Predicting Product Purchase and Usage: the Role of Perceived Control, Past Behavior and Product Involvement

Citation:

Flora Kokkinaki (1999) ,"Predicting Product Purchase and Usage: the Role of Perceived Control, Past Behavior and Product Involvement", in NA - Advances in Consumer Research Volume 26, eds. Eric J. Arnould and Linda M. Scott, Provo, UT : Association for Consumer Research, Pages: 576-583.

Advances in Consumer Research Volume 26, 1999      Pages 576-583

PREDICTING PRODUCT PURCHASE AND USAGE: THE ROLE OF PERCEIVED CONTROL, PAST BEHAVIOR AND PRODUCT INVOLVEMENT

Flora Kokkinaki, London Business School

INTRODUCTION

The concept of attitudes occupies a central position in the study of consumer behavior. Attitudes are thought to underlie and motivate behavior and thus to contribute to its explanation and prediction. The Theory of Reasoned Action (TRA, Ajzen and Fishbein 1980, Figure 1) is a dominant approach to modeling attitudinal influence on behavior (Bagozzi, Baumgartner and Yi 1992). Based on the assumption that individuals make rational use of available information when making behavioral decisions, the TRA views behavioral intentions as the immediate determinants of behavior. Intentions, in turn, are determined by attitudes toward the behavior (the individual’s evaluation of the behavior) and subjective norms (the perceived social pressure to perform or not to perform the behavior). Despite the successful application of the TRA in a variety of contexts and behaviors (for meta-analyses see Farley, Lehmann and Ryan 1981; Sheppard, Hartwick and Warshaw 1988) several modifications and extensions have been suggsted.

The Theory of Planned Behavior (TPB, Ajzen 1985, 1991) was developed in order to broaden the range of behaviors that can be explained by the TRA, to include those that are under incomplete volitional control. This model integrates perceived behavioral control (the perceived ease or difficulty of performing the behavior) as an additional determinant of intentions and behavior (Figure 1). Several studies comparing the predictive validity of the TRA and the TPB show that the former model performs just as well as the latter (e.g., Fishbein and Stasson 1990). Other studies, however, indicate that the inclusion of a control variable improves substantially the prediction of intentions and behavior (see Ajzen 1991).

A number of researchers have further questioned the sufficiency of the TRA in predicting habitual behaviors and have suggested the integration of past behavior and habit in the model (e.g., Landis, Triandis and Adamopoulos 1978). The TRA (and the TPB) maintain that past behavior is reflected on attitudes, subjective norms and perceived control, as the beliefs underlying these constructs are, at least in part, based on past experience (Ajzen 1991). However, Triandis (1980) argues that for frequently performed, overlearned behaviors, one might act more or less automatically, as a consequence of habit alone, without the mediation of cognitive variables. Empirical tests offer mixed support to Triandis’ approach as the inclusion of past behavior does not consistently improve the prediction of behavior (see Triandis 1980). Although Triandis did not address the dependence of intentions on past behavior, other researchers have shown that past behavior can also influence intentions and hence affect behavior indirectly (e.g., Bagozzi and Kimmel 1995).

Despite the extensive empirical scrutiny that the TRA has undergone, the issue whether perceived control and past behavior should be integrated in the model remains unresolved (Ajzen 1991). The first objective of the present study is to examine the predictive validity of the TRA and the incremental contribution of perceived control and past behavior, in the context of two behaviors thought to differ in terms of volitional control and frequency of performance.

The study applies the TRA and its extended versions (see Figure 1) in the prediction of intentions to purchase and use a durable, technical product. More specifically, the models are applied in the prediction of intentions to "buy a personal computer within the next nine months" and in the prediction of intentions to "use a computer in the following two weeks" and of actual product usage. According to the TPB, the impact of perceived control on intentions and behavior becomes stronger as volitional control over the behavior decreases (Madden, Ellen and Ajzen 1992). The purchase behavior under study is quite problematic in terms of control (e.g., affordability of the product, availability of technical knowledge/advice required for product selection), especially for the specific, student population involved in the study. It is therefore expected that perceived control contributes significantly to the prediction of purchase intentions. The following hypothesis is tested:

H1: Perceived control improves significantly the prediction of purchase intentions.

In contrast, respondents are not expected to perceive usage behavior to pose problems of control (e.g., in terms of availability of the product and technical knowledge/expertise required for product usage). Hence, usage intentions are expected to be based on attitudes and subjective norms and not to be influenced by perceived control.

Further, following Triandis’ (1980) rationale, it is assumed that the frequency of ast performance of a behavior determines the sufficiency of intentions to predict it. It is expected that past behavior contributes significantly to the prediction of usage intentions and/or behavior, as using the product is a repeatedly performed behavior and therefore, habit and past behavior come into play. The following hypothesis is tested:

H2: Past behavior improves significantly the prediction of usage intentions and/or behavior.

In contrast, because purchase of expensive, durable products, such as personal computers, constitutes a discrete, infrequent behavior, the influence of past behavior (e.g., past purchase of the product) is expected to be fully anticipated in attitudes, subjective norms and perceived control and the beliefs on which these constructs are based.

The TRA and the TPB assume that the predictive weight of attitudes and other behavioral determinants varies across individuals, behaviors and situations (Ajzen 1991). However, the models do not specify what factors moderate their relative impact on behavior. A large body of research, conducted outside the boundaries of the models, has identified a number of variables (e.g., involvement with the attitude object, attitude accessibility and attitude certainty) that determine the magnitude of the attitude-behavior relation (for a meta-analysis see Kraus 1995). Recently, these variables have been integrated into the more general construct of attitude strength (Krosnick and Petty 1995). Strong attitudes have a strong impact on behavior, are temporally stable and resistant to change.

The Elaboration Likelihood Model (ELM) of attitude strength (Petty, Haugtvedt and Smith 1995) approaches attitude strength from a persuasion perspective and views involvement as a critical determinant of attitude strength. According to the model, high involvement during attitude formation leads to extensive elaboration of persuasive information, which results in attitudes that are more accessible and structurally consistent, are held with greater certainty/confidence, are associated with more extensive object-related knowledge and thus have stronger behavioral consequences.

FIGURE 1

THE TRA AND ITS EXTENDED VARIANTS (ADAPTED FROM AJZEN AND FISHBEIN 1980; AJZEN 1991; BAGOZZI AND KIMMEL 1995)

Despite the importance and extensive investigation of the effects of involvement on attitude strength and attitude-behavior consistency, very limited attention has been paid to the effects of the variable within more complex attitude models. The second objective of the present research is to examine the moderating role of involvement within the TRA and its extended variants, and thus, to contribute to the identification of conditions that promote or restrain the relative importance of behavioral determinants.

The present conceptualization of involvement resembles Houston and Rothschild’s (1978) enduring product-class involvement, which refers to an ongoing concern with a product class and reflects the (relatively stable over time) feelings of interest, enthusiasm, and excitement consumers might have about a specific product category (Zaichkowsky 1985). The selected product category of personal computers is expected to induce a wide range of involvement levels across individuals and thus to provide an appropriate research opportunity. It is assumed that product involvement also reflects involvement with product related behaviors. In other words, it is assumed that individuals who are highly involved with a product class also attach greater importance to their product related behaviors, as compared to less involved individuals.

Several studies have shown that high involvement enhances attitude-behavior consistency (e.g., Nederhