You are on page 1of 17

Understanding the Effect of Customer Relationship Management Efforts on Customer Retention and Customer Share Development Author(s): Peter

C. Verhoef Source: The Journal of Marketing, Vol. 67, No. 4 (Oct., 2003), pp. 30-45 Published by: American Marketing Association Stable URL: http://www.jstor.org/stable/30040548 . Accessed: 13/10/2011 06:48
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Marketing.

http://www.jstor.org

PeterC. Verhoef

the Effect Understanding of Customer Relationship Managemen

Efforts

on

Customer Retention and

CustomerShare

Development

Scholars have questioned the effectiveness of several customer relationshipmanagement strategies. The author effects of customer relationship perceptionsand relationship marketinginstrumentson investigates the differential customer retentionand customer share developmentover time. Customerrelationship perceptionsare considered evaluations of relationshipstrength and a supplier'sofferings,and customer share development is the change in customer share between two periods. The results show that affective commitmentand loyaltyprogramsthat provide economic incentives positively affect both customer retentionand customer share development, whereas directmailingsinfluencecustomer share development. However,the effect of these variables is rathersmall. The results also indicatethat firmscan use the same strategies to affect both customer retentionand customer share development.

relationships have been increasingly studied

in the academic marketing literature(Berry 1995; Customer Dwyer, Schurr, and Oh 1987; Morgan and Hunt 1994; Sheth and Parvatiyar 1995). An intense interestin customer relationships is also apparentin marketingpractice and is most evident in firms' significant investmentsin customer relationshipmanagement(CRM) systems (Kerstetter 2001; Reinartz and Kumar 2002; Winer 2001). Customer retentionrates and customer share are importantmetrics in CRM (Hoekstra, Leeflang, and Wittink 1999; Reichheld 1996). Customershare is defined as the ratioof a customer's purchases of a particularcategory of products or services from supplierX to the customer'stotal purchasesof thatcategory of productsor services from all suppliers(Peppersand Rogers 1999). To maximize these metrics, firms use relationshipmarketing instruments(RMIs), such as loyalty programs and direct mailings (Hartet al. 1999; Robertsand Berger 1999). Firmsalso aim to build close relationshipswith customersto enhance customers' relationship perceptions (CRPs). Although the impact of these tactics on customer retention has been reported(e.g., Bolton 1998; Bolton, Kannan,and Bramlett2000), there is skepticism about whethersuch tacof Peter Verhoef Assistant C. is Professor Marketing, of DepartmentMarUniof Erasmus and Rotterdam School Economics, keting Organization, and the Rotterdam. author The acknowledges financial versity, gratefully thanks datasupport a Dutch of financial services The company. author VerBasDonkers, Langerak, Leeflang, Fred Peter Loren Peeter Lemon, and JM for helpful Wittink, thefour anonymous reviewers their legh,Dick the of The also suggestions. author acknowledges comments research of Yale of seminar at the participants UniversityGroningen, School Manhe of and Finally, agement, Tilburg University, the University Maryland. and HansFranses his advisers, Philip acknowledges twodissertation for enduring Hoekstra, their Janny support.

tics can succeed in developing customer share in consumer markets(Dowling 2002; Dowling and Uncles 1997). Several studies have considered the impact of CRP on either customerretentionor customershare,but not on both (e.g., Anderson and Sullivan 1993; Bolton 1998; Bowman and and Narayandas2001; De Wulf, Odekerken-Schrider, Iacobucci2001). A few studies have consideredthe effect of RMIs on customer retention (e.g., Bolton, Kannan, and Bramlett2000). In contrast,the effect of RMIs on customer sharehas been overlooked.Furthermore, most studies focus on customer share in a particularproduct category (e.g., Bowman and Narayandas 2001). Higher sales of more of the same product or brand can increase this share; however, firms that sell multiple products or services achieve share increases by cross-selling other products. Moreover, no study has consideredthe effect of CRPs and RMIs on both customer retentionand customer share. It is often assumed in the literature thatthe same strategiesused for maximizing customer share can be used to retain customers; however, recent studies indicate that increasingcustomershare might require different strategies than retaining customers (Blattberg,Getz, and Thomas 2001; Bolton, Lemon, and Verhoef 2002; Reinartzand Kumar2003). Prior studies have used self-reported, cross-sectional data that describe both CRPs and customer share (e.g., De and Wulf, Odekerken-Schrider, Iacobucci2001). The use of such data may have led to overestimationof the considered associations because of methodological problems such as carryoverand backfireeffects and common methodvariance (Bickart 1993). Such datacannotestablish a causal relationship; indeed, the argument could be made that causality works the other way (i.e., I am loyal, therefore I like the company) (Ehrenberg1997). Longitudinaldata ratherthan cross-sectional data should be used to establish the causal relationshipbetween customershare and its antecedents.
Journal of Marketing Vol. 67 (October 2003), 30-45

30/ Journal Marketing, of October 2003

I have the following researchobjectives: First, I aim to understand effect of CRPs and RMIs on customerretenthe tion and customer share development over time. Second, I examine whetherthe effect of CRPs and RMIs on customer retentionand customer share development is different.My dataon CRPs, operationaldata study analyzes questionnaire on the applied RMIs, and longitudinal data on customer retention and customer share of a (multiservice) financial service provider.

Literature Review
CRPsand CustomerBehavior
Table 1 providesan overview of studies thatreportthe effect of CRPs on customerbehavior,and it describes the dependent variables,the design and context of the study,the CRPs studied, and the effect of CRPs on behavioralcustomerloyalty measures(which can be self-reportedor actualobserved loyalty measures).Table 1 shows that the results of studies that relate CRPs to actual customerbehaviorare mixed. RMIs and Customer Behavior Table 2 provides an overview of the limited numberof academic studies thatconsiderthe effect of RMIs. The majority of the studies have focused on loyalty or preferentialtreatment programs,and the results show mixed effects of these programson customer loyalty. Despite the intensive use of direct mailings in practice, their effect on customer loyalty has almost been ignored.More important, effect of RMIs the on customer share development over time has not been investigated.

tomer relationship largely depends on the applied RMIs and Bolton 2000; Christy, Oliver, and Penn (Bhattacharya and 1996; De Wulf, Odekerken-Schridder, Iacobucci 2001). Moreover, because of the increasing popularity of CRM among businesses, an increasing numberof firms are using RMIs. In the model, I also include customers'past behaviorin the relationship as control variables, which might capture inertiaeffects that are consideredimportantdeterminants of customerloyalty in business-to-consumer markets(Dowling and Uncles 1997; Rust, Zeithaml, and Lemon 2000). Past customer behavioral variables (e.g., relationshipage, prior customer share) can also be indicators of past behavioral loyalty, which often translates into future loyalty. Prior research suggests that the type of productpurchasedin the past is an indicator of future cross-selling potential (e.g., Kamakura, Ramaswami,and Srivastava1991).

Hypotheses
CRPs Relationship marketingtheory and customer equity theory posit that customers' perceptionsof the intrinsic quality of the relationship(i.e., strengthof the relationship)and customers' evaluations of a supplier's offerings shape customers'behaviorin the relationship(Garbarino Johnson and 1999; Rust, Zeithaml, and Lemon 2000; Woodruff 1997). The most prominentperceptionrepresentingthe strengthof the relationshipis (affective) commitment(Moorman,Zaltman, and Desphandd 1992; Morgan and Hunt 1994). Because satisfactionand paymentequity are importantconstructswith respectto the evaluationof a supplier'sofferings (Bolton and Lemon 1999), I included these three constructs in the model. The two categories of constructs differ in terms of both content and time orientation:Affective commitment is forwardlooking, whereas satisfaction and payment equity are retrospectiveevaluations. In the customerequity and relationshipmarketingliterature, other CRPs that are not included in my model are often studied. Trust and brand perceptions are the most prominentof these variables(Morganand Hunt 1994; Rust, Zeithaml,and Lemon 2000). I did not include brandperceptions because the focus is on currentcustomers. My contentionis thatthe brandis especially significantin attracting new customers.During the relationship,the brandprobably influences affective commitment(Bolton, Lemon, and Verhoef 2002). I did not include trust, because trust should be considered merely an antecedent of satisfaction and commitment (Geyskens, Steenkamp, and Kumar 1998). No direct effect on customerbehaviorshould be expected. Affective Commitment Commitmentis usually defined as the extent to which an exchange partnerdesires to continue a valued relationship (Moorman,Zaltman, and Desphand6 1992). I focus on the affective componentof commitment,that is, the psychological attachment, based on loyalty and affiliation, of one exchange partner to the other (Bhattacharya,Rao, and Glynn 1995; Gundlach,Achrol, and Mentzer 1995). CustomerRelationship Efforts/ 31 Management

