This action might not be possible to undo. Are you sure you want to continue?

BooksAudiobooksComicsSheet Music### Categories

### Categories

### Categories

Editors' Picks Books

Hand-picked favorites from

our editors

our editors

Editors' Picks Audiobooks

Hand-picked favorites from

our editors

our editors

Editors' Picks Comics

Hand-picked favorites from

our editors

our editors

Editors' Picks Sheet Music

Hand-picked favorites from

our editors

our editors

Top Books

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Audiobooks

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Comics

What's trending, bestsellers,

award-winners & more

award-winners & more

Top Sheet Music

What's trending, bestsellers,

award-winners & more

award-winners & more

Welcome to Scribd! Start your free trial and access books, documents and more.Find out more

Stretching Firm and Brand Reputation Author(s): Luís M. B. Cabral Source: The RAND Journal of Economics, Vol. 31, No. 4 (Winter, 2000), pp. 658-673 Published by: Blackwell Publishing on behalf of The RAND Corporation Stable URL: http://www.jstor.org/stable/2696353 Accessed: 21/05/2010 05:21

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=black. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

The RAND Corporation and Blackwell Publishing are collaborating with JSTOR to digitize, preserve and extend access to The RAND Journal of Economics.

http://www.jstor.org

MIT. and especially Steve Tadelis and Birger Wernerfelt for useful comments and suggestions. 4. Joe Farrell. CentER. Tel Aviv. consumers should expect the quality of a Canon photocopier to be approximately at the level of Canon cameras. Introduction Canon established its reputation as a maker of photographic cameras. No. or should it create a new name (and start a new reputation history)? I show that for a given level of past performance (reputation). should it use the same name as its base product (reputation stretching). by adding a photocopier to its product portfolio.B. using the same name saves part of the cost of launching a new product (Tauber. three referees. Haifa. Mike Katz. IESE. 31. A different approach-the one I follow in this article-is to look at the firm's decision as an informational problem. Consumers observe the performance of the firm's products. seminar participants at QM&W.edu. My purpose in this article is to analyze these tradeoffs in a framework of optimal firm strategy and rational consumer behavior.' Continuing with the same example. then the reputation of the Canon brand suffers. derive its equilibrium. if the photocopier turns out to be of poor quality (or perceived as such). I thank Severin Borenstein. 658-673 Stretching firm and brand reputation Luis M. Naturally. Canon uses its reputation to profit from the sale of the new product. Yale. selling its new offering under the same brand name. 1988). Was this a good move by Canon? One justification for Canon's strategy is that because creating new brand names is costly. David Dranove. PennState." 658 Copyright ( 2000. 1 As the Economist (1994) argues. Pompeu Fabra. Canon enjoys a reputation for product quality as a result of its good track record in selling cameras. a related but different product. Sridhar Moorthy. I develop a simple adverse selection model. Canon risks squandering its reputation. and product performance is positively related to the firm's (privately observed) quality level. Nova. Since what it takes to produce a good camera is similar to what it takes to produce a good photocopier. firms stretch if and only if quality is sufficiently high. Brands offer a route through the confusion. In other words. Duke. Berkeley. Editor Jeniffer Reinganum. Jerusalem. Winter 2000 pp. LSE. and Kellogg. it entered-into the market for photocopiers. lcabral@stem. Cabral* I consider an adverse selection model of product quality. 1. Tore Nilssen. In the mid1970s. They see a huge range of products that look the same and seem to perform similar. Stretching thus signals high quality. Carl Shapiro.nyu. I alone am responsible for remaining errors and shortcomings. In this context. "brands are created because buyers crave information. and characterize the main informational effects involved U * New York University. If a firm is to launch a new product. WZB.RAND Journal of Economics Vol. RAND .

Klein and Leffler (1981). I assume that reputation is associated to the firm's brand name. the model can also be interpreted as depicting the decision of stretching a firm's reputation. In this context. The feedback reputation effect implies that higher-quality firms are more confident that stretching will consolidate their reputation. Some correspond to existing empirical work. In this case. the strategy of brand stretching in this context is explored in Choi (1998). the decision would be to sell a second product versus creating a new firm to sell the new product. Finally. Strictly speaking. In addition to these basic results. the firms that stretch their reputation are the firms of higher quality level.3 When the firm's actions are not observable (moral hazard). the new product is of marginal importance with respect to the base product. firms prefer to "protect" their reputation.4 2 Throughout most of the article. See Pepall and Richards (2000) for an analysis of brand stretching in this context. so for a given level of quality. some to possible statistical tests that relate reputation. ( RAND 2000. First. I consider three effects in the decision to stretch a reputation. If. I call this the signalling effect. I consider a number of possible extensions of the model. 3In this article. An alternative approach is to assume that brands have an intrinsic value. a firm's reputation from its base product influences the consumers' willingness to pay for a new product sold under the same name (and for additional sales of the base product). Second. and Shapiro (1983). in an attempt to "build" their reputation. Consumers observe the performance of the firm's products. stretching takes place if and only if reputation is sufficiently high. Other articles that take the moral-hazard approach to explain brand stretching include Andersson (1998) and Cabral (2000). The direct reputation effect is of ambiguous sign. Choi's (1998) model features both moral hazard and adverse selection. Finally. and the decision to stretch. reputation corresponds to an implicit contract between seller and buyers whereby the former supplies high-quality experience goods and the latter pay a high price-until the seller cheats on buyers and reputation breaks down. As a result. the performance of the new product.2 In the model. I only consider brands as vehicles of information. then firms will "exploit" their reputation. I will return to this work later. which is related to the firm's quality level. . If the new product is very profitable compared to the base product. I derive empirical implications of the model's assumptions and results. when reputation is high. insofar as the decision to stretch is related to the firm's quality. in the context of moral hazard. reputation is the consumers' posterior on the firm's quality level given the firm's performance history. however. by contrast. I call this the feedback reputation effect. each firm is characterized by an exogenously given quality level. then firms will only stretch if reputation is sufficiently low. if sold under the same name. the model is one of brand stretching. the simple fact that a firm uses the same name to launch a second product influences the consumers' willingness to pay for the new product (and for additional sales of the base product). including the endogenous decision of whether to launch a new product and the choice of which product to stretch to. which is the firm's private information and applies to any product it supplies. Finally. influences the consumers' willingness to pay for future sales of the base product. I call this the direct reputation effect. for a given reputation level. 4 Actually.CABRAL / 659 in the decision of stretching a reputation. More on this later. performance. However. Brand names can be vehicles of reputation in two ways: they may incorporate information about the firm's actions or information about firm characteristics. Kreps (1990) put forward the idea of the firm's name (or the firm's brand name) as the carrier of reputation. The economic analysis (and the economics literature) on reputation stretching can be divided according to the nature of the branding effects it is based upon. The seminal economics articles on this approach are Telser (1980).

