Tarun Kushwaha & Venkatesh Shankar Are Multichannel ... · Venkatesh Shankar is Professor of Marketing and Coleman Chair in Marketing, Mays Business Schooi, Texas A&M Uni-versity

Tarun Kushwaha & Venkatesh Shankar

Are Multichannel Customers ReallyMore Valuable? The Moderating Roleof Product Category Characteristics

How does the monetary value of customer purchases vary by customer preference for purchase channels (e.g.,traditional, electronic, multichannel) and product category? The authors develop a conceptual model andhypotheses on the moderating effects of two key product category characteristics—the utilitarian versus hedonicnature of the product category and perceived risk—on the channel preference-monetary value relationship. Theytest the hypotheses on a unique large-scale, empirically generalizable data set in the retailing context. Contrary toconventional wisdom that all multichannel customers are more valuable than single-channel customers, the resultsshow that multichannel customers are the most valuable segment only for hedonic product categories. The findingsreveal that traditional channel customers of low-risk categories provide higher monetary value than othercustomers. Moreover, for utilitarian product categories perceived as high (low) risk, web-only (catalog- or store-only) shoppers constitute the most valuable segment. The findings offer managers guidelines for targeting andmigrating different types of customers for different product categories through different channels.

Keywords: customer relationship management, channels, multichannel marketing, retailing

M anaging customers according to their channel pref-erence—that is, whether they purchase from a tradi-tional channel (e.g., catalog, store), an electronic/

digital channel (e.g., web, mobile), or multiple channels—has become a cornerstone of marketing strategy (Neslin etal. 2006). Multichannel marketing refers to the practice ofsimultaneously offering customers information, goods, ser-vices, and support through two or more synchronized chan-nels (Rangaswamy and Van Bruggen 2005). The highgrowth in retail sales through electronic and multiple chan-nels indicates a need for marketing managers and scholarsto develop a deeper understanding of this important topic ina retailing context that includes direct marketing (catalog,web, mobile) retailers and brick-and-mortar stores.

Conventional wisdom, shaped by anecdotal evidenceand initial research studies, suggests that multichannel cus-tomers constitute the most valuable segment for marketers

Tarun Kushwaha is Assistant Professor of Marketing, Kenan-FlaglerSchool of Business, University of North Carolina at Chapei Hiii (e-maii:[email protected]). Venkatesh Shankar is Professor of Marketingand Coleman Chair in Marketing, Mays Business Schooi, Texas A&M Uni-versity (e-mail: [email protected]). The authors thank i-Behaviorfor data and the Direct Marketing Education Foundation for doctoral sup-port. This article is based on an essay from the first author's dissertation.The authors thank three anonymous reviewers; seminar participants atthe Marketing Science Conference, Haring Symposium, University ofHouston Doctoral Symposium, University of North Carolina at Chapel Hill,Syracuse University, Georgia Institute of Technology, University of Illinois,and McGill University; Roger Kerin; Bill Perreault; Alina Sorescu; RajanVaradarajan; and Manjit Yadav for helpful comments. Address all corre-spondence to Venkatesh Shankar. Aric Rindfleisch served as area editorfor this article.

regardless of the product category. For example, the U.S.-based multichannel retailer Nordstrom finds that acrosscategories, customers who use more than one channel spendfour times as much as those who shop only through onechannel (Clifford 2010). The limited relevant scholarly arti-cles that typically analyze a single category demonstratethat multichannel customers purchase more often and spenda larger share of wallet than single-channel customers(Kumar and Venkatesan 2005; Venkatesan, Kumar, andRavishanker 2007).

It is unclear, however, whether these results generalizeto all product categories, which are typically classifiedalong two key characteristics: utilitarian (e.g., office sup-plies, garden supplies) versus hedonic (e.g., apparel, cos-metics) and low perceived risk (e.g., books, home furnish-ings) versus high perceived risk (e.g., computers, jewelry).Are multichannel customers the most valuable for utilitar-ian, hedonic, high-risk, or low-risk categories? This issue isimportant because customer behavior fundamentally variesby these product category types (Ailawadi et al. 2006;Ailawadi, Lehmann, and Neslin 2003; Inman, Winer, andFerraro 2009; Kamakura and Du 2012; Narasimhan, Neslin,and Sen 1996). For example, customers of a hedonic prod-uct category may seek variety and spend more on items inthat category across different channels. In contrast, cus-tomers of utilitarian categories may want to shop efficientlyin one channel and spend more in that channel. Similarly,low- (high-) risk product categories may attract customersof traditional (electronic) channels and induce them tospend more in those channels. However, because muchresearch on multichannel customer behavior is based ondata from a single product category or firm, it precludes the

© 2013, American Marketing AssociationISSN: 0022-2429 (print), 1547-7185 (electronic) 67

Journal of MarketingVoiume 77 (July 2013), 67-85

study of the product category's role in the monetary valueof shoppers' purchases by channel preference.

We define a multichannel customer of a broad productcategory as a customer who buys items in that categoryfrom more than one channel.' B> viewing multichannelshopping from the customer angle, our approach provides aholistic view of a customer's behavior. We address twoimportant research questions in the .retailing context:

1. How does the monetary value of purchases by multichannelcustomers differ from that of single-channel customers?

2. How does the relationship between a customer's channelpreference and monetary value -'ary by key product cate-gory characteristics (utilitarian v;. hedonic nature and per-ceived risk)?

The answers to these questions are critical from boththeoretical and managerial standpoints. For example, if themonetary value of multichannel customers is higher thanthat of single-channel customers across all categories, mar-keters should reach customers of all categories through dif-ferent channels. Similarly, if web-only shoppers are themost valuable channel segment for high-risk/utilitariancategories (e.g., computers, electronics), marketers shouldtarget these shoppers. Finally, if traditional channel shop-pers of low-risk categories (e.g., office supplies, gardensupplies) provide higher monetary value than multichannelor web-only shoppers, marketers should focus on theseshoppers.

To address these research questions, we develop a con-ceptual framework and important hypotheses related to themoderating role of the two key product category character-istics—utilitarian versus hedonic nature and perceivedrisk—on the link between channel preference and customermonetary value. We test our hypotheses and obtain empiri-cally generalizable insights by andyzing a unique large-scale, cross-sectional data set of 1 million customers ran-

'In our subsequent empirical analysis, we test for alternativedefmitions of multichannel customers and show that our resultsare robust to alternative definitions.

domly drawn from 96 million customers of 750 direct mar-keting retailers, spanning 22 product categories across thecatalog and web channels over a four-year period. We gen-eralize the results to the store channel with a longitudinalanalysis of transaction data from a large multiproductretailer with the store channel in addition to the catalog andweb channels.

Our results show that contrary to conventional wisdom,multichannel customers form the most valuable segmentonly for hedonic product categories. We also find that tradi-tional channel customers of low-risk product categoriesprovide higher monetary value than other customers. More-over, web-only (store- or catalog-only) customers of high-(low-) risk/utilitarian categories offer higher monetaryvalue than other single-channel or multichannel customers.Our findings provide valuable managerial guidelines forshopping in different channels.

This article contributes to the literature in at least twokey ways. First, it offers a theoretical understanding of themoderating effects of category characteristics on the chan-nel preference-monetary value relationship. Second, it pro-vides empirically generalizable counterintuitive findingsabout this relationship. Furthermore, as Table 1 shows,unlike related research (e.g., Ansari, Mela, and Neslin 2008;Kumar and Venkatesan 2005), the current research exam-ines the moderating effects of the product category and ana-lyzes data from multiple categories and firms in an inte-grated framework.

Conceptual DevelopmentWe first develop a conceptual model of the relationshipsbetween channel preference, product category characteris-tics, and monetary value (see Figure 1). We focus on bothtraditional and electronic channels. We classify the storeand catalog channels under the unifying banner of a tradi-tional channel because they have a much longer historythan electronic channels and are perceived as close substi-tutes (Avery et al. 2012). Any combination of these chan-nels constitutes a multichannel. Among the outcomes, we

FIGURE 1A Conceptual Model of Relationships Between Channel Preference, Monetary Value, and Product

Category Characteristics

Product Category CharacteristicUtilitarian Versus Hedonic Nature

Interactions: H2, Hg

Channel Preference•Trad•Elec

•Mult

itional (catalog, store)tronic (web)channel Main Effect: H,

Product Category CharacteristicLow Versus High Perceived Risk

Interactions: H4, H5

tMonetary Value

— Hypothesized effects — Controlled effects

68 / Journal of Marketing, July 2013

.co(0a>(0irc(0

QC

2

UJ (01 à)

t3U

co3

trib

oO

M(/)

AC

I

SmQ

ross

u

o

Î3

EO)

tind

era

oS

co

peti

EoO

ries

oS"

Cat

ta T

ype

o

(/)O

'S5*coo

oO)

wU•5

SioQ.

Q.

Art

i

.1Q.

CO

cou03CO

CO• ocoü

CO

CO• a

ooOJCO

CO

03

OC

:3

oO)COCO

•aeuCO

eucCOx:o

= ^ O

0}

CO

O

>O)eu

§1

f l2 E

CO .̂...̂

> o2!

cuX)

E

IIJ5 o

m Ü

ieu

oc

O

T3

g

s.o

COü

E

r co

mO

ffele

rsre

tai

cucCOsz

uate

1E

val

o

CO

E

CO

OüCU

CO

o i_•r 0)CO c:

iIS

c

11b m

E

CO

E

o i_V 0)COCO0}

CO

enc

o•55

cCO

cuo

CO ü

O OJ

OJ o

cu CO

oj.g£ CO

o

O)

E

cCO

eoeu

ES

-oç

m

1 ï

c co eo

cCOm03

ICO

"BCO

T3

CO

X I

CO o

s OJ

E5— o0 >

s OJ

OJ -hCO t :CO 3

c"en

CO

i.i'C0-2Í

^ CO

" Ea? c.E oE«CO ÜX cu

LU T3

' Lf) CO LO; o 5 o' 2 E o

•o

eo

r: o

O OZ z

CO CO•D T3C CO OÜ OOJ OJ

CO CO

OJ OJ

oc oc

o o

euÜ

I Ieu

05

ICO

I

•o ^

CO CO

CO o

E "

D)ça3 =CO £

ET3

t3

IIQ. C

M- eoO .C

litE c ^— o —c « >

Ü C 'S T:s rv( CD Tñ

•5 p <J T3 s OJ-5 .£ ^ O o C> m ® Ê 2 O

••- O «i

eu c £ï - '•- eo o £

^ CO CO ^ C3J (1)

LU £ en UJ S Q.T3

CO

OJ

Oo

CM

OC\J

C P C -iT.en en

c êneu

^ Z OJÜOJ

Are Multichannel Customers Really More Valuable? / 69

focus on a key managerial variable—that is, the monetaryvalue of customer purchases—measured as the dollar valueof customer transactions.

Prior research has examined the role of the productcategory in influencing different facets of customer behav-ior in traditional and electronic channels (Yadav andVaradarajan 2005b). Consistent with this research, weexamine the moderating role of two key product categorycharacteristics (i.e., utilitarian vs. hedonic nature and per-ceived risk) in shaping the relationship between channelpreference and customer monetary \alue.

We define a utilitarian category as a category dominanton attributes such as functionality, practicality, cognition,and instrumental orientation, consistent with Dhar andWertenbroch (2000). Computing equipment, consumerelectronics, office supplies, home appliances, and gardenequipment are examples of utilitarian categories. We definea hedonic category as a category dominant on attributessuch as experiential benefits, affect, enjoyment, enduringinvolvement, intrinsic motivation, and aesthetics (Dhar andWertenbroch 2000). Examples cf hedonic categoriesinclude CDs, DVDs, antiques, and ipparel. Unlike hedonicproducts, utilitarian products can be easily compared andevaluated along different attributes.

We refer to perceived risk of a product category as the"customers' (overall) perceptions of uncertainty andadverse consequences of buying a good (or service)"(Dowling and Staelin 1994, p. 119; see also Bart et al.2005). Perceived risk of a product category is evaluated onfive dimensions of uncertainty: functional (not performingto expectation), financial (losing money), safety (causingphysical harm), psychological (tarnishing self-image), andsocial (lowering others' perceptions of the user) (Jacobyand Kaplan 1972). Office supplies and books are examplesof low-risk categories, while jewelry and computers areexamples of high-risk categories.

We chose these two product category characteristics aspotential moderators from two maia theoretical considera-tions. First, both category characteristics are grounded inregulatory focus theory (RFT), the basis for our hypothesesdevelopment. Prior research in marketing (see Chemev2004; Yeo and Park 2006) treats hedonic (utilitarian) andhigh (low) perceived risk attributes as consistent with a pro-motion (prevention) focus in goal orientation, the key ingre-dients of RFT.

Second, utilitarian versus hedonic nature and perceivedrisk constitute fundamental bases for consumer purchaseand consumption. Batra and Ahotla (1990, p. 159) state that"consumers purchase goods and ser/ices and perform con-sumption behaviors for two reasons: (1) consummatoryaffective (hedonic) gratification (from sensory attributes)and (2) instrumental (utilitarian) reasons." Likewise, thestudy of perceived risk as an inherent product categorycharacteristic behind purchase and consumption behaviorhas a long tradition in the marketing literature (e.g.. Coxand Rich 1964; Sheth and Venkatesai 1968).

Our conceptual model includes t!ie direct/main effect ofchannel preference and the interaction effects of channelpreference and product category characteristics on mone-tary value. We develop hypotheses pertaining to these

effects. Our overarching argument is that different shoppershave different foci based on RFT, and if shoppers' focus fitswith their channel preference based on the product categorycharacteristics, they will experience greater regulatory fit.In tum, stronger regulatory fit will lead to higher spendingin their preferred channel on the product categories thatexhibit those characteristics.

Main Effect

We first develop our hypothesis about the main effect ofcustomer channel preference on the monetary value of pur-chases across categories. An entity (e.g., customer) evalu-ates the outcome of an exchange process with another entity(e.g., firm) by comparing the perceived benefits with theperceived costs related to the exchange, consistent with thequid pro quo notion (Bagozzi 1975). In addition to eco-nomic aspects, social and psychological aspects (e.g.,mutual respect, commitment, trust) play an important rolein determining the entities' perceived overall benefits, costs,and value in an exchange (Frazier 1999).

Depending on their perceived value of an exchangethrough a channel, customers prefer to use different chan-nels and spend different amounts in different channels. Cus-tomers who perceive exchanges in a channel as being ofhigh value become frequent customers with a high degreeof trust and commitment to purchase through that channel.Customers with a stronger commitment spend more on theirpurchases than other customers (Venkatesan, Kumar, andRavishanker2007).

The use of multiple channels is associated with a highlevel of monetary value for customers across all productcategories for several reasons. First, additional channelsprovide greater convenience value for customers, increasingtheir purchase frequency and accelerating purchases acrossmultiple items and categories. Second, the wide assortmentof products across different channels offers multiple oppor-tunities for customers to buy and increase their spending.Third, customers can combine the benefits from differentchannels, realize greater value, and increase their spending(Frazier 1999). The web and traditional channels are comple-mentary rather than cannibalistic with regard to the moneyspent on shopping (Deleersnyder et al. 2002). Thus, channelsegment membership (single-channel or multichannel) is aproxy for customers' perceived value of and commitment tothat channel. This commitment is positively related to cus-tomers' spending in that channel across products. Therefore:

H]! Across all product categories, multichannel customershave a higher monetary value of purchases than single-channel customers.

Moderating impact of Product CategoryCharacteristics

We now develop hypotheses on how the hedonic versusutilitarian nature and perceived risk moderate the strengthof the relationship between channel preference and mone-tary value. We present a summary of the hypothesestogether with the associated rationale in Table 2.

We adopt RFT to motivate our hypotheses. According toRFT, people can be classified into two types on the basis of


U 0)

.ii iS« c£ oOSa

(A

SIa>

IQ.

i

áO>

•ao

nE

0)

i illlili

ü E

"- «o fr

111

c

0)•oo

"îâc

tio

21 -

o

g"

atai

U

u Q. ™C « ^ 3

S .2 ^

ê«5

O

=Ü

A

ce

DJoc"

Ace

ceIUJ

ce"X

3en

encO

55

O)

in .2 .^to CO c 50 ü 0 g,O c u 0c CO JE >-

_(Ô o

^ I i •̂ E

l l l l

i55

g £ .1 £•co e n « . TS o3 .£ en" SJ, 0 CO

£ E S | «? ?J ÇU m ä

3 COQ O• K O -s; D. O

tÓ E .O

CO0

0

en

o

0

O O C C

^5 -"-H i rE 0. O)

eo c. *^ Al « co 0' ^ .y J3 S: i l

IS jaP S:-ti

oJO

ICO

ora3en£O)co55

t-o553en0CO

I

0 .E Q.

enw ô.

,O R 0lA.« E.9 0 .£c - S

S o gQ. oc .2

il!

Co• o

Xcô

X

oen0

Ü

5Äo to-I o

CO0

X uin

O)

Are iVluitichannei Customers Reaiiy More Vaiuabie? / 71

their regulatory orientation in pursuing a goal: preventionfocused and promotion focused (Avnet and Higgins 2006).A prevention focus stresses safety, security, and responsibil-ity, whereas a promotion focus emphasizes hope, advance-ment, and achievement. Thus, a promotion focus involvesmaximizing positive outcomes, whereas a prevention focusmeans minimizing negative outcomes. As RFT postulates,people make choices that are consistent with their regula-tory orientation (promotion or prevention focus) in goalpursuit (Avnet and Higgins 2006). When such choices sus-tain their regulatory orientation, people experience a regula-tory fit, leading them to continue their pursuits (Aaker andLee 2006). Thus, a customer is likely to engage repeatedlyin his or her preferred buying process (channel) if doing sois consistent with his or her regulatory orientation (Avnetand Higgins 2006). Regulatory fit leads to greater customerengagement through two well-documented processes(Cesario, Higgins, and Scholer 2008): (I) feeling rightabout the task and (2) increased information processing.These engaged customers tend to value and pay more forproducts than customers lacking in regulatory fit (Avnet andHiggins 2006). Therefore, promotion- and prevention-focused customers will tend to spend more on their pur-chases in their preferred channel.

Some channels are closely associated with a preventionfocus, whereas others are aligned with a promotion focus.Because of their long history, traditional channels (i.e., cata-log and store) offer high levels of familiarity, safety, confi-dence, and trust. In the case of physical stores, customerscan browse, touch, and feel products before purchase. Fur-thermore, many catalog companies and physical stores havehad a long-standing practice of accepting returns from cus-tomers without asking questions. Therefore, traditionalchannels offer customers high confidence and trust in theirpurchases. Thus, catalog- and store-only customers arelikely to have a high prevention focus as they repeatedlypatronize traditional channels that offer high levels ofsafety, minimizing negative outcomes.

In contrast, the relatively newer electronic channels(e.g., the web) evoke high behavioral and environmentaluncertainty (Schlosser et al. 2006; Van Noort, Kerkhof, andFennis 2008). A large-scale survey by the Pew Internet Project(2008) shows that the web is perceived as highly uncertainsuch that approximately three-quarters of participants wereunwilling to provide personal and credit card informationover the Internet. Furthermore, the web entails the risk ofidentity theft, a significant deterrent to online channel adop-tion (Garver 2012). Despite more than a decade since theadvent of electronic commerce, the adoption of the web as atransaction channel is still limited because of a high level ofperceived risk. Therefore, web-only customers who repeat-edly patronize the electronic channel are likely to be drivenby adventure and the need to signal advancement—that is,by focusing on positive outcomes. Moreover, web-onlyshoppers tend to be younger, better educated, and moreprone to search on the web than other shoppers. Thus, thesecustomers are likely to have a greater promotion focus.

Customers who adopt multiple channels seek greaterconvenience and display boldness associated with the adop-tion of the electronic channel. They seek greater enjoyment

and adventure through the use of different channels. There-fore, multichannel customers are also likely to have agreater promotion focus.

In summary, customers patronizing traditional modes oftransaction (catalog and store) are likely to have a greaterprevention focus. In contrast, customers adopting nontradi-tional modes (web and multiple channels) are likely to havea greater promotion focus. However, product category char-acteristics moderate the monetary values of customers bychannel preference.

Utilitarian versus hedonic nature and channel preferenceinteraction. Because utilitarian products (e.g., computers,garden equipment, sports equipment) have clear and well-defined attributes, they are relatively easy to compare andevaluate. Thus, for utilitarian product categories, shoppingtasks involve planned purchases, goal-directed choice, andcognitive involvement (Novak, Hoffman, and Duhacheck2003). Goal-directed behavior can lead to habit formationand automatic behavior (Aarts and Dijksterhuis 2000). Inaddition, utilitarian categories are typically high on search-dominant attributes. People allocate time, a scarce resource,to different activities, including search (Becker 1965).Scarcity of time is negatively linked to search efforts(Beatty and Smith 1987). Therefore, consumers of utilitar-ian products value efficiency in shopping (Mathwick, Mal-hotra, and Rigdon 2002). In general, efficiency attributesare associated with a greater prevention focus (Chernev2004). Goal-oriented shopping behavior associated withutilitarian product categories is efficient in time utilizationwhen both search and purchase are done habitually andrepeatedly in a single channel. Because efficiency is para-mount, these customers prefer using a single channel tomultiple channels.

Customers of traditional channels have a greater preven-tion focus, which maps with the prevention-focus attributesof utilitarian product categories, providing a strong channel-category fit. In contrast, because multichannel customersare promotion focused on utilitarian products, they experi-ence a relatively weak regulatory fit. A stronger regulatoryfit is associated with greater engagement and higher spend-ing, and thus we expect traditional channel customers ofutilitarian products to spend more than their multichannelcounterparts.

