake shoe bottoms using cow skin. Some companieseven adopted nanotechnologies.13 We first construct adummy variable for each type of equipment and then

12 Italy has a famous cluster of shoe manufacturers.13 It is claimed that nanotechnology can be applied to making shoesthat absorb foot sweat and increase comfort.

add them up to generate the equipment variable.We generate the variable workmanship to indicatewhether the shoes were made with detailed and carefulcraftsmanship. Finally, we add up the values of thesedifferent characteristic variables to obtain the overallquality proxy.

3.4. Analyses and ResultsWe study shoe characteristics before and after thebrands were infringed by counterfeiters because ofthe unexpected loosening of government enforce-ment. Wilcoxon ranksum tests on the set of product-characteristics variables pre and post entry by

Table 1 Wilcoxon Ranksum Tests and Regression on Leather ShoeQuality

Panel A

Dependent variable Functionality Appearance Appearance

Entryt2 1035 9037 8050410595 420385 410515

c 2005

410015Company and year FE Y Y YNo. obs. 3,336 3,336 3,336

Panel B

Medium value Ranksum test(1) (2) (3)

Pre-entry Post-entry p-value

Surface material Regular cow Precious cow 00000

Side material Regular cow Sheep 00000

Bottom material Regular Regular (5% cow skin) 0072Appearance Fine Elegant, patterns 00000

Functionality 6006 6008 0083Workmanship 0093 0095 0026Versatility 00096 0010 0077Cushioning 00096 001 0088Quality 1806 2206 00000

Notes. Panel A reports the IV estimate on the counterfeit entry effect onauthentic quality, estimated by log deflated costs and the sum of various shoecharacteristics, with the interaction of government enforcement change andrelationship proxy as the main IV.

Panel B is an abridged excerpt of summary statistics from Table A.6 ofthe online appendix of Qian (2008). It tabulates the ranksum test statisticsfor the leather shoe characteristics pre and post counterfeit entry. All of thecharacteristics variables are categorical. Surface material includes 14 varieties,ranging from plastic leather to crocodile skin; side material includes sevenvarieties, ranging from plastic leather to baby cow skin; bottom materialincludes five varieties, ranging from inferior polysynthetic leather to cow skin.The Appearance variable is the sum of the dummies indicating whether a pairof shoes is fine, elegant, and with decorative patterns. Functionality is anaggregate variable for various functional attributes, including adroit (a dummyequals 1 if a pair of shoes is adroit), absorption (a dummy equals 1 if a pair ofshoes absorbs sweat), athletic feet (a dummy equals 1 if a pair of shoes helpstreat athletic feet), soft, cushion (a dummy equals 1 if a pair of shoes hascushion effects), comfort, sturdy, warm, and friction (a dummy equals 1 if apair of shoes protects from slippery grounds). Workmanship is a dummy thatequals 1 if a pair of shoes is carefully and finely manufactured. Quality is thesum of all dimensions of shoe characteristics.

Statistically significant at 5% level; statistically significant at 1% level.

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


Table 2 Wilcoxon Ranksum Tests and Regression on Sport Shoe Quality

Panel A

Dependent variable Functionality Appearance Appearance

Entryt2 1043 7055 6035410555 410485 420295

c 1015

400405Company and year FE Y Y YNo. obs. 3,335 3,335 3,335

Panel B

Medium value Ranksum test(1) (2) (3)

Quality dimensions Pre-entry Post-entry p-value

Surface material PU; net Syn. leather; light net 00000

Bottom material MD TPR 00229Air pumps None/middle Top/middle/back 0002

Appearance Fine Fine/elegant 0009

Functionality 12 12Workmanship 1 1Versatility 1 1Supportiveness 1 1Quality 21.3 24.1 00000

Notes. Panel A reports the IV estimates of the counterfeit entry effect onauthentic quality, estimated by log deflated costs and the sum of various shoecharacteristics, with the interaction of government enforcement change andrelationship proxy as the main IV.

