Decoding Customer-Firm Relationships - Mende Et Al

MARTIN MENDE, RUTH N.
BOLTON, and MARY JO BITNER*
Many firms strive to create relationships with customers, but not all
customers are motivated to build close commercial relationships. This
article introduces a theoretical framework that explains how relationship-
specific attachment styles account for customers’ distinct preferences for
closeness and how both attachment styles and preferences for
closeness influence loyalty. The authors test their predictions with survey
data from 1199 insurance customers and three years of purchase
records for 975 of these customers. They find that attachment styles
predict customers’ preferences for closeness better than established
marketing variables do. Moreover, attachment styles and preferences for
closeness influence loyalty intentions and behavior, controlling for
established antecedents (e.g., relationship quality). Finally, exploring the
underlying process, the authors show that preference for closeness
partially mediates the effect of attachment styles on cross-buying
behavior. This research provides novel customer segmentation criteria
and actionable guidelines that managers can use to improve their ability
to tailor relationship marketing activities and more effectively allocate
resources to match customer preferences.
Keywords: attachment styles, services, relationship marketing, loyalty,

relational preferences
Decoding Customer–Firm Relationships:

How Attachment Styles Help Explain
Customers’ Preferences for Closeness,
Repurchase Intentions, and Changes in
Relationship Breadth
The field of marketing “is supposed to be concerned

*Martin Mende is Assistant Professor of Marketing, Gatton College of
Business and Economics, University of Kentucky (e-mail: martin.mende@
about the connection of the firm to its customers” (Reib-
uky.edu). Ruth N. Bolton is Professor of Marketing (e-mail: ruth.bolton@ stein, Day, and Wind 2009, p. 1). Many firms invest heavily
asu.edu), and Mary Jo Bitner is Professor of Marketing and PetSmart Chair in relationship marketing (RM) to create, sustain, and
in Services Leadership (e-mail: [email protected]), W.P. Carey
School of Business, Arizona State University. This research was supported
enhance close relationships with their customers, assuming
by a grant from the Marketing Science Institute and from the Center for such investments lead to positive financial outcomes. How-
Services Leadership at Arizona State University. The authors gratefully ever, some customers are indifferent or averse to a firm’s
acknowledge this generous support. The authors are grateful to the two
anonymous JMR reviewers for helpful and constructive feedback that has
relationship-building efforts (Godfrey, Seiders, and Voss
significantly improved the article. Finally, the authors thank Maura Scott, 2011). Firms should allocate their resources to customers
Kathleen Seiders, Kay Lemon, Linda Salisbury, and Yoko Kusunose for who are likely to be receptive to RM (Palmatier 2008) and
helpful comments on earlier versions of this article. Jeffrey Inman served profitable for the firm. This strategy is difficult to execute
as associate editor for this article.
because little is known about how customers differ in their
© 2013, American Marketing Association Journal of Marketing Research