Although web-only and multichannel customers arelikely to have a greater promotion focus, which provides aweak regulatory fit with the utilitarian attributes, the greaterefficiency of the web maps well with the shopping goalsassociated with utilitarian categories. Thus, web-only cus-tomers of utilitarian categories are also likely to have bettergoal-attribute fit than their multichannel counterparts.Taken together, we expect that single-channel customers ofutilitarian product categories have higher monetary valuesthan multichannel customers. Thus:

H2: The monetary value of purchases by single-channel cus-tomers of utilitarian product categories is higher than thatby multichannel customers of these categories.

Hedonic categories, such as apparel, cosmetics, andDVDs, are conducive to unplanned or impulse buying andvariety seeking (Novak, Hoffman, and Duhacek 2003).


Impulse purchases are characterized by spontaneity, com-pulsion, excitement, and disregard for consequence (Kou-faris 2002). Variety-seeking behavior is, in part, driven byfactors such as product category-specific differences (VanTrijp, Hoyer, and Inman 1996). Customers of hedonic prod-uct categories likely have high goal ambiguity because ofaffect-dominant attributes and the salience of the experien-tial value of hedonic products. Such goal ambiguity leadscustomers of hedonic categories to include disparate prod-ucts in their consideration set (Ratneshwar, Pechmann, andShocker 1996) and seek variety. In addition, perceiveduncertainty about future preferences is likely to be higherfor hedonic products, leading to variety seeking as a choiceheuristic (Simonson 1990). In general, hedonic attributes areassociated with a greater promotion focus (Chemev 2004).Thus, customers are more likely to engage in variety-seekingbehavior for hedonic categories and spend more (Garg,Inman, and Mittal 2005; Kurt, Inman, and Argo 2011; Ratnerand Kahn 2002; Stilley, Inman, and Wakefield 2010a, b).

Multichannel customers are likely to have a greater pro-motion focus, which provides a strong regulatory fit withhedonic attributes. Multiple channels provide a greaterassortment of products than a single channel. More hedonicproducts across multiple channels offer customers moreopportunities to engage in impulse purchases, enhance cus-tomers' consideration set, and promote greater variety seek-ing. In contrast, the prevention focus of traditional channelcustomers has a weak regulatory fit with hedonic attributes.Thus, multichannel customers are likely to be more stronglyengaged and have a higher spending level than their coun-terparts from traditional channels.

Web-only customers are also likely to have a greaterpromotion focus, which provides a strong regulatory fitwith hedonic attributes. However, the use of multiple chan-nels offers greater convenience and variety, which are moresatisfying for variety-seeking and impulse purchase behav-iors commonly involved in hedonic categories. Thus, multi-channel customers of hedonic categories are likely to have abetter goal-attribute fit than their web-only counterparts.

In summary, we expect that multichannel customers ofhedonic product categories have higher monetary valuethan their single-channel counterparts. These argumentslead to the following hypothesis:

H3: The monetary value of purchases by multichannel cus-tomers of hedonic product categories is higher than thatby single-channel customers of these categories.

Perceived risk and channel preference interaction. Forproduct categories with high perceived risk, such as elec-tronics, telecommunications equipment, and musical instru-ments, customers face considerable uncertainty. Therefore,product categories with high perceived risk fit the goal ori-entation of promotion-focused customers (Yeo and Park2006). A promotion focus is consistent with risk-seekingbehavior (Avnet and Higgins 2006). Conversely, a preven-tion focus is synonymous with risk-averse behavior and fitswith low-risk product categories, such as office supplies,books, and toys.

Because of the channel-category fit, low-risk categorieslikely attract prevention-focused customers who shop in tra-

ditional channels, whereas high-risk categories may drawpromotion-focused customers, such as those shopping on theweb or through multiple channels. Consistent with the notionthat the perceived risk of the web is due to the relative new-ness and impersonal nature of the channel (Montoya-Weiss,Voss, and Grewal 2003), we expect that web-only cus-tomers have a greater promotion focus than a preventionfocus. Similarly, because multichannel customers may seekgreater enjoyment and adventure through different chan-nels, they are likely to have a greater promotion focus. Incontrast, because of the low-risk profiles of the store andcatalog channels, single-channel customers are likely tohave a greater prevention focus.

With the high degree of fit among prevention focus,low-risk categories, and traditional channels, traditionalchannel customers tend to spend more than web-only ormultichannel customers in low-risk categories. In contrast,web-only and multichannel customers of high-risk categorieslikely spend more than customers of traditional channelsbecause of the high degree of fit between these categoriesand channels. These arguments lead to following hypotheses:

H4: The monetary value of purchases by traditional channelcustomers of low-risk product categories is higher thanthat by electronic channel and multichannel customers ofthese categories.

H5: The monetary value of purchases by electronic channeland multichannel customers of high-risk product cate-gories is higher than that by traditional channel customersof these categories.

Empirical AnalysesWe examine an empirical context comprising a carefullycompiled, unique, and large cross-sectional database ofapproximately 1 million U.S. customers who were ran-domly selected from a cooperative database of 96 millioncustomers of 750 direct marketers covering 22 product cate-gories and several subcategories during a four-year period(2001-2004). We obtained data from i-Behavior, a syndi-cated data aggregator firm. Firms in the cooperative databasehave only the web and catalog channels (no physical storesexist), so the catalog is their offiine channel. The data containcustomers' demographic characteristics, shopping experi-ences, preferred purchase channels, order details, and productcategories purchased. This period adequately captures thegrowth phase of the web as a distribution channel. Detailson these 22 product categories appear in the Web Appendix(WAI; www.marketingpower.com/jm_webappendix).

Our data set is neither firm nor industry specific, and itcaptures customer purchases across a comprehensive set ofproduct categories and competing firms. Such data are highlyrepresentative of the population and allow for empiricalgeneralizability. Data sets from prior research are primarilyfrom a single firm across one or a few product categories.Our database covers a wide range of product categories,such as apparel, accessories, gifts, hobby items, electronics,and musical instruments, for 750 multichannel direct mar-keters, which enables us to develop a richer understandingof a customer's channel preference and behavior than whenanalyzing data from a single firm or a few categories.


Operationalization of Variabies

The operationalization of the variab.es in our data appearsin Table 3. The exogenous classes of variables, such asdemographic characteristics, shopping experience, high-endcatalog usage, and the number of unique mailing lists thatcontain the customer's name, are based on the customers'past transaction history.

Table, 4 provides summary s:atistics for the keyvariables in our model. Of the usable sample (customerswith data on every variable in the database), 71.8% pur-chased only through the catalog channel, 5.3% purchasedonly through the web, and the remaining 22.9% purchasedthrough both channels.2 Although the number of purchasesof web-only shoppers is much smaller than that of catalog-

2We compared the usable sample (n = 412,424) with the unus-able sample on each of the five dependent variables analyzed inthe subsequent sections. The means of the key dependent variablesof the usable and unusable samples are similar.

only and riiultichannel shoppers, web-only retail sales grewby approximately 12% during the 2006-2010 period(Jupiter Research 2011). The summary snapshot suggeststhat multichannel customers spend approximately one and ahalf times more than catalog-only customers and approxi-mately five and a half times more than web-only customers.Similarly, multichannel customers buy more often (higherfrequency) than single-channel customers. However, arethese initial summary observations true for all product cate-gories? We address this question in our empirical analysis.

Measurement of Product Category Ciiaracteristics

We use data from exogenous sources to measure the twokey category characteristics of the 22 product categories.We classify product categories with a higher utilitarianscore than hedonic score as utilitarian, and vice versa forthe purchased product category basket. We classify the cate-gories as high or low risk on the basis of a median splitalong these dimensions for the purchased product category

TABLE 3Operationalization of Variables in the Data

Variable Operationalization

Focal Dependent VariableMonetary value (DLLR)

Other Dependent VariablesChannel preference (WEB, CTLG)

Mailers (MAIL)

Frequency (ORDR)

Product Category Characteristics

Utilitarian versus hedonic (UTL, HED)

Perceived risk (LR, HR)

Control Variables/InstrumentsAge (LAGE)

Family size (FSI2E)

Education (EDU)

Number of high-end catalogs (HIGH)a

Largest past spending (HISPEND)

Relative credit card use (RCCU)^

Returns (RET)

Shopping experience (EXP)

Number of categories bought (CAT}

Unique mailing lists on (UNQML)

Target customer net worth (NTWTH)

Lists responded to (UNQRS)

Total dollars spent by the customer in the four-year data window.

Dummy variables representing web-only and catalog-only with multichannel (bothweb and catalog) as the base. Based on the customer's purchase channel over thedata window.

Number of marketing mailers sent to the customer in the past four yearstransformed to a near normal distribution using Anscombe's (1948) transformation.

Number of orders by the customer in the four-year window transformed to nearnormal distribution using an Anscombe's (1948) transformation.

Dummy variables representing utilitarian and hedonic categories with all categoriesas the base.

Dummy variables representing low- and high-risk categories with all categories asthe base.

The midpoint of the age range to which the customer belongs (seven intervals). Forthe last age range, which is open-ended (75+ years), the lower bound of the rangeis taken as the measure.Number of adults and children in the customer's household.Number of years of education of the customer.Number of high-end catalogs from which the customer ordered.Dollar value of the customer's largest order in the data window.

Percentage of occasions during which the customer used a major credit card.

Number of items the customer returned.

Number of weeks since the customer placed the first order before start of the dataperiod.

Number of different product categories the customer bought.

Number of unique mailing lists on which customer is listed.

The net worth score of a target customer as reported by Claritas on a ten-pointscale.

Percentage of unique mailing lists to which a customer responded.

aData aggregators in the direct marketing industry classify catalogs into five categories on a continuum from "low-scale" to "high-scale" cata-logs. The number of times a customer orders from the highest category of the "high-scale" catalogs is the operationalization of the variable.

t M C d and Visa issued by major banks, American Express, and Discover are classified as major credit cards.


TABLE 4Summary Statistics of Key Variables in the Data

Variable/Item

Sample size (n)Channel preference (%)Monetary value ($)FrequencyaMailersaAgeFamily sizeEducation (years)Number of high-end catalogsLargest past spend in an order ($)Relative credit card useReturnsShopping experience (weeks)Number of categories boughtUnique mailing lists onTarget customer net worthUnique lists responded to (%)

Catalog Only

296,07971.79

1,123.926.895.12

57.222.42

13.31.68

124.54.32.03

157.674.925.776.41

.27

Web Only

21,7765.28

477.693.542.73

45.992.63

13.96.55

201.56.44.01

80.632.161.396.56

.02

Multichannel

94,56922.93

1,542.037.415.53

48.992.67

13.781.07

158.74.45.03

166.965.417.066.61

.27

»The reported summary statistics are before performing an Anscombe transformation.

basket. We dummy-code consumer shopping baskets onfour variables: hedonic dummy, utilitarian dummy, high-risk dummy, and low-risk dummy. The mixed shoppingbasket, containing both hedonic and utilitarian and high-risk and low-risk products, serves as the base case scenario.

