Factors Influencing the Effectiveness of Relationship Marketing: A Meta-Analysis

Robert W. Palmatier
Assistant Professor
University of Cincinnati
PO Box 210145
Cincinnati, OH 45221
Voice: 513.556.3935
Facsimile: 513.556.0979
Email: rob.palmatier@uc.edu

Rajiv P. Dant
Associate Professor of Marketing
University of South Florida
College of Business Administration
4202 E. Fowler Avenue, BSN 3403
Tampa, FL 33620
Voice: 813.974.6227 or 5978
Facsimile: 813.974.6175
Email: rdant@coba.usf.edu

Dhruv Grewal
Toyota Professor of Marketing
Babson College
Malloy Hall
Babson Park, MA 02457
Voice: 781.239.3902
Email: dgrewal@babson.edu

Kenneth R. Evans
Professor of Marketing & Associate Dean
110B Cornell Hall, University of Missouri
Columbia, MO 65211
Voice: 573.882.1039
Email: evans@missouri.edu

January 13, 2005

Prepared for:
Journal of Marketing

The authors thank the Marketing Science Institute for its financial support of this research. They
also thank Steven P. Brown and Kent B. Monroe for their assistance on this project, as well as
the editor and reviewers for their constructive comments.
 Factors Influencing the Effectiveness of Relationship Marketing: A Meta-Analysis

Abstract
Relationship marketing (RM) has emerged as one of the dominant mantras in business

strategy circles, though RM investigations often yield mixed results. To help managers and

researchers improve the effectiveness of their efforts, the authors synthesize RM empirical

research in a meta-analytic framework. Although the fundamental premise that RM positively

affects performance is well supported, many of the authors’ findings have significant

implications for research and practice. Relationship investment has a large direct effect on seller

objective performance, which implies that additional meditated pathways may explain the impact

of RM on performance. Objective performance is influenced most by relationship quality (a

composite measure of relationship strength) and least by commitment. The results suggest also

that RM is more effective when relationships are more critical to customers (e.g., service

offerings, channel exchanges, business markets) and built with an individual person rather than a

selling firm (which partially explains the mixed effects between RM and performance reported in

previous studies).

1
 Introduction

Relationship marketing (RM), in both business practice and as a focus of academic

research, has “experienced explosive growth” in the past decade (Srinivasan and Moorman

2005), where RM has been defined as “all marketing activities directed towards establishing,

developing, and maintaining successful relational exchanges” (Morgan and Hunt 1994, p. 22).

Most research and practice assumes that RM efforts generate stronger customer relationships that

enhance seller performance outcomes, including sales growth, share, and profits (Crosby, Evans,

and Cowles 1990; Morgan and Hunt 1994), but some business executives have been

disappointed in the effectiveness of their RM efforts (Colgate and Danaher 2000). Researchers

also have suggested that in certain situations, RM may have a negative impact on performance

(De Wulf, Odekerken-Schröder, and Iacobucci 2001; Hibbard et al. 2001).

Overall, these findings suggest that the effectiveness of RM efforts may vary depending

on the specific RM strategy and exchange context; this inconsistency with regard to performance

suggests the need for a meta-analysis to integrate the abundance of accumulated empirical

research and better understand the RM strategies that are most effective for building strong

relationships, the outcomes most affected by customer relationships, and the conditions in which

RM is most effective for generating positive seller outcomes. Advancing understanding of the

primary drivers of RM effectiveness can increase the return on firms’ RM investments

dramatically and provide researchers with insights into ways to build more comprehensive

models of the influence of RM on performance (Reinartz and Kumar 2003).

Using Dwyer, Schurr, and Oh’s (1987) seminal article on relationships; Crosby, Evans,

and Cowles’s (1990) introduction of relationship quality; and Morgan and Hunt’s (1994) key

mediating variable theory of RM, most research has conceptualized the effects of RM on

outcomes as fully mediated by one or more of the relational constructs of trust, commitment,

2
relationship satisfaction, and/or relationship quality. The existing literature offers a wide range of

antecedents for these relational mediators, and researchers disagree about which best captures the

characteristics of a relational exchange that influence performance. For example, Morgan and

Hunt (1994) propose that trust and commitment are both key to predicting exchange

performance, whereas others suggest that either trust (e.g., Doney and Cannon 1997;

Sirdeshmukh, Singh, and Sabol 2002) or commitment (e.g., Anderson and Weitz 1992; Gruen,

Summers, and Acito 2000; Jap and Ganesan 2000) alone is the critical relational construct.

Another school of thought suggests that the global construct of relationship quality, as

reflected by a combination of commitment, trust, and relationship satisfaction, offers the best

assessment of relationship strength and provides the most insight into exchange performance

(e.g., De Wulf, Odekerken-Schröder, and Iacobucci 2001; Kumar, Scheer, and Steenkamp 1995).

These different relational mediators have been linked empirically to many antecedents and

outcomes, which leads to the critical question: How does the relational mediated model vary

across different relational perspectives?

In this article, we systematically review and analyze the literature on relational mediators

in a meta-analytic framework (Figure 1) to provide insight into four research questions: (1)

which RM strategies are most effective for building customer relationships, (2) what outcomes

are most affected by customer relationships, (3) which moderators are most effective in

influencing relationship–outcome linkages, and (4) how does the RM strategy → mediator →

outcome linkage vary across different mediators?

—Insert Figure 1 about here—

Conceptual Framework

In reviewing the literature pertaining to relational mediators, we have identified many

constructs with similar definitions that operate under different aliases and constructs with similar

3
names but different operationalizations. Thus, we use a single construct definition (see Table 1)

to code existing research and include a construct in the conceptual framework only if at least 10

effects emerge to support its empirical analysis. Of the many constructs investigated, only 18 met

these criteria and appear in the model. Our nomological placement of each construct is driven by

both theory and the frequency of placement in extant research. Of those studies that include

hypothesized relationships with relational mediators, more than 90% are consistent with the

causal ordering of constructs in our framework, with the exceptions of conflict and cooperation,

which agree with our nomological framework in approximately 70% of extant studies.

—Insert Table 1 about here—

Although a relationship is, by its very nature, two sided, and though both parties typically

share in the benefits of a strong relationship, some antecedents and outcomes may have

differential effects according to the measurement perspective (e.g., dependence). Thus, we adopt

terminology to identify the perspective of each construct relative to its relational mediators. In

this framework, “seller” refers to the party that implements the RM effort in the hope of

strengthening its relationship with the “customer,” and the relational mediator captures the

customer’s perception of its relationship with the seller. For clarity and consistency, we use these

customer and seller perspectives even when the two parties may not be engaged in a typical

exchange transaction (e.g., a strategic alliance). Thus, we classify antecedents and outcomes as

“customer focal” when they share the same perspective as the relational mediator and “seller

focal” when they adopt a perspective opposite of that of the evaluation of the relational mediator.

We develop our conceptual framework in four parts, which roughly parallel our research

questions, by first reviewing the literature on relational mediators, then investigating the

antecedents and outcomes of these mediators, and finally studying potential moderators of the

impact of relational mediators on outcomes.

4
Relational Mediators

Successful RM efforts improve customer loyalty and firm performance through stronger

relational bonds (e.g., De Wulf, Odekerken-Schröder, and Iacobucci 2001; Sirdeshmukh, Singh,

and Sabol 2002), but the literature offers varied perspectives on which relational constructs

mediate the effects of RM efforts on outcomes. Commitment and trust are most often studied,

where commitment is “an enduring desire to maintain a valued relationship” (Moorman, Zaltman,

and Deshpandé 1992, p. 316) and trust is “confidence in an exchange partner’s reliability and

integrity” (Morgan and Hunt 1994, p. 23). Another relationship mediator, relationship

satisfaction, is a customer’s affective or emotional state toward a relationship. Relationship

satisfaction reflects exclusively the customer’s satisfaction with the relationship and differs from

the customer’s satisfaction with the overall exchange. Other researchers have suggested that

these mediators are merely indicators of the global mediator relationship quality, which is “an

overall assessment of the strength of a relationship” conceptualized as a multidimensional

construct that captures the many different facets of an exchange relationship (De Wulf,

Odekerken-Schröder, and Iacobucci 2001, p. 36; see also Crosby, Evans, and Cowles 1990). Its

structure and underlying dimensions vary across empirical studies, but central to the

conceptualization is the belief that no single dimension or relational construct can fully define

the “overall depth or climate” of an exchange relationship (Johnson 1999, p. 6).

Thus, whereas the literature consistently conceptualizes a mediating model for the effects

of RM on performance, the specific relational mediator(s) or composite of mediators appears to

be driven mainly by researcher discretion; empirical comparisons of the differential effects of

these relational mediators are noticeably absent. For example, some propose trust as the critical

relational mediator: Berry (1996, p. 42) offers “trust as perhaps the single most powerful

relationship marketing tool available to a company,” and Spekman (1988, p. 79) suggests that

5
trust is the “cornerstone” of long-term relationships. Alternatively, Gundlach, Achrol, and

Mentzer (1995, p. 78) propose commitment as the “essential ingredient for successful long-term

relationships,” and Morgan and Hunt (1994, p. 23) suggest “commitment among exchange

partners as key to achieving valuable outcomes.” More recently, De Wulf, Odekerken-Schröder,

and Iacobucci (2001) prefer relationship quality to any specific component. In summary, there is

little agreement among researchers as to which individual or composite relational mediator best

captures the key aspect(s) of a relationship that most affects outcomes. To address this issue

empirically, our meta-analytic framework compares the relative effects of the different

perspectives by analyzing relational mediators separately and as a group.

Antecedents to Relational Mediators

Customer-Focal Antecedents. Customers may perceive value in a relationship when they

receive relationship benefits from an exchange partner (e.g., time savings, convenience,

companionship), which increases their willingness to develop relational bonds. Relationship

benefits also have been shown to affect relational mediators positively (Morgan and Hunt 1994;

Reynolds and Beatty 1999). In addition, dependence on the seller reflects the customer’s

evaluation of the value of those seller-provided resources for which few alternatives are available

(Hibbard, Kumar, and Stern 2001). The literature is mixed regarding the effect of a customer’s

dependence, or relative dependence (i.e., customer’s dependence reduced by the seller’s

dependence), on relational mediators. Researchers find empirical support for both positive and

negative influences of relative dependence on relational mediators (Anderson and Weitz 1989;

Morgan and Hunt 1994), which suggests its impact may be contingent on the context.

Seller-Focal Antecedents. Researchers have investigated various RM strategies that

sellers can employ to strengthen relationships. Relationship investment refers to the time, effort,

and resources sellers invest in building stronger relationships. Such investments often generate

6
expectations of reciprocation that can help strengthen and maintain a relationship and therefore

positively influence relational mediators (Anderson and Weitz 1989; Ganesan 1994). Seller

expertise reflects the knowledge, experience, and overall competence of the seller. When

customers interact with a competent seller, they receive increased value, their relationship

becomes more important, and they invest more effort to strengthen and maintain it (Crosby,

Evans, and Cowles 1990; Lagace, Dahlstrom, and Gassenheimer 1991).

Dyadic Antecedents. Customer- and seller-focal antecedents are meaningful from one

side of the exchange dyad, but other antecedents require the active involvement of both exchange

partners and are equally meaningful from both perspectives. For example, communication, or the

amount, frequency, and quality of information shared between exchange partners (Mohr, Fisher,

and Nevin 1996), requires both parties to exchange information. Communication builds stronger

relationships in an exchange by helping resolve disputes, align goals, and uncover new value-

creating opportunities (Morgan and Hunt 1994). Similarity is the commonality in appearance,

lifestyle, and status between individual boundary spanners or the similar cultures, values, and

goals between organizations. Such similarities between people or organizations may provide

cues that the exchange partner will help facilitate important goals and has been shown to affect

relational mediators positively (Crosby, Evans, and Cowles 1990; Doney and Cannon 1997).

Relationship duration is the length of time the relationship between the exchange partners has

existed, whereas interaction frequency refers to the number of interactions per unit of time

between partners. Both provide trading partners with more behavioral information in varied

contexts, which allows for better predictions that should increase each party’s confidence in its

partner’s behavior (Anderson and Weitz 1989; Doney and Cannon 1997). Finally, conflict entails

the overall level of disagreement between exchange partners, often termed “perceived” or

“manifest” conflict. As conflict increases, the customer is less likely to have confidence in the

7
long-term orientation of the seller or invest in building or maintaining a relationship; thus,

conflict should negatively influence the customer’s trust and commitment toward the seller

(Anderson and Weitz 1992).

Consequences of Relational Mediators

Customer-Focal Outcomes. Increased customer loyalty is one of the most common

outcomes expected from RM efforts, but loyalty has been defined and operationalized in many

different ways. An expectation of continuity reflects the customer’s intention to maintain the

relationship in the future and captures the likelihood of continued purchases. However,

researchers have criticized this measure of loyalty because customers with weak relational bonds

and little loyalty may report high continuity expectations as a result of their perceptions of high

switching costs or their lack of time to evaluate alternatives (Oliver 1999). Word of mouth

(WOM) captures the likelihood that a customer will refer a seller positively to another potential

customer and thereby indicates both attitudinal and behavioral dimensions of loyalty. Some

studies operationalize customer loyalty as a composite or multidimensional construct that

includes groupings of intentions, attitudes, and seller performance indicators. We note that

relational mediators positively influence global measures of customer loyalty, just as they do its

individual components (De Wulf, Odekerken-Schröder, and Iacobucci 2001; Sirdeshmukh,

Singh, and Sabol 2002).

Seller-Focal Outcomes. Possibly the most important outcome of RM efforts is seller

objective performance, which captures the seller’s actual performance enhancements, including

sales, profit, and share of wallet. Some researchers have found empirical support for the

influence of relational mediators on seller objective outcomes (e.g., Doney and Cannon 1997;

Siguaw, Simpson, and Baker 1998), but several other studies have failed to find any significant

8
effects, which suggests that the effect of RM on performance may be context dependent (Crosby,

Evans, and Cowles 1990; Gruen, Summers, and Acito 2000).

Dyadic Outcomes. Cooperation captures the level of coordinated and complementary

actions between exchange partners in their efforts to achieve mutual goals. Cooperation

promotes value creation beyond that which each party could achieve separately, but because one

party often receives its portion of the value earlier, the other party must have enough trust in the

relationship to wait for its future reciprocation. Researchers have shown that trust and

commitment between exchange partners are critical for cooperation (Anderson and Narus 1990;

Morgan and Hunt 1994).

Moderators of Relational Mediators’ Influence on Outcomes

The RM model (RM strategies → relational mediator → outcomes) we conceptualize

herein can be applied across many different contexts in which business strategies may have

varying effects. One objective of our meta-analysis therefore is to identify and empirically test

the influence of potential moderators on the linkages in the RM model.

Contexts Influencing Relationship Importance. Relationship marketing is based on the

premise that building strong relationships positively influences exchange outcomes, and

researchers recognize that exchanges vary across a spectrum from transactional to relational

(Anderson and Narus 1991). For exchanges in which relationships are more important, we expect

that the relational mediators will have a greater impact on outcomes, whereas in highly

transactional exchanges, the relationships between buyers and sellers may have little influence

on outcomes. Extant literature identifies three situations in which relationships may be more

important for the success of an exchange. First, services generally are perceived as less tangible,

less consistent, and more perishable, and customers and sellers are more involved in the

production and consumption of services than they are for products (Zeithaml, Parasuraman, and

9
Berry 1985). This closer interaction between customers and sellers may make customer–seller

relationships more critical for services, and the intangibility of the offering may make the

benefits of trust more critical because evaluations often are ambiguous.

Second, channel researchers tend to distinguish between channel partner exchanges and

direct seller–customer transactions. Exchanges between channel partners have higher levels of

interdependence, require coordinated action, and rely on the prevention of opportunistic behavior

(Anderson and Weitz 1989). Thus, coordination improvements and the reduction of opportunistic

behaviors through strong relationships should be more important in a channel context, which

should lead to a greater impact of relational mediators on performance compared with their

impact in direct exchanges.

Third, Anderson and Narus (2004, p. 21) differentiate consumer and business markets on

the basis of the importance of relationships, maintaining that a “firm’s success in business

markets depends directly on its working relationships.” If a working relationship is more critical

for a customer’s success in business markets, relationships should have a greater impact on

exchange outcomes in business than in consumer markets.

Individual Versus Organizational Relationships. Customers may form a relationship

with an individual boundary spanner in the selling organization and/or with the selling

organization as a whole. This issue of individual versus organizational relationships also has

significant managerial implications as firms continue to seek to increase their service efficiencies

through the use of technology (e.g., customer relationship management). Experimental research

shows that when evaluating another individual, people make stronger, quicker, and more

confident judgments than when they evaluate a group; those judgments also are more strongly

related to outcomes and behaviors (Hamilton and Sherman 1996). Accordingly, we expect

customers’ judgments about an individual boundary spanner, when based on relational

10
characteristics (e.g., trust in the salesperson), will be stronger, more confident, and more strongly

linked to outcomes than their judgments based on the relational characteristics of a selling firm

(e.g., trust in the firm). Post hoc findings support this premise, including those of Doney and

Cannon (1997, p. 45), who report that “the process by which trust develops appears to differ

when the target is an organization … as opposed to an individual salesperson,” and those of

Iacobucci and Ostrom (1996, p. 69), who find that “[i]ndividual-to-firm relationships [are] also

typically short-term and less intense in comparison to individual-level dyads.” Thus, the positive

effect of relational mediators on outcomes will be greater when the relational mediator is

targeted toward an individual member of the selling organization than toward the organization.

Method

Collection and Coding of Studies

The key impetus for RM research occurred after Dwyer, Schurr, and Oh’s (1987) seminal

article, so we searched empirical research for the mediators of interest during the period 1987–

2004. We employed various methods in our literature search, including (1) a search of the

ABI/Informs, PsycINFO, and Business Source Premier databases for each relational mediator;

(2) a search of the Social Sciences Citation Index using the seminal articles for these constructs;

(3) manual shelf searches of journals that contain research on relational mediators; and (4) e-

mails sent to researchers in the domain asking for their published and unpublished works. Our

search generated more than 100 published and unpublished papers, each of which we evaluated

for measures of the relationships among antecedents, outcomes, and the four relational

mediators. Because correlations were the most common metric included in these studies (>95%),

we e-mailed authors to request the correlation matrices for any studies in which they were not

provided. Two independent coders, not familiar with the study, used the definitions in Table 1 to

code the studies, and any differences (> 95% agreement) were resolved through discussion

11
(Szymanski and Henard 2001). When a single study provided more than one effect size estimate

for the same relationship, we calculated an average. In cases in which the multiple effect size

estimates from the same study were independent, we included them as separate effect size

estimates. This procedure prevents the bias that may occur due to multiple counts of dependent

effect size estimates and enables us to code moderators that vary across subsets of a sample in a

single study (e.g., Brown and Peterson 1993). Ultimately, we combined 637 correlations from

111 independent samples drawn from 94 different manuscripts, to yield a combined N of 38,077

with which to calculate the pairwise effect size estimates.1

Univariate Analyses

We began our analysis by adjusting our basic input measure, correlations (r), for

corrections due to measurement error (scale reliability differences); we report the correlation

adjusted for reliability (Hunter and Schmidt 1990). We then adjusted for sampling error (sample

size differences) and report the sample-weighted reliability adjusted r and its 95% confidence

intervals.2 We calculated the χ2 test (degrees of freedom [df] = 1) for association, then addressed

the file-drawer problem by computing the classical file drawer N (Rosenthal 1979) and the Q

statistic test of homogeneity (cf. Cheung and Chan 2004; Hunter and Schmidt 1990) for each

relationship.3

We performed such analyses for the influence of each antecedent on the four relational

mediators (providing there were four effects for each antecedent), which enables us to compare

the influence of each antecedent on each mediator. Although these mediators measure different

aspects of a relationship, researchers have argued that they are highly related and difficult to

distinguish and therefore can be combined into a composite construct (Crosby, Evans, and

Cowles 1990; De Wulf, Odekerken-Schröder, and Iacobucci 2001; Smith 1998). To facilitate the

12
comparison of relative effects among the different antecedents on the overall relationship, we

duplicate these analyses for the effects of each antecedent on all four mediators as a group.