ISSN: 0022-2437 (print), 1547-7193 (electronic) 125 Vol. L (February 2013), 125–142
126 JOURNAL OF MARKETING RESEARCH, FEBRUARY 2013
relational preferences (Palmatier et al. 2006). Thus, Keller aforementioned loyalty antecedents established in market-
and Lehmann (2006, p. 742) urge researchers to examine ing—predict changes in customers’ relationship breadth
how “a customer’s desired relationship [can] be deter- over a three-year period. Exploring the underlying process,
mined,” whether “customers still desire close relationships we also find that the effects of customer attachment styles
with companies,” and how “a desired customer relationship on cross-buying behavior are partially mediated by prefer-
[can] be cultivated by the company through marketing ence for closeness. Finally, this research provides firms with
activities.” a novel and important enhancement to existing market seg-
This article addresses these questions by developing a mentation criteria. The two conceptual components of cus-
conceptual framework based on relationship-specific customer attachment styles and preference for closeness enrich
tomer attachment styles. The framework builds on attach- customer segmentation and portfolio management, improve
ment theory, a comprehensive paradigm for explaining how firms’ ability to tailor marketing activities to customers’
and why people (dis)engage in close relationships (Mikulin- relational profiles, and help allocate marketing resources
cer and Shaver 2007). Attachment theory is a major founda- more effectively.
tion for research in psychology that studies interpersonal In the remainder of this article, we present our framework
relationships (Hazan and Shaver 1994). In this study, we of customer attachment styles and review research relevant
build on attachment theory to reveal the underlying theoreti- to understanding customers’ relational preferences. Then,
cal mechanisms that explain customers’ relationship prefer- we hypothesize how customers’ attachment styles influence
ences and loyalty in their relationships with firms. Specifi- their preferences for closeness and how both constructs
cally, we develop and estimate a model that addresses the influence loyalty to the firm. We test our hypotheses by esti-
following questions: mating equations that describe customers’ preferences for
closeness, repurchase intentions, and objective changes in
1. How can understanding relationship-specific customer their relationship breadth over time. We estimate the equa-
attachment styles help marketers (a) predict a customer’s
tions with cross-sectional survey data from 1199 insurance
preference for closeness to the firm and (b) customize
relationship-building efforts to match relational preferences?
customers and associated purchase records for 975 of these
2. Do customers’ attachment styles and preferences for close- customers over a three-year period. Last, we present our
ness influence loyalty to the firm (i.e., repurchase intentions results, discuss their implications for marketing theory and
and changes in relationship breadth), beyond established loy- practice, and identify future research opportunities.
alty drivers, such as satisfaction?
CUSTOMER ATTACHMENT STYLES
By answering these questions, we make five contributions to An attachment style is the systematic pattern of relational
the marketing literature. First, customers’ relational prefer- expectations, needs, emotions, and social behaviors that
ences must be understood and measured for firms to manage results from the internalization of a particular history of
relationships effectively. We define and measure customers’ attachment experiences (Mikulincer and Shaver 2007).
preferences for closeness as their systematic preference for Research in psychology has shown that attachment styles
frequent, diverse, and mutually influential interactions with are best conceptualized and measured along two continu-
a firm. Second, we show how relationship-specific attach- ous, quasi-orthogonal dimensions called “attachment anxi-
ment styles explain customers’ preferences for closeness ety” and “attachment avoidance” (Brennan, Clark, and
and their response to closeness-enhancing RM activities Shaver 1998). Attachment anxiety is the extent to which a
(e.g., invitation to join a firm’s customer recommendation person worries that relationship partners might not be avail-
program, regular face-to-face consultations with employ- able in times of need, has an excessive need for approval,
ees). These insights help answer Keller and Lehmann’s and fears rejection and abandonment. Attachment avoidance
(2006) questions about whether customers desire close rela- is the extent to which a person has an excessive need for
tionships and how firms can cultivate closeness with mar- self-reliance, fears depending on others, distrusts relation-
keting activities. ship partners’ goodwill, and strives for emotional and cog-
Third, this study shows how customers’ attachment styles nitive distance from partners. Together, these two dimen-
and their preferences for closeness influence loyalty inten- sions account for consistent and profound differences in the
tions. Our investigation of insurance customers demon- nature of close relationships (Simpson 1990).
strates that customers’ attachment styles and preferences for Attachment theory is important in psychology, but only
closeness influence repurchase intentions after controlling four marketing studies have examined attachment styles.
for prominent antecedents of repurchase intentions (e.g., Swaminathan, Stilley, and Ahluwalia (2009) conduct a lab-
relationship quality, price fairness, switching barriers, oratory experiment in which they primed (rather than meas-
tenure) (Palmatier et al. 2006; Verhoef 2003). ured) interpersonal attachment styles; that is, they asked stu-
Fourth, because it is well established that loyalty inten- dents to think and write about a close interpersonal
tions and loyalty behavior are not always aligned (Oliver relationship that was characterized by a particular attach-
1999; Seiders et al. 2005), we examine the impact of cus- ment style (high/low anxiety and high/low avoidance). The
tomers’ attachment styles and their preferences for closeness authors show that priming interpersonal attachment styles
on loyalty behavior over time—specifically, on objectively moderates the influence of brand personality on brand out-
measured changes in relationship breadth (i.e., the varying comes, such as purchase likelihood. However, attachment-
number of different types of insurances a customer has, related priming effects “are likely to be short-lived and
such as car, life, or home insurance) (Bolton, Lemon, and unstable” (Mikulincer and Shaver 2007, p. 67). Our study
Verhoef 2004). The analyses reveal that customer attach- measures (rather than primes) attachment styles; further-
ment styles and preference for closeness—but none of the more, rather than focusing on interpersonal, noncommercial
Decoding Customer–Firm Relationships 127
attachment targets, we consider customers’ attachment styles, but it did not investigate customer preferences for
styles toward a focal firm (i.e., within the context of an closeness or loyalty.
actual commercial relationship). In summary, our work extends prior research in marketing
Second, Thomson and Johnson (2006) conduct a cross- that draws on attachment theory. It investigates relationship-
sectional survey of 118 undergraduate students to study specific, firm-focused customer attachment styles, rather
general attachment styles—that is, a person’s attachment than general or interpersonal attachment styles. Thus, we
anxiety and avoidance toward other people in general. They use the following definitions (Brennan, Clark, and Shaver
elicited responses to items such as “I want to get close to 1998):
people, but I worry about being hurt” (anxiety) and “people
are never there when you need them” (avoidance) (empha- •Customer attachment anxiety is the extent to which a customer
sis added). Then, students rated a commercial relationship worries that the firm might not be available in times of need,
of their choice (i.e., with any firm or brand). Students’ gen- has an excessive need for approval, and fears rejection and
eral attachment anxiety and avoidance negatively affected abandonment from this firm.
their perceptions of reciprocity (e.g., “I give to this firm/ •Customer attachment avoidance is the extent to which a cus-
brand about as much as it gives to me”), and avoidance ele- tomer distrusts the firm’s goodwill, is characterized by an
excessive need for self-reliance, fears depending on the firm,
vated the students’ focus on financial motives in market-
and strives for emotional and cognitive distance from the firm.
place relationships. That study did not find any direct
effects of general attachment styles on customer involve- Our research extends the existing literature by investigating
ment or commitment but indicates an indirect effect on sat- the influence of customer attachment styles on three
isfaction (through reciprocity). Extending these insights, we dependent variables in commercial relationships: cus-
reveal direct effects and an indirect effect of customer tomers’ preferences for closeness, repurchase intentions,
attachment styles on relationship and loyalty measures from and their loyalty behavior over time.
actual customers of a focal firm.
Third, Paulssen (2009) surveys business-to-business cus- CLOSENESS: OVERLOOKED BUT RELEVANT FOR RM
tomers to study how general business attachment styles Prior research has identified customers with distinct rela-
influence commercial relationships. This cross-sectional tional orientations (De Wulf, Odekerken-Schröder, and
study reveals that attachment avoidance might influence
Iacobucci 2001), but this work tends to be descriptive and
satisfaction, trust, and repurchase intent. Note that the sur-
context-specific rather than based on a comprehensive
vey instructed respondents to think about all their “business
relationships in general” but not to think about a specific theory. In general, the theoretical mechanisms that influence
business partner; in contrast, we study customer relationships which customers seek or avoid closeness have been neg-
with a focal firm. Notably, Paulssen also does not follow the lected. This section presents research relevant to the under-
dominant conceptualization of attachment styles of attach- standing of customer preference for closeness.
ment anxiety and avoidance (Mikulincer and Shaver 2007). Conceptualizing Customer Preference for Closeness
His measures reflect aspects of avoidance (e.g., willingness
to depend) but do not capture attachment anxiety (e.g., worry Psychological research has established interdependence
about rejection). In contrast, we examine customer attach- theory as a major framework for examining closeness. Inter-
ment anxiety and avoidance and show that both dimensions dependence theory considers relational closeness a function
have distinctive effects in commercial relationships. of the partners’ interactional patterns (Kelley et al. 1983).
Fourth, Mende and Bolton (2011) find that their conceptu- Berscheid, Snyder, and Omoto (1989) build on this theory
alization and measurement of customer attachment styles is to develop the prominent Relationship Closeness Inventory
consistent with recent findings in psychology. It is now well (RCI), a 60-plus-item self-report measure. They conceptu-
established that people develop multiple attachment styles alize closeness as mirrored in (1) a high frequency of inter-
that are organized hierarchically, from general to relationship- actions between partners, (2) diverse forms of interaction
specific attachment styles. Attachment styles can emerge as with each other, and (3) a reasonably strong influence on
relationship-specific constructs that may or may not be con- each other. This characterization of closeness applies to all
gruent with the person’s general or higher-level attachment relationships; it is not restricted to close interpersonal part-
styles (Klohnen et al. 2005). Relationship-specific attach- nerships (Berscheid, Snyder, and Omoto 2004). Thus, we
ment styles are better predictors of relationship outcomes believe that closeness between a customer and a firm is
than general attachment styles; thus, “researchers trying to
indicated by analogous interactional patterns. Some
predict outcomes within a specific relationship domain
research is consistent with the notion that the frequency
should measure individuals’ corresponding attachment
models to maximize their predictive ability and validity” with which a firm interacts with its customers, as well as the
(Klohnen et al. 2005, p. 1678). Mende and Bolton (2011) are breadth and nature of firm-initiated communication, influ-
the first researchers in marketing to investigate relationship- ences closeness (Crosby, Evans, and Cowles 1990; Mohr,
specific, firm-focused attachment styles. Their work pro- Fisher, and Nevin 1996). The bidirectionality of interactions
vides empirical evidence that customers with low levels of and a noncoercive influence on each other have also been
attachment anxiety, avoidance, or both perceive a service identified as important components of close relationships in
firm and service employee more positively in terms of satis- marketing research (Mohr, Fisher, and Nevin 1996). Thus,
faction, trust, and affective commitment than customers we define preference for closeness as a customer’s system-
with high levels. However, their study focused on the con- atic preference for frequent, diverse, and mutually influen-
ceptualization and measurement of customer attachment tial RM-related interactions with a firm.
Revealing Antecedents of Customer Preference for attachment strategy derived from a comprehensive denial or
Closeness Through Attachment Theory dismissal of their attachment needs, which leads to the inhi-
Marketers’ lack of attention to customers’ preferences for bition of closeness enhancements. Consequently, avoidant
closeness is detrimental because RM posits that (some) cus- people tend to respond to closeness-triggering behavior
tomers are receptive to marketing efforts that build customer– with “a repertoire of defensive strategies” aimed at distanc-
firm bonds. Moreover, RM is guided by the principle of tailor- ing themselves from others (Edelstein and Shaver 2004, p.
ing marketing activities to the individual customer (Palmatier 409). In summary, we expect the two attachment dimen-
2008). Thus, marketers’ inability to identify customers who sions to affect a customer’s preference for closeness in
are more (or less) receptive to the formation of close bonds opposing ways:
conflicts with RM’s central mission and hurts its efficiency H1a: Customer attachment anxiety is positively related to the
and effectiveness. Palmatier (2008, p. 54) writes that man- preference for closeness to the firm.
agers face a quandary: “They know that building strong H1b: Customer attachment avoidance is negatively related to
relationships is important, but they have little guidance on the preference for closeness to the firm.
how to build and maintain strong relationships or target and Although empirical studies in psychology rarely address
adapt their [RM] strategies on the basis of customer … fac- the interaction between anxiety and avoidance (derived from
tors.” We believe that much can be learned from attachment continuous anxiety and avoidance variables), theoretical
theory, which is one of the most viable and comprehensive reasoning suggests that some customers have high levels of
theories of close relationships in psychology (Mikulincer anxiety and avoidance—a so-called fearful attachment style
and Shaver 2007). Thus, we draw on the concept of cus- (Brennan, Clark, and Shaver 1998). A fearful attachment
tomer attachment styles to develop a model that explains the style results from people’s failure to reach their attachment
psychological mechanisms that influence why and how cus- goals through either a hyperactivation of their attachment
tomers react differently to closeness-triggering RM activi- system (anxiety) or a defensive deactivation of their attach-
ties and that reveals how these mechanisms relate to cus- ment system (avoidance). Fearfully attached people fluctu-
tomer loyalty. ate and show characteristics of both attachment dimensions;
HYPOTHESES they desire closeness but fear the potentially negative con-
sequences of closeness and reliance on others, though they
We hypothesize that customers’ attachment styles (i.e., “wish they did not have to feel this way” (Mikulincer and
varying levels of attachment anxiety and avoidance) affect Shaver 2007, p. 42). Collins and Feeney (2004) argue that
their preferences for a close relationship with the firm (H1). this dilemma is typically resolved in favor of avoidance
Furthermore, we predict that both preference for closeness because fearful people’s cognitions about the risks of close-
and customer attachment styles influence loyalty (H2 and ness override their emotional desire for closeness.
H3). These insights suggest that fearful customers value close-
Customer Attachment Styles and Preference for Closeness ness but tend to avoid situations in which they feel vulnera-
ble to rejection and ultimately remain withdrawn and dis-
Swaminathan, Stilley, and Ahluwalia (2009) show that tant from the firm. For preference for closeness, we expect
consumers’ interpersonal attachment styles influence their the positive effect of attachment anxiety to interact with the
bonding with brand personalities. Specifically, consumers negative effect of avoidance, such that this negative asso-
with high interpersonal attachment anxiety prefer brands ciation can outweigh the anxiety-driven desire for close-
with a sincere (vs. an exciting) personality, whereas con- ness; that is, the net effect of these two forces can become
sumers with high interpersonal avoidance prefer an exciting negative. This interplay does not allow a prediction on the
(vs. a sincere) brand personality. Expanding the idea that direction of the net effect.
consumers with different attachment styles are not equally
sensitive to bonding with a brand, we propose that attach- H1c: Customer attachment anxiety and customer attachment
ment theory is useful to understand closeness in commer- avoidance interact to influence the preference for closeness
cial relationships because it emphasizes the distinct ways to the firm, such that the positive relationship between
anxiety and preference for closeness becomes less positive
attachment styles influence the regulation of (physical and
(and potentially negative) as avoidance increases.
psychological) closeness in interpersonal relationships
(Collins and Feeney 2004). Although both attachment anxi-
ety and avoidance are tied to a person’s preference for Customer Preference for Closeness and Loyalty to the Firm
closeness, they are two distinct mechanisms that function as Interdependence theory posits that close relationships are
quasi-opposing effects. Anxiously attached people desire a those in which both parties have a frequent and strong
high level of relational intimacy, display greater care seek- impact on each other in diverse types of activities across
ing, and require an elevated level of responsiveness and time. We believe that a customer’s preference for repeated
emotional rapport in a partnership (Hazan and Shaver 1994). customer–firm interactions over time resembles a form of
They ([un]consciously) adopt a so-called hyperactivated social bonding that helps maintain relationship commit-
attachment strategy that constantly focuses on the search for ment. This rationale is consistent with marketing research
being appreciated and valued by a partner, while vigilantly on the frequency of customer–firm interactions. For exam-
screening for potential threats to the relationship (Mikulin- ple, Dagger, Danaher, and Gibbs (2009) focus on relation-
cer and Shaver 2007). Avoidantly attached people prefer to ship strength (rather than loyalty) as the outcome variable
maximize cognitive, emotional, and physical distance from and show that customers who interact more frequently with
others. They ([un]consciously) adopt a so-called deactivated a service firm report stronger relationships. While the posi-
tive effect of frequency of customer–firm interactions on Research on the distinct associations of both attachment
loyalty has received some attention, little marketing dimensions with relational commitment suggests that the
research has studied how the diversity of such interactions respective effect sizes of customer attachment anxiety and
or how the mutual influence between customer and firm avoidance on loyalty will differ systematically. Insights into
affects loyalty. However, our rationale that preference for personal relationships show that attachment anxiety
closeness is positively associated with loyalty is consistent decreases relational satisfaction but, paradoxically, might
with psychological research on how the three facets of also decrease the likelihood that a person will leave an
closeness predict relationship stability. Specifically, there is unhappy partnership (Davila and Bradbury 2001). The theo-
a positive association between the frequency of contact retical reasoning is that their concerns about self-worth and
between partners and relationship stability (Cate, Levin, and abandonment, their dependent manner of bonding, and their
Richmond 2002). Moreover, partners who jointly engage in focus on proximity seeking and on gaining approval from
diverse activities and have higher levels of mutual impact the partner motivate anxious people to make special efforts
(e.g., on their plans and goals) have lower levels of breakups to maintain the relationship status quo, “even though the
(Cate, Levin, and Richmond 2002). Finally, a person’s (self- relationship was not what one had hoped for” (Kirkpatrick
reported) subjective closeness to the partner helps explain and Davis 1994, p. 504). A similar tendency to maintain the
relationship stability (Berscheid, Snyder, and Omoto 1989; status quo is not to be expected from avoidant people,
Simpson 1990). Thus, we hypothesize the following: because attachment avoidance is not associated with any of
the previously mentioned characteristics (e.g., concern
H2: Customer preference for closeness is positively related to about abandonment) (Davila and Bradbury 2001). Rather,
loyalty to the firm.
avoidant people are typically less invested in relationships,
Customer Attachment Styles and Loyalty to the Firm are less upset when they end, and report relatively low lev-
Customer loyalty is a major objective for marketers els of commitment (Mikulincer and Shaver 2007). In sum-
mary, although we expect a negative association between
because it yields favorable business outcomes. Paulssen
attachment anxiety and loyalty, we hypothesize that the
(2009) hypothesized that general business attachment styles
relative size of this effect in customer–firm relationships is
would have a direct effect on loyalty intentions but fails to
smaller for anxious than for avoidant customers:
find empirical support for this prediction in a sample of
business-to-business customers. Unlike Paulssen (2009), we H3c: The negative relationship between customer attachment
study relationship-specific attachment styles, so we expect anxiety and loyalty is smaller (less negative) than the nega-
that customers’ attachment anxiety and avoidance influence tive relationship between customer attachment avoidance
their loyalty, similar to the influence that attachment styles and loyalty to the firm.
have on commitment in interpersonal relationships (Simp- In addition, we propose that customer attachment anxiety
son 1990). Our rationale is based on the observation that and avoidance interact to influence loyalty. Our theoretical
attachment researchers and marketing scholars define com- rationale is based on the fearful attachment style of cus-
mitment in the same way: in terms of a person’s intention to tomers who have high levels of both attachment anxiety and
remain in a relationship and invest in it (e.g., Simpson 1990). avoidance. Recall that fearful customers fluctuate between
Adults with low attachment anxiety and avoidance have a hyperactivated attachment strategy (associated with anxi-
longer-lasting romantic relationships than highly anxious ety) and a deactivated attachment strategy (associated with
and avoidant people (Feeney and Noller 1990). However, avoidance). Customers with a fearful attachment style either
the underlying theoretical reasons for these patterns are dif- are preoccupied with screening for indicators of the firm’s
ferent for anxiety than for avoidance. Anxious people desire lack of availability, caring, or responsiveness or strive to be
a committed relationship with the partner and tend to form self-protective and remain detached from the firm. Either
high levels of commitment relatively quickly; because they strategy can distort perceptions in a relationship and hinder
are strongly committed, often before they know a partner the consolidation of long-lasting bonds (Mikulincer and
well, they are more vulnerable to being disappointed in and Shaver 2007). Thus, we believe that a customer’s fluctua-
feeling hurt by a partner. In contrast, avoidant people strive tion between both strategies interferes with customer–firm
for a self-protective deactivation of their attachment needs; relationship maintenance and hypothesize a negative inter-
therefore, they have higher threshold levels for making action effect on loyalty:
commitments and commit less frequently and, when they
H3d: Customer attachment anxiety and customer attachment
are in a relationship, to a lower degree (Feeney and Noller avoidance interact to decrease loyalty to the firm.
1990). Extending these patterns to commercial relation-
ships, we predict that anxious customers’ lack of commit- Next, we describe the study context and research design,
ment stems from disappointment in or frustration with the describe our measures, assess the reliability and validity of
firm, whereas avoidant customers’ lack of commitment our measures, and show descriptive statistics for all
stems from their unwillingness to invest in any long-term variables.
relationship with the firm. In summary, customers with
RESEARCH DESIGN AND METHODOLOGY
lower (higher) attachment anxiety and avoidance are more
(less) loyal in commercial relationships: Study Context, Survey Design, and Sample
H3a: Customer attachment anxiety is negatively related to loy- We collaborated with the insurance division of a large
alty to the firm. financial services company that offers its products under
H3b: Customer attachment avoidance is negatively related to a single brand name in North America. The business-to-
loyalty to the firm. consumer insurance context is suitable for our research for
two reasons: First, a core benefit of insurance services is a median education of “some college.” Their tenure with the
security provided to customers by their commercial partner firm ranged from less than 1 year to 56 years, with a median
in times of need, which fits our attachment theory perspec- of 4 years (M = 7.85, SD = 9.02). Because 1.95% of item
tive. Second, insurance is a highly abstract and complex responses were missing, we replaced them with the mean
service that entails future benefits, resulting in a strong value. All scales in the survey were distributed normally, so
focus on RM by insurance companies (Crosby, Evans, and we conducted hypothesis tests using untransformed data
Cowles 1990; Verhoef 2003). The company cooperating in (Curran, West, and Finch 1994). Tables 1 and 2 show the
this research strives to build close bonds with its clients focal constructs, measures, and their sources.
through a network of more than 600 retail outlets operated
by exclusive agents and supported by customer contact/call Measurement of Customer Attachment Styles
centers and an interactive website. The firm offers six major As Table 1 displays, we measured customer attachment
product categories of insurance policies (life, home, car, styles with an eight-item self-report scale (using seven-
health, travel, and special risks), with a strategic emphasis point Likert-type items). Mende and Bolton (2011) devel-
on life and property insurance. Approximately 45% of the oped this scale, basing it on the dominant attachment style
firm’s customers purchase two or more product (insurance) instrument in psychology, the Experiences in Close Rela-
categories. Because the firm operates in regions with pub- tionships scale (Brennan, Clark, and Shaver 1998; Fraley,
licly funded health care, our sample did not contain any cus- Waller, and Brennan 2000). During the formal scale devel-
tomers with health insurance policies. opment process, Mende and Bolton examined reliability and
A market research firm telephoned a probability sample validity statistics of the customer attachment styles meas-
of 7500 customers of this firm. Although no incentive was ures using three samples in three service settings. Across
offered for participation, 1223 people agreed to participate these three studies, the coefficient alpha scores for the cus-
in the study. We excluded 24 respondents from our analyses tomer attachment anxiety and avoidance subscales ranged
because of erroneous information (compared with firm from .77 to .87, consistently exceeding the threshold level
records), so the final sample contained 1199 customers, of .70 (Nunnally and Bernstein 1994). Indicating discrimi-
yielding a response rate of 16%. Respondents completed a nant validity (Fornell and Larcker 1981), the average vari-
structured 15–20 minute interview that began with respon- ance extracted of customer attachment anxiety and attach-
dents’ assessments of the firm (e.g., relationship quality) ment avoidance was greater than the square of the
before eliciting their preference for closeness, attachment correlation between them. Finally, using confirmatory fac-
style, and classification variables. They ranged in age from tor analyses, the authors demonstrated discriminant validity
20 to 91 years, with a median age of 45 years (M = 47.09, between customer attachment styles and attachment
SD = 17.02); in addition, 44.50% were women and reported strength as well as relationship quality (satisfaction, trust,
Table 1
MEASUREMENT OF DEPENDENT VARIABLES AND THEORETICAL ANTECEDENTS
Construct Measures Scale Statistics Source