ConceptualModel
Figure 1 shows the conceptualmodel. In this model, I consider customer retention and customer share development between two periods (T1 andTo) as the dependentvariables, which are affected by CRPs and RMIs. Because I consider customerretentionand customer share developmentas two separate processes, relationshipmaintenance and relationship development, the underlyinghypotheses of the model explicitly predictthat differentconstructsof CRPs, and different RMIs influence customer retention and customer share development.The rationalefor this distinction is that a customer's decision to stay in a relationshipwith a firm may be differentfrom his or her incrementaldecision to add or drop existing products.Consistentwith this notion, Blattberg, Getz, and Thomas (2001) argue that customerretention is not the same as customer share, because two firms could retainthe same customer.Reinartzand Kumar(2003) suggest thatrelationshipdurationand customershareshould be considered as two separatedimensions of the customer relationship. Bolton, Lemon, and Verhoef (2002) propose thatthe antecedentsof customerretentionmight be different from the antecedentsof cross-buying behavior.I explicitly addressthese differences in the hypotheses. The inclusion of CRPs as antecedentsof retention and customer share development is based on relationshipmarketing theory, which suggests that CRPs affect behavioral customerloyalty. I includedRMIs because a successful cus-

relationship customer onof Additional Results/Comments depends orientation Effect

satisfaction of effect

Quadratic

and service mediated age enhanced age positively important important consumer enhanced are by are differences differences satisfaction satisfaction usage equity satisfaction of expected of equity satisfaction relationship characteristics firms satisfaction firms relationship of of by Effect future affects other othermoderated Payment Effect by payment PerformanceEffect Performance Effect

with

with

by

(+) Perceptions (Effect) Loyalty Included

(+) (+) (+),

(-) (+), (+)

(+) (+)

(+)

(0)

(+) (+) (+), (+) equity

(0)

(+)

(+) (0) (+), (0), equity equity

quality

quality Benefits Relationship Satisfaction Satisfaction Satisfaction Satisfaction Satisfaction Satisfaction commitment commitment Satisfaction CommitmentSatisfaction Satisfaction Satisfaction CommitmentSatisfaction Service payment payment payment

Behavioral card card services brands visitors Contextindustries industries market market Retailing Retailing Channels Credit Car association Car Professional Credit Entertainment 1 CRPs Study entertainment Grocery Theater Financial Various Various Telecommunications Telecommunications, of on TABLE Effect on Design Longitudinal Longitudinal LongitudinalLongitudinalLongitudinal Longitudinal Longitudinal Longitudinal StudiesStudy Experiment Cross-sectional Cross-sectional Cross-sectional Cross-sectional Cross-sectional Cross-sectional Cross-sectional of

and and and and Overview (1994) and and and (1996) (2001) Johnson and and (2001) Studies Sullivan Lemon (2000) (1997) (2001) (2000) Hunt Berry, (1998) Franses, Kamakura of (1999) (2002) (2000) and and Kumar, OdekerkenWhite, and Kannan, Kannan, (1999) (1993) (2001) (1999) Summers, and and Hoekstra Tsiros Wulf, Bowman AcitoBolton Bramlett Winer Schri6der, Bramlett Macintosch Lockshin lacobucci Verhoef, Mittal, Bolton Narayandas Bolton, Bolton, and Parasuraman Mittal Lemon, De Examples Gruen, Anderson Zeithaml, Morgan Garbarino

Loyalty

intentions

share

retention relationship and/or duration Customer Observed

usage Service Cross-buying

Purchase Behavioral Measurement Self-Reported

Customer

32/ Journal Marketing, of October 2003

Effect on customer retention.Given the previous definition of affective commitment,it might be expected that this type of commitmentaffects customerretentionpositively.In line with this, researcherswho relate commitment to selfreportedbehavior,such as purchaseintentions,usually find that commitment positively affects customer loyalty (e.g., Garbarinoand Johnson 1999; Morgan and Hunt 1994). of However,the appearance such an effect has recentlybeen questioned (Gruen,Summers,and Acito 2000; MacKenzie, Podsakoff, and Ahearne 1998). Despite this, I hypothesize the following: H1:Affective commitment positively affects customer retention. Effect on customer share development. Relationship marketingtheory posits that because affectively committed customers believe they are connected to the firm, they display positive behavior toward the firm. As a consequence, affectively committedcustomersare less likely to patronize other firms (Dick and Basu 1994; Morgan and Hunt 1994; Sheth and Parvatiyar1995). In other words, committedcustomersare more (less) likely to increase(decrease)theircustomer share for the focal supplierover a period of time. share H2:Affectivecommitment positivelyaffectscustomer development time. over

from a particularsupplier (Oliver and Winer 1987). This depends on, among otherthings, the customer'ssatisfaction level. As a consequence, customers who are more satisfied are more likely to remaincustomers.Thus: affectscustomer retention. H3:Satisfaction positively Effecton customershare development.Although a positive relationship between satisfaction and customer share has been demonstratedin a single product category (Bowman and Narayandas2001), this does not necessarily imply that satisfactionalso positively affects customersharedevelopment for a multiservice provider.A theoretical explanation for the absence of such an effect could be that positive evaluations of currentlyconsumed products or services do not necessarilytransferto otheroffered productsor services. In otherwords, satisfiedcustomersare not necessarily more likely to purchaseadditionalproductsor services (Verhoef, Franses, and Hoekstra 2001). Another explanation is that thoughcustomerretentionrelatesto the focal supplieralone, customer share development also involves competing suppliers. As a result, developmentof a customer'sshare might be affected more by the actions of competing suppliersthan by the focal firm's priorperformance.Thus, I do not expect satisfaction to have a positive effect on customer share development.

Satisfaction
I define satisfaction in this study as the emotional state that occurs as a result of a customer'sinteractionswith the firm over time (Anderson,Fornell, and Lehmann 1994; Crosby, Evans, and Cowles 1990). Szymanski and Henard's(2001) meta-analysis shows that satisfactionhas a positive impact on self-reportedcustomerloyalty. Despite such positive results in the literature,the link between satisfaction and actual customer loyalty has been questioned (e.g., Jones and Sasser 1995). Researchershave searchedfor a betterunderstanding this link andhave proof a nonlinearrelationshipbetween satisfactionand cusposed tomer behavior (e.g., Anderson and Mittal 2000; Bowman and Narayandas2001). Other studies have shown that relationship age, product usage, variety seeking, switching costs, consumer knowledge, and sociodemographics(e.g., age, income, gender)moderatethe link between satisfaction and customerloyalty (Bolton 1998; Bowman and Narayandas 2001; Capraro, Broniarczyck, and Srivastava 2003; Homburg and Giering 2001; Jones, Mothersbaugh, and 2001). Finally,dynamics Beatty 2001; Mittaland Kamakura duringthe relationshipmay also affect this link. Customers update their satisfaction levels using informationgathered during new interactionexperiences with the firm, and this new informationmay diminish the effect of prior satisfaction levels (Mazurskyand Geva 1989; Mittal, Kumar,and Tsiros 1999). Effect on customer retention. Despite the apparent absence of an empiricallink between satisfactionandbehavioral customerloyalty, several studies show that satisfaction affects customer retention (Bolton 1998; Bolton, Kannan, and Bramlett 2000). The underlying rationale is that customers aim to maximize the subjective utility they obtain

PaymentEquity
Paymentequityis defined as a customer'sperceivedfairness of the price paid for the firm's productsor services (Bolton and Lemon 1999, p. 173) and is closely related to the customer'sprice perceptions.Paymentequity is mainly affected by the firm's pricing policy. As a result of its groundingin fairness, a firm's payment equity also depends on competitors' pricing policies and the relative quality of the offered services or products. Effect on customer retention. Higher payment equity (i.e., price perceptions)leads to greaterperceived utility of the purchased products or services (Bolton and Lemon 1999). As a resultof this greaterperceivedutility,customers should be more likely to remain with the firm. Consequently, payment equity should have a positive effect on customerretention.This is consistentwith empiricalstudies that show that payment equity positively affects customer retention (Bolton, Kannan,and Bramlett 2000; Varukiand Colgate 2001). Thus: affectscustomer retention. H4:Payment equitypositively Effect on customer share development. Although I expect paymentequity to have a positive effect on customer retention,I do not necessarily expect this to be true for customer share development. There are two reasons payment equity may have no effect on customer share development. on First, literature price perceptionssuggests that customers with higher price perceptionsare more likely to search for betterprices (Lichtenstein,Ridgway, and Netemeyer 1993). Intuitively,the suggestion that such customersare less loyal makes sense. For example, customers of discounters (with high scores on price perceptions) are known to visit the greatestnumberof stores in their searchfor the best bargain.

Efforts Customer Relationship Management / 33

of effect rates effect positive Results No purchase on Short-term effect effect programs retention usage on effect effect No service and Positive

convincing Positive loyalty No

card Study Grocery brands Context Retailing Retailing Airlines Credit Retailing

design Design Loyalty Panel Study

direct survey, onuse mail

data panel

on use survey

on survey treatment

program perceptions perceptions programs Longitudinal

loyalty data, data, perceptions Aggregated Cross-sectional Cross-sectional preferential Cross-sectional Behavioral 2 on RMIs TABLE of share, share purchase Measure shares data service

share penetration, purchase intentions buyers customer usage retention, Effect empirical sole Customer Customer Loyalty No on average Purchase Aggregated Aggregated frequency, Customer

Studies

(1987) (2001)

(1997) (1997)

Lemon Bramlett and (2000) and

(2001)

Uncles Sharp Study Shoemaker and Odekerken-Schrider, lacobucci and and and Wulf, Sharp Dowling Bawa De

(2000) lacobucci Odekerken-Schrider, Zeithaml,Kannan, and Wulf, Rust, Bolton,De

mail RMIDirect

programs Loyalty

34/ Journal Marketing, of October 2003

FIGURE1 Conceptual Model


CRPs Satisfaction H3 Ha H1
/H2

RMIs

Customer retention To T,

H6a
H6b

Loyalty program

Payment equity

A Customer share TO T,

IH5

Directmailings

Affectivecommitment

ControlVariables Customer shareTo age Relationship To Typeof servicepurchased TO Accordingto this reasoning,customerswith betterprice perceptions are more likely to decrease customer share over time. Second, as is satisfaction,a customer'spaymentequity is based on the customer's awareness of the prices of services or productspurchasedfrom the focal firm in the past (Bolton, Lemon, and Verhoef2002). However,the prices of additionalservices or productsfrom the focal suppliermight be different from the currentlypurchasedservices or products. Therefore,a high payment equity score may not indicate that the customer will purchase other products or services from the same supplier.As a consequence, I do not expect payment equity to affect customer share development. RMIs Bhattacharyaand Bolton (2000) suggest that RMIs are a subset of other marketinginstrumentsthat are specifically aimed at facilitating the relationship,and they distinguish between loyalty or reward programs and tailored promotions. In addition, RMIs can be classified according to Berry's (1995) first two levels of relationshipmarketing.At the first level (Type I), firms use economic incentives, such as rewards and pricing discounts, to develop the relationinclude more ship. At the second level (TypeII), instruments firms attempt social attributes. using Type II instruments, By to give the customerrelationshipa personaltouch. In this study,I focus on two specific Type I RMIs:direct mailings and loyalty programs.Direct mailings usually are personallycustomizedoffers on productsor services thatthe customer currentlydoes not purchase.In most cases, price discounts or other sales promotions(e.g., gadgets) are used to entice the customerto buy. I focus on direct mailings that are a "call to action"ratherthan only a reinforcingmechanism for the relationship(e.g., thank-youletters). The loyalty programI include in the study is a rewardprogramthat providesprice discountsbased on the numberof productsor services purchasedand the length of the relationship. Direct Mailings Direct mailings have some unique characteristics:enablement of personalized offers, no direct competition for the attentionof the customer from other advertisements,and a capacity to involve the respondent (Roberts and Berger 1999). Because direct mailings focus on creatingadditional sales, I do not expect them to influence customerretention. Moreover,the datado not enable me to relatedirectmailings to customerretention. Effect on customershare development.There are several theoretical reasons direct mailings should positively influence customersharedevelopment.First, direct mailings can create interestin a (new) service and therebylead to a final purchase(Robertsand Berger 1999). Second, the personalization affordedby direct mailings may increase perceived with relationshipquality,because customersare approached individualizedcommunicationsthat appeal to their specific needs and desired manner of fulfilling them (De Wulf, Odekerken-Schriider, and Iacobucci 2001; Hoekstra, Leeflang, and Wittink 1999). Third, according to the sales the rewards(i.e., price dispromotionsliterature, short-term counts) offered by direct mailings may motivate customers to purchaseadditionalservices and thus increase customer share.In supportof this claim, Bawa and Shoemaker(1987) report short-termgains in redemptionrates of direct mail coupons. I hypothesize the following: sharedevelopaffectcustomer positively Hs:Directmailings mentovertime. Loyalty Programs Effect on customer retentionand customer share development. There are several theoretical reasons the rewardbased loyalty programbeing studiedshouldpositively affect both customer retention and customer share development. First, psychological investigationsshow that rewardscan be highly motivating(Lathamand Locke 1991). Researchalso shows that people possess a strong drive to behave in whatCustomerRelationship Efforts135 Management

ever mannernecessary to achieve future rewards(Nicholls 1989). According to Roehm, Pullins, and Roehm (2002, p. 203), it is reasonableto assume that duringparticipationin a loyalty program,a customer might be motivatedby program incentives to purchase the program sponsor's brand repeatedly. Second, because the program'srewardstructureusually depends on prior customer behavior, loyalty programscan providebarriersto customers'switching to anothersupplier. For example, when the reward structuredepends on the length of the relationship,customersare less likely to switch (because of a time lag before the same level of rewardscan be received from another supplier). It is well known that switching costs are an important antecedent of customer loyalty (Dick and Basu 1994; Klemperer1995). Despite the theoreticalargumentsin favorof the positive effect of loyalty programson customer retention and customer share development, several researchershave questioned this effect (e.g., Dowling and Uncles 1997; Sharpand Sharp 1997). In contrast, Bolton, Kannan, and Bramlett (2000) and Rust, Zeithaml, and Lemon (2000) show that loyalty programshave a significant, positive effect on customerretentionand/orservice usage. In this study,I build on the theoreticalargumentin favor of the positive effect that loyalty programshave on customer retentionand customer share development. H6:Loyaltyprogram membership positivelyaffects(a) customerretention (b) customer and share development.

products, and customer characteristics.In the second (T1) survey, I collected data on customer ownership of various insuranceproducts. Although the company whose data I used offers other products,such as loans, I limited the studyto the categoryof insuranceproducts.The rationalefor this limitation is that customersusually buy each type of insuranceproductfrom a single insurancecarrier(i.e., insurancetype X [life insurance] from insurance carrier Y [i.e., Allianz Life Company]), but this does not necessarily hold for otherfinancial products or services. For example, it is well known that many customers have savings accounts at several financial institutions. Moreover, the insurance market is the most importantmarketfor this company in terms of the number of customers and customer turnover(approximately90%). As a result of this choice, the sample is restrictedto those customers who purchaseinsuranceproductsonly from the company.This resultedin a usable sample size of 1677 customersfor the firstmeasurement(To)and 918 for the second measurement(T1).

Contentsof the CompanyCustomerDatabase


The company'scustomerdatabaseprovideddata on the past behavior of individual customers and the company RMIs directed at individualcustomers. The past customerbehavior data cover two periods. The first period starts at the beginning of a relationshipbetween the company and the customer and ends at To (this period differs among customers). The data on past customerbehavior included variables such as numberof insurancepolicies purchased,type of insurance policies purchased, and relationship length. The second period covers the interval between To and T1. For this period,the databaseprovideddata aboutwhich customers left the company and the numberof company insurance policies a customerowned at T1. The company's customer databasecontains the following informationon RMIs: loyalty programmembershipat To and the numberof direct mailings sent between To and T1. Every customer who purchases one or more financial

ResearchMethodology
Research Design
I combined survey data from customersof a Dutch financial services company with data from that company's customer database.I used a panel design, displayedin Figure2, to collect the data. I collected the survey data at two points in time: To andT1. I used the first (To)surveyto measureCRPs of the company, customer ownership of various insurance

FIGURE 2

Panel Design
DatafromCustomer Database

Startof Relationship

T0 0 (Survey1Among Customers)

T1 (Survey2 Among Customers Interviewed in Survey1)

361Journal Marketing, of October 2003

services from the company can become a member of the loyalty program (an opt-in program).At the end of each year, the programgives customersa monetaryrewardbased on the numberof services purchasedand the age of the relationship. Because the companyuses regression-typemodels to select the customers with the highest probability of respondingto direct mailings, the numberof directmailings sent differs among customers.

Validation CRPs of
The final measuresare reportedin the Appendix.The scales for commitmentand satisfactionhave reasonablecoefficient alphas.For paymentequity,I reporta correlationcoefficient of .49, which is not considerablyhigh.2 However,note that the reportedcomposite reliabilities of all scales are sufficient (Bagozzi and Yi 1988). I applied confirmatoryfactor analysis in Lisrel 83 to furtherassess the quality of the measures (Jdreskogand Sorbom 1993), and I achieved the fol= lowing model fit: X2 217.4 (degrees of freedom [d.f.]) = 51,p <.01), X2/d.f.= 4.26 (d.f. = 1,p < .05), goodness-of-fit index = .98, adjustedgoodness-of-fit index = .97, comparative fit index = .98, and root mean squareerrorof approximation = .04. These fit indexes satisfy the criteriafor a good model fit (Bagozzi andYi 1988; Baumgartner Homburg and difference tests on the respective fac1996). A series of X2 tor correlationsprovided furtherevidence for discriminant validity (Andersonand Gerbing 1988). On the basis of these results, I summed the scores on the items of each construct. The means, standarddeviations, and correlationmatrix are shown in Table 3.