Panel B is an abridged excerpt of summary statistics from Table A.7 ofthe online appendix of Qian (2008). It tabulates the ranksum test statis-tics for the authentic sport shoe characteristics pre and post counterfeitentry. All of the characteristics variables are categorical. Surface and sidematerial includes six varieties, ranging from inferior polysynthetic leather toreal leather with materials for nets; bottom material includes four varieties,ranging from thermoplastic polyurethane to special rubber. The Appearancevariable is the sum of the dummies indicating whether a pair of shoes isfine, and elegant. Functionality is an aggregate variable for various functionalattributes, including adroit (a dummy equals 1 if a pair of shoes is adroit),absorption (a dummy equals 1 if a pair of shoes absorbs sweat), athleticfeet (a dummy equals 1 if a pair of shoes helps treat athletic feet), soft,cushion (a dummy equals 1 if a pair of shoes has cushion effects), comfort,sturdy, warm, friction (a dummy equals 1 if a pair of shoes protects fromslippery grounds), lasting (a dummy equals 1 if a pair of shoes lasts a longtime), support (a dummy equals 1 if a pair of shoes supports the ankle well),and flexibility. Workmanship is a dummy that equals 1 if a pair of shoes iscarefully and finely manufactured. Quality is the sum of all dimensions of shoecharacteristics.

Statistically significant at 10% level; statistically significant at 5% level;statistically significant at 1% level.

counterfeiters provide summary statistics of the inno-vation patterns. Results (Panel B in Tables 1 and 2)clearly show that the authentic producers used fanciersurface and side materials and improved the shoeappearance tremendously (especially for leather shoes)after counterfeit entry. The improvements in thesesearchable quality dimensions were not matched alongthe experiential quality dimensions. There was no sig-nificant improvement in the functionality (both at theaggregate level and the detailed-characteristics level):mobility, versatility, cushioning effects, etc. The overall

quality is shown to have improved after entry, withan extremely low p-value in the Wilcoxon ranksumtest. These findings directly support the predictions onsearchable quality innovations in Propositions 3 and 6.

For a more formal test, we generate continuousmeasures of searchable and experiential qualities bysumming the shoe characteristics along relevant dimen-sions of qualities, respectively. We carry out regressionsfor the continuous variable for searchable and expe-riential quality in the leather and sport shoe sectors,controlling for year and brand fixed effects.

Qualitya1t =0 +1Counterfeita1tk+24cc1tca1t5+3YearDt+4BrandDa+a1t1 (2)

Counterfeita1tk=0 +1ISOEnforcea1tk+2ISOa1tk+3Enforcea1tk+4t+a1t1 (3)

where Qualitya1 t refers to quality measures of brand ain year t and takes on two alternative measurements asdetailed in the previous subsection: searchable quality(appearance) and experiential quality (functionality).Subscript k stands for the number of lagged years ittakes for the entry of counterfeits to take full effecton the price outcome. Based on Akaikes informationcriterion (AIC) and Bayesian information criterion(BIC), we select a model with a two-year lag (k= 2).ca1 t is the unit materials cost of the authentic product inthe current year, t. cc1 t is the unit cost of the counterfeitproduct in the current year. The variable 4cc1 t ca1 t5 isthen the empirical measurement of in the theoreticalmodel. To address the potential endogeneity of coun-terfeit entry, we simultaneously estimate Equations (1)and (2) using two-stage least squares.

We find statistically significant coefficients on thesearchable quality, significant at the 1% level (Panel Ain Tables 1 and 2). Although experiential (functional)quality did not experience any significant changes,the searchable quality (appearance dimensions) ofshoes improved by 20% (the median of the appearancedistribution was 12). This provides rigorous empiricalsupport for the key predictions in Proposition 3.

4. ConclusionCounterfeiting is an unavoidable issue for policy mak-ers and business managers operating in emergingmarkets. Although counterfeit entry is a form of compet-itive entry, it differs from other regular competition inthree key ways. First, counterfeiters attempt to imitateauthentic products and free ride on the investments intrademarks and quality by branded companies. Second,counterfeiting unbundles the searchable quality (e.g.,appearance) and experiential quality (e.g., functionality)of a product. Third, counterfeiting has the defining traitof bearing the intent to deceive. Its deceptive nature

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


often gives rise to asymmetric information betweenthe counterfeiters and consumers in transactions. Suchasymmetric information can aggravate competition andinfluence market incentives in nonconventional ways.The media frequently report stories where confusedconsumers are conned into buying counterfeits. Ourmodel is adept to addressing these scenarios.