Dependent Variables
Preference for •[Firm] should contact me every three months just to “stay in touch.” N.A. (formative Based on Barnes (1997);
closeness •In a typical month, I spend a lot of time reading [firm’s] material, visiting its indicators) Berscheid, Snyder, and
website, interacting with its employees, or thinking about [firm]. Omoto (1989); Crosby,
•I am open to [firm] guiding my plans regarding my financial security. Evans, and Cowles (1990);
•[Firm] should actively offer me additional financial services that fit my needs. Doney and Cannon (1997);
•I do not like it when [firm] asks me to recommend it to other people. (R) Mohr, Fisher, and Nevin
•If [firm] asked me, I would discuss my views about its service quality. (1996)
•I prefer to meet my agent in person rather than talking to him/her on the phone.
•I would like to have a closer relationship with [firm] than I do right now.
Repurchase intention •I am very likely to continue buying my insurance from [firm]. N.A. Adapted from Mittal and
Kamakura (2001)
Change in •Difference score: “number of product categories in 2010 – number of product N.A.
relationship categories in 2007” (data from collaborating firm’s data base); difference score
breadth </=/> 0, indicating change in customer’s relationship breadth over time.
Theoretical Antecedents
Customer attachment •I worry about being abandoned by [firm] as a customer. Coefficient Mende and Bolton (2011)
anxiety •[Firm] changes how it treats me for no apparent reason. alpha: .77
•I worry that [firm] doesn’t really like me as a customer.
•I worry that [firm] doesn’t care about me as much as I care about it.
Customer attachment •It is a comfortable feeling to depend on [firm]. (R) Coefficient Mende and Bolton (2011)
avoidance •I am comfortable having a close relationship with [firm]. (R) alpha: .80
•It’s easy for me to feel warm and friendly toward [firm]. (R)
•It helps to turn to [firm] in times of need. (R)
Notes: All items were measured on seven-point Likert-type scales (1 = “strongly disagree,” 4 = “neutral,” and 7 = “strongly agree”). (R) = Item was reverse
keyed. N.A. = not applicable.
Table 2
MEASUREMENT OF CONTROL VARIABLES
Construct Measures Source

Relationship quality3 •Satisfaction: I am satisfied with [firm]; I am content with [firm]; I am happy with [firm].b Aaker, Fournier, and Brasel (2004);
Thomson (2006)
•Trust: [Firm] is trustworthy; [Firm] keeps its promises; [Firm] is truly concerned about my Doney and Cannon (1997)
welfare.b
•Affective commitment: I enjoy being a customer of firm]; I have positive feelings about Coulter, Price, and Feick (2003);
[firm]; I feel attached to [firm].b Gruen, Summers, and Acito (2000)
Anticipated regret3 •If I were to switch away from [firm], I might regret it.b Tokman, Davis, and Lemon (2007)
Switching barriers3 •Switching costs: Switching from [firm] would require more time and effort than I am willing Bougie, Pieters, and Zeelenberg
to put forth.b (2003)
•Customer lock-in: I would easily find another satisfactory insurance provider if I left [firm]. Johnson, Barksdale, and Boles
[R]b (2001)
Marketing levers •Price equity: My insurance premium is fair; Premiums are reasonable considering the service Bolton and Lemon (1999); Verhoef,
“price” and I receive; [Firm] gives me my money’s worth.b Langerak, and Donkers (2007)
“advertising”3 •Advertising affect: I like [firm]’s advertising.b Bergkvist and Rossiter (2007)
Critical incidents3 •Claim (dummy coded: 1 = yes, 0 = noa)
•Complaint (dummy coded: 1 = yes, 0 = noa)
Relationship •Relationship length: tenure (in years)
characteristics2 •Relationship breadth: number of product categories client has with insurance company [1, 2, or 3]
Customer •Age2 (in years)
characteristics1, 2, 3 •Gender1 [dummy: female = 0a, male = 1]
•Education3 [dummy: low (e.g., 12th grade), medium (e.g., college), and higha (e.g., MS)]
•Income per household p.a.3[dummy: low < 50, medium = 51–75, and higha > 76, in 000]
aThis category served as the dummy reference category.