product performance is i. a fact that is common knowledge to firm and consumers. then most profits are coming from the new product. my qualitative results would still hold if I assumed a small. Conditional on quality. The firm's second product (if it exists) is of the same quality as the first one. a production batch). In Sections 4 and 5. In this context. If 8 is close to one. Such beliefs are updated based on the observation of the firm's past performance as well as on other actions by the firm that might signal its private information. the strategy of brand stretching in an adverse selection context was proposed by Wernerfelt (1988). though I look at a different set of issues: whereas Tadelis examines the equilibrium in the market for names. 7The same qualitative results go through if the correlation between quality levels is positive and sufficiently high.d.g.660 / THE RAND JOURNAL OF ECONOMICS An alternative information-based approach to brand effects is that of adverse selection. product performance is public information. a continuum of units that perform equally (e. See Section 7.i. 1]). The basic framework for the analysis of this type of reputation was laid down in Milgrom and Roberts (1982) and Kreps and Wilson (1982). which only the firm can observe. 8 The model can be interpreted either as one where firms sell one unit per period or. alternatively. reputation corresponds to the consumers' posterior beliefs regarding firm quality.8 units of the basic product at ages 0 and 2. though the approach I take is somewhat different: in Wernerfelt. The firm sells 1 . which measures the relative importance of the new product with respect to the base product.. Each firm is endowed with quality q. 2. I consider the case when names are not traded but can be used to sell more than one product. 2). r]. The problem I consider is the same as in Wernerfelt (1988). Specifically. I develop a model that shares some of the features of Tadelis' (1999) basic framework. if 8 is close to zero. In each period. Suppose that quality is a firm attribute that consumers cannot observe ex ante. I present the basic model and derive its Bayesian equilibria.e. C) RAND 2000. 1. across periods. then most profits come from the base product. brand stretching signals quality because stretching is more costly than creating a new name. q]. In Sections 2 and 3. is related to its quality. rt E [r. of which it sells 8 units (8 E [0. . strictly positive probability of new product development. who considers a model where names are traded. 6 The assumption that only a countable set of firms develop a second product greatly simplifies the analysis.. a fixed measure of new firms is born. whereas I assume that brand stretching is cost neutral.7 Notice that payoffs are paramaterized by the value of 8. assuming that each product is sold under a different name.8 Consumers derive utility u(r) from the product's performance. Each firm is endowed with a basic product and lives for three periods (ages 0. I discuss the equilibrium stretching strategy and analyze the particular case when utility is linear and both quality and performance are normally distributed. according to the cdf F(r Iq) (density f).6 The firm's main decision is whether to sell its second product under the same name as the first one or under a different name. However. but this would unnecessarily make the model more cumbersome. A product's performance in period t. The empirical implications of the model are discussed in Section 6. Moreover. q E [q. where u(r) is strictly increasing 5 Icould also assume sales of the base product at age 1. My approach is one of adverse selection. The article is structured as follows. The idea of names as carriers of reputation (in the context of adverse selection) was explored by Tadelis (1999). All proofs are in the Appendix.5 A countable (i. Model * I consider an economy with overlapping generations of firms. Finally. measure zero) set of firms is endowed with a second product at age 1. Section 7 suggests a number of extensions of the basic framework.

Consumers do not directly observe q. it implies that v(q) is increasing. ro is a sufficient indicator for the firm's reputation level. I denote by x(q.10 As mentioned above.12 w(fI) is also firm profit per unit in each period. This assumption implies that higher-q firms produce better products. 12 The assumption that consumers bid for each firm's output allows me to set aside the question of signalling through prices and instead focus on the firm's decision of naming its second product. I assume that consumers do not observe the ownership of each brand. Consumers therefore cannot distinguish between a new product launched by a new firm and a new product launched by an old firm. the assumption implies that higher performance is "good news" regarding the firm's quality (Milgrom. The consumers' information set includes past product performance and the observation that the firm has stretched its reputation. the better the posterior distribution of q. fl = {rO. then [f(r Iq")]I[f(r Iq')] is strictly increasing in r. ? RAND 2000. they form posterior beliefs H(q If) (density h). as well as the assumption of perfect correlation of product quality. ro) the probability that the firm stretches its name to the new product. however. the consumers' willingness to pay. in case it stretches. and il = {rO} at the beginning of the firm's third year.CABRAL / 661 u(r) dF(r Iq) is the consumers' willingness to pay for a in r. 1981). fl should explicitly include the information that the firm did not stretch. the items under a = 1 do not apply and the analysis is trivial. However. the firm's strategy consists of whether or not to stretch given that it is endowed with a new product. } at the beginning of a firm's third year. the information that the firm does not stretch is irrelevant. if q" > q'. in equilibrium x(q. as such firm has no decisions to make. given my assumption that the measure of firms that develop a second product is zero. It follows that each unit is sold for w(Q). 10Strictly speaking. For this reason. the higher ro. that is. consumers would be willing to pay more for a higher-q firm product. It follows that v(q) = product of quality q. They hold a (correct) prior G(q) (density g) and. result. based on information Q. f(r Iq) > 0 for all q. r.'1 Finally. I will below refer to ro as the firm's reputation level. That is. consumers are willing to pay up to w(f) = Iq v(q) dH(q If).9 Specifically. I assume the market is short on the sellers' side and that consumers bid for each firm's output. f . that is.s} for a firm that stretches and at the time it does so. ro) is either zero or one. The timing of the model.s. if that is the case. It follows that at each information set Q. Formally. from the point of view of a firm that is endowed with two products. Moreover. il = 0 for a new product. This decision is based on the firm's quality level q (which is the firm's private information) as well as on the first-period product performance ro (which is public information). 11As we will see. in the sense of first-order stochastic dominance. since I assume production costs to be zero. but pure strategies are a derived. in case it does not stretch. In other words. 13 For a firm not endowed with a second product. Finally. il = {rO. notice that in the first period. The family of densities {f(. is summarized in Table 1. r.|q) } has the strict monotone likelihood ratio property (SMLRP). not assumed.13 The model's notation is summarized in Table 2. 9Later I relax this assumption. I make the following assumption regarding the function F(r Iq): Assumption 1.

s. ro). A Bayesian equilibrium (BE) is a pair (x(q. rj) dF(r Iq) (1) N(ro) = Sw(0) + (1 8)w(rO)- where. Firm sells second product under the same name with probability x(q. To determine the firm's optimal strategy. In words. The equilibrium concept I will use is that of Bayesian equilibrium: U Definition 1. as well as the main features of each equilibrium. Consumers buy one unit of first product. r) Ql F(r Iq) G(q) H(q IfQ) Posterior distribution of quality (density h) ?) RAND 2000. Consumers buy one unit of second product and observe performance r . . s) + (1-8) w(ro. ro) = Sw(ro. and (ii) H(q | Q) is Bayesian consistent given x(q. ro) =S(q. ro) - N(rO) S(q. H(qjfl)) such that (i) x(q. ro). as an abuse of notation. ro) is optimal given H(q | Q). s}. TABLE 2 q rt s 8 Notation Firm quality Product performance in period t Event of stretching Relative size of new product (with respect to base product) Consumer utility given performance Consumer indirect expected utility given quality Consumer willingness to pay given belief H about quality Probability of stretching Consumer's information set Probability distribution of performance given quality (density f) Prior distribution of quality (density g) u(r) v(q) w(H) x(q. drawn from F(r1 | q). s) denotes w(fl) where il = {ro. w(ro. drawn from F (r0 Iq). and so forth.662 / THE RAND JOURNAL Timing OF ECONOMICS TABLE 1 a = 0 a = 1 a = 2 Firm is born and endowed with quality q. N(ro) is the profit from selling a new product under a new name. I need to compute the expected value from launching the new product under the same name as the first product. We then have M(q. 3. Sw(0). Consumers buy one unit and observe performance ro. The difference between the former and the latter is the marginal payoff from stretching. S(q. ro). ro). M(q. and compare it to the payoff from using a new name.8)w(ro). (1 . Equilibria In this section I derive the set of equilibria in the model. N(ro). plus the profit from selling the base product for the second time. ro).

When there are zeroprobability events. then we can support a pooling equilibrium. . if consumers observe a product with no history. The implication is that in equilibrium it is the firms with higher q that stretch. s) = f f(rOIq)x(4. In the present case. I show that there exists a pooling equilibrium whereby firms never stretch. Proposition I (pooling equilibrium). Next I consider an equilibrium where there is some separation. s). some of which (a measure zero) are endowed with a second product. ro) h(q ro. but this is a measure zero event. they assume this is a new product. note that although I focus on a particular firm. The probability of stretching is therefore increasing in q. So. rj) = g(q)f(rO jq)f(rj | q)x(q. hence h(q 10) = g(q). Bayesian equilibria are consistent with any kind of beliefs. Proposition 2 (semiseparating equilibrium). a new product could also be launched by an old firm. In fact. ro) = 0. This implies that h(qjO) = g(q) h(q Iro) = g(q)f(ro Iq) F q f(ro Iq) dG(4) (2) g(q)f(rO j q)x(q. s. First. M(q. a firm stretches with probability one if and only if q > q*(ro). We can therefore associate such an event with the belief that the firm is of the lowest quality. rj) dF(r Iq). and does not stretch otherwise. stretching is a zero-probability event. There exists a pooling equilibrium whereby no firm ever sells a second product under the same name: x(q. If we make those beliefs sufficiently unfavorable to the first player. ro) dG(q) h(q I ro. for example. To obtain a BE. is increasing in q. ro) dG(4) To understand these expressions. ? RAND 2000. Proposition 1 is a common result in signalling games. In this equilibrium. ro) is the profit from selling the second product under the base product's name. s. thus making "no stretching" an equilibrium. the model assumes overlapping generations of continuums of firms.CABRAL / 663 S(q. Vq. (1 - 8) jw w(rO. The thrust of the proof of Proposition 2 is to show that the marginal payoff from stretching. Then I show there exists a unique semiseparating equilibrium (in the sense that no complete pooling takes place for any value of ro). I impose that H(q Ifl) be consistent with Bayes' theorem. Sw(ro.14 The first results in this section are the following. The reason for this result is that the only part of the 14 See Section 7 for a discussion of how to relax this assumption. plus the expected profit from selling the base product for the second time. ro) f(ro I )f(r 1q)x(q. There exists a unique equilibrium such that for all ro. ro). rO. the probability of stretching is strictly positive.