CustomerSurvey DataCollection
At To, customer survey data were collected by telephone from a randomsample of 6525 customersof the company.A quotasamplingapproachwas used to obtaina representative sample. I received datafrom 2300 customers(35% response rate). After those responses with too many missing values were deleted, a sample size of 1986 customersremained.At T1, I again collected data from those customers, except for those who left the company between To and T1. In the second data collection effort, 1128 customers were willing to cooperate (65% response rate). To assess nonresponsebias at T1, I tested whetherrespondentsand nonrespondents differed significantlywith respect to customershare at To.A ttest does not reveal a significantdifference (p = .36). Thus, I conclude that there is no nonresponsebias.

Measurement DependentVariable of
An often-used method of measuringcustomer share is asking customersto reportthe numberof purchasesof the focal brandthey normallymake (Bowman and Narayandas2001; De Wulf, Odekerken-Schrider, Iacobucci 2001). In this and study, I sought a more objective measure. In line with the of conceptualization customershare,I define customershare of customeri for supplierj in category k at time t as

Measurement CRPs of
For the measurementof CRPs (i.e., affective commitment, satisfaction, and payment equity), I adaptedexisting scales to fit the context of financialservices. For the affective commitmentscale, I adapteditems from the studies of Anderson and Weitz (1992), Garbarino and Johnson (1999), and Kumar,Scheer,and Steenkamp(1995). To measuresatisfaction, I adapted Singh's (1990) scale and added four new items. Finally, I based the payment equity scale on items adapted from Bolton and Lemon's (1999) and Singh's (1990) studies. To assess construct validity and clarify wording, the original scales were tested by a group of 12 marketingacademics and 3 marketingpractitioners familiarwith customer relationships.Subsequently,the scales were tested by a random sample of 200 customersof the company.On the basis of interitem correlations,item-to-total correlations,coefficient alpha, and exploratoryand confirmatoryfactor analysis, I reducedthe set of items in each scale.1 and method, (1991) proposed 11follow Steenkamp vanTrijp's factoranalysisand then confirmatory factor using exploratory to constructs. analysis validate marketing

21 reportcorrelation coefficientratherthan Cronbach's alpha because usedonlytwoitems.Cronbach's I is to alpha designed test theinteritem of reliability a scaleby comparing everycombination of eachitemwithall otheritemsin the scaleas a group.Because thereis no group withwhicheachitemcanbe compared a twoin item scale (onlythe otheritem),Cronbach's alphais meaningless for two-item scales.It mightalsobe argued one of the single that itemswouldbe bettersuitedfor measuring construct the froma
content validity perspective. To check this, I also estimated the

models(see Tables and5) with a singleitem as an antecedent. 4 Forbothitems,theeffectof payment remained equity insignificant
in the two models. Because in general multiple-itemmeasurement is preferred over single-item measurement, I report the model results of the summatedtwo-item scores.

TABLE 3 Means (Standard Deviation) and Correlation Matrix Independent Variables Mean X1 Commitment X2 Satisfaction X3 Paymentequity X4 Directmail X5 Loyaltyprogram X6 Log customer share To
*p < .05. **p< .01.

X1 1.00 .37** .14** .01 .09** .12**

X2 1.00 .21** .02 .14** .09**

X3

X4

X5

X6

2.96 3.75 3.41 3.51 .30 -.152

(.77) (.44) (.56) (2.12) (.46) (.66)

1.00 .01 .03 .06*

1.00 .56** .48**

1.00 .53**

1.00

Customer Efforts /37 Relationship Management

(1)

Number services of in category k purchased atsupplier at timet j = Customersharei~j~k,t Number services of in k purchased category at fromall suppliers timet

Data for the numeratorwere available from the company customerdatabase;however, data for the denominatorwere generally not stored in the company customer database. Therefore,I asked customers in the survey which insurance products(of both the company and competitors)they owned at To and at T1.

Analysis
The theoreticaldistinction between customer retentionand customersharedevelopmenthas implicationsfor my analysis. As a result of this distinction, I use a dual approach.I firstestimatea probitmodel to explain customerretentionor defection for the remainingsample afterTo (N = 1677). Second, I use a regression model to explain customer share developmentover time for the customers who remain with the company.A serious issue with this type of approachis that the explanatoryvariablesexplaining customerretention also explain customer share development. As a consequence, the regression parametersmay be biased (Franses and Paap 2001). I apply the Heckman (1976) two-step procedure to correct for this bias. Using this procedure, I include the so-called Heckman correction term (or inverse Mills ratio) in the regression model for customer share development.This correctionterm is calculatedby means of outcomes of the probit model for customer retention.This modeling approach is also known as the Tobit2 model (Fransesand Paap 2001). Because the inclusion of this correction term may cause heteroskedasticity,I apply White's (1980) method to adjust for heteroskedasticity.Another issue with the approachis that restrictingthe sample in the customersharedevelopmentregressionmodel to remaining customersmight restrictthe potentialvariancein the dependent variable,thus affecting the estimationresults.To assess whetherthis is true, I calculatedthe standarddeviationsfor the restricted and unrestricted sample. The differences between standard deviationsin customersharedevelopment are small: .10 for the unrestrictedsample, including defectors, and .09 for the restrictedsample. In the empiricalmodeling, I furtherassess this issue by estimatingthe customer share development model for the unrestrictedsample and comparingthe results with those of the restrictedsample. Because I am interestedin the changes in customershare over time, I use a difference model to test the hypotheses (Bowman and Narayandas2001). In line with the literature on marketshare models, the difference between the logs of customer share at T1 and To (CSo, CS1) is the dependent variablein the regressionmodel. This variablecan be interpreted as the percentagechange in customer share over the measuredperiod. In both the probit model for customer retentionand the regression model for customer share development of the customers who remain with the company,I use a hierarchical modeling approach.I include the past customerbehavior

covariates (past behavior)as independentvariablesand the mean-centeredcomposites of the items in the relationship perception scales (perceptions;e.g., affective commitment, satisfaction, payment equity). Finally, I include RMIs. For the loyalty program,I constructed a dummy variable that indicatedwhetherthe customerwas a memberof the loyalty programat To. I dealt with the number of direct mailings sent to a customer as follows: Because the company stops direct mailing customers when they defect, the numberof directmailings was not includedin the probitmodel for customer retention.Because customersleave duringthe period covered in the study, the numberof mailings could be correlated with defection. However, this correlationis not due to the positive effect of direct mailings on customer retention; rather,it is the result of the company's mailing policy. The foregoing results in the following two equations:
(2) P(retention= 1) = + alpast behavioro+ a2perceptionso o0 + a3RMIso 1, and -

(3)

Log(CS1)- log(CSo) = 1o + 1lpast behavioro + f2perceptionso+ 33RMIso 1 + P4Heckmancorrection. -

In Equations2 and 3, I providethe formulationof the model in the form of matricesin which each a or Pmay comprise several separateparameters. example, in the case of [2, For there are three different parametersfor the effect of commitment,satisfaction,and paymentequity.

HypothesisTesting
CustomerRetention
Approximately6.4% of the 1677 customers in the sample defected duringthe period of the study.3I reportthe estimation results of Equation3 in Table4. The first model (which only includes controlvariableswith respect to past customer behavior)explains approximately17%of the varianceand is significant(p < .01). The coefficients of the includedcontrol variables intuitively have the expected signs. Customers with high prior customer shares and lengthy relationships are less likely to defect. Furthermore,the ownership of a coinsurance, damage insurance, car insurance, and/or life insuranceproducthas a positive effect (p < .05). In the second model, including CRPs, McFadden R2 increases by approximately1%(p = .06). Only affectivecommitmenthas a significant,positive effect (p < .01) on customerretention, in supportof H1. I found no effect for either satisfactionor paymentequity.These results do not supportH3 or H4. Following Bolton (1998), I also explored whetherrelationship age moderates the effect of satisfaction. The estimation resultsindicatethatthe interactiontermbetween satisfaction

3Thesample 1677for the analysis the antecedents cusof of of tomerretention muchlargerthanthe sampleused in the cusis tomersharedevelopment dataabout model,becausebehavioral customers' purchase behavior were unnecessary the cusin past tomerretention customers who did not analysis.Consequently, in can in respond the secondsurvey be included thisanalysis.

38 / Journalof Marketing, October2003

TABLE4 Probit Model Results for Customer Retention (N = 1677)


Variable Constant Log customer share To Log relationship age Coinsurance Damage insurance Car insurance Lifeinsurance Perceptions Commitment Satisfaction Paymentequity RMIs Loyaltyprogram McFaddenR2 AdjustedMcFaddenR2 Likelihood ratiostatistic (d.f.) Akaikeinformation criterion < .05. *p **p<.01. H1(+) H3(+) H4(+) H6a(+) .168 .165 127.64** (6) .384 .178 .173 135.20** (9) .383 Hypothesis (Sign) Model 1 (z-Value) 1.66 .34 .11 .12 .78 .36 1.02 (5.10)** (2.23)* (2.32)* (1.21) (4.13)** (2.97)** (4.00)** Model 2 (z-Value) 1.68 .33 .11 .12 .78 .33 1.01 (5.03)** (2.13)** (2.14)* (1.22) (4.06)** (2.70)** (3.99)** Model 3 (z-Value) 1.58 .30 .09 .11 .74 .33 1.00 (4.58)** (1.89) (1.79) (1.04) (3.92)** (2.72)** (3.95)**

.21 (2.66)** -.21 (1.52) -.03 (.26)

.20 (2.58)** -.22 (1.63) -.03 (.30) .38 (2.02)* .184 .179 139.68** (10) .382

and relationshipage is significant(c = .28; p = .01), in support of the idea that relationshipage enhances the effect of satisfaction. In the third model, with the loyalty program included,McFaddenR2 increasesby approximately1%(p < .05). I found the loyalty programto have a significant,positive effect (p < .05), in supportof H6a. Customer Share Development for Remaining Customers Figure 3 shows the changes in customer share for the customers who did not defect. Although on averagechanges in customer share are almost zero, I observed changes in customer share for approximately68% of the customersin the in sample (N = 918). The distribution Figure 3 is symmetrical. For 34% of customers in the sample, I observed negative changes, and for approximately34%, their customer shares increased.As a logical consequence, the average for changes in customer share is zero (i.e., the mean values for customer share at To and T1 have approximatelythe same value of .285). The regression results of Equation 3 are reported in Table 5. The first model (including past customerbehavior) explains approximately 10% of the variance in customer share changes. The log of customer share at To has a negative effect on changes in customershare(p < .01). Thus, customers with large (small) customersharesare more likely to decrease (increase) their customer share in the next period. Customers who own damage insurance, car insurance, or coinsuranceare more likely to increase their customershare (p < .01). The estimationresults of the second model (which includes CRPs) show that affective commitmenthas a sig-

3 FIGURE Customer Share Development (N = 918)


500

400

300

200

100

0 -.25 .00 .25 .50 .75

nificant,positive effect on customersharedevelopment(p < .05). Thus, I find supportfor H2. However, I found no significant effect for either satisfaction or payment equity. These results are in line with my expectations that such CRPs do not directly affect customershare development.In the thirdmodel (which includes RMIs), the loyalty program has a significant,positive effect on customershare development (p < .05). Direct mailings also positively affect cus-

CustomerRelationship Efforts/ 39 Management

TABLE5 Regression Model Results of Changes in Customer Share (N = 918)


Variable Constant Heckmancorrection Log customershare To Coinsurance Damage insurance Car insurance Legal insurance Perceptions Commitment Satisfaction Paymentequity RMIs Loyaltyprogram Directmailing R2 AdjustedR2 F-value
*p < .05. **p< .01.

Hypothesis (Sign)

Model 1 (t-Value) -.44 .06 -.17 .02 .14 .04 .03 (6.84)** (.90) (9.97)** (3.83)** (6.35)** (2.69)** (1.15)

Model 2 (t-Value) -.46 (7.09)** .07 (1.27) -.19 (10.3)** .02 (3.92)** .15 (6.52)** .01 (2.29)* .03 (1.16) .03 (2.55)* .00 (.01) -.01 (.85)

Model 3 (t-Value) -.52 (7.80)** .10 (1.48) -.20(11.0)** .02 (3.26)** .14 (6.09)** .04 (2.52)** .03 (1.16) .03 (2.58)** -.00 (.21) -.01 (.66) .04 (2.22)* .01 (2.31)*

H2(+) H3(+) H4(+) H5(+) H6(+) .10 .10 16.95**

.11 .10 12.21**

.13 .12 11.72**

tomer share development(p < .05).4 Thus, both H5 and H6b are supported. The Heckman (1976) correctionterm is not significant, which implies that selecting only the remainingcustomers does not affect the estimation results (Franses and Paap 2001). It might be arguedthat leaving out defectors would reducevariancein the customersharedevelopmentmeasure, which in turn might affect the estimation results. To assess this issue further,I also estimateda model that included the defectors.5However,thereare two problemswith the model. 4Anissue in estimating effectof directmailings thatthe the is whosedataareuseddoesnotrandomly selectcustomers company
to receive such mailings; the company uses models to target the most receptivecustomers.These models are not known. The company's use of such models might lead to an endogeneity problem, which could result in (upwardly biased) inconsistent parameter estimates for direct mailings. To test for possible endogeneity, I used the Hausmantest that Davidson and MacKinnon(1989) propose. This test does not reveal any evidence for endogeneity (p =

First, I cannot include direct mailings as an explanatory variablebecause, as I noted previously,no mailings are sent to defectors.Second, because the log of 0 does not exist, the differencesin logs of customersharebetween T1 and To for defectors cannotbe calculated.A solution to this problemis to impute a share value that is close to 0 (e.g., .001). I used this approachand imputedseveral differentvalues to assess the stabilityof the results, and the results remainedthe same for the different imputations.The estimation results for an imputedvalue for customershareat T1 for defectorsof .001 show that the coefficients of affective commitmentand the loyalty programremain significant,but there is no effect of satisfactionor paymentequity. The R2 of the model is .09, which is lower than the R2 of .12 of the model that includes only the remaining customers reported in Table 5. Given
these results, I conclude that restricting the sample to remaining customers does not affect the hypotheses-testing results.

.88).
this 5Notwithstanding result, I also used two approachesto correct for possible endogeneity. The first approachapplied instrumental variablesusing two-stage least squaresin the estimationof a system of two equations (Pindyck and Rubinfeld 1998). I used two sociodemographicvariablesas instrumental variables:income and age. I selected these variablesbecause they are often included in CRM models (Verhoefet al. 2003). The estimationof this model results in the same parameterestimate for direct mailings (.04); however, this parameteris only marginally significant (p = .10). The second approachestimateda system of equationsin which two separateequations are estimated:one with customer share development as a dependentvariable and the other with the numberof direct mailings as a dependent variable. With this approach,the effect of direct mailings remained significant (p < .05); however, the parameter estimate decreasedfrom .04 to .013. On the basis of these analyses, I conclude thatendogeneity of directmailings does not affect the hypothesis testing.

AdditionalAnalysis
Effectof CommitmentS Mediating
In the relationship marketing literature, there has been a debate about the mediating role of commitment (Garbarino and Johnson 1999; Morgan and Hunt 1994). In this study, commitment may mediate the effect of payment equity and

satisfaction on customer share development, which in turn may explain the nonsignificanteffects of both satisfaction and paymentequity.To test for this mediatingeffect, I used Baron and Kenny's(1986) proposedmediationtest. I reestimated Model 2 (Column 3, Tables 4 and 5) in both the customer retentionand the customer share developmentappli6Areviewer this suggested analysis.

40/ Journal Marketing, of October 2003

estimates cations, but I left out commitment.The parameter for satisfaction and payment equity remain insignificantin both models (customerretention:a = -.10, p > .10; a = .03, p > .10; customer share development:[ = .01, p > .10; P = -.01, p > .10). In addition,I reestimatedboth models, leavestiing out satisfactionand paymentequity.The parameter mates for commitmentwere significantin both models (customer retention: a = .17, p < .05; customer share development:[ = .02, p < .01). Finally,I estimateda regression model in which I related satisfaction and payment equity to commitment.The parametersof both satisfaction and payment equity were positive and significant (y = .61, p < .01; y = .09, p <.05). These results show that satisfaction and payment equity should be considered antecedents of affective commitment;however,affective commitmentdoes not function as a mediatingvariable.

Effectof CRPsand RMIson CustomerRetention and CustomerShare Development


The first notable finding of this research is that affective commitmentis an antecedentof both customerretentionand customer share development.This result is not in line with recent findings that commitment does not influence customer retention (e.g., Gruen, Summers, and Acito 2000). However, it confirms previous claims in the relationship marketingliteraturethat commitment is a significant variable in customer relationships (Morgan and Hunt 1994; Sheth and Parvatiyar1995); more precisely, it affects both relationshipmaintenanceand relationshipdevelopment.At the same time, the absence of an effect of satisfaction and payment equity raises some notable issues. This result contradictspreviousfindings in the literature (e.g., Bowman and Narayandas2001; Szymanski and Henard 2001); several reasons may explain this. First, priorresearchhas typically relied on surveymeasuresfor which self-reporteddependent variables are correlatedas a result of common method of measures.This study uses behavioraldata based (partially) on internalcompany data. Second, unlike prior studies on customer share (e.g., Bowman and Narayandas2001; De Wulf, Odekerken-Schrider,and Iacobucci 2000) in which causalityis problematic,this study focuses on the change in customer share.An understanding customer share develof opment may require a deeper understandingof the role of CRPs and RMIs. Third, prior studies focus on customer share of a single brandin a single productcategory (Bowman and Narayandas2001), but this study focuses on customer share across multiple differentservices. Customer share changes occur over time when customers add (or drop) new (current)productsor services to (from) their portfolio of purchasedproducts or services at the focal supplieror at competing suppliers.In this underly-

Conclusions
Summaryof Findings
In this article, I contributedto the marketingliteratureby studying the effect of CRPs and RMIs on both customer retentionand customersharedevelopmentin a single study. The objectives of this article were twofold. First, I aimed to understand effect of CRPs and RMIs on customerretenthe tion and customer share development. Second, I examined whether different variables of CRPs and RMIs influence customer retentionand customer share development.Using a longitudinalresearchdesign, I relatedCRPs and RMIs to actual customerretentionand customer share development. An overview of the hypotheses, those that were supported and those that were not supported,is provided in Table 6. For the remainderof this discussion, I focus on the notable findings.

TABLE 6

Summary of Hypothesis-Testing Results


Customer Retention Antecedents Affective commitment Satisfaction Hypothesis (Sign) H1(+) H3(+) Effect
+

Customer Share Development Support Yes No Hypothesis (Sign) H2(+) No effect Effect
+

Support Yes Yes

0; positively moderatedby relationship age 0 N.A.

Payment equity Direct mailings

H4 (+) No effect

No N.A.

No effect H5(+) H6b(+)

0 + +

Yes Yes Yes

+ Yes Loyalty H6a(+) program Notes:N.A.= notavailable; effectcouldnotbe estimated this becauseof datalimitations.

CustomerRelationship Efforts/ 41 Management

ing decision process, satisfaction and payment equity play only a marginalrole for several reasons. First, satisfaction and payment equity are based on one's currentexperiences with the focal supplier.These experiences do not necessarily transferto other products or services of that supplier: New events may occur during the relationshipthat could change these perceptions (e.g., Mazursky and Geva 1989; Mittal, Kumar, and Tsiros 1999), thereby limiting the explanatorypower currentperceptions. Second, in a competitive environment,firms attempt to maximize customer share. Although customers may be satisfied with the focal firm's offering, they may be equally satisfied with competing offerings from other suppliers. This again limits the explanatorypower of satisfaction and payment equity. In contrast,affectivecommitmentseems less vulnerableto new experiences in the relationship;it is also unlikely that customers will considerthemselves committedto multiple suppliers. Instead of satisfaction and payment equity being considereddirectantecedentsof customerretentionandcustomer share development, they should be considered variables that shapecommitment(e.g., Morganand Hunt 1994). A second notablefinding is thatRMIs can influence customer retention and customer share development. Direct mailings with a "call to action"are suitableto enhance customer share over time. Loyalty programsthat provide economic rewards are useful both to lengthen customer relationships and to enhance customer share. Bolton, Kannan, and Bramlett(2000) reportthat loyalty programsfor credit card customers have a strong, positive effect on customer retention;however,no studies have yet consideredthe effect of loyalty programsand direct mailings on customer share development.The repeatedlyreportedpositive effect of the loyalty program counters the contention of Dowling and Uncles (1997, p. 75) that "it is difficult to increase brand loyalty above the market norms with an easy-to-replicate 'add on' customerloyalty program." The third relevant finding pertains to the explanatory power of both CRPs and RMIs. For both customerretention and customer share development, past customer behavior explains the largestpartof the variance(CRPs and RMIs are responsible only for approximately 10% of the total explained variancesin both the customer retentionand the customersharedevelopmentmodels). This finding seems to supportthe claims of skepticsof CRM thatthereis not much a firm can do to affect customerloyalty in consumermarkets (Dowling 2002). Duringreflection on the results of the customer share developmentmodel, it might also be perceived thatEhrenberg's (1997, p. 19) remarkson the antecedentsof marketsharealso hold for the antecedentsof customershare his development;in particular, claim "thatmost marketsare near stationaryand that everybody has to run hard to stand still" might also be applicable to customer share development. In the shortrun,my resultspoint to the effect of RMIs as only marginal.For example, stopping direct mailings for one year may not necessarily severely harmcustomershare development in that year. In a long-term perspective, the effects might be different. The effect of both CRPs and RMIs on customer purchase behavior could result in

increased relationshipage, increased customer shares, and purchasesof certainadditionalproductsor services (e.g., car insurance,life insurance).Some of these variablespositively affect customer retention and customer share development in later stages of the customerrelationship.

DifferencesBetween the Antecedents of CustomerRetentionand CustomerShare Development


Another research objective was to examine whether the antecedentsof customerretentionand customersharedevelopment are different.Theoretically,there is a clear distinction between relationship maintenance and relationship development;however, this has not been empiricallyinvesa tigated.Unfortunately, statisticalcomparisonof the coefficients in the customer retentionmodel and customer share developmentmodel is not possible (Fransesand Paap2001, Ch. 4). Thus, the only possible comparison is whether the significantpredictorsare different.The results show thatthe consistent significant variables(see Table6) are remarkably across the two models (i.e., affective commitmentand loyalty programs are significant predictors of both customer retentionand customersharedevelopment).The only exception is the interactioneffect between satisfaction and relationship age. However, with consideration of the effect of the past customer behavior control variables, there are some differences. For example, whereas high priorcustomer sharehas a positive effect on customer retention, it has a negative effect on customer share development. Likewise, relationship age has a positive effect on customer retentionbut no effect on customer share development. The latter results confirm that different variables affect customer retention and customer share development. However, from a CRM this difference is not as importantas it seems, .perspective, because the same CRM variablesaffect both customerretention and customersharedevelopment. Management Implications This research provides implications for effective management of customer relationships.First, if managersstrive to affect customer retention, they should focus on creating committed customers. In addition, a loyalty programwith economic incentives leads to greater customer retention. These results contrast with recent recommendationsthat creating close ties with customers is a better strategy for enhancing customer loyalty than using economically oriented programs(Braum 2002); firms should do both. Both affective commitmentand economically orientedRMI programs (direct mailings and loyalty programs)enhance customer retention and customer share development. Enhancing satisfactionand using attractivepricingpolicies can also increaseaffective commitment.OtherType II RMIs, such as affinity programsand other socially orientedprograms,may help as well (Rust, Zeithaml, and Lemon 2000). If firms strive for immediate results, economically based loyalty programsand direct mailings are preferable.

42/ Journal Marketing, of October 2003

Second, if firms strive to maximize customershare,creating affectively committed customers using a loyalty program and sending direct mailings that provide economic incentives are recommended.However,the short-term positive effects of such approachesare rathersmall. This might supportthe claim of expertsof CRM thattryingto maximize customerretentionand customersharedevelopmentis difficult. However,this does not mean that firms should not use such strategies.In the long run, the positive effects of such strategies may be larger. The short-term small positive effects of these strategies on customer retention and customer share development could result in larger positive effects in the long run as a result of the positive effects of past customerbehavioralvariables,such as relationshipage and priorcustomershare. Third, my analysis suggests that, in general, firms can use the same strategiesto affect customerretentionand customer share development.Fourth,a principle of CRM is to focus efforts on the most loyal customers.However,improving share for loyal customers is much more difficult, because they have a greatertendency to reduce their shares in the future. Research Limitations This study has the following limitations: First, it is conducted for one company in the financial services market.I chose the financial services marketbecause it is an important segment of the economy and because thereis a long tradition of customerdata storagein this market,which makes it relativelyeasy to collect behavioralcustomerloyalty data. However, the financial services market has some unique characteristics. Customers purchase insurance products infrequently,and as a result changes in customer share are not observedas frequentlyas in otherindustries.Because of relatively high switching costs, switching behavior is not common. These characteristicsmay have limited the variance in the customer share development measure. These characteristicsmay also explain some of the results and may, to some extent, threaten the generalizability of the results. Thus, there is a need to extend this study to other markets, especially markets in which more switching is observed. Second, although the study applied a longitudinal research design, the causality question remains difficult. Because of the dynamic nature of customer relationships, in multiple measurements time (includingchanges in CRPs) are needed in the model. Third, modeling the effect of RMIs is ratherdifficult, particularlyif the RMIs are self-selected or based on customers' purchasebehavior.In the loyalty programI studied, customerscan choose whetherto become a member.It could be arguedthat customers who expect to purchasenew services are more inclined to join. I chose not to correctfor this in the analysis at this time. Furtherresearchcould develop models to correctfor possible endogeneity of the RMIs.

The last researchlimitationpertainsto the measurement of payment equity. In this research, I used only two items the (see the Appendix), which could have undermined reliability of the measurement.Furtherresearch could develop more extensive scales.

FurtherResearch
Furtherresearchshould focus on the following issues: First, the results show that the effect of CRPs and RMIs on customer retention and customer share is not large. Perhaps other variables, such as service calls or sales visits, are important antecedents. In addition, competing marketing variables, such as competitive loyalty programsand direct mailings, have not been included here. Further research could investigate the effect of these variables. A second avenue for furtherresearchis the effect of RMIs on CRPs and in turn on customerbehavior.A simultaneousequation approach, with an appropriatetest for mediating effects, would be necessary to addressthis issue. In this respect, the interactionsbetween CRPs and RMIs could also be investigated. Finally, furtherresearchcould develop decision supporttype models (using dataavailablein customerdatabases and data from questionnaires)that would demonstratethe impact of variousCRM strategies.

Appendix of Description Scales for Perceptions


Commitment (Cronbach's Alpha [CA]= .77; Composite Reliability [CR] = .78) I am a loyal customerof XYZ. Because I feel a strong attachmentto XYZ, I remain a customer of XYZ. Because I feel a strongsense of belonging with XYZ, I want to remaina customerof XYZ. Satisfaction (CA = .83; CR = .83) How satisfied (1 = "very dissatisfied"and 5 = "very satisfied") are you about *thepersonal attention XYZ. of *thewillingness XYZto explain of procedures. of *theservicequality XYZ. *theresponding XYZto claims. by *theexpertise thepersonnel XYZ. of of with *your relationship XYZ. of *thealertness XYZ. Payment Equity (r = .49; CR = .88) How satisfied (1 = "very dissatisfied"and 5 = "very satisfied") are you about the insurancepremium? Do you think the insurancepremium of your insurance is too high, high, normal,low, or too low?

CustomerRelationship Efforts/ 43 Management

REFERENCES
Anderson, Erin and BartonWeitz (1992), "The Use of Pledges to Build and Sustain Commitment in Distribution Channels," Journal of MarketingResearch, 29 (February),18-34. Anderson, Eugene W., Claes Fornell, and Donald R. Lehmann (1994), "CustomerSatisfaction, Market Share, and Profitability: Findings from Sweden,"Journal of Marketing,58 (July), 53-66. and Vikas Mittal (2000), "Strengthening Satisfactionthe Profit Chain,"Journal of Service Research, 3 (2), 107-120. and MaryW. Sullivan (1993), "TheAntecedentsand Consequences of CustomerSatisfactionfor Firms,"MarketingScience, 12 (Spring), 125-43. Anderson, James C. and David W. Gerbing (1988), "Structural Equation Modeling Practice: A Review and Recommended Two-StepApproach," Psychological Bulletin, 103 (3), 411-23. Bagozzi, RichardP. and YoujaeYi (1988), "On the Evaluationof Structural EquationModels,"Journal of the Academy of MarketingScience, 16 (Winter),74-94. Baron, Reuben M. and David A. Kenny (1986), "The ModeratorMediator Variable Distinction in Social Psychological Research,"Journal of Personality and Social Psychology, 51 (6), 1173-82. Hans and ChristianHomburg(1996), "Applications Baumgartner, of StructuralEquationModeling in Marketingand Consumer Research,"InternationalJournal of Research in Marketing,13 (2), 139-61. Bawa, Kapil and RobertW. Shoemaker(1987), "The Effects of a Direct Mail Coupon on Brand Choice Behavior,"Journal of MarketingResearch, 24 (November),370-76. Berry, Leonard L. (1995), "RelationshipMarketingof Services: Journalof theAcadGrowingInterest,EmergingPerspectives," emy of MarketingScience, 23 (Fall), 236-45. C.B. andRuthN. Bolton (2000), "Relationship MarBhattacharya, in keting in Mass Markets," Handbookof RelationshipMarketeds. ing, JagdishN. Sheth andAtul Parvatiyar, ThousandOaks, CA: Sage Publications,327-54. , HayagreevaRao, and Mary Ann Glynn (1995), "UnderAn standingthe Bond of Identification: Investigationof Its Correlates Among Art Museum Members,"Journal of Marketing, 59 (October),46-57. Bickart, BarbaraA. (1993), "Carryoverand Backfire Effects in Journal of MarketingResearch,30 (FebMarketingResearch," ruary),52-62. Blattberg,RobertC., Gary Getz, and JacquelynS. Thomas (2001), Customer Equity: Building and Managing Relationships as Valuable Assets. Boston: HarvardBusiness School Press. Bolton, RuthN. (1998), "A Dynamic Model of the Durationof the Customer'sRelationship with a Continuous Service Provider: The Role of Satisfaction,"Marketing Science, 17 (Winter), 45-65. -, P.K. Kannan,and Mathew D. Bramlett(2000), "Implications of Loyalty ProgramMembershipand Service Experiences for CustomerRetentionand Value,"Journal of the Academyof MarketingScience, 28 (Winter),95-108. - and KatherineN. Lemon (1999), "A Dynamic Model of Customers' Usage of Services: Usage as an Antecedent and Consequence of Satisfaction,"Journal of MarketingResearch, 36 (May), 171-86. , - , and Peter C. Verhoef (2002), "The Theoretical Underpinningsof CustomerAsset Management:A Framework and Propositionsfor FutureResearch," ErasmusResearchInstitute in ManagementWorking Paper No. ERS-2002-80-MKT, ErasmusUniversity,Rotterdam. Bowman, Douglas and Das Narayandas (2001), "Managing Customer-Initiated Contacts with Manufacturers: The Impact on Share of Category Requirements and Word-of-Mouth Behavior," Journal of Marketing Research, 38 (August), 281-97. Braum,Lutz (2002), "InvolvedCustomersLead to Loyalty Gains," MarketingNews, 36 (2), 16. Capraro,Anthony J., Susan Broniarczyk,and RajendraK. Srivastava (2003), "FactorsInfluencing the Likelihood of Customer Defection: The Role of ConsumerKnowledge,"Journal of the Academyof MarketingScience, 31 (Spring), 164-75. Christy,Richard,GordonOliver, and Joe Penn (1996), "Relationship Marketingin Consumer Markets,"Journal of Marketing Management,12 (1), 175-88. Crosby, Lawrence A., Kenneth R. Evans, and Deborah Cowles (1990), "RelationshipQuality in Services Selling: An Interpersonal Influence Perspective," Journal of Marketing,54 (July), 68-81. Davidson, Russell and James G. MacKinnon (1989), "Testingfor Consistency Using Artificial Regressions,"Econometric Theory, 5 (3), 363-84. De Wulf, Kristof,Gaby Odekerken-Schrider, Dawn Iacobucci and (2001), "Investmentsin Consumer Relationships: A CrossCountry and Cross-Industry Exploration,"Journal of Marketing, 65 (October),33-50. Dick, Alan S. and KunalBasu (1994), "Customer Loyalty:Toward an Integrated Journal of the Academy ConceptualFramework," of MarketingScience, 22 (Spring), 99-113. Dowling, GrahameW. (2002), "CustomerRelationship Management: In B2C Markets,Often Less Is More," CaliforniaManagementReview,44 (Spring), 87-104. and MarkUncles (1997), "Do CustomerLoyalty Programs Really Work?"Sloan ManagementReview, 38 (Fall), 71-82. Dwyer, Robert F., Paul H. Schurr,and Sejo Oh (1987), "Developing Buyer-Seller Relationships,"Journal of Marketing, 51 (April), 11-27. JourEhrenberg,Andrew (1997), "Descriptionand Prescription," nal of AdvertisingResearch, 37 (6), 17-22. Franses,Philip Hans and RichardPaap (2001), QuantitativeModels in MarketingResearch. Cambridge,UK: CambridgeUniversity Press. Garbarino,Ellen and Mark S. Johnson (1999), "The Different Roles of Satisfaction, Trust, and Commitment in Customer Journal ofMarketing, 63 (April), 70-87. Relationships," Geyskens, Inge, Jan-Benedict E.M. Steenkamp, and Nirmalya Kumar (1998), "GeneralizationsAbout Trust in Marketing Channel Relationships Using Meta-Analysis," International Journal of Research in Marketing,15 (3), 223-48. Gruen, Thomas W., John O. Summers, and Frank Acito (2000), "RelationshipMarketingActivities, Commitment, and Membership Behaviors in Professional Associations," Journal of Marketing,64 (July), 34-49. Gundlach,GregoryT., Ravi Achrol, and John T. Mentzer (1995), "The Structureof Commitmentin Exchange,"Journal of Mar78-92. keting, 59 (January), Hart, Susan, Andrew Smith, Leigh Sparks, and Nikos Tzokas (1999), "AreLoyalty Schemes a Manifestationof Relationship Marketing?" Journal of Marketing Management, 15 (6), 541-62. Heckman,James J. (1976), "The Common Structureof Statistical Models of Truncation,Sample Selection, and Limited Dependent Variables and a Simple Estimator for Such Models," Annals of Economic and Social Measurement,5, 475-92. Hoekstra, Janny C., Peter S.H. Leeflang, and Dick R. Wittink (1999), "The CustomerConcept:The Basis for a New MarketJournal ofMarket-FocusedManagement,4 (1), ing Paradigm," 43-76. Homburg,Christianand Annette Giering (2001), "PersonalCharacteristics as Moderatorsof the Relationship Between Cus-