bVariable measured on seven-point Likert-type scales (1 = “strongly disagree,” and 7 = “strongly agree”). Coefficient alphas for multi-item measure: rela-
tionship quality = .94; price equity = .87.
Notes: Data sources: 1 = interviewer, 2 = firm’s data base, and 3 = customer self-reported.
and affective commitment). In summary, the customer explicitly assesses a partner’s influence on the respondent’s
attachment measures exceed thresholds for reliability and plans for achieving a particular financial standard and mak-
validity, correctly reflect customer attachment anxiety and ing major investments. Therefore, two of our items refer to
avoidance, and capture all defining content facets of the being guided in the area of financial planning and allowing
attachment dimensions. the firm to actively offer financial services. The RCI cap-
tures the extent to which a partner influences the respon-
Measurement of Preference for Closeness dent’s activities within his or her social network and the
No scale in the marketing literature measures customer extent to which the respondent accepts responsibilities
preference for closeness, so we derived an eight-item scale within the focal relationship. Thus, we selected two items
from Berscheid, Snyder, and Omoto’s (1989) 60-plus-item examining customers’ openness to actively help acquire
RCI (see Table 1). Consistent with prior RM research new customers through referrals and customers’ willingness
(Mohr, Fisher, and Nevin 1996), we consider these eight to provide feedback to the firm about its service quality. Our
items formative measures. They consist of combinations of final two measures accounted for customers’ need for
indicators that represent different customer–firm inter- human interaction (i.e., preference for face-to-face meetings
actions. They are necessary but separate facets of closeness. rather than talking on the phone) and their global desire to
Thus, the items are not highly correlated with one another, develop a closer relationship with the firm (Barnes 1997;
and it is inappropriate to report internal scale statistics. Web Dabholkar 1996).
Appendix A (www.marketingpower.com/jmr_webappendix)
provides further details regarding the construction and Measurement of Loyalty Intentions and Loyalty Behavior
assessment of this formative indicator in light of established We measured repurchase, or loyalty, intentions with one
guidelines (Diamantopoulos and Winklhofer 2001); here, item (“I am likely to continue buying from [firm]”; seven-
we provide only a brief overview. point Likert-type item) because Bergkvist and Rossiter
Berscheid, Snyder, and Omoto (1989) emphasize the (2007) show that single-item measures achieve the same
need to adjust closeness items to ensure that they address predictive validity as multi-item measures, provided the
the frequency and diversity of interacting and the influence focal construct is concrete and singular. We captured loyalty
of partners on each other in the population under study. behavior through changes in customers’ relationship
Therefore, we adjusted item wordings to apply to customers breadth, defined as the variation in the number of different
in a financial services context. Two items refer to the fre- offerings customers purchased from the firm (Bolton,
quency of interacting and the amount of time customers Lemon, and Verhoef 2004). We identified changes in rela-
spend interacting with the firm (Berscheid, Snyder, and tionship breadth using purchase records for 975 of the 1199
Omoto 1989; Crosby, Evans, and Cowles 1990). The RCI customers who participated in our survey; the firm’s pur-
chase records for the remaining 224 were incomplete or the number of product categories a customer held in 2007
lacking.1 Specifically, in 2007, the insurance company pro- (e.g., life, car). Customers can have up to six product cate-
vided us with data on how many of the six product categories; customers in our sample had one (N = 464), two (N =
gories customers held at that point. In addition, the firm pro- 530), and three (N = 205) product categories. Finally, we
vided data on the number of product categories customers controlled for critical incidents (claims and complaints;
who had participated in our survey (in 2007) held in Smith and Bolton 1998) and demographics (age, education,
December 2010 (i.e., three years later). We computed dif- income, and gender; Cooil et al. 2007).
ference scores for individual clients as “number of product We measured both relationship quality and price equity
categories2010 – number of product categories2007.” This with multiple items; they were represented by an index
score is positive for customers who increased their relation- (computed as the mean of the respective items) as described
ship breadth by cross-buying, and it is negative for those in Table 2. The coefficient alpha scores for these scales
who cancelled existing insurance categories between 2007 (relationship quality = .94; price equity = .87) were above
and 2010. The difference score is zero for customers who the minimum level of .70 (Nunnally and Bernstein 1994).
did not change the number of categories. To have enough To assess convergent validity for the relationship quality
observations for each level of the difference score, we col- and price equity constructs, we ran separate principal com-
lapsed customers with positive scores into one group (N = ponent analyses for both groups of items. In both analyses,
96) and customers with negative scores into another group only one component was extracted. All factor loadings were
(N = 118); the third group included the customers who did .75 or higher, in support of convergent validity.
not change their relationship breadth (N = 761).
Prior research has studied customer cross-buying of Preliminary Descriptive Statistics
financial services as a binary outcome variable (e.g., Salazar, Table 3 provides the descriptive statistics and correlations
Harrison, and Ansell 2007). Our analytical focus on changes among the various constructs. Means and standard devia-
in relationship breadth includes the notion of cross-buying tions are consistent with prior research. Item correlations
because cross-buying is an indicator for relationship exten- across constructs are substantially lower than item correla-
sion. Our investigation is conceptually richer because it not tions within constructs, in support of convergent and dis-
only addresses the antonym of cross-buying (i.e., “no cross- criminant validity. As we expected, customer attachment
buying”) but also splits the “no cross-buying” category to anxiety and avoidance had modest positive correlations
distinguish (1) customers who did not change their relation- with each other (r = .18).
ship breadth from (2) customers who reduced their relation-
ship breadth. Understanding these three groups is relevant MODEL SPECIFICATION AND ESTIMATION
from both theoretical and managerial perspectives. We hypothesized that customers’ attachment styles would
affect their preferences for a close relationship with the firm
Measurement of Control Variables (H1) and that both (preference for closeness and attachment
We included a broad set of covariates that have been styles) would influence loyalty (H2 and H3). We tested these
shown both theoretically and empirically to influence loy- hypotheses by estimating separate equations for customer
alty (Table 2). A first set of covariates included satisfaction, preferences for closeness, repurchase intentions, and behav-
trust, and affective commitment (Garbarino and Johnson ioral loyalty (i.e., changes in relationship breadth). The pre-
1999; Verhoef 2003), which we modeled as the higher-order dictor variables in each equation included customer attach-
construct relationship quality (De Wulf, Odekerken- ment anxiety, attachment avoidance, and their two-way
Schröder, and Iacobucci 2001). We also included antici- interaction, which we computed by multiplying mean-
pated regret (Inman, Dyer, and Jia 1997), switching barriers centered anxiety and avoidance variables (Aiken and West
(Palmatier et al. 2006), price equity (i.e., price fairness) 1991). The three equations also included all the covariates
(Bolton and Lemon 1999), advertising affect (Bergkvist and listed in Table 2. We estimated the closeness and intentions
Rossiter 2007), as well as the length and breadth of the rela- equations using ordinary least squares (OLS). We estimated
tionship (from the firm’s purchase records) (Bolton, Lemon, the behavioral loyalty equation using multinomial logistic
and Verhoef 2004). We measured relationship breadth by regression (MLR) (in Stata), which fits our analytical goals
for three reasons: First, MLR is well established (Lilien,
1Note that we could not infer that these 224 customers defected from the
Kotler, and Moorthy 1992); indeed, it is “perhaps the most
firm. Limitations in the firm’s database prevented us from tracking certain frequently used choice model on marketing” (Leeflang et al.
customer identification numbers over time; for example, when a customer 2000, p. 241). Second, Kumar and Shah (2009) underscore
moves into a different market/geographical district or gets a new phone that multinomial logit models are insightful for questions
number, the system is prone to lose track of the focal customer identifica- related to customer segmentation, which is at the heart of our
tion number (and may erroneously assign a new client identification num-
ber). As a result, these customers may still be with the firm, and specifying
investigation. Third, logit models are useful to examine cus-
them as “defected” might introduce considerable error to the analyses. tomers’ relationship breadth (Bolton, Lemon, and Verhoef
Therefore, our subsequent analyses exclude customers with incomplete 2008; Shah et al. 2012), which is our focus.2 Tables 4, 5, and
information. This analytical approach notwithstanding, we also estimated a
Heckman-type selection model to account for potential sample selection
bias resulting from the 224 missing customers. We specified the selection 2We also explored an ordered logit model. We tested the parallel regression
equation as a bivariate probit model and the outcome equation as a multi- assumption (i.e., the proportional odds assumption or parallel lines assump-
nomial probit model (Roodman 2011). The results appear in Web Appen- tion) underlying this model using a Brant test (Long and Freese 2006). This
dix B (www.marketingpower.com/jmr_webappendix). These additional Brant test was significant, showing that various variables in this model vio-
analyses show only small changes compared with the results we report sub- lated the parallel regression assumption. This indicates that a more flexible
sequently. Overall, the insights from the selection model support our theo- model—one that does not impose the constraint of parallel regressions—
retical framework and the robustness of our hypothesized effects. should be considered, such as the MLR (Long and Freese 2006).
Table 3
CORRELATION MATRIX
Decoding Customer–Firm Relationships
Correlations
Measures M SD 1 2 3 4 5 6 7 8 9 10 11 12 13
1. Attachment anxiety 2.46 1.38 1.00
2. Attachment avoidance 3.06 1.37 .18* 1.00
3. Preference for closeness 4.04 .96 .13* –.40* 1.00
4. Repurchase intention 5.88 1.41 –.30* –.49* .15* 1.00
5. Satisfaction 5.83 1.37 –.31* –.50* .24* .56* 1.00
6. Trust 5.63 1.29 –.28* –.61* .28* .47* .77* 1.00
7. Affective commitment 5.43 1.39 –.27* –.65* .30* .53* .81* .80* 1.00
8. Relationship quality 5.62 1.33 –.31* –.66* .29* .56* .93* .92* .94* 1.00
9. Switching costs 5.61 1.25 .02 –.21* .09* .29* .21* .19* .21* .22* 1.00
10. Customer lock-in 4.41 2.11 –.21* –.22* .14* .22* .25* .24* .28* .28* .11* 1.00
11. Anticipated regret 3.48 1.93 –.09* –.39* .27* .42* .37* .38* .41* .41* .39* .31* 1.00
12. Price equity 4.48 1.90 –.25* –.54* .25* .50* .75* .69* .74* .78* .20* .25* .39* 1.00
13. Advertising affect 5.36 1.41 –.10* –.40* .27* .27* .42* .47* .48* .49* .08* .19* .26* .42* 1.00
*Correlation is significant at .01 level (two-tailed); N = 1199.
Notes: Consistent with prior marketing research (De Wulf, Odekerken-Schröder, and Iacobucci 2001; Palmatier et al. 2006) and in light of the results from a confirmatory factor analysis, we calculated an overall
index of “relationship quality” based on the customers’ respective ratings of satisfaction, trust, and affective commitment. We used this overall index as an independent variable in the subsequent analyses.
133
Table 4
CUSTOMER ATTACHMENT STYLES AND THE PREFERENCE FOR CLOSENESS TO THE FIRM (H1)
Variable Predicted Sign Unstandardized Coefficient (SE) Standardized Coefficient

Customer attachment anxiety + .16 (.02) .24***
Customer attachment avoidance – –.23 (.02) –.34***
Anxiety ¥ avoidance +/– –.00 (.01) –.01
Relationship quality N.A. .01 (.01) .04
Price equity N.A. .01 (.03) .01
Anticipated regret N.A. .06 (.02) .13***
Switching costs N.A. –.02 (.01) –.05*
Customer lock-in N.A. .02 (.01) .05†
Advertising affect N.A. .08 (.02) .11***
Claim (yes) (no = reference category) N.A. –.15 (.05) –.08**
Complaint (yes) (no = reference category) N.A. .24 (.08) .07*
Tenure N.A. –.01 (.00) –.06*
Product categories N.A. –.07 (.04) –.05*
Gender (M) (female = reference category) N.A. .12 (.05) .07*
Age N.A. .00 (.00) –.03
Income low N.A. .14 (.06) .07*
Income medium N.A. .03 (.06) .01
Education low N.A. .01 (.07) .00
Education medium N.A. –.03 (.07) –.02
Intercept N.A. 3.72 (.21)
†p < .10.
*p < .05.
**p < .01.
***p < .001.
Notes: N.A. = not applicable. Dependent variable: preference for closeness; estimation method: OLS regression. R2: .276; adjusted R2: .264; F (19, 1179) =
23.64, p < .001.
Table 5
CUSTOMER PREFERENCE FOR CLOSENESS, ATTACHMENT STYLES, AND REPURCHASE INTENTIONS (H2, H3)
Variable Predicted Sign Unstandardized Coefficient (SE) Standardized Coefficient

