Seiders, Voss, Grewal and Godfrey.pdf

7/21/2019 Seiders, Voss, Grewal and Godfrey.pdf

1/23

Do Satisfied Customers Buy More? ExaminingModerating Influences in a Retailing Context

K. Seiders

G.B. Voss

D. Grewal &

A.L. Godfrey

ABSTRACT

In this research, the authors propose that the relationship between satisfaction and repurchasebehavior is moderated by customer, relational, and marketplace characteristics. They furtherhypothesize that the moderating effects emerge if repurchase is measured as objective behavior butnot if it is measured as repurchase intentions. To test for systematic differences in effects, the authorsestimate identical models using both longitudinal repurchase measures and survey measures as thedependent variable. The results suggest that the relationship between customer satisfaction andrepurchase behavior is contingent on the moderating effects of convenience, competitive intensity,customer involvement, and household income. As the authors predicted, the results are significantlydifferent for self-reported repurchase intentions and objective repurchase behavior. The conceptualframework and empirical findings reinforce the importance of moderating influences and offer newinsights that enhance the understanding of what drives repurchase behavior.

Marketing literature consistently identifiescustomer satisfaction as a key antecedent toloyalty and repurchase, but current knowledgefails to explainfully the prevalence of satisfiedcustomers who defect and dissatisfiedcustomers who do not (Bendapudi and Berry1997; Ganesh, Arnold, and Reynolds 2000;Jones and Sasser 1995; Keaveney 1995).Although prior research points to severalvariables that may moderate the satisfactionrepurchase relationship, empirical results are

equivocal and difficult to reconcile.

Many empirical studies examining direct andmoderated satisfactionrepurchase effectsmeasure repurchase intentions rather thanobjective repurchase behavior. Studies canproduce erroneous inferences if there aresignificant differences between intentions andsubsequent behavior (Bolton 1998; Kamakuraet al. 2002; Mittal and Kamakura 2001;Morwitz, Steckel, and Gupta 1997) or ifcommon method variance inflates estimates of

the association between self-reportedsatisfaction and intentions (Bolton 1998; Gruen,

Summers, and Acito 2000; Morwitz andSchmittlein 1992). Satisfaction levels at whichcustomers report a positive intent can differconsiderably from those at which customersengage in the corresponding behavior (Mittaland Kamakura 2001). Therefore, additionalresearch is necessary that explicitly examinesthe extent to which results converge when usingrepurchase intentions versus objectiverepurchase behavior as the dependent measure.

In response to calls for deeper insight intofactors that may moderate the satisfactionrepurchase relationship (e.g., Bolton, Lemon,and Verhoef 2004), we propose a conceptualframework that explains why two customerswith the same (different) levels of satisfactionengage in different (the same) patterns ofrepurchase behavior. We use consumer resourceallocation theory to support our prediction that,after we control for main effects established inprior research (Anderson and Sullivan 1993;Bolton 1998; Boulding et al. 1993; Rust,

Zahorik, and Keiningham 1995), customer,relational, and marketplace characteristics


2/23

moderate the relationship between satisfactionand repurchase behavior but do not moderatethe relationship between satisfaction andrepurchase intentions. For example, conve-nience (a marketplace characteristic) conservescustomers time and effort and therebyfacilitates a satisfied customers ability to fulfillhis or her intent.

We test the conceptual framework in anunderstudied retail context that is characterizedby low switching costs and comparisonshopping behavior. This context is noteworthybecause no known research has examineddifferences in intentions and objectiverepatronage behavior in a retail shoppingcategory marked by moderate repurchase

frequency. Research suggests that the predictivevalidity of repurchase intentions varies widelyfrom frequently purchased convenience goodsto infrequently purchased durables (e.g.,Chandon, Morwitz, and Reinartz 2005). Inaddition, the satisfactionrepurchaserelationship can differ significantly betweencontractual services and discrete, recurringpurchases (Lemon, White, and Winer 2002;Reinartz and Kumar 2003), for which switchingcosts are lower and customers typically are not

obligated to give all their business to any onefirm (e.g., Rust, Lemon, and Zeithaml 2004).Thus, our research extends current knowledgeby capturing the complexity of the satisfactionrepurchase relationship in a context marked bydiscrete recurring transactions.