Causal Model

In addition to the pairwise analyses, we aggregated the studies to test the nomological

causal model implicit in Figure 1. This multivariate technique has the advantage of analyzing all

linkages simultaneously, but it also needs significantly more data because the effects (i.e.,

correlation coefficients) must be available between each construct in the model and all other

constructs, not just the pairwise effects for constructs with proposed relationships (Brown and

Peterson 1993). Thus, causal models typically are limited to only the most commonly studied

constructs. We determined the average-adjusted intercorrelation among all constructs in the

framework whose required correlation coefficients were reported in three or more studies (Table

2). Furthermore, to increase the number of constructs that met this requirement and provide a

concise synthesis of the literature, we grouped all relational mediators together and thus propose

a fully mediated model (Morgan and Hunt 1994). Of the 14 antecedent and outcome constructs

included in Figure 1, only 6 met this criterion and could be evaluated in the causal model.

—Insert Table 2 about here—

Results

After we report the results of our causal model estimation procedure (Cheung and Chan

2005; Furlow and Beretvas 2005), we provide the results of the pairwise and casual model

analyses structured around our four focal research questions. Because the first two questions

focus on the effectiveness of antecedents that influence relational mediators (Table 3) and

relational mediators that influence outcomes (Table 4), we report the findings starting with the

most influential constructs and concentrate on the aggregate results for all mediators (i.e., last

row of each construct in Tables 3 and 4). Next, we report the results of the moderator analyses

13
(Table 5) to understand in which context RM is most effective. Finally, we concentrate on the

last research question, namely, how the RM mediating model varies across different mediators.

For this question, we no longer focus on the aggregate results but instead evaluate the effects that

pertain to each mediator and thereby provide insight into how the effects vary across mediators.

Causal Model Estimation

The fit indices from the structural model testing of the causal model indicate this model

fits the data poorly: χ2(5) = 322.27 (p < .01), comparative fit index (CFI) = .87, goodness-of-fit

index (GFI) = .95, and root mean square of approximation (RSMEA) = .19 (for an example of

this technique, see Brown and Peterson 1993). Modification indices suggest a revised causal

model that includes direct paths from dependence on seller and relationship investment to seller

objective performance. The revised model results in fit indices that indicate a good fit to the data:

χ2(3) = 12.15 (p < .01), CFI = .89, GFI = .99, and RSMEA = .04.4

—Insert Tables 3 and 4 about here—

Which RM Strategies Are Most Effective for Building Customer Relationships?

As we clearly show in Table 3, not all RM strategies (antecedents) are equally effective

for building relationships. The average sample-weighted, reliability-adjusted correlations among

antecedents and mediators is .41, and they range from .13 for relationship duration to the largest

absolute effect of –.67 for conflict. All paths from antecedents to relational mediators are

supported in the pairwise analyses, except for the path from interaction frequency to relationship

satisfaction. Most of these findings appear robust with regard to the number of null studies

needed to render the observed effects zero (mean file drawer N is 3,152). Only three linkages

appear susceptible to a file-drawer problem: interaction frequency → relationship satisfaction (N

= 2), relationship duration → relationship satisfaction (N = 18), and relationship duration →

14
relationship quality (N = 25). In the Q-statistic test for homogeneity, with one exception (seller

expertise → relationship satisfaction), all the tests for homogeneity are significant.

Several insights can be drawn from evaluating the relative impact of different RM

strategies on building strong customer relationships. Conflict (r = –.67) has the largest absolute

impact on the relational mediators of all antecedents, in support of the importance of resolving

problems and disagreements to prevent relationship-damaging conflicts (alternatively, the

presence of conflict may seriously undermine the effect of other RM antecedents). That the

largest effect is negative extends to the RM domain the finding that people pay more attention to

negatives than to positives (Shiv, Edell, and Payne 1997) and warrants further investigation. The

seller expertise (r = .62) and communication (r = .54) antecedents have the greatest positive

influence on relational mediators. The great impact of seller expertise suggests the importance of

training boundary spanners and the potential detriments of staffing call centers with

inexperienced or unskilled employees. The influence of seller expertise also seems to apply

across all four relational mediators, in support of Vargo and Lusch’s (2004, p. 3) claim that

“skills and knowledge are the fundamental unit of exchange,” such that sellers’ skill and

knowledge are the most important value-creating attributes. Similarly, the large positive effect of

communication on all mediators is consistent with its role in both uncovering value-creating

opportunities and resolving problems.

Relationship investment (r = .46), similarity (r = .44), and relationship benefits (r = .42)

are the next most influential RM strategies. The strong positive impact of the seller’s relationship

investments and customer relationship benefits indicates that managers should engage in

proactive RM spending. The importance of similarities between buyers and sellers suggests that

without common reference points, exchange partners may find it difficult to move the exchange

from a purely economic or transactional basis to a relational basis. The last three antecedents—

15
dependence on seller (r = .26), interaction frequency (r = .16), and relationship duration (r =

.13)—have notably smaller effects on relational mediators.

The causal model analysis generates the same rank order of relative effects of the

antecedents on relational mediators as the pairwise analyses, which increases our confidence in

the univariate results. Communication (β = .29, p < .01), relationship investment (β = .23, p <

.01), relationship benefit (β = .18, p < .01), and dependence on seller (β = .05, p < .01) all have

significant positive effects on relational mediators, but relationship duration (β = .02) fails to

influence relational mediators significantly in the multivariate analysis.

What Outcomes Are Most Affected by Customer Relationships?

In Table 4, we show that customer relationships do not equally influence all exchange

outcomes. The average correlations among relational mediators and outcomes is .55, ranging

from a low of .35 for seller objective performance to a high of .70 for cooperation. All paths

from relational mediators to outcomes are supported. None of these results appears susceptible to

a file-drawer problem; all paths would require more than 375 null studies to generate a zero

effect, with a mean file drawer N of 6,153. All the Q-statistic tests for homogeneity are

significant, demonstrating statistical heterogeneity and supporting a moderator analysis.

Relational mediators have the largest combined influence on the dyadic outcome of

cooperation (r = .70), followed by WOM (r = .61). This finding reinforces the importance of

relationship building for a high level of customer cooperation. The greater impact of relational

mediators on WOM (r = .61) than on the expectation of continuity (r = .56) or customer loyalty

(r = .52) lends support to Reichheld’s (2003, p. 48) premise that WOM may be the best indicator

of “intense loyalty.” Only customers who have strong relationships with sellers are willing to

risk their own reputation by giving a referral.

16
 Of the five outcomes, relational mediators have the least influence on seller objective

performance (r = .35). Thus, though customer relationships positively influence performance

outcomes, in support of efforts put into RM strategies, the actual effect on performance is lower

than that on the other four outcomes. This finding is not surprising; relational mediators are more

closely related to loyalty and cooperation than is objective performance, which often depends on

other, nonrelational factors (e.g., the economy).

The causal model includes only one outcome, but we confirm the significant influence of

relational mediators on seller objective performance (β = .16, p < .01). In addition, though the

impact of relationship benefits and communication strategies on seller objective performance are

fully mediated by the relational mediators, the influence of dependence and relationship

investment are only partially mediated; both dependence on seller (β = .22, p < .01) and

relationship investment (β = .34, p < .01) also have large direct effects on seller performance.

—Insert Table 5 about here—

Which Moderators Are Most Effective in Influencing Relationship–Outcome Linkages?

In Table 5, we present the influence of moderators on the linkage between relational

mediators and outcomes.5 The premise that customer relationships have a greater impact on

exchange outcomes in situations in which relationships are more critical to the success of the

exchange is supported for the impact of all mediators on customer loyalty among services,

channels, and business customers. The correlation of all mediators with customer loyalty is .58

for service versus .43 for product-based exchanges (p < .05), .65 for channel versus .46 for direct

interactions (p < .01), and .56 for business versus .46 for consumer markets (p < .05). We find a

similar effect in business markets for the impact of relationships on seller objective performance,

for which the influence of all mediators is r = .36 in business markets versus r = .25 in consumer

markets (p < .01). In summary, the significant moderation of the influence of relationships on

17
customer loyalty across services, channels, and business markets, as well as on performance in

business markets, provides support for our premise that customer relationships have a greater

impact on exchange outcomes in situations in which relationships are more critical to success.

Contrary to our expectations, relational mediators’ influence on the expectation of

continuity is greater in consumer than in business markets, mostly because of commitment’s

influence on the expectation of continuity (r = .46 business, r = .71 consumer, p < .05). Because

commitment taps a customer’s desire to maintain a valued relationship, whereas the expectation

of continuity captures a customer’s intent to maintain the relationship, consumers may be better

able to convert their attitude or desire into an intention than are business buyers because they

have a higher degree of control over their actions. Consistent with the theory of planned

behavior, the link between an attitude and an intention should be stronger as control increases

(cf. Ajzen and Fishbein 1980). Thus, the stronger impact of commitment on the expectation of

continuity (which results from higher levels of control) in consumer than in business markets

may offset the typically greater importance of relationships in business markets.

As we proposed, relationships have a greater impact on customer loyalty when the target

of the relationship is an individual person (r = .56) rather than an organization (r = .46) (p < .05).

Similarly, the impact of relational mediators on cooperation is greater (r = .68 for interpersonal, r

= .55 for interorganizational, p < .05) when the customer’s relationship is targeted toward a

person employed by the seller rather than toward the seller overall. We provide additional

support for this finding in Table 5, in which we show that of the 16 moderation tests, 81% are in

the expected direction, and the impact of all mediators on seller objective performance is

significant at the p < .10 level (r = .40 for interpersonal, r = .31 for interorganizational).

18
How Does the RM Strategy → Mediator → Outcome Linkage Vary Across Mediators?

The preceding research questions focus on the effects of the four mediators as a group.

