Printer Friendly

Determinants of consumers' decisions to seek third party redress: an empirical study of dissatisfied patients.

Determinants of Consumers' Decisions to Seek Third Party Redress: An Empirical Study of Dissatisfied Patients

One of the major thrusts of the consumerism movement stems from what Peterson (1974) has articulated as the consumer's right to recourse and redress, that is, a fair settlement of just claims. However, researchers have consistently found evidence to support the "silent majority" hypothesis: a majority of dissatisfied consumers are either unable to or do not seek redress from sellers directly. Aaker and Day (1982, p. 7) observe that "it has become clear, nonetheless, that businesses have difficulty listening to their customers." In part because of this "difficulty," several third party agencies have mushroomed with the goal of mediation between consumers and businesses. In this paper attempts to understand the reasons behind, or the antecedents of, a consumer's decision to use one or more third parties to seek redress are discussed.

The study of the antecedents of third party actions is important for at least four reasons. First, the number of litigation cases and complaints to consumer associations has continued to increase through the 1980s, especially for service-related problems (Bernhardt, Robinson, and Emmons 1983; Brown and Swartz 1984). For instance, malpractice suits involving medical services increased 34 percent from 1976 through 1981 (Gibson 1981; Easterbook 1987), and the size of the awards increased nearly 400 percent over the same period. Second, researchers and practitioners appear to agree that seeking redress from third parties (rather than from providers) almost always increases costs to society in general and the focal industry in particular. In a recent study Easterbook (1987) reported that because of malpractice suits alone doctors will spend $15 billion per year in "defensive medicine." Third, although several researchers have noted that third party actions are, in general, expressed by a "very small" percentage of consumers (Day and Landon 1976) who do not obtain satisfactory redress from the seller (Barnes and Kelloway 1980), recent studies by Brown and Swartz (1984) and Andreasen (1984) suggest that this result is less valid when dissatisfaction involves professional services (e.g., medical care). In particular, Brown and Swartz observe that dissatisfied patients have a feeling of "helplessness" and "often the only perceived course of redress is litigation" (Brown and Swartz 1984, p. 92; Andreasen 1984). This underscores the importance of studying third party actions as a consumer complaint behavior (CCB) phenomenon in its own right, at least for professional services. Fourth, most previous studies have not treated third party CCB as an independent set of complaint behaviors. Instead, these actions have been treated as a part of "public actions" that also include complaint behavior directed at sellers (Day and Landon 1976; Bearden 1983; Barnes and Kelloway 1980). More recent studies, however, provide compelling evidence for considering third party CCB as a distinct phenomenon (Feick 1987; Singh 1988). Thus, third party actions can neither be ignored as legitimate complaint behaviors nor aggregated to a higher level of analysis (e.g., public actions) as much of past research has tended to do.

Despite its importance, the understanding of the factors that underlie consumers' decisions to seek third party redress is an underresearched area in CCB literature. In a recent review of this literature, Ursic observed that "no studies were found that examine the consumer decision to go to court. Thus there is virtually a complete lack of information concerning an important consumer behavior phenomenon" (Ursic 1985a, p. 21).

This paper presents an initial step in the systematic inquiry into consumers' decisions to seek redress from third parties. Drawing upon a recent framework by Ursic (1985a, 1985b) and relevant research from the CCB area, a theoretical model of redress decisions is proposed. Its determinant constructs and the relationships among them are readily amenable to operationalization and empirical verification. A portion of the model is tested using data from patients dissatisfied with health care. In so doing the measurement error in constructs is explicitly considered by the use of the latent variable structural equations (LVSE) modeling approach. This offers significant advantages over traditional methods (e.g., regression, path analysis) in that the relationships recovered from data are among theoretic latent constructs rather than among some linear combination of observables (Bagozzi 1980, 1984; Bentler 1980). Further, in testing the proposed model, polychoric correlations(1) among the observed variables rather than the traditional methods for computing correlations (e.g., Pearson moment) are utilized. This strategy allows for an unbiased test of the model and provides more concrete suggestions for future development of the model.

This paper is organized around four key areas. First, the theoretical basis for the proposed model is discussed. Next, the research method utilized to test a portion of the proposed model is presented. Then, results obtained from the LVSE analysis of polychoric correlations are discussed. Finally, the limitations and implications of the study to researchers, practitioners, and public policy officials are enumerated.


The proposed model for understanding a dissatisfied consumer's decision to seek redress from third parties is depicted in Figure 1. The model is discussed from the perspective of a dissatisfied patient. The health care industry is distinguished by some unique characteristics. It is a high involvement service. Consumers are often hesitant to switch or voice problems to physicians, and they often lack knowledge to judge the quality of service (Andreasen 1984, 1985).

A consumer's perception of some dissatisfaction (DISS) triggers the whole model in Figure 1. The theory of disconfirmation of expectations probably explains why some situations are perceived as dissatisfying (Oliver 1980; Bearden and Teel 1983). Several studies consistently show that the dissatisfaction level has only a marginal influence on the specific complaint response selected by a dissatisfied consumer (Day 1984; Oliver 1986; Richins 1979; Singh and Howell 1985). Thus, it is hypothesized that DISS is a necessary precondition for the decision process of Figure 1 but bears no direct effect on the specific choice of third party actions. Accordingly, DISS is not depicted in Figure 1 as a potential determinant.

Third party actions (ACT) are defined as complaint behaviors that are directed toward one or more agencies that are not directly involved in the exchange relationship. As such, ACT does not imply a decision to engage in a single, specific behavior (e.g., complain/not complain). Instead, several diverse and multiple options are available to a dissatisfied consumer, such as contacting the Better Business Bureau (BBB), the newspaper, and/or a lawyer. Thus, it is not surprising that most studies tend to utilize diverse approaches for conceptualizing ACT. For instance, Haefner and Leckenby (1975) focus on complaints addressed to consumer protection agencies; Brown and Swartz (1984) examine malpractice litigation; Bernhardt, Robinson, and Semans (1983) attempt to study actions involving "consumer action panels"; and Ursic (1985a, 1985b) considers "court actions" as the third party CCB of interest. Such diverse approaches to defining the dependent variable do not allow one to relate and build upon the different studies.

Two recent studies appear to suggest the possibility of a more general approach. First, Feick (1987) utilized the notion of "sets" of behaviors(2) (i.e., multiple actions that are interrelated) to study complaint behaviors. Feick posits that such behavior sets can be ordered hierarchically so that the subset of behaviors at any level would be similar to but distinct from behavior subsets at other levels. Using complaint responses of 2,849 dissatisfied consumers, Feick found empirical support for hierarchically ordered "easy" and "hard" complaint actions. Specifically, "easy" actions included complaining to the salesperson, store, or company and not buying the product or service or dealing with the company. By contrast, "hard" actions include complaining to a consumer agency and taking some legal action. This suggests that CCB actions directed at third parties may constitute a set of behaviors that are located at a higher hierarchical level than other types of CCB responses.

Singh (1988) empirically investigated the distinct dimensions that underlie consumers' intentions to engage in various complaint behaviors. The intentions construct (INTEN) is defined as a dissatisfied consumer's predisposition for one or more complaint behaviors. Using four different data sets, Singh's analysis uncovered three distinct dimensions: (1) voice CCB, i.e., actions directed at the seller/manufacturer; (2) private CCB, i.e., informal complaint actions involving friends and relatives; and (3) third party CCB, i.e., formal complaints directed toward agencies not directly involved in the exchange relationship. Thus, consumers in different situations consistently appear to view third party CCB as a relatively similar set of behaviors that are distinguishable from other forms of CCB.