CONCEPTUALIZING A MODERATEDSATISFACTION REPURCHASEBEHAVIOUR RELATIONSHIP

In Figure 1, we present a conceptual frameworkthat proposes satisfaction and customer,relational, and marketplace characteristics asantecedents to repurchase intentions andbehavior. We conceptualize customersatisfaction as a cumulative, global evaluation

based on experience with a firm over time(Homburg, Koschate, and Hoyer 2005).Repurchase intentions represent thecustomers self-reported likelihood of engagingin future repurchase behavior, whereasrepurchase behavior is the objectively observedlevel of repurchase activity. The defaultexpectation is that satisfaction positivelyinfluences both repurchase intentions andbehavior, and we offer no formal hypothesis forthis well-established relationship.

Figure 1:A framework of examining moderators of the relationship between customersatisfaction and repurchase


3/23

The dotted lines in Figure 1 capture directrelationships that have been previouslyestablished in the literature (e.g., Beatty andSmith 1987; Bolton, Kannan, and Bramlett,2000; Rust, Lemon, and Zeithaml 2004;Soberon-Ferrer and Dardis 1991). In thesections that follow, we provide brief reviews ofthe relevant literature for these direct effects,but we do not offer formal hypotheses for them.Instead, we focus on the moderating effectsdepicted by the solid lines in Figure 1.Specifically, we predict that customer,relational, and marketplace characteristicsmoderate the relationship between satisfactionand objective repurchase behavior after weexplicitly control for their direct (i.e., main)effects. Moreover, we believe that these

variables do not moderate the relationshipbetween satisfaction and repurchase intentionsafter we control for direct effects.

For conceptual, methodological, and empiricalreasons, we believe that customer, relational,and marketplace characteristics moderate theeffect of satisfaction on objective repurchasebehavior but not on repurchase intentions. First,consumer resource allocation theory suggeststhat repurchase behavior reflects interveningcontingencies that measures of repurchase

intentions do not. Consumers allocate a varietyof resources to purchase decisions (Batshell1980; Roberts and Dant 1991; Zeithaml 1988),including money (Marmorstein, Grewal, andFishe 1992), time and effort (Becker 1965;Feldman and Hornik 1981; Jacoby, Szybillo,and Berning 1976), motivation, opportunity,and cognitive ability (e.g., MacInnis andJaworski 1989; MacInnis, Moorman, andJaworski 1991; Peracchio and Meyers-Levy1997). One stream of research depictsconsumers as cognitive misers (e.g., Shugan

1980) who lack the motivation and cognitiveability to incorporate intervening contingenciesinto their predicted repurchase probabilities.Because consumers are not motivated toconsider simple intervening characteristics(e.g., how different levels of income mightfacilitate or constrain future repurchase activity)or capable of foreseeing complex interveningfactors (e.g., competitive interactions amongfirms), they routinely provide inaccuratepredictions of their future behavior(Kahneman and Snell 1992; Morwitz 1997;Morwitz, Steckel, and Gupta 1997). Thus,consumer resource allocation theory explains

why people fail to consider interveningcontingency effects in predicting their futurebehavior and predicts subsequent differences intheir motivation and capability to engage inrepurchase behavior.

Second, from a methodological perspective, weexpect systematic differences in themeasurement properties of repurchaseintentions and behavior. Because intentionsmeasures typically use five- or seven-pointscales, information lost as a result of rangerestrictions and coarseness can attenuateresearchers ability to detect significantinteraction effects that truly exist in thepopulation (Russell and Bobko 1992). Rangerestriction occurs when information is lost

because the highest or lowest point on the scaledoes not accurately capture extreme variationsin the construct of interest. Similarly,coarseness refers to information that is lostwhen one-point scale variations do notaccurately capture within-range variation in theconstruct of interest. Range-restricted andcoarse scales may capture direct linearrelationships with other constructs, especially ifthe two measures share common methodvariance and response bias (Bolton 1998;Morwitz and Schmittlein 1992). Measurement

theory suggests that intentions measures do notcapture the nuanced, complex variations thatare provided by objective repurchase behaviormeasures, even if respondents could makeaccurate predictions.