If q*(ro) is decreasing. It follows that at each round of the iterated elimination process. firms stretch if and only if ro is greater than the inverse of q*(ro). However. however. In other words. (i) F(r Iq) = N(q. only the performance levels ro and rl. then firms stretch when their reputation is high. 15 It is also the only equilibrium that survives the Fudenberg and Tirole (1991) version of D1. u(r) is linear. In the remainder of the article I shall focus on the semiseparating equilibrium of Proposition 2. (ii) G(q) = N(O. for what realizations of ro should a firm stretch? This is a more complicated question: higher levels of ro imply that the firm can get more from selling a new product with the same name as its base product. then for a given value of q. 8. there actually exists a continuum of equilibria.15 To be more rigorous. The direct reputation effect then implies that higher reputation leads to higher benefits from stretching. is the following: Given a value of q. though it is generally not the unique divine equilibrium. higher-quality firms expect higher values of r1. o>r). For each ro. Formally. Assumption 1 implies such payoff is increasing in q. I assume that the ex ante expected value of q is zero. With no further loss of generality. Assumption 2. in terms of characterizing the optimal firm strategy. Assumption 3. In other words. the equilibrium of Proposition 2 is the only universally divine equilibrium in a discretized version of the game where there is a finite number of types (which. increase the consumers' willingness to pay. provides an answer to the question above: for a given ro. ? RAND 2000. Proposition 3 (reputation exploiting). it is the firms with higher q that stretch. This is the only universally divine equilibrium." Proposition 2 provides an answer to the following question: Given a value of ro.664 / THE RAND JOURNAL OF ECONOMICS payoff function that depends on q is expected future sales at age 2 in case of stretching. This in turn eliminates any equilibrium where all types pool at "no stretching. then q*(ro) is decreasing. the payoff from stretching is increasing in q (see the proof of Proposition 2). This effect. there exists a 8 such that if 8> then q*(ro) is decreasing. 4. The linear-normal case * To get a better idea of the nature of the equilibrium. but it also means that the firm has more to lose from squandering its reputation. for each value of ro I can select as equilibrium strategies and beliefs either those in Proposition 1 or those in Proposition 2. It follows that the payoff from stretching is increasing in q. the only equilibrium where stretching takes place with positive probability for all values of ro. which I call the feed-back reputation effect. Uq). Propositions 1 and 2 show that there exist two different equilibria. the direct reputation effect may cut both ways. By Assumption 1. I now consider the particular case when utility u(r) is linear and both q and rt are normally distributed. the D2 criterion eliminates the lowest surviving type. how does the optimal strategy depend on the value of q? Consumers do not observe q. . In fact. This is fairly intuitive: a high value of 8 effectively implies that the firm is only concerned with the sales of its new product. If 8 is close to one. the optimal strategy is one of reputation exploiting: use your brand name if and only if it has value. The reason is that for any consumer belief. This leads to a continuum of possible combinations between the two. higher values of r. The complementary question. My next result states that if the new product is relatively profitable. can be arbitrarily large).

CABRAL / FIGURE 1 (Trq WHENFAND G ARE NORMAL = EQUILIBRIUM = 665 1) q \q**(r0) . (iii) it follows that stretching is optimal if and only if profit at age 2 in the case of stretching is positive. 16 . q**(ro). The first thing to notice from Figure 1 is that q**(ro) is decreasing. age 2 profit is proportional to ro + rl.17 As shown in the proof of Proposition 2. by assumption. . (iv) since. the equilibrium strategy consists of a critical value q*(ro) above which the firm stretches. corresponding to the case when consumers do not include s in their updating of h(q). the value q**(ro) corresponds to the case when consumers rationally process the information on performance (ro) but not the information provided by the firm's choice of whether or not to stretch (s). is ro + q. 8 = ?/2. Figure 1 depicts the equilibrium value q*(ro) in the (ro. l . . Straightforward computation shows that Assumption 3 is consistent with Assumption 1. 17 Details ? RAND 2000. of the computation are available upon request. and (v) the firm's expected value of ro + r. Normality of the prior and the performance distributions allows for a simple solution to the updating problem. ro) = I f q f(ro Ii) dG(4) f q f(ro I4)x(4. that is. note that (i) profit at age 1 in the case of stretching is equal to profit at age 2 in the case of no stretching (in both cases. so that h(q) = g (q)f(ro Iq) and not h(q) g(q)f(r0 q)x(q. I . Some of the details are included in the proof of Proposition 4 below. then q**(ro) = -ro. consumers observe ro).16 Figure 1 illustrates the equilibrium of Proposition 2 in the linear-normal example and for the parameter values o-q= o-r = 1. consumers do not process the information from the observation of s. the figure also depicts the critical value. (ii) profit at age 1 in the case of no stretching is zero (the profit of a new product launched under a new name). if 8 = 1/2. Specifically. To understand this fact. To get a better grasp of the effects at work. q) map. r> -4 -3 -2 -1 0 1 2 Linearity of the utility function implies that consumers only care about expected performance. ro) dG(4) as is the case in a Bayesian equilibrium.