Taken together, the preceding studies provide a starting point for a theoretically based conceptualization of the third party CCB phenomenon. In particular, third party CCB intentions (INTEN) can be construed to be a set of predispositions to engage in one or more third party behaviors. This is consistent with the results of Feick (1987) and Singh (1988). However, ACT is posited to be a function of INTEN as well as several situational factors. For instance, a consumer who is predisposed toward third party CCB per se may choose to complain to a Better Business Bureau (BBB) because the BBB is located near his or her office. Another consumer similarly predisposed may choose legal action instead because of his or her intimacy with a lawyer. Such situational factors may vary from individual to individual and from episode to episode. For this reason, Figure 1 depicts ACT as composed of several specific options that are determined by one's INTEN and unspecified situational factors. By thus defining ACT and INTEN and the relationship between them, the focus of the proposed model is to explain and to predict consumers' intentions to engage in one or more third party actions and not any one specific ACT. This approach is not only consistent with the findings of Feick and Singh, but also advances the goal of developing model(s) that encompass the wide range of third party CCB.

The proposed model posits three constructs as direct antecedents of third party intentions: ATTI, PROB, and VALUE. Attitude toward complaining (ATTI) is conceptualized as an individual's stable evaluations regarding the "goodness" or "badness" of complaining per se, irrespective of the specifics of the dissatisfaction episode (Richins 1982). Richins notes that some individuals may consider complaining to be an "appropriate" behavior, while others may deem this behavior undesirable. Furthermore, such feelings may be multidimensional; that is, complaining may be evaluated as appropriate/inappropriate due to normative considerations (e.g., "I should complain") or for anticipation of societal benefits (e.g., "Others would benefit"). Although Ursic (1985a) does not propose ATTI as a potential predictor, several studies in CCB literature support the role of ATTI as a direct positive antecedent of INTEN (Richins 1982; Bearden and Mason 1983; Bearden and Crockett 1981). For instance, in Richins' study attitudes explained 25 percent of the variance in complaint intentions and about 15 percent of the variance in complaint behaviors.

Few studies, however, have investigated the role of ATTI for third party CCB. In an exceptional study, Richins found that third party complainers were significantly different from the general population on two dimensions of the attitude construct. In particular, third party complainers perceived businesses to be less responsive to complaints and believed more strongly that people should complain when dissatisfied (i.e., normative dimension). The "societal" consideration was not evaluated differently, however.

PROB, or the consumer's subjective probability of the success of third party CCB (e.g., in obtaining desirable outcomes), is expected to be positively related to INTEN. In accord with Ursic (1985a), PROB measures an individual's perception about the chances that satisfying outcome(s) (e.g., refund) would result if one or more third party CCB are exercised. Several studies in the CCB literature support this role of the PROB construct (Day 1984; Forbes et al. 1985). However, only a few studies have examined this effect empirically. Instead, considerable indirect evidence exists that dissatisfied consumers who do not complain perceive that the chances of satisfying outcomes are meager even if they did complain, that is, lower PROB (Day 1986; Day and Ash 1979). In the context of third party CCB, Ursic's (1985a) study provides the only direct empirical evidence of the path between PROB and a consumer's decision to take some court action.

VALUE is hypothesized to affect INTEN directly and positively. It is conceptualized as an individual's evaluation of the tradeoffs between benefits stemming from one or more third party actions and the costs associated with taking those actions. In other words, it represents an individual's assessment of the question: "Is a specific third party action worth the effort?" Benefits include rewards such as compensation for bad service or poor performance, while costs incurred may include special trips to consumer agencies and legal fees.

Although Ursic (1985a) posits benefits and costs as separate constructs, there are theoretical and pragmatic reasons for conceptualizing them jointly as a single construct. Theoretically, Hirschman (1970) in his theory of voice and exit has suggested that "worthwhileness of complaint," that is a cost/benefit evaluation, should be considered as a major predictor of complaint behavior. Hirschman's theory has been described as one of "great fertility and synthesizing power" (Barry 1974, p. 106) and has been successfully applied in several dissatisfaction situations (Fornell and Didow 1980; Andreasen 1985; Farrell 1983; Spencer 1986; Graham 1986). Thus, it is noteworthy that costs and benefits are not considered individually in Hirschman's theory. Within the CCB literature, Landon (1977), in proposing a model of consumer complaint behavior, also posits that a joint consideration of payoff (i.e., benefits) minus the costs is an important predictor of CCB.

Pragmatically, benefits are always associated with some costs. More importantly, this association is generally inverse. That is, the possibility of greater benefits (e.g., filing a suit for damages and compensation instead of accepting a compromise through a consumer agency) would almost always involve greater costs (e.g., time, effort, and lawyer fees). Thus, consumers end up making tradeoffs between costs and benefits, thereby choosing actions that they perceive represent the "best" tradeoff. Empirical support for the hypothesized role of VALUE comes from the studies of Day (1986) and his associates. These studies consistently show that a significant proportion (28 percent to 50 percent) of dissatisfied consumers did not complain because they perceived "that it was not worth the time and effort." Direct evidence of this relationship is also provided by Richins (1979).

Three indirect antecedents of INTEN are posited: GLOB, PRIOR, and ATTRIB. They are indirect antecedents in the sense that these constructs are hypothesized to affect INTEN only through their effects on ATTI, PROB, and VALUE. In accord with Westbrook (1980), GLOB represents an individual's global feelings about the focal marketplace and the behavior of its players (e.g., firms). Because the proposed model focuses on dissatisfied patients, GLOB implies feelings about the health care system. These feelings encompass an individual's response to general questions such as: "Do hospitals and physicians care for the needs of their patients, and are the hospitals and doctors really concerned about patients' satisfaction?"(3) Most CCB studies have utilized consumer discontent as a measure for GLOB, so that higher values of GLOB actually imply negative feelings toward the market system. Using such a measure, Westbrook argues that these global feelings are sources of several specific feelings and evaluations about the focal market system. In particular, consumers' feelings about the health care system are expected to influence ATTI, PROB, and VALUE. Precise directional hypotheses cannot be offered, however, due in part to a lack of previous research.

Previous experience of complaining to third parties (PRIOR) would be positively related to ATTI, PROB, and VALUE. Consistent with Ursic's (1985a) model for court actions, and with Day's (1984) framework for CCB in general, PRIOR is conceptualized to include the nature (good/bad) as well as the extent (frequency) of previous third party CCB experiences. It is apparent that the more positive and the more frequent the PRIOR, the higher the PROB for future dissatisfactions and the greater the VALUE due to heightened expectations of benefits in future third party complaint episodes. In addition, higher PRIOR is expected to reinforce positive attitudes toward complaining. Ursic's (1985a) results provide evidence in support of the PRIOR - PROB link.

The attribute of blame (ATTRIB) is also hypothesized to influence PROB, VALUE, and ATTI directly. The ATTRIB construct is conceptualized as the patient's subjective evaluation of whether the self or the provider (patient/hospital) is to blame for the dissatisfaction. When the provider is blamed, the ATTRIB measure is usually coded to be more positive. Within the CCB literature, studies by Folkes (1984) and Krishnan and Valle (1979) are generally supportive of the hypothesized relationships. In the context of third party actions, Ursic (1985a) has advanced the notion of "anger at seller" as a potential predictor of PROB and VALUE. When problems or dissatisfactions are strongly attributed to a health care provider, they most likely would be reflected in "anger" toward the provider. Thus, the two constructs appear exchangeable. However, because the ATTRIB construct stems from attribution theory (Weiner 1980), it provides a stronger theoretical framework for its justification, conceptualization, and intended role.

Three major exogenous characteristics are specified in the proposed model: demographic, personality, and environmental. Because none of these variables is defined as specific to a particular dissatisfaction episode, each is positioned away from the dependent variable. Consistent with the CCB literature, several demographic characteristics can be identified as potential predictors: education, income, age, sex, race, and occupation (Robinson 1979; Singh and Howell 1985). For instance, it is expected that patients who are better educated, earn higher incomes, are younger, and occupy professional jobs would tend to have more positive feelings about the health care system (GLOB) and possess greater experience of complaining (PRIOR).