44 / Journalof Marketing, October2003

tomer Satisfactionand Loyalty-An EmpiricalAnalysis,"Psychology and Marketing,18 (1), 43-66. Jones, Michael A., David L. Mothersbaugh,and SharonE. Beatty (2001), "SwitchingBarriersand RepurchaseIntentionsin Services," Journal of Retailing, 76 (2), 259-74. Jones, Thomas O. and W. Earl Sasser Jr. (1995), "Why Satisfied Customers Defect," Harvard Business Review, 73 (November-December),88-100. Jdreskig, Karl and Dag Sorbom (1993), LISREL83 Structural Equation Modeling with the SIMPLIS Command Language. Chicago: Scientific SoftwareInternational. Kamakura,WagnerA., SridharN. Ramaswami, and RajendraK. Srivastava(1991), "ApplyingLatentTraitAnalysis in the Evaluation of Prospects for Cross Selling of Financial Services," InternationalJournal of Research in Marketing,8 (4), 329-49. Jim Kerstetter, (2001), "SoftwareHighfliers,"BusinessWeek, (June 18), 108-109. Klemperer, Paul (1995), "CompetitionWhen Consumers Have Switching Costs: An Overview with Applications to Industrial Trade,"The Organization,Macro Economics, and International Review of Economic Studies, 62 (4), 515-39. Kumar, Nirmalya, Lisa K. Scheer, and Jan-Benedict E.M. Steenkamp (1995), "The Effects of Supplier Fairness on Vulnerable Resellers,"Journal of MarketingResearch, 32 (February), 54-65. Latham, Garry P. and Edwin A. Locke (1991), "Self Regulation Through Goal Setting," OrganizationalBehavior and Human Decision Processes, 50 (2), 212-47. Lemon, KatherineN., TiffanyBarnettWhite, and Russell S. Winer (2002), "DynamicCustomerRelationshipManagement:Incorinto the Service RetentionDeciporatingFutureConsiderations sion,"Journal of Marketing,66 (January),1-14. Lichtenstein,Donald R., Nancy M. Ridgway,and RichardG. Netemeyer (1993), "Price Perceptions and Consumer Shopping Behavior: A Field Study,"Journal of MarketingResearch, 30 (May), 234-45. Macintosh, Gerrardand Lawrence S. Lockshin (1997), "Retail Relationships and Store Loyalty: A Multi-Level Perspective," International Journal of Research in Marketing, 14 (5), 487-97. MacKenzie,Scott B., Philip M. Podsakoff,and Michael J. Ahearne (1998), "Some Possible Antecedents and Consequences of InRole and Extra-Role Salesperson Performance,"Journal of Marketing,62 (July), 87-98. Mazursky,David andAviva Geva (1989), "Temporal Decay in Satisfaction-Purchase Intention Relationship," Psychology and Marketing,6 (3), 211-27. Mittal, Vikas and Wagner A. Kamakura (2001), "Satisfaction, RepurchaseIntent,and RepurchaseBehavior:Investigatingthe Journal of Moderating Effect of Customer Characteristics," MarketingResearch, 38 (February),131-42. , Pankaj Kumar, and Michael Tsiros (1999), "AttributeLevel Performance, Satisfaction, and Behavioral Intentions over Time:A Consumption-System Journalof MarApproach," keting, 63 (April), 88-101. Moorman, Christine, Gerald Zaltman, and Rohit Deshpand6 (1992), "RelationshipsBetween Providers and Users of Marketing Research:The Dynamics of TrustWithin and Between Journal of MarketingResearch, 29 (August), Organizations," 314-28. Morgan, Robert M. and Shelby D. Hunt (1994), "The Commitment-TrustTheory of Relationship Marketing," Journal of Marketing,58 (July), 20-38. Nicholls, John G. (1989), The CompetitiveEthos and Democratic Education.Cambridge,MA: HarvardUniversityPress. Oliver, Rick and Russel S. Winer (1987), "A Frameworkfor the Formation and Structureof Consumer Expectations: Review

and Propositions,"Journal of Economic Psychology, 8 (4), 469-99. Peppers,Don and MarthaRogers (1999), EnterpriseOne-to-One: Tools for Competing in the Interactive Age. New York: Doubleday. Pindyck, Robert S. and Daniel L. Rubinfeld (1998), Econometric Models and Economic Forecasts.Boston: Irwin/McGraw-Hill. Reichheld, FrederickF. (1996), The Loyalty Effect. Boston: HarvardBusiness School Press. Reinartz,WernerJ. andV. Kumar(2002), "TheMismanagementof Customer Loyalty," Harvard Business Review, 80 (July), 86-94. and (2003), "TheImpactof CustomerRelationship Characteristicson Profitable Lifetime Duration,"Journal of Marketing,67 (January),77-99. Roberts, Mary Lou and Paul D. Berger (1999), Direct Marketing Management.Englewood Cliffs, NJ: PrenticeHall. Roehm, Michelle L., Ellen Bolman Pullins, and HarperA. Roehm Jr. (2002), "Designing Loyalty-Building Programs for Packaged Goods Brands," Journal of Marketing Research, 39 (May), 202-213. Rust, Roland T., Valarie A. Zeithaml, and KatherineN. Lemon (2000), Driving Customer Equity: How Customer Lifetime Value Is Reshaping Corporate Strategy. New York: The Free Press. Sharp, Byron and Anne Sharp (1997), "Loyalty Programs and Their Impact on Repeat-PurchaseLoyalty Patterns,"International Journal of Research in Marketing,14 (5), 473-86. Sheth, JagdishN. and Atul Parvatiyar(1995), "RelationshipMarketing in ConsumerMarkets:Antecedentsand Consequences," Journal of the Academy of Marketing Science, 23 (Fall), 255-71. Singh, Jagdip (1990), "Voice, Exit, and Negative Word-of-Mouth Behaviors:An InvestigationAcross Three Service Categories," Journal of the Academy of Marketing Science, 18 (Winter), 1-15. Steenkamp,Jan-BenedictE.M. and Hans C.M. van Trijp (1991), "The Use of LISREL in Validating Marketing Constructs," InternationalJournal of Research in Marketing,8 (4), 283-99. Szymanski, David M. and David H. Henard (2001), "Customer Satisfaction: A Meta-Analysis of the Empirical Evidence," Journal of the Academy of Marketing Science, 29 (Winter), 16-35. Varuki,Sajeev and Mark Colgate (2001), "The Role of Price Perceptions in an Integrated Model of Behavioral Intentions," Journal of Service Research,3 (3), 232-40. Verhoef, Peter C., Philip Hans Franses, and Janny C. Hoekstra (2001), "The Impact of Satisfaction and Payment Equity on Cross Buying:A Dynamic Model for a Multi-ServiceProvider," Journal ofRetailing, 77 (3), 359-78. , Penny N. Spring, Janny C. Hoekstra, and Peter S.H. Leeflang (2003), "The Commercial Use of Segmentation and PredictiveModeling Techniquesfor DatabaseMarketingin the Decision SupportSystems,34 (4), 471-81. Netherlands," CovariWhite, Halbert (1980), "A Heteroskedasticity-Consistent ance Matrix and a Direct Test for Heteroskedasticity," Econometrica, 48 (4), 817-38. Winer, Russell S. (2001), "A Frameworkfor CustomerRelationCaliforniaManagementReview, 43 (Sumship Management," mer), 89-108. Woodruff,Robert B. (1997), "CustomerValue: The Next Source for CompetitiveAdvantage,"Journal of the Academy of Marketing Science, 25 (Spring), 139-53. Zeithaml, Valarie A., Leonard L. Berry, and A. Parasuraman (1996), "The Behavioral Consequences of Service Quality," Journal of Marketing,60 (April), 31-46.

CustomerRelationship Efforts/ 45 Management

You might also like