Factors Influencing the Likelihood of Customer Defection: The Role of Consumer Knowledge

Page created by James Dixon
 
CONTINUE READING
JOURNAL OF THE ACADEMY
                 Capraro
                 10.1177/0092070302250900
                         etOF
                           al. MARKETING
                               / CUSTOMER SCIENCE
                                          DEFECTION   ARTICLE                SPRING 2003

                                             Factors Influencing the Likelihood
                                             of Customer Defection:
                                             The Role of Consumer Knowledge
                                             Anthony J. Capraro
                                             University of North Carolina at Asheville

                                             Susan Broniarczyk
                                             University of Texas at Austin

                                             Rajendra K. Srivastava
                                             University of Texas at Austin

                                             Customer satisfaction is the predominant metric firms use           Today, most firms’programs to control customer defec-
                                             for detecting and managing customers’ likelihood to de-          tion center heavily on the management of customer satis-
                                             fect. But while satisfaction and defection are related, satis-   faction. There is, of course, good support for this
                                             faction is only a weak predictor of whether a customer will      approach. A substantial body of research suggests that
                                             defect. This article suggests that for repurchase decisions      both repurchase intent (e.g., Anderson and Sullivan 1993;
                                             that involve an information-based evaluation of alterna-         Coyne 1989; Cronin and Taylor 1992; LaBarbera and
                                             tives to the incumbent, likelihood of defection will be influ-   Mazursky 1983; Reichheld and Sasser 1990) and repur-
                                             enced by “how much” customers know about those                   chase behavior (e.g., Bolton 1998; LaBarbera and
                                             alternatives. The relationship between level of knowledge        Mazursky 1983; Newman and Werbel 1973; Sambandam
                                             about alternatives and defection is examined in the context      and Lord 1995) are linked to customer satisfaction.
                                             of actual health insurance choices. Results suggest that the        However, a closer review raises the question as to how
                                             level of objective and subjective knowledge about alterna-       well defection can be controlled by focusing solely on
                                             tives has a direct effect on likelihood of defection—above       managing satisfaction. A number of academic studies
                                             and beyond satisfaction level. The view of defection for-        found that satisfaction explains a relatively small propor-
                                             warded in this article suggests that managers may be able        tion of the variance (less than 8%) in repurchasing behav-
                                             to gain additional control over customer defection through       iors (Bolton 1998; LaBarbera and Mazursky 1983;
                                             actions aimed at influencing how much customers know             Newman and Werbel 1973). Reichheld (1996) cited
                                             (or come to know) about alternative vendors.                     research indicating that 65 to 85 percent of defecting cus-
                                                                                                              tomers do so despite being “satisfied” or “highly satis-
                                             Keywords: customer defection; knowledge about alterna-           fied.” Thus, high satisfaction levels do not guarantee that
                                                       tives; satisfaction; missing information; evalu-       customers will not defect. And dissatisfaction does not
                                                       ation of alternatives                                  necessarily lead to defection—customers may continue to
                                                                                                              purchase from a vendor that has been a source of dissatis-
                                                                                                              faction (Hennig-Thurau and Klee 1997).
                                                                                                                 Some observe that the relationship between satisfaction
                                             Journal of the Academy of Marketing Science.                     and defection is nonlinear—that satisfaction has a stronger
                                             Volume 31, No. 2, pages 164-175.
                                             DOI: 10.1177/0092070302250900                                    effect on defection when customers are “extremely satis-
                                             Copyright © 2003 by Academy of Marketing Science.                fied” (Coyne 1989; Jones and Sasser 1995; Mittal and
Capraro et al. / CUSTOMER DEFECTION              165

Kamakura 2001). However, reaching uniform “extreme                                     FIGURE 1
satisfaction” across a diverse customer base can be costly               The Relationship Between Consumer
(Fornell 1992)—at some point, efforts to reduce defection                   Knowledge, Satisfaction Level,
by improving satisfaction may yield diminishing (or nega-                    and Likelihood of Defection—
tive) returns.                                                            Specifications of Competing Models
   The above observations suggest value in broadening
our understanding of the factors that can influence a cus-      Figure 1A.        Main Effects of Knowledge (Model 1)
tomer’s likelihood to defect. To this end, Oliver (1999)
                                                                 Objective Knowledge                      (+)
explored the notion of a psychological state of loyalty. At      at Initiation of
the highest states of such loyalty, a customer will block out    Prepurchase Search

communications from the incumbent’s competitors, mak-            Satisfaction
                                                                                                          (-)               Likelihood of
                                                                 Level                                                      Defection
ing him or her less likely to defect. Other authors have
                                                                                                          (+)
explored switching barriers as a deterrent of defection          Subjective Knowledge
                                                                 at Time of Choice
(e.g., Fornell 1992; Heide and Weiss 1995; Jones,
Mothersbaugh, and Beatty 2000; Oliva, Oliver, and
MacMillan 1992).                                                Figure 1B.        Interaction Between Satisfaction and Knowledge (Model 2)
   This article attempts to further our understanding of
defection, focusing on repurchase decisions that involve         Objective Knowledge
                                                                 at Initiation of
                                                                                              (+)

an information-based evaluation of alternatives to the           Prepurchase Search                         (-)
                                                                                                                      (-)   Likelihood of
incumbent (alternatives). It examines a factor previously                                           Satisfaction
                                                                                                    Level                   Defection
noted as a switching barrier (Klemperer 1987), but whose                                                    (+/-)
                                                                                              (+)
relationship to likelihood of defection has not been sys-        Subjective Knowledge
                                                                 at Time of Choice
tematically studied—customers’ level of knowledge about
alternatives. Controlling for satisfaction level, it explores
the relationship between likelihood of defection and the
                                                                Figure 1C.        Knowledge as Mediator Between Satisfaction and Likelihood
level of objective knowledge (Hypothesis 1) and subjec-                           of Defection (Model 3)
tive knowledge (Hypothesis 2). Level of objective knowl-
                                                                            (-)      Objective Knowledge
edge about alternatives is defined as the number of                                  at Initiation of
                                                                                                                    (+)
                                                                                     Prepurchase Search
instances of accurate information about alternatives (e.g.,
product features) stored in memory. Level of subjective          Satisfaction                 (-)                           Likelihood of
                                                                 Level                                                      Defection
knowledge (Brucks 1985; Park, Mothersbaugh, and Feick
                                                                                     Subjective Knowledge           (+)
1994) is defined in terms of how much individuals per-                               at Time of Choice
                                                                            (-)
ceive they know about alternatives. The article also tests
competing models of the interrelationship between knowl-
edge and satisfaction (Hypothesis 3), examining whether
they have independent effects on likelihood of defection        customer knows about alternatives. For instance, informa-
(Figure 1A), whether the effect of knowledge is moderated       tion-processing theories of choice (e.g., Bettman 1979;
by satisfaction level (Figure 1B), or whether knowledge         Fishbein and Ajzen 1975) suggest that a consumer’s evalu-
mediates the effect of satisfaction level on likelihood of      ation of an alternative will depend on the content of the
defection (Figure 1C).                                          consumer’s knowledge (i.e., information pertaining to
                                                                how the alternative performs on decision-relevant
Objective Knowledge About Alternatives                          attributes).
and Likelihood of Defection                                         Another stream suggests that simply having less
                                                                knowledge can influence evaluation. Consumers com-
   Dick and Basu (1994) conceptualized a customer’s             monly make decisions with incomplete knowledge about
decision to defect or not defect as depending on the relative   alternatives (Kivetz and Simonson 2000). Most research
evaluation of the incumbent versus alternatives. This per-      has examined the situation where consumers are missing
spective has been useful in explaining the linkage between      information about a single attribute. Under these condi-
satisfaction and defection—reduced satisfaction dimin-          tions, consumers may infer a value for the missing attrib-
ishes the evaluation of the incumbent, which in turn influ-     ute based on (1) the average value of the attribute across
ences relative evaluation in favor of alternatives, making      other competitors (Ross and Creyer 1992), (2) the overall
defection more likely (Oliver 1997).                            evaluation of the product based on known common attrib-
   Research suggests that the relative evaluation of incum-     utes (Kivetz and Simonson 2000), or (3) the value of a
bent and alternatives can also be influenced by what a          known attribute of the alternative that is perceived to be
166   JOURNAL OF THE ACADEMY OF MARKETING SCIENCE                                                                 SPRING 2003