Similarly, several personality variables appear useful for further investigation in the proposed model of third party CCB: specifically, aggressiveness, assertiveness, self-confidence, locus-of-control, dogmatism, and self-monitoring (Robinson 1979; Singh and Howell 1985). The specific hypothesis would depend upon the particular personality variable utilized. For instance, more aggressive patients are more likely to blame the provider (ATTRIB), to possess greater experience in complaining (PRIOR), and to be disposed negatively toward the health care system (GLOB).

Environmental influences include effects due to the general economic climate, the type of health care conditions in the patient's geographic area, the availability of third party agencies in the area, and the regulation climate of the state, among other factors. Although empirical studies have not explored these effects directly for the case of third party CCB, studies by Fornell and Didow (1980) and Fornell and Robinson (1983) in the context of CCB suggest that environmental effects are tenable.

The preceding model is essentially an initial attempt toward an understanding of the third party phenomenon. Further development can result from empirical investigation of the hypothesized linkages. The following study was conducted with the motivation to investigate empirically a portion of the proposed model.


Data were collected by utilizing a cross-sectional design and a self-administered mail survey. Although most previous studies are based on recall of a past dissatisfaction experience (Robinson 1979; Singh and Howell 1985), Landon (1980) raises some cautions in the use of this procedure. He notes that "after a lengthy questioning on a survey about dissatisfaction experiences, respondents are likely, when asked, `what did you do about your dissatisfaction?' to respond that `I complained,' when in fact the consumer may have done nothing about the problem" (Landon 1980, p. 337). Recognizing this caution, the survey procedure was modified somewhat. In particular, the questionnaire was designed so that two questions were posed at the outset. First, an open ended question asked respondents to explain briefly a dissatisfying experience with a doctor and/or hospital or its staff (e.g., nurses) that they could remember most clearly. Second, respondents were asked to report the complaint actions they had taken following the dissatisfaction. Detailed questions regarding the dissatisfying experience were not elicited before the complaint actions section.

Instead, in the later sections of the instrument respondents were asked to imagine that a dissatisfying experience similar to the one they had described occurred again. Various measures about what respondents think/feel/intend to do were obtained for this future incident. Use of a future incident to measure the study constructs is more appealing than mere recall of past feelings and intentions. In the analysis that follows, first a consistency check is conducted between the intentions of third party redress (i.e., based on future incident) and actual behaviors elicited at the outset. Following this, the proposed model is confronted with empirical data using third party intentions as the dependent construct.

Because third party CCB is an underresearched area, standard measures for various constructs were not available. Most measures had to be adapted from relevant consumer behavior studies. In addition, PROB and VALUE required identification of consequences that are perceived to result from third party actions. Further, these consequences had to be specific to medical care dissatisfaction. In order to accomplish this, a focus group of faculty and staff was employed. Consequences were retained using criteria of saliency and parsimony. Based on pretest results and initial analysis, the consequence items were refined. In addition, the complete questionnaire was pretested. Items were revised based on pretest input regarding confusing and/or ambiguous wording and instructions.

Final Questionnaire

A self-administered questionnaire containing the following measures and several other questions was completed by the respondents. Table 1 provides the specific operationalization employed, and the Appendix lists the specific items utilized.

ACT was measured by a two-item dichotomous (Yes/No) scale. Respondents were asked if they had taken one or both of these actions following the dissatisfaction: (1) complained to a third party, such as the Medical Association or the Better Business Bureau, and (2) took some legal action against the hospital and/or physician.

DISS was assessed by a ten-point "Not-at-all-dissatisfied-Completely dissatified" scale. Respondents were asked: "Overall how dissatisfied were you before you did anything about the problem?"

INTEN was measured by a four-item, six-category Likert scale adapted from Day et al. (1981). The respondents were asked to report the likelihood of taking each of the four actions on a six-point "Very Likely-Very Unlikely" scale. The composite reliability of this measure is 0.85.

PROB was measured by an adaptation of Day's (1984) study. It is a three-item, six-category scale. The PROB construct is conceived as the effectiveness of third party actions; that is, the probability that the third party actions, if pursued, would result in some desired consequences (i.e., redress). This likelihood is measured by a six-point "Very Likely-Very Unlikely" scale. The composite reliability of this three-item measure is 0.70.

VALUE is often measured by a good/bad scale. Initial pretests indicated that such a measure neither provided any significant variation in responses nor evoked cost/benefit tradeoffs. A different approach suggested by Bagozzi (1982) was adopted for developing a measure for VALUE. Specifically, this measure asks the respondents to evaluate if they are likely to take some action given that some particular benefit (e.g., redress) was sure to occur. Some patients may choose not to engage in third party actions even if benefits were sure to occur merely due to the prohibitive costs (e.g., time/effort) involved. Such notions were captured by a three-item, six-category scale. The composite reliability of this measure of VALUE is 0.60. However, an analysis of the three items revealed that one of the items had poor item-total correlation. Deleting this particular item (VALUE1 in the Appendix) increased the reliability from 0.60 to 0.87. This suggests that the item lacks consistency, and its deletion appears justified.

ATTI is based on Richins' (1982) study. It is a four-item, six-category scale. Specifically, the normative dimension of the attitude construct was tapped. The normative dimension was selected because earlier research suggests that its role in understanding third party intentions may be significant (Richins 1982). The ATTI items achieved a composite reliability of 0.50. This value is low and suggests that more scale development work is needed. However, because of the stability of these measures during estimation, it was decided to retain all four items for this initial study (see more discussion of this under measurement model analysis).

GLOB is a six-item measure. Note that this is not specific to the dissatisfaction episode. Rather, it pertains to overall feelings toward the medical care constituents. The items were scored so that higher values represented negative feelings. Responses were obtained on a six-category "Strongly Agree-Strongly Disagree" Likert scale. The composite reliability of the six-item scale is 0.83.

PRIOR was obtained as follows. Respondents were first asked if they had ever engaged in third party actions (e.g., legal). For those responding "Yes," an open-ended question measured the number of times they had done so. The PRIOR construct was coded simply by the number of times a respondent had taken third party actions previously.

DEMOGRAPHIC includes four characteristics: age, sex, education, and income. These particular characteristics were selected because previous research indicated that these variables may have significant effects. Specific hypotheses and operationalizations are provided in Table 1. [Tabular Data Omitted]


The population of interest is patients who had a recent dissatisfying experience. However, sampling frames for such a population are not readily available (Robinson 1979). A substitute procedure suggested by Robinson, wherein a random sample of households is asked to preselect themselves if they can recall a recent dissatisfying experience, was adopted. A total of 1,000 households in the Southwest was selected to receive the questionnaire packet. Followup was accomplished with reminder cards and telephone callbacks.

A total of 166 households returned the questionnaire. The true response rate cannot be estimated precisely because this would have involved estimating the proportion of the number of respondents to the number of households who had experienced a recent dissatisfying medical care experience. As noted above, the denominator of this proportion is an elusive number. However, the telephone callbacks provided an estimate. Five hundred telephone callbacks were made. Telephone numbers were obtained from the crisscross directory. A contact rate of about 80 percent was achieved (up to three calls were made to those not initially reached). Of those contacted, at least 70 percent stated that they had not responded because they could not recall a dissatisfying experience. In contrast, those who had experienced a "recent" problem were eager to participate and "let someone know" about their dissatisfaction. Note that the dissatisfaction experience rate of 30 percent based on the telephone callbacks is consistent with previous research in CCB (Best and Andreasen 1977; Day and Bodur 1977). Based on this estimate for the percentage of dissatisfied households, the estimate for the response rate is 55 percent.

Because the survey instructions asked that the questionnaire be completed by the person who recently underwent a dissatisfying experience, it is difficult to compare sample demographic characteristics with some given population (e.g., census information). Of the respondents, 30 percent are male, 70 percent are married, and about 70 percent had at least college education. The median income is around $30,000. Over 68 percent of the respondents are 35 years in age or more. Some of the returned surveys had to be deleted because of incomplete responses. One hundred twenty-five usable responses were returned.

Method of Analysis

Initially, the correspondence between intentions (i.e., INTEN) and actual third party behavior was examined using analysis of variance procedures. The actual behaviors were coded as either a 1 (Yes) or 0 (No) and entered as a treatment variable. The strength of the intention-behavior correspondence was inferred from the F-statistic (i.e., ratio of within to error mean squares). Also, the mean values for the INTEN were examined across the two levels of behavior.

Next, the hypothesis that the dissatisfaction level (DISS) has no direct effect on third party intentions (INTEN) was examined. Following this, the empirical model displayed in Figure 2 was analyzed using the LISREL VI software for the analysis of latent variable structural equations (Joreskog and Sorbom 1985). Because of this methodology, Figures 2 and 3 are depicted in path analysis conventions. However, for the sake of clarity, errors in structural equations and correlations among the exogenous constructs are depicted only in relevant tables. The use of a latent variable model is desirable in that the estimated structural coefficients reflect relationships among estimates of the theoretic observables. Also, goodness-of-fit measures are provided for the model.

A matrix of polychoric correlations was used as input to LISREL. Note that polychoric correlations are preferable to other methods for computing intercorrelations because they yield the "best results" in terms of unbiased estimates and smaller squared errors of these estimates (Babakus, Ferguson, and Joreskog 1987). Also, for each latent variable, the first lambda was set equal to unity to fix the scale of measurement. This convention merely scales the latent variables in the same units of measurement as one of their respective operationalizations.

Although maximum likelihood (ML) estimates are frequently computed when using LISREL, Joreskog and Sorbom state that the use of such estimates is less defensible when polychoric correlations are utilized. First, the matrix of polychoric correlations may not be positive definite rendering ML estimation inapplicable. Second, these correlations do not behave as ordinary sample moments so that the standard errors and chi square values resulting from an ML estimation may be grossly in error. Instead, Joreskog and Sorbom suggest that the unweighted least squares (ULS) estimation is more appropriate. Thus, ULS estimation procedure was utilized.

The following steps were used to evaluate the ULS solution. First, the goodness-of-fit statistics were examined. In particular, the goodness-of-fit index (GFI) and the root mean square residual (RMR) were used to assess the relative number of correlations explained by the model. GFI values close to 1.00 and RMR close to 0.0 are indicative of good model fits. Second, the coefficient of determination, [R.sup.2], for the structural equations and the dependent construct, INTEN, was examined. This measure represents the amount of variance explained by the model. Obviously, values close to 1 are desirable. Third, fitted residuals were examined relative to the estimated polychoric correlations. Large absolute residuals (e.g., [is greater than] 0.10) imply that the model does not satisfactorily reproduce observed correlations. Finally, the proposed model was compared with respect to two trivial models: namely, the null and the full models. The null model includes none of the structural paths, but consists of all measurement relationships (i.e., between the latent construct and its indicants). The full model, in contrast, includes all possible structural paths as well as the hypothesized measurement relationships. These models are trivial in the sense that they are not theoretically interesting; the null model posits no effects, and the full model is fully saturated. However, these models provide useful standards for comparison. Recall that the proposed model is an unsaturated model in that it hypothesizes the presence of some (not all) specific structural relationships. The "goodness" of the hypothesized model, thus, can be evaluated by noting: (1) the improvement over the null model and (2) its closeness to the full model. That is, if an unsaturated model (e.g., the proposed model) yields goodness-of-fit characteristics similar to a fully saturated model (i.e., the full model), then our confidence in the proposed model is enhanced.

Note, the ULS procedure does not provide standard errors for each estimate. Thus, the conventional t-test for the significance of path coefficients cannot be conducted. However, these coefficients can be evaluated for substantive significance. In other words, based on its magnitude, an individual path coefficient can be evaluated for its substantive importance in predicting a dependent variable. Typically, for a standardized coefficient, effect sizes greater than 0.20 are considered to be large enough for substantive purposes. Also, Table 2 shows that correlations exceeding 0.20 were significant at p = 0.01 (see also O'Grady 1982). [Tabular Data Omitted]


Summary Statistics

Mean and standard deviations for the various constructs of the study are provided in Table 2. Also provided are Pearson moment correlations computed among the summated indices of the respective constructs. Several preliminary inferences can be drawn based on Table 2. Standard deviations for most measures exceed one. This suggests that the measures appear to be able to capture individual differences. Intercorrelations among the summated indices provide further insight into the data. Note, of the 36 correlations among the predictors (i.e., excluding INTEN), about 75 percent are no larger than 0.15, the value significant at 0.05. This suggests that multicollinearity problems are not severe. However, the relationships between INTEN and key predictors are significant and in the expected direction.

Consistency Check

The analysis of variance procedure with INTEN as the dependent construct and the behavioral measure ACT as the two-level treatment produced an F-statistic of 11.03 (p [is less than] 0.00). This suggests that differences in behaviors can significantly explain differences in intentions to engage in third party actions. In addition, mean values for the INTEN measure are compared for the two levels of third party behaviors. INTEN means of 16.57 and 9.44 are obtained corresponding to the "Yes" and "No" cells respectively. The difference in means of 7.13 is significant for a one-tailed t-test (p [is less than] 0.01). This clearly supports the notion of correspondence between intentions and behaviors in the data.


Role of Dissatisfaction

Initially, the hypothesis that DISS has no direct effect on INTEN was examined. Polychoric correlations were input to LISREL, and the ULS method of estimation was utilized.

In overall fit, the DISS - INTEN model yielded the following statistics: GFI=0.998, adjusted GFI=0.993, and RMR=0.027. These statistics are indicative of a "good" fit between the model and data. However, the estimate of the DISS - INTEN effect is only 0.17, and the corresponding [R.sup.2] is only 0.04.

Based on the criterion for substantive significance when paths exceed 0.20, the effect of DISS does not approach significance. Further, the low [R.sup.2] indicates that DISS probably plays a marginal role, if any, in determining INTEN. These results appear supportive of the hypothesized role of DISS.

Model Fit Analysis

Next, three different models were estimated: (1) the proposed model of Figure 2, (2) the null model with no structural relationships, and (3) the full model with all possible structural relationships. Further, because the model in Figure 2 was not introduced as a definitive framework, directions were sought for improving the model. Specifically, the plausibility of additional structural paths was evaluated using multiple criteria, namely: (1) improvement in goodness-of-fit statistics (i.e., GFI and RMR), (2) increase in R-square, (3) theoretical support for the relationship, and (4) parsimony, that is, fewest number of additional paths.

The final model selected (Figure 3) had four more paths than the proposed model. These are (1) PRIOR and INTEN, (2) SEX and VALUE, (3) INCOME and PROB, and (4) AGE and ATTI. Theoretical justification and implications of these paths are discussed in the structural model analysis section below. This final model produced the following statistics(4): GFI=0.91, RMR=0.093, and R-square for INTEN=0.65. Also, the residuals for the four INTEN measures were small and less than 0.10 in absolute value. In particular, 75 percent of these residuals are less than 0.04, and about 38 percent are no greater than 0.01. Note that the final model has fourteen fewer paths than the full model, yet it yields comparable fit statistics. In particular, the GFI and RMR differ by less than 3 percent in the two models, and the R-square values for the INTEN construct differ by no more than 6 percent. Such small differences are suggestive of the adequacy of the final model. Further, because (1) the final model has only four more paths than the proposed model, (2) about 65 percent of the variance in the dependent construct is explained, and (3) residuals are small, it appears safe to suggest that the final model is a reasonable representation of data.(5)

The measurement and structural parameter estimates of the final model are provided in Table 3 and Table 4, respectively. A discussion of these structural paths and the measurement paths follows. [Tabular Data Omitted]

Measurement Model Analysis

Both the ULS and standardized measurement parameter estimates are depicted in Table 3. The standardized estimate is akin to a factor loading. An examination of Table 3 shows that all parameter estimates exceed 0.30. The only exception is the third indicant of ATTI, which has an estimate of 0.27. This suggests that the various measures of study are reasonably well specified. Composite reliabilities for the latent constructs provide further evidence of satisfactory measurements. In particular, reliabilities range between 0.70 and 0.85, which appears reasonable for an initial study of third party CCB. The only exception is ATTI with a reliability of 0.50. Clearly, more scale development work is required for this construct. However, in the context of the present model, parameter estimates for ATTI depicted remarkable stability. For instance, the change in lambda estimates was less than 15 percent between the models of Figure 2 and Figure 3. This suggests that although ATTI has less than desirable reliability, its indicators are stable. This seems to be a reasonable basis to retain ATTI for further analysis.

Structural Model Analysis

Table 4 reports the ULS parameter estimates for the structural paths in the final model. Also included in Table 4 are the corresponding standardized estimates that can be utilized for the comparison of effects.

As hypothesized, PROB, VALUE and ATTI positively and significantly influence a consumer's intentions to seek redress from third parties. A comparison of these effects reveals that a person's attitude toward complaining (ATTI) is the most influential factor. By contrast, the effect of VALUE is borderline. Thus, it appears that expectations of success of third party CCB (PROB) and consumer's attitude (ATTI) play a central role in understanding INTEN. This finding is consistent with Ursic's finding (1985a) as far as PROB is concerned. To the author's knowledge, a direct test of ATTI's influence on third party INTEN has not been reported earlier.

A consumer's prior experience of third party complaining (PRIOR) was found to influence positively intentions to engage in future third party actions. This direct effect was not hypothesized. Instead, PRIOR was expected to influence INTEN indirectly through its effect on PROB, VALUE, and ATTI (which are also substantively significant). Post hoc explanations for PRIOR's direct effect can be advanced based on script and reinforcement theory. That is, previous experiences provide scripts for future behaviors when the situation is somewhat similar. Additionally, earlier experiences of complaining to third parties may be positively reinforcing per se (irrespective of the outcome) because consumers may feel that they acted responsibly. However, as with all post hoc explanations, this rationale should be considered tentative and subject to further investigations.

GLOB and PRIOR were hypothesized to influence PROB, VALUE, and ATTI. Results in Table 4 afford mixed support for these hypotheses. Consider the effects on PROB. Although GLOB negatively affects PROB (as expected), this path is not large enough to be substantively significant. Prior experiences (PRIOR), however, have a significant positive effect (path=0.28). This suggests that as consumers have greater experience with third party systems, they probably learn to become more effective and perceive greater success in future third party actions (PROB). This finding is consistent with results obtained in previous studies (Ursic 1985a).

In addition, consumer's income (INC) is found to influence PROB (path=-0.25) negatively, implying that as income level increases, expectations of success in third party actions decreases. This counter-intuitive relationship was not hypothesized. Can this finding be explained on the grounds that consumers in the lower income bracket have greater experience with third party channels, thus resulting in higher PROB? Probably not. Note in Table 4 that income in fact has a positive and not negative relationship with PRIOR. Clearly, more research is needed to understand this anomaly.(6)

VALUE is influenced positively by GLOB (path=0.23), PRIOR (path=0.63), and the sex of the respondent (path=0.42). In other words, the more discontent the patients have with their health care systems, the greater their prior experiences with third party actions, and women in particular perceive greater VALUE of third party channels. Note, a directional hypothesis was proposed only for PRIOR (which is supported), and the effect of sex was not hypothesized. Thus, women may find third party channels very beneficial (Value) if they assist them in obtaining redress. A similar rationale can be offered for those patients who feel that the health care system cares "nothing at all" about them (high GLOB).

ATTI is positively influenced by GLOB (path=0.37), but negatively affected by PRIOR (path=-0.41) and age of the respondent (path=-0.23). The more discontented the consumer, the more positive the attitude toward complaining when dissatisfied. This is consistent with the general notion of "discontent" as developed by Lundstrom and Lamont (1976). However, the effect of PRIOR is counterintuitive. Patients with greater prior experience of complaining are found to have less positive attitudes toward complaining. A potential explanation for this unexpected finding could be that the lengthy and drawn-out legal battles leave patients less positively inclined toward complaining. There is some indirect support for this explanation. Sarat (1977) found that greater contact with the legal system (e.g., small claims courts) resulted in a diminished opinion of its operation.(7)

Finally, younger respondents tend to have more positive attitudes toward complaining. This appears consistent with several earlier studies that report that consumers who complain frequently (i.e., across product categories) tend to be younger (Singh and Howell 1985; Robinson 1979).

The effects of exogenous variables are mixed. None of the tested variables (i.e., age, sex, education, and income) is found to affect GLOB substantively. In contrast, all these variables significantly affect PRIOR. Of these relationships, note that sex of the respondent has the largest path coefficient (value=0.70) implying that men rather than women have significantly greater previous experience of complaining to third parties. Also, third party complainers are younger, less educated, and come from higher income brackets. This profile is similar to that for the "complainer" group reported in the literature, with the exception of the education relationship (i.e., "complainers" are generally more educated). A potential reason for the inverse relationship in our data could be due to the specific operationalization used. A four-category education scale (high school, trade school, college, graduate school) was used in the present study. Other studies have used a continuous scale based on the number of years. Categorization of education years into four specific groups could have affected the relationship. The absence of a strong relationship with global feelings toward health care is an unexpected finding. Potential candidates for further research include race, occupation, and, more importantly, personality variables (e.g., locus-of-control, self-monitoring).


This research has attempted to achieve three major objectives: (1) to contend that it is theoretically and empirically useful to examine the multiple options of third party CCB within a single framework, (2) to propose a theoretical framework that explains the processes that consumers probably undergo in making a decision to engage in one or more third party behaviors, and (3) to examine empirically the correspondence between the proposed model and data in the context of health care services.


The study is based on correlation data from a cross-sectional research design. Although such designs are customary, caution is necessary in inferring sequential effects from this data. A more realistic view could be obtained from longitudinal investigations.

Also, measurements may have been biased due to respondents' recall and memory abilities. In particular, because respondents are asked to recall a dissatisfying experience, it is possible that the more serious problems may have been reported. In addition, systematic effects due to memory lapses and reconstruction of past events are bound to occur.

This study was designed so that it would be less susceptible to such effects by including two sections. In the first the respondents were asked to recall a dissatisfaction episode and record the complaint actions they had taken. Note that no questions were asked about their beliefs/thoughts/feelings. In the second section, they were asked to imagine that an episode similar to the one they had described occurred again. All measures of study, except third party behaviors, were collected for this future incident. The results for the consistency check suggest that this "future" incident approach appears to be a reasonable strategy for obtaining valid data.

Another source of bias stems from sampling considerations. Geographic restriction of the sample and the possibility that the more dissatisfied patients may have responded are contributing factors. In addition, less-than-desirable reliability for the ATTI measure may have affected the quality of results. Note also that only a portion of the proposed model was investigated. A complete test would have necessitated an unduly long questionnaire that in turn was felt to degrade response quality significantly. Future research may address this deficiency by systematically investigating different subsets of the proposed model.