consumers would be willing to pay a high price for the firm's output. the probability of stretching goes to one. As suggested by Proposition 4. for a given reputation level. q*(ro) is increasing if 8 is small enough: Proposition 4 (reputation building).g. then q*(ro) is increasing.accordingly. there exists a 8 such that if 8 < 8. firms stretch if and only if ro is greater than the inverse of q*(ro). ? RAND 2000. that is. firms with higher quality decide to stretch (Proposition 2). In fact. 18 An implicit assumption of my analysis is that firms cannot change the name of their existing product. the incumbent's true ability (e. Even an average-quality firm expects that its reputation will improve starting from a low reputation level: things can only get better. As a result. 5. and consequently the signalling effect becomes less and less important. then we get an equilibrium threshold q*(ro) that falls below q**(ro): the signalling effect increases the likelihood of stretching. If we factor in the signalling effect. a firm of quality lower than q** (but not much lower) would have a big incentive to stretch. As suggested by Proposition 3.18 That is. At the equilibrium threshold q* (ro) and for a given ro. 19Dranove and Tan (1990) show that information spillovers may lead an incumbent to be overconservative in its decision to enter a new market: by doing so. More generally. stretching implies a lower payoff than starting a new name. q*(ro) . As the level of reputation increases. For each ro. In the short run. then consumers should expect such a firm to be of very high quality (q > -ro in the case when 8 = 1/2). the firm follows a strategy of reputation protection. 1/2. in contrast with the large-8 case.666 / THE RAND JOURNAL OF ECONOMICS q > q**(ro) cannot be an equilibrium. . Discussion * As suggested in the previous sections. then the firm stretches only if its quality is very high. What about the stretching strategy for a given q? Proposition 3 suggests that if 8 is high enough. then the optimal strategy is to stretch if ro is very low or very high. quality) may be revealed and induce entry into the base market. the firm's strategy for low 8 is one of reputation building. it is the firms with higher q that stretch (Proposition 2). This is the essence of the signalling effect: the mere fact that a firm stretches improves the consumers' assessment of quality. To put it differently: If 8 is low. The feedback reputation effect implies that higher-quality firms are more confident that stretching will consolidate their reputation. firms stretch if and only if ro is lower than the inverse of q*(ro). then it is the firms with highest ro that stretch.19 Figure 2 illustrates Propositions 3 and 4: it depicts the equilibrium threshold q*(ro) for three different values of 8: 0. but in the long run. for 8 = 1/2 and for low values of ro. 8 = 1 implies that q*(ro) is decreasing: for a given q. however. In fact.. The idea of Proposition 4 is that if future sales of the base product are sufficiently profitable with respect to the stretched product. if 8 = 1/2and q is one standard deviation below average.q**(ro) tends to zero. As a result. Figure 1 shows that this is not a general result.and 1. 8 = 0 implies that q*(ro) is increasing: for a given q. then firms should stretch if and only if their reputation is very low. the optimal strategy may be more complicated. all I need to assume is that consumers are able to identify the product even if the firm changes its name. in effect "free riding" on the highquality firms' stretching strategy. there are essentially three effects at work in the decision of stretching a reputation. such a strategy pays off. For example. q*(ro) is increasing. If that were the case and a finn with low rowere to stretch. For intermediate values of 8. The intuition is that as the level of reputation increases.

whereas Wernerfelt's does not (see Wernerfelt. 1986). when reputation is high. He considers the case when the seller offers a series of products under the same brand name. . An alterative signal- ling model of reputation stretching is given by Wernerfelt (1988). however. I assume that brand stretching is cost neutral. then firms will stretch only if reputation is sufficiently low. that is. Choi's (1998) theory of brand extension (still another term for the same practice) features both adverse selection and moral hazard. This extra cost allows for a separating equilibrium where a firm umbrella brands if and only if it is a good type (high quality). so that for a given level of quality. Each new product is launched at a price that signals its quality. the mere fact that a firm stretches signals that its quality is relatively high. in an attempt to "build" their reputation. then the price distortion necessary to signal quality is lower. the cost of launching a new product is the same regardless of the name the firm chooses for its new product. My model differs from Wernerfelt in several respects. by contrast. but Wernerfelt shows it to be the only equilibrium that survives the Cho-Kreps intuitive criterion. In particular. firms prefer to "protect" their reputation (Proposition 4). He assumes that umbrella branding (an alternative term for brand stretching) is more costly than creating a new brand. his result depends on a positive cost of umbrella branding. because it is the firms of higher quality that stretch in equilibrium. stretching takes place if and only if reputation is sufficiently high (Proposition 3). Another important difference is that my result depends (crucially) on a positive correlation of qualities across a firm's product offerings. By contrast. as is the case with brand-extended products. Finally. If the new product is relatively profitable compared to the base product.CABRAL FIGURE 2 EQUILIBRIUM THRESHOLD q*(r0)AS A FUNCTION OF 8. This is one of many Bayesian equilibria. If the new product is of marginal importance. If consumers later find out that the product is of low quality. If consumers expect the product to be of high quality. then the firm's reputation breaks down and future high-quality launches will need to be signalled ? RAND 2000. a Comparison with alternative theories of brand stretching. THE RELATIVEWEIGHT OF THE 0-= 1) q / 667 STRETCHED PRODUCT = (Trq 5=0 31-r -2 -1 0 2~~- 1 The direct reputation effect implies that firms with higher reputation are able to sell for a higher price than firms with lower reputation. then firms will "exploit" their reputation.