The main contribution of this paper is to develop atheoretical model that incorporates all of these uniquefeatures of counterfeiting competition and to uncoverthe direction and nature of authentic product innova-tion in response to counterfeiting. Although Nelson(1970) introduces the concepts of searchable and experi-ential goods in economics, we argue that searchableand experiential dimensions of quality can coexistwithin the same good. We propose a novel model thatincludes both of these quality dimensions to analyzethe competitive effects of imitators and counterfeiters.In the theoretical literature, this study contributes byoffering a tractable theoretical framework to cover twodistinct quality dimensions in a vertical differentiationsetting and unveiling the entry effects of counterfeits.By examining the equilibrium conditions and allowingthe authentic producer to endogenously determineexperiential quality and searchable quality as well asprices, we are able to shed new light on the natureof innovations along different quality dimensions asbrand protection strategies. The results therefore haveimmediate managerial implications.

In the case of deceptive counterfeiting, our theorypredicts a noteworthy phenomenon that the marketcan serve as the invisible hand even for regulatingcounterfeits. We first discuss the results for scenarioswhen the brand has a production cost advantage overcounterfeits. When there is a moderate amount ofcounterfeits in the market that confuses a small fractionof consumers, the authentic firm has an incentive tosplit profits with the counterfeiters in a pooling equi-librium. The brand will improve experiential qualityto capture the high-valuation expert consumers if theinnovation cost is not too high. When counterfeitinggrows too rampant and fools too many consumers,the market incentives turn from favoring a poolingequilibrium to favoring a separating equilibrium. Theauthentic brand will invest to innovate and differentiateitself from counterfeiters in terms of searchable quality.In this type of separating equilibrium, counterfeitinginduces the authentic producer to invest more in theproducts searchable quality (e.g., appearance) andless in experiential quality (e.g., the functionality ofthe shoes, not apparent to consumers at the time ofpurchase). This could be socially wasteful if searchablequality does not contribute to consumer utility as muchas experiential quality. We next turn to the scenariowhen counterfeiters enjoy a cost advantage over theauthentic producer. No searchable quality upgrade

is possible here because counterfeiters would readilycopy it at a lower cost. The only hope for the brand tosurvive in the market is to innovate its experientialquality for the expert consumers. Otherwise, the lemonproblem occurs, with counterfeits driving the authenticproducts out of the market.

These predictions are empirically validated by ourfindings based on a sample of authentic and counter-feit shoes for 31 brands operating in China at a timewhen counterfeit entry surged because of a suddenchange in government enforcement policy. In particu-lar, the branded companies that survived counterfeitinfringements all significantly improved their shoessurface, side materials, and appearance, whereas func-tional quality such as sturdiness and flexibility didnot witness any significant changes. These innovationspushed up marginal production costs and hence raisedauthentic prices after counterfeit entry.

The findings bring to light the nature and extent ofproduct differentiation in the face of counterfeiting, incontrast to the generally low-quality competition withperfect information, and provide managerial guidanceas to which product attributes to differentiate on. Thisresearch enriches a Teecean perspective: a substantialportion of the rents from innovation arise not fromtechnological novelty but from embedding innovationin brands (and other searchable quality dimensions)insulated from fringe competition.

This paper is a stepping stone for exploring thecomplex impacts of counterfeits compared to genericcompetition. Although the current analyses shed lighton the key quality and pricing strategies against coun-terfeits, other strategies (e.g., product-line proliferation)and dimensions of asymmetric information and impli-cations can also be associated with counterfeiting. Itwould be a natural next step to extend the model toa dynamic multiperiod model, where one can studythe reputation effects of brands and explore whetherbrands would invest in experiential quality in anticipa-tion of future benefits. We are making further attemptsto better understand counterfeiters decisions to entermarkets and any potential complementary effects coun-terfeits may have on authentic products, in line withthe analytical framework in Chen and Xie (2007).

Supplemental MaterialSupplemental material to this paper is available at http://dx.doi.org/10.1287/mksc.2014.0867.

AcknowledgmentsThe authors are grateful to Philippe Aghion, Eric Anderson,Kevin Bryan, Richard Caves, Anne Coughlan, Josh Lerner,Debu Purohit, Jagmohan Raju, K. Sudhir, and Miguel Villas-Boas for helpful advice and comments and to Maja Kos forcopyediting. The first author also gratefully acknowledgescooperation from the Chinese Quality and Technology Super-vision Bureau (QTSB), the companies she interviewed and

Dow

nloa

ded

from

info

rms.o

rg b

y [1

42.10

3.180

.205]

on 19

June

2015

, at 1

3:59 .

For p

erson

al us

e only

, all r

ights

reserv

ed.


surveyed, and generous funding from the Chinese Na

Documents

mksc%2E2014%2E0867