Now we investigate the individual linkages to identify when mediators operate differently,

starting with the front half of the model: the RM strategy → relational mediator linkage (Table

3). The effectiveness of RM strategies varies across different relational mediators. We consider

the differential effects of relationship investments and benefits on mediators together because

they are logically related. Sellers’ relationship investments normally generate customer

relationship benefits, but in some cases, an investment may not be desired or generate any actual

benefit. Relationship investment has the least impact on commitment (r = .34) of all the

relational mediators (mean of other relational mediators, r = .52), with no overlap in the

confidence intervals (CI). Thus, sellers can strengthen their overall relationships through

investments (possibly by generating feelings of reciprocity), but the relative impact on customer

commitment is minimal. Alternatively, customer relationship benefits have the greatest impact

on customer commitment (r = .51), especially compared with customer trust (r = .33, no overlap

in CI), which suggests that customers value these benefits and want to maintain them. This

discrepancy may occur because many relationship investments do not generate value for the

customer and therefore do not lead to customer commitment. Although investments that do not

generate customer value may strengthen relationships by generating debts of reciprocity, they

will not necessarily generate an enduring desire to maintain a valued relationship.

As we might have expected, dependence has a greater positive effect on commitment (r =

.37) than the other mediators (mean of other relational mediators, r = .19, no overlap in CI),

which reflects customers’ desire to maintain a relationship with the seller on which they are

dependent. The relatively limited effect of dependence on customer trust (r = .21), may be due to

their concern that the seller will take advantage of their dependence.

19
 Although similarity often is hypothesized to influence trust by reducing uncertainty and

serving as a cue to facilitate goals, we find that similarity actually has a greater impact on

commitment (r = .63) than on trust (r = .41). This greater impact on commitment might be

explained by research on stereotype behaviors, which suggests that people want to strengthen

and maintain relationships with “in-group” members and that similarity is a proxy for customers’

perceptions of a seller’s fit with their in-group (Devine 1995).

The influence of interaction frequency on trust is relatively much greater (r = .30) than

that of the three other mediators (mean of other relational mediators, r = –.01, no overlap in CI).

As customers interact more frequently with sellers, they appear to gain more information about

their partner, which reduces their uncertainty about future behaviors and improves trust;

however, the frequency of their interaction has little effect on other relational mediators.

Now we turn our attention to the back half of the model, the relational mediator →

outcome linkage (Table 4), for which we find that relational mediators have differential effects

on most of the outcomes studied. Commitment (r = .58) has the greatest influence on customer

loyalty (mean of all other relational mediators, r = .47, no overlap of CI), as we might expect

from these two similar constructs.

Moreover, relationship quality has the greatest influence on objective performance (r =

.63), followed by trust (r = .35), relationship satisfaction (r = .32), and last by commitment (r =

.27), and the CIs of relationship quality, trust, and commitment do not overlap. These findings

suggest that RM researchers may need to take a multiple mediator or composite view when they

measure customers’ relationships to capture their impacts on objective performance. Different

dimensions of a relationship may be synergistic, and superior performance may be possible only

when the relationship is sufficiently strong on all critical aspects. Trust (r = .73) is most critical

for cooperation relative to the other mediators (mean of other relational mediators, r = .66, no

20
overlap of CI), in support of its role in coordinating actions among partners to create value and

achieve mutual outcomes.

Discussion

We provide evidence that the intervening role of relational mediators between RM

strategies and exchange outcomes is more complex than that suggested by extant research, but

the fundamental premise that RM and strong relationships positively affect performance is well

supported. Several of our findings offer important implications for improving the effectiveness

of RM research and practice (for a summary of key findings and implications, see Table 6).

—Insert Table 6 about here—

First, RM strategies/antecedents have a wide range of effectiveness in terms of generating

strong relationships, though specific strategies appear to be most effective for strengthening

specific aspects of a relationship. Overall, expertise and communication are the most effective

relationship-building strategies across all elements of a relationship, whereas the other strategies

often have differential effects across the different mediators. For example, generating

relationship benefits, promoting customer dependency, and increasing similarity to customers are

more effective strategies for increasing customer commitment than for building trust, whereas

relationship investment and interaction frequency do the opposite. Therefore, when comparing

the relative effectiveness of RM strategies, the results depend on the relational mediator

investigated. These findings suggest that RM may be improved by taking a more fine-grained

approach in which managers target RM strategies at specific relational weaknesses.

Second, we find that objective performance is influenced most by relationship quality (a

composite measure of relationship strength) and least by commitment, which supports a

multidimensional perspective of relationships in which no single or “best” relational mediator

can capture the full essence or depth of a customer–seller relationship (Hennig-Thurau,

21
Gwinner, and Gremler 2002; Johnson 1999). Previous research (Berry 1996; Doney and Cannon

1997; Spekman 1988) that offers either commitment or trust as the key, central, or cornerstone

relational mediator may focus too narrowly; a relationship may be truly effective only when most

or all of its key aspects are strong. Research that focuses only on commitment and generalizes

from its impact on customer intention or intermediate behaviors to its effect on seller

performance therefore may prove misleading. For example, commitment has the greatest effect

on customer loyalty and the smallest impact on objective performance.

Third, the large direct effects of dependence and relationship investment on seller

objective performance suggest that these antecedents influence performance through alternative,

mediated pathways. Dependence, though not very effective at building relationships, can

improve performance by increasing switching costs and barriers to exit, which may make it an

effective performance-enhancing strategy but not an effective RM strategy. However,

relationship investment both builds customer relationships and directly improves performance,

which suggests that the extant relational-mediated framework is not comprehensive and that

additional mediators (e.g., reciprocity) must be investigated to explain the impact of RM on

performance fully. The importance of capturing the direct effect of relationship investment (β =

.34) is reinforced by its greater impact on objective performance compared with the effect of the

relational mediators (β = .16) in the causal model.

Fourth, the findings that strong relationships appear more effective for building customer

loyalty and improving seller performance for (1) service versus product offerings, (2) channel

versus direct exchanges, and (3) business versus consumer markets lend support to the premise

that RM may be a more effective strategy in situations in which relationships are more critical.

This finding calls into question efforts by sellers to force RM strategies in contexts in which the

customer’s relational needs are unclear; it also may explain the less-than-desirable results of RM

22
on performance documented in previous studies (e.g., Reinartz and Kumar 2003). Because these

situational moderators are coarse proxies for customers’ relationship needs, RM effectiveness

likely varies across other factors that influence customers’ need for strong relationships. In turn,

researchers must take care when extending RM research to these different contexts.

Fifth, the results suggest that customer relationships have stronger effects on exchange

outcomes when their target is an individual person rather than a selling firm. Thus, RM

strategies focused on building interpersonal relationships between boundary spanners (e.g.,

dedicated salesperson, social entertaining) may be more effective than those focused on building

customer-to-firm relationships (e.g., team selling, frequency-driven loyalty programs). Social

psychology’s individual and group judgment theory (Hamilton and Sherman 1996, p. 336),

which posits “differences in the outcomes of impressions formed of individual and group targets,

even when those impressions are based on the very same behavioral information,” has several

implications for the marketing domain and may provide a parsimonious explanation for previous

marketing research (Doney and Cannon 1997; Iacobucci and Ostrom 1996). The post hoc finding

that conflict has a more negative impact (p < .01) on customer–firm relationships than on

customer–individual relationships also is consistent with this theory, because judgments about

persons are more resilient to disconfirming events than are judgments about groups (Hamilton

and Sherman 1996). Thus, managers may want to use boundary spanners or salespeople rather

then centralized service centers to resolve conflicts, because customers’ relationships with

salespeople may withstand conflict better than their relationships with selling firms.

Managerial Implications

Most promising for managers is that five of the strategies with the greatest impact are

either seller focal or dyadic, in support of the effectiveness of proactive relationship-building

strategies undertaken by sellers. Business executives focused on building and maintaining strong

23
customer relationships should note that the selection and training of boundary spanners is

critical; expertise, communication, and similarity to customers are the most effective

relationship-building strategies. The next most effective strategy is for managers to make

relationship investments and generate relationship-based benefits for customers; furthermore,

relationship investment has the added benefit of influencing performance directly. However,

managers must recognize that these proactive efforts will be wasted if they leave customer

conflict unresolved, because the negative influence of conflict on customer relationships is

greater in magnitude than that of any other strategy. Some firms thus could generate higher

returns by reallocating their RM investments to conflict resolution. Extending service recovery

research into the RM domain to develop strategies for “relationship recovery” also might be

worthwhile. A strategy of increasing customer dependence does not appear to be an effective

way to build relationships but does seem to influence seller performance directly. Neither

relationship duration nor interaction frequency is a good driver of strong customer relationships.

Of all the outcomes we analyze, relationships have the greatest influence on cooperation

and WOM. The impact on cooperation implies that RM efforts may be effectively extended

across many nontraditional buyer–seller interactions (e.g., alliances, interdepartmental groups);

in these situations, cooperation is often critical for success. Quite simply, firms that depend on

WOM strategies for new customers should implement effective RM programs.

Some results suggest that a more targeted effort may improve RM efficiency. Because

RM strategies appear to operate through different mediators that affect outcomes differentially, a

manager who desires cooperation between two groups after a merger and who recognizes that

trust is the relational mediator with the greatest influence on cooperation should select those RM

strategies that influence trust best (i.e., communication, interaction frequency). Marketers with a

portfolio of customers, channels, and products could improve the return on their RM

24
expenditures by targeting their spending toward segments in which RM is more likely to pay off,

such as those customers who purchase more services, channel versus direct customers, and

business versus consumer segments.

Managers also may want to leverage the potentially stronger impact on customer loyalty

and seller performance in relationships that involve an individual boundary spanner. For firms

that experience low turnover, focusing their RM efforts on building customer–salesperson bonds

may be a productive strategy, though developing strong relationships may prove difficult for

firms that want to move customers from dedicated salespeople to offshore call centers for various

reasons. These firms should recognize that a lack of seller expertise, dissimilarities between

boundary spanners and customers, ineffective communication, and shifts from interpersonal to

person-to-firm relationships can negatively affect customer–seller relationships.

Limitations

Meta-analyses have several strengths, but they also contain inherent limitations. First, the

constructs we include are constrained to those variables for which sufficient primary data are

available. Thus, our framework should be considered a summary of the most commonly studied

RM constructs, not an exhaustive list or even a list of the most important constructs. For

example, mutual dependence, seller disclosure, and functional conflict have been shown to be

important constructs for RM, but due to data unavailability, we could not include them in our

meta-analysis. Second, heterogeneity in effect sizes remained even after we accounted for any

variability due to the moderator variables in the study, which suggests that the effect sizes we

report should be considered averages and may vary with the inclusion of unmeasured moderating

conditions. Third, because of the limited number of studies for some moderator variables, our

study has limited power to reject null hypotheses.

25
Future Research Directions

After nearly two decades of RM research, marketers’ efforts may need to shift from

significant testing to identifying which RM strategies, and in what conditions, generate the

highest return on RM investment. Our synthesis of the extant literature identifies several avenues

that require further study.

Research should expand the constructs included in our RM-mediated framework and

determine which aspects or dimensions should be included to obtain a multifaceted view of

relational exchanges. Although commitment and trust clearly play critical roles, other candidates

might include relationship satisfaction, exchange efficiency, equity, relational norms, or

reciprocity. The absence of any measure of reciprocity between exchange partners is especially

notable because it has been identified as “the core of marketing relationships” (Bagozzi 1995, p.

275) and may help explain the pattern of effects surrounding the impact of relationship

investments and benefits on relational mediators. Integrating reciprocity into the relational-

mediating framework also may explain the large direct effect of relationship investment on

performance, such that people’s inherent desire to repay “debts” generated by sellers’

investments may lead to performance-enhancing behaviors, independent of trust or commitment.

In addition to taking a multidimensional perspective of relationships, the scope of RM

research should expand to investigate potential interactions among the relational mediators and

identify relational synergies. For example, the strong linkage between relationship quality and

objective performance may be due to interactions among the different facets of a relationship.

Even some of the high-impact antecedents and important outcomes in our framework

appear in relatively few primary studies (i.e., conflict, seller expertise, WOM), which suggests

the need for additional research. Seller expertise, beyond product-specific expertise, might

include overall customer knowledge, industry expertise, creativity, process knowledge, and

26
intraorganizational facilitation. Strategies that remedy conflict-laden events, such as service or

relationship recoveries, also are critical to incorporate into practice as well as further research.

The relatively small correlations between customer focal antecedents (relationship

benefit, dependence on seller) and relational mediators are surprising because we have taken the

relational mediators from the customer’s perspective as well. This finding may be due to a

misspecification; we may not have studied some critical customer focal antecedents. Thus,

researchers should investigate other customer-focal antecedents, such as perceived exchange

efficiency, perceived relationship investments, and liking, to identify any other key drivers of a

strong relationship from the customer’s perspective.

The heterogeneity across nearly all linkages, even after we account for the moderators we

included, demands research to determine other moderators that may influence RM effectiveness

(e.g., relationship age, customer control, customer involvement, relationship orientation of the

customer). For example, as the customer’s need for a relationship increases, RM strategies may

become more effective. Thus, researchers should develop a measure of the relationship

orientation of the customer to support the segmentation of RM efforts. Marketers then could

target their RM efforts toward those customers with the highest susceptibility for RM. Contrary

to most existing RM research, our results and social psychology theory suggest researchers

should differentiate between individual-to-individual and individual-to-firm relationships.

In brief, we provide insight into the most effective RM strategies, the conditions that

moderate that effectiveness, and how the links between both antecedents and consequences of

relational mediators depend on the mediator being investigated. These insights provide managers

with opportunities to improve the returns on their RM investments and researchers with

directions to build more robust models of the influence of RM on outcomes.

27
 TABLE 1
Review of Construct Definitions, Aliases, and Representative Papers

Constructs Definitions Common Aliases Representative Papers

Relational Mediators
Commitment An enduring desire to maintain a valued relationship Affective, behavioral, obligation, and Anderson and Weitz 1992; Jap and Ganesan 2000; Moorman, Zaltman,
normative commitment and Deshpande 1992; Morgan and Hunt 1994
Trust Confidence in an exchange partner’s reliability and integrity Trustworthiness, credibility, benevolence, and Doney and Cannon 1997; Hibbard et al. 2001; Sirdeshmukh, Singh, and
honesty Sabol 2002
Relationship Customer’s affective or emotional state toward a relationship, typically evaluated cumulatively Satisfaction with relationship, but not overall Crosby, Evans, and Cowles 1990; Reynolds and Beatty 1999
satisfaction over the history of the exchange satisfaction
Relationship Overall assessment of the strength of a relationship, conceptualized as a composite or Relationship closeness and strength Crosby, Evans, and Cowles 1990; De Wulf, Odekerken-Schröder, and
quality multidimensional construct capturing the different but related facets of an relationship Iacobucci 2001
Antecedents
Relationship Benefits received, including time saving, convenience, companionship, and improved decision Functional and social benefits and rewards Hennig-Thurau, Gwinner, and Gremler 2002; Morgan and Hunt 1994;
benefits making Reynolds and Beatty 1999
Dependence on Customer's evaluation of the value of seller-provided resources for which few alternatives are Relative and asymmetric dependence, Hibbard, Kumar, and Stern 2001; Morgan and Hunt 1994
seller available from other sellers switching cost, and imbalance of power
Relationship Seller's investment of time, effort, spending, and resources focused on building a stronger Support, gifts, resources, investments, and De Wulf, Odekerken-Schröder, and Iacobucci 2001; Ganesan 1994
investment relationship loyalty programs
Seller expertise Knowledge, experience, and overall competency of seller Competence, skill, knowledge, and ability Crosby, Evans, and Cowles 1990; Lagace, Dahlstrom, and Gassenheimer
1991
Communication Amount, frequency, and quality of information shared between exchange partners Bilateral or collaborative communication, Anderson and Weitz 1992; Mohr, Fisher, and Nevin 1996; Morgan and
information exchange, and sharing Hunt 1994
Similarity Commonality in appearance, lifestyle, and status between individual boundary spanners or Salesperson or cultural similarity, shared Crosby, Evans, and Cowles 1990; Doney and Cannon 1997; Morgan and
similar cultures, values, and goals between buying and selling organizations values, and compatibility Hunt 1994
Relationship Length of time that the relationship between the exchange partners has existed Relationship age or length, continuity, and Anderson and Weitz 1989; Doney and Cannon 1997; Kumar, Scheer, and
duration duration with firm or salesperson Steenkamp 1995
Interaction Number of interactions or number of interactions per unit time between exchange partners Frequency of business contact and interaction Crosby, Evans, and Cowles 1990; Doney and Cannon 1997
frequency intensity
Conflict Overall level of disagreement between exchange partners Manifest and perceived conflict or level of Anderson and Weitz 1992; Kumar, Scheer, and Steenkamp 1995
conflict, but not functional conflict
Outcomes
Expectation of Customer’s intention to maintain the relationship in the future, which captures the likelihood of Purchase intentions, likelihood to leave Crosby, Evans, and Cowles 1990; Doney and Cannon 1997
continuity continued purchases from the seller (reverse), and relationship continuity
Word of mouth Likelihood of a customer positively referring the seller to another potential customer Referrals and customer referrals Hennig-Thurau, Gwinner, and Gremler 2002; Reynolds and Beatty 1999

Customer loyalty Composite or multidimensional construct combining different groupings of intentions, attitudes, Behavioral loyalty and loyalty De Wulf, Odekerken-Schröder, and Iacobucci 2001; Hennig-Thurau,
and seller performance indicators Gwinner, and Gremler 2002; Sirdeshmukh, Singh, and Sabol 2002
Seller objective Actual seller performance enhancements including sales, share of wallet, profit performance, Sales, share, sales effectiveness, profit, and Reynolds and Beatty 1999; Siguaw, Simpson, and Baker 1998
performance and other measurable changes to the seller’s business sales performance
Cooperation Coordinated and complementary actions between exchange partners to achieve mutual goals Coordination and joint actions Anderson and Narus 1990; Morgan and Hunt 1994

28
 TABLE 2
Average Reliability-Adjusted Intercorrelations Among Constructs in Causal Model1, 2
Constructs RBEN DEPS RINV COMM RDUR RMED SOP
Relation Benefits (RBEN) [0.87]
Standard deviation
Number of studies
Cumulative sample size
Dependence on Seller (DEPS) 0.12 [0.85]
Standard deviation 0.29
Number of studies 4
Cumulative sample size 886
Relationship Investment (RINV) 0.42 0.13 [0.82]
Standard deviation 0.08 0.21
Number of studies 7 5
Cumulative sample size 1911 1273
Communication (COMM) 0.49 0.28 0.47 [0.85]
Standard deviation 0.15 0.21 0.13
Number of studies 10 12 9
Cumulative sample size 2380 3260 2893
Relationship Duration (RDUR) 0.15 0.14 0.02 0.18 [0.99]
Standard deviation 0.03 0.17 0.06 0.13
Number of studies 3 7 7 6
Cumulative sample size 1097 2150 3496 2282
Relational Mediator (RMED) 0.43 0.19 0.45 0.51 0.11 [0.85]
Standard deviation 0.18 0.25 0.18 0.19 0.13
Number of studies 18 33 24 38 26
Cumulative sample size 5108 9296 8564 9803 10720
Seller Objective Performance (SOP) 0.23 0.29 0.44 0.30 0.12 0.35 [0.92]
Standard deviation 0.04 0.31 0.31 0.30 0.12 0.22
Number of studies 3 8 7 12 8 47
Cumulative sample size 600 2860 2813 3149 4293 16469
1
Entries on the diagonal in brackets [ ] are weighted mean Cronbach's alpha coefficients.
2
Constructs were included in the causal model when three or more correlation coefficients were available among that construct
and all other constructs in the model.