The empirical evidence is consistent with a positive correlation between stretching and performance (see Court. then P(s Iro) is unambiguously increasing in ro. Empirical evidence and empirical implications * Empirical evidence seems to support the model's prediction that bad news in one product leads consumers to revise their expectations about the quality of other products sold under the same name. firms that stretch end up doing better. toothpaste and toothbrushes. the probability of stretching conditional on ro. Although Choi's model features adverse selection and signalling. in the linear-normal example. as is the case when 8 is small. not causality: it is not the case that by stretching firms perform better. 1999). is given by Jil(r f(ro Iq) (ro) dG(q) (3) P(s ro) f f(ro Ia) dG(q) P(sojr) is increasing in ro (by Assumption 1) and decreasing in q*(ro). P(s Iro) is increasing for all values of 8. In fact. For example. the strategy of brand extension is supported by a moral-hazard equilibrium in the tradition of the Klein and Leffler (1981) and Shapiro (1983) models. which plots C) RAND 2000. then it is conceivable that P(s Iro) is decreasing in ro. it is important to notice that this is a case of correlation. 6. while performance is also increasing in quality.668 / THE RAND JOURNAL OF ECONOMICS with a large price distortion. then the likelihood of stretching is increasing in reputation. Another implication is that if the stretched product is relatively profitable (high 8). . This is not an obvious result. Proposition 2 also has implications for reputation. However. This is illustrated by Figure 3. First. stretching. the average performance of firms that stretch is greater than the average performance of firms that do not. if q*(ro) is increasing. Leiter. but less so for lower values of 8. Specifically. P(s Iro). This is so because the probability of stretching is increasing in quality. but not generally. Jarrell and Peltzman (1985) find that a dangerous drug recall by the Food and Drug Administration lowers the manufacturer's stock price by more than can be attributed to the lost profits from the recalled drug. However. Erdem (1998) estimates positive (if small in magnitude) cross-category effects between two oral hygiene products. Too often one hears the argument that firms should stretch because it improves their performance. as is the case when 8 is high enough (Proposition 3). As it turns out. If q*(ro) is decreasing. 2 q qj*(ro) P(s Iro) = 1 -a (D i2 q + ca2 -~~~ro where (F is the standardized-normal cdf. rather. and performance. Sullivan (1990) shows that the 1986 sudden-acceleration incident with the Audi 5000 shrunk the demand for the Audi 5000 and the Audi Quattro as well. and Loch. My results show that this is true for high-quality firms.

ro)] dF(ro Iq). Specifically. Extensions * The model I have presented is very simple in many respects. Suppose. I believe that the main results are robust and would hold in more general settings as well. I also consider three possible extensions of my basic framework. In what follows. I have assumed that these firms form a set of measure zero. if consumers observe a new product with a new name. However. The equations for the posterior distribution of q are now more complicated. I specifically consider the robustness of the model with respect to the assumption that the set of firms developing a new product has measure zero.CABRAL I 669 FIGURE 3 PROBABILITYOF STRETCHING CONDITIONALON REPUTATION. For example. I have considered an overlapping-generations model of three-period-lived firms. Firms are endowed with a product in the first period. however. We thus have. Also. instead of (2). they must consider the possibility that this product is being launched by a new firm or by an old firm that decided to create a new name. 7. I could have considered more than three periods. (y(q) is the probability that a firm of quality q does not stretch. The figure suggests that a regression with P(s Iro) as a dependent variable would show a positive coefficient in the ro08interaction term. FOR DIFFERENT VALUES OF 8 1) P(sIre) (0q9 = at = 8I=0 '5~ ~~~I -4 -3 ~ -2 ~~= I I _ _ I _ _ I _ > _ _ _ -1 0 1 2 P(s ro) for three values of 8. instead of assigning all of the market power to the seller. I could have assumed some sharing with buyers. Some firms are also endowed with a second product in the second period. h(q|0) = g(q)[1 + ay(q)] f [1 + ay(q)] dG(q) where y(q) = [1. f Positive probability of new product development.x(q. that independently of q. a firm is endowed with a second product with probability a.) 0 RAND 2000. .

