Purchasing Generic Prescriptions Drugs: an Analysis Using Two Behavioral Intention Models

ABSTRACT - The present study used Fishbein's theory of reasoned action and a subjective probability model to examine the decision to purchase generic prescription drugs. Two samples were used to assess the robustness of each model's predictions In addition, those factors that differentiate intenders from nonintenders were examined. Both models were found to predict intention in both a student and nonstudent sample. Using the components of the Fishbein model, several differences were found between: (1) individuals who intend to purchase generic prescription drugs with those who do not and (2) the student and nonstudent sample The implications of these findings for both basic and applied researchers were discussed.

Citation:

David Brinberg and Vicki Cummings (1984) ,"Purchasing Generic Prescriptions Drugs: an Analysis Using Two Behavioral Intention Models", in NA - Advances in Consumer Research Volume 11, eds. Thomas C. Kinnear, Provo, UT : Association for Consumer Research, Pages: 229-234.

Advances in Consumer Research Volume 11, 1984      Pages 229-234

PURCHASING GENERIC PRESCRIPTIONS DRUGS: AN ANALYSIS USING TWO BEHAVIORAL INTENTION MODELS

David Brinberg, Baruch College, CUNY

Vicki Cummings, University of Maryland

[The computer time for this project was supported through the facilities of the Computer Science Center at the University of Maryland.]

ABSTRACT -

The present study used Fishbein's theory of reasoned action and a subjective probability model to examine the decision to purchase generic prescription drugs. Two samples were used to assess the robustness of each model's predictions In addition, those factors that differentiate intenders from nonintenders were examined. Both models were found to predict intention in both a student and nonstudent sample. Using the components of the Fishbein model, several differences were found between: (1) individuals who intend to purchase generic prescription drugs with those who do not and (2) the student and nonstudent sample The implications of these findings for both basic and applied researchers were discussed.

INTRODUCTION

Until recently, consumers have had little or no input into the type of prescription drug product purchased. With the support of consumer groups and federal government agencies, many states are repealing anti-substitution laws, thereby increasing the consumer's access to generic prescription drugs A model law was presented by the Federal Trade Commission in 1979, but most states adopted their own individual versions (Drug Topics 1979) Currently, forty-eight states have adopted some type of generic substitution law

Many state anti-substitution laws have several common features. The Drug Product Selection Law enacted in 1977 to replace New York's previous anti-substitution laws is useful to illustrate these common features: (1) an approved formulary (positive substitution), (2) a two-sided prescription form where the physician signs that substitution is or is not permitted, (3) mandatory substitution by the pharmacist if substitution is approved by the physician, and (4) the pharmacist having the option to pass on to the consumer the difference in the price between the brand and generic drug.

By 1984, many leading brand name drugs (2.g., valium) will go off patent, dramatically increasing the number of pharmaceutical companies that will produce generic equivalents. Given the rapid growth in the consumers' use of generic prescription drugs, the application of different decision models is likely to increase our understanding of those factors that influence a consumer's decision to purchase a generic prescription drug. At the same time, the use of different decision models is likely to increase our understanding of the scope and limits of these models. Several models have been developed to predict an individual's decision and to identify those factors that determine this decision (e g , Fishbein and Ajzen 1975; Jaccard and King 1977). The two models used in the current study are Fishbein's theory of reasoned action (Fishbein 1979) and a subjective probability model (Jaccard and King 1977).

Fishbein's Theory of Reasoned Action

Fishbein and Ajzen (1975) have developed a model of reasoned action that is used: (1) to predict an individual's intention and behavior and (2) to identify those factors that determine intention. Intention is viewed as the best predictor of behavior and is determined by an individual's attitude toward the act plus the perceived social pressure to perform the behavior (subjective norm) The relationship among the components may be expressed as follows:

Behavior = Intention = Attitude(w₁) + Subjective norm(w₂)

Attitude is determined by the salient beliefs multiplied by the evaluative aspect of the belief, and the subjective norm is determined by the normative beliefs multiplied by the motivation to comply. These relationships may be expressed as follows :

Attitude = Sb_ie_i         Equation 2

where b_i = perceived likelihood that the behavior in question will result in some outcome and e_i = evaluation of that outcome

Subjective norm = SNbMc         Equation 3

where Nb = belief that a particular referent thinks one should or should not perform the behavior in question and Mc = motivation to comply with that referent.

w₁ and w₂ are theoretical parameters reflecting the relative importance of each component in determining intention. These parameters are generally determined through regression analysis.