This paper offers a theoretical model of third party redress processes that can be utilized to understand the focal phenomenon in different dissatisfaction contexts. In addition, the model is not limited to a specific act involving one and only one third party channel. Instead, it posits that it is possible to predict intentions to engage in one or more actions from a set of third party actions. The empirical investigation of a portion of the proposed model reveals that 65 percent of the variance in the dependent construct is explained. Previous work in the area has achieved explanation levels of only around 17 percent (Ursic 1985a). Further, the key hypothesized relationships (e.g., involving PROB, ATTI, and VALUE) are supported by the data. Thus, sufficient evidence exists to suggest that the proposed model is a reasonable framework and warrants the serious attention of researchers. This is not to say that the framework is a definitive model. Rather, it should be the starting point for further theoretical development and empirical work in the area.

Initially, it is recommended that the model be investigated for service dissatisfactions because third party actions are more common for such episodes (Brown and Swartz 1984; Bernhardt, Robinson, and Semans 1983). Such investigations will also test the generalizability of the proposed model. In addition, although it is not expected that the specific path coefficients obtained in this study (i.e., for health care) are in and of themselves generalizable to other services, it is fruitful to examine how these coefficients vary across service categories.

This research also identifies several areas where more theoretical work is needed. For instance, consider the PRIOR - ATTI link. Most previous studies contend that this link should be positive. Yet in our data this path coefficient was negative. The role of income is also elusive. Income was found to be negatively associated with PROB, but positively related to PRIOR. In part, these anomalies may stem from pooling consumers with varying prior third party experiences. This noise in the data may be reduced by analyzing more homogenous subgroups, such as partitioning on the basis of successful or unsuccessful PRIOR outcomes. More empirical work in improving the operationalization of some constructs, particularly ATTI, would be helpful. Likewise, other approaches for measuring the cost/benefit tradeoffs (VALUE) warrant careful examination.

More importantly, it is critical to understand not only why dissatisfied patients complain to third parties but also why dissatisfied patients choose third party CCB from the set of entire CCB options (e.g., voice). Several possibilities exist. Consumers may elect third party actions (1) only when all other CCB options have failed, (2) when they perceive that the success of voice CCB is minimal, or (3) unrelated to the choice of other CCB actions. Sorting through these possibilities can potentially shed greater insight into the focal phenomenon.

The results also send some clear and compelling signals to public policy officials. Greater use of third party channels to clear marketplace dissatisfaction appears to hinge on consumers' normative beliefs about complaining (ATTI) and perceived probability of success (PROB). Note that the mean values for ATTI items are greater than the midpoint of the scale. By contrast, the means for PROB items are below this midpoint (Table 2). Thus, while consumers have positive ATTI, they do not perceive that consumer agencies would be successful in obtaining redress. This sheds light on the kind of impediments that stand in the way of consumer agencies' efforts to deliver redress to dissatisfied consumers. Consumer agencies, therefore, should consider specific programs for enhancing their effectiveness (i.e., PROB) and communicating this to consumers.

For health care services, the results of this study afford some specific suggestions. It has become painfully clear that third party channels are inefficient for redress of patient dissatisfactions. For instance, recently advertisements showing a surgeon bound at the wrists were placed in several Florida newspapers with the following caption: "Malpractice Insurance is Making it Impossible for Physicians to Practice" (Marketing News, September 11, 1987). Although in part the problem is structural, a critical factor is the ability of health care providers to redress patient problems directly. Specific programs that facilitate complaining and a responsive system to redress just complaints can serve to alleviate the fears of providers (e.g., malpractice suits). Unquestionably, the identification of patient problems and redress thereof is critical. However, our results help in profiling patients who are most likely to go to third parties for redress. Both demographic (sex, age, and income) and psychographic (PRIOR, PROB, and ATTI) characteristics are implicated. Because third party complaints are detrimental from a provider's standpoint, providers may use this profile to develop communication or other programs that facilitate patient complaint handling and satisfaction. This profile can be especially beneficial in improving the effectiveness of such programs through appropriate media channels and message content. Following this initial step, a broader program for wider dissemination of complaint handling and responsiveness may be adopted.


This study provides a comprehensive and testable model of the processes that underlie third party actions that is consistent with previous research in the area. Overall, the tested model explains about 65 percent of the variance in the dependent construct. This value is significantly higher than explanation levels achieved in previous empirical research. It suggests that the proposed model is a reasonable perspective on third party actions and can be a starting point for further refinement and development. Also, it supports the notion that considering all third party actions within the proposed model affords future researchers with a firmer ground for systematic and cumulative research.

However, the proposed model is an initial step toward understanding the "why" of third party CCB. Future replications could provide guidelines for improving the model. In addition, several avenues for future research are provided. Further studies along these lines have the potential to enhance significantly understanding about the third party redress phenomenon.


Items Used to Measure the Various Constructs



If a similar incident occurred again, how likely is it that you would:

INTEN1 write a letter to the local newspaper about your bad experience? (Y[sub.1]) INTEN2 report to a third party so that they can warn other consumers? (Y[sub.2]) INTEN3 complain to a consumer agency and ask them to make the hospital and/or physician take care of your problem? (Y[sub.3]) INTEN4 take some legal action against the hospital and/or physician? (Y[sub.4])


Assume you reported the incident to a third party, such as the Medical Association or the Better Business Bureau, how likely is it that they would:

PROB1 take no action? (reverse scored) ([Y.sub.5]) PROB2 make the physician/hospital take care of the problem? ([Y.sub.6]) PROB3 solve your problem and ensure that the physician and/or hospital is careful in the future? ([Y.sub.7])


How likely is it that you would report the incident to a third party such as the Medical Association or the Better Business Bureau if you were pretty sure that they would: VALUE1 take no action? (reverse scored)* VALUE2 make the hospital and/or physician take care of your problem? ([Y.sub.8]) VALUE3 solve the problem and ensure that the hospital and/or physician is careful in the future? ([Y.sub.9])


ATTI1 It bothers me quite a bit if I don't complain about poor medical treatment. ([Y.sub.10]) ATTI2 I often complain when I am dissatisfied with medical care because I feel it is my duty to do so. ([Y.sub.11]) ATTI3 People are bound to end up with unsatisfactory medical treatment once in a while, so they should not complain (reverse scored). ([Y.sub.12]) ATTI4 I don't like people who complain to physicians, because usually their complaints are unreasonable (reverse scored). ([Y.sub.13])

Global Affect

GLOB1 Getting satisfactory medical care is a real problem. ([Y.sub.14]) GLOB2 Patients do not have any influence on the medicines that are prescribed to them. ([Y.sub.15]) GLOB3 Most hospitals care nothing at all about the patients. ([Y.sub.16]) GLOB4 The patient is usually the least important consideration to most hospitals. ([Y.sub.17]) GLOB5 As soon as they discharge a patient, most hospitals forget about the patient. ([Y.sub.18]) GLOB6 In general, hospitals are plain dishonest in their dealing with the patient. ([Y.sub.19]) (a)The item lable in parenthesis corresponds to the specific labels used in Figures 2 and 3. (*)This item was deleted from analysis because it correlated poorly with theremaining two measures.

(1)When data are collected on a categorical scale (e.g., Likert), Babakus, Ferguson and Joreskog (1987, p. 226) report that the polychoric correlation provides the best results "producing virtually unbiased estimates and the smallest squared errors on the average." Note, this result is applicable to the usual studies in social sciences that obtain perceptual data on a Likert (or some variation thereof) scale. The polychoric correlation is defined as follows. If [v.sub.1] and [v.sub.2] are two ordinal level variables (e.g., items on a Likert scale) with [v.sub.1*] and [v.sub.2*] as the corresponding underlying continuous, bivariate normal variables, then the correlation coefficient between [v.sub.1*] and [v.sub.2*] is the polychoric correlation. This can be computed using LISREL or other statistical software (e.g., IMSL).