Measurement scale. Consistent with prior research, weoperationalize the hedonic versus utilitarian nature of aproduct category using the hedonic utility (HEDUT) scale(Voss, Spangenberg, and Grohmann 2003). The scale mea-sures the strength of a product category on utilitarian andhedonic aspects using an equal number of items. The detailsof the scale items and anchors used for measuring each ofthese two aspects appear in the Web Appendix (WA2; www.marketingpower.com/jm_webappendix). We operationalizeperceived risk using the scale developed by Jacoby andKaplan (1972) and used by others (e.g., Chaudhuri 1998).Product category risk has five components: functional (notperforming to expectation), financial (losing money), safety(causing physical harm), psychological (tarnishing self-image), and social (lowering others' perceptions of theuser). Figure 2 depicts the relative positions of categorieson these two dimensions. We calibrate the axes in the mapon deviations from the median score. Details of the scaleitems appear in WA2.

Data collection. We collected data on these measuresfrom students in a nationally ranked business program of alarge well-known university in the eastern United States. Toreduce cognitive fatigue associated with long question-naires, we used a split sample approach in which eachrespondent evaluated only 11 product categories. We ran-domly assigned the product categories to the two types ofquestionnaires and randomly distributed the questionnairesto the respondents. Of the 78 questionnaires, we received67 usable responses.

Scale properties. Both the HEDUT and Perceived Riskscales possess excellent convergent validity, divergentvalidity, and reliability. All five items from the Perceived

Risk scale load on one factor, measuring the underlyingconcept of perceived risk. Details on the scale propertiesand computation of thé composite measures of HEDUT andPerceived Risk appear in the Web Appendix (WA2; www.marketingpower.com/jm_webappendix). Table 5 reports themean composite scores for each product category on eachof these three underlying dimensions across our sample.

Measurement validation. The measures have consider-able face validity. The categories with higher technologicalcomplexity and higher prices or those used more in socialsettings scored higher on perceived risk (e.g., electronics,photography and video equipment, jewelry, apparel andaccessories). Similarly, the respondents perceived thebeauty and cosmetics, wines, and home furnishing cate-gories as largely hedonic and the computing, telecommuni-cations, and office supplies categories as mainly utilitarian.We cross-validated these findings through the ratings offive experts who classified all 22 categories into this 2 x 2matrix. The interrater reliability was .93, suggesting highexternal validity of our survey results.

Model Formulation

Monetary value. We model the key dependent variableof interest, monetary value (DLLR¡) of a customer i, as afunction of his or her channel preference, purchase fre-quency (ORDRj), and number of marketing mailersreceived (MAILj), as follows:

(1) DLLRi = ßo + ßiCTLGi + ßaWEBj + ßjUTLi +

+ ßjLRi + ßeHR; + ß7-26PCIi +

where CTLG (catalog only) and WEB (web only) aredummy variables for the use of the catalog and web chan-nel, respectively, with multichannel as the base channel. Inaddition, UTL (utilitarian categories only), HED (hedoniccategories only), LR (low-risk categories only), and HR(high-risk categories only) are dummy variables represent-

Are Multichannei Customers Realiy iVIore Valuable? / 75

FIGURE 2Relative Positions of Product Categories Along Key Category Characteristics

2 -

(O

£ 1 -

£o

-1 -

- 2 -

IComputingl

<3

OJTelecomniunication Equipmej

|MugicnlIn8trument8|

[Photography & Video]

[Automotive Acceigorie»

|P et Items [ [Craft Supplieg]

Home&GardenI

[office Suppliei|

[jeveliy

[Beauty &Cosnietic»[

[Gifts &Holiday8|c

Booki

-3 -2 -1

Utilitarian Hedonic

Notes: X- and y-axes are calibrated as deviations from the median {0, 0).

ing category characteristics with all categories as the base.Finally, PCI is a vector of 12 two-way and 8 three-wayinteraction variables of category characteristics and chan-nels, ID is a vector of instruments for the monetary valueequation, ßs are the response parameters, and \|/ is a nor-mally distributed random error component. To isolate theimpact of channel preference on monetary value throughEquation 1, we account for the endogeneity or simultaneityof channel preference, purchase frequency, and number ofmailers .3

Channel preference. A customer is likely to prefer achannel (or combination of channels) that provides thehighest utility. Let the utility U of customer i's preference ofchannel j be given by

probit model for the probability of channel preference (P)on the basis of the following equation:

(3) p,,=j. , , j . (

where O is the probability density function of normal distri-bution and V is the deterministic component of utility.

Purchase frequency. The purchase frequency of a cus-tomer is given by

(4) ORDRj = j + Y2WEBÍ + + Y4MA1L¡

(2) Uij = ij + e¡j,

where IC is a vector of instruments for channel preference;j G 1,2,3 such that 1 = catalog, 2 = web, and 3 = multi-channel; OjS are channel-specific response parameters, e is anormally distributed random error component, and the otherterms are as defined previously. We specify a multinomial

example, a customer may prefer to use multiple channelsbecause he or she has a larger shopping basket or purchases moreexpensive items. Similarly, a customer may have a high monetaryvalue because he or she receives many marketing mailers or has ahigh purchase frequency. We model and account for such simul-taneity and endogeneity.

where Y is a parameter vector, IO is a vector of instrumentsfor the purchase frequency equation, Ç is a normally distrib-uted random error component, and the other terms are asdefined previously.

Mailers. The number of marketing mailers a customerreceives is given by

(5) MAIL; = Ô0 + Ô,CTLGi + ôsWEBj + 03DLLR¡ + O4ORDRÍ

where IM is a vector of instruments for the mailer equation, 5is a parameter vector, ö is a normally distributed error com-ponent, and the rest of the terms are as defined previously.


Identification and Instrumentai Variables

To identify the four equations, each with three endogenousvariables, we require at least three excluded exogenousvariables or instruments for each one. Theoretically, a goodinstrument should be correlated with the left-hand-sideendogenous variable but uncorrelated with the independentvariables.^ We propose nine excluded exogenous variablesfor each equation that constitute appropriate instrumentsaccording to theoretical considerations examined in themarketing literature.

Customer-ordering characteristics (ID). Three customer-ordering characteristics—namely, the number of high-endcatalogs used, the value of the highest basket, and the rela-tive use of a credit card—may influence the monetary valueof purchases for the following reasons. Customers whobrowse high-end products (HIGH) are likely to spend morebecause the average unit price of such items is higher thanthat for other items. Similarly, the dollar value of the cus-tomer's largest order (HISPEND) will be strongly corre-lated with the monetary value of customers but only weaklyrelated to the other endogenous variables such as purchasefrequency. Finally, consistent with previous research thatshows that customers who purchase with a credit card arelikely to spend more than those who use other paymentmodes (Soman and Cheema 2002), we use relative creditcard usage (RCCU) as an instrument for monetary value.

Customer demographics (IC). A customer's demograph-ics may significantly influence his or her channel preference

subsequently test for the quality of instruments in the"Robustness Checks" section.

TABLE 5Summary Scores of Product Categories on

HEDUT and Perceived Risk Scaies

CategoryUtilitarian Hedonic Risl<

Score Score Score

Apparel and accessories 5.13 5.84 3.91Arts and antiques 2.96 4.38 3.54Automotive accessories 6.19 2.57 3.50Beauty and cosmetics 4.30 5.92 4.85Books and magazines 5.04 5.67 3.18CDs and DVDs 4.65 5.64 3.47Collectibles and memorabilia 3.08 5.03 3.99Computing equipment 6.66 5.60 5.22Craft supplies 4.77 4.18 3.03Electronics 6.16 5.07 4.79Gifts and holidays 4.24 5.23 3.45Home and garden equipment 5.65 3.19 2.77Home furnishing 3.56 5.80 3.48Jewelry 4.12 5.03 4.87Musical instruments 5.14 4.74 4.55Office supplies 6.77 2.84 2.23Pet supplies and items 6.14 3.42 3.07Photography and video 6.14 5.27 4.53Sports equipment 5.70 5.11 4.08Telecommunication equipment 6.63 5.13 4.98Toys and games 3.95 4.93 3.44Wines 4.56 5.53 4.09

Notes: The scores are based on a seven-point scale in which highernumbers indicate greater strength of the measured attribute.

behavior. Different socioeconomic classes may have differ-ent predispositions to buy from different types of channels,and customer demographics play an important role in thechoice of the information channel and the resulting share ofvolume for a channel (Inman, Shankar, and Ferraro 2004).Age, family size, and education are key demographicvariables influencing channel preference .5 The literature onstore choice behavior (e.g., Popkowski Leszczyc, Sinha,and Timmermans 2000) and channel-category associations(Inman, Shankar, and Ferraro 2004) suggests that three cus-tomer demographic variables —age (LAGE), family size(FSIZE), and education (EDU) —most likely influencechannel preference.

Customer shopping experience (10). We expect thatcustomer shopping experience, which includes years ofshopping (EXP), number of product categories purchased(CAT), and number of items returned (RET), influencespurchase frequency for the following reasons. Typically,customers who have shopped longer are more knowledge-able about selling practices and channels, have greatershopping involvement, and order more frequently (Bolton1998). The number of categories and purchase frequencyare positively related because a customer tends to buy asso-ciated categories on a given occasion (Kumar and Venkate-san 2005).6 Customers with higher returns also likely havehigher purchase frequencies (Kumar and Venkatesan 2005).

Customer marketing profile (IM). Firms send marketingmailers to prospects according to their marketing profiles.Three key variables that constitute marketing profile are (1)the number of unique mailing lists containing the cus-tomer's name (UNQML), (2) the net worth of the targetcustomer (NTWTH), and (3) the number of unique mailinglists to which the customer has responded (UNQRS). Directmarketers typically use these factors when selecting newcustomers to target in their direct mail campaigns (DirectMarketing Association 2005).

Estimation

The proposed simultaneous system comprises an observedendogenous discrete choice variable (channel preference),endogenous count variables (frequency and mailers), and anendogenous continuous variable (monetary value). Becausewe have a combination of discrete and continuous variables

5We do not include gender in our analysis because there are nostrong theoretical reasons to expect differences in channel prefer-ence due to gender differences and because a subsequent empiricalanalysis involving gender showed that gender has a nonsignificanteffect on channel preference (p > .10). This finding is consistentwith the result that there are no significant differences betweenmale and female web-only shoppers (Jupiter Research 2011). Wealso exclude income because it is highly correlated with educationin the data we subsequently analyze.

Âlthough the number of categories may seem positively corre-lated with monetary value, there is no theoretical reason to believethis is so. A customer buying one low-value item each from sev-eral categories may have a lower monetary value than another cus-tomer buying several high-value items from a single category.Indeed, the correlation between these two variables is not high inour data (.39).


in the system, traditional two-stage or three-stage leastsquares estimation will lead to biased estimates. To estimatethis system, we extend the generalized probit framework(Amemiya 1978), which assumes that the random errorcomponent in each equation is normally distributed. Tomake the joint estimation tractable, we transform the nega-tive binomially distributed count variables (frequency andmarketing mailers) into near-normal distributed variables,using Anscombe transformation (Anscombe 1948). Thisprocedure ensures that our system has equations with onlytwo types of dependent variables with normally distributederrors. We follow the two-step estimation approach detailedin the Web Appendix (WA3; www.marketingpower.com/jm_webappendix). In Step 1, we regress each endogenousvariable on all the exogenous instruments. In Step 2, weregress monetary value on the included exogenous variablesand predicted values of endogenous variables (from Step 1).We use the ordinary least squares estimation for the mone-tary value, purchase frequency, and marketing mailers mod-els and the multinomial probit estimation for the channelpreference model.

Results and Discussioni\/lain Effect

We present the results of the monetar>' value model in Table6.'^ Consistent with Hi, across all product categories, multi-channel customers have a significantly higher monetaryvalue than single-channel customers (p < .01). The averagemultichannel customer outspends the average catalog- andweb-only customers by $60.13 (ß,) and $108.92 (ßj),respectively. In addition, the intercept is positive and signifi-cantly high ($436.76; ßo), highlighting the expected highspending level of multicategory, multichannel shoppers.

Hypothesized interaction Effects

H2 suggests that single-channel customers of utilitarianproduct categories spend more than other customers. Afterwe control for the effects of other variables, the averagecatalog-only, web-only, and multichannel customers ofutilitarian product categories spend $438.70 (ßo + ßi + ß3 +ß7), $431.65 (ßo + ß2 + ß3 + ßs), and $435.05 (ßo + ßs),respectively.8 However, the difference in spending amongthe average catalog-only, web-only» and multichannel cus-tomers is not significant (p > .10). Thus, we fmd that themonetary values of utilitarian category purchases do notsignificantly differ among traditional, electronic, and multi-ple channel customers.

H3 proposes that multichannel customers of hedonicproduct categories will outspend their single-channel coun-terparts. We fmd that the average catalog-only, web-only,and multichannel customers of hedonic product categoriesspend $187.43 (ßo + ßi + ß4 + ßg), $112.77 (ßo + ß2 + ß4 +

TABLE 6Results of Monetary Value Model

Coefficient

Main EffectsInterceptCatalog-only dummyaWeb-only dummyaUtilitarianHedonicLow-riskHigh-risk

Two-Way InteractionsCatalog only x utilitarianWeb only x utilitarianCatalog only x hedonicWeb only x hedonicCatalog only x low riskWeb only x low-riskCatalog only x high-riskWeb only x high-riskUtilitarian x low-riskUtilitarian x high-riskHedonic x low-riskHedonic x high-risk

Three-Way InteractionsCatalog only x utilitarian x

low-riskWeb only x utilitarian x

low-riskCatalog only x utilitarian x

high-riskWeb only x utilitarian x

high-riskCatalog only x hedonic x

low-riskWeb only x hedonic x

low-riskCatalog only x hedonic x

high-riskWeb only x hedonic x

high-riskOther Controls

Purchase frequencyMailersHigh-end catalogLargest past spendRelative use of credit card

Model fit (R-square)

ßoßiß2ßsß4ßsße

ß7ßsßgßioßiißl2ßi3ßi4ßi5ßi6ßl7ßi8

ßi9

ß20

ß21

ß22

ß23

ß24

ß25

ß26

ß27ß28ß29ß30ßsi

Estimate

436.76-60.13

-108.92-1.71

-105.23-24.89

59.34

63.79105.53-83.97

-109.84107.5447.7549.27

114.12-5.64

-98.50112.49-37.68

69.86

-84.36

6.06

47.55

-150.86

-67.67

-65.64

-103.91

149.8315.1523.272.01

20.4962.45%

SE

20.7715.9816.9910.6815.5118.6615.78

15.0618.379.92

16.3113.1727.2513.7831.5912.6117.8224.2126.35

17.86

21.77

10.88

11.46

23.72

12.49

22.85

18.19

9.306.693.63

.537.05

model results for the other endogenous variables not usedfor hypotheses testing appear in the Web Appendix (WA4;www.markelingpower.com/jm_webappendix).

^We tested the differences between the effects by accounting forthe standard errors and covariances of the parameter estimates.

^Multichannel is the base case.Notes: All boldfaced coefficients have ps < .05.

ßio), and $331.53 (ßo + ß4), respectively. The difference inspending between the average multichannel and catalog-only customers ($144.09) and between the average multi-channel and web-only customers ($218.76) is positive andsignificant (p < .01). These findings suggest that multichan-nel customers of hedonic product categories significantlyoutspend their single-channel counterparts, in support ofH3.

H4 proposes that traditional channel customers of low-risk categories have higher monetary value than that ofother customers. The average catalog-only, web-only, andmultichannel customers of low-risk product categories


spend $459.27 (ßo + ßi + ßs + ßii)- $350.70 (ßo + ßa + ßs +ß,2), and $411.87 (ßo + ßs), respectively. The difference inspending between the average catalog- and web-only cus-tomers ($108.57,p < .01) and between the average catalog-only and multichannel customers ($47.41,/? < .10) is posi-tive and signiflcant, suggesting that traditional customers oflow-risk categories offer higher monetary value than othercustomers. These results are consistent with H4.

According to H5, multichannel and web-only customersof high-risk categories have higher monetary value than thatof other customers. The average catalog-only, web-only,and multichannel customers of high-risk product categoriesspend $485.24 (ßo + ßi + ße + ßis). $501.30 (ßo + ßa + ße +ßi4), and $496.10 (ßo + ßo), respectively. The difference inspending among the average multichannel, catalog-only,and web-only customers is not significant (p > .10). Thus,these results do not support H5.

Other interaction Effects

We did not have formal hypotheses for the effects of three-way interactions among channel preference, utilitarian ver-sus hedonic nature, and perceived risk because we treatthese effects as empirical issues. We now discuss the resultsof these interaction effects.

Low-risk/utilitarian. We find that the average catalog-only, web-only, and multichannel customers of low-risk/utilitarian product categories spend $585.57 (ßo + ßj + ß3 +ß5 + ß? + ßii + ßi5 + ßi9). $322.78 (ßo + ß2 + ß3 + ßs + ßs +ßi2 + ßi5 + ß2o)> and $404.51 (ßo -n ßj + ßs + ßu), respec-tively. The average catalog-only customers of low-risk/utilitarian product categories spend $262.79 {p < .01) and$181.05 (p < .01) more than their web-only and multichannelcounterparts, respectively. Therefore, for low-risk/utilitariancategories, traditional channel customers outspend non-traditional channel customers. Although the interactioneffects of channel preference and a utilitarian nature suggestno significant difference between the monetary values ofsingle-channel and multichannel customers (lack of supportfor H2), the interaction effects of channel preference and thelow-risk nature of the category indicate a higher spendinglevel by customers of traditional channels (H4). Becausetraditional channel customers are a subset of single-channelcustomers, for low-risk/utilitarian categories, traditionalchannel customers (who tend to have a prevention focus)experience a stronger regulatory fit than their web-only andmultichannel counterparts.

High-risk/utilitarian. We find that the average catalog-only, web-only, and multichannel customers of high-risk/utilitarian product categories spend $454.88 (ßo + ßi + ß3 +ßö + ß, + ß,3 + ß,6 + ß2,), $554.16 (ßo + ß2 + ß3 + ße + ß8 +ßi4 + ßi6 + ß22). and $395.88 (ßo + ß3 + ße + ßie). respec-tively. The average web-only customers of high-risk/utili-tarian product categories spend $99.28 (p < .05) and$158.27 (p < .01) more than their catalog-only and multi-channel counterparts, respectively. In addition, we find thatthe difference between the monetary values of the averagemultichannel and catalog-only customers is not statisticallysignificant (p > .10). These results suggest that for high-risk/utilitarian product categories, web-only customers provide

the highest monetary value, but the monetary values ofcatalog-only and multichannel customers do not differ.

This result is consistent with the arguments used fortheorizing two-way interaction effects. Promotion-focusedweb-only customers are often comfortable buying high-value items from high-risk/utilitarian categories (e.g., con-sumer electronics, computing equipment) (Van Noort,Kerkhof, and Fennis 2008). Furthermore, utilitarian cate-gories typically require a high degree of information search.For such categories, the web is conducive for informationgathering and offers a high level of convenience for shop-ping and ordering items (Yadav and Varadarajan 2005a).Moreover, as we suggested previously, web-only shopperstend to be younger, better educated, more risk taking, andmore prone to obtaining information on utilitarian productson the web than other shoppers. Satisfaction and enjoyableexperience with the information search through an onlinechannel can lead to positive outcomes (Mathwick and Rig-don 2004). Consequently, web-only customers of high-risk/utilitarian categories buy more often and spend morethan other single-channel customers. Because the utilitariannature of the categories leads to efficient shopping througha single channel, web-only customers also outspend multi-channel shoppers.

Low-risk/hedonic. We now turn to the effects of theinteractions of perceived risk with the hedonic nature of theproduct category. For low-risk/hedonic product categories,the average catalog-only, web-only, and multichannel cus-tomers spend $231.72 (ßo -t- ßi + ß4 + ßs + ßg + ß,, H- ßn +ß23), $180.46 (ßo + ß2 + ß4 + ßs + ßlO + ßl2 + ßl7 + ß24),and $419.13 (ßo + ß4 + ßs + ßn), respectively. The averagemultichannel customers of low-risk/hedonic product cate-gories spend $187.41 (p < .01) and $238.68 (p < .01) morethan their catalog- and web-only counterparts, respectively.The result of the test of H4 shows that traditional channelcustomers of low-risk product categories outspend othercustomers of these categories. The result of the test of H3shows that multichannel customers of hedonic categoriesspend more than other customers. However, for both thelow-risk and the hedonic nature categories, multichannelcustomers have higher monetary value than other cus-tomers. This finding suggests that the hedonic nature has astronger effect than low risk on monetary value.

High-risk/hedonic. The average catalog-only, web-only,and multichannel customers of high-risk/hedonic productcategories spend $192.73 (ßo -i- ß, + ß4 + ße + ßg + ßn +ß,8 + ß25), $144.64 (ßo + ß2 + ß4 + ße + ßio + ßi4 + ßi8 +ß2e), and $353.19 (ßo + ß4 -H ße + ßig), respectively. Theaverage multichannel customers of high-risk/hedonic prod-uct categories spend $160.47 (jj < .01) and $208.55 (p <.01) more than their catalog- and web-only counterparts,respectively. The two-way interaction effects of channelpreference and hedonic nature (H3) and of channel prefer-ence and high-risk nature (Hg) indicate a higher spendinglevel by multichannel customers than other customers. Con-sequently, the interaction of the hedonic nature with highrisk has a more positive effect on the spending of multi-channel customers than that of single-channel customers.Taken together, the results show that for hedonic categories.


multichannel customers provide the highest monetary valueregardless of the risk level of the category.

Extension and Generaiization of Resuits to theStore Channei

The large data set in our study helps uncover empiricallygeneralizable findings across multiple product categoriesand direct marketers with catalog, web, and multiple chan-nels. To generalize these findings to the store channel, weextend this study with an analysis of a U.S. multiproductretailer's data set that includes (1) lime-series data and (2)data from physical stores. Time-series data facilitate thestudy of the potentially causal relationship between channelpreference and monetary value. Because store purchasesconstitute a majority of transactions for many multichannelretailers, analysis of physical store data enables us to gener-alize our results.

The details of this analysis appear in the Web Appendix(WA5; www.marketingpower.com/jm_webappendix). Thefmdings from this analysis are consistent with those fromthe larger cross-sectional data set and reinforce our conclu-sions. In addition, the fmdings bolster the temporal linksamong the variables in our model and extend the generaliz-ability of our results for the catalog-only channel segmentto all traditional channel segments, including the store-onlychannel segment.

Robustness Checi(S

Out-of-sample predictions. We validate our findingswith an out-of-sample prediction on a randomly selectedholdout sample of 50,000 customers. Using the parameterestimates from our estimated model on a sample of theremaining 362,424 customers, we predict the monetaryvalue of customers in the holdout sample. The meanabsolute deviation (MAD) is 250.14, the mean monetaryvalue of the holdout sample is $1,254, and the MAD isapproximately 19.95% of the sample mean. These values arereasonable for cross-sectional out-of-sample validation .9

Quality of instruments. We test the validity and strengthof our instruments in multiple ways. In the "Identificationand Instrumental Variables" subsection, we argue that ourchoice of instrumental variables is based on theory. We testfor the strength of the instrumental variables using Staigerand Stock's (1997) approach. In this approach, we test thefirst-stage F-statistic for each equation with the instrumentalvariables. The bias introduced by the v/eak instruments is ofthe order of the inverse of the F-statistic. We follow Stockand Watson's (2003) rule of thumb; that is, an F-statisticgreater than 10 is acceptable because it corresponds to abias of less than 10% in the estimates. Staiger and Stock's(1997) test for the first-stage regression in our data does notindicate the presence of poor instruments. The F-statisticsof the monetary value, frequency, and mailers equations are26,747, 28,033, and 31,058, respectively. Thus, any weakinstrument introduces, at worst, a less than .0001% bias.

MAD percentage values are comparable to those Jen,Chou, and Allenby (2009) report in a similar direct marketing con-text for predicting monetary value using linear regression.

The adjusted R-square for these regression equations is alsohealthy (at least 60%), and the instrumental variables usedin each equation are significant (p < .001), suggesting thatthe instruments are strong.

We also perform two formal tests to evaluate the validityof our instruments. First, consistent with Bound, Jaeger, andBaker (1995), we examine the validity of our instrumentsby using the correlation test. The correlation matrix reportedin the Web Appendix (WA6; www.marketingpower.com/jm_webappendix) suggests that the correlations of instru-mental variables with the associated endogenous variablesare high, whereas those with other endogenous variables aremoderate to low. Second, to ensure that our choice ofinstruments does not drive the directions of our results, wecompare the observed average values of monetary valueacross the different baskets. The directions of theseobserved differences are similar to those from the resultsestimated through our simultaneous system. However, byaccounting for endogeneity and simultaneity in our model,we can determine the correct magnitudes of the differencesin monetary values across the baskets. Together with thetheoretical arguments, these tests support the appropriate-ness of our instruments.

The instruments are exogenous or predetermined withrespect to the variables studied. Nevertheless, to ensurecomplete independence from the variables, we estimatedthe model with values of the instruments from a matchedsample of customers in the database. We used the propen-sity score matching method to select the matched sample,consistent with Rosenbaum and Rubin (1983). The resultsof this analysis, reported in the Web Appendix (WA7;www.marketingpower.com/jm_webappendix), are consis-tent with those reported in Table 6.

Operationalization of category characteristics as con-tinuous variables. We test the robustness of our findings toaltemative (continuous) measures of category characteris-tics. We generate continuous measures for each of the utili-tarian, hedonic, and perceived risk characteristics for a cus-tomer's basket by averaging the scores of eachcharacteristic across the product categories bought. Forexample, if a customer purchased only apparel and acces-sories and beauty and cosmetics, his or her hedonic, utilitar-ian, and perceived risk scores wouid be 4.72, 5.88, and4.38, respectively. The results of the monetary value modelfrom this analysis appear in the Web Appendix (WA8;www.marketingpower.com/jm_webappendix) and are con-sistent with our main model results.

Alternative definitions of multichannel customer. Weperform robustness checks for altemative definitions of amultichannel shopper. First, we define a multichannel shop-per as someone who shops across channels but within onespecific category (e.g., shoes) and across firms. The resultsare largely consistent with those reported in Table 6. Second,we define a multichannel shopper as someone who shopsacross channels and across categories but within a firm. Wediscuss this analysis in detail in the previous section onextension and generalization of results to the store channel(see WA5 in the Web Appendix; www.marketingpower.com/jm_webappendix). Finally, we define multichannel


shoppers as those who shop across channels within a cate-gory and within a firm. We use the same data as those in theprevious robustness check. However, we classify a customeras a multichannel shopper if he or she purchased acrosschannels within a given product category of the multiproductretailer. The results are largely consistent with thosereported in Table 6.

ImplicationsWe summarize our findings in Table 7. Our main effectfinding is that across product categories, an average multi-channel customer provides higher monetary value than anaverage single-channel customer. As discussed in the "Con-ceptual Development" section, customers who prefer multiplechannels may become more engaged in the purchase processas they shop across channels. Greater engagement may leadto more frequent purchases, a greater order quantity, greaterspending, or a combination of all these outcomes.

Importantly, our results show that product categorycharacteristics moderate the relationship between channelpreference and monetary value. The results of the two-wayinteractions show that the positive relationship between thepreference for multiple channels and monetary value isstronger for hedonic product categories than for utilitariancategories. A plausible explanation is that hedonic productcategories are likely to evoke impulse purchase and variety-seeking behaviors, and multiple channels provide greateropportunity and convenience to engage in those behaviors.

A key finding is that for low-risk categories, traditionalchannel customers have higher monetary value than otherchannel customers. This may be because low-risk productcategories attract prevention-focused shoppers, who pur-chase mainly from traditional channels and spend morethan their electronic and multichannel counterparts.

We also find that the perceived risk of a product cate-gory moderates the relationship between channel preferenceand monetary value for utilitarian product categories. Aplausible rationale follows. According to RFT, for high-risk/utilitarian product categories, promotion-focused cus-tomers spend more in the web channel, whereas for low-risk/utilitarian product categories, prevention-focusedcustomers spend more in the catalog or store channel.Because of the regulatory fit of their orientation with theproduct category and the channel, these customers spendmore in the respective channels than their other single-channel or multichannel counterparts.

Theoretical Implications

We contribute to the literature in several ways. First, weextend prior research on the value of multichannel shoppers(e.g., Kumar and Venkatesan 2005) and offer new insightsiiito the moderating effects of the product category on thechannel preference-customer monetary value relationship.Contrary to conventional wisdom and prior research, weshow that multichannel customers are not the most valuablesegment for all product categories. Our results demonstratethat traditional channel customers of low-risk/utilitariancategories outspend multichannel customers and that web-only customers who buy only high-risk/utilitarian categoriesoffer higher monetary value than multichannel customers.

Second, we extend prior research on the importance ofproduct category characteristics on outcomes of managerialrelevance. Prior research has examined the importance ofcategory characteristics on variables such as unplanned pur-chases (Inman, Winer, and Ferraro 2009), category manage-ment (Dhar, Hoch, and Kumar 2001), sales promotion(Ailawadi et al. 2006; Narasimhan, Neslin, and Sen 1996),revenue premium (Ailawadi, Lehmann, and Neslin 2003),and spending during economically difficult times(Kamakura and Du 2012). We extend this research by

TABLE 7Summary of Results

Hypothesis and Finding Retailers and Their Target Channel Segment

Strategie Question: Which Customer-Channel Segment to Target for High Monetary Value?Large mass-merchandise retailers (e.g.. Target, Sears) -^ multichannel customersSpecialty retailers of utilitarian products (e.g.. Best Buy, Staples) -> all channel segmentsSpecialty retailers of hedonic products (e.g.. Pottery Barn, Ulta, GameStop) -^ multichannelcustomersSpecialty retailers of low-risk products (e.g.. Office Depot, Tractor Supply Co.) -> traditionalchannel customersSpecialty retailers of high-risk products (e.g.. Pier 1 Imports, Kay Jewelers) -^ all channelsegmentsSpecialty retailers of low-risk/utilitarian products (e.g., PetSmart, Office Depot) -^ store-only orcatalog-only customersSpecialty retailers of high-risk/utilitarian products (e.g.. Wolf Camera, Crutchfield) -> web-onlycustomersSpecialty retailers of low-risk/hedonic products (e.g.. Toys "R" Us, Ulta) -> multichannelcustomersSpecialty retailers of high-risk/hedonic products (e.g., J.C. Penney, Pier 1 Imports) ->multichannel customers

Notes: M = multichannel, T = traditional channel (store/catalog), and E = electronic; superscripts: H = hedonic, U = utilitarian, HR = high-risk,and LR = low-risk.

IEHg: TU = EU = MU

H4: TLR > MLR, ELF

Hgi MHR = EHR =

JU-LR >


showing the effects of product category characteristics onthe relationship between channel preference and monetaryvalue.

Third, we illustrate the importance of the utilitarian ver-sus hedonic nature of a product category in determining thevalue of shopper channel segments. The finding that web-only (catalog- or store-only) customers spend more thanmultichannel customers on high-risk/(low-risk)/utilitariancategories suggests that the value of shopper channel seg-ments depends on whether the category is utilitarian orhedonic. For utilitarian categories, ii is highly efficient toshop in a single channel and realize the best value. How-ever, for hedonic categories, customers shopping in multi-ple channels have multiple opportunities to spend, seekvariety, or purchase on impulse. Our fmdings add to the lit-erature on the importance of the utilitarian versus hedonicnature of product categories (Chitturi, Raghunathan, andMahajan 2008; Dhar and Wertenbroch 2000; Inman, Winer,and Ferraro 2009).

Fourth, our findings highlight the role of perceived riskof a product category in determining the value of shoppersby their channel preference. The amount of money shoppersspend on a product category in thîr preferred channeldepends on their perceptions of the risks associated with thecategory. These findings supplement prior conceptual andempirical research on consumer behavior in different chan-nels (Balasubramanian, Raghunathan, and Mahajan 2005;Van Noort, Kerkhof, and Fennis 2008; Yadav and Varadara-jan 2005b).