The Fishbein model has been tested in a wide variety of cultures (cf. Ajzen and Fishbein 1980) and with a wide variety of behaviors. For instance, fertility-related behaviors (Davidson and Jaccard 1975), blood donation (Pomazal and Jaccard 1976), voting behavior (Fishbein and Coombs 1974), the use of credit unions (Ryan and Bonfield 1980) as well as preventive health care (Olivier and Berger 1979) have all been predicted successfully by using the extended Fishbein model. Beardon and Mason (1978) used Fishbein's attitude model to predict the intention to purchase generic prescription drugs and found a disaggregate model to predict accurately a personal intention

Subjective Probability Model

Wyer and Goldberg (1970) extended McGuire's (1960) work on syllogistic reasoning and postulated that subjective probabilities combine in a manner consistent with the laws of objective probability theory. This model may be expressed as follows:

P_B = P_"P_B|A + P_"P_B|A        Equation 4

Where: P_" = the probability that proposition A is true;

P_B|A = the probability that proposition B is true given that proposition A is true; P_" = the probability that proposition A is not true and P_B|A = the probability that proposition B is true given that proposition A is not true.

This model was extended by Jaccard and King (1977) to predict behavioral intention. The definitions of the components of this model and their interrelationships are presented below:

a) A belief is the relation between two objects; with objects being used in its most generic sense

b) The belief strength is the subjective probability linking the two objects. It is determined on a scale from 1.00 (extremely likely) to 0 0 (extremely unlikely).

c) The Behavioral Intention is the likelihoot of performing some future behavior and is assessed using a subjective probability estimate varying from .00 to 1.0. Intention does not basically differ from other types of beliefs, although it does possess certain distinctive characteristics; it always links oneself and some action, refers to future behavior and usually correlated with overt behavior.

d) The conditional probability (PI|B) that is, given the belief is true (or false), what is the likelihood of an intention" (Jaccard, Knox and Brinberg (1979)

This model may be expressed as follows

P_I = P_BP_I|B + P_BP_I|B        Equation 5

where P_I = an individual's intention to perform a certain behavior; P_I|B= an individual's perceived probability that performing a certain behavior will result in a certain outcome; P_B = an individual's perceived probability of performing a certain behavior given that B is true; P_B = is the perceived probability that the belief is not true (determined by 1-P_B ) and P_I|B = is the individual's perceived probability of performing the behavior given that it does not result in that outcome All probabilities are estimated by the individual except for P_B (i.e., 1 - P_B)

The difference between the two conditional probabilities (i.e., P_I|B - P_I|B) is described as the "psychological relevance" of a belief When the difference is small, the belief is said to be psychologically irrelevant (e.g., a person would purchase generic prescription drugs regardless of whether the manufacturer is easy to identify) However, when the difference between the two conditional probabilities is large, the belief is considered to be psychologically relevant (e.g., a person would purchase a generic prescription drug given it has been adequately tested, but would not purchase the drug if it had not been adequately tested).

The subjective probability model is a normative model and indicates "what man ought to do" Jaccard and King (1977) pointed out that deviations from the model indicate that the logical relations may be interfered with by irrational psychological processes. With this in mint, the model can provide a baseline from which the deviations can be investigated

Research using the subjective probability model does "support the proposition that the relationship between beliefs and behavioral intention can be described according to mathematical probabilities" (Jaccard and King 1977 p. 33). The model has been found to be a significant predictor of intention, with correlations ranging from .69 to 82 for consequences of smoking cigarettes (Jaccard and King 1977), from 76 to .88 for voting intention (Jaccard, Knox and Brinberg 1979), and from .35 to 61 for eating at a fast-food restaurant (Durand 1981). Additional evidence in support of equation 5 is that the total model predicted intention better than any of the parts (Jaccard, Knox and Brinberg 1979)

Comparison of the Fishbein and the Subjective Probability Motel

In both models, intention is viewed as the best predictor of behavior and is hypothesized to mediate the relationship between all external variables and behavior. In addition, beliefs are included in both motels, although related to intention differently. Within the probability model, the relationship between a belief and intention is described by the laws of objective probability, and research using this model has typically related a single belief to intention. However, Jaccard and King (1977) have noted that more than one belief may be included in equation 5 to predict intention. Fishbein's model uses an expectancy-value formulation and includes the sum of the "salient" beliefs (weighted by the evaluation of each belief or the motivation to comply) to predict a person's attitude, subjective norm, and subsequently their intention.

Fishbein makes a distinction between attitudinal and normative beliefs and treats these two types of beliefs (weighted by the evaluation of the belief and the motivation to comply, respectively) as the determinants of intention. In the probability model, no distinction is made between attitudinal and normative beliefs and both may be used to predict intention (see Miniard and Cohen 1981 and Fishbein and Ajzen 1981 for a more detailed discussion concerning the relationship between attitude and subjective norm). In addition, a measure of overall evaluation (i.e., attitude) and overall social pressure (i.e., subjective norm) is included in Fishbein's model whereas the probability model only includes measures of belief strength.

Finally, researchers testing the Fishbein model have typically used regression weights to determine the psychological importance of attitude and subjective norm in predicting intention. Several researchers, however, (e.g., Dawes and Corrigan 1974) have questioned the use of regression coefficients as an index of importance. In the probability model, the "relevance index" is used as an estimate of psychological importance.

Many researchers (e.g., Ferber 1977; Brinberg and McGrath 1983) have expressed concern in using only one type of population to examine a model or set of models. A great deal of psychological and consumer behavior research has been limited to college students and few attempts have been made to determine the robustness of some model with respect to the sample examined. In the present study, the robustness of each behavioral intention model was assessed across two different samples: a college sample from Maryland and a community sample from Rochester, New York.

In summary, the purpose of this study was threefold: (a) to determine the accuracy of each theory in predicting the individual's intention to purchase generic prescription drugs, (b) to determine the robustness of each model across two samples (i e., college and community respondents) and (c) to describe those factors that influence intention, by comparing individuals who intend to purchase generic prescription drugs with those who to not.

METHOD

Sampling Procedure

The community sample was selected from suburban Rochester, New York. City data were used to divide the suburban area into seventeen towns and communities Six of these communities were randomly selected to participate in the study and then were divided into sampling blocks, each consisting of residential areas of approximately fifty homes.

Within each community, a block was randomly selected from which to collect data. All units within a block were included as elements in the sample. Areas sampled included single family homes, multi-family units, condominium complexes, and apartment complexes. The resulting sample consisted of a total of six neighborhoods from six different communities; each composed of approximately fifty households.

The six communities provided a diverse sample with respect to educational levels, ethnic background, and income. Interviews were conducted at different times during the day and evening; up to three trips were made to each neighborhood to increase response rate. The interviewer asked to speak with the individual in the household most likely to purchase a prescription drug if one were needed by some member of the family (household). The respondent then was asked to answer sample questions to get a "feel" for the instrument. Once these were answered, the individual completed the questionnaire. The interviews ranged in length from twenty minutes to one hour (X= 30 minutes). Over the course of three visits, a total of 109 useable questionnaires were completed by respondents. [There were 31 refusals and 13 respondents with incomplete questionnaires. A total of 153 households were contacted.]

Community Sample Characteristics

Respondents' ages ranged from 19 to 84 (X = 38). A total of 52 people (48%) reported drug expenses of less than $50.00 per year, and only 33 (30%) had annual expenses of over $100.00 The sample was well educated with less than 2: of the sample having received less than a high school education. A total of 37 (34%) of the individuals had earned a college degree and 25 (23%) of the respondents had some postgraduate work. The median income ranged between $20,000 to $25,000.

College Sample Characteristics

A college sample was obtained from an introductory consumer economics class at a large mid-Atlantic university. A brief explanation concerning the purpose of the study was provided by the interviewer. The format and items for the student questionnaire were the same as those used for the community sample. The questionnaire was administered to small groups of respondents. A total of 96 respondents were obtained.

Respondents' ages ranged from 18 to 58 (X = 22.8). The median income ranged between 0 to $5,000. Their prescription drug expenditures were also relatively low, with 50 (52%) spending less than $50 per year and only 25 (26%) with greater than $100.

Instrument Development

A pretest was conducted to determine the salient beliefs for both the community and college sample A total of 41 college students were selected from an introductory consumer behavior class, and a total of 47 respondents were randomly selected in a telephone survey from the greater Rochester area. The salient beliefs were determined by asking the individuals to list the advantages, disadvantages and other possible consequences of purchasing a generic prescription drug as well as what relevant others think they should or should not do. Both samples had similar modal salient beliefs (i.e., cost, safety, efficacy, quality, testing of the drug, ease in obtaining drug, ease in identifying the manufacturer, and comprehension about usage of the product). This set of beliefs was similar to those used in related studies (e.g., Beardon and Mason 1978). The relevant others were close friends, relatives, doctors and pharmacists

Based on the salient beliefs obtained in the pretest, a questionnaire was constructed that operationalized the components of both the Fishbein and the subjective probability model. Below are examples of measures obtained for each component of the two models.

Belief = Purchasing a generic prescription drug means that I will be getting a drug that has been adequately tested

Likely _ _ _ _ _ _ _ unlikely

Evaluation = Getting a drug that has been adequately tested is:

good _ _ _ _ _ _ _ bad

Normative Belief = My doctor thinks I should purchase generic prescription drugs.

likely _ _ _ _ _ _ _ unlikely

Motivation to comply = Generally speaking, I want to do or I want not to do what my doctor wants me to do:

want to do _ _ _ _ _ _ _ want not to do

Attitude = My purchasing generic prescription drugs is:

good _ _ _ _ _ _ _ bad

Subjective Norm = Most people who are important to me think I should purchase generic prescription drugs.

likely _ _ _ _ _ _ _ unlikely

PI|B Suppose that in purchasing generic prescription drugs, you are in fact getting high quality drugs. How likely is it that you would purchase generic prescription drugs?

likely _ _ _ _ _ _ _ unlikely

PI|B = Suppose that in purchasing a generic prescription drug, you are, in fact not getting high quality drugs. How likely is it that you would purchase generic prescription drugs?

likely _ _ _ _ _ _ _ unlikely

Intention = I intend to purchase generic prescription drugs:

likely _ _ _ _ _ _ _ unlikely

RESULTS

Predictive Accuracy of the Fishbein Model

For the community sample, the multiple correlation of attitude and subjective norm with intention was 65 ( p < .01), and the standardized regression coefficients were .56 (p < .01) and .16 (p < .05) for the attitude and subjective norm respectively. For the college sample, the multiple correlation of attitude and subjective norm with intention was .63 (p < .01), and the standardized regression coefficients were .44 (p < 01) and 30 (p < .01) for the attitude and subjective norm respectively. The percent of explained variance was highly similar for the two samples (i.e., 42.2: and 39.7% for the community and college sample, respectively). In addition, the zero-order correlations between attitude and subjective- norm with intention were not significantly different for the two samples. These findings provide evidence for the robustness (i.e., generalizability) of this theory across diverse samples.

Predictive Accuracy of the Subjective Probability Model

The predictive accuracy of the probability model was determined by correlating the predicted and obtained intention score. A summary of this analysis may be found in Table 1. The correlations ranged from 35 to .70 (p < .01) for the community sample, with an average correlation of .55. For the college sample, the correlations ranged from .26 to 68 (p < .01), with an average correlation of .48.

The robustness of this model was determined by comparing the correlation of each belief with intention across the two samples Using a Fisher's r-z to compare these correlations,there were no significant differences in the predictive accuracy of the model. As with the Fishbein model, one implication of this finding is that the subjective probability model is robust across both a nonstudent and student population.

TABLE 1

PREDICTIVE ACCURACY OF THE SUBJECTIVE PROBABILITY MODEL

Diagnostic Use of Fishbein's Model

A strategy presented by Ajzen & Fishbein (1980) to determine those factors most likely to influence intention is to identify differences between individuals who intent to perform a behavior with those who to not. Given the four basic components of Fishbein's model (i.e., beliefs, evaluations, normative beliefs, and motivation to comply), each needs to be examined to determine the extent to which intenders are different than nonintenders. To accomplish this comparison, a Hotelling's T2 analysis was conducted for each component of Fishbein's model

When comparing beliefs, intenders were significantly different than nonintenders for both the community (12 = 77 0; p < 01) and college sample (T2 = 27.9; p < .01). A summary of these comparisons may be found in Table 2.

TABLE 2

COMPARISON OF BELIEFS FOR INTENDERS AND NONINTENDERS

TABLE 2

COMPARISON OF BELIEFS FOR INTENDERS AND NONINTENDERS

In both the community and college sample, intenders believed that generic drugs were safer, more adequately tested and of higher quality than nonintenders. Intenders also were less confused about the usage of generic drugs than nonintenders. There were significant differences between the beliefs of the college and community sample (T2 = 17.6; P < .05). The two beliefs significantly different for the two groups were: the ease of identifying the manufacturer (t(203) = 2.86; p < .01; X = -.46 and .28 for the community and college sample, respectively) and the safety of the drug (t(203) = 2 14; p < 05; X = 2.03 and 1.68 for tile community and college sample, respectively).

When comparing the evaluation of the beliefs, intenders were significantly different than nonintenders for both the community (Ti = 29.9; P < .01) and college sample (T2 = 38.9; p < .01). A summary of these comparisons may be found in Table 3.

In the community sample, intenders evaluated the cost of the generic drugs as more positive than the nonintenders. No other evaluations significantly differed between the two groups. For the college sample, intenders evaluated the cost, quality, safety, and effectiveness of generic drugs more positively than nonintenders There were no significant differences between the evaluation of the beliefs when comparing the college and community sample.

For the community sample, the normative beliefs of intenders were significantly different than nonintenders (T2 = 18.8; p < .01). For the college sample, there was a similar trend between intenders and nonintenders (T2 = 10.2; p < 10). In both samples, intenders believed that both close friends and the pharmacist were more likely to think they should purchase generic prescription drugs than nonintenders. A summary of these comparisons may be found in Table 4 There were no significant differences between the college and community sample with respect to the normative beliefs.

Finally, for the community sample, the motivation to comply of intenders was significantly different than nonintenders (T2 = 15.9; p < .01). In the college sample, there was a trend for intenders and nonintenders to differ significantly (T2 = 8 4; p < .10) Intenders in the community sample were more likely to comply with their pharmacist (p < 01) For the college sample, intenders were more likely to comply with their close friends (p < .01). A summary of these comparisons may be found in Table 5.

TABLE 3

COMPARISON OF THE EVALUATION OF THE BELIEFS FOR INTENDERS AND NONINTENDERS

TABLE 4

COMPARISON OF THE NORMATIVE BELIEFS FOR INTENDERS AND NONINTENDERS

TABLE 5

COMPARISON OF THE MOTIVATION TO COMPLY OF INTENDERS AND NONINTENDERS

There were significant differences between the community and college sample with respect to their motivation to comply (T = 14.9; p < .01). The community respondents were less likely to comply with close friends (c(203) = 1.97; D < .05; X = 13) and more likely to comply with a doctor (t(203) = 1.74; p < .10; X = 1.14) than college students (X = 54 and .79 for close friends and doctor, respectively).

Diagnostic Use of the Subjective Probability Model

As noted earlier, the difference between the two conditional probabilities is used as an index of psychological importance Table 5 contains a summary of the relevance scores for each belief for the college and community sample

TABLE 6

RELEVANCE SCORES FOR THE COMMUNITY AND COLLEGE SAMPLES

For each belief, the relevance score is relatively high, suggesting that all beliefs may be considered at least somewhat relevant. In addition, there is a high degree of similarity between the relevance of the beliefs for the community and college sample (r = .99; p < .01).

Both the probability model and the Fishbein model purport to identify beliefs chat, when modified, are likely to change intention. One strategy for identifying these beliefs is to use the relevance index (from the probability model) and the correlation between a belief and intention (i e., the degree to which a belief can differentiate intenders from nonintenders; Ajzen and Fishbein 1980). Convergence between the beliefs identified by using each model can provide support for the use of either approach in identifying psychologically important beliefs. To determine this convergence, the correlation between a belief and intention was converted to a z score and correlated with the mean relevance score for that belief. This correlation was significant (r =.77 and .93; p < .01, for the community and college sample, respectively), indicating high convergence between the two approaches for identifying "psychologically important" beliefs This finding suggests that for this behavior both techniques measure the same underlying construct i.e., psychological importance Thus, either technique may be useful to a researcher interested in identifying psychologically important beliefs in order to change an individual's intention to purchase generic prescription drugs.

DISCUSSION

Both the Fishbein theory of reasoned action and the subjective probability model were significant predictors of an individual ' s intention to purchase generic prescription drugs. In addition, the predictive accuracy of both models was highly similar across two different samples of respondents, i.e., college students in Maryland and a nonstudent sample in suburban Rochester, New York This robustness across two samples increases our confidence in the use of these two models

The findings from the present study have implications for both basic and applied researchers. For basic researchers, the robustness of these models across samples should increase a researchers confidence that these two behavioral intention models are able to accurately predict intention for a sample other than college students. In addition, when indices of importance were derived from both models (i e., the relevance index from the probability model and the correlation between beliefs and intention in Fishbein's motel), these measures were highly correlated, indicating convergence in the measure of psychologically import and beliefs. Thus, either approach may be used to measure important beliefs. Future work needs to be conducted to examine in greater detail the measurement of importance and which (if either) technique is more effective

For applied researcher, these findings also have a number of interesting implications Several concepts within Fishbein's model were found to differ for those individuals who intend to purchase generic prescription drugs with those who do not. These detailed comparisons between intenders and nonintenders provide a profile for each group that educators and marketers may use to develop educational programs or advertisements. For instance, in the community sample, nonintenders believed it was less likely that generic drugs were a safe product than intenders Given this difference a marketer would want to focus their advertisement on the safety of the drug since this belief differentiates nonintenders from intenders. Differences between intenders and nonintenders for the other components of Fishbein's model also may provide useful information in the development of any change program Since the present study only examined generic prescription drugs, future research would be needed to determine whether similar marketing strategies would be developed for branded drugs.

In sum, the present study used two behavioral intention models to examine the decision to purchase generic prescription drugs. Two samples were used to assess the robustness of each model's predictions. In addition, those factors that differentiate intenders from nonintenders were examined. Both models were found to predict intention in both a student and nonstudent sample. Several differences were found between:(l) individuals who intent to purchase generic prescription drugs with those who do not and (2) the student and nonstudent sample. Finally, the implications of these findings for both basic and applied researchers were discussed.

REFERENCES

Ajzen, Icek. and Fishbein, Martin. (1980), "Understanding attitudes and predicting social behavior," Englewood, N.J., Prentice-Hall.

Beardon, William O. & Mason, John B. (1978) "Consumer-perceived risk and attitudes toward generically prescribed drugs." Journal of Applied Psychology, 63, 741-746.

Brinberg, David and McGrath, Joseph E. (1983), "A validity network schema" Paper presented at the AERA.

Davidson, Andrew R. and Jaccard, James (1975), "Population psychology: A new look at an old problem." Journal of Personality and Social Psychology, 31, 1073-1082.

Dawes, Robin M. and Corrigan, Bernard. (1974), "Linear models in decision making." Psychological Bulletin, 81, 95-106.

Drug Topics (1979), Model Law, p. 40.

Durand, Judy. (1981), "Decisions to eat at a fast food restaurant: Can they be predicted and changed." Unpublished Master's thesis. University of Maryland.

Ferber, Robert (1977), "Research by convenience." Journal of Consumer Research, 4, 57-58.

Fishbein, Martin. 1979), "A theory of reasoned action: Some applications and implications." In H. Howe and M. Page (Eds.) Nebraska Symposium of Motivation, Lincoln, Nebraska: University of Nebraska Press.

Fishbein, Martin and Ajzen, Icek (1975), "Beliefs, attitudes, intention and behavior: An introduction to theory and research." Reading, Mass.

Fishbein, Martin and Ajzen, Icek. (1981), "On construct validity: A critique of Miniard and Cohen's paper." Journal of Experimental Social Psychology,17,340-350.

Fishbein, Martin and Coombs, Fred S. (1974), "Basis for decision: An attitudinal analysis of voting behavior." Journal of Applied Social Psychology,4,95-124.

Jaccard, James and King, George W. (1977), "The relation between behavioral intentions and beliefs: A probabilistic model." Human Communications Research, 3, 326-334.

Jaccard, James, Knox, Richard and Brinberg, David. (1979), "Prediction of behavior from beliefs: An extension and test of a subjective probability model," Journal of Personality and Social Psychology, 37, 1239-1248.

McGuire, William J. ( 1960) A syllogistic analysis of cognitive relationships. In Carl Hovland & Milton Rosenberg (Eds.) "Attitude organization and change." New Haven, Conn.: Yale University Press.

Miniard, Paul W. and Cohen, Joel B. (1981), "An examination of the Fishbein-Ajzen behavioral intention model's concepts and measures. Journal of Experimental Social Psychology, 17, 309-339.

Olivier, Richard and Berger, Phillip. (1979), "A path analysis of preventive health care decision motels," Journal of Consumer Research, 6, 113-122.

Pomazal, Richard and Jaccard, James (1976), "An informational approach to altruistic behavior." Journal of Personality and Social Psychology, 33, 317-326.

Ryan, Michael and Bonfield, E.H. (1980), "Fishbein's intention model: A test of external and pragmatic validity," Journal of Marketing, 44, 82-95.

Wyer, Robert S. & Goldberg, Lee. (1970), "A probabilistic analysis of relations among beliefs and attitudes." Psychological Review, 77, 100-120.

----------------------------------------