glamour brands and glamour stocks

G ModelJ

G

Ma

b

a

ARRAA

JGGG

KGBH

migiMdtt

a

QsHAs

0

ARTICLE IN PRESSEBO-3326; No. of Pages 16

Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Journal of Economic Behavior & Organization

j ourna l h om epa ge: w ww.elsev ier .com/ locate / jebo

lamour brands and glamour stocks�

atthew T. Billetta,1, Zhan Jiangb,∗, Lopo L. Regoa,2

Kelley School of Business, Indiana University, 1309 E 10th Street, Bloomington, IN 47405-1701, USAShanghai Advanced Institute of Finance, Shanghai Jiaotong University, 211 West Huaihai Road, Datong Plaza, Shanghai 200030, China

r t i c l e i n f o

rticle history:eceived 22 January 2013eceived in revised form 27 February 2014ccepted 15 March 2014vailable online xxx

EL classification:021214

eywords:lamour stocksrand equityML loadings

a b s t r a c t

We explore the influence of customer perceptions from the product market on firms’ returncharacteristics in the stock market. Using customers’ opinions on over 1200 brands, we findthat stocks of companies with prestigious brands have high market-to-book ratios and largenegative loadings on the Fama-French HML factor. This relation is not explained by distressrisk, asset irreversibility/growth, or information asymmetry. The HML loadings are mostpronounced when retail investor ownership is high (when institutional ownership is low),when the brand is less familiar, and when market-wide investor sentiment is high. Weconclude glamour in the product market is an important component of glamour in thestock market.

© 2014 Elsevier B.V. All rights reserved.

Numerous studies suggest that customer relationships in the product market influence investment decisions in financialarkets. Grullon et al. (2004) find the breadth of ownership, in terms of the number of institutional and retail investors,

ncreases in firm advertising expense. Frieder and Subrahmanyam (2005) find that firms with more familiar brands havereater retail investor bases. Keloharju et al. (2012) explore the stock holdings of retail investors and find that individualsnvest a disproportionate amount of their portfolio in stocks where they have a significant product market relationship.

oreover, they find this product market bias does not reflect any information advantage and conclude that retail investorserive utility from such ownership “. . .just as a fan of a sports team could derive direct utility from her investment in theeam.” While these links establish that a firm’s products influences its ownership structure, little evidence exists on how

Please cite this article in press as: Billett, M.T., et al., Glamour brands and glamour stocks. J. Econ. Behav. Organ. (2014),http://dx.doi.org/10.1016/j.jebo.2014.03.014

his ultimately influences firm value and stock returns.Taking brand equity metrics from the marketing literature, we examine how customer perceptions of companies’ brands

ssociates with equity valuation and stock return characteristics. We measure customer perception using a brand equity

� We thank David Barker, Werner De Bondt, Bob Chirinko, Burcu Esmer, Lily Fang, Jon Garfinkel, David Hirshleifer, Kenneth Kim, Igor Kozhanov, Yimingian, Fatma Sonmez, Inho Suk, Cristian Tiu, Ashish Tiwari, K.C. John Wei, Tong Yao, the editors, and three anonymous referees as well as conference and

eminar participants at the 2011 American Finance Association, the 2010 Behavioral Finance Conference at DePaul University, 2010 City University ofong Kong International Conference on Corporate Finance and Financial Markets, Beijing University, Georgetown University, Indiana University, Shanghaidvanced Institute of Finance, SUNY-Buffalo, University of Iowa, University of South Carolina, and University of South Florida for helpful comments anduggestions. The authors are grateful to HarrisInteractive for access to the EquiTrend© database. All remaining errors are our own.∗ Corresponding author. Tel.: +86 21 6293 2079.

E-mail addresses: [email protected] (M.T. Billett), [email protected] (Z. Jiang), [email protected] (L.L. Rego).1 Tel.: +1 812 855 3366.2 Tel.: +1 812 855 1202.

http://dx.doi.org/10.1016/j.jebo.2014.03.014167-2681/© 2014 Elsevier B.V. All rights reserved.

dx.doi.org/10.1016/j.jebo.2014.03.014

dx.doi.org/10.1016/j.jebo.2014.03.014

http://www.sciencedirect.com/science/journal/01672681

http://www.elsevier.com/locate/jebo

mailto:[email protected]



dx.doi.org/10.1016/j.jebo.2014.03.014

G Model
ARTICLE IN PRESSJEBO-3326; No. of Pages 16
2 M.T. Billett et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx

database containing information gathered annually from over 20,000 customers, and we label our primary measure Prestige.The marketing literature demonstrates that Prestige captures how attached and emotionally connected customers are tothe brands.1 While we empirically explore rational information based and risk channels, our primary focus is on discerningwhether potential mispricing channels may link the product markets and financial markets.2

We begin by examining the relation between Prestige of the firms’ brands and their financial characteristics. We finda strong positive relation between market-to-book and Prestige, consistent with findings in the marketing literature thatstronger brands associate with greater firm value (Gruca and Rego, 2005). We further decompose firm value into funda-mental value and a misvaluation measure proposed by Rhodes-Kropf et al. (2005). We find Prestige is significantly positivelycorrelated to this measure of misvaluation, suggesting the influence of Prestige may not only influence fundamental value,but may also relate to potential misvaluation.

We next explore whether stock return characteristics relate to Prestige. Lakonishok et al. (1994) argue “glamour” stockshave overly enthusiastic investors which leads to overvalued stocks. They suggest the value premium, measured by theloading on the HML factor, reflects this mispricing rather than a fundamental risk. Similarly, Daniel et al. (2005) suggestovervaluation caused by investor overconfidence can result in a low loading on the HML factor.3 Using this notion we askwhether glamour brands (high Prestige) contribute to glamour stocks. To test this we form portfolios stratified by whethercustomer perceptions place the firm’s brands in the top, middle, or bottom third in Prestige. We find this is indeed the case.Namely, the portfolio of low Prestige firms has a positive and significant loading on the HML factor (0.263), while the loadingon high Prestige brand firms is a significantly negative −0.370.

However, as pointed out by Daniel et al. (2005), misvaluation as well as priced risk give rise to HML loadings. In attemptsto distinguish the risk and mispricing channels we conduct additional tests using commonly accepted risk interpretations ofHML loadings. For example, Chen and Zhang (1998) suggest the HML factor may capture distress risk, and Zhang (2005) andCooper (2006) argue the degree to which a firm has growth opportunities versus assets in place may be an important riskfactor that explains the value premium. In this case the difference in HML loadings between high Prestige and low Prestigeportfolios could just capture underlying risks associated with Prestige. To see if this is the case we use double sorts wherewe first sort on distress risk (asset growth) and then on Prestige. We continue to find significant HML loading differencesacross Prestige terciles after controlling for distress risk (asset growth).

We ask whether Prestige relates to HML simply as a proxy for financial characteristics known to associate with HMLloadings. In the spirit of Lemmon and Portniaguina (2006), we orthogonalize Prestige to firm financial characteristics includingmarket-to-book, firm size, profitability, asset growth, analyst coverage and other controls. We then re-sort the firms and formportfolios based on the residuals from this regression. Portfolios based on high residual Prestige continue to have significantlynegative HML loadings while low residual Prestige firms have positive loadings.

Our third set of tests exploits the idea that mispricing should attenuate as information costs and arbitrage costs decline.Theoretically, mispricing is caused by misguided investors who overestimate the precision of their private information, andtherefore hold incorrect beliefs about the fundamental value of an asset. We hypothesize that less information asymme-try should reduce the incorrect belief and reduce the impact of sentiment on overvaluation. In simple terms, if customerfamiliarity increases investor awareness and information, then we expect Prestige effects due to misvaluation to diminishas the brands familiarity increases. Our test supports this prediction. We first sort firms into portfolios based on Familiarity(a measure from our customer database) and then on Prestige, and we find that the influence of Prestige on HML loadings ispronounced in the unfamiliar brands and diminishes as brands become more familiar. Moreover, Prestige has no effect onHML loadings for the most familiar brands.

We conjecture that if the effect of Prestige on HML reflects mispricing, then this effect will be pronounced when overallmarket-wide sentiment is high. In other words, when market-wide investor sentiment is high we would expect to seethe product-market channel fueling sentiment for particular stocks. Using the market-wide sentiment index of Baker andWurgler (2006), we find that the loading on HML for the high-Prestige low-Familiarity is only significantly negative duringperiods where market-wide sentiment is high. This suggests that, while brand Prestige may be relatively stable, its’ influenceis confined to periods of high overall market sentiment.

Our last set of HML based tests examine whether the HML factor loadings vary with institutional ownership. Presumably,institutional investors should be less affected by Prestige than retail investors if it is indeed indicative of misvaluation. In


this case we would expect to see the relation between Prestige and HML loadings to dissipate as institutional ownershipincreases. This is precisely what we find. Forming portfolios on double sorts where we sort firms into high, medium and lowinstitutional holdings and then by Prestige we find that the HML loading for the high Prestige portfolio is −0.754 for the low

1 We measure brand perceptions using a unique dataset, EquiTrend©. The marketing literature shows brands can create unique images and memoryassociations in the eyes and minds of customers. Moreover, brands can create emotional ties between customers and products (Berthon et al., 1999;Chaudhuri and Holbrook, 2001; Keller, 2003). This dimension is captured by our prestige measure. We also discuss results using consumer awareness,labelled Familiarity. It is important to note that Prestige does not equate to luxury or other price point definitions of goods. The Marketing literature alsorefers to this brand dimension (Prestige) as brand identity or brand emotional connection (Keller, 1993).

2 We also document that a firm’s systematic risk (Beta) decreases in Familiarity as well as prestige, consistent with Rego et al. (2009). This relation isconsistent with the notion that brand equity lowers firm risk by delivering more stable and less volatile profits.

3 They point out that overvaluation as well as priced risk factors can lead to significant HML loadings. This suggests that HML loadings alone cannotdistinguish between misevaluation and risk. We conduct additional tests below in attempts to distinguish these two sources of HML loadings.

dx.doi.org/10.1016/j.jebo.2014.03.014

G ModelJ

iaoHo

CCaceAPr

baal

csf

t4

1

sstw

1

b2aefs

ii1lt1a

aI

l


M.T. Billett et al. / Journal of Economic Behavior & Organization xxx (2014) xxx–xxx 3

nstitutional holding group and 0.480 for the high institutional holding group. Moreover, the difference between the highnd low Prestige HML loadings declines from 1.22 to 0.715 to −0.251 as we move from low to medium to high institutionalwnership. Thus, the glamour effect of Prestige disappears as institutional ownership increases. Given a risk associated withML should not depend on retail versus institutional ownership, these results lend credence to the notion that the influencef Prestige operates, at least to some degree, through misvaluation.

If the HML loadings are indicative of misvaluation, we would expect to see Prestige relate to future returns. Brown andliff (2005) and Baker and Wurgler (2006) examine returns following periods where overvaluation has likely occurred.onsistent with overvaluation, their findings show low overall stock returns follow periods of high market sentiment. In

similar vein, Daniel and Titman (2006) show that intangible past returns (the portion of returns that is not explained byontemporaneous financial performance) portend negative future returns. Thus, following Baker and Wurgler (2006), wexamine the correlation between firm-level Prestige and future stock returns in a multivariate cross-sectional framework.fter controlling for firm characteristics as well as market-wide sentiment, we find that stocks in the top third of customerrestige ratings experience future annual returns that are 5–8% lower than the firms in the bottom third of customer Prestigeatings.

We also explore how customer familiarity associates with stock returns. If the influence of our Prestige measure is driveny an information effect,4 then higher customer familiarity should lead to lower or negative loadings on the HML factor, or on

liquidity factor, and should lead to lower future returns. Similarly, if other behaviour-based alternative explanations suchs familiarity bias or investor attention bias5 exist, then greater customer Familiarity should also lead to lower or negativeoadings on HML factor, and lead to lower future returns. We find no such evidence.

Our paper contributes to the literature on three dimensions: first, we establish a link between how customers view aompany from a product market perspective and how investors view the company from a financial market perspective. Putuccinctly, we find glamour brands associate with glamour stocks. Our findings shed light on the interpretations on the HMLactor and suggest existing risk based explanations for HML do not appear to capture the influence of this customer channel.

We motivate the potential linkages between product market familiarity and sentiment and stock returns by reviewinghe related literature in Section 1. Section 2 describes the data and sample procedures. Section 3 presents results, and Section

presents conclusions.

. Potential influence of product market familiarity and prestige on stock returns

As mentioned above, there are a number of ways information and opinions in the product market may spillover to thetock market. We begin by reviewing the marketing literature to help understand the product market metrics we use in thistudy, customer Familiarity and Prestige. We then discuss rational channels connecting the product market and stock markethat focus on investor awareness, attention, and information effects. Last, we examine how the Prestige of a company’s brandsill likely affect the returns on the company’s stock.

.1. Brand equity and firm characteristics: theory and evidence from marketing

One of the most important marketing functions in a firm is to develop and maintain brand equity. Brand equity cane loosely defined as the value added to a product or service by its association with a brand name or symbol (see Aaker,004; Keller, 1993). Brand equity value is theorized to be a function of customers’ awareness of the brand and the imagessociations of the brand in customers’ memory (Berthon et al., 1999; Lane and Jacobson, 1995). Thus we can think of brandquity as having two primary components: (1) brand awareness, given that customers choose to buy from a feasible set ofamiliar goods and services; and (2) brand image, since customers perceptions of the brand influence their choice among theet of familiar goods and services.

Strong brands are characterized by high levels of customer awareness and strong, favourable, and unique associationsn customers’ memory (Keller, 1993). Strong customer-based brand equity is believed to enable stronger product/servicedentification and facilitate search, suggesting that strong brands will likely exhibit higher repeat purchasing (Berthon et al.,999; Keller, 2003). Strong brands are also more likely to evoke an emotional connection resulting in greater loyalty and

owered susceptibility to rivals’ marketing efforts (Chaudhuri and Holbrook, 2001). In turn, loyal customers are more likelyo repurchase the brand, consider only that brand, and engage in no brand-related information search (Newman and Werbel,973). Finally, brand differentiation (i.e., uniqueness of the brand associations) is believed to reduce product substitutionnd thus protect future cash flows (McAlister et al., 2007).


For the purposes of this study we argue that if brands exhibiting higher levels of awareness (i.e. the first component)re more easily understood and less opaque, then we would expect lower asymmetric information for familiar brands.n addition, if these emotionally connected or “attached” customers become “attached” investors then we would expect

4 I.e. high prestige firms have more information available to the public or less information asymmetry.5 Related literature on home bias and familiarity bias includes Cooper and Kaplanis (1994), Huberman (2001) and Barker and Loughran (2007). Related

iterature on investor attention includes but not limited to Chen et al. (2004) and Barber and Odean (2008).

dx.doi.org/10.1016/j.jebo.2014.03.014

G Model

to find that investor sentiment relates to customers’ perception of the firm’s brands or the brand Prestige (i.e. the secondcomponent).

Independent of these hypotheses that connect brand equity to stock returns, we expect brand equity characteristics torelate to firm fundamentals. The literature documents strong relations between customer-based brand equity and financialperformance and characteristics. Srivastava et al. (1998, 1999) developed an extensive theoretical framework, linking marketbased assets (such as brand equity or customer satisfaction) to shareholder value. Their theoretical framework has beenempirically tested extensively (see Anderson et al., 2004; Gruca and Rego, 2005; Tuli and Bharadwaj, 2009) and findingsgenerally provide supporting evidence of a positive association between investments in market based assets and superiorfirm performance. There is also ample evidence that customer opinions/brand equity influence firm risk (Rego et al., 2009;Madden et al., 2006). Specifically, Rego et al. (2009) show that brand equity is negatively related to firm risk, where theydocument higher brand equity associates with higher credit ratings and lower equity risk. Thus, it is important that our testscontrol for the influences of these first order channels (i.e., the direct economic link between brand characteristics and firmperformance, value, and risk).

1.2. Investor awareness, attention and information spillover

Both anecdotal and academic evidence suggests that the awareness/familiarity/visibility of a firm among investors mat-ters. Investor awareness comes from a number of sources. For example, investor awareness is related to the geographicvicinity. French and Poterba (1991) find that investors overweight their portfolios with domestic stocks. Coval and Moskowitz(1999) find that US portfolio fund managers prefer investing in locally headquartered firms. One of the explanations for thishome bias (Cooper and Kaplanis, 1994) is the investor’s preference for familiarity (Huberman, 2001). Barker and Loughran(2007) find that the stock returns of geographically close companies are more correlated and that this increased correlationis likely due to trading activity by local investors. The literature also shows that the inclusion in the S&P 500 index enhancesinvestor awareness (Chen et al., 2004). Similarly, media exposure could attract investors. Barber and Odean (2008) showthat that individual investors are net buyers of attention grabbing stocks, e.g., stocks in the news, and stocks experiencinghigh abnormal trading volume. Fang and Peress (2009) find stocks in the media earn significantly lower returns than stockslacking media attention.

Interestingly, recent evidence shows that visibility in the product market is positively associated with investor awareness.Keloharju et al. (2012) show individuals that are customers of a company’s products are more likely to own the company’sstock. Grullon et al. (2004) show greater advertising expense, which likely drives consumer awareness and opinion, associateswith greater breadth of ownership. Frieder and Subrahmanyam (2005) found that Individuals prefer investing in visible,brand-name stocks. Moreover the findings of these studies support rational explanations rather than behavioural ones.Grullon et al. (2004) argue that investor awareness reduces information asymmetry and therefore increases the liquidity.Similarly, Chemmanur and Yan (2009) argue product market advertising can signal the true value to both product marketand financial market prior to equity issuance, therefore reducing the need for underpricing.

However, the rational information effect documented above may coexist with behavioural influences. Increased liquiditymay associate with lower asymmetric information and higher firm value; however, as Baker and Stein (2004) point out,increased liquidity can also be a sign of investor sentiment. Lou (2010) examines the relation between advertising and returnsand finds increased advertising leads to increased individual investor buying and positive excess returns, which reverse inlater years. He concludes this effect is likely due to short-sales constraints or behavioural biases. Moreover, Keloharju et al.(2012), who document that an individual who is a customer of the company’s products is more likely to be an investorin the company’s stock, fail to find a financial motive for the investments. In fact, they conclude the only explanation forthe observed customer–investor relation is that individuals derive utility from owning those stocks for which they are alsocustomers, beyond that which can be explained by risk-return tradeoffs. In addition, while it is very likely that advertisingcan increase visibility, it is also possible that less visible firms spend more on advertising to gain visibility. Advertisingexpense might not be the most direct measure of consumer awareness or investor awareness. Our study employs a directmeasure of product market awareness, customer familiarity, to investigate the relation between firm visibility and firm risk.

1.3. Customer prestige, glamour, and HML factor loadings

Many studies argue that the HML (High Minus Low book-to-market) factor is a result of misvaluation of “glamour” stocks,relative to “value” stocks, suggesting glamour stocks are overvalued relative to their fundamental value. Daniel et al. (2005)suggest overvaluation caused by investor overconfidence results in a low loading on the HML factor, thereby making the pointthat both risk and misvaluation may drive HML loadings. De Bondt and Thaler (1985), for example, find that a contrarianstrategy based on past return information earns significant abnormal returns. Lakonishok et al. (1994) suggest the valuepremium, measured by the HML factor, reflects this mispricing rather than a fundamental risk. Daniel and Titman (1997)argue that the book-to-market ratio factor is nothing but a proxy for overpricing and underpricing phenomena observed in


stock prices. Finally, Chou et al. (2007) claim that the book-to-market ratio reflects mispricing using Japanese stock data.However, other studies provide rational explanations for this observed premium arguing the book-to-market ratio proxies

for an underlying risk. One prominent explanation is that high book-to-market firms are assigned a higher risk premiumbecause of the greater risk of financial distress. Fama and French (1995) find that a high book-to-market ratio signals

dx.doi.org/10.1016/j.jebo.2014.03.014

G ModelJ

pmwlbb(

alf

2

aapft

2fohtI

sbaTSc

Wasuy

a9h

2bbG

FplbsbPw



ersistent low earnings. Chen and Zhang (1998) show that value stocks (high book-to-market) have higher financial leverage,ore earnings uncertainty, and thus are riskier. However, counter evidence is found by Dichev (1998) who shows that firmsith a high likelihood of financial distress, measured by high O-score (Ohlson, 1980 and Z-score (Altman, 1968) tend to have

ow average stock returns. Griffin and Lemmon (2002) further point out the difference in returns between high and lowook-to-market ratio is largest among firms with the highest distress risk and this large differential cannot be explainedy economic fundamentals. Instead, they find support for the mispricing argument, that is, low book-to-market stocks“glamour stocks”) are overpriced and high book-to-market stocks (“value stocks”) are underpriced.

In this paper, we examine the link between HML loadings and customer prestige for a company’s brands. If customerttachment in the product market spills over to the financial market, then we expect that glamour brands (high Prestige) mayead to a glamour stock with positive investor sentiment and a more negative loading on the HML factor, after controllingor the common risk interpretations of the HML factor.

. Data and sample construction

Following prior studies in the marketing literature (see Aaker and Jacobson, 1994; Aaker and Joachimsthaler, 2000; Hellofsnd Jacobson, 1999; Rego et al., 2009) we use the EquiTrend© database developed by the Brand & Strategy Consulting Practicet HarrisInteractive to collect our information on customer familiarity with a company’s brand(s) and to collect customererceptions of the brand’s image. EquiTrend© is a unique database that measures and compares customer-based brand equityor over 1000 brands in 35 industry categories. EquiTrend© provides survey-based measures that are gathered from morehan 20,000 customers each year. We were able to secure complete access to the database for the years 2000–2006.

We measure Familiarity using EquiTrend©’s familiarity score, captured on a 1–5 scale where 1 = “never heard of the brand”, = “just know of the brand”, 3 = “somewhat familiar with the brand”, 4 = “very familiar with the brand”, and 5 = “extremelyamiliar with the brand”. For comparability with Prestige (see below), we rescale brand Familiarity to a 1–10 point scale. Inrder for a brand to be able to sustain a presence in the marketplace, customers must be familiar with it. Companies with theighest Familiarity scores include MCDONALD’S, COCA-COLA and GILLETTE, which are brands that are well recognized byhe customers. Companies with the lowest Familiarity scores include AMGEN, DOMINICKS SUPERMARKETS and INFOSPACENC.

Similarly, brand image, which we label Prestige, is proxied using EquiTrend©’s brand quality score, using a 0–10 scale withcale labels of 0 = “unacceptable/poor”, 5 = “quite acceptable”, and 10 = “outstanding/extraordinary”. This score provides aenchmark of the general population’s opinion about a brand, which can be influenced by exposure to advertising, nostalgicssociations with a brand, etc. The examples of companies with highest brand Prestige score include VOLKSWAGEN, SONY,IFFANY and COACH. The examples of companies with low Prestige scores include ALTRIA, REYNOLDS AMERICAN, PAYLESSHOES and INSIGHT COMMUNICATIONS. Note, that Prestige does not necessarily measure “luxury”, but rather it capturesustomer attachment and connectedness.

Proprietary databases like the EquiTrend© database often require some cleaning. For example, Jim Beam and Jim Beamhiskey are listed as two unique brands when they actually refer to the same brand. However, Reynolds clear plastic wrap

nd Reynolds wrap aluminium foil are correctly listed as two unique brands. Furthermore, during the period under analysis,ome brands were transferred across firms. In order to preserve data quality, we manually check and corrected brand namessed in the EquiTrend© database, and we also manually check the association between brands and companies at any givenear.

We required non-missing GVKEY and PERMNO for all the companies with brand equity indices so that we can collectccounting information from COMPUSTAT and CRSP. After applying these filters, we have 4355 brand-year observations,92 unique brand and 318 unique firms. To construct a firm measure of Familiarity and Prestige we use the brand with theighest score given that brand is likely to be the dominant brand for the firm.6

Although we have only 318 firms in our sample, they represent a significant portion of the US economy. For sample period000–2006, the total market cap of our sample firms ranges from $6.74 trillion in 2002 to $10.4 trillion in 2006, which isetween 67% and 75% of total NYSE market cap during the same period. Our sample firms include the top 10 largest US firmsy market capitalization, such as the Exxon Mobil, Microsoft, Wal-Mart, Procter and Gamble, Apple, Johnson & Johnson,oogle, General Electric, IBM, and JP Morgan.

Table 1 panel A shows the year distribution of brand Familiarity and brand Prestige scores. We observe a fairly stableamiliarity score over the sample period 2000–2006, while there is a slight decrease in overall Prestige during our sampleeriod. Even though brand Familiarity and brand Prestige capture different customer perspectives, these two measures are

ikely to be positively correlated. A simple example would be Hershey’s products are not only well known but also are likedy customers. The sample correlation between Prestige and Familiarity is 0.360. It confirms with our expectation but at theame time demonstrates the significant difference between these two metrics: a well-known brand might not necessarily


e a prestigious brand. We also present the industry distribution and industry mean values of Prestige and Familiarity inanel B. Overall our sample spans many industries and we do not see any large industry effects on Prestige, although weill explicitly control for industry in some of our tests, below.

6 For example, Coca Cola owns both Coke® and Dassani® water. As a robustness check we take an average across all brands and find similar results.

dx.doi.org/10.1016/j.jebo.2014.03.014

ARTICLE IN PRESSG ModelJEBO-3326; No. of Pages 16


Table 1Descriptive statistics: brand equity indices. The table presents distribution of brand equity scores of our sample firms by years (panel A) over our sampleperiod 2000–2006 and by 12 Fama-French industries (panel B). Our brand equity measures, brand Familiarity and brand Prestige, are obtained from EquiTrenddatabase. The Prestige score, ranging from 0 to 10, provides a benchmark of the general population’s opinion about a brand’s quality. The Familiarity score,ranging from 1 to 10, measures the customer awareness of a specific brand. In Panel B we report scores based on the 12 Fama-French industry classifications.

Panel A: year distribution

Year No. of firms Familiarity Familiarity Prestige PrestigeMean Median Mean Median

2000 218 6.941 7.265 7.426 7.5032001 224 6.521 6.852 7.305 7.3462002 284 6.105 6.508 7.271 7.3082003 290 6.029 6.402 7.227 7.2832004 281 6.687 7.206 6.997 7.0072005 282 6.784 7.313 6.975 6.9502006 286 6.547 7.171 6.873 6.891

Panel B: industry distribution

FF industry classifications No. of firms Familiarity Familiarity Prestige PrestigeMean Median Mean Median

Consumer non-durables 314 7.329 7.546 7.255 7.382Consumer durables 97 7.692 7.853 7.818 7.898Manufacturing 106 7.841 7.944 7.544 7.513Energy 42 6.530 7.045 6.867 6.858Chemicals and allied products 68 8.091 8.767 7.633 7.780Business equipment 194 5.856 6.175 7.184 7.255Telephone and television transmission 142 6.200 6.637 6.796 6.831Wholesale, retail, and some services 383 6.619 6.810 7.062 7.140Healthcare, medical equipment, and drugs 76 6.046 6.570 7.324 7.470Finance 205 4.525 4.190 6.603 6.575Other 238 6.214 6.640 7.130 7.043

* Significance at the 10% level.** Significance at the 5%, level.*** Significance at the 1%, level.

We recognize that there is a selection bias for our sample since the firms covered by the EquiTrend© database are alreadyto some extent recognizable and therefore bigger and more mature than the general population. Table 2 examines the firmcharacteristics of our sample firms as well as three subsamples sorted by terciles of Prestige (panel A). We observe that firmswith more prestigious brands are significantly more profitable (i.e. ROA), have higher cash flow and lower market volatilitythan firms with less prestigious brands. One interpretation could be these high Prestige firms are bigger, more mature andless risky. However, panel A shows that the difference in firm size is contrary to this explanation although not statisticallysignificant. Similarly, we observe no significant difference in either capital structure or asset growth among firms withdifferent Prestige levels. We see capital expenditures, R&D, and advertising increase in Prestige, suggesting it takes greaterinvestment to create prestigious brands, presumably by creating differentiable products and via advertising.

The market-to-book ratios are higher in firms with prestigious brands than those with less prestigious brands, whichcan be interpreted as higher growth rate and/or more overvaluation in those high Prestige firms. We further analyze themarket-to-book ratio in Panel B where we report mean values by terciles of Prestige and Familiarity. Interestingly market tobook increases in both dimensions.

We also explore how Familiarity in the product market associates with analyst coverage in the stock market. We computetwo measures of analyst information: the number of analysts covering the stock and analyst earnings forecast dispersion.And compute the correlations between these measures and Familiarity and Prestige. These results are reported in the InternetAppendix, Table A.1. We find Familiarity is significantly positively correlated with analyst coverage. The correlation betweenPrestige and analyst coverage is also significant, albeit weaker both statistically and economically. We also find that Prestigeis significantly negatively correlated with analyst forecast dispersion, while there is no correlation between Familiarity andanalyst forecast dispersion. These results suggest we may need to control for the information environment of the firm whenwe attempt to isolate the effects of Prestige on stock returns (see below).

At this point, we do not have any conclusive findings on the relation between risk (i.e. market risk, financial risk, liquidityrisk) and Prestige. However, our univariate analysis so far seems to indicate that Prestige is not a simple derivative of firmlife cycle – namely, bigger firms with more cash flow and lower growth rates do not necessarily have higher brand Prestige.Overall, the results in Table 2 support many of the findings in the marketing literature that relate brand equity to firmfundamentals.


dx.doi.org/10.1016/j.jebo.2014.03.014



Table 2Summary statistics for sample firm characteristics. The table presents sample firm characteristics across 3 terciles (low, medium and high) of Prestige score(panel A) and market-to-book ratio by terciles of Prestige and Familiarity (panel B). The firm total asset (TA) is directly obtained from COMPUSTAT. Othervariables are defined as follows: market-to-book = market value of equity/book value of equity; Book leverage = (book value of long-term debt + book value ofshort-term debt)/TA; Cash flow = (Net income + depreciation and amortization)/TA; ROA = EBIT/prior year TA; Tobin’s Q = (book value of total asset + marketvalue of equity-book value of equity)/book value of total asset; The construction of Z-score follows Altman (1977). The calculation of asset growth oppor-tunity follows Cooper et al. (2008) and is the annual change in total assets; Institutional holdings are defined as the percentage institutional ownershipcompared with total shares outstanding. The institutional holdings are obtained from 13F; Capital investment is CAPEX/TA; Advertising expenditure andR&D expenditure are scaled by TA. The significance level of the difference is based on a Wilcoxon sum-rank test.

Panel A: across different levels of Prestige

Variables Prestige

Low Medium High All Diff (high-low)

Total assets 14,690 9032 11,936 11,864 −2753Market-to-book 2.143 2.669 3.534 2.746 1.390***

Book leverage 0.241 0.226 0.252 0.241 0.011Cash flow 0.074 0.096 0.099 0.092 0.026***

ROA 0.072 0.111 0.117 0.102 0.045***

Altman’s Z-score 1.508 2.093 2.025 1.923 0.517***

Asset growth 0.064 0.069 0.067 0.066 0.003Institutional holdings 0.648 0.682 0.642 0.653 −0.007Capital investment 0.037 0.041 0.043 0.041 0.006***

Advertising expenditure 0.021 0.038 0.036 0.032 0.015***

R&D expenditure 0.005 0.009 0.020 0.014 0.015***

No. of obs 618 624 623 1865

Panel B: market-to-book by Prestige and Familiarity terciles

Low Familiarity Median Familiarity High Familiarity

Low prestige 2.133 2.022 2.559Median prestige 2.030 2.619 3.722High prestige 3.286 2.901 3.774

* Significance at the 10% level.** Significance at the 5%, level.

3

3

aattcIi

eofloa

hm

*** Significance at the 1%, level.

. Results

.1. Customer familiarity, prestige and market-to-book valuation

Srivastava et al. (1998) formalize the link between firm value and market-based assets, which include customer awarenessnd perception of a firm’s brands. We explore the relation between customer characteristics and the market-to-book ratio in

multivariate setting. We regress the firm’s market-to-book ratio on Prestige and Familiarity including firm characteristicshat likely associate with firm value (size, leverage profitability, risk, and asset growth) and we include time effects. We reporthe results in Table 3. In the first specification we see Prestige has a positive and significant coefficient, as does Familiarity,onsistent with the predictions of Srivastava et al. (1998). In the next two specifications we explore transforms of Prestige.n specification 2 we include a dummy variable, High prestige, equal to one when the firm’s Prestige score is in the top tercilen a given year. We see this transform of Prestige also carries a positive and significant coefficient.

One concern is that Prestige may associate with other firm characteristics that influence firm value. To better isolate theffects of prestige we compute Residual prestige. Specifically, we follow Lemmon and Portniaguina (2006) and regress Prestigen analyst coverage, analyst forecast dispersion, institutional ownership, distress risk, asset growth, firm size, ROA, Cashow and year dummies. Then we define Residual prestige as the residual from this first stage regression.7 We see in column 3f Table 3 that the coefficient on Residual prestige is positive and also significant at the 5% level. Now that we have established

strong link between firm value and Prestige, we next explore whether prestige relates to potential misvaluation.To isolate the portion of the market-to-book ratio that may be due to misvaluation we follow Rhodes-Kropf et al. (2005),

enceforth RRV. RRV express the market-to-book ratio as the ratio of current market value, M, to “true” market value, V ,ultiplied by the ratio of “true” market value to the book value, B.

( )( )


M

B≡ M

V

V

B(1)

7 See internet appendix Table A.2 for first stage regression results.

dx.doi.org/10.1016/j.jebo.2014.03.014



Table 3Prestige and market-to-book ratio. This table reports panel regression of market-to-book ratio and total misvaluation on Prestige and Familiarity scores andother predictor variables. Total misvaluation is constructed from a decomposition of the market-to-book ratio as in Rhodes-Kropf et al. (2005). Prestige isequal to one, two or three when the Prestige score is in the bottom, middle and top third of all firms at specific year respectively. High Prestige dummy isequal to one if the Prestige score is the top third of all firms in that specific year. Residual Prestige is the residual from the first stage regression of Prestigeon a set of control variables including analyst coverage, analyst forecast dispersion, institutional ownership, distress risk, asset growth, firm size, ROA, Cashflow and year dummies. Familiarity is equal to one, two or three when the Familiarity score is in the bottom, middle and top third of all firms at specific yearrespectively. The residual Prestige is equal to one, two or three when the residual Prestige is in the bottom, middle and top third of all firms at specific yearrespectively. High Prestige dummy is equal to 1 if the Prestige score is in the top third of all firms, 0 otherwise. The heteroscedasticity-robust t-statistics arereported in parentheses under the estimates.

Variables Market-to-book ratio Total misvaluation by decomposingmarket-to-book ratio

1 2 3 4 5 6

Prestige 1.629** 0.093***

(2.092) (2.640)High prestige (dummy) 2.874** 0.110**

(2.504) (2.029)Residual prestige 1.823** 0.118***

(2.169) (3.385)Familiarity 0.939** 0.893** 0.648 0.064* 0.069* 0.033

(2.375) (2.212) (1.495) (1.788) (1.899) (0.872)Log 0.059 0.031 0.233 −0.030 −0.031 −0.013(Total asset) (0.121) (0.063) (0.420) (−1.460) (−1.483) (−0.564)Book −4.360 −4.300 −4.685 1.247*** 1.251*** 1.346***

Leverage (−1.249) (−1.229) (−1.197) (6.302) (6.272) (6.387)ROA 14.458*** 14.056*** 17.189*** −0.082 −0.090 0.097

(3.048) (3.003) (2.896) (−0.325) (−0.347) (0.355)Z score −0.148 −0.038 0.101 −0.017 −0.010 0.020

(−0.171) (−0.046) (0.104) (−0.505) (−0.284) (0.720)Asset growth −0.134 −0.141 −0.125 −0.018** −0.018** −0.011

(−1.466) (−1.532) (−1.063) (−2.427) (−2.444) (−1.217)Momentum 0.489 0.409 0.094 0.091 0.085 0.111(Last 1 year) (0.669) (0.543) (0.114) (1.430) (1.319) (1.379)Constant 0.633 3.126 −0.815 0.429* 0.567** 0.151

(0.104) (0.495) (−0.116) (1.750) (2.337) (0.593)Year dummy Yes Yes Yes Yes Yes YesObservations 1231 1231 1009 1035 1035 841R-square 0.043 0.0446 0.047 0.188 0.182 0.224



In log terms this expression is

m − b ≡ (m − b) + (v − b) (2)

RRV then further decompose this difference between the market value and book value of a firm into three components:

mit − bit = [mit − v(�it; ˛jt)] + [v(�it; ˛jt) − v(�it; ˛j)] + [v(�it; ˛j) − bit] (3)

where mit is the market value of firm i at time t, bit is the book value of firm i at time t, v(�it; ˛jt) is the fundamental “true”value estimated using firm accounting data, �it , and contemporaneous (time t) sector j accounting multiples, ˛it, v(�it; ˛j) isthe fundamental “true” value estimated using firm accounting data, �it , and the long-run sector accounting multiples, ˛j .

Given these definitions, RRV interpret the first bracketed term in Eq. (3) as firm specific misvaluation, the second bracketedterm as sector specific misvaluation, and the third bracketed term as the deviation of true market value and book value.RRV find that these components help explain why specific firms as well as specific sectors experience takeover activityand merger waves. We adopt this decomposition to see whether Prestige and Familiarity relate to these components. Wehypothesize that if Prestige is related to overvaluation then it should be positively correlated with either or both the first andsecond components of Eq. (3). RRV argue that the estimate of long-run value, v(�it; ˛j), could reflect information possessedby firm managers but unknown to the market at time t. This would imply that the estimate of time-series sector error,[v(�it; ˛jt) − v(�it; ˛j)], could be a form of misvaluation due to information asymmetry and not necessarily a reflection ofbehavioural biases. If more familiar firms have less asymmetric information then we would predict less misvaluation for


more familiar firms.We implement RRV’s decomposition following their methodology and implementation.8 First, we find that Prestige is pos-

itively correlated with both components of misvaluation, especially the firm-specific misvaluation. The correlation between

8 Hertzel and Li (2010) also adopt this technique to explore behavioral explanations of stock price performance around SEOs.

dx.doi.org/10.1016/j.jebo.2014.03.014

G ModelJ

PmtRtoo

3

md“oi2f

Ptttbad

LmfiPa

iWsw

tdco2mP2

sP

asi

imgi



restige and firm-specific (total) misvaluation is 0.148 (0.182). We also find a positive correlation between Familiarity andisvaluation, although the correlation between firm-specific misvaluation and Familiarity is neither economically nor sta-

istically significant. In the last three columns of Table 3 we report results of regressions where the left hand side variable isRV’s total misvaluation and using the specifications from columns 1–3. We see Prestige is significantly positively related tootal misvaluation. In all three specifications the coefficient on the prestige variable is positive and significant at the 5% levelr better. Familiarity is only weakly related to misvaluation, suggesting Prestige’s influence on the misvaluation componentf firm value more pronounced than that of familiarity.

.2. Customer familiarity, prestige, and stock return factor loadings

We next examine how Familiarity and Prestige associate with stock returns. Given the relation between Prestige and botharket-to-book and total misvaluation, we would expect Prestige to relate to HML factor for both risk and mispricing, as

iscussed in Section 1. We form annual portfolios of firms based on their prior years’ characteristic. We first form threePrestige” portfolios based on whether customer opinions about the Prestige of the company’s brands are in the top, middle,r bottom third of the sample in year t − 1. We then calculate the value weighted returns for each portfolio for the monthsn year t. Thus our sample spans 84 months from 2001 to 2007 corresponding to our brand equity data from years 2000 to006. We regress these monthly returns on the three-factor model of Fama and French.9 We report results using the threeactor models in Table 4.

In Panel A we see low Prestige firms have a loading on HML of 0.263, significant at the 1% level. In contrast, the highrestige firms have a negative loading on HML, −0.370, also significant at the 1% level. Moreover the difference betweenhese two coefficients is significant at the 1% level. In contrast, we find no significant differences between the loadings onhe market factor or SMB factor for the high versus low Prestige portfolios. We find no such difference in HML loadings acrosshe portfolios stratified by Familiarity (see internet appendix Table A.4). We also estimate the results using five portfoliosased on Prestige quintiles and we find the HML loadings for the lowest to highest quantiles are 0.334, 0.081, −0.194, −0.230,nd −0.395 (see internet appendix Table A.5). Both the lowest and highest to loadings are significant at the 5% level and theifference on the loadings of the highest and lowest quintile is significant at the 1% level.

To help disentangle the influence of Prestige that may be attributed to firm fundamental characteristics we again followemmon and Portniaguina (2006) and use residual Prestige using all the aforementioned control variables as well as thearket-to-book ratio in the first stage regression of Prestige. We then take the residuals from this regression and re-sort

rms based on this residual Prestige.10 Panel B reports the results sorting the portfolios by residual Prestige. Given residualrestige is a generated regressor, we report z-statistics based on bootstrap standard errors following Pagan (1984). Here wegain see a negative and significant loading on the high portfolio that is statistically different from that of the low portfolio.

One possible explanation could be that Prestige is simply capturing an industry effect. To see if this is the case we constructndustry-adjusted Prestige by subtracting the average Prestige score for all firms in the same 2-digit SIC code in a given year.

e then resort based on this measure and report the results in Panel C. As in the prior two panels, we see a negative andignificant loading on the HML for the high group. We also see a positive and significant loading on HML for the low group,ith the difference in loadings significant at the 1% level.

We find Prestige and Familiarity are fairly stable over time, and this may be even more so at the brand level givenhe persistence of customer opinions. Therefore, our tests on the sentiment effect may be driven more by cross-sectionalifferences rather than time series. To focus on the time series dimension we recognize that over a relatively longer horizon,ustomer sentiment is more likely to experience significant changes. To see whether this time series dimension supportsur prior findings, we construct a portfolio of all firms that experience a significant decrease or increase in Prestige from000 to 2006. A significant increase is defined as a change in Prestige from (1) the bottom third to the top third, or (2) theiddle third to top third, or (3) the bottom third to middle third. Similarly, a significant decrease is defined as a change in

restige that leads to a similar drop in categorization. For each portfolio, we compare the HML loadings for the subperiod11


000–2004 and for the subperiod 2004–2006.An advantage of this time series analysis is that it can reduce the concern that cross-sectional differences in customer

entiment could be driven by some omitted firm characteristics.12 We find that the portfolio with a significant decrease inrestige experiences a corresponding statistically significant increase in HML loading (from significantly −0.854 to insignif-

9 We also use a five factor model that adds in a momentum factor (Carhart, 1997) and a liquidity factor (Pastor and Stambaugh, 2003), see internetppendix Table A.3. We find similar results. We tried using the UMO factor developed by Hirshleifer and Jiang (2010) instead of HML factor and find aimilar difference in loadings between high prestige and low prestige portfolios. As Hirshleifer and Jiang (2010) argue, UMO factor captures the misvaluationn the stock market. Our result here is consistent with our findings based on HML factor and with sentiment hypothesis.10 To further test that Prestige is not just picking up other factors that also contribute to overvaluation, we first sort firms by the market-to-book rationto high, medium, and low groupings. Then within these three categories we further sort by customer prestige. We find that the HML loading for high

arket-to-book firms is 0.118, insignificantly different from zero, for the low prestige group and −0.460, significant at the 1% level, for the high prestigeroup. Moreover the difference between these two loading is significant at the 5% level. Thus, the effect of prestige does not seem to be entirely throughts relation to the market-to-book ratio.11 The choice of year 2004 is fairly ad hoc, the results are consistent when we split our sample period to 2000–2003 and 2003–2006.12 We recognize that this time series analysis still cannot fully exclude this possibility, for example the omitted factor may be time varying as well.

dx.doi.org/10.1016/j.jebo.2014.03.014



Table 4Prestige and HML loadings. This table presents three-factor regressions of value-weighted monthly excess returns for portfolios of firms sorted by Prestige(panel A), Residual Prestige (Panel B) and industry-adjusted Prestige (Panel C) in the prior year. The excess return is defined as the difference betweenportfolio return and risk free return. Both Prestige and Familiarity are defined in Table 1. The residual Prestige is the residual from the first stage regressionof Prestige on a set of firm characteristics that include firm size, market-to-book ratio, ROA, Cash flow, analyst coverage, analyst forecast dispersion,institutional ownership, distress risk, asset growth and year dummies. The residual Prestige is equal to one, two or three when the residual Prestige scoreis in the bottom, middle and top third of all firms at specific year respectively. The industry-adjusted Prestige is constructed by subtracting the averagePrestige score for all firms in the same 2-digit SIC code in a given year. The explanatory variables in this table are MKT, SMB, HML. These variables arethe returns to zero-investment portfolios designed to capture market, size, and book-to-market, respectively. These factors are collected from KennethFrench website; the constructions of these factors are given in Fama and French (1993). The sample period is from January 2000 through December 2006.The heteroscedasticity-robust t-statistics are reported in parentheses under the estimates. In panel B, z-statistics based on bootstrap standard errors arereported in parentheses under the estimates. The significance of the difference in coefficients is based on Wald tests.

Panel A: sorts based on Prestige

Variables Prestige

Low Medium High Difference between low and high

MKT 1.020*** 1.071*** 0.968*** 0.052(16.50) (12.52) (14.78)

SMB −0.335*** −0.264** −0.194** −0.141(−3.78) (−2.54) (−2.01)

HML 0.263*** −0.141 −0.370*** 0.633***

(3.15) (−0.95) (−3.27)Constant −0.000 0.003 −0.000

(−0.21) (1.12) (−0.01)Observations 84 84 84R-squared 0.84 0.84 0.86

Panel B: sorts based on Residual prestige

Variables Residual prestige

Low Medium High Difference between low and high

MKT 0.944*** 0.984*** 1.008*** −0.064(9.10) (12.09) (10.19)

SMB −0.365*** −0.258** −0.206 −0.159(−3.14) (−2.16) (−1.52)

HML 0.113 −0.310** −0.353** 0.466**

(1.01) (−2.11) (−2.40)Constant 0.005** 0.005* 0.002

(2.34) (1.88) (1.06)Observations 84 84 84R-squared 0.74 0.75 0.80

Panel C: sorts based on industry-adjusted Prestige

Prestige

Variables Low Medium High Difference between low and high

MKT 1.015*** 0.955*** 1.087*** −0.072(18.23) (19.42) (17.87)

SMB −0.448*** −0.220*** −0.261*** −0.187(−5.65) (−2.88) (−2.84)

HML 0.253*** 0.038 −0.476*** 0.725***

(2.70) (0.40) (−4.13)Constant 0.002 0.003* 0.003*

(1.11) (1.81) (1.69)Observations 84 84 84R-squared 0.83 0.87 0.89

*
Significance at the 10% level.** Significance at the 5%, level.

icantly −0.110), suggesting a decrease of glamour in the product market associates with a decrease of glamour in the stockmarket and therefore less overvaluation (see internet appendix Table A.6). Similarly, an increase of glamour in the prod-uct market associates with an increase in the glamour in the stock market and therefore more overvaluation. However,the increase is not statistically significant in this case. The asymmetry could be due to several reasons: first, there might a


downward shift in the market-wide sentiment over the sample period which attenuates the increase in individual firm-levelsentiment. Second, these tests simply lack sufficient power given the sample size.

These results support the “glamour” effect of customer prestige. However, this may be driven by differences in risk,misvaluation, or both, which we explore in detail below. As noted earlier, Prestige could reflect the information risk associated

dx.doi.org/10.1016/j.jebo.2014.03.014



Table 5Prestige and future returns. This table reports panel regression of annual stock return on Prestige and Familiarity scores and other predictor variables duringthe previous year. In the full sample analysis (models 1 and 2), Prestige is equal to one, two or three when the Prestige score is in the bottom, middle andtop third of all firms at specific year respectively. High Prestige dummy is equal to one when the Prestige is in the top third of all firms at specific yearand Middle Prestige dummy is equal to one when the Prestige is in the middle third of all firms at specific year. Familiarity is equal to one, two or threewhen the Familiarity score is in the bottom, middle and top third of all firms at specific year respectively. The residual Prestige is the residual from thefirst stage regression of Prestige on a set of firm characteristics that include firm size, market-to-book ratio, ROA, Cash flow, analyst coverage; analystforecast dispersion, institutional ownership, distress risk, asset growth and year dummies. Market sentiment index is an annual index, obtained directlyfrom Wurgler’s website. For the details, please refer to Baker and Wurgler (2006, 2007). In the subsample analysis, we include the firms with only topand bottom third of Prestige scores (model 3). High Prestige dummy is equal to 1 if the Prestige score is in the top third of all firms, 0 otherwise. Theheteroscedasticity-robust t-statistics are reported in parentheses under the estimates.

Variables 1 2 3

Prestige −0.039**

(−2.55)Residual −0.032*

Prestige (−1.83)High Prestige (dummy) −0.078**

(−2.58)Middle Prestige (dummy) −0.058**

(−2.01)Familiarity 0.000 0.016 −0.000

(0.03) (0.95) (−0.00)Market −0.036* −0.010 −0.036*

Sentiment index (−1.83) (−0.45) (−1.85)Log −0.030*** −0.042*** −0.031***

(Total asset) (−4.02) (−3.86) (−4.09)Market-to-book −0.001** −0.001** −0.001**

(−2.25) (−2.33) (−2.30)Momentum (last 1 year) −0.056 −0.046 −0.057

(−1.27) (−0.91) (−1.28)Constant 0.563*** 0.604*** 0.540***

(5.26) (4.65) (5.11)Year dummy YES YES YESOBS 1244 982 1244R-squared 0.19 0.18 0.19

wPc

stdoF(lwI

3

besFmo

sPr



ith each firm, especially given the strong positive correlation between Prestige and Familiarity reported in Table 1. Ourrestige measure could also capture other effects, such as investor attention, awareness, as well as information effects thatould exhibit themselves in HML loadings.

To test these possibilities, we rerun our tests of portfolios where membership is based on whether customer Familiarityurveys placed the company in the top, middle, or bottom third. Customer Familiarity should be more directly related tohese alternative explanations than customer Prestige. We find that the factor loadings for these portfolios are strikinglyifferent from those for Prestige. We find that the low Familiarity portfolio has a statistically significant negative loadingn HML, −0.238, in contrast to the positive loading for low Prestige firms. The HML loading is insignificant for the highamiliarity group. Moreover, the difference between the HML loadings for the low and high Familiarity groups is insignificantsee internet appendix Table A.4). This suggests that the correlation between Familiarity and Prestige is not driving our HMLoadings results for the Prestige portfolios. However, it may be that Prestige simply associates with greater risks that associate

ith HML loadings. While we explore this possibility below, first we examine the relation between Prestige and future returns.f the negative loadings represent mispricing then we would expect Prestige to associate with lower future returns.

.3. Customer sentiment and future returns

Following Fama and French (1992), Hirshleifer et al. (2004), and Baker and Wurgler (2006), we explore the relationetween a firm’s annual returns and the Prestige of its brands. If Prestige does spillover to investor sentiment, then we shouldxpect a negative correlation between customer sentiment and future return. As Baker and Wurgler (2006) point out, investorentiment leads to overvaluation and therefore lower future returns. Therefore, we regress annual stock returns on Prestige,amiliarity as well as firm characteristics in the previous year. The firm characteristics include size, market-to-book, andomentum. In these regressions we define Prestige and Familiarity equal −1, 0, and 1 if they are in the bottom, middle or top

f third based on the customer surveys. We also include Baker and Wurgler’s market sentiment index and year dummies.


The results are reported in Table 5. In specification one, we find that the Prestige variable has a coefficient of −0.039,ignificant at the 5% level. This suggests annual returns are lower by 7.8% per year for the high Prestige versus the lowrestige firms. In contrast, the coefficient on Familiarity is insignificant, suggesting no relation between Familiarity and futureeturns. In specification two we replace Prestige with residual Prestige, designed to separate sentiment effect from firm

dx.doi.org/10.1016/j.jebo.2014.03.014

G Model

fundamentals. The coefficient on residual Prestige remains significantly negative (−0.032) at the 10% level.13 In specificationthree we include dummy variables, High Prestige and Middle Prestige, which equal one if the firm is the top or middle tercileof Prestige, respectively. The coefficient on the dummy variable High Prestige is −0.078, significant at the 5% level and theMiddle Prestige coefficient is 2% higher, −0.058, and also significant at the 5% level. We also test the difference in the returnsbetween the high and middle groups or between the middle and low groups. The differences are in the same direction, butnot statistically significant. In addition, in robustness checks we include the ratio of advertising expense to total assets (wealso include lagged advertising to total assets). We find a significantly negative coefficient on advertising and continue tofind a negative and significant coefficient on Prestige (−0.039).

We also estimate these regressions using the Fama-Macbeth (FM) approach. This should help see to what extent theresults are driven by cross-sectional variation versus time-series. First we estimate yearly cross-sectional regressions andthen average the seven resulting coefficients. While we find the resulting coefficient on Prestige is negative, we also find itto be statistically insignificant. This could indicate that our results are driven more by time-series effects; however, it mayalso be due to the lack of statistical power given our limited of data.

Taken as a whole, these results are consistent with the glamour results discussed earlier. This suggests that the negativeHML loadings may at least in part be driven by misvaluation and may not be entirely attributable to risk. Next we explorethe HML loadings using double sorts in an attempt to further distinguish risk from mispricing.

3.4. Financial distress risk, asset growth, and HML factor loadings

There are a number of risk-based explanations for varying HML loadings that may be at work along with, or perhapseven through, our customer prestige measure. In fact, we know Prestige relates to firm fundamentals, including risk, fromTable 1 and other studies. As we have shown in the univariate analysis, firms with higher Prestige are more profitable, havehigher market-to-book ratios and Tobin’s Q. In the three-factor model, we also show that the portfolio consisting of higherprestigious firms exhibits a slightly lower market beta. However, since the most significant difference between differentPrestige portfolios comes from the HML loadings, we examine whether this difference captures overvaluation in higherPrestige portfolio or lower fundamental risk in this portfolio.

As noted earlier, we recognize different interpretations of HML factor loadings in the literature: HML factor may capturefinancial distress risk (Chen and Zhang, 1998) or asset growth (Zhang, 2005; Cooper, 2006). We have also discussed thepossibility that Prestige may be inversely related to distress risk (see Section 1.1). Rego et al. (2009) show a compositemeasure of brand equity associates with lower creditor risk and lower equity risk. It is possible that the negative HMLloadings for firms with high Prestige brands could be capturing lower distress risk.14

In contrast, the prediction on firm Prestige, asset growth and the HML loading is less clear. On the one hand, it is possiblethat high Prestige indicates high growth opportunities and, therefore, leads to negative HML loadings. On the other hand,Zhang (2005) and Cooper (2006) argue that assets in place are inherently riskier than growth options given that investmentin assets is irreversible. This difference in the risk of assets in place versus growth options will then be captured in the loadingon HML (where a positive loading reflects that assets in place are riskier than growth options). Therefore, we would expectthat the large investment required to establish a high Prestige brand or a familiar brand is sunk and highly irreversible. Thiswould suggest high Prestige firms should have positive HML loadings.

To formally test these alternative interpretations of loadings on the HML factor, we form portfolios based on double sorts.We form nine portfolios based on the whether the firm’s financial distress risk, measured by Altman’s Z score, is in the top,middle, or bottom third of the sample and based on whether the firm’s brands are in the top, middle or bottom terciles ofPrestige. We then conduct a similar analysis using nine portfolios based on whether the firm’s asset growth, calculated as inCooper et al. (2008), is in the top, middle, or bottom third of the sample. We then run three and five factor return regressionson these subsample portfolios to see how their risk and return characteristics differ. We discuss but do no tabulate theresults.

When we stratify the sample by both financial distress risk and Prestige. For the low Z score group (i.e., high distress)none of the HML loadings are significant. For the medium Z score group, we see the low Prestige portfolio has a significantlypositive loading and the high Prestige portfolio has an insignificantly negative loading. Perhaps more importantly these twoloadings are significantly different at the 5% level. Turning to the high Z score group (i.e., low distress) we see the low Prestigeportfolio has a significantly positive HML loading and the high Prestige portfolio has a significantly negative loading. The


difference is also highly significant. These results suggest that Prestige is not just capturing distress risk.We find similar relations between Prestige and HML loadings, consistent with a glamour effect, after controlling for asset

growth. We find, across each tercile of asset growth, that a positive HML loading on the low Prestige portfolio and a negative

13 We also compute the z-statistic based on bootstrap standard errors following Pagan (1984) given residual Prestige is a generated regressor. The z-statistic associated with the residual Prestige coefficient is −2.14, significant at the 5% level. Another concern may be that returns drive Prestige. Soaringstock returns could raise customer opinions of the firm’s products and we could have a bias due to reverse causality. To test whether such a relation existswe regress Prestige on lagged Prestige, lagged annual stock return, lagged advertising expense, the natural log of assets, and the market-to-book ratio. Wefind the coefficient on the lagged annual stock return is insignificant. Thus reverse causality does not appear to be a big concern.

14 However, this would not explain the findings of an interaction between our prestige and familiarity results and additional results based on double sorts,which we discuss in detail below.

dx.doi.org/10.1016/j.jebo.2014.03.014

G ModelJ

lstw

etmm

psl

3

rtam

bcoMdsb

3

lttai

wso−itv

3

wasr

ia

−



oading on the high Prestige portfolio. In each case the negative loading on the HML factor for the high Prestige portfolio isignificant at the 5% level and the difference between the loadings on the high and low Prestige portfolios is significant athe 1% level in all three cases. These findings suggest that asset growth does not fully explain the glamour affect associatedith Prestige.

Our findings so far are consistent with the notion that glamour effect in the product market contributes to the glamourffect in the stock market in ways that are not fully captured by firm risk. It is important to note that, we by no means arguehat the value premium, more appropriately, the glamour discount in the stock market is entirely explained by the product

arket glamour. As previously mentioned, brand Prestige associates with many firm fundamentals, such as profitability,arket value, etc.In Table 5, we see a large negative loading on the high residual Prestige portfolio and the loading on the low residual

ortfolio is positive and insignificant, with the difference between the two loadings significant at the 1% level. These resultsuggest the correlation of Prestige with the aforementioned firm characteristics is not completely accounting for the HMLoadings.

.5. The influence of the information environment and ownership structure on HML loadings

In this section we implement double sorts to see if the influence of Prestige on varies with the firm’s information envi-onment and varies with the types of investors that own the stock. As we have shown, the information effect does not seemo drive our findings on sentiment. However, we might expect the information and sentiment effects to interact. As Bakernd Wurgler (2007) point out, stocks that are hard to value tend to be most affected by sentiment and that misvaluationay be more likely in less well known/understood stocks.In an attempt to test this interaction, we stratify the sample by both Familiarity and Prestige. As a company’s brands

ecome more familiar it is possible that investors become more informed, which should attenuate any misvaluation. In thisase we would expect more familiar firms to be less subject to mispricing. Panel A of Table 6 reports the results. The loadingn the HML factor for the high Prestige firms goes from −0.870 to 0.100 as we move from low Familiarity to high Familiarity.oreover, the difference between the HML loading on low Prestige and high Prestige firms declines with Familiarity. This

ifference is 0.982, significant at the 1% level, for the low Familiarity subsample, and reduces to 0.281 for the high Familiarityubsample. We interpret this as evidence that the sentiment effect is concentrated in less well known firms and that as firmsecome more commonly known, the sentiment effect diminishes.

.6. The interaction of market-level sentiment and prestige

We observe from the previous section that information (Familiarity) affects how much influence Prestige has on HMLoadings. In this section, we investigate whether market-wide sentiment can also influence the exposure of individual firmso firm-level customer sentiment via Prestige. For the portfolio of high-Prestige and low-Familiarity, which has been showno be most susceptible to sentiment, we compare the glamour effect between two periods: high market sentiment periodsnd low market sentiment periods. We define market sentiment as high (low) when Baker and Wurgler’s market sentimentndex is above (below) the median across all years.

If market-wide sentiment is high, then we predict that the portfolio of firms with high firm-level customer sentimentill be even more overvalued, resulting in more negative loadings on the HML factor. On the other hand, if market-wide

entiment is low, the high firm-level customer sentiment effect will be offset by the low market sentiment, resulting in lessvervaluation and less negative loadings on HML factor. Our findings support this prediction. The loading on HML factor is1.076, significant at 1% level, when market sentiment is high. However, the loading on HML factor becomes a statistically

nsignificant −0.090 when market sentiment is low. The difference between these two periods is also significant, even thoughhe sample size is small (see internet appendix Table A.7).15 Thus, while brand Prestige may be somewhat persistent, its effectaries across time with overall market-wide sentiment.

.7. The influence of the ownership structure on HML loadings

We next see whether the effect of Prestige varies by the ownership structure of the firm. If the glamour affect associatedith Prestige is predominantly driven by retail investors, then we would expect the HML loading on high Prestige stocks to

ttenuate as institutional ownership increases. Using Thomson Reuters’ Institutional (13F) holdings data, we double sorttocks into nine portfolios base on terciles of institutional ownership and Prestige and rerun our factor regressions. Theesults are reported in Panel B of Table 6.


Focusing on the HML loadings for the high Prestige portfolios we see the loading is significantly negative for the lownstitutional holdings group and significantly positive for the high institutional holdings group. Put simply, the glamourffect associated with high Prestige firms dissipates, and even reverses, as institutional holdings increase. Presumably this

15 We also run the regressions pooled and use “high sentiment dummy” interacted with each factor. We find the interactive of this dummy and HML is0.986 and significant at the 5% level, while the interactions with MKT and SMB are insignificant.

dx.doi.org/10.1016/j.jebo.2014.03.014



Table 6Factors that influence the link between prestige and HML loadings. This table presents three-factor regressions of value-weighted monthly excess returnsfor portfolios of firms sorted first by either Familiarity (Panel A) or institutional holdings (IH) (Panel B) and then second by Prestige (panel A) or by Familiarity(panel B). Excess return is defined in Table 3. The definition of Prestige and Familiarity is described in Table 1. Institutional holdings are defined as thepercentage institutional ownership compared with total shares outstanding. The institutional holdings are obtained from 13F. The explanatory variablesare defined in previous tables. The sample period is from January 2000 through December 2006. The heteroscedasticity-robust t-statistics are reported inparentheses under the estimates. The significance of the difference of HML loadings is based on Wald test.

Panel A

Familiarity Low Medium High

Prestige Low Medium High Low Medium High Low Medium High

MKT 1.180*** 1.516*** 1.228*** 0.863*** 1.053*** 1.318*** 1.106*** 0.977*** 0.77***

(11.960) (9.410) (8.540) (10.077) (12.464) (7.624) (8.529) (9.865) (15.581)SMB −0.255* −0.572** 0.190 −0.243** −0.156* −0.156 −0.707*** −0.197 −0.260***

(−1.687) (−2.518) (0.894) (−2.394) (−1.867) (−0.758) (−4.204) (−1.599) (−3.815)HML 0.117 −0.094 −0.865*** 0.356*** −0.117 −0.766*** 0.381** −0.173 0.100

(0.725) (−0.378) (−3.129) (3.011) (−1.039) (−3.428) (2.193) (−0.961) (0.845)Difference between low and high 0.982*** 1.122*** 0.281Constant −0.001 0.000 −0.002 0.000 0.005** 0.002 0.002 0.004 −0.000

(−0.354) (0.060) (−0.365) (0.091) (2.575) (0.451) (0.471) (1.272) (−0.625)Obs. 84 84 84 84 84 84 84 84 84R-squared 0.755 0.653 0.630 0.621 0.813 0.701 0.581 0.739 0.758

Panel B: institutional holdings and Prestige

IH Low Medium High

Prestige Low Medium High Low Medium High Low Medium High

MKT 1.108*** 1.040*** 0.886*** 0.878*** 1.130*** 1.023*** 1.156*** 1.077*** 0.910***

(7.629) (6.996) (6.098) (8.871) (12.338) (10.587) (11.997) (11.658) (12.194)SMB −0.689*** −0.169 −0.082 −0.366*** −0.468*** −0.260** 0.043 0.053 0.020

(−3.143) (−0.708) (−0.495) (−3.509) (−3.794) (−1.995) (0.360) (0.453) (0.171)HML 0.463** −0.547* −0.754*** 0.320*** −0.010 −0.395** 0.229** 0.273** 0.480***

(2.270) (−1.970) (−5.649) (2.872) (−0.054) (−2.582) (2.105) (2.537) (3.816)Difference between low and high 1.217*** 0.715*** −0.251Constant −0.000 0.009* 0.001 −0.000 0.001 0.001 −0.001 −0.005 −0.000

(−0.021) (1.954) (0.377) (−0.081) (0.491) (0.216) (−0.190) (−1.529) (−1.285)Obs. 84 84 84 84 84 84 84 84 84R-squared 0.572 0.637 0.666 0.711 0.766 0.796 0.730 0.708 0.572



is due to the preponderance of retail investors in the low institutional holdings group and perhaps due to the institutionsavoiding high Prestige stocks when the valuation is unattractive. When we examine the difference between the loadings onthe high and low Prestige portfolios we see it is significant for the low and medium institutional holdings groups but not forthe high institutional holdings group. These results are unlikely to be driven by a risk factor associated with Prestige giventhis risk would also have to explain the differences across institutional holdings.

4. Conclusions

We posit that customer awareness and perceptions in the product market may influence investors in the stock market.Using a unique dataset based on customer opinions about their perception of a company’s brands, we partition firms basedon whether their brands have high or low Familiarity and Prestige in the product market. We find that Prestige in the productmarket associates with glamour in the stock market, as evidenced by the HML factor loadings. We explore common riskexplanations for the HML loadings, which fail to fully explain this association between glamour brands and glamour stocks.