the same result would follow for small values of a. but then decides whether to launch it (under the same name as the first product) or not to launch it at all. rO. Recall that in equilibrium and for the marginal type q*(ro). instead of (2). s. ro)] f(ro. To compensate for this. Moreover. q) dG(q) q)[1 - x(q. I conjecture that the lower the degree of correlation between products. the result is robust to small perturbations in the value of a. the curve q**(ro) in Figure 1 corresponds to the condition S(q**(ro). the greater the probability of stretching. when consumers make a second purchase from a firm that did not stretch.670 / THE RAND JOURNAL OF ECONOMICS Likewise. having developed one. the positive signalling effect exactly balances a negative reputation effect. that is. decided to sell it under a new name. the higher the threshold q*(ro) above which firms stretch.p) . O Imperfect correlation. the greater the degree of correlation. Does the above conjecture imply that we should observe firms stretching mainly into unrelated product lines? Not necessarily. The above conjecture is for the comparative statics with respect to an exogenous change in p. (That is. It seems reasonable to assume that (a/ap)lS(q. Formally. the greater the correlation between q and q'. the higher p is. this model is 20 The exact statement would then be: for a given ro. Along the same lines. ro)] dG(q) (The expressions for h(q jro.2O Endogenous choice of correlation. r. then it has less to choose from. the proof of Proposition 2 is based on strict monotonicity with respect to q. Let S(q.) Notice that the above expressions are continuous in a at a = 0. ro) be the expected payoff from stretching given that consumers do not process the information given by the signal that the firm stretched. the more important the ro news. it seems reasonable to assume that the cost of stretching is increasing in the degree of correlation: if the firm wants to choose a product that is very closely correlated to its current offering.N(r0)j > 0. Consequently. as in the model above. ?) RANDw2000. g (q)f aY) r+ f Jq h (q Iro) = (1 - (ro Iq) + a Jq g(q)f(rOlq)[1 f(ro. Suppose that p is common knowledge to firm and consumers.) remain unchanged. they must consider the possibility that the firm did not develop a second product or.x(q. A greater degree of correlation implies that the reputation effect is more negative. and that consumers can never observe q or q'. one could also endogenize the decision of launching a new product. In words. which implies a greater O q*. A tantalizing possibility is that there exists a separating equilibrium whereby the higher the quality and/or reputation. . The argument runs as follows. O Endogenous choice of launching new product. the signalling effect must also be more positive. Also. ro) = N(ro). that the firm can only observe q' after it decides whether or not to stretch. both good news and bad news.) Let q' be the quality of the second product and p an indicator of the correlation between q and q'. . But it probably makes more sense to think of p as the result of an endogenous choice by the firm. Suppose that a firm develops a new product. We thus have. One interesting question is how the equilibrium strategy depends on the degree of correlation between the quality of the base product and the quality of the second product. s) and h(q jro.

From (1). In fact. This induces a posterior distribution H that is dominated by both G and H(q Iro). Finally. s) and w(0). if w(0) = 0.e. s) and H(q Iro. the fact that M(q. Q. M(q. in the sense of first-order stochastic dominance. r). High values of r.CABRAL / 671 isomorphic to the model in Section 2. Lemma Al.D.. the result follows. In fact. In fact. the proof of which is obtained by straightforward differentiation. ro) > 0 if and only if q = q. For 8 large enough. Proof of Proposition 1. Since M(q. r) be the marginal benefit from stretching given that consumers hold beliefs corresponding to q*(ro) and q*(ro). But then we have a contradiction: 0 ML(q*(ro). q*(ro) < q. the comparison between stretching and not stretching corresponds to the comparison between w(ro.E. ro) is strictly increasing in q. if and only if x' > x". then we get the same model. r) < MH(q*(rO). f(ro Iq) > 0. Finally. through the second-period payoff in the case of stretching.E. r) < MH(q. Suppose that 5(x Ix) = tr(x)/j tfr(x)dx (x > x). Since w increases when H increases in the sense of first-order stochastic dominance. where the last inequality follows from monotonicity of M(q. we only have to show that the probability of stretching is strictly positive. r) with respect to q. that is. we see that the marginal payoff from stretching only depends on r. Then l(xIx') dominates 5(x Ix"). ro) is increasing in q and the belief q*(ro). for all q and ro. high values of r.r) < MH(q*(rO). suppose there exist two equilibria identified by the functions q*(ro) and q*(ro). Appendix * The proofs of Propositions 1-4 follow. and a theory of how reputation affects the firm's expansion strategy rather than its product naming strategy. Consider the following result.q. we can assign any posterior belief upon the event of stretching. Proof of Proposition 3. it follows that the marginal firm's payoff is positive for sufficiently high q*(ro). It follows that the realized marginal payoff from stretching is strictly increasing in rl. r) and MH(q. ro) = w(rOs) = T f(ro |q) dG(4) N(ro) = w(0) f v(q) dG(q)- The indifferent type q*(ro) is therefore determined by Q) RAND 2000. respectively. In the limit when 8 = 1. The alternative interpretation of the model is interesting because it dispenses with the assumption of unobservability of brand ownership. both H(q Iro. We would then have a model of firm reputation rather than brand reputation. . rj) dominate H(q I0) and H(q Iro) (in the sense of first-order stochastic dominance). the firm's unit profit at age 1 depending on whether it stretches or does not. Also. r0)= if q > q*(rO) O otherwise. in the limit as q*(ro) -. are good news. Let that belief be that x(q. we have f(ro I q)v(q) dG(q) S(q. respectively.D. i. Since in the proposed equilibrium stretching occurs with probability zero and. let ML(q. ro) does not depend on q. as shown above.r) = 0. where q*(ro) < q*(ro). Assumption 1 then implies that M(q. Proof of Proposition 2. where 5 and fr are densities and x E Rl. This implies that ML(q. shift the posterior distribution of q in the direction of first-order stochastic dominance. s. Q. To show that there exists a unique q*(ro). ro) is strictly increasing in q implies that the equilibrium strategy must be f1 x(q. other than through rl.

CHOI. J.. By the same argument as in the proof of Proposition 2. 332. Vol. March 2000.q) Assumption 1 implies that v(q) is strictly increasing. "An Empirical Analysis of Umbrella Branding. LEITER. the condition for the indifferent type q*(ro) is given by fw(r.B. AND WILSON. CABRAL. New York University.. Q." Review of Economic Studies. respectively). the left-hand side is deceasing in ro. D. ? RAND 2000. COURT. 253-279. I will now argue that under Assumptions 2 and 3. Cambridge. Mass. Vol. 89 (1981). 4(x) and (D(x) are the standardized normal density and cdf. AND LOCH. is normal + -2). 1994. 1999. Incumbency. the left-hand side is strictly increasing in q*. S. pp. 575-585.. "The Role of Market Forces in Assuring Contractual Performance." Mimeo. where Z(x) = 0(x)/[l . References ANDERSSON. D. D. the posterior distribution of q given signals ro.P "Brand Extension and Informational Leverage. T. Assumption 1 and the fact that v(q) is strictly increasing imply that the left-hand side is strictly increasing in ro.u + o-2Z[(q* .: MIT Press. No. AND LEFFLER.D. 655-669. pp." International Journal of Industrial Organization.asp." In J. Assumption 3 (normality) implies that the posterior distribution of q given a signal ro is normal with mean oq2r/(oq2 + o-2) and variance o-202/-q2 + o-2). 615-641. "Reputation and Imperfect Information. AND TAN.E. Assumption 2 (linearity) implies that with mean o'2(ro + rl)/(2o-2 + o-2) and variance 0o20-2/(2o-2 consumers only care about the expected value of q. Available at http://205. Vol. Lund University. 655-669. July 2. 100-110. ERDEM.A. "Corporate Culture and Economic Theory. 65 (1998)." Journal of Marketing Research. pp. The result then follows from the implicit function theorem and continuity in 8. AND PELTZMAN. Vol. "The Impact of Product Recalls on the Wealth of Sellers. 512-536." McKinsey Quarterly. eds. 27 -~ (1982). Shepsle. Vol.123/marketin/brle99. pp.253.. Proof of Proposition 4. pp. Lemma Al and the fact that v(q) is strictly increasing imply that the left-hand side is strictly increasing in q*. Perspectives on Positive Political Economy. "Brand Leverage. Moreover. T f(ro Iq) dG(. rj) dF(r1 Iq*) - w(rO) = 0." Vol. Vol." Mimeo. . M.G. KLEIN. "Optimal Brand Umbrella Size.M." Journal of Economic Theory. T.D. then the expected value of q given that q > q* is equal to . DRANOVE. ECONOMIST. AND TIROLE. 8 (1990). Q. L. pp. 1990. it is known that if q N(.. Alt and K. 2. and Conservativism.J. these facts imply that the above equation can be rewritten as (F2 22 +~~~~~-~ * q*- O'q 2o-2 + q (ro0q*)+ 2(ro r + q 2)+ r 2 + o2 q rJ- 2 2 Oq --or ( ro+ r1) dF(rq*) - Oq q r +*r2 r = 0.D." Journal of Political Economy.C. the left-hand side is strictly decreasing in ro. Finally. August 1998. F "The Firm as a Pool of Reputations. B. "Don't Get Left on the Shelf: Why Brands Will Survive. 93 (1985). G. R. 1991. Game Theory. JARRELL.1(x)] is the standardized-normal hazard rate (i.128. New York: Cambridge University Press.)/lo-. Z(x) is strictly increasing. FUDENBERG. In the limit when 8 = 0." Journal of Political Economy.E. K. The result then follows from the implicit function theorem and continuity in 8.e. pp.672 / THE RAND JOURNAL OF ECONOMICS f f(ro I q)v(q) __________________ dG(q) -J f v(q) dG(q) = 0. o-). Together. "Information Spillovers. r. 2of + o'2 Straightforward computation shows that since Z(x) is strictly increasing. s. M. 65 (1998). KREPs.

Vol. Vol. "Good News and Bad News: Representation Theorems and Applications. 380-391. Vol. PEPALL. 98 (1983). and Entry Deterrence. TADELIS. 280-312." Journal of Business." Bell Journal of Economics. December 1986. SHAPIRO. Reputation.CABRAL MILGROM. 26-30. J. January 2000. pp." Tufts University. 548-563. WERNERFELT. "What's in a Name? Reputation as a Tradeable Asset." Northwestern University. S. C. AND ROBERTS. "Measuring Image Spillovers in Umbrella-Branded Products. pp. SULLIVAN. pp." Journal of Economic Theory. "A Theory of Self-Enforcing Agreements. 53 (1980). AND RICHARDS. TAUBER. "Premiums for High Quality Products as Returns to Reputation. "Umbrella Branding as a Signal of New Product Quality: An Example of Reputational Economics of Scope. 458-466. M. 659-680. 27-44. 63 (1990). TELSER. 715. Vol. 309-329.R. . Vol. D. B." Journal of Advertising Research. Vol. 27 (1982).G. 19 (1988). ? RAND 2000. L. 89 (1999). "The Simple Economics of 'Brand Stretching'. pp. "Brand Leverage: Strategy for Growth in a Cost-Control World. pp." American Economic Review." Quarterly Journal of Economics. 28 (1988)." Journal of Business. Kellogg GSM DP no. pp. "Predation." RAND Journal of Economics. . L. / 673 P. pp. Vol. E. Vol.M. "Umbrella Branding as a Signal of New Product Quality: An Example of Signalling by Posting a Bond. pp. 12 (1981).

Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

We've moved you to where you read on your other device.

Get the full title to continue

Get the full title to continue listening from where you left off, or restart the preview.

scribd