Finally, prior empirical research demonstratesthat the conversion of intent into repurchase ismoderated by various factors, including the typeof product (Jamieson and Bass 1989; Kalwaniand Silk 1982; Young, DeSarbo, and Morwitz1998), demographics (Morwitz and Schmittlein,

1992), experience (Bentler and Speckart 1979;Morwitz and Schmittlein 1992), and time lapse(Chandon, Morwitz, and Reinartz 2005; Mittaland Kamakura 2001; Young, DeSarbo, andMorwitz 1998). Studies that Chandon,Morwitz, and Reinartz (2005) conductedsuggest that consumers provide relatively moreaccurate predictions of frequent, routinepurchase decisions, such as those involvinggrocery items, than of infrequent, complexpurchase decisions, such as those involvingcomputers or automobiles. We attribute this


4/23

Table 1

Moderators of the Association Between Satisfaction and Repurchase

StudyDependent Variable(s) Context

and DesignCustomer

CharacteristicsRelational

CharacteristicsMarketplace

Characteristics

Bolton (1998) Relationship Duration (OM)TelecommunicationsLongitudinalContractual service

Length ofexperience (+)

Bowman andNarayandas(2001)

Share of Category Requirements(SR)

Consumer package goodsCross-sectionalNoncontractual goods

Heavy user (+) Loyalty (+)

Bowman andNarayandas(2004)

Share of Customer Wallet (SR)Processed metalLongitudinalNoncontractual industrial goods

Size () Accountmanagement tenure

(+)

Satisfaction withcompetitor (+)

Burnham, Frels,and Mahajan(2003)

Intention to Stay with Provider(SR)

Credit card and telephoneservice

Cross-sectionalContractual service

Relational switchingcosts (n.s.)

Proceduralswitching costs

(n.s.)Financial switching

costs (n.s.)

Capraro,Broniarczyk,and Srivastava(2003)

Garbarino andJohnson (1999)

Homburg andGiering (2001)

Defection/Repurchase (SR)Health insuranceLongitudinalContractual service

Future Intentions (SR)Professional theater

Cross-sectionalContractual and noncontractual

service

a. Recommendation Intentions(SR)

b. Brand Repurchase Intentions(SR)

c. Dealer Repurchase Intentions(SR)

Auto manufacturer/dealerCross-sectionalContractual goods andservices

Objectiveknowledge (n.s.)

Subjectiveknowledge (n.s.)

IncomeSP (: a, b,c) SSP (+:a, b)

Involvement: SSP(: b)

Gender:SP (+m:c) SSP(+f: b)

AgeSP (+: a, b,c) SSP (:b)

Relationalorientation ()

Variety seekingSP (: a, b, c)

Jones,Mothersbaugh,and Beatty(2000)

Repurchase Intentions (SR)Banking and hair salonCross-sectionalContractual and noncontractual

services

Interpersonalrelationships ()

Switching costs ()Attractiveness of

alternatives (+)


5/23

Table 1Continued

Study Dependent Variable(s)Context and Design

CustomerCharacteristics

RelationalCharacteristics

MarketplaceCharacteristics

Magi (2003) a. Share of Purchase (SR)b. Share of Visits (SR)

Grocery storesLongitudinalNoncontractual consumption

goods

Economicorientation

(: a; n.s.: b)Personalizing

orientation (: a, b)Apathetic shopping

orientation(n.s.: a, b) Age

(n.s.: a, b)

Purchase volume(+: a; n.s.: b)

Mittal andKamakura(2001)

Repurchase Behavior (OM)Automobile manufacturerLongitudinalContractual durable goods

Sex (+) Education(+) Marital status

(n.s.) Age (+)Children (+)

Urban versussuburban (n.s.)