Fifth, our results suggest important implications for theinteraction of perceived risk with the DtiIJtarian nature of thecategory. The finding that a single-channel segment offershigher monetary value than the multichannel segment andthe result that different single-channel segments providehigher monetary values of utilitarian categories for differentlevels of risk suggest that there are some commonalities butimportant differences in the underlying mechanism thatmay induce high spending. Because utilitarian categoriesare typically associated with a prevention focus, consumersmay be emphasizing purchasing efficiency, which is betterrealized in a single channel than in multiple channels. Con-sequently, single-channel customers of utilitarian categoriesmay be buying more items at higher spending levels. How-ever, at the same time, if the risk levels are high, promotion-focused consumers using the web can obtain more informa-tion and buy utilitarian items more often with higherspending levels than those using other channels. In contrast,if the risk levels are low, prevention-focused consumers canroutinize their shopping and spend more on traditionalchannels (e.g., catalog, store) than on other channels.

Managerial Implications

The results offer several actionable managerial implica-tions. First, managers can use the finding about the directeffect of channel preference on monetary value to makechannel-specific investments. Our finding reveals that ingeneral, multichannel customers who buy in multiple cate-gories are most valuable, so retail firms that sell multipleproduct categories (e.g., mass merchandisers such as Target

and Sears) should induce multichannel customers to buymore by investing in all the channels.

Second, specialty retailers of hedonic products, such asPottery Bam, Ulta, GameStop, and J.C. Penney couldincentivize their single-channel customers to shop in otherchannels, because our findings show that multichannelshoppers provide the highest monetary value for such prod-ucts. Shopping in multiple channels provides shoppers withmore opportunities to indulge in favorable experiencesoffered by hedonic products, increasing their spending onthose products. For example, when a web-only shopper pur-chases a fashion clothing item on the web, a retailer such asJ.C. Penney could invite that shopper to visit its brick-and-mortar store by offering a gift or a preferred item at a dis-count that can be collected only at the store. When theshopper visits the store to pick up the item, he or she mighttry more hedonic products, perhaps prompting the purchaseof more items.

Third, our results show that traditional channel cus-tomers of low-risk categories provide high monetary valuedue to a strong channel-category fit in prevention focus.Specialty retailers of low-risk products (e.g.. Office Depot,Tractor Supply Co.) could induce traditional channel cus-tomers to spend more at their physical stores or throughtheir catalogs by emphasizing items that are consistent withprevention focus. They could group similar products (e.g.,surge protectors with cables and batteries, livestock feedwith dog food and dog collar) through displays at the physi-cal stores or in catalogs to remind prevention-focused cus-tomers to buy more items on each purchase occasion.

Fourth, our findings demonstrate that traditional channelcustomers of low-risk/utilitarian products spend more thanother customers. Specialty retailers of low-risk/utilitarianproducts (e.g.. Office Depot, PetSmart) could help tradi-tional channel customers routinize their shopping and pur-chase more efficiently and repeatedly at their stores orthrough their catalogs. They could track the purchase histo-ries of these customers and prompt them to buy more of thesame items on a periodic basis.

Fifth, our findings reveal that electronic channel cus-tomers of high-risk/utilitarian products tend to outspend othercustomers. Specialty retailers of high-risk/utilitarian products(e.g.. Wolf Camera, Crutchfield) could make their websites"sticky" through features such as single-click ordering, prod-uct reviews, and new item recommendations. In this way,these retailers could make it convenient for promotion-focused customers who typically prefer the electronic chan-nel to continue shopping and spend more in their preferredchannel.

Sixth, specialty retailers of high-risk/utilitarian productscould also educate their prevention-focused customers whoprefer to shop through catalogs or at physical stores aboutthe high trust levels at their websites. In this way, marketerscan help such customers reduce their risk perceptions andbuy more from the web channel. For example, the staff at aBest Buy store could provide reassurance to prevention-focused store customers by demonstrating the ease andtrustworthiness of ordering online through computers at thestore and by enabling them to purchase online. Customerswho become accustomed to the online channel might shop


for more high-risk/utilitarian items online, leading them toprovide a higher monetary value in the future.

Seventh, retailers could use the insights from ourresearch to make more effective targeting decisions. Ourfindings imply that retailers of hedonic product categories(e.g., J.C. Penney, Pottery Bam, Pier 1 Imports) should tar-get multichannel customers. The results also suggest thatretailers of low-risk/utilitarian products (e.g.. Office Depot,Tractor Supply Co., PetSmart) should target customers whoprefer traditional channels. Similarly, retailers of high-risk/utilitarian products (e.g.. Best Buy, Wolf Camera,Crutchfield) should target competitors' web-only customersfor switching and offer incentives to their own web-onlycustomers to enhance retention.

Limitations, Further Research, andConclusion

This study has limitations that further research couldaddress. First, we examined observed purchase behavior.We do not have data on how customers use the channels forinformation search. Although such data are difficult to col-lect, analyzing them together with transaction data couldshed additional light on single- versus multiple-channelshopping, extending the work of Verhoef, Neslin, andVroomen (2007).

Second, if data on customer referrals are available, ourmodel of customer value could be expanded to includereferral value, extending Kumar, Petersen, and Leone's(2010) study to the multichannel context. Such an analysiscould provide a richer understanding of customer value.

Third, if data on price promotions are available, aninvestigation of the differences in the effectiveness of pricepromotions across different channel shoppers would be a

fruitful research avenue. Such an investigation would pro-vide a deeper understanding of the role of discounts in cre-ating differences in monetary values by channel preference.

Fourth, if longitudinal customer purchase data on abroad array of categories across firms are available, adeeper analysis of channel switching across product cate-gories could be undertaken to obtain greater insights intomultichannel shopping. Such an analysis would offer anuanced understanding of changes in monetary values dueto channel switching.

Fifth, although our conceptual arguments are rooted inindividual motivation, we use behavioral outcome data(spending)—not data at the decision process level. Supple-menting our study with behavioral experiments at the indi-vidual level would bolster the validity of the findings.

Finally, with the surge in the sales of mobile devices,such as smartphones and tablets, customer use of the mobilechannel is growing rapidly. As data on mobile channelbecome available, it would be useful to extend our study tothe mobile channel.

In conclusion, contrary to conventional wisdom that allmultichannel customers are valuable, our results show thatmultichannel customers are the most valuable segment onlyfor hedonic product categories; single-channel customers ofutilitarian categories and traditional channel customers oflow-risk categories provide higher monetary value thanother customers. The results reveal that for utilitarian prod-uct categories involving high (low) risk, electronic (tradi-tional) channel shoppers constitute the most valuable seg-ment. Our findings offer managers guidelines for targetingand migrating different types of customers for differentproduct categories through different channels. They alsoserve as an impetus for further research on the growing phe-nomenon of multichannel marketing.

REFERENCESAaker, Jennifer L. and Angela Y. Lee (2006), "Understanding

Regulatory Fit," Journal of Marketing Research, 43 (Febru-ary), 15-19.

Aarts, Henk and Ap Dijksterhuis (2000), "Habits as KnowledgeStructures: Automaticity in Goal-Directed Behavior," Journalof Personality and Social Psychology, 78 (1), 53-63.

Ailawadi, Kusum L., Bad Harlam, Jacques Cesar, and DavidTrounce (2006), "Retailer Promotion Profitability: The Role ofPromotion, Brand, Category, and Market Characteristics,"Journal of Marketing Research, 43 (November), 518-35.

, Donald R. Lehmann, and Scott A. Neslin (2003), "RevenuePremium as an Outcome Measure of Brand Equity," Journal ofMarketing, 67 (October), 1-17.

Amemiya, Takeshi T. (1978), "The Estimation of SimultaneousEquation Generalized Probit Model," Econometrica, 46 (Sep-tember), 1193-205.

Ansari, Asim, Carl Mela, and Scott Neslin (2008), "CustomerChannel Migration," Journal of Marketing Research, 45 (Feb-ruary), 60-76.

Anscombe, F.J. (1948), "The Transformation of Poisson, Binomial,and Negative Binomial Data," Biometrika, 35 (June), 246-53.

Avery, Jill, Thomas J. Steenburgh, John Deighton, and Mary Cara-vella (2012), "Adding Bricks to Clicks: Predicting the Patternsof Cross-Channel Elasticities over Time," Journal of Market-ing,16 (May),96-in.

Avnet, Tamar and E. Tory Higgins (2006), "How Regulatory FitAffects Value in Consumer Choices and Opinions," Journal ofMarketing Research, 43 (February), 1-10.

Balasubramanian, Sridhar, Raj Raghunathan, and Vijay Mahajan(2005), "Consumers in a Multichannel Environment: ProductUtility, Process Utility, and Channel Choice," Journal of Inter-active Marketing, 19 (2), 12-30.

Bagozzi, Richard (1975), "Marketing as Exchange," Journal ofMarketing, 39 (October), 32-39.

Bart, Yakov, Venkatesh Shankar, Fareena Sultan, and Glen L.Urban (2005), "Are the Drivers and Role of Online Trust theSame for All Web Sites and Consumers? A Large ScaleExploratory Empirical Study," Journal of Marketing, 69 (Octo-ber), 133-52.

Batra, Rajeev and Olli T. Ahtola (1990), "Measuring the Hedonicand Utilitarian Sources of Consumer Attitudes," MarketingLetters, 2 (2), \59-10.

Beatty, Sharon and Scott M. Smith (1987), "External SearchEffort: An Investigation Across Product Categories," Journalof Consumer Research, 14 (1), 83-95.

Becker, Gary S. (1965), "A Theory of the Allocation of Time,"Economic Journal, 75 (299), 493-517.

Bolton, Ruth (1998), "A Dynamic Model of the Duration of theCustomer's Relationship with a Continuous Service Provider:The Role of Satisfaction," Marketing Science, 17 (1), 45-65.

Are Multichannei Customers Reaiiy More Vaiuabie? / 83

Bound, John, David A. Jaeger, and Regina M. Baker (1995),"Problems with Instrumental Variables Estimation When theCorrelation Between the Instruments and the EndogenousExplanatory Variable Is Weak," Journal of the American Statis-tical Association, 90 (430), 443-50.

Cesario, Joseph, E. Troy Higgins, and Abigail A. Scholer (2008),"Regulatory Fit and Persuasion: Basic Principles and Remain-ing Questions," Social and Personality Psychology Compass, 2(I), 444-63.

Chaudhuri, Arjun (1998), "Product Class Effects on PerceivedRisk: The Role of Emotion," ¡nternaticnal Journal of Researchin Marketing, 15 (2), 157-68.

Chernev, Alexander (2004), "Goal-Attnbute Compatibility onConsumer Choice," Journal of Consumer Psychology, 14(1/2), 141-50.

Chitturi, Ravindra, Rajagopal Raghunathan, and Vijay Mahajan(2008), "Delight by Design: The Roie of Hedonic Versus Utili-tarian Benefits," Journal of Marketing 72 (May), 48-63.

Clifford, Stephanie (2010), "Nordstrom Links Online Inventory toReal World," The New York Times, (August 24), (accessedMarch I, 2013), [available at http://uww.nytimes.com/2010/08/24/business/24shop.html?fta=y].

Cox, Donald F. and Stuart U. Rich (1964), "Perceived Risk andConsumer Decision-Making: The Case of Telephone Shop-ping," Journal of Marketing, 1 (November), 32-39.

Deleersnyder, Barbara, Inge Geyskens. Katrijn Gielens, andMamik G. Dekimpe (2002), "How Cannibalistic Is the InternetChannel? A Study of the Newspaper Industry in the UnitedKingdom and the Netherlands," International Journal ofResearch in Marketing, 19 (4), 337-48L

Dhar, Ravi and Klaus Wertenbroch (2000), "Consumer ChoiceBetween Hedonic and Utilitarian Goods," Journal of Market-ing Research, 37 (February), 60-71.

Dhar, Sanjay, Stephen Hoch, and Nanda Kumar (2001), "EffectiveCategory Management Depends on the Role of the Category,"Journal of Retailing, 11 (2), 165-84.

Direct Marketing Association (2005). ¡»Multichannel MarketingReport. New York: Direct Marketing Association.

Dowling, Grahame and Richard Staelin (1994), "A Model of Per-ceived Risk and Intended Risk-Handlmg Activity," Journal ofConsumer Research, 2\ (June), 119-34.

Frazier, Gary L. (1999), "Organizing and Managing Channels ofDistribution," Journal of the Academy of Marketing Science,27 (Spring), 2 2 6 - ^ .

Garg, Nitika, J. Jeffrey Inman, and Vikas Mittal (2005), "Inciden-tal and Task-Related Affect: A Re-lnqiHr>' and Extension of theInfluence of Affect on Choice," Journal of ConsumerResearch, 32 (\), 154-59.

Garver, Abe (2012), "Top 5 Dangers of Online Shopping and Pre-cautions to Take," Forbes, (September 11), (accessed March 1,2013), [available at http://www.forbe3.com/sites/abegarver/2012/09/11 /the-top-five-dangers-of-online -shopping-precautions-to-take/].

Inman, J. Jeffrey, Venkatesh Shankar, and Rosellina Ferraro(2004), "The Roles of Channel-Category Associations andGeodemographics in Channel Patronage;" Journal of Market-ing, 6SiApr\\),5\-l\.

, Russell S. Winer, and Rosellina Ferraro (2009), "The Inter-play Among Category Characteristics. Customer Characteris-tics, and Customer Activities on In-Siore Decision Making,"Journal of Marketing, 73 (September),. 19-29.

Jacoby, J. and L.B. Kaplan (1972), "The Components of PerceivedRisk," in Proceedings, Third Annual Conference of Associationfor Consumer Research, M. Venkatesao, ed. Chicago: Associa-tion for Consumer Research, 382-93.

Jen, Lichung, Chien-Heng Chou, and Greg M. Allenby (2009),"The Importance of Modeling Tempoiai Dependence of Tim-

ing and Quantity in Direct Marketing," Journal of MarketingResearch, 46 (November), 482-93.

Jupiter Research (2011), Jupiter Research Internet ShoppingReport. New York: Jupiter Research.

Kamakura, Wagner and Rex Du (2012), "How Economic Contrac-tions and Expansions Affect Expenditure Patterns," Journal ofConsumer Research, 39 (2), 229-47.

Konus, Umut, Peter C. Verhoef, and Scott A. Neslin (2008),"Multi-Channel Customer Segmentation," Journal of Retail-ing, 84 (December), 398-413.

Koufaris, Marios (2002), "Applying the Technology AcceptanceModel and Flow Theory to Online Consumer Behavior," Infor-mation Systems Research, 13 (2), 205-223.

Kumar, V., Andrew Petersen, and Robert Leone (2010), "DrivingProfitability by Encouraging Referrals: Who, When, andHow," Journal of Marketing, 74 (September), 1-17.

and Rajkumar Venkatesan (2005), "Who Are MultichannelShoppers and How Do They Perform? Correlates of Multi-channel Shopping Behavior," Journal of Interactive Marketing,19 (Spring), 44-61.

Kurt, Didem, J. Jeffrey Inman, and Jennifer J. Argo (2011), "TheInfluence of Friends on Consumer Spending: The Role ofAgency-Communion Orientation and Self-Monitoring," Jour-nal of Marketing Research, 48 (August), 741-54.

Mathwick, Charla, Naresh Malhotra, and Edward Rigdon (2002),"The Effect of Dynamic Retail Experiences on ExperientialPerceptions of Value: An Internet and Catalog Comparison,"Journal of Retailing, 78 ( 1 ), 51-60.

and Edward Rigdon (2004), "Pay, Flow, and Online SearchExperience," Journal of Consumer Research, 31 (2), 324-32.

Montoya-Weiss, Mitzi M., Glenn B. Voss, and Dhruv Grewal(2003), "Determinants of Online Channel Use and Overall Sat-isfaction with a Relational, Multichannel Service Provider,"Journal of the Academy of Marketing Science, 31 (October),448-58.

Narasimhan, Chakravarthi, Scott A. Neslin, and Subrata Sen(1996), "Promotional Elasticities and Category Characteris-tics," Journal of Marketing, 60 (April), 17-30.

Neslin, Scott A., Dhruv Grewal, Robert Leghorn, VenkateshShankar, Marije L. Teerling, Jacquiline S. Thomas, et al.(2006), "Challenges and Opportunities in Multichannel Man-agement," Journal of Service Research, 9 (2), 95-112.

Novak, Thomas, Donna Hoffman, and Adam Duhachek (2003),"The Influence of Goal-Directed and Experiential Activities onOnline Flow Experiences," Journal of Consumer Psychology,13(0,3-16.

Pew Internet Project (2008), "Online Shopping," (February 13),(accessed September 18, 2012), [available at http://www.pewintemet.org/Press-Releases/2008/Online-Shopping.aspx].

Popkowski Leszczyc, Peter T.L., Ashish Sinha, and Harry J.P.Timmermans (2000), "Consumer Store Choice Dynamics: AnAnalysis of the Competitive Market Structure for GroceryStores," Journal of Retailing, 76 (Autumn), 323-45.

Rangaswamy, Arvind and Gerrit H. Van Bruggen (2005), "Oppor-tunities and Challenges in Multichannel Marketing: An Intro-duction to the Special Issue," Journal of Interactive Marketing,19 (Spring), 5- i l .

Ratner, Rebecca and Barbara Kahn (2002), "The Impact of PrivateVersus Public Consumption on Variety-Seeking Behavior,"Journal of Consumer Research, 29 (2), 246-57.

Ratneshwar,S., Cornelia Pechmann, and Allan D. Shocker (1996),"Goal-Derived Categories and the Antecedents of Across-Category Consideration," Journal of Consumer Research, 23(3), 240-50.

Rosenbaum, Paul R. and Donald B. Rubin (1983), "The CentralRole of the Propensity Score in Observational Studies forCausal Effects," Biometrika, 70 (I), 41-55.


Schlosser, Ann E., Tiffany Bamett-White, and Susan M. Lloyd(2006), "Converting Website Visitors into Buyers: How WebSite Investments Increase Consumer Trusting Beliefs andOnline Purchase Intentions," Jourruil of Marketing, 70 (April),133^8.

Sheth, Jagdish N. and M. Venkatesan (1968), "Risk ReductionProcesses in Repetitive Consumer Behavior," Journal of Mar-keting Research, 5 (August), 307-310.

Simonson, Itamar (1990), "The Effect of Purchase Quantity andTiming on Variety-Seeking Behavior," Journal of MarketingResearch, 27 (May), 150-62.

Soman, Dilip and Amar Cheema (2002), "The Effect of Credit onSpending Decisions: The Role of Credit Limit and Credibility,"Marketing Science, 21 (Winter), 32-53.

Staiger, Douglas and James H. Stock (1997), "InstrumentalVariables Regression with Weak Instruments," Econometrica,65 (3), 557-86.

Stilley, Karen M., J. Jeffrey Inman, and Kirk L. Wakefield(2010a), "Planning to Make Unplanned Purchases? The Roleof In-Store Slack in Budget Deviation," Journal of ConsumerResearch, 31 (2), 264-1^.

, , and (2010b), "Spending on the Fly: MentalBudgets, Promotions, and Spending Behavior," Journal ofMarketing, 74 (May), 34-47.

Stock, James and Mark Watson (2003), Introduction to Economet-rics. New York: Pearson Addison Wesley.

Thomas, Jacquelyn S. and Ursula Sullivan (2005), "ManagingMarketing Communication with Multichannel Customers,"Journal of Marketing, 69 (October), 239-51.

Valentini, Sara, Elisa Montaguti, and Scott A. Neslin (2011),"Decision Process Evolution in Customer Channel Choice,"Journal of Marketing, 75 (November), 72-86.

Van Noort, Guda, Peter Kerkhof, and Bob M. Fennis (2008), "ThePersuasiveness of Online Safety Cues: The Impact of Preven-tion Focus Compatibility of Web Content on Consumers' RiskPerceptions, Attitudes, and Intentions," Journal of InteractiveMarketing, 22 (Autumn), 58-72.

Van Trijp, Hans CM., Wayne Hoyer, and J. Jeffrey Inman (1996),"Why Switch? Product Category-Level Explanations for TrueVariety-Seeking Behavior," Journal of Marketing Research, 33(August), 281-92.

Venkatesan, Rajkumar, V. Kumar, and N. Ravishanker (2007),"Multichannel Shopping: Causes and Consequences," Journalof Marketing, 11 (April), 114-32.

Verhoef, Peter C , Scott A. Neslin, and Bjom Vroomen (2007),"Multichannel Customer Management: Understanding theResearch-Shopper Phenomenon," International Journal ofResearch in Marketing, 24 (2), 129-48.

Voss, Kevin E., Eric R. Spangenberg, and Bianca Grohmann(2003), "Measuring the Hedonic and Utilitarian Dimensions ofConsumer Attitude," Journal of Marketing Research, 40 (Feb-ruary), 310-20.

Yadav, Manjit S. and Rajan Varadarajan (2005a), "Interactivity inthe Electronic Marketplace: An Exposition of the Concept andImplications for Research," Journal of the Academy of Market-ing Science, 33 (4), 585-603.

and (2005b), "Understanding Product Migration tothe Electronic Marketplace: A Conceptual Framework," Jour-nal of Retailing, 81 (2), 125^0.

Yeo, Junsang and Jong won Park (2006), "Effects of Parent-Extension Similarity and Self Regulatory Focus on Evaluationsof Brand Extensions," Journal of Consumer Psychology, 16(3), 272-82.


Copyright of Journal of Marketing is the property of American Marketing Association and itscontent may not be copied or emailed to multiple sites or posted to a listserv without thecopyright holder's express written permission. However, users may print, download, or emailarticles for individual use.

Tarun Kushwaha & Venkatesh Shankar Are Multichannel ... · Venkatesh Shankar is Professor of Marketing and Coleman Chair in Marketing, Mays Business Schooi, Texas A&M Uni-versity

Documents