29
 TABLE 3
Results: Descriptive Statistics and Influence of Antecedents on Relational Mediators
Number
Simple
Average r Sample Weighted χ2 for 95% Confidence Interval1 File Drawer Q-Statistic for
Proposed Relationships of Raw Total N Adjusted for Reliability Adjusted Association N (using two- Homogeneity
Average r
(df = 1)1 Lower Bound Upper Bound tailed test)
1
Effects Reliability Average r Test (df)1
Relationship benefits → Commitment 11 3,162 0.37 0.45 0.51 987.50 0.48 0.53 2,670 501.29 (10)
Relationship benefits → Trust 13 3,633 0.31 0.34 0.33 420.99 0.30 0.36 1,433 66.61 (12)
Relationship benefits → Relationship satisfaction 7 2,057 0.41 0.46 0.45 490.59 0.43 0.49 916 53.56 (6)
Relationship benefits → Relationship quality 8 2,091 0.33 0.39 0.36 353.47 0.35 0.43 749 60.27 (7)
Relationship benefits → All mediators 39 10,943 0.35 0.40 0.42
Dependence on seller → Commitment 16 4,670 0.33 0.40 0.37 700.04 0.35 0.40 3,206 530.30 (15)
Dependence on seller → Trust 26 5,935 0.14 0.17 0.21 261.01 0.18 0.23 1,354 452.24 (25)
Dependence on seller → Relationship satisfaction 3 1,076 0.13 0.13 0.27 ― ― ― ― ―
Dependence on seller → Relationship quality 2 1,604 0.07 0.08 0.08 ― ― ― ― ―
Dependence on seller → All mediators 47 13,285 0.20 0.24 0.26
Relationship investment → Commitment 15 6,544 0.36 0.43 0.34 821.86 0.32 0.36 4,200 381.19 (14)
Relationship investment → Trust 17 4,601 0.38 0.45 0.45 1,053.73 0.42 0.47 4,455 161.70 (16)
Relationship investment → Relationship satisfaction 10 2,691 0.42 0.47 0.52 882.85 0.49 0.55 2,217 276.14 (9)
Relationship investment → Relationship quality 9 2,635 0.49 0.57 0.58 1,125.80 0.55 0.60 2,750 245.40 (8)
Relationship investments → All mediators 51 16,471 0.40 0.47 0.46
Seller expertise → Commitment 1 177 0.78 0.90 0.90 ― ― ― ― ―
Seller expertise → Trust 12 3,464 0.49 0.59 0.52 1,121.22 0.49 0.54 3,988 279.45 (11)
Seller expertise → Relationship satisfaction 5 1,049 0.49 0.56 0.56 413.64 0.52 0.60 473 1.20 (4)
Seller expertise → Relationship quality 1 1,009 0.90 0.98 0.98 ― ― ― ― ―
Seller expertise → All mediators 19 5,699 0.53 0.62 0.62
Communication → Commitment 25 5,840 0.45 0.53 0.55 2,244.96 0.54 0.57 13,712 323.37 (24)
Communication → Trust 29 7,948 0.43 0.51 0.56 3,146.74 0.54 0.57 21,962 939.32 (28)
Communication → Relationship satisfaction 6 1,727 0.46 0.51 0.51 546.15 0.48 0.55 870 34.78 (5)
Communication → Relationship quality 7 1,907 0.35 0.40 0.43 407.26 0.40 0.47 709 82.39 (6)
Communication → All mediators 67 17,422 0.43 0.51 0.54
Similarity → Commitment 3 386 0.54 0.66 0.63 ― ― ― ― ―
Similarity → Trust 10 2,562 0.50 0.56 0.41 475.20 0.38 0.44 1,884 347.42 (9)
Similarity → Relationship satisfaction 1 151 0.30 0.34 0.34 ― ― ― ― ―
Similarity → Relationship quality 1 1,009 0.22 0.24 0.24 ― ― ― ― ―
Similarity → All mediators 15 4,108 0.48 0.54 0.44
Relationship duration → Commitment 13 6,638 0.12 0.13 0.11 82.43 0.09 0.14 312 241.99 (12)
Relationship duration → Trust 20 8,201 0.12 0.12 0.14 162.63 0.12 0.16 674 79.73 (19)
Relationship duration → Relationship satisfaction 5 1,542 0.09 0.09 0.13 24.15 0.08 0.17 18 11.14 (4)
Relationship duration → Relationship quality 5 1,830 0.11 0.11 0.11 22.95 0.07 0.16 25 25.15 (4)
Relationship duration → All mediators 43 18,211 0.11 0.12 0.13
Interaction frequency → Commitment 2 724 -0.03 -0.03 -0.03 ― ― ― ― ―
Interaction frequency → Trust 10 2,198 0.30 0.33 0.30 210.97 0.26 0.35 680 294.63 (9)
Interaction frequency → Relationship satisfaction 4 965 0.11 0.11 0.04 1.47 -0.02 0.10 2 27.81 (3)
Interaction frequency → Relationship quality 3 1,124 -0.03 -0.03 -0.03 ― ― ― ― ―
Interaction frequency → All mediators 19 5,011 0.17 0.19 0.16
Conflict → Commitment 10 4,339 -0.41 -0.49 -0.71 3,331.03 -0.72 -0.69 5,496 1,145.07 (9)
Conflict → Trust 9 2,906 -0.54 -0.63 -0.65 1,769.52 -0.68 -0.63 4,033 206.98 (8)
Conflict → Relationship satisfaction 1 95 -0.27 -0.31 -0.31 ― ― ― ― ―
Conflict → Relationship quality 0 NA NA NA NA ― ― ― ― ―
Conflict → All mediators 20 7,340 -0.46 -0.55 -0.67
1 Operationally, we attempted calculations only when there were a minimum of four raw effects associated with a relationship. A dash ( ― ) indicates this condition was not met.

30
 TABLE 4
Results: Descriptive Statistics and Influence of Relational Mediators on Outcomes
1
Number Average r Sample Weighted χ2 for 95% Confidence Interval File Drawer Q-Statistic for
Simple
Proposed Relationships of Raw Total N Adjusted for Reliability Association N (using two- Homogeneity
Average r
Effects Reliability Adjusted Average r (df = 1)1 Lower Bound Upper Bound tailed test)1 Test (df)1

Commitment → Expectation of continuity 16 4,215 0.45 0.54 0.53 1,447.81 0.51 0.55 5,895 210.88 (15)
Trust → Expectation of continuity 24 6,632 0.47 0.55 0.58 2,889.95 0.56 0.60 15,456 274.10 (23)
Relationship satisfaction → Expectation of continuity 5 1,879 0.50 0.58 0.57 778.08 0.54 0.60 949 21.36 (4)
Relationship quality → Expectation of continuity 3 1,733 0.50 0.55 0.54 ― ― ― ― ―
All mediators → Expectation of continuity 48 14,459 0.47 0.55 0.56

Commitment → Word of mouth 6 3,674 0.52 0.61 0.64 2,111.52 0.62 0.66 2,707 147.41 (5)
Trust → Word of mouth 5 3,507 0.48 0.56 0.62 1,833.48 0.60 0.64 1,804 61.05 (4)
Relationship satisfaction → Word of mouth 3 1,054 0.48 0.50 0.53 ― ― ― ― ―
Relationship quality → Word of mouth 3 1,733 0.58 0.61 0.60 ― ― ― ― ―
All mediators → Word of mouth 17 9,968 0.51 0.58 0.61

Commitment → Customer loyalty 12 4,588 0.45 0.54 0.58 1,996.01 0.56 0.60 5,447 151.79 (11)
Trust → Customer loyalty 20 6,328 0.44 0.51 0.54 2,248.51 0.52 0.55 10,572 308.98 (19)
Relationship satisfaction → Customer loyalty 9 2,781 0.35 0.39 0.41 522.61 0.38 0.44 1,188 129.89 (8)
Relationship quality → Customer loyalty 9 2,851 0.40 0.46 0.47 750.86 0.45 0.50 1,722 55.04 (8)
All mediators → Customer loyalty 50 16,548 0.42 0.48 0.52

Commitment → Seller objective performance 20 7,342 0.30 0.35 0.27 549.90 0.25 0.29 3,489 407.99 (19)
Trust → Seller objective performance 32 10,306 0.29 0.33 0.35 1,333.28 0.33 0.36 10,108 924.64 (31)
Relationship satisfaction → Seller objective performance 7 1,605 0.32 0.37 0.32 172.58 0.27 0.36 376 84.62 (6)
Relationship quality → Seller objective performance 6 3,517 0.28 0.31 0.63 1,930.18 0.61 0.65 1,986 2,641.29 (5)
All mediators → Seller objective performance 65 22,770 0.29 0.34 0.35

Commitment → Cooperation 16 4,436 0.41 0.50 0.64 2,509.07 0.62 0.66 7,385 418.42 (15)
Trust → Cooperation 24 6,192 0.56 0.67 0.73 5,340.80 0.72 0.74 28,898 476.68 (23)
Relationship satisfaction → Cooperation 5 931 0.45 0.55 0.68 630.30 0.64 0.71 468 17.33 (4)
Relationship quality → Cooperation 0 NA NA NA NA ― ― ― ― ―
All mediators → Cooperation 45 11,559 0.50 0.60 0.70
1 Operationally, we attempted calculations only when there were a minimum of four raw effects associated with a relationship. A dash ( ― ) indicates this condition was not met.

31
 TABLE 5
Influence of Moderators on Relational Mediators' Effects on Outcomes1
Total Service vs. Product- Channel vs. Direct Business vs. Consumer Individual vs.
Moderated Relationships Number of Based Exchanges Exchanges Markets Organizational
Raw Effects Service Product Channel Direct Business Consumer Individual Organizational
Commitment → Expectation of continuity 16 0.63 (6) 0.46 (8) 2
0.54 (4) 0.53 (10) 0.46 (11) 0.71 (5) * 0.54 (4) 0.53 (12)
Trust → Expectation of continuity 24 0.51 (6) 0.56 (15) 0.57 (7) 0.55 (13) 0.54 (15) 0.57 (9) 0.50 (4) 0.55 (17)
Relationship satisfaction → Expectation of continuity 5 ― ― ― ― ― ― ― ―
Relationship quality → Expectation of continuity 3 ― ― ― ― ― ― ― ―
All mediators → Expectation of continuity 48 0.56 (15) 0.53 (24) 0.54 (26) 0.57 (12) 0.52 (33) 0.61 (15) * 0.52 (13) 0.55 (32)

Commitment → Customer loyalty 12 0.70 (2) 0.49 (8) * 0.63 (1) 0.52 (9) 2 0.59 (3) 0.52 (9) 2
0.62 (2) 0.52 (10) *
2
Trust → Customer loyalty 20 0.53 (5) 0.47 (12) 0.68 (1) 0.49 (17) 0.49 (4) 0.51 (16) 0.51 (5) 0.50 (14)
Relationship satisfaction → Customer loyalty 9 ― ― ― ― 0.60 (2) 0.33 (7) * 0.59 (2) 0.33 (7) *
Relationship quality → Customer loyalty 9 0.54 (1) 0.40 (6) 2
― ― 0.60 (2) 0.42 (7) * 0.61 (1) 0.44 (8) *
* * * *
All mediators → Customer loyalty 50 0.58 (8) 0.43 (33) 0.65 (2) 0.46 (40) 0.56 (11) 0.46 (39) 0.56 (10) 0.46 (39)

Commitment → Seller objective performance 20 0.26 (6) 0.40 (10) 2

0.37 (7) 0.37 (11) 0.39 (16) 0.21 (4) * 0.28 (2) 0.36 (18)
Trust → Seller objective performance 32 0.27 (4) 0.34 (24) 0.31 (7) 0.30 (16) 0.35 (26) 0.26 (6) 0.38 (13) 0.31 (17)
2
Relationship satisfaction → Seller objective performance 7 ― ― ― ― 0.40 (5) 0.29 (2) 0.44 (5) 0.19 (2)
2
Relationship quality → Seller objective performance 6 ― ― ― ― ― ― 0.59 (2) 0.17 (4)
* 2
All mediators → Seller objective performance 65 0.32 (12) 0.36 (41) 0.36 (19) 0.34 (31) 0.36 (53) 0.25 (12) 0.40 (22) 0.31 (41)

Commitment → Cooperation 16 0.63 (3) 0.47 (11) 2

0.49 (7) 0.51 (9) 0.50 (15) 0.54 (1) 0.79 (2) 0.46 (14) *
Trust → Cooperation 24 0.73 (3) 0.67 (21) 0.70 (10) 0.66 (14) 0.67 (23) 0.66 (1) 0.70 (9) 0.64 (14)
Relationship satisfaction → Cooperation 5 ― ― ― ― ― ― ― ―
Relationship quality → Cooperation 0 ― ― ― ― ― ― ― ―
All mediators → Cooperation 45 0.67 (7) 0.59 (36) 2
0.60 (21) 0.60 (24) 0.60 (42) 0.61 (3) 0.68 (14) 0.55 (30) *
* p < .05, one-tailed.
1 The cell entries show the average effects encountered for each moderator level, with the total number of effects in parenthesis (), subjected to t-test comparisons (cf. Brown 1996). We dropped those
studies that could not be coded into subgroups from the comparisons. Also, the limited number of effects suggested dropping word of mouth from the analysis. Operationally, we only carry out
comparisons when the total number of raw effects is six or more to ensure the a priori probability of finding at least three effects at each level of moderator. When this cutoff is not met or the number
of effects for one level of moderator is less than 1, we use a dash ( ― ) to indicate that we did not perform that particular moderator analysis.
2
Nonsignificant results should be interpreted cautiously. In a majority of the cases for which a moderator variable does not significantly moderate the effect of a given relationship, the power
associated with the test is relatively small. On the basis of the expected effect size, power, and number of studies required to move the power to 80%, we identify (with superscript 2) those
relationships that researchers may be premature in dismissing as not significantly moderated (cf. Cohen 1977; Fern and Monroe 1996).

32
 TABLE 6
Summary of Key Findings and Implications
Key Findings Research and Managerial Implications
Antecedents
Relationship marketing (RM) strategies/antecedents have a wide range of Selection and training of boundary spanners is critical; expertise, communication, and similarity
effectiveness for generating strong relationships. Expertise and communication are to customers are some of the most effective relationship-building strategies. Expertise's impact
most effective, then relationship investment, similarity, and relationship benefits; supports Vargo and Lusch’s (2004) premise that “skills and knowledge" are the most important
dependence, frequency, and duration are relatively ineffective. seller value-creation attributes.

The negative impact of conflict is larger in magnitude than the positive effect of any All proactive RM efforts may be wasted if customer conflict is left unresolved.
other RM strategy.

Specific RM strategies appear most effective for strengthening one aspect of a RM may be improved through a fine-grained approach that targets specific relational weaknesses.
relationship. Relationship benefits, customer dependency, and similarity are more The relative effectiveness of RM strategies depends on the relational mediator investigated.
effective for increasing commitment than for building trust; the opposite is true for
relationship investment and frequency.
Outcomes
Relationship quality (a composite measure of relationship strength) has the largest No single relational mediator captures the full essence or depth of a customer–seller relationship;
influence on objective performance and commitment the least. the findings support a multidimensional perspective of relationships. Extant research focused on a
single relational mediator may provide misleading guidance.

Surprisingly, relationship investment has a large direct effect on seller objective The classic mediating model of RM (Morgan and Hunt 1994) needs to be adapted to include
performance, in addition to its frequently hypothesized indirect effect. alternative mediated pathways (e.g., reciprocity).

Dependence has a large direct effect on seller objective performance but a relatively Dependence is not a very effective relationship-building strategy but can improve performance in
small impact on relational mediators. other ways, possibly by increasing switching costs and barriers to exit.

Of all outcomes, relationships have the greatest influence on cooperation and WOM RM efforts may be effectively extended across many other nontraditional buyer–seller
and the least on objective performance. interactions (e.g., interdepartmental groups) for which cooperation is often critical for success.
WOM behaviors may be the best discriminator of true customer loyalty (Reichheld 2003).

Moderators
Relationship marketing is typically more effective when relationships are more Researchers must take care when extending findings across contexts in which relationship
critical to customers, such as for (1) service versus product offerings, (2) channel importance may vary. Managers might target RM expenditures to customer segments with the
versus direct exchanges, and (3) business versus consumer markets. highest desire for strong relationships to improve returns.

Customer relationships often have stronger effects on exchange outcomes when Researchers should differentiate the effects of customer relationships with boundary spanners
their target is an individual person rather than a selling firm. from those with firms. Strategies such as team selling, salesperson disintermediation, and the use
of call centers should be evaluated in light of the impact of interpersonal relationships.

33
 FIGURE 1
Relational Mediator Meta-Analytic Framework

Customer Focal Antecedents Customer Focal Outcomes

Relationship benefits1 Expectation of continuity
Dependence on seller1 Word of mouth
Customer Focal
Customer loyalty
Relational Mediators1
Seller Focal Antecedents Seller Focal Outcomes
•Commitment
Relationship investment1 •Trust Seller objective performance1
•Relationship satisfaction
Seller expertise
•Relationship quality

Dyadic Antecedents Dyadic Outcomes

Communication1 Cooperation
Moderators
Similarity
•Service vs. product-based exchanges
Relationship duration1 •Channel vs. direct exchanges
Interaction frequency •Business vs. consumer markets
•Individual vs. organizational relationships
Conflict

1 Construct had sufficient reported effects to be included in the multivariate causal model.

34
 REFERENCES

Ajzen, Icek and Martin Fishbein (1980), Understanding Attitudes and Predicting Social Behavior.
Englewood Cliffs, NJ: Prentice-Hall.

Anderson, Erin and Barton A. Weitz (1989), "Determinants of Continuity in Conventional

Industrial Channel Dyads," Marketing Science, 8 (Fall), 310-23.

——— and ——— (1992), "The Use of Pledges to Build and Sustain Commitment in Distribution
Channels," Journal of Marketing Research, 29 (February), 18-34.

Anderson, James C. and James A. Narus (1990), "A Model of Distributor Firm and Manufacturer
Firm Working Partnerships," Journal of Marketing, 54 (January), 42 - 58.

——— and ——— (1991), "Partnering as a Focused Market Strategy," California Management
Review, 33 (Spring), 95-113.

——— and ——— (2004), Business Market Management: Understanding, Creating, and
Delivering Value. Upper Saddle River, NJ: Prentice Hall.

Bagozzi, Richard P. (1995), "Reflections on Relationship Marketing in Consumer Markets,"

Journal of the Academy of Marketing Science, 23 (4), 272-77.

Berry, Leonard L. (1996), "Retailers with a Future," Marketing Management, 5 (Spring), 39-46.

Brown, Steven P. (1996), "A Meta-Analysis and Review of Organizational Research on Job
Involvement," Psychological Bulletin, 120 (2), 234-49.

——— and Robert A. Peterson (1993), "Antecedents and Consequences of Salesperson Job
Satisfaction: Meta-Analysis and Assessment of Causal Effects," Journal of Marketing
Research, 30 (February), 63-77.

Cheung, Mike W. and Wai Chan (2005), "Meta-Analytic Structural Equation Modeling: A Two-
Stage Approach," Psychological Methods, 10 (1), 40-64.

Cheung, Shu Fai and Darius K. S. Chan (2004), "Dependent Effect Sizes in Meta-Analysis:
Incorporating the Degree of Interdependence," Journal of Applied Psychology, 89 (5), 780-
91.

Cohen, Jacob (1977), Statistical Power Analysis for the Behavioral Sciences. New York: Academic
Press.

Colgate, Mark R. and Peter J. Danaher (2000), "Implementing a Customer Relationship Strategy:
The Asymmetric Impact of Poor Versus Excellent Execution," Journal of the Academy of
Marketing Science, 28 (3), 375-87.

35
Crosby, Lawrence A., Kenneth R. Evans, and Deborah Cowles (1990), "Relationship Quality in
Services Selling: An Interpersonal Influence Perspective," Journal of Marketing, 54 (July),
68-81.

Devine, Patricia G. (1995), "Prejudice and Out-Group Perception," in Advanced Social Psychology,
Abraham Tesser, ed. New York: McGraw-Hill, 467-524.