Verhoef (2003) a. Customer Retention (OM)b. Customer Share Development

(OM) InsuranceLongitudinalContractual service

Relationship age(+: a; n.s.: b.)

Verhoef, Franses,

and Hoekstra(2002)

a. Customer Referrals (SR)

b. Number of Services Purchased(OM)

InsuranceCross-sectional and

longitudinalContractual service

Relationship age

(n.s.: a; +: b)

Current Study Repurchase Intentions (SR),Repurchase Visits (OM), andSpending (OM)

Apparel and home furnishingsCross-sectional and

longitudinalNoncontractual fashion goods

InvolvementHousehold income

Relationship ageRelationship

programparticipation

Competitiveintensity

Convenience ofoffering


6/23

lower accuracy in the prediction of infrequent,complex purchase decisions to unforeseencontingency effects that emerge betweenintentions measurement and subsequentrepurchase (Kalwani and Silk 1982).

In the following section, we rely on thisconceptual, methodological, and empiricalevidence to develop specific hypotheses thatbuild on prior research that has examinedmoderators of the satisfactionrepurchaserelationship. In Table 1, we summarize thestudies that support moderating effects ofvarious customer, relational, and marketplacecharacteristics. We report the results only formoderating effects; that is, we do not include

results for main effects. A review of Table 1shows that our study makes uniquecontributions by testing formerly unexaminedmoderating variables; linking survey data toself-reported intentions and objective,longitudinal repurchase behavior; andinvestigating a previously understudied contextmarked by low exit barriers.

DEVELOPMENT OF HYPOTHESES

The conceptual framework we present in Figure

1 proposes three categories of moderators thatoperate at different levels. Customercharacteristics explain variations in thesatisfactionrepurchase relationship due toindividual differences, relational charac-teristics capture customers investments inbuilding or formalizing relationships with aspecific firm, and marketplace characteristicsaccount for variations related to market-levelcompetition. For each category of moderator,we propose and subsequently test two specificmoderating variables. In each case, we predict

an interaction effect after we control for maineffects.

Customer Characteristics

Customer characteristics explain variations inpeoples purchase levels for an entire purchasecategory. We expect that customer-levelvariables have a direct influence on repurchaseintentions and behavior and moderate therelationship between satisfaction andrepurchase behavior. We examine involvement,

a motivational resource, and householdincome, a monetary resource. Because both

moderators are closely linked to key resources,they are likely to be among the most significantcustomer-level influences.

Involvement. Involvement is the importance ofthe purchase category to the consumer and isbased on the consumers inherent needs, values,and interests (Mittal 1995). From a resourceperspective, highly involved customers allocatemore time and effort to search (Beatty andSmith, 1987; Bloch, Sherrell, and Ridgway1986; Maheswaran and Meyers-Levy 1990) andreport higher levels of repatronage intentions(Wakefield and Baker 1998), which suggests apositive direct link between involvement andrepurchase intentions and behavior. We

acknowledge an alternative view thatinvolvement could negatively affect repurchaseintentions. More involved consumers may bemore likely to search and potentially identifymore preferred alternatives in the market,regardless of their level of satisfaction.

We also expect that involvement enhances thepositive effect of satisfaction on actualrepurchase behavior but not on repurchaseintentions. Involved shoppers should allocatemore time, effort, and money to retailers that

provide exceptional satisfaction. They shouldalso be more discriminating among offeringsand more responsive and committed to superiorofferings (Beatty, Kahle, and Homer 1988).This positive moderating effect would extend torepurchase intentions if involved customersaccurately incorporated these complex effectsinto their predictions, but because we do notexpect such incorporation to occur, we formallyhypothesize the following:

H1: Involvement (a) moderates (enhances) the

positive relationship between customersatisfaction and objective repurchase

behaviour but (b) does not moderate thepositive relationship between customersatisfaction and repurchase intention.

Household income. Household income ispositively related to consumers routineexpenditures for multiple types of services(Nichols and Fox 1983; Soberon-Ferrer andDardis 1991), loyalty among online shoppers(Keaveney and Parthasarathy 2001), and

profitable lifetime customer duration (Reinartzand Kumar 2000). On the basis of these


7/23

findings, we expect that household income hasa positive influence on repurchase intentionsand behavior.

Household income should also intensify therelationship between satisfaction andrepurchase behavior. The conversion of intentinto purchase varies across groups that differ intheir ability to fulfill that intent (Morwitz andSchmittlein 1992), and lower-income customersmay be constrained in their purchases. Becausehigher-income customers place a higher valueon time and are more discriminating in howthey allocate their time (Marmorstein,Grewal, and Fishe 1992), they should visit andspend less at retailers that offer low satisfactionand more at retailers that offer high satisfaction.

This positive moderating effect would extend tointentions only if higher- and lower-incomecustomers accurately incorporated the enablingand constraining effect of income. Because wedo not expect such incorporation to occur, weformally hypothesize the following:

H2: Household income (a) moderates (enhances)

the positive relationship between customersatisfaction and objective repurchase

behavior but (b) does not moderate thepositive relationship between customersatisfaction and repurchase intentions.

Relational Characteristics

Relational characteristics represent formal andinformal bonds between the firm and itscustomers; relational bonds can create socialand financial switching barriers that providefirms with an advantage insulated fromcompetitor actions. Although relationalmoderators have been examined primarily inthe context of contractual services, relationalstrategies designed to encourage discrete,ongoing repurchase are widespread. Proposedrelational moderators include relationship agewith the focal firm and participation in thefirms relationship program.

Relationship age. Prior experience influencesintent, repurchase behavior (Anderson,Fornell, and Lehmann 1994; Morwitz andSchmittlein 1992), and loyalty (Ganesh, Arnold,and Reynolds 2000). Relationship age ispositively related to customer profitability

(Reinartz and Kumar 2000, 2003), retention(Bolton 1998), number of services purchased

(Verhoef, Franses, and Hoekstra 2002),continued museum membership (Bhattacharya1998; Bhattacharya, Rao, and Glynn 1995), and(we expect) repurchase intentions and behavior.

Empirical results indicate that length of priorexperience enhances the positive associationbetween satisfaction and subsequentrelationship duration (Bolton, 1998) and thatrelationship age enhances the link betweensatisfaction and retention and the number ofservices purchased (Verhoef 2003; Verhoef,Franses, and Hoekstra 2002). This effect wouldextend to intentions only if relationalcustomers accurately incorporated themoderating effect of prior relationalinvestments, but because we do not expect such

incorporation to occur, we hypothesize thefollowing:

H3: Relationship age (a) moderates (enhances)

the positive relationship between customersatisfaction and objective repurchase behavior

but (b) does not moderate the positiverelationship between customer satisfactionand repurchase intentions.

Relationship program participation. Relation-ship programs represent company initiativesthat target individual customers who agree toexchanges that may be complementary orancillary to their purchase transactions. Thesepro- grams promote retention by enhancingcustomers perceptions of the relationshipinvestment and increasing their trust andcommitment (De Wulf, Odekerken-Schroder,and Iacobucci 2001; Rust, Lemon, and Zeithaml2004). Participants may receive personalizedcommunications that keep them informed ofnew offerings or preferential treatment andrewards for past loyalty. Empirical findings

indicate that relationship program participationhas positive direct effects on intentions, usagelevels, retention, and customer sharedevelopment (Bolton, Kannan, and Bramlett2000; Garbarino and Johnson 1999; Verhoef2003).We also expect that relationship programparticipation enhances the positive effect ofsatisfaction on repurchase behavior. Customersenter relationships in part to reduce the timeand effort required for purchase decisions(Bhattacharya and Bolton 2000; Sheth and

Parvatiyar 1995), which suggests thatrelationship program participants should be


8/23

less inclined to shop around and more inclinedto allocate purchases to relational providers thatoffer superior satisfaction. This positivemoderating effect would extend to intentionsonly if customers accurately incorporated themoderating effect of relational programparticipation. Because we do not expect suchincorporation to occur, we hypothesize thefollowing:

H4: Relationship program participation (a)

moderates (enhances) the positiverelationship between customer satisfactionand objective repurchase behavior but (b)does not moderate the positive relationship

between customer satisfaction andrepurchase intentions.

Marketplace Characteristics

Marketplace moderators feature interactionsamong customers, the focal firm, andcompeting firms. For example, intensecompetition that spurs price promotions mayincrease switching behavior and overallpurchase volume, or new firms entering themarketplace may steal customers and marketshare from entrenched competitors. We examinethe convenience of the focal firms offering andits interaction with competitive intensity in the

marketplace.

Convenience. Overall convenience is a second-order construct that consists of five types oftime and effort costs involved in serviceexperiences (Berry, Seiders, and Grewal 2002).Empirical findings indicate that convenience issignificantly related to customer satisfactionand behavioral intentions (Andaleeb and Basu1994), consumer switching behavior (Keaveney1995), and customer perceptions and retention(Rust, Lemon, and Zeithaml 2004).

In addition to its direct effects, we propose thatconvenience enhances the positive effect ofsatisfaction on repurchase behavior but not onintentions. From a resource allocationperspective, a convenient offering conservescustomers time and effort and therebyfacilitates a satisfied customers ability to fulfillhis or her intent. In this capacity, conveniencefunctions less as an input to evaluation andmore as an ongoing barrier that encourages ordiscourages repurchase behavior. This is likely

to be particularly relevant for repatronagebehavior, for which access to geo- graphically

based retailers or other service firms is a majordecision factor, and can produce both plannedand unplanned trade-offs between degree ofconvenience and level of satisfaction. Thus:

H5: Convenience (a) moderates (enhances) the

positive relationship between customersatisfaction and objective repurchase

behavior but (b) does not moderate thepositive relationship between customersatisfaction and repurchase intentions.

Competitive intensity. We define competitiveintensity as the level of direct competition thatthe focal firm faces within its immediatebusiness domain. Competitive intensity canattenuate competitive advantage and influencerepurchase behavior over time because

competition erodes customers perceptions ofdifferential advantage along unsustainabledimensions. For example, conveniencerepresents a characteristic that can be readilyreplicated in many marketplaces; thus, therelative advantage it offers when competition islow is eroded as competition intensifies.

We illustrate the expected interaction using ananecdote about gas station competition andrepurchase. A consumer routinely travels threedistinct routes along which he or she makes

repurchase decisions. On the first route, there isonly one gas station; the convenience of theoffering may be paramount, so the travelerrepurchases at this gas station, especially if heor she is satisfied with the service station but,when necessary, even if he or she is not. On thesecond route, there are two gas stations onopposite sides of the road, both of which areopen with no waiting line; convenience maylead the traveler to repurchase at whicheverstation is on the side of the road in the directionhe or she is traveling. Alternatively, one of thecompetitors may deliver higher satisfaction onanother dimension, which would lead thetraveler to cross the road if necessary torepurchase from the same gas station. On thethird route, there are four gas stations locatedon the four corners of an intersection; each isopen without a waiting line. Convenience maycontinue to play a key role (e.g., stop at the firstone on the same side of the road that does nothave a line), but an alternative decision rulecould lead to convenience becoming

irrelevant.


9/23

This anecdote suggests a three-way interactionamong satisfaction, convenience, andcompetitive intensity. When competitiveintensity is low, convenience prevents defectionand facilitates repurchase behavior, thusexerting both a direct and a moderatinginfluence on repurchase. However, ascompetitive intensity increases, convenienceplays a less important role in the repurchasedecision. It is not clear whether competitiveintensity will have a significant direct effect,which would depend on whether customersperceive shopping synergies associated with alarge number of competitors in a singledestination, such as at a regional shop- pingmall. We do not expect that customers willincorporate these complex interactions into

repurchase intentions, which suggests thefollowing hypothesis:

H6: Competitive intensity (a) moderates the

relationships among customer satisfaction,convenience, and repurchase behavior suchthat convenience enhances the relationship

between satisfaction and repurchase behaviorwhen competitive intensity is low but notwhen competitive intensity is high but (b)does not moderate the relationship amongcustomer satisfaction, convenience, andrepurchase intentions.