We test whether better informed customers are less likely to suffer from sentiment bias by examining the Prestigeeffect on relatively unfamiliar versus highly familiar brands. We find that the Prestige effect diminishes in Familiarity and isabsent in the high Familiarity sub group. Moreover, the effect appears limited to periods where overall market-wide investorsentiment is high. We also find this Prestige channel dissipates as Prestige changes over time and as institutional holdingsincrease and retail investors are more likely to influence valuation. Turning to tests not based on factor loadings, we find that


a negative correlation between Prestige and future returns and we find Prestige associates with overvaluation based on thevalue decomposition of Rhodes-Kropf et al. (2005). Overall these results point in the direction of product market sentimentspilling over into investor sentiment.

dx.doi.org/10.1016/j.jebo.2014.03.014

G ModelJ

it

A

j

R

AAAAAABBBB

BBBCC

CCC

CC

CCCDDDDDFFFFFFGGGH

H

H

HHK

KKLL

LLM

MNOPP



Our results have important implications in the debate about whether the explanatory power of the HML factor on returnss driven by an underlying risk factor or by investor misvaluation. While both effects are likely at work, our results suggesthe product market channel may be one avenue where investor sentiment develops.

ppendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.ebo.2014.03.014.

eferences

aker, D.A., 2004. Brand Portfolio Strategy. Free Press, New York.aker, D.A., Jacobson, R., 1994. The financial information content of perceived quality. J. Market. Res. 31, 191–201.aker, D.A., Joachimsthaler, E., 2000. Brand Leadership. Free Press, London.ltman, E.I., 1968. Financial ratios, discriminant analysis, and the prediction of corporate bankruptcy. J. Finance 23, 589–609.ltman, E.I., 1977. The Z-Score Bankruptcy Model: Past, Present, and Future. New York, Wiley.nderson, E.W., Fornell, C., Mazvancheryl, S.K., 2004. Customer satisfaction and shareholder value. J. Market. 68, 172–185.aker, M., Stein, J., 2004. Market liquidity as a sentiment indicator. J. Financial Markets 7, 271–299.aker, M., Wurgler, J., 2006. Investor sentiment and the cross-section of stock returns. J. Finance 61, 651–707.aker, M., Wurgler, J., 2007. Investor sentiment in the stock market. J. Econ. Perspect. 21, 129–151.arber, B., Odean, T., 2008. All that glitters: the effect of attention and news on the buying behavior of individual and institutional investors. Rev. Financial

Stud. 21, 785–818.arker, D., Loughran, T., 2007. The geography of stock market returns. J. Behav. Finance 8, 177–190.erthon, P., Hulbert, J.M., Leyland, F.P., 1999. Brand management prognostications. Sloan Manage. Rev. 40, 56–65.rown, G.W., Cliff, M.T., 2005. Investor sentiment and asset valuation. J. Bus. 78, 405–440.arhart, M., 1997. On persistence in mutual fund performance. J. Finance 52, 57–82.haudhuri, A., Holbrook, M.B., 2001. The chain of effects from brand trust and brand affect to brand performance: the role of brand loyalty. J. Market. 65,

81–93.hemmanur, T.J., Yan, A., 2009. Advertising, Attention, and Stock Returns. SSRN Working Paper.hen, N., Zhang, F., 1998. Risk and return of value stocks. J. Bus. 71, 501–535.hen, H., Noronha, G., Singal, V., 2004. The price response to S&P 500 index additions and deletions: evidence of asymmetry and a new explanation. J.

Finance 59, 1901–1930.hou, P., Wei, K.C., Chung, H., 2007. Sources of contrarian profits in the Japanese stock market. J. Empiric. Finance 14, 261–286.ooper, I., Kaplanis, E., 1994. Home bias in equity portfolios, inflation hedging, and international capital market equilibrium. Rev. Financial Stud. 7,

45–60.ooper, I., 2006. Asset pricing implications of non-convex adjustment cost of investment. J. Finance 61, 139–170.ooper, M.J., Gulen, H., Schill, M.J., 2008. Asset growth and the cross-section of stock returns. J. Finance 53, 1609–1651.oval, J.D., Moskowitz, T.J., 1999. Home bias at home: local equity preference in domestic portfolios. J. Finance 54, 1–39.aniel, K., Hirshleifer, D., Subrahmanyam, A., 2005. Investor Psychology and Tests of Factor Pricing Models. Working Paper.aniel, K., Titman, S., 1997. Evidence on the characteristics of cross-sectional variation in common stock returns. J. Finance 52, 1–33.aniel, K., Titman, S., 2006. Market reactions to tangible and intangible information. J. Finance 61, 1605–1643.e Bondt, W.F.M., Thaler, R., 1985. Does the stock market overreact? J. Finance 40, 793–805.ichev, I.D., 1998. Is the risk of bankruptcy a systematic risk? J. Finance 53, 1131–1147.ama, E.F., French, K.R., 1992. The cross-section of expected stock returns. J. Finance 47, 427–465.ama, E.F., French, K.R., 1993. Common risk factors in the return on bonds and stocks. J. Financial Econ. 33, 3–53.ama, E.F., French, K.R., 1995. Size and book-to-market factors in earnings and returns. J. Finance 50, 131–155.ang, V.H., Peress, J., 2009. Media coverage and the cross-section of stock returns. J. Finance 64, 2023–2052.rench, K.R., Poterba, J.M., 1991. Investor diversification and international equity markets. Am. Econ. Rev. 81, 222–226.rieder, L., Subrahmanyam, A., 2005. Brand perceptions and the market for common stock. J. Financial Quant. Anal. 40, 57–85.riffin, J.M., Lemmon, M.L., 2002. Book-to-market equity, distress risk, and stock returns. J. Finance 57, 2317–2336.ruca, T.S., Rego, L.L., 2005. Customer satisfaction, cash flow, and shareholder value. J. Market. 69, 115–130.rullon, G., Kanatas, G., Weston, J.P., 2004. Advertising, breadth of ownership, and liquidity. Rev. Financial Stud. 17, 439–461.ellofs, L., Jacobson, R., 1999. Market share and customers’ perceptions of quality: when can firms grow their way to higher versus lower quality? J. Market.

63, 16–25.ertzel, M.G., Li, Z., 2010. Behavioral and rational explanations of stock price performance around SEOs: evidence from a decomposition of market-to-book

ratios. J. Financial Quant. Anal. 45, 935–958.irshleifer, D., Jiang, D., 2010. Commonality in misvaluation, equity financing, and the cross-section of stock returns. Rev. Financial Stud. 23,

3401–3436.irshleifer, D., Hou, K., Teoh, S., Zhang, Y., 2004. Do investors overvalue firms with bloated balance sheets? J. Account. Econ. 38, 297–331.uberman, G., 2001. Familiarity breeds investment. Rev. Financial Stud. 14, 313–357.eloharju, M., Knüpfer, S., Linnainmaa, J.T., 2012. Do investors buy what they know? Product market choices and investment decisions. Rev. Financial Stud.

25, 2921–2958.eller, K.L., 1993. Conceptualizing, measuring, and managing customer-based brand equity. J. Market. 57, 1–22.eller, K.L., 2003. Strategic Brand Management: Building, Measuring, and Managing Brand Equity, 2nd ed. Prentice Hall, Englewood Cliffs, NJ.akonishok, J., Shleifer, A., Vishny, R.W., 1994. Contrarian investment, extrapolation, and risk. J. Finance 49, 1541–1578.ane, V., Jacobson, R., 1995. Stock market reactions to brand extension announcements: the effects of brand attitude and familiarity. J. Market. 59,

63–77.emmon, M., Portniaguina, E., 2006. Consumer confidence and asset prices: some empirical evidence. Rev. Financial Stud. 19, 1499–1529.ou, D., 2010. Attracting Investor Attention Through Advertising. SSRN Working Paper 1571947.adden, T.J., Fehle, F., Fournier, S.M., 2006. Brands matter: an empirical investigation of brand building activities and the creation of shareholder value. J.

Acad. Market. Sci. 34, 224–235.


cAlister, L., Srinivasan, R., Kim, M., 2007. Advertising, research and development, and systematic equity risk of the firm. J. Market. 71, 35–48.ewman, J.W., Werbel, R.A., 1973. Multivariate analysis of brand loyalty for major household appliances. J. Market. Res. 10, 363–389.hlson, J.A., 1980. Financial ratios and the probabilistic prediction of bankruptcy. J. Account. Res. 18, 109–131.agan, A., 1984. Economic issues in the analysis of regressions with generated regressors. Int. Econ Rev. 25, 154–193.ástor, L’., Stambaugh, R.F., 2003. Liquidity risk and expected stock returns. J. Polit. Econ. 111, 642–685.

dx.doi.org/10.1016/j.jebo.2014.03.014

http://dx.doi.org/10.1016/j.jebo.2014.03.014

http://dx.doi.org/10.1016/j.jebo.2014.03.014

http://refhub.elsevier.com/S0167-2681(14)00084-5/sbref0005









































































































































































































































































































































































































































































































































































































































































































































































































































G Model

Rego, L.L., Billett, M.T., Morgan, N.A., 2009. Consumer-based brand equity and firm risk. J. Market. 73, 47–60.Rhodes-Kropf, M., Robinson, D.T., Viswanathan, S., 2005. Valuation waves and merger activity: the empirical evidence. J. Financial Econ. 77, 561–603.


Srivastava, R., Shervani, T., Fahey, L., 1998. Market-based assets and shareholder value: a framework for analysis. J. Market. 62, 2–18.Srivastava, R., Shervani, T., Fahey, L., 1999. Marketing, business processes, and shareholder value: an organizationally embedded view of marketing activities

and the discipline of marketing. J. Market. 63, 168–179 (Special Issue).Tuli, K.R., Bharadwaj, S., 2009. Customer satisfaction and stock returns risk. J. Market. 73, 184–193.Zhang, L., 2005. The value premium. J. Finance 60, 67–103.

dx.doi.org/10.1016/j.jebo.2014.03.014


























































































glamour brands and glamour stocks

Documents