Customer attachment anxiety – –.17 (.03) –.16***
Customer attachment avoidance – – .21 (.03) –.19***
Anxiety ¥ avoidance – –.07 (.02) –.10***
Preference for closeness + –.07 (.04) –.04†
Relationship quality N.A. .08 (.02) .21***
Price equity N.A. .12 (.04) .12**
Anticipated regret N.A. .11 (.02) .15***
Switching costs N.A. .09 (.02) .12***
Customer lock-in N.A. .01 (.02) .01
Advertising affect N.A. –.02 (.03) –.02
Claim (dummy) (no = reference category) N.A. .01 (.07) .00
Complaint (dummy) (no = reference category) N.A. .15 (.12) .03
Tenure N.A. .00 (.00) .01
Product categories N.A. .08 (.05) .04†
Gender (M) (female = reference category) N.A. –.09 (.07) –.03
Age N.A. .01 (.00) .06*
Income low N.A. –.01 (.10) .00
Income medium N.A. –.06 (.09) –.02
Education low N.A. –.01 (.09) .00
Education medium N.A. .06 (.09) .02
Intercept N.A. 2.88 (.32)
†p < .10.
*p < .05.
**p < .01.
***p < .001.
Notes: N.A. = not applicable. Dependent variable: repurchase intentions; estimation method: OLS regression. R2: .442; adjusted R2: .432; F (20, 1178) =
46.60, p < .001.
6 present the corresponding estimated equations. Finally, as (R2 = .276, adjusted R2 = .264, F(19, 1179) = 23.64, p <
we further address in the following section, we explored a .001; Table 4). The OLS equation for repurchase intentions
potential mediating role of preference for closeness using a explained 43% in the dependent variable (R2 = .442,
bootstrapping approach (Hayes and Preacher 2011). adjusted R2 = .432, F(20, 1178) = 46.60, p < .001; Table 5).
Closeness and Repurchase Intentions Equations Change in Relationship Breadth Equation
The OLS equation for preference for closeness to the firm We explored the statistical fit for the MLR model describ-
accounted for 26% of the variance in the dependent variable ing change in relationship breadth with three (pseudo-R-
Table 6
CUSTOMER PREFERENCE FOR CLOSENESS, ATTACHMENT STYLES, AND LOYALTY BEHAVIOR OVER TIME (H2, H3)
Parameter Estimates
Logistic Function 1: Increasing Logistic Function 2: Decreasing Relationship
Relationship Breadth Versus No Change Relationship Breadth Versus No Change
Variable Coefficient (SE) z pr ± SD/2 rrr Coefficient (SE) z pr ± SD/2 rrr
Customer attachment anxiety –.27 (.10)*** –2.69 –.026 .76 –.23 (.10)** –2.30 –.020 .80
Customer attachment avoidance .15 (.12) 1.27 .015 1.17 .11 (.11) .96 .010 1.12
Anxiety ¥ avoidance .00 (.07) –.03 –.002 1.00 .09 (.05)* 1.67 .014 1.09
Preference for closeness .44 (.14)*** 3.13 .031 1.56 .08 (.13) .65 .003 1.09
Relationship quality –.01 (.06) –.20 –.002 .99 –.04 (.06) –.79 –.012 .96
Price equity .15 (.14) 1.02 .015 1.16 –.01 (.12) –.05 –.002 .99
Switching costs .05 (.06) .86 .007 1.05 .04 (.06) .77 .006 1.04
Customer lock-in –.01 (.06) –.02 –.002 .99 –.02 (.06) –.35 –.003 .98
Anticipated regret –.01 (.07) –.10 –.001 .99 .02 (.07) .25 .003 1.02
Advertising affect .05 (.10) .47 .005 1.05 –.01 (.09) –.07 –.001 .99
Claim (dummy) .12 (.24) .49 .009 1.12 –.05 (.22) –.21 –.004 .96
Complaint (dummy) –.39 (.45) –.86 –.027 .68 .39 (.35) 1.11 .035 1.47
Tenure –.01 (.02) –.40 –.006 .99 .02 (.01) 1.36 .001 1.02
Product categories (2007) –.44 (.18)** –2.49 –.031 .64 1.59 (.18)*** 8.72 .088 4.89
Gender (dummy) .21 (.24) .90 .016 1.24 –.11 (.22) –.51 –.010 .89
Age –.03 (.01)*** –3.63 –.039 .97 .01 (.01) 1.14 .014 1.01
Income low –.41 (.34) –1.22 –.027 .66 –.31 (.31) –1.01 –.020 .73
Income medium –.14 (.30) –.48 –.010 .87 –.05 (.27) –.02 –.002 .96
Education low .16 (.32) .52 .010 1.18 .37 (.30) 1.23 .026 1.45
Education medium .01 (.32) .01 .000 1.01 .01 (.31) .04 .001 1.01
Intercept –2.53 (1.14) –5.63 (1.10)
Overall Model Evaluation: Likelihood ratio test: 2(40) = 170.00, p < .001
Fit: Cox and Snell = .16, Nagelkerke = .22, McFadden = .13
*p < .10.
**p < .05.
***p < .01.
Notes: Dependent variable: change in relationship breadth (2007–2010); estimation method: multinomial logistic regression. The column “ pr ± SD/2”
reports the change in probabilities as the independent variable varies from one-half standard deviation below to one-half standard deviation above its mean
(holding all other variables constant at their mean); for dummy variables, it shows the change in probability as the variable changes from 0 to 1 (Long and
Freese 2006). The column “rrr” reports the relative risk ratio, which is defined as the exponentiated coefficients exp (b) (Long and Freese 2006). Statistically
significant coefficients are displayed in boldface.
square) statistics: Cox and Snell (.16), Nagelkerke (.22), tomers who increased or decreased their relationship
and McFadden (.13) (Table 6). Although considering a breadth. This feature is relevant because prior research sug-
model’s hit rate is not always useful when a majority of gests that customer cross-buying may be influenced, at least
observations falls in one category (Iyengar and Gupta in part, by different antecedents compared with their inac-
2006), we note that the model equation correctly classified tion or decreases in relationship breadth (Kumar, Morris, and
78% of the customers into the three groups (increase, Pancras 2008; Verhoef 2003). Logistic function 1 compares
decrease, and no change in relationship breadth). That is, customers who increased their relationship breadth with
our model significantly outperforms the benchmark model customers who did not change it (reference group). Logistic
according to the proportional chance criterion with its function 2 compares customers who decreased their rela-
chance hit rate of 63.4% (z = 9.59, p < .001) (Hair et al. tionship breadth with customers who did not change it.
2010). In addition, Press’s Q of 881.40 exceeds the critical
chi-square value of 10.83 (d.f. = 1, p < .001) (Hair et al. RESULTS OF HYPOTHESES TESTS
2010). Both the proportional chance criterion and Press’s Q H1: The Influence of Customer Attachment Styles on
indicate that the classification accuracy is greater than that Preference for Closeness
expected by chance. Moreover, a significant likelihood ratio
test (2 = 170.00, d.f. = 40, p < .001) indicated that our The results support our hypotheses that anxiety is posi-
model was more effective than the null model. Likelihood tively related (H1a) and avoidance is negatively related
ratio tests showed that four variables contributed signifi- (H1b) to customer preference for closeness (p < .001) (Table
cantly to the overall model: preference for closeness (2 = 4). H1c predicted an anxiety-by-avoidance interaction with-
10.11, d.f. = 2, p < .01), attachment anxiety (2 = 12.22, d.f. = out specifying the direction of the net effect; it was not sup-
2, p < .01), product categories2007 (2 = 107.91, d.f. = 2, p < ported (p > .05). Because this is the first study to report on
.001), and age (2 = 16.51, d.f. = 2, p < .001). antecedents of customer preference for closeness, we briefly
For a dependent variable with three groups, the MLR examine the results for the covariates.
estimates two logistic functions—that is, one function for Anticipated regret (p < .001), advertising affect (p <
each group relative to the reference group. We defined the .001), and customer lock-in (p < .10) had a positive effect
reference group as customers who did not change the on preference for closeness, whereas switching costs had a
breadth of their relationship and compared them with cus- negative effect (p < .05). While an insurance claim had a
negative effect (p < .01) on preference for closeness, a com- H3: The Influence of Customer Attachment Styles on
plaint had a positive effect (p < .01), consistent with the Loyalty
service recovery paradox whereby a well-managed com- Repurchase intentions. In support of our hypotheses, cus-
plaint can improve a complainant’s overall assessment of tomer attachment anxiety (H3a) and avoidance (H3b) were
the firm to levels higher than it was before the failure negatively related to repurchase intentions, and this nega-
(Smith and Bolton 1998). Customers with longer tenures or tive effect of attachment anxiety on loyalty was smaller than
who held more product categories had lower levels of pref- the effect of avoidance (H3c). The anxiety ¥ avoidance inter-
erence for closeness to the insurance company (p < .05). action was negatively related to repurchase intentions (H3d)
Customers with a low annual household income reported a (all effects p < .001) (Table 5). To interpret the interaction
higher preference for closeness than customers with a high term, we calculated simple slopes of repurchase intentions
income (p < .05). Finally, men reported a higher preference regressed on anxiety for the mean avoidance and one stan-
for closeness than women (p < .05), consistent with research
dard deviation above and below the mean of avoidance
showing that men are more likely to engage in financial
(Aiken and West 1991). The slope of the relationship
planning than women (Hira and Mugenda 2000).
between repurchase intention and anxiety became more
H2: The Influence of Customer Preference for Closeness on negative with increasing levels of avoidance. Thus, there is
Loyalty a deleterious interplay of the two attachment dimensions on
repurchase intentions. Finally, consistent with prior
Repurchase intentions. Customer preference for close-
research, relationship quality (p < .001), price equity (p <
ness had a small, marginally significant (p < .10), negative
.01), anticipated regret (p < .001), and switching costs (p <
effect on repurchase intentions (Table 5), contrary to H2.
.001) positively influenced repurchase intentions. Age (p <
Multicollinearity could be one reason a coefficient emerges
.05) and product categories (p < .10) were positively associ-
in the opposite direction than predicted. Thus, we examined
ated with repurchase intentions.
variance inflation factors (VIFs) as indicators of multi-
Change in relationship breadth: Effects of customer
collinearity in our survey data. A VIF approaching or
exceeding 10 indicates a multicollinearity problem (Hair et attachment styles. Recall that logistic function 1 compared
al. 2010; O’Brien 2007). Examining our variables, we found customers with increases in relationship breadth with cus-
that the highest VIF was for relationship quality (3.72), well tomers with no change. The MLR coefficient estimate for
below the established threshold. The second highest VIF attachment anxiety was negative and significant (b = –.27,
was price equity (2.64), and 18 of the 20 variables in the p < .01; pr = –.026; rrr = .76). This result supports H3a and
model had VIFs of less than 2.07. These insights suggest is consistent with the repurchase intentions equation. If
that multicollinearity is not the explanation for this negative attachment anxiety increases by one standard deviation, the
coefficient3; rather, we provide theory-based arguments for probability of a customer engaging in cross-buying
this effect in the discussion. decreases by approximately 3% (all else being equal).
Change in relationship breadth. In support of H2, the Logistic function 2 compared customers with decreases
MLR estimate for the effect of preference for closeness on in relationship breadth with customers with no change.
cross-buying behavior was positive and significant (b = .44, Again, the MLR coefficient estimate for anxiety was nega-
p < .01) (Table 6, function 1). If a customer’s preference for tive and significant (b = –.23, p < .05; pr = –.020; rrr =
closeness score increases by one point, the multinomial log- .80). If a customer’s attachment anxiety increases by one
odds for increasing the number of product categories (vs. standard deviation, the probability of decreasing the num-
not changing it) increases by .44 units (all else being equal). ber of product categories2007 decreases by approximately
In Table 6 (column “pr ± SD/2”), we also report the 2% (all else being equal). The notable pattern of these three
change in probabilities as the independent variable varies anxiety-related results (from the two logistic functions and
from one-half standard deviation below to one-half standard the OLS equation on repurchase intentions) is consistent
deviation above its mean (holding all other variables con- with the notion that anxious people are reluctant to actually
stant) (Inman, Winer, and Ferraro 2009; Long and Freese leave a relationship; that is, they tend to keep the status quo
2006). With such a one-standard-deviation increase in pref- and remain in the focal relationship.
erence for closeness (pr = .031), the probability of a cus- Customer attachment avoidance was not statistically sig-
tomer cross-buying increases by approximately 3%, hold- nificant; thus, H3b and, consequently, H3c were not supported
ing all other variables constant. The relative risk ratio (rrr) in the loyalty behavior model. Logistic function 2 showed a
in Table 6 shows that the relative risk of the customer falling positive and marginally significant anxiety-by-avoidance
into the cross-buying group (vs. the “no change” group) is interaction (b = .09, p < .10; pr = .014; rrr = 1.09), in sup-
expected to increase by a factor of 1.56 as the preference for port of H3d and consistent with the intentions results.
closeness increases by one unit, holding all other variables Change in relationship breadth: other antecedents. We
constant. In the “Discussion” section, we present a theoreti- briefly discuss our results regarding the other two antecedents
cal explanation for why preference for closeness had a that significantly influence this outcome variable. In logis-
negative effect on repurchase intentions but a positive effect tic function 1, the MLR estimate for product categories2007
on cross-buying behavior. regarding cross-buying was negative and significant (b =
–.44, p < .05; pr = –.031; rrr = .64); in function 2, the esti-
mate for product categories2007 was positive and significant
(b = 1.59, p < .01; pr = .088; rrr = 4.89); this effect in
3As an additional robustness check, we also conducted all our analyses
without the two variables with the highest VIFs (relationship quality and
price equity) in the models. We found that the pattern of the hypothesized explaining a decrease of relationship breadth was large.
effects remained virtually the same in the absence of these variables. Moreover, in logistic function 1, the MLR estimate for age
was negative and significant (ß = –.03, p < .01; pr = –.039; measuring the frequency, diversity, and mutually influential
rrr = .97). This negative effect contrasts the positive effect nature of customers’ interactions with a firm. Customer
the intentions-based regression model reported for age. attachment anxiety is positively related to the preference for
However, cross-buying is a different facet of customer loy- closeness, but attachment avoidance is negatively related to
alty than a mere repurchase (i.e., prolonging existing insur- it. Attachment styles explain substantial variation (approxi-
ance policies). mately 10%) in preferences for closeness after accounting
for a broad set of covariates drawn from the marketing and
Ancillary Analysis: Exploring the Mediating Role of psychology literature. These insights, which are based on
Preference for Closeness linking two major theories (attachment theory and inter-
We hypothesized that customer attachment styles affect dependence theory), help marketers identify customers who
both preference for closeness (H1) and loyalty (H3); in par- seek or shun closeness in commercial relationships.
allel, we expected that preference for closeness influences loy-
Customer Preference for Closeness Influences Loyalty
alty (H2). Taken together, these hypotheses may suggest that
preference for closeness mediates the effects of attachment In the intentions equation, preference for closeness has
styles on loyalty (i.e., repurchase intentions and change in an unexpected negative association with repurchase intent.
relationship breadth). Although we did not introduce a formal In the loyalty behavior model, there is a large positive effect
mediation hypothesis, exploring such a transmitting effect of preference for closeness on the likelihood that customers