related to the missing attribute (Broniarczyk and Alba           buyer will not initiate purchase behavior toward an alter-
1994).                                                           native until some requisite level of information about that
    A number of authors report that the inferred value is        alternative has been gathered. Translated to a defection
negatively discounted in evaluating the alternative              context, this suggests that a customer will need to have
(Jaccard and Wood 1988; Johnson and Levin 1985; Meyer            some requisite level of knowledge about an alternative at
1981). Others find that consumers bypass inferring a value       time of choice in order for him or her to consider that alter-
for missing information and simply assign a negative value       native a viable option to defect to. While a customer may
to the missing information (Simmons and Lynch 1991).             have a positive evaluation of an alternative, a perception of
Presumably the likelihood of making inferences decreases         insufficient knowledge about an alternative will preclude
as the amount of missing information increases, making           switching to it.
the assignment of a negative value more likely as the               Factors that result in customers perceiving that there are
amount of missing information increases. Furthermore,            more alternatives about which they have sufficient knowl-
Johnson and Levin (1985) found that when inferences are          edge to defect should increase the likelihood of defection.
made, the magnitude of negative discount increases with          Customers with higher levels of subjective knowledge
the amount of “missing” information. Thus, as the level of       about alternatives will perceive themselves as being more
missing information about alternatives increases, one            knowledgeable about alternatives (Brucks 1985) and thus
would expect a lowering of their evaluation and conse-           should perceive that there are more alternatives for which
quently a reduction in likelihood of defection.1                 they have sufficient knowledge to defect to.
    The amount of “missing” information about alterna-
tives at time of choice will be substantially influenced by       Hypothesis 2: Level of subjective knowledge about alter-
how much the customer knows (objective knowledge)                    natives to the incumbent at the time of choice will be
                                                                     positively associated with likelihood of defection.
about alternatives at initiation of prepurchase search. This
prior knowledge, which reflects the history of the cus-
                                                                 The Relationship Between
tomer’s experience with, passive exposure to, and prior in-
                                                                 Satisfaction and Knowledge
vestigation of alternatives, is a base on which prepurchase
search will build. The larger this base, the less there is to       The above hypotheses implicitly treat knowledge and
learn about alternatives (Brucks 1985; Moorthy,                  satisfaction level as having independent effects on defec-
Ratchford, and Talukdar 1997; Punj and Staelin 1983). In         tion (Model 1, Figure 1A). However, conceivably, the rela-
addition, customers with more prior knowledge more               tionship of these variables to each other and to likelihood
tightly focus their search on gathering information that is      of defection may be more complex.
decision relevant (Alba and Hutchinson 1987; Johnson
                                                                    Model 2: Satisfaction as Moderator Between Level of
and Russo 1984). Having less to learn about alternatives,
                                                                    Knowledge and Likelihood of Defection
and more strongly focused on gathering information rele-
vant to evaluating alternatives, customers with more objec-          Another potential model is that satisfaction level might
tive knowledge about alternatives at initiation of prepurchase   moderate the strength of the relationship between knowl-
search should be “missing” less decision-relevant infor-         edge about alternatives and likelihood of defection (Figure
mation about alternatives at time of choice and therefore        1B). Dick and Basu (1994) pointed out that a customer’s
more likely to defect.                                           attitude toward the incumbent may influence how compet-
                                                                 itor information will be gathered (Alba and Hutchinson
 Hypothesis 1: Level of objective knowledge about alter-         1987), interpreted (Fazio 1990), or evaluated (Cacioppo
    natives at initiation of prepurchase search will be          and Petty 1985). One implication of this is that customers
    positively associated with likelihood of defection.          with a higher satisfaction level (an influence on attitude)
                                                                 might exhibit greater encoding and interpretation biases
Subjective Knowledge About                                       with regard to information about the incumbent’s competi-
Alternatives and Likelihood of Defection                         tors, resulting in the development of knowledge about
                                                                 alternatives that is more biased in favor of the incumbent.
   We have argued that level of objective knowledge about        If so, the relationship between objective knowledge about
alternatives can affect likelihood of defection by influenc-     alternatives and defection might be weaker among cus-
ing evaluation of alternatives. We now turn to level of sub-     tomers who are more satisfied—in other words, an interac-
jective knowledge about alternatives, suggesting that this       tion effect.
type of knowledge will affect defection by influencing the           An interesting further question is how subjective
set of alternatives that consumers will perceive as viable       knowledge and satisfaction might interact in determining
candidates for purchase.                                         likelihood of defection. Customers who are more dissatis-
   Marketing’s seminal studies in buying behavior (e.g.,         fied with the incumbent might be willing to switch to
Engel, Kollat, and Blackwell 1973) have observed that a          another vendor without having to feel as knowledgeable
Capraro et al. / CUSTOMER DEFECTION    167

about alternatives—in other words, dissatisfied customers      subjective knowledge about alternatives at time of choice.
might have a lower requisite level of knowledge needed to      If the level of such knowledge is a predictor of likelihood
defect. Such a reduction in the requisite level of knowledge   of defection, then subjective knowledge at time of choice
needed to defect would have little effect on likelihood of     may mediate the relationship between satisfaction and
defection among customers with a high level of subjective      likelihood to defect. Model 3 (Figure 1C) reflects these
knowledge. However, for customers with a low level of          posited mediating roles of subjective and objective knowl-
subjective knowledge, such a reduction should increase         edge about alternatives.
the likelihood to defect. This reasoning would lead one to         We will test these three competing models of the rela-
posit that the relationship between subjective knowledge       tionship between satisfaction, objective knowledge at the
about alternatives and likelihood to defect will be weaker     initiation of prepurchase search on defection, and subjec-
among more dissatisfied customers.                             tive knowledge at the time of choice on likelihood of de-
   Alternatively, dissatisfaction might make customers         fection. The relationship between knowledge and
more aware of the potential for negative consequences that     satisfaction on likelihood of defection is as follows:
could result from purchasing alternatives about which one
perceives having too little knowledge. Such a reaction          Hypothesis 3a: Model 1: Knowledge and satisfaction
could make accuracy goals (Bettman, Luce, and Payne                have independent effects on likelihood of defection.
1995) more salient and thus elevate the requisite level of      Hypothesis 3b: Model 2: Satisfaction moderates the ef-
                                                                   fect of knowledge on likelihood of defection.
knowledge that customers need in order to switch vendors.
                                                                Hypothesis 3c: Model 3: Knowledge mediates the rela-
In contrast to the previous reasoning, this would suggest a        tionship between satisfaction and likelihood of
moderating effect in which customers with low levels of            defection.
subjective knowledge would be less likely to defect when
they are less satisfied—in other words, the relationship
between subjective knowledge and likelihood to defect          METHOD
would be stronger among the more dissatisfied. We inves-
tigate potential interactions between satisfaction and the         Hypotheses were tested in the context of a choice of
level of both objective knowledge and subjective knowl-        health insurance plan at a large university. This context
edge in Model 2 (see Figure 1B).                               provided an appropriate test environment for several rea-
                                                               sons. First, a choice of health plan is a decision for which
   Model 3: Level of Knowledge as a Mediator Between Satis-
   faction and Likelihood of Defection                         consumers have been observed to seek out information
                                                               about available alternatives (Gibbs, Sangl, and Burrus
   Recent studies have begun to consider mediators in the      1996). Thus, it appears to be a decision that involves an
relationship between satisfactions and repurchase behav-       information-based evaluation of alternatives.
ior (e.g., Nijssen, Singh, Sirdeshmukh, and Holzmüeller            Second, the highly controlled health plan renewal pro-
2003; Oliver 1999). Here we consider the level of objective    cess at this university offered an opportunity to study
knowledge at initiation of prepurchase search and level of     defection in a context where respondents could be sur-
subjective knowledge at time of choice in such a mediating     veyed at times close to their initiation of prepurchase
role.                                                          search and to their making a final choice. At this univer-
   Bloch, Sherrell, and Ridgeway (1986) observed that          sity, employees select health insurance plans annually dur-
consumers conduct ongoing search between purchases,            ing a prescribed deliberation period that lasts for exactly 1
for instance, to build a store of knowledge for later use.     month. Information about the next year’s plans is unveiled
Given that satisfaction level has been found to be nega-       only at the beginning of this deliberation period—prior to
tively related to search effort (Newman and Staelin 1972),     this, no official information is available. Given these con-
one would expect less satisfied customers to expend more       ditions, the unveiling of information about the next year’s
effort in ongoing search. In turn, greater ongoing search      plan seemed to provide a relatively well-defined popula-
should result in higher levels of objective knowledge about    tion-level marker of the initiation of prepurchase search
alternatives at initiation of prepurchase search. If such      and the closing of the deliberation period a well-defined
knowledge is a determinant of likelihood of defection,         marker as to time of choice.
then some of the effect of satisfaction on likelihood of           Third, the context was one in which consumers’ knowl-
defection may be mediated by the level of objective knowl-     edge about alternatives prior to initiating prepurchase
edge about alternatives at initiation of prepurchase search.   search would represent a base on which knowledge about
   Greater prepurchase search effort has been associated       the coming year’s plan could be built. For instance, an
with higher levels of subjective knowledge (Park et al.        understanding of the previous year’s HMO coverage
1994). Given the above-described relationship between          guidelines and/or procedures would facilitate understand-
satisfaction level and search effort, one would expect that    ing for the next year. However, knowledge of past year’s
less satisfied customers should exhibit higher levels of       plans would not be completely predictive for the following
168   JOURNAL OF THE ACADEMY OF MARKETING SCIENCE                                                                 SPRING 2003

year—benefits offered by a particular provider (e.g.,             Responses were coded correct if the answer was right
deductible levels, situations covered, etc.) changed from         (e.g., answered “Y” when the correct answer was “Y”).
year to year.                                                     Incorrect, “?”, or blank responses were coded as not cor-
    Data were collected in two stages. The first stage began      rect. Objective knowledge about alternative health plans
about 1 month before the “unveiling” of the new health            was operationalized as the total number of correct answers
plans for the upcoming year. Surveys were mailed to a ran-        for the three plans to which the respondent was not cur-
dom sample of 1,000 university staff and faculty members,         rently a subscriber. Raw scores were collapsed into high-
stratified to provide approximately equal numbers of sub-         knowledge and low-knowledge groupings (split at the
scribers to each of the four health insurance providers.          median).
This first stage survey measured respondents’ satisfaction
with their current plan, their perceptions about the general      Subjective Knowledge About Alternatives
risk associated with switching health plans, and their level
of objective knowledge about alternative plans. A follow-            Subjective knowledge was measured by a single item
up was mailed 1 week after the initial mailing. Only              that asked respondents to assess “how well do you under-
responses returned before the unveiling of the new health         stand the coverage provided” for each of the health plans
plan (385 responses) were used.                                   (1 = well, 7 = not well). Level of subjective knowledge
    The second stage consisted of recontacting these 385          about alternatives was operationalized as the sum of sub-
respondents immediately after the close of the health plan        jective knowledge scores for the three plans to which the
deliberation period (a follow-up was sent 1 week later). In       respondent was not a subscriber.
this second stage, respondents were asked how much they
felt they knew about the new plans that had been offered—         Switching Risk
an indicator of their subjective knowledge about alterna-
tives at the time of choice. Again, a large majority of               Burnham (1998) reported that perceptions of risk asso-
responses were received within 3 weeks of the initial mail-       ciated with switching to a new vendor can act as a barrier to
ing. Two hundred thirty-five responses were received,             switching. To control for this effect, we included a single
resulting in a total response rate of more than 23 percent.       item to measure customers’ perceptions of the general
                                                                  riskiness of switching to a different health plan. The
                                                                  degree of this perceived risk was measured by agreement
MEASUREMENT
                                                                  to a single item: “There is risk in changing health plans—
                                                                  you never know what you forgot to ask” (1 = strongly
   Measures were developed either by adapting scales              agree, 7 = strongly disagree).
from the existing literature (e.g., satisfaction) or in concert
                                                                      The measures used for each construct were evaluated
with university health plan administrators and existing
                                                                  according to the procedure outlined by Gerbing and
theory. The items used to measure the following constructs
                                                                  Anderson (1988). Results (see Appendix A) suggest that
are listed in Appendix B.
                                                                  the measures were adequate in terms of unidimensionality,
                                                                  convergent or discriminant validity, and reliability.
Satisfaction                                                      Response bias was analyzed from several perspectives
                                                                  (Armstrong and Overton 1977). Defection rates in
   Satisfaction with the current health insurance plan was        returned responses (~4%) were similar to those in the sub-
measured using a four-item scale adapted from Oliver              scriber population (~5%). Furthermore, the results suggest
(1980). Each item used a 7-point scale, ranging from              little early or late response bias. Response rates varied
strongly agree to strongly disagree. A composite satisfac-        somewhat across health plans—the PPO plan (all others
tion score was developed by summing the responses to the          were HMO) had the lowest response rate (13%) and the
four items.                                                       lowest defection rate (~1%). Satisfaction levels across
                                                                  plans were indistinguishable.
Objective Knowledge About Alternatives

   Consistent with Brucks (1985) and Park et al. (1994),          RESULTS3
level of objective knowledge about alternatives was mea-
sured as the number of correct responses to 13 questions             There were 13 defections in the 222 responses used in
asked about features in the various health plans offered for      the following analysis. Descriptive data for the variables in
the current year.2                                                this study are displayed in Appendix C. The data suggest
   Respondents were asked to answer “Y,” “N,” or “?” to           that most respondents were satisfied with their current
indicate whether each plan offered a particular feature.          health plan (median score of 20 out of a possible 28) and
Capraro et al. / CUSTOMER DEFECTION           169

                                                         TABLE 1
                              Relationship Between Level of Knowledge About Alternatives,
                                    Satisfaction, and Likelihood of Defection (N = 222)
                                                                                          b             p                     Model Fit
Baseline model
   Satisfaction                                                                       –.09             .09             –2 LL = 92.8, p = .045
                                                                                                                        2
   Switching risk                                                                     –.38             .03             R = .08
Model 1—main effects of knowledge
   Satisfaction                                                                       –.10             .07             –2 LL = 78.0, p = .000
                                                                                                                        2
   Switching risk                                                                     –.41             .02             R = .25
                        a
   Objective knowledge                                                                1.47             .04
                          b
   Subjective knowledge                                                                .20             .008
Model 2—interaction between satisfaction and knowledge
   Satisfaction                                                                       –.22             .47             –2 LL = 76.7, p = .000
                                                                                                                        2
   Switching risk                                                                     –.39             .04             R = .27
                        a
   Objective knowledge                                                                3.45             .25
   Subjective knowledgeb                                                               .52             .17
                         a
   Objective Knowledge × Satisfaction                                                 –.10             .48
   Subjective Knowledgeb × Satisfaction                                               –.02             .19
Model 3—knowledge as mediator between satisfaction and likelihood of defection
   Satisfaction → Likelihood of defection                                             –.09             .09
                                               a
   Satisfaction      → Objective knowledge                                            –.002            .95
   Satisfaction      → Subjective knowledgeb                                           .11             .16
                                                                                                           c

   Satisfaction             →                                                         –.10             .07
   Objective knowledgea → → Likelihood of defection                                   1.47             .04
                          b
   Subjective knowledge →                                                              .20             .008

NOTE: LL = log likelihood.
a. At initiation of prepurchase search.
b. At time of choice.
c. Two-tailed test.

that respondents view the action of switching to a new                                        TABLE 2
health plan as risky (median score of 6 out of 7). As                      Distribution of Defections Across Satisfaction
expected, respondents had less than complete objective                             Level and Level of Knowledge
knowledge about current alternatives (median score of 4                                                A. Frequency of Defection Across
out of a possible 39) but had greater objective knowledge                                            Satisfaction and Objective Knowledge
about their own plan (median of 6.0 out of a possible score                                         High Objective           Low Objective
of 13). Consistent with Brucks (1985), subjective and                                              Knowledge About          Knowledge About
objective knowledge were positively correlated (r = .2, p =                                         Alternatives at          Alternatives at
.004). None of the other variables were significantly corre-                                         Initiation of            Initiation of
                                                                                                  Prepurchase Search       Prepurchase Search
lated with each other.
                                                                          Not satisfied/neutral       4/23 (17%)a               0/26 (0%)
                                                                                                                                          a
    A hierarchical logistic regression procedure was used                                                       a                         a
                                                                          Satisfied                   6/81 (7%)                 3/92 (3%)
to test Hypotheses 1 and 2. The first step estimated a base-
line model using satisfaction and switching risk as predic-                                            B. Frequency of Defection Across
tors of defection. A second step (the main effects model or                                          Satisfaction and Subjective Knowledge
Model 1) added level of objective knowledge at initiation                                           High Subjective         Low Subjective
of prepurchase search and level of subjective knowledge at                                         Knowledge About         Knowledge About
time of choice to the baseline model.                                                               Alternatives at         Alternatives at
                                                                                                    Time of Choice          Time of Choice
    As shown in Table 1, the baseline model4 adequately
fits the data (–2 log likelihood = 92.8, p = .045), explaining            Not satisfied/neutral       4/29 (14%)a               0/20 (0%)
                                                                                                                                          a
                                                                                                                a                         a
                                                                          Satisfied                   9/99 (9%)                 0/74 (0%)
8 percent of the variation in likelihood of defection
(Nagelkerke 1991). Consistent with past research, the                     a. Number of defections/number of respondents in cell (% defections).
defection rate (see Table 2, part A) is higher among the
unsatisfied (4 out of 49, or 8%) than the satisfied (9 out of
173, or 5%). Furthermore, as shown in Table 1, the coeffi-                switching risk (recoded so that 7 represents high per-
cient for satisfaction (bsatisfaction = –.09) is negative and sig-        ceived risk) is negative and significant (bswitching risk = –.38,
nificant at the p = .09 level. The coefficient for perceived              p = .03). Thus, higher satisfaction and greater perceived
170    JOURNAL OF THE ACADEMY OF MARKETING SCIENCE                                                                             SPRING 2003

switching risk are associated with lower likelihood of                  likelihood of defection and as a moderator in the relation-
defection.                                                              ship between subjective knowledge at time of choice and
                                                                        likelihood of defection.
Knowledge About Alternatives                                                The results do not support Hypothesis 3b (see Model 2
and Likelihood of Defection                                             in Table 1). Neither the interaction term between satisfac-
                                                                        tion and objective knowledge (bobjective knowledge × satisfaction =
     Model 1 tests (see Table 1) for main effects of satisfac-          –.10, p = .48) nor the interaction term between satisfaction
tion, perceived switching risk, objective knowledge, and                and subjective knowledge (bsubjective knowledge × satisfaction = –.02,
subjective knowledge. As with the baseline model, satis-                p = .19) is significant. Furthermore, the fit of Model 2,
faction and perceived switching risk are negatively related             while adequate (–2 LL = 76.7, p = .000), is not signifi-
to likelihood of defection (bsatisfaction = –.10, p = .07; bswitching   cantly better than the main effects model (Ξ2difference = 1.3, 2
risk = –.41, p = .02). Supporting Hypothesis 1, the coeffi-             df, p = .43). Based on moderated regression analysis crite-
cient for objective knowledge (bobjective knowledge = 1.47, p =         ria (Aiken and West 1991), we conclude that Model 2
.04) is positive and significant. Thus, after controlling for           (Hypothesis 3b) is not supported.
satisfaction and perceived switching risk, higher levels of                 Model 3 in Table 1 reports on the test of Hypothesis 3c
objective knowledge at initiation of prepurchase search                 (consumer knowledge mediates the effect of satisfaction
are associated with greater likelihood to defect.                       on likelihood of defection). Baron and Kenny’s (1986)
     The distribution of defections with regard to satisfac-            procedure, adapted for two mediators (Shapiro and Spence
tion level and level of objective knowledge about alterna-              2002) was used. First, satisfaction5 was regressed against
tives prior to prepurchase search is reported in Table 2, part          likelihood to defect. As shown, satisfaction is a predictor
A. The defection rate among those with a high level of                  of likelihood of defection (bsatisfaction = –.09), significant at
objective knowledge about alternatives (10 out of 104, or               the p = .09 level. In the next stage of the test, satisfaction
10%) is higher than among those with a low level (3 out of              was regressed against objective knowledge at initiation of
118, or 3%). Among the unsatisfied, all defectors exhib-                prepurchase search and subjective knowledge at time of
ited a high level of objective knowledge about alternatives.            choice in separate analyses. As can be seen from the mid-
     Supporting Hypothesis 2 (Model 1, Table 1), the coeffi-            dle section of the results pertaining to Model 3, satisfac-
cient for subjective knowledge (bsubjective knowledge = .20, p =        tion is not a significant predictor of either level of objective
.008) is positive and significant, indicating that after con-           knowledge at initiation of prepurchase search (bobjective knowl-
trolling for satisfaction and perceived switching risk,                 edge = –.002, p = .95) or level of subjective knowledge at
higher levels of subjective knowledge at time of choice are             time of choice (bsubjective knowledge = .11, p = .16). This finding
associated with greater likelihood to defect. Table 2, part B           violates a necessary condition for mediation—that satis-
reports the distribution of defections with regard to satis-            faction must be a predictor of the posited mediating
faction level and subjective knowledge at the time of                   variables.
choice (above or below the median). Note that only those                    Finally, comparing the top and bottom sections
with a high level of subjective knowledge defected.                     depicted in Model 3, the coefficient for satisfaction is
     Furthermore, adding objective and subjective knowl-                essentially the same whether both objective knowledge at
edge improves the model fit (–2LL = 78.0, p = .000) rela-               initiation of prepurchase search and subjective knowledge
tive to the baseline model (Ξ2difference = 14.8, 2 df, p < .000)        at time of choice are included as predictors of likelihood of
and raises the R2 from .08 to .25. Thus, going beyond con-              defection (bsatisfaction = –.1, p = .07) or whether they are omit-
sumers’ satisfaction and perceived switching risk to con-               ted (bsatisfaction = –.09, p = .09). This again violates the condi-
sider consumers’ objective and subjective knowledge sig-                tions necessary for mediation. Thus, Model 3 (Hypothesis
nificantly enhances our ability to predict likelihood of                3c) is not supported.
defection.                                                                  In summary, our competing model tests of the relation-
                                                                        ship between knowledge, satisfaction, and likelihood of
Relationship Between Knowledge and                                      defection support Model 1 (Hypothesis 3a)—objective
Satisfaction and Likelihood of Defection                                and subjective knowledge appear to have independent
                                                                        direct effects on consumers’ likelihood of defecting,
   The support for Hypotheses 1 and 2 provides evidence                 beyond effects associated with satisfaction level.
for independent effects of knowledge and satisfaction on
likelihood of defection, as suggested by Hypothesis 3a.
Competing models of the relationship between knowledge                  DISCUSSION
and satisfaction on likelihood of defection were also
tested. Model 2 tested Hypothesis 3b, adding satisfaction                  Recently, researchers have begun to consider factors
as a moderator in the relationship between objective                    beyond satisfaction that may influence whether a customer
knowledge at initiation of prepurchase search and                       will defect. This article contributes to this stream, focusing
Capraro et al. / CUSTOMER DEFECTION      171

on “how much customers know about alternatives” as a            satisfaction, these two customers would be equally likely
determinant of likelihood of defection. Specifically, it sug-   to defect. However, the first customer’s drop in satisfac-
gests that for repurchase decisions where customers con-        tion is likely to have motivated him or her to investigate
sider alternatives and make information-based decisions,        (search) alternatives during the period of dissatisfaction
likelihood of defection will be influenced by the level of      (Newman and Staelin 1972). That fluctuation will have
two kinds of knowledge—subjective knowledge and                 left that customer with more knowledge about alternatives
objective knowledge about alternatives. These two types         as prepurchase search begins—a position that our findings
of knowledge appear to have independent effects and             suggest will make the customer more vulnerable at the
together account for about twice as much variance in like-      next time of choice. And even if the customer does not
lihood of defection as satisfaction and perceived switching     defect at that time, such knowledge about alternatives will
risk.                                                           render customers more capable of learning about alterna-
   The mediation hypotheses were not supported. While a         tives for future purchase opportunities (Alba and Hutchin-
satisfaction-knowledge-likelihood of defection link is the-     son 1987).
oretically plausible, little of the effect of satisfaction on       Previous authors have suggested that it may be possible
likelihood of defection is mediated by knowledge about          to influence customers’ buying behavior by influencing
alternatives. Furthermore, no evidence was found for an         the content of the information that customers attend to and
interaction between satisfaction and level of knowledge         perceive (e.g., Hoch and Deighton 1989; Kivetz and
about alternatives. Of course, additional power may have        Simonson 2000). We broaden this thinking to suggest that
allowed us to detect the posited mediation and interaction      even if it is not possible to influence the content of what
effects. But from a practical perspective, our results sug-     customers know about alternatives, defection may be
gest that both level of objective knowledge at initiation of    reduced if one can simply influence customers to know
prepurchase search and level of subjective knowledge at         less about alternatives.
time of choice primarily operate on likelihood to defect as         Interestingly, there may be potential to do so. Research
main effects.                                                   suggests that the search effort that a consumer exerts
   Although we believe this study is the first to systemati-    depends on the trade-off between the projected costs and
cally investigate how a customer’s level of knowledge           the benefits of search (Stigler 1961). To the extent that
about alternatives influences likelihood of defection, pre-     managers could induce customer perceptions that it will be
vious studies have recognized that “level of knowledge          difficult (and/or less rewarding) to gather information
about competitors” plays a role in defection. Klemperer’s       about alternatives, customers should reduce their investi-
(1987) and Fornell’s (1992) characterizations of a lack of      gation of alternatives, know less about alternatives, and
knowledge about competitors as a switching barrier seem         ultimately be less likely to defect. While specific manage-
to recognize that level of knowledge about alternatives         rial actions that could influence customers’ investigation
plays a role in defection. Similarly, Oliver’s (1999) asser-    of alternatives are a matter for future research, it seems that
tion that highly loyal customers become less vulnerable, in     actions such as designing one’s offering to make direct
part because they “tune out” competitive overtures, also        comparison with alternatives more difficult (increase the
seems consistent with our view that customers who know          cost of search) or actions to ensconce oneself as the cus-
less about alternatives will be less likely to defect.          tomer’s trusted and primary source of product class infor-
   Consideration of level of knowledge about alternatives       mation (reduce the projected benefit of search) seem
advances our understanding of defection. For instance, it       promising.
helps to explain why dissatisfied customers sometimes do
not defect. Although dissatisfaction may tend to shift rela-
tive evaluation in a way that disfavors the incumbent, if a     LIMITATIONS AND FUTURE RESEARCH
customer does not know enough about alternatives at the
time of choice to defect or discounts alternatives due to          There are some limitations that should be noted in inter-
missing information, defection may not occur.                   preting this study. For instance, while our view of the role
   It also carries implications as to how managers might        of knowledge about alternatives in defection is intended to
assess and manage their customers’ vulnerability to defec-      be generally relevant to repurchase decisions that involve
tion. For instance, consideration of level of knowledge         an information-based evaluation of alternatives, such
about alternatives suggests that a customer’s vulnerability     decisions are more likely to occur with purchases that con-
will depend not only on satisfaction level but also on how a    sumers consider important. Knowledge about alternatives,
customer has reached that level of satisfaction. Consider       for instance, would not be expected to play a role in con-
two customers about to initiate prepurchase search—one          texts where consumers use low-effort decision strategies
whose satisfaction level has fluctuated from high to low to     such as choice tactics (Hoyer 1984).
high since the last purchase and a second whose satisfac-          Furthermore, our results reflect only a single context.
tion level has been consistently high. Looking only at          Certainly that context is substantial, with more than $250
172   JOURNAL OF THE ACADEMY OF MARKETING SCIENCE                                                                             SPRING 2003

billion paid in private health care insurance premiums in                impact on likelihood of defection in a pure credence or
the United States in 1997 (National Committee for Qual-                  experience good than in health insurance—a hybrid of
ity Assurance 2001). Of course, not every health plan deci-              experience and search qualities.
sion involves an evaluation of alternatives. However, a                     Three measurement limitations should be noted. First,
substantial proportion apparently does (Gibbs et al. 1996).              subjective knowledge and switching risk were measured
Given the high customer acquisition costs in the health                  by a single item. While our single items seem to tap into
insurance industry ($200 to $400 per member), even if our                the central construct sampled by other studies’ multi-item
findings could lead to only a small reduction in rate of                 scales (Campbell and Goodstein 2001; Park et al. 1994),
defection, that reduction would have a substantial profit                multiple items would have been preferable. Second, it
impact (Wood 1999).                                                      would be valuable to reproduce this study in a context
    Replication in other contexts is needed to assess the                where there is a higher rate of defection. Although our
generalizability of our findings and conclusions. For                    200+ sample size is adequate for logistic regression tech-
instance, our findings reflect an all-or-none defection/                 niques, the small number of observed defections limits our
nondefection decision. While such decisions are common                   statistical power to detect effects. This lack of power, how-
(e.g., automobile purchases, choice of school), there are                ever, seems to suggest robustness in the relationships we find.
other kinds of decisions for which customers can switch                     Finally, our design leaves open the possibility that satis-
vendors for only a portion of their purchases (e.g., opening             faction level might have changed between the measure of
an account with a new stock broker). If customers view                   satisfaction and time of choice (~1.5 months) and that a
such a decision as less consequential, knowledge about                   satisfaction measure taken closer to time of choice would
alternatives may play a smaller role in defection.                       account for more of the variance in defection. However,
    It would also be interesting to explore the role of knowl-           given the historically low defection rates in health plans at
edge about alternatives in defection for more purely expe-               this institution (~5%), it seems likely that satisfaction level
rience or credence goods. By definition, there is more                   would not have substantially changed in the 1 to 2 months
uncertainty about the performance of an experience or cre-               between the measure of satisfaction and the choice of the
dence good than a search good. If customers are accus-                   next year’s health plan. Consequently, we believe that a
tomed to more uncertainty in choosing the former kinds of                later measure satisfaction of level would have had little
goods, then a lack of subjective knowledge might have less               effect on our results.

                                                          APPENDIX A
                                                        Measurement Model

                                                                                     Subjective           Objective
                               Satisfaction                       Risk                                    Knowledge
                                                                                     Knowledge

                    S1          S2         S3          S4           R1                   SK                  OK

  Unidimensionality: LISREL modification indexes suggested that one satisfaction indicator might also load on the subjective knowl-
     edge construct as well. However, the standardized loading of this indicator on subjective knowledge was eight times weaker than on
     satisfaction.
  Reliability: Cronbach’s alpha for the satisfaction measure was .85—substantially above the recommended level of .7 (Nunnally 1978).
     Reliability cannot be assessed for the single-item measures.
  Discriminant validity: Examination of the Phi correlations indicated that all were significantly different from 1, with the largest corre-
     lation plus twice the standard error being .35.
  Convergent validity: All satisfaction items had a significant loading on the satisfaction construct, with the lowest t-value in the Lambda
     matrix being 10.0—evidence of adequate convergent validity (Sujan, Weitz, and Kumar 1994) of the satisfaction items.
  Overall fit: Error variances for satisfaction are calculated by LISREL. Estimates of error for the remaining constructs are set at 10 per-
                                                                                                       2
     cent of the sample variance (Hayduk 1987). Although the goodness-of-fit test is significant (Ξ = 20.0, 11 df, p = .045), other indica-
     tors suggest adequate fit, with Goodness-of-Fit Index, Adjusted Goodness-of-Fit Index, Nonnormed Fit Index, Normed Fit Index,
     and Comparative Fit Index at .9 or above.
Capraro et al. / CUSTOMER DEFECTION   173

                                                          APPENDIX B
                                                       Items Used in Study

Satisfaction Items (α = .85)

          Strongly                                                                                                   Strongly
            Agree                                                                                                    Disagree
               1                  2              3               4                5               6                      7
I am satisfied with my current health plan.
I am satisfied with the way my plan handles financial matters (e.g., billings, reimbursements).
I have been pleased with my health plan’s response when I have a question or complaint.
I am confident that my health plan will provide the care I need whenever I need it.

Objective-Knowledge Questions (predeliberation period)

Indicate whether each plan has the following features (circle “Y,” “N,” or “?”)
Annual Well Woman Exam
Prescription Coverage Through Caremark
Access to Emergency Room Treatment, but Must Notify Primary Care Physician Within 48 Hours
Access to Doctors Affiliated With Austin Regional Clinic (ARC)
24-Hour Access to Medical Personnel
Access to Doctors Affiliated With Austin Diagnostic Clinic (ADC)
Use of HMO Facilities or Choice of Any Doctor or Hospital
Obtain a 90-Day Supply of Prescription Medicine and Incur Only a Single Co-Pay
Ability to Use Austin Diagnostic Medical Center as a Hospital
Preexisting Condition Restrictions Waived During Enrollment Period
Use of Seton Medical Center as a Hospital
Counseling for Mental Health and Substance Abuse Problems
Use of St. David’s as a Hospital Facility

Subjective Knowledge About Alternatives (postpurchase)

             Well                                                                                                    Not Well
              1                  2             3              4              5                6                         7
Please rate how well you understand the coverage provided by each of the following health plans.

Switching Risk

           Strongly                                                                                                  Strongly
            Agree                                                                                                    Disagree
               1                   2           3             4              5                     6                      7
There is risk in changing health plans—you never know what you forgot to ask.

                                                         APPENDIX C
                                                 Descriptive Statistics (N = 222)

                                          Low               High Response/        Median               Mean        Standard
                                        Response           Maximum Possible      Response             Response     Deviation
Satisfaction level                          4                    28/28               20                  19.6         5.0
Switching risk                              1                     7/7                 6                   5.8         1.4
Objective knowledge of alternatives         0                    28/39                4                   5.7         6.0
Subjective knowledge of alternatives        3                    21/21               12                  12.0         5.8
Objective knowledge of own plan             0                    11/13                6                   5.8         2.6
Subjective knowledge of own plan            1                     7/7                 6                   5.8         1.4
174    JOURNAL OF THE ACADEMY OF MARKETING SCIENCE                                                                                       SPRING 2003

NOTES                                                                          Baron, Reuben M. and David A. Kenney. 1986. “The Moderator-Mediator
                                                                                   Variable Distinction in Social Psychological Research: Conceptual,
                                                                                   Strategic and Statistical Considerations.” Journal of Personality and
    1. The effect of missing information on likelihood of defection may            Social Psychology 51 (6): 1173-1182.
not be the same for all customers. In a situation where preferences are het-   Bettman, James R. 1979. An Information Processing Theory of Con-
erogeneous and offerings are differentiated, we expect customers to fall           sumer Choice. Reading, MA: Addison-Wesley.
into two groups.                                                               , Mary Frances Luce, and John W. Payne. 1995. “Constructive
    For some, a reduction in missing information about alternatives will           Consumer Choice Processes.” Journal of Consumer Research 25
lead to improved evaluation of at least some alternatives. While custom-           (December): 187-217.
ers missing substantial amounts of information about alternatives will be      Bloch, Peter H., Daniel L. Sherrell, and Nancy M. Ridgeway. 1986.
unlikely to defect (due to discounting), those missing less will be more           “Consumer Search: An Extended Framework.” Journal of Consumer
likely to defect. Overall, the level of missing information at time of             Research 13 (June): 119-127.
                                                                               Bolton, Ruth N. 1998. “A Dynamic Model of the Duration of the Cus-
choice will be negatively related to likelihood of defection.
                                                                                   tomer’s Relationship With a Continuous Service Provider: The Role
    For others, a reduction in missing information about alternatives may          of Satisfaction.” Marketing Science 17 (1): 45-65.
not lead to improved evaluations of alternatives. This could occur among       Broniarczyk, Susan M. and Joseph W. Alba. 1994. “The Role of Con-
customers whose evaluation of what is learned about alternatives as the            sumers’ Intuitions in Inference Making.” Journal of Consumer Re-
level of missing information declines is negative enough to balance or             search 21 (December): 393-407.
outweigh any positive effect due to reduced discounting. Again, custom-        Brucks, Merrie. 1985. “The Effects of Product Class Knowledge on In-
ers missing a lot of information about alternatives will be unlikely to de-        formation Search Behavior.” Journal of Consumer Research 12
fect; however, here, neither will those with higher levels. Thus, we expect        (June): 1-16.
no relationship between level of missing information at time of choice         Burnham, Thomas Adams. 1998. “Measuring and Managing Consumer
                                                                                   Switching Costs to Improve Customer Retention in Continuous Ser-
and likelihood to defect.
                                                                                   vices.” Unpublished doctoral dissertation. The University of Texas at
    An absence of relationship for one group of customers and a negative           Austin.
relationship for the other group will, in the aggregate, manifest as an        Cacioppo, John T. and Richard E. Petty. 1985. “Central and Peripheral
overall negative relationship. Thus, in aggregate, we expect amount of             Routes to Persuasion: The Role of Message Repetition.” In Psycho-
missing information about alternatives at time of choice to be negatively          logical Processes and Advertising Effects: Theory, Research and Ap-
related to likelihood of defection.                                                plication. Eds. Linda F. Alwilt and Andrew A. Mitchell. Hillsdale,
    2. Features were chosen as test items based on two considerations.             NJ: Lawrence Erlbaum, 91-111.
First, according to university health plan administrators, each feature rep-   Campbell, Margaret C. and Ronald C. Goodstein. 2001. “The Moder-
                                                                                   ating Effect of Perceived Risk on Consumers’ Evaluations of Product
resented an important consideration in subscribers’ evaluations of health
                                                                                   Incongruity: Preference for the Norm.” Journal of Consumer Re-
plans. Second, the features chosen were differentially represented in the          search 28 (December): 439-449.
various health plans—plans differed on 9 out of the 13 attributes used in      Coyne, Kevin. 1989. “Beyond Service Fads—Meaningful Strategies for
this measure. No two plans were the same on all features. A tabulation of          the Real World.” Sloan Management Review 30 (summer): 69-76.
how plans compared on the features used to test objective knowledge is         Cronin, J. J. and S. A. Taylor. 1992. “Measuring Service Quality: A Reex-
available from the first author of this article. The set of available health       amination and Extension.” Journal of Marketing 56 (July): 55-68.
care providers remained stable during the course of this study.                Dick, Alan S. and Kunal Basu. 1994. “Customer Loyalty: Toward an Inte-
    3. For clarity’s sake, the results presented below reflect a recoding          grated Conceptual Framework.” Journal of the Academy of Mar-
such that higher levels of satisfaction, switching risk, and subjective            keting Science 22 (winter): 99-113.
knowledge are reflected by higher ratings.                                     Engel, James F., David T. Kollat, and Roger D. Blackwell. 1973. Con-
                                                                                   sumer Behavior. New York: Holt, Rinehart & Winston.
    4. One of our reviewers suggested that, given the nonlinearities that      Fazio, Russell H. 1990. “Multiple Processes by Which Attitudes Guide
have been observed between satisfaction and defection, we might try to             Behavior: The MODE Model as an Integrative Framework.” In Ad-
use a quadratic term in the regression models. Adding this term to the             vances in Experimental Social Psychology, Vol. 23. Ed. Mark P.
models does improve the variance explained by about 7 percent (signifi-            Zanna. New York: Academic Press, 75-109.
cant at the .01 level). However, the addition of the satisfaction-squared      Fishbein, Martin and Icek Ajzen. 1975. Belief, Attitude, Intention and Be-
term drove the coefficient for the linear satisfaction term to insignifi-          havior: an Introduction to Theory and Research. Reading, MA: Ad-
cance—perhaps due to multicollinearity. Since adding the quadratic term            dison-Wesley.
had no impact on the coefficients for either objective knowledge or per-       Fornell, Claes. 1992. “A National Customer Satisfaction Barometer: The
ceived risk of switching, we present models with only the linear term.             Swedish Experience.” Journal of Marketing 56 (January): 6-21.
                                                                               Gerbing, David W. and James C. Anderson. 1988. “An Updated Para-
    5. All regressions included perceived switching risk as a control vari-        digm for Scale Development Incorporating Unidimensionality and
able. The coefficients for perceived switching risk are not included since         Its Assessment.” Journal of Marketing Research 25 (May): 186-192.
they are not relevant to the test for mediation.                               Gibbs, Deborah A., Judith A. Sangl, and Judith A. Burrus. 1996. “Con-
                                                                                   sumer Perspectives on Information Needs for Health Plan Choice.”
                                                                                   Health Care Financing Review 18 (1): 55-73.
                                                                               Hayduk, L. A. 1987. Structural Equation Modeling With LISREL. Balti-
REFERENCES                                                                         more, MD: Johns Hopkins University.
                                                                               Heide, Jan B. and Allen M. Weiss. 1995. “Vendor Consideration and
Aiken, Leona S. and Steven G. West. 1991. Multiple Regression: Testing             Switching Behavior for Buyers in High-Technology Markets.” Jour-
   and Interpreting Interactions. Newbury Park, CA: Sage.                          nal of Marketing 59 (July): 30-43.
Alba, Joseph W. and Wesley Hutchinson. 1987. “Dimensions of Con-               Hennig-Thurau, Thorsten and Alexander Klee. 1997. “The Impact of
   sumer Expertise.” Journal of Consumer Research 13 (March): 411-                 Customer Satisfaction and Relationship Quality on Customer Reten-
   454.                                                                            tion: A Critical Reassessment and Model Development.” Psychology
Anderson, E. W. and M. W. Sullivan. 1993. “The Antecedents and Conse-              and Marketing 14 (8): 737-764.
   quences of Customer Satisfaction for Firms.” Marketing Science 12           Hoch, Stephen J. and John Deighton. 1989. “Managing What Consumers
   (2): 125-143.                                                                   Learn From Experience.” Journal of Marketing 53 (2): 1-20.
Armstrong, J. Scott and Terry S. Overton. 1977. “Estimating                    Hoyer, Wayne D. 1984. “An examination of Consumer Decision Making
   Nonresponse Bias in Mail Surveys.” Journal of Marketing Research                for a Common Repeat Purchase Product.” Journal of Consumer Re-
   14 (August): 396-402.                                                           search 11 (December): 822-829.
Capraro et al. / CUSTOMER DEFECTION            175