RESEARCH DESIGN AND EMPIRICALRESULTS

To examine our hypotheses, we worked with anational specialty retail chain that sells its ownbrand of upscale womens apparel and homefurnishings in approximately 100 NorthAmerican locations. The company providedcontact information for 3117 customers andoffered a $20 coupon to customers whoresponded to the four-page questionnaire. Thecustomer list included randomly selectednames of customers who had purchasedmerchandise from any store during the 12weeks before the generation of the list. Thus,the sampling frame represents currentcustomers.

Contact information included names andaddresses for all 3117 customers and e-mailaddresses for 1150 customers who had joinedthe relationship program, which featured

frequent e-mails announcing newly arrivedmerchandise and promotions. We sent e-mail

messages to all 1150 e-mail addresses, invitingpotential respondents to click through to anonline survey. Of these 1150 addresses, 264 e-mails were returned as undeliverable, leaving aneffective sampling frame of 886. After twoweeks, we sent an additional e-mail tononrespondents, offering them another chanceto participate. We ultimately received 285surveys, for an effective response rate of 32%.We eliminated 12 respondents who providedincomplete information from subsequentanalyses, leaving a total of 276 usableresponses.

We sent postal mail to the other 1967 names onthe customer list. Of these, 28 were returned asundeliverable, leaving an effective sampling

frame of 1939. After four weeks, we sent afollow-up letter and survey to the non-respondents, offering them another chance toparticipate. A total of 721 people responded, foran effective response rate of 37%. Of these, 52incomplete surveys were unusable, leaving atotal of 669 usable responses. The 945respondents to both surveys were primarilywomen (99%) between the ages of 35 and 54years (66%) with at least some collegeeducation (96%) and an average householdincome exceeding $58,000.

Construct Measurement

We operationalized repurchase behavior usingtwo measures from the companys records: thenumber of repurchase visits and the amount ofrepurchase spending during the 52 weeks aftercompletion of the survey. The use of objectiverepurchase data for the year following thesurvey eliminates concerns of common methodvariance, simultaneity, or endogeneity. We logtransformed the repurchase behavior measures

to improve distribution normality.

Several independent measures were objectivesecondary data or single-item, self-reportedmeasures. We measured household income asthe median household income reported in the2000 census for the respondents zip code.Relationship age was a single-item measure(i.e., How long have you been a customer?). Relationship programparticipation was a dichotomous variableindicating whether the customer had opted in to

the companys e-mail program. Tooperationalize competitive intensity, we used


10/23

Census Bureau Zip Code Business Patterns datathat report the number of establishmentscompeting in each North American IndustryClassification System (NAIC; http://censtats.census.gov/cbpnaic/cbpnaic.shtml);using the respondents zip code, we includedthe total number of competitors in womensclothing (NAIC code 448120) and other homefurnishings (NAIC code 442299).

We adapted multi-item scales to measurerepurchase intentions (Parasuraman, Zeithaml,and Berry 1994), satisfaction (Voss,Parasuraman, and Grewal 1998), and involve-ment (Beatty and Talpade 1994). Because nocomprehensive convenience scale existed, wefollowed standard procedures to develop scale

items for each of the five convenience types(Berry, Seiders, and Grewal 2002). The multi-group confirmatory factor analysis that wereport in the Appendix supports the reliabilityand consistency of the scales (Voss andParasuraman 2003). We used mean scores forthe latent constructs in subsequent regressionanalyses.

In Table 2, we present descriptive statistics andconstruct correlations for the variables ofinterest. Comparison of the means for the postal

mail and e-mail samples indicates thatrelationship program participants are moreinvolved, have lower relationship ages, andengage in more repurchase visits and spending.These mean differences raise questions as towhether there are differences in the structuralrelationships of interest across the two samples.We conducted an exploratory analysis toaddress this. Of the 15 possible structuraldifferences across the three models (five foreach model: repurchase visits, spending, andintentions), only one was significant at the p

Documents

Seiders, Voss, Grewal and Godfrey.pdf