seems fruitful, because customer attachment styles and pref- increase their relationship breadth (as hypothesized). That is,
erence for closeness are novel constructs to the RM arena. the effect of preference for closeness on repurchase inten-
Repurchase intentions. We tested for mediation using the tions is the opposite of its effect on cross-buying. We showed
“mediate” bootstrapping approach (Hayes and Preacher previously that multicollinearity is not a likely explanation
2011). Our model included the three customer attachment for this effect; rather, this pattern is consistent with two con-
style variables (anxiety, avoidance, and anxiety ¥ avoid- ceptual insights. First, consumers’ self-reported loyalty
ance) as independent variables, all covariates listed in Table intentions and their loyalty behavior over time are theoreti-
2, and repurchase intention as the dependent variable. The cally different phenomena that frequently deviate from each
potential mediator in the model was preference for close- other (Seiders et al. 2005). In his seminal study, Oliver (1999)
ness. We found that the mean indirect effect from the boot- develops numerous conceptual arguments for why loyalty
strap analysis for preference for closeness was not signifi- intentions will not always translate into loyalty behavior
cant, with the 95% confidence intervals (CIs) including zero (e.g., changes of need, variety seeking). Moreover, Seiders
for attachment anxiety (a ¥ b = –.01, 95% CI = –.03 to .00), et al. (2005, p. 26) examine the satisfaction–repurchase link
for avoidance (a ¥ b = .02, 95% CI = –.01 to .04), and for and show that “the results are significantly different for self-
the anxiety ¥ avoidance interaction (a × b = .00, 95% CI = reported repurchase intentions and objective repurchase
–.002 to .003). That is, although the corresponding values behavior” (as a function of customer, relational, and mar-
are close to the threshold, these results suggest that prefer- ketplace factors). Such intention–behavior discrepancies are
ence for closeness did not mediate the effects of customer particularly likely when the focal context involves infre-
attachment styles on repurchase intentions. quent and relatively complex purchase decisions (Seiders et
Change in relationship breadth. As we show in Table 6, al. 2005), as is the case in our insurance context (e.g., in
preference for closeness was significant in logistic function 2007, some consumers may not have been able to accurately
1 but not in logistic function 2. Thus, our mediation analy- predict their future insurance needs over multiple years). In
sis focused on preference for closeness using the two groups summary, prior RM research has shown that (1) intentions
compared in logistic function 1 (customers who cross- and behavior are, at least in part, driven by distinct theoreti-
bought vs. customers who did not change their relationship cal mechanisms and antecedents and (2) deviations between
consumers’ repurchase intentions and their objective pur-
breadth). Consequently, the dependent variable was bivari-
chase behavior occur frequently and, to a certain degree, are
ate, and we tested the hypothesized mediation effect using
to be expected. Nevertheless, the finding that preference for
the indirect bootstrapping approach (Preacher and Hayes
closeness had a negative (albeit small and only marginal)
2008). Our model included customer attachment anxiety as
effect on repurchase intentions suggests that it would be
the independent variable, preference for closeness as the
worthwhile to further reflect on it; this leads to the second
potential mediator, all covariates listed in Table 2, and
conceptual insight. We believe that a negative effect of pref-
cross-buying as the binary outcome variable. The mean
erence for closeness on repurchase intentions is consistent
indirect effect from the bootstrap analysis for preference for
with a need fulfillment explanation.
closeness was significant, with the 95% CIs excluding zero
Self-determination theory (SDT) holds that need fulfill-
for attachment anxiety (a ¥ b = .08, 95% CI = .03 to .14).
ment is the major motivator for people to build and main-
This finding shows that preference for closeness mediates tain relationships (La Guardia et al. 2000). According to
the effect of customers’ attachment anxiety on their cross- SDT, relatedness describes a person’s feeling of being con-
buying behavior over time. nected with a partner and is widely considered a crucial
DISCUSSION relational need. Indeed, research on SDT has found the need
for relatedness to be a particularly strong predictor of rela-
Customer Attachment Styles Explain Preference for tionship well-being and functioning; accordingly, people
Closeness to the Firm expect their partners to be sensitive and responsive to this
This study introduces a way to reveal customer prefer- relational need (La Guardia et al. 2000). Notably, people
ences for closeness based on interdependence theory by who cannot fulfill their relational needs in a focal relation-
ship are motivated to seek fulfillment in other relationships; indirect effect of interpersonal anxiety and avoidance on
indeed, such people reported higher infidelity intentions in satisfaction (mediated by reciprocity in commercial rela-
the context of romantic relationships (Lewandowksi and tionships). In our context, preference for closeness partially
Ackerman 2006). mediated the effect of attachment anxiety on cross-buying
These theoretical insights suggest that customers should behavior, but it did not mediate regarding repurchase inten-
expect a company to be sensitive and responsive to their tions. Although these initial results are not conclusive, the
preference for closeness. The negative effect of preference prior marketing research and related theoretical support in
for closeness on repurchase intentions may reveal that, at two additional literature streams warrant exploring the find-
the time of our survey in 2007, some customers felt that this ings further.
need was not (fully) fulfilled. These customers may have Specifically, we can link attachment styles with the afore-
considered alternative commercial relationships and, conse- mentioned SDT and its focus on need fulfillment as a prom-
quently, reported lower repurchase intentions. At this point, ising platform for the study of such mediation. Similar to
it is important to recall that we observed change in relation- SDT, attachment theory recognizes the fundamental need
ship breadth three years after customers reported their for relatedness as a platform for attachment security (a
repurchase intentions. In that period, the insurance company proxy for low levels of anxiety and avoidance) (Hazan and
used insights from our 2007 survey to inform its marketing. Shaver 1994). Consistent with this theorizing, empirical
By responding to clients who preferred close(r) relation- work has shown that fulfilling the need for relatedness (both
ships, the firm may have encouraged cross-buying (indicated partially and fully) mediates the relationship between peo-
by the positive effect in the behavioral loyalty equation). ple’s attachment security and their psychological well-being
Overall, this finding is consistent with the idea that the orga- (e.g., risk of depression) (La Guardia et al. 2000). Second,
nizational ability to understand, identify, and respond to the research on employee–environment fit also builds on a need
preference for closeness provides a mechanism for firms to fulfillment lens to examine how employee attitudes are
encourage customer retention and cross-buying. influenced by the perceived fit between their desires and the
Customer Attachment Styles Influence Loyalty resources in the organizational environment available to
meet these desires (Greguras and Diefendorff 2009). This
As customer attachment anxiety and avoidance increase work has demonstrated that employee–organization fit pre-
(and interact), repurchase intentions decrease. In terms of dicts commitment to an organization and, importantly, that
their predictive weight (i.e., standardized coefficients), this effect is mediated by need fulfillment.
attachment styles are second only to relationship quality and While neither of the two literature streams has examined
are more influential than some established loyalty attachment anxiety and avoidance in detail, it is reasonable
antecedents (e.g., price equity, switching barriers). Cus- to extend the preceding theorizing to our context by propos-
tomer attachment styles also influence loyalty behavior. ing the following. Attachment styles predict customer pref-
First, there is a detrimental effect of attachment anxiety erence for closeness; in light of their preferences, customers
because the probability of a customer increasing the rela- then reflect on the extent to which a firm’s approach to cre-
tionship breadth (vs. not changing it) decreases with higher ating closeness fits their own preference. In turn, the extent
anxiety (Table 6, function 1). This effect is partially medi- of this perceived fit (i.e., need fulfillment) positively or
ated by preference for closeness. Second, there is a benefi- negatively influences customer loyalty to the firm. This
cial effect of attachment anxiety because the likelihood of a chain of effects suggests that a standardized approach to
customer decreasing the relationship breadth (vs. not chang- RM—which many firms have adopted—might backfire
ing it) decreases with higher anxiety (Table 6, function 2). twofold. Anxious customers will perceive a standardized
Third, there is a negative (but only marginally significant) format of customer-firm interactions as “not close enough.”
effect whereby the likelihood of customers reducing the
In parallel, avoidant customers will assess it as “too close,”
relationship breadth (vs. not changing it) increases with the
which can trigger negative responses (e.g., annoyance),
anxiety-by-avoidance effect (Table 6, function 2).
consistent with recent RM research (Godfrey, Seiders, and
Notably, these results reveal that customer attachment
Voss 2011). This final insight alone suggests that studying
anxiety reduces both cross-buying and disloyalty behaviors.
the potential mediating link among attachment styles, pref-
That is, consistent with research on interpersonal relation-
erence for closeness, and loyalty is important for marketing
ships (Davila and Bradbury 2001), and as proposed by H3c,
scholars.
customer attachment anxiety seems to be associated with a
motivation to maintain the status quo. These customers pre- CONCLUDING REMARKS
fer the status quo over reducing their relational investments
Managerial Implications
by shrinking the relationship breadth (switching costs or
customer lock-in cannot explain this effect, because we con- Our study provides multiple managerial implications.
trolled for both variables). First, firms should regularly measure customer attachment
styles and preferences for closeness. While not every cus-
Exploring the Mediating Role of Preference for Closeness tomer can be surveyed, it is common practice to survey
While it was not at the heart of our research, we explored samples of customers and extrapolate results to the cus-
whether preference for closeness mediates the effects of tomer population. This approach is reasonable to learn
customer attachment styles on loyalty. The idea that attach- about customer attachment styles. For example, the insur-
ment style effects may be mediated has been considered in ance company supporting this research conducts annual
marketing. For example, Thomson and Johnson (2006)— interviews with its customers about their financial service
using data on students’ general attachment styles—show an needs; it now administers our scales in these interviews.
Second, attachment styles help firms identify which cus- generate stable revenues (because increases and decreases
tomers prefer high or low levels of closeness. Our concep- in relationship breadth are mitigated). Stable revenue
tualization of closeness helps managers tailor the interac- streams (even if they are lower) can be desirable in a cus-
tional frequency, diversity, and the mutual influence of tomer portfolio (Tarasi et al. 2011).
customer–firm interactions to match customer preferences. Fifth, managers should understand that anxious customers
For example, anxious customers welcome more frequent may be more sensitive to relational cues. For example, anx-
contacts and a variety of touch points with the firm (e.g., ious customers are more likely to respond positively to
interactive websites, newsletters, phone calls, face-to-face being recognized by a loyalty program and equally sensitive
meetings). Such closeness-enhancing marketing activities to losing status (e.g., a downgrade). Similarly, attachment
are not equally effective for avoidant customers. Third, anxiety can lead customers to overreact to critical incidents,
firms should examine how preferences for closeness and such as a service failure or being “dropped” by a firm (e.g.,
attachment styles are related to customer profitability. For because of being unprofitable). Here, anxious customers
example, the firm that sponsored this study improved the seem likely to (over)react with third-party complaints and
profitability of avoidant clients by reducing RM efforts tar- negative word of mouth.
geted at them (e.g., visits by the agent). Firms can maintain Sixth, this study enriches customer segmentation and RM
closeness with anxious customers using less expensive mar- approaches by accounting for distinct configurations of our
keting tools (e.g., birthday cards, e-mails, and phone calls outcome variables. Figure 1 shows a cross-tabulation of
are cheaper than personal visits). Thus, an attachment- customers’ preference for closeness and loyalty, based on a
informed approach to RM can improve firms’ effectiveness median split. These two variables can be closely associated
and efficiency. with each other (Cells 2 and 3 in Figure 1), but they are not
Fourth, attachment styles help firms identify customer always directly linked (Cells 1 and 4). A simple thought
segments that have the potential to be more or less loyal. experiment explains why. Customers may be loyal because
Customers with low attachment anxiety and avoidance of high satisfaction with a firm’s offering (e.g., network reli-
report the highest loyalty intentions, whereas avoidant cus- ability leads to loyalty to a cell phone provider). Yet these
tomers report the lowest. Moreover, high levels of attach- same customers do not necessarily desire frequent, diverse,
ment anxiety reduce cross-buying. Therefore, managers and influential interactions with the firm. In addition, some
could focus cross-selling efforts on customers of low anxi- customers may report low levels of loyalty because the firm
ety and avoidance for whom they can leverage higher loy- has not identified and appropriately addressed their prefer-
alty potentials. An attachment-informed firm could also use ences for closeness (Cell 4). Because customer loyalty
high attachment avoidance as an early indicator of loyalty- sometimes serves as a useful proxy for long-term profitabil-
averse customers. Moreover, anxious customers are likely to ity, Figure 1 also suggests that considering customer prefer-
Figure 1
ENRICHING CUSTOMER SEGMENTATION BY CROSSING CUSTOMER PREFERENCE FOR CLOSENESS AND LOYALTY TO THE FIRM
Notes: Customer preference for closeness and loyalty to the firm are not always strongly linked with each other. To illustrate this, by demonstrating that the
nondiagonal boxes in the matrix are nonzero, we calculated the number of customers in the cells on the basis of the predicted closeness scores and the pre-
!dicted (averaged) loyalty scores for every respondent. We chose the respective median of the two predicted variables as the cutoff point and cross-tabulated
the newly formed categorical variables.
ences for closeness can lead to a more nuanced understand- REFERENCES

ing of how resources should be allocated according to cus- Aaker, Jennifer L., Susan Fournier, and Adam S. Brasel (2004),
tomer lifetime value. “When Good Brands Do Bad,” Journal of Consumer Research,
Finally, understanding customer attachment styles and 31 (1), 1–16.
preferences for closeness can help with new customer Aiken, Leona S. and Stephen G. West (1991), Multiple Regression:
acquisition. Many firms strive to attract new customers by Testing and Interpreting Interactions. Newbury Park, CA: Sage
promising to be close to them (e.g., Darden’s Olive Garden Publications.
Aksoy, Lerzan, Timothy L. Keiningham, Alexander Buoye, Bart
restaurants advertise “When you’re here, you’re family”). Larivière, Luke Williams, and Ian Wilson (2012), “Does Loy-
Such indiscriminate acquisition approaches (falsely) pre- alty Span Domains? Examining the Relationship Between Con-
sume that all customers desire closeness, but they might sumer Loyalty and Other Loyalties and Its Influence on Happi-
backfire for avoidant customers. We believe that firms that ness,” working paper, Gabelli School of Business, Fordham
can tailor marketing efforts to match customer relational University.
preferences attract a broader set of new clients. Barnes, James G. (1997), “Closeness, Strength, and Satisfaction:
Examining the Nature of Relationships Between Providers of
Limitations and Further Research Financial Services and Their Customers,” Psychology and Mar-
keting, 14 (8), 765–90.
This research is not without limitations. First, we col- Bergkvist, Lars and John R. Rossiter (2007), “The Predictive
lected data from customers from a single insurance firm. Validity of Multiple-Item Versus Single-Item Measures of the
Further research should examine other insurance contexts Same Constructs,” Journal of Marketing Research, 44 (May),
(e.g., health insurance) and additional industries to assess 175–84.
the extent to which our results are generalizable and to iden- Berscheid, Ellen, Mark Snyder, and Allen M. Omoto (1989), “The
tify relevant moderating variables. Second, our data, while Relationship Closeness Inventory: Assessing the Closeness of
extensive, were not without limitations. For example, the Interpersonal Relationships,” Journal of Personality and Social
Psychology, 57 (5), 792–807.
data did not allow us to account for situational triggers in ———, ———, and ——— (2004), “Measuring Closeness: The
customers’ lives (e.g., purchasing a home, having children) Relationship Closeness Inventory (RCI) Revisited,” in Hand-
(Gustafsson, Johnson, and Roos 2005). Accounting for such book of Closeness and Intimacy, Debra J. Mashek and Arthur P.
situational triggers is an important extension of our frame- Aron, eds. Mahwah, NJ: Lawrence Erlbaum Associates,
work.4 In addition, we could not track 224 of the customers 81–101.
in our sample over time because of technical limitations in Bolton, Ruth N. and Katherine N. Lemon (1999), “A Dynamic
the firm’s database. However, we accounted for a potential Model of Customers’ Usage of Services: Usage as an Antecedent
and Consequence of Satisfaction,” Journal of Marketing Research,
selection bias resulting from these missing clients by esti-
36 (May), 171–86.
mating a Heckman-type selection model (see Web Appen- ———, ———, and Peter C. Verhoef (2004), “The Theoretical
dix B at www.marketingpower.com/jmr_webappendix). Underpinnings of Customer Asset Management: A Framework
This selection model produced highly similar results to the and Propositions for Future Research,” Journal of the Academy
ones discussed here. (The only hypothesized effect that was of Marketing Science, 32 (3), 271–93.
not replicated was the marginally significant attachment ¥ ———, ———, and ——— (2008), “Expanding Business-to-Business
avoidance interaction in the behavioral loyalty model.) Customer Relationships: Modeling the Customer’s Upgrade
Despite this slight deviation, we believe that these addi- Decision,” Journal of Marketing, 72 (January), 46–64.
Bougie, Roger, Rik Pieters, and Marcel Zeelenberg (2003),
tional analyses support the robustness of our framework and “Angry Customers Don’t Come Back, They Get Back,” Journal
hypothesized effects. Third, we did not track how customer of the Academy of Marketing Science, 31 (4), 377–93.
attachment styles vary over time and across attachment tar- Brennan, Kelly A., Catherine L. Clark, and Phillip R. Shaver (1998),
gets. Interpersonal attachment styles are relatively stable “Self-Report Measurement of Adult Romantic Attachment,” in
over months and years, but they can be influenced by a part- Attachment Theory and Close Relationships, J.A. Simpson and
ner’s actions (Mikulincer and Shaver 2007). Similarly, S.W. Rholes, eds. New York: Guilford Press, 46–76.
attachment styles can also vary across concrete and abstract Cate, Rodney M., Lauren A. Levin, and Lucinda S. Richmond
(2002), “Premarital Relationship Stability: A Review of Recent
attachment targets (e.g., firm vs. employee) (Aksoy et al. Research,” Journal of Social and Personal Relationships, 19
2012; Mende and Bolton 2011). Thus, rather than viewing (2), 261–84.
customer attachment styles as a given, marketers should Collins, Nancy L. and Brooke C. Feeney (2004), “Working Mod-
examine attachment anxiety and avoidance over time to els of Attachment Shape Perceptions of Social Support: Evi-
understand their interplay with firms’ marketing activities dence from Experimental and Observational Studies,” Journal
and examine distinctive attachment styles related to specific of Personality and Social Psychology, 87 (3), 363–83.
attachment targets. Finally, we used formative measures to Cooil, Bruce, Timothy L. Keiningham, Lerzan Aksoy, and Michael
Hsu (2007), “A Longitudinal Analysis of Customer Satisfaction
capture customer preference for closeness. Additional
and Share of Wallet: Investigating the Moderating Effect of
research might investigate the different dimensions of close- Customer Characteristics,” Journal of Marketing, 71 (January),
ness by developing a scale that has multiple reflective meas- 67–83.
ures of frequency, diversity, and influence. Such a scale Coulter, Robin A., Linda L. Price, and Lawrence Feick (2003),
would also lend researchers a deeper investigation into the “Rethinking the Origins of Involvement and Brand Commit-
antecedents of customers’ preference for closeness beyond ment,” Journal of Consumer Research, 30 (2), 151–69.
customer attachment styles. Crosby, Lawrence A., Kenneth R. Evans, and Deborah Cowles
(1990), “Relationship Quality in Services Selling: An Interper-
sonal Influence Perspective,” Journal of Marketing, 54 (July),
4We are grateful to the two anonymous reviewers for these suggestions. 68–81.
Curran, Patrick J., Stephen G. West, and John F. Finch (1994), October 5, 2012), [available at http://www.afhayes.com/spss-
“The Robustness of Test Statistics to Nonnormality and Specifi- sas-and-mplus-macros-and-code.html#medcurve].
cation Error in Confirmatory Factor Analysis,” Psychological Hazan, Cynthia and Phillip R. Shaver (1994), “Attachment as an
Methods, 1 (1), 16–29. Organizational Framework on Close Relationships,” Psycho-
Dabholkar, Pratibha A. (1996), “Consumer Evaluations of New logical Inquiry, 5 (1), 1–22.
Technology-Based Self-Service Options,” International Journal Hira, Tahira K. and Olive M. Mugenda (2000), “Gender Differ-
of Research in Marketing, 13 (1), 29–51. ences in Financial Perceptions, Behaviors and Satisfaction,”
Dagger, Tracey, Peter J. Danaher, and Brian Gibbs (2009), “How Journal of Financial Planning, 13 (2), 86–92.
Often Versus How Long: The Interplay of Contact Frequency Inman, J. Jeffrey, James S. Dyer, and Jianmin Jia (1997), “A Gen-
and Relationship Duration in Customer Perceptions of Service eralized Utility Theory Model of Disappointment and Regret
Relationship Strength,” Journal of Service Research, 11 (May), Effects on Post-Choice Valuation,” Marketing Science, 16 (2),
371–88. 97–111.
Davila, Joanne and Thomas N. Bradbury (2001) “Attachment Inse- ———, Russell S. Winer, and Rosellina Ferraro (2009), “The Inter-
curity and the Distinction Between Unhappy Spouses Who Do play Among Category Characteristics, Customer Characteris-
and Do Not Divorce,” Journal of Family Psychology, 15 (3), tics, and Customer Activities on In-Store Decision Making,”
371–93. Journal of Marketing, 73 (September), 19–29.
De Wulf, Kristof, Gaby Odekerken-Schröder, and Dawn Iacobucci Iyengar, Raghuram and Sunil Gupta (2006), “Advanced Regres-
(2001), “Investments in Consumer Relationships: A Cross- sion Models,” in The Handbook of Marketing Research, Rajiv
Country and Cross-Industry Exploration,” Journal of Market- Grover and Marco Vriens, eds. Thousand Oaks, CA: Sage Pub-
ing, 65 (October), 33–50. lications, 267–87.
Diamantopoulos, Adamantios and Heidi M. Winklhofer (2001), Johnson, Julie T., Hiram C. Barksdale Jr., and James S. Boles
“Index Construction with Formative Indicators: An Alternative (2001), “The Strategic Role of the Salesperson in Reducing
to Scale Development,” Journal of Marketing Research, 38 Customer Defection in Business Relationships,” Journal of Per-
(May), 269–77. sonal Selling & Sales Management, 21 (2), 123–34.
Doney, Patricia M. and Joseph P. Cannon (1997), “An Examina- Keller, Kevin L. and Donald R. Lehmann (2006), “Brands and
tion of the Nature of Trust in Buyer–Seller Relationships,” Jour- Branding: Research Findings and Future Priorities,” Marketing
nal of Marketing, 61 (April), 35–51. Science, 25 (6), 740–59.
Edelstein, Robin S. and Phillip R. Shaver (2004), “Avoidant Kelley, Harold H., Ellen Berscheid, Andrew Christensen, John
Attachment: Exploration of an Oxymoron,” in Handbook of Harvey, Ted L. Huston, George Levinger, et al. (1983), Close
Closeness and Intimacy, Debra J. Mashek and Arthur Aron, eds.
Relationships. San Francisco: Freeman.
Mahwah, NJ: Lawrence Erlbaum Associates, 397–412.
Kirkpatrick, Lee A. and Keith E. Davis (1994), “Attachment Style,
Feeney, Judith A. and Patricia Noller (1990), “Attachment Style as
Gender, and Relationship Stability,” Journal of Personality and
a Predictor of Adult Romantic Relationships,” Journal of Per-
Social Psychology, 66 (3), 502–512.
sonality and Social Psychology, 58 (2), 281–91.
Klohnen, Eva C., Joshua A. Weller, Shanhong Luo, and Mary
Fornell, Claes and David F. Larcker (1981), “Evaluating Structural
Choe (2005), “Organization and Predictive Power of General
Equation Models with Unobservable Variables and Measure-
and Relationship-Specific Attachment Models,” Personality and
ment Error,” Journal of Marketing Research, 18 (February),
Social Psychology Bulletin, 31 (12), 1665–82.
39–50.
Fraley, R. Chris, Neils G. Waller, and Kelly A. Brennan (2000), Kumar, V., George Morris, and Joseph Pancras (2008), “Cross-
“An Item Response Theory Analysis of Self-Report Measures Buying in Retailing: Drivers and Consequences,” Journal of
of Adult Attachment,” Journal of Personality and Social Psy- Retailing, 84 (1), 15–25.
chology, 78 (2), 350–65. ——— and Denish Shah (2009), “Expanding the Role of Market-
Garbarino, Ellen and Mark S. Johnson (1999), “The Different ing: From Customer Equity to Market Capitalization,” Journal
Roles of Satisfaction, Trust, and Commitment in Customer of Marketing, 73 (November), 119–36.
Relationships,” Journal of Marketing, 63 (April), 70–87. La Guardia, Jennifer G., Richard M. Ryan, Charles E. Couchman,
Godfrey, Andrea, Kathleen Seiders, and Glenn B. Voss (2011), and Edward L. Deci (2000), “Within-Person Variation in Secu-
“Enough Is Enough! The Fine Line in Executing Multichannel rity of Attachment: A Self-Determination Theory Perspective on
Relational Communication,” Journal of Marketing, 75 (July), Attachment, Need Fulfillment, and Well-Being,” Journal of
94–109. Personality and Social Psychology, 79 (3), 367–84.
Greguras, Gary J. and James M. Diefendorff (2009), “Different Fits Leeflang, Peter S.H., Dick R. Wittink, Michel Wedel, and Philippe
Satisfy Different Needs: Linking Person–Environment Fit to A. Naert (2000), Building Models for Marketing Decisions.
Employee Attitudes and Performance Using Self-Determination Boston: Kluwer Academic Publishers.
Theory,” Journal of Applied Psychology, 94 (2), 465–77. Lewandowski, Gary W. and Robert A. Ackerman (2006), “Some-
Gruen, Thomas W., John O. Summers, and Frank Acito (2000), thing Is Missing: Need Fulfillment and Self-Expansion as Pre-
“Relationship Marketing Activities, Commitment, and Member- dictors of Susceptibility to Infidelity,” The Journal of Social
ship Behaviors in Professional Associations,” Journal of Mar- Psychology, 146 (4), 389–403.
keting, 64 (July), 34–49. Lilien, Gary L., Philip Kotler, and K. Sridhar Moorthy (1992),
Gustafsson, Anders, Michael D. Johnson, and Inger Roos (2005), Marketing Models. Englewood Cliffs, NJ: Prentice Hall.
“The Effects of Customer Satisfaction, Relationship Commit- Long, Scott J. and Jeremy Freese (2006), Regression Models for
ment Dimensions and Triggers on Customer Retention,” Jour- Categorical Dependent Variables Using Stata, 2d ed. College
nal of Marketing, 69 (October), 210–18. Station, TX: Stata Press.
Hair, Joseph F., William C. Black, Barry J. Babin, and Rolph E. Mende, Martin and Ruth N. Bolton (2011), “Why Attachment
Anderson (2010), Multivariate Data Analysis, 7th ed. Upper Security Matters: How Customers’ Attachment Styles Influence
Saddle River, NJ: Prentice Hall. Their Relationships with Service Firms and Service Employ-
Hayes, Andrew F. and Kristopher J. Preacher (2011), “Indirect and ees,” Journal of Service Research, 14 (3), 285–301.
Direct Effects of a Multicategorical Causal Agent in Statistical Mikulincer, Mario and Phillip R. Shaver (2007), Attachment in
Mediation Analysis,” working paper, School of Communication Adulthood: Structure, Dynamics, and Change. New York: Guil-
and Department of Psychology, Ohio State University, (accessed ford Press.
Mittal, Vikas and Wagner A. Kamakura (2001), “Satisfaction and Seiders, Kathleen, Glenn B. Voss, Dhruv Grewal, and Andrea L.
Repurchase Behavior: The Moderating Influence of Customer Godfrey (2005), “Do Satisfied Customers Really Buy More?
and Market Characteristics,” Journal of Marketing Research, 38 Examining Moderating Influences in a Retailing Context,”
(February), 131–42. Journal of Marketing, 69 (October), 26–43.
Mohr, Jakki J., Robert J. Fisher, and John R. Nevin (1996), “Col- Shah, Denish, V. Kumar, Yingge Qu, and Sylia Chen (2012),
laborative Communication in Interfirm Relationships: Moderat- “Unprofitable Cross-Buying: Evidence from Consumer and
ing Effects of Integration and Control,” Journal of Marketing, Business Markets,” Journal of Marketing, 76 (May), 78–95.
60 (July), 103–115. Simpson, Jeffry A. (1990), “Influence of Attachment Styles on
Nunnally, Jum C. and Ira H. Bernstein (1994), Psychometric Romantic Relationships,” Journal of Personality and Social
Theory, 3d ed. New York: McGraw-Hill. Psychology, 59 (5), 971–80.
O’Brien, Robert M. (2007), “A Caution Regarding Rules of Smith, Amy K. and Ruth N. Bolton (1998), “An Experimental
Thumb for Variance Inflation Factors,” Quality & Quantity, 41 Investigation of Service Failure and Recovery: Paradox or
(5), 673–90. Peril?” Journal of Service Research, 1 (1), 65–81.
Oliver, Richard L. (1999), “Whence Consumer Loyalty?” Journal Swaminathan, Vanitha, Karen Stilley, and Rohini Ahluwalia
of Marketing, 63 (Special Issue), 33–44. (2009), “When Brand Personality Matters: The Moderating
Palmatier, Robert W. (2008), Relationship Marketing, Marketing Role of Attachment Styles,” Journal of Consumer Research, 35
Science Institute Relevant Knowledge Series. Cambridge, MA: (6), 985–1002.
Marketing Science Institute. Tarasi, Crina, Ruth N. Bolton, Michael D. Hutt, and Beth A.
———, Rajiv P. Dant, Dhruv Grewal, and Kenneth R. Evans Walker (2011), “Balancing Risk and Return in a Customer Port-
(2006), “Factors Influencing the Effectiveness of Relationship folio,” Journal of Marketing, 75 (May), 1–17.
Marketing: A Meta-Analysis,” Journal of Marketing, 70 (Octo- Thomson, Matthew (2006), “Human Brands: Investigating
ber), 136–53. Antecedents to Consumers’ Strong Attachments to Celebrities,”
Paulssen, Marcel (2009), “Attachment Orientations in Business- Journal of Marketing, 70 (July), 104–119.
to-Business Relationships,” Psychology and Marketing, 26 (6), ——— and Allison Johnson (2006), “Marketplace and Personal
507–533. Space: Investigating the Differential Effects of Attachment Style
Preacher, Kristopher J. and Andrew F. Hayes (2008), “Asymptotic Across Relationship Contexts,” Psychology and Marketing, 23
and Resampling Strategies for Assessing and Comparing Indi- (8), 711–26.
rect Effects in Multiple Mediator Models,” Behavior Research Tokman, Mert, Lenita M. Davis, and Katherine N. Lemon (2007),
Methods, 40 (3), 879–91. “The WOW Factor: Creating Value Through Win-Back Offers
Reibstein, David, George Day, and Jerry (Yoram) Wind (2009), “Is to Reacquire Lost Customers,” Journal of Retailing, 83 (1),
Marketing Academia Losing Its Way?” Journal of Marketing, 47–64.
73 (July), 1–3. Verhoef, Peter C. (2003), “Understanding the Effect of Customer
Roodman, David (2011), “Fitting Fully Observed Recursive Relationship Management Efforts on Customer Retention and
Mixed-Process Models with cmp,” The Stata Journal, 11 (2), Customer Share Development,” Journal of Marketing, 67
159–206. (October), 30–45.
Salazar, Maria T., Tina Harrison, and Jake Ansell (2007), “An ———, Fred Langerak, and Bas Donkers (2007), “Understanding
Approach for the Identification of Cross-Sell and Up-Sell Brand and Dealer Retention in the New Car Market,” Journal of
Opportunities Using a Financial Services Customer Database,” Retailing, 83 (1), 97–113.
Journal of Financial Services Marketing, 12 (2), 115–31.

Decoding Customer-Firm Relationships - Mende Et Al

Uploaded by

Copyright:

Available Formats

Decoding Customer-Firm Relationships - Mende Et Al

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Decoding Customer-Firm Relationships - Mende Et Al

Uploaded by

Copyright:

Available Formats

MARTIN MENDE, RUTH N.

BOLTON, and MARY JO BITNER*

Keywords: attachment styles, services, relationship marketing, loyalty,

Decoding Customer–Firm Relationships:

The field of marketing “is supposed to be concerned

© 2013, American Marketing Association Journal of Marketing Research

Construct Measures Scale Statistics Source

Construct Measures Source

aThis category served as the dummy reference category.

Variable Predicted Sign Unstandardized Coefficient (SE) Standardized Coefficient

Variable Predicted Sign Unstandardized Coefficient (SE) Standardized Coefficient

ences for closeness can lead to a more nuanced understand- REFERENCES

You might also like