Jaccard, James and Gregory Wood. 1988. “The Effects of Incomplete In-          and W. Earl Sasser. 1990. “Zero Defections: Quality Comes to
    formation on the Formation of Attitudes Toward Behavioral Alterna-            Services.” Harvard Business Review 68 (September): 105-111.
    tives.” Journal of Personality and Social Psychology 54 (4): 580-591.     Ross, William T. and Elizabeth H. Creyer. 1992. “Making Inferences
Johnson, Eric J. and Edward J. Russo. 1984. “Product Familiarity and              About Missing Information: The Effects of Existing Information.”
    Learning New Information.” Journal of Consumer Research 11 (1):               Journal of Consumer Research 19 (June): 14-25.
    542-551.
                                                                              Sambandam, Rajan and Kenneth R. Lord. 1995. “Switching Behavior in
Johnson, Richard D. and Irwin P. Levin. 1985. “More Than Meets the
                                                                                  Automobile Markets: A Consideration Sets Model.” Journal of the
    Eye: The Effect of Missing Information on Purchase Evaluations.”
                                                                                  Academy of Marketing Science 23 (winter): 57-65.
    Journal of Consumer Research 12 (September): 169-177.
Jones, Michael A., David L. Mothersbaugh, and Sharon E. Beatty. 2000.         Shapiro, Stewart and Mark T. Spence. 2002. “Factors Affecting En-
    “Switching Barriers and Repurchase Intentions in Services.” Journal           coding, Retrieval, and Alignment of Sensory Attributes in a Memory-
    of Retailing 76 (2): 259-274.                                                 Based Brand Choice Task.” Journal of Consumer Research 28 (4):
Jones, Thomas O. and W. Earl Sasser Jr. 1995. “Why Satisfied Customers            603-617.
    Defect.” Harvard Business Review 73 (November-December): 88-              Simmons, Carolyn J. and John G. Lynch. 1991. “Inference Effects With-
    99.                                                                           out Inference Making? Effects of Missing Information on Dis-
Kivetz, Ran and Itamar Simonson. 2000. “The Effects of Incomplete In-             counting and Use of Presented Information.” Journal of Consumer
    formation on Consumer Choice.” Journal of Marketing Research 37               Research 17 (March): 477-491.
    (November): 427-448.                                                      Stigler, George. 1961. “The Economics of Information.” Journal of Polit-
Klemperer, Paul. 1987. “Markets With Consumer Switching Costs.” The               ical Economy 69 (June): 213-225.
    Quarterly Journal of Economics 102 (May): 375-394.                        Sujan, Harish, Barton A. Weitz, and Nirmalya Kumar. 1994. “Learning
LaBarbera, Priscilla A. and David. Mazursky. 1983. “A Longitudinal As-            Orientation, Working Smart and Effective Selling.” Journal of Mar-
    sessment of Consumer Satisfaction/Dissatisfaction: The Dynamic                keting 58 (July): 39-52.
    Aspect of the Cognitive Process.” Journal of Marketing Research 20
    (November): 393-404.                                                      Wood, Steven D. 1999. “Strategies for Improving Health Plan Reten-
Meyer, Robert J. 1981. “A Model of Multiattribute Judgments Under At-             tion.” Journal of the Healthcare Financial Management Association
    tribute Uncertainty and Informational Constraint.” Journal of Mar-            Resource Guide:1-5.
    keting Research 18 (November): 428-441.
Mittal, Vikas and Wagner A. Kamakura. 2001. “Satisfaction, Repurchase
    Intent, and Repurchase Behavior: Investigating the Moderating Ef-         ABOUT THE AUTHORS
    fect of Customer Characteristics.” Journal of Marketing Research 38
    (February): 131-143.
Moorthy, Dhar, Brian T. Ratchford, and Sabrata Talukdar. 1997. “Con-          Anthony J. Capraro (tcapraro@unca.edu), an assistant profes-
    sumer Information Search Revisited: Theory and Empirical Analy-           sor at the University of North Carolina at Asheville, earned his
    sis.” Journal of Consumer Research 23 (March): 263-277.
Nagelkerke, N. J. D. 1991. “A Note on a General Definition of the Coeffi-     Ph.D. in marketing in 1999 from the University of Texas after
    cient of Determination.” Biometrika 78:691-692.                           having spent 20 years in industry in marketing and marketing
National Committee for Quality Assurance. 2001. “Managed Care and             management positions. His current research interest focuses on
    the U.S. Health Care Industry.” Retrieved July 22, 2002, from http://     developing and enhancing the value of a firm’s customer base.
    www.ncqa.org/somc2001/intro/somc_2001_industry.htm
Newman, Joseph W. and Richard Staelin. 1972. “Prepurchase Informa-
    tion Seeking for New Cars and Major Household Appliances.” Jour-          Susan Broniarczyk (Susan.Broniarczyk@bus.utexas.edu), an
    nal of Marketing Research 9 (August): 249-257.                            associate professor at the University of Texas at Austin, earned
 and Richard A. Werbel. 1973. “Multivariate Analysis of Brand              her Ph.D. in marketing from the University of Florida. She serves
    Loyalty for Major Household Appliances.” Journal of Marketing Re-         on the editorial boards of the Journal of Consumer Research and
    search 10 (November): 404-409.
Nijssen, Edwin, Jagdip Singh, Deepak Sirdeshmukh, and Hartmut                 the Journal of Marketing Research and the advisory board for the
    Holzmüeller. 2003. “Investigating Industry Context Effects in Consumer-   Association for Consumer Research. Her research, which exam-
    Firm Relationships: Preliminary Results From a Dispositional Ap-          ines consumer decision making and how consumers’ knowledge
    proach.” Journal of the Academy of Marketing Science 31 (winter):         structures affect their reaction to missing or conflicting product
    46-60.
Nunnally, Jum. 1978. Psychometric Theory. New York: McGraw-Hill.
                                                                              information, appears in the Journal of Consumer Research, the
Oliva, Terrence, Richard L. Oliver, and Ian C. MacMillan. 1992. “A Ca-        Journal of Marketing Research, and Organizational Behavior
    tastrophe Model for Developing Service Satisfaction Strategies.”          and Human Decision Processes.
    Journal of Marketing 56 (July): 83-95.
Oliver, Richard L. 1980. “A Cognitive Model of the Antecedents and
    Consequences of Satisfaction Decisions.” Journal of Marketing Re-
                                                                              Rajendra K. Srivastava (Rajendra.Srivastava@bus.utexas.edu)
    search 17 (November): 460-469.                                            is the Jack R. Crosby Regent’s Chair in Business and a professor
. 1997. Satisfaction: A Behavioral Perspective on the Consumer.            of marketing and management science and information systems
    Boston: Irwin.                                                            (MSIS) in the McCombs School of Business at the University
. 1999. “Whence Consumer Loyalty.” Journal of Marketing 63                 Texas at Austin. He is also the Daniel J. Jordan Research Scholar
    (Special Issue): 33-44.
                                                                              at Emory University. He earned his doctorate from the University
Park, Whan C., David L. Mothersbaugh, and Lawrence Feick. 1994.
    “Consumer Knowledge Assessment.” Journal of Consumer Re-                  of Pittsburgh. His research, which spans marketing and finance,
    search 21 (June): 71-82.                                                  has been published in the Journal of Marketing, the Journal of
Punj, Girish N. and Richard Staelin. 1983. “A Model of Consumer Infor-        Marketing Research, Marketing Science, and the Journal of
    mation Search for New Automobiles.” Journal of Consumer Re-               Banking and Finance. His current research interests focus on the
    search 9 (March): 366-380.
Reichheld, Frederick F. 1996. The Loyalty Effect: the Hidden Force Be-        impact of marketing strategy and market-based assets on corpo-
    hind Growth, Profits, and Lasting Value. Boston: Harvard Business         rate financial performance, particularly in the context of technol-
    School Press.                                                             ogy-intensive products and services.
You can also read