De Wulf, Kristof, Gaby Odekerken-Schröder, and Dawn Iacobucci (2001), "Investments in

Consumer Relationships: A Cross-Country and Cross-Industry Exploration," Journal of
Marketing, 65 (October), 33-50.

Doney, Patricia M. and Joseph P. Cannon (1997), "An Examination of the Nature of Trust in Buyer-
Seller Relationships," Journal of Marketing, 61 (April), 35-51.

Dwyer, Robert F., Paul H. Schurr, and Sejo Oh (1987), "Developing Buyer-Seller Relationships,"
Journal of Marketing, 51 (April), 11-27.

Fern, Edward F. and Kent B. Monroe (1996), "Effect Size Estimates: Issues and Problems in
Interpretation," Journal of Consumer Research, 25 (September), 89-105.

Furlow, Carolyn F. and S. Natasha Beretvas (2005), "Meta-Analytic Methods of Pooling

Correlation Matrices for Structural Equation Modeling Under Different Patterns of Missing
Data," Psychological Methods, 10 (2), 227-54.

Ganesan, Shankar (1994), "Determinants of Long-Term Orientation in Buyer-Seller Relationships,"

Journal of Marketing, 58 (April), 1-19.

Grewal, Dhruv, Sukumar Kavanoor, Edward F. Fern, Caroly Costley, and James Barnes (1997),
"Comparative Versus Noncomparative Advertising: A Meta-Analysis," Journal of
Marketing, 61 (Oct), 1-15.

Gruen, Thomas W., John O. Summers, and Frank Acito (2000), "Relationship Marketing Activities,
Commitment, and Membership Behaviors in Professional Associations," Journal of
Marketing, 64 (July), 34-49.

Gundlach, Gregory T., Ravi S. Achrol, and John T. Mentzer (1995), "The Structure of Commitment
in Exchange," Journal of Marketing, 59 (January), 78-92.

Hamilton, David L. and Steven J. Sherman (1996), "Perceiving Persons and Groups," Psychological
Review, 103 (2), 336-55.

Hennig-Thurau, Thorsten, Kevin P. Gwinner, and Dwayne D. Gremler (2002), "Understanding

Relationship Marketing Outcomes: An Integration of Relational Benefits and Relationship
Quality," Journal of Service Research, 4 (February), 230-47.

Hibbard, Jonathan D., Frederic F. Brunel, Rajiv P. Dant, and Dawn Iacobucci (2001), "Does
Relationship Marketing Age Well?" Business Strategy Review, 12 (4), 29-35.

36
———, Nirmalya Kumar, and Louis W. Stern (2001), "Examining the Impact of Destructive Acts
in Marketing Channels Relationship," Journal of Marketing Research, 38 (February), 25-61.

Hunter, John E. and Frank L. Schmidt (1990), Methods of Meta Analysis. Newbury Park, CA: Sage
Publications.

Iacobucci, Dawn and Amy Ostrom (1996), "Commercial and Interpersonal Relationships; Using the
Structure of Interpersonal Relationships to Understand Individual-to-Individual, Individual-
to-Firm, and Firm-to-Firm Relationships in Commerce," International Journal of Research
in Marketing, 13 (1), 53-72.

Jap, Sandy D. and Shankar Ganesan (2000), "Control Mechanisms and Relationship Life Cycle:
Implications for Safeguarding Specific Investments and Developing Commitment," Journal
of Marketing Research, 37 (May), 227-45.

Johnson, Jean L. (1999), "Strategic Integration in Industrial Distribution Channels: Managing the
Interfirm Relationship as a Strategic Asset," Journal of the Academy of Marketing Science,
27 (1), 4-18.

Kumar, Nirmalya, Lisa K. Scheer, and Jan-Benedict E. M. Steenkamp (1995), "The Effects of
Supplier Fairness on Vulnerable Resellers," Journal of Marketing Research, 32 (February),
54-65.

Lagace, Rosemary R., Robert Dahlstrom, and Jule B. Gassenheimer (1991), "The Relevance of
Ethical Salesperson Behavior on Relationship Quality: The Pharmaceutical Industry,"
Journal of Personal Selling and Sales Management, 11 (Fall), 39-47.

Mohr, Jakki J., Robert J. Fisher, and John R. Nevin (1996), "Collaborative Communication in
Interfirm Relationships: Moderating Effects of Integration and Control," Journal of
Marketing, 60 (July), 103-15.

Moorman, Christine, Gerald Zaltman, and Rohit Deshpandé (1992), "Relationships Between
Providers and Users of Market Research: The Dynamics of Trust Within and Between
Organizations," Journal of Marketing Research, 29 (August), 314-29.

Morgan, Robert M. and Shelby D. Hunt (1994), "The Commitment-Trust Theory of Relationship
Marketing," Journal of Marketing, 58 (July), 20-38.

Oliver, Richard L. (1999), "Whence Consumer Loyalty?" Journal of Marketing, 63 (Special Issue),
33-44.

Orwin, Robert G. (1983), "A Fail-Safe N for Effect Size in Meta-Analysis," Journal of Educational
Statistics, 8, 157-59.

Reichheld, Fredrick F. (2003), "The One Number You Need," Harvard Business Review, 81
(December), 46-54.

37
Reinartz, Werner J. and V. Kumar (2003), "The Impact of Customer Relationship Characteristics on
Profitable Lifetime Duration," Journal of Marketing, 67 (January), 77-99.

Reynolds, Kristy E. and Sharon E. Beatty (1999), "Customer Benefits and Company Consequences
of Customer-Salesperson Relationships in Retailing," Journal of Retailing, 75 (1), 11-32.

Rosenthal, Robert (1979), "The 'File-Drawer Problem' and Tolerance for Null Results,"
Psychological Bulletin, 86, 638-41.

——— (1994), "Parametric Measures of Effect Size," in The Handbook of Research Synthesis,
Harris Cooper and Larry V. Hedges, eds. New York: Russell Sage Foundation, 231-44.

Shadish, William R. and Keith C. Haddock (1994), "Combining Estimates of Effect Size," in The
Handbook of Research Synthesis, Harris Cooper and Larry V. Hedges, eds. New York:
Russell Sage Foundation, 261-84.

Shiv, Baba, Julie A. Edell, and John W. Payne (1997), "Factors Affecting the Impact of Negatively
and Positively Framed Ad Messages," Journal of Consumer Research, 24 (December), 285-
94.

Siguaw, Judy A., Penny M. Simpson, and Thomas L. Baker (1998), "Effects of Supplier Market
Orientation on Distributor Market Orientation and the Channel Relationship: The
Distributor Perspective," Journal of Marketing, 62 (July), 99-111.

Sirdeshmukh, Deepak, Jagdip Singh, and Barry Sabol (2002), "Consumer Trust, Value, and Loyalty
in Relational Exchanges," Journal of Marketing, 66 (January), 15-37.

Smith, J. Brock (1998), "Buyer-Seller Relationships: Bonds, Relationship Management, and Sex-
Type," Canadian Journal of Administrative Sciences, 15 (1), 76-92.

Spekman, Robert E. (1988), "Strategic Supplier Selection: Understanding Long-Term

Relationships," Business Horizons, 31 (July/August), 75-81.

Srinivasan, Raji and Christine Moorman (2005), "Strategic Firm Commitments and Rewards for
Customer Relationship Management in Online Retailing," Journal of Marketing, 69
(October), 193-200.

Sterne, Jonathan A. C. and Matthias Egger (2001), "Funnel Plots for Detecting Bias in Meta-
Analysis: Guidelines on Choice of Axis," Journal of Clinical Epidemiology, 54, 1046-55.

Szymanski, David M. and David H. Henard (2001), "Customer Satisfaction: A Meta-Analysis of the
Empirical Evidence," Journal of the Academy of Marketing Science, 29 (1), 16-35.

Vargo, Stephen L. and Robert F. Lusch (2004), "Evolving to a New Dominant Logic for
Marketing," Journal of Marketing, 68 (January), 1-17.

Zeithaml, Valarie A., A. Parasuraman, and Leonard L. Berry (1985), "Problems and Strategies in
Services Marketing," Journal of Marketing, 49 (Spring), 33-46.

38
1
Upon request, we can provide a list of the articles used in our empirical meta-analysis.
2
Prior to applying the sample weights, we first converted the reliability-adjusted rs to variance-stabilizing Fisher’s z
scores (Rosenthal 1994; Shadish and Haddock 1994). Following standard procedures (Shadish and Haddock 1994, p.
268), we reconverted them back to rs to report the sample-weighted, reliability-adjusted r and the 95% confidence
intervals.
3
To estimate the publication bias associated with published studies, we employed multiple methods: (1) Rosenthal’s
(1979) well-known file drawer method; (2) Orwin’s (1983) failsafe N, which represents the number of missing studies
(set to .05) that would bring the effect to .075, or less than the .10 effect level classified by Cohen (1977) as signifying a
low effect; and (3) comprehensive meta-analysis software (cf. http://www.meta-
analysis.com/html/stat_analysis_overview.html) to compute funnel plots for the various relationships. Funnel plots offer
a simple scatter plot–based visual tool for investigating publication bias in meta-analyses (cf. Sterne and Egger 2001).
Overall, the funnel plots corroborate the inferences we drew from the file drawer N and Orwin’s failsafe N; namely, the
data we use in the meta-analyses do not display any evidence of publication bias.
4
The median sample size from our meta-analysis of the studies included in the causal model is 2,839. Modification
indices indicate that a direct path from relationship duration to seller objective performance could improve the model
fit, but our evaluation of the parsimony-adjusted fit indices suggests that the slight improvement in fit is more than
offset by a loss in parsimony. The significance and pattern of effects do not change with this additional path, so we do
not add it.
5
We carried out the moderator analysis using the procedure employed by Brown (1996) and Grewal and colleagues
(1997). The results we report in Table 5 must be interpreted cautiously, because in the majority of the nonsignificant
cases, the power of the test is relatively small (Cohen 1977; Fern and Monroe 1996). On the basis of our power analysis,
we have flagged the relationships in Table 5 that we believe researchers would be premature in dismissing.

39