(2)A reviewer has pointed out that the notion of behavior "sets" is not "new" to the social sciences. In particular, it is conceptually similar to Alderson's (1965) concept of "assortment." Likewise, Sheth (1970) and others have noted that some consumer-related phenomena, such as brand loyalty, may in fact be ordered in "sets" or groups of behaviors.

(3)It can be argued that such feelings are, perhaps, manifestations of the relative power/influence of consumers over sellers. In a very competitive industry structure, sellers are intimately dependent on customer loyalty through market share, thus enhancing consumers' relative influence (Hirshman 1970). In such markets, consumers may perceive sellers to be responsive to their problems and concernced about their satisfaction. In contrast, monopolistic or less competitive markets may be characterized by less favorable conditions for the consumer resulting in negative GLOB (Andreasen 1984).

(4)In overall fit, the proposed model of Figure 2 produced following statistics: GFI=0.89, RMR-0.13, R-squared for all structured equations=0.44, and R-squared for INTEN=0.59. By contrast, the null model yielded: GFI=0.727, RMR=0.162, and zero values for R-squared since the structural paths were not included. The corresponding statistics for the full model were GFI=0.916, RMR=0.09, R-squared for all structural equations=0.73, and R-squared for INTEN=0.71. Note, however, that the full model included 18 more structural paths than the proposed model.

(5)In making this claim, the multivariate distribution of the variable is a legitimate concern. Only if this distribution does not depart greatly from a normal distribution can one place confidence in the statement that the "final model is a reasonable representation of data." To evaluate this possibility, the Q-Q plot of normalized residuals obtained form LISREL was examined. This plot shows that the residents follow a linear curve with slope slightly greater than 1.0. This is indicative of trivial departures from normality (Joreskog and Sorbom 1985,p. 17, Section III).

(6)One reviwer has pointed out that a probable reason for this negative relationship may be due to the nature of the health care service. In particular, two unique features of this service may play a role, namely: (1) it is often difficult for patients to switch services, and (2) voicing complaints to physicians is both socially inhibiting and less fruitful (because of the attitude of the physicians). Thus, consumers with higher income may be more aware of these features and therefore perceive that the probability of success (PROB) is low.

(7)Likewise, the preceding reviewer posits the nature of the health care service as another plausible explanation of the negative relationship between PRIOR and ATTI. Prior experience of complaining about health care providers may be so frustrating that it results in negative attitudes toward complaining itself. Future research across different service contexts can examine this possibility.


Aaker, David and George S. Day (1982), "A Guide to Consumerism" in Consumerism: Search

for the Consumer Interest, Aaker and Day (eds.), Free Press: 2-20. Alderson, Wroe (1965), Dynamic Marketing Behavior, Homewood, IL: Irwin. Allison, Neil K. (1978), "A Psychometric Development of a Test for Consumer Alienation,"

Journal of Marketing Research 15 (November): 565-575. Andreasen, Alan R. (1985), "Consumer Responses to Dissatisfaction in Loose Monopolies,"

Journal of Consumer Research, 12 (September): 135-141. Andreasen, Alan R. (1984), "Consumer Satisfaction in Loose Monopolies: The Case of

Medical Care," Journal of Public Policy and Marketing, 2: 122-135. Babakus, E., Carl Ferguson and Karl Joreskog (1987), "The Sensitivity of Confirmatory

Maximum Likelihood Factor Analysis to Violations of Measurement Scale and Distributional

Assumptions," Journal of Marketing Research, 24 (May): 222-228. Bagozzi, Richard P. (1984), "A Paradigm for Theory Construction in Marketing," Journal of

Marketing, (Winter): 11-29. Bagozzi, Richard P. (1982), "A Field Investigation of Causal Relations Among Affect, Intentions

and Behavior," Journal of Marketing Research, 17 (November): 562-584.

Bagozzi, Richard (1980), Causal Modelling in Marketing, New York, NY: John Wiley and

Sons. Barnes, James and Karen R. Kelloway (1980), "Consumerists: Complaining Behavior and

Attitudes Toward Social and Consumer Issues," Advances in Consumer Research, Jerry

Olson (ed.) 329-334. Barry, Brian (1974), "Review Article: Exit, Voice and Loyalty," British Journal of Political

Science, 4: 79-107. Bearden, William (1983), "Profiling Consumers Who Register Complaints Against Auto

Repair Services," Journal of Consumer Affairs, 17(2) (Winter): 315-335. Bearden, William O. and Jesse E. Teel (1983), "Selected Determinants of Consumer Satisfaction

and Complaint Reports," Journal of Marketing Research, 20 (February): 21-28. Bearden, William O. and Melissa Crockett (1981), "Self Monitoring, Norms and Attitudes as

Influences on Consumer Complaining," Journal of Business Research, 9: 255-266. Bearden, William O., and Barry Mason (1983), "Empirical Evidence on Marketplace Alienation,"

Journal of Macro-Marketing, (Fall): 6-20. Bentler, Peter (1980), "Multivariate Analysis with Latent Variables: Causal Modelling,"

Annual Review of Psychology 31: 419-456. Bernhardt, Kenneth, Larry Robinson and Debra Semans (1983), "Consumer Complaint

Handling: Self Regulations vs. Government Regulations," AMA Educators Proceedings:

369-373. Best, Arthur and Alan R. Andreasen (1977), "Consumer Response to Unsatisfactory Purchases:

A Survey of Perceiving Defects, Voicing Complaints and Obtaining Redress," Law

and Society Review, 11 (Spring): 701-742. Brown, Stephen and Teresa Swartz (1984), "Consumer Medical Complaint Behavior: Determinants

of and Alternatives to Malpractice Litigation," Journal of Public Policy and

Marketing, 3: 85-98. Business Week (1984), "Making Service a Potent Marketing Tool," June 11, 1984: 164-170. Day, Ralph (1986), "A Post-Dissatisfaction View of Complaining Behavior," Paper presented

at the Association of Consumer Research Meetings. Day, Ralph L. (1984), "Modeling Choices Among Alternative Responses to Dissatisfaction,"

Advances in Consumer Research, Kinner (ed.), 11: 469-499. Day, Ralph, K. Gabricke, T. Schaetzle, and F. Staubach (1981), "The Hidden Agenda of

Consumer Complaining," Journal of Retailing 57 (Fall): 86-106. Day, Ralph and Muzzaffar Bodur (1977), "A Comprehensive Study of Satisfaction with

Consumer Services," Consumer Satisfaction, Dissatisfaction and Complaining Behavior, Day (ed.): 64-70. Day, Ralph L. and Stephen B. Ash (1979), "Consumer Responses to Dissatisfaction with Durable Products," Advances in Consumer Research, Vol. 6, William Wilkie (ed.):

438-444. Day, Ralph, and E. Laird Landon, Jr. (1976), "Collecting Comprehensive Consumer Complaint Data By Survey Research," Advances in Consumer Research, Vol. 3, B.B. Andersen

(ed.): 263-268. Easterbook, Gregg (1987), "The Revolution in Medicine," Newsweek, January 26,1987:

40-74. Engel, James F. and Roger D. Blackwell (1982), Consumer Behavior, 4th Edition, Dryden

Press. Farrel, D. (1983), "Exit, Voice, Loyalty and Neglect as Responses to Job Dissatisfaction: A

Multidimensional Scaling Study," Academy of Management Journal, 26: 596-607. Feick, Lawrence (1987), "Latent Class Models for the Analysis of Behavioral Hierarchies,"

Journal of Marketing Research, (May): 174-186. Folkes, Valerie S. (1984), "Consumer Reactions to Product Failure: An Attributional

Approach," Journal of Consumer Research, 10 (March): 393-409. Forbes, J. D., David Tje and Shirley Taylor (1985), "Toward a Model of Consumer Post-Choice

Response Behavior," Paper presented at Association of Consumer Research

Meetings. Fornell, Claes and David Larcker (1981), "Evaluating Structural Equations Models with

Unobservable Variables and Measurement Error," Journal of Marketing Research, 18

(February): 39-50. Fornell, Claes and Nicholas M. Didow (1980), "Economic Constraints on Consumer Complaining

Behavior," Advances in Consumer Research, 7: 318-323. Fornell, Claes and William T. Robinson (1983), "Industrial Organization and Consumer

Satisfaction/Dissatisfaction," Journal of Consumer Research, 9 (March): 403-412. Gaedeke, Ralph (1972), "Filing and Disposition of Consumer Complaints: Some Empirical

Evidence," Journal of Consumer Affairs, 9 (Winter): 205-211. Gibson, Williams (1981), "Medical Malpractice: A Reoccurring Crisis," Journalof Insurance,

42 (November-December): 12-16. Graham, J. W. (1986), "Principled Organizational Dissent: A Theoretical Essay," Research

in Organizational Behavior, Staw and Cummings (eds.), 8: 1-52. Haefner, James and John Lackenby (1975), "Consumers' Use and Awareness of Consumer

Protection Agencies," Journal of Consumer Affairs, 9 (Winter): 205-211. Hermann, Robert O. (1982), "The Consumer Movement in Historical Perspective" in

Consumerism: Search for Consumer Interest, Aaker and Day (eds.), Free Press. Hirschman, Albert (1970), Exit, Voice and Loyalty: Response to Decline in Firms, Organizations

and States, Cambridge, MA: Harvard University Press. Howard, John and Jagdish N. Seth (1969), The Theory of Buyer Behavior, New York: John

Wiley and Sons. Joreskog, Karl and Dag Sorbom (1985), LISREL VI: Analysis of Structural Relationships by

the Method of Maximum Likelihood, University of Uppsala, Sweden. Krishnan, S. and Valerie A. Valle (1979), "Dissatisfaction-Attributions and Consumer Complaint

Behavior," Advances in Consumer Research, Wilkie (ed.), 6: 445-449. Landon, E. Laird Jr. (1980), "The Direction of Consumer Complaint Research," Advances in

Consumer Research, Olson (ed.): 335-338. Landon, E. Laird Jr. (1977), "A Model of Consumer Complaint Research," Consumer Satisfaction

Dissatisfaction and Complaining Behavior, Day (ed.): 31-35. Lunderstorm, William and Lawrence Lamont (1976), "The Development of a Scale to Measure

Consumer Discontent," Journal of Marketing Research, 13 (November): 373-381. Marketing News (1987), "Medics' Ads Appeal for Public Support in Malpractice Crisis,"

September 11, 1987: 32. O'Grady, Kevin (1982), "Measures of Explained Variance: Cautions and Limitations,"

Psychological Bulletin, 92 (November): 766-777. Oliver, Richard (1986), "An Investigation of the Interrelationship Between Consumer

(Dis)Satisfaction and Complaint Reports," Advances in Consumer Research, Wallendorf

and Andersen (eds.): 218-222. Oliver, Richard L. (1980), "A Cognitive Model of the Antecedents and Consequences of

Satisfaction Decisions," Journal of Marketing Research, 17 (November): 460-469. Peterson, Esther (1974), "Consumerism as a Retailer's Asset," Harvard Business Review, 52

(May-June): 91-101. Resnik, Alan and Robert Harmon (1983), "Consumer Complaints and Managerial Response:

A Holistic Approach," Journal of Marketing Research, 47 (March): 86-97. Richins, Marsha L. (1982), "An Investigation of Consumer's Attitude Toward Complaining,"

Advances in Consumer Research, Andrew Mitchell (ed.): 502-506. Richins, Marsha L. (1979), "Consumer Complaining Process: A Comprehensive Model,"

New Dimensions of Consumer Satisfaction and Complaining Behavior, Day and Hunt

(eds.), 9: 502-506. Robinson, Larry (1979), "Consumer Complaint Behavior: A Review with Implications for

Further Research," New Dimensions of Consumer Satisfaction and Complaining Behavior,

Day and Hunt (eds.): 41-50. Sarat, A. (1977), "Studying American Legal Culture: An Assessment of Survey Evidence,"

Law and Society Review, 11 (Fall): 427-488. Schutz, Howard and Roger Dickinson (1987), "Small Claims Court Litigant Satisfaction: It's

How They Play the Game That Counts," Consumer Satisfaction, Dissatisfaction and

Complaining Behavior, Hunt and Day (eds.) (forthcoming). Sheth, Jagdish (1970), "Measurement of Multidimensional Brand Loyalty of a Consumer,"

Journal of Marketing Research, 7 (August): 348-354. Singh, Jagdip (1988), "Consumer Complaint Intentions and Behavior: Definitional and

Taxonomical Issues," Journal of Marketing, 52 (January): 93-107. Singh, Jagdip and Roy D. Howell (1985), "Consumer Complaining Behavior: A Review and

Prospectus," Consumer Satisfaction/Dissatisfaction and Consumer Complaint Behavior

Proceedings, Day and Hunt (eds.), Indiana University. Spencer, Daniel (1986), "Employee Voice and Employee Retention," Academy of Management

Journal, 29 (3): 488-502. TARP (1986), Consumer Complaint Handling in America: An Update Study, White House

Office of Consumer Affairs, Washington, D.C. Thorelli, Hans (1975), "Consumer Redress/Directions and Options," in Consumer Complaints:

Public Policy Alternatives, Davita and McLaughlin (eds.), Acropolis Books. Triandis, Harry C. (1977), Interpersonal Behavior, Monterey, CA: Brooks/Cole.

Ursic, Michael (1985a), "A Model of the Consumer Decision to Seek Legal Redress," Journal

of Consumer Affairs, 19: 20-36. Ursic, Michael L. (1985b), "Consumer Use of the Legal Redress System," Southern Marketing

Association Proceedings: 273-275. Weiner, Bernard (1980), Human Motivation, New York: Holt, Rinehart, and Winston. Westbrook, Robert A. (1980), "Intra-Personal Affective Influences on Consumer Satisfaction.

with Products," Journal of Consumer Research, 7 (June): 49-54. Wilkie, William (1985), Consumer Behavior, New York: John Wiley and Sons.

PHOTO : FIGURE 1 The Proposed Model of Third Party Redness Actions

PHOTO : FIGURE 2 The Empirical Model

PHOTO : FIGURE 3 The Estimated Final Model (only significant paths are shown)

Jagdip Singh is an Assistant Professor of Marketing at the Weatherhead School of Management, Case Western Reserve University, Cleveland, Ohio 44106.
COPYRIGHT 1989 American Council on Consumer Interests
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1989 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Singh, Jagdip
Publication:Journal of Consumer Affairs
Date:Dec 22, 1989
Previous Article:Using content analysis to understand the consumer movement.
Next Article:Consumer correspondence: an exploratory investigation of consistency between business policy and practice.

Related Articles
Complaint behavior of Mexican-American consumers to a third-party agency.
Industry characteristics and consumer dissatisfaction.
Retail attributes: influence on shopping mode choice behavior.
Usefulness of economics in explaining consumer complaints.
'Lack of teeth' in patients' complaints.
Media myth-making to be studied. (News Briefs/Nouvelles en Bref).
Research Program on Pharmaceutical Policy Issues.
Consumer satisfaction with community mental health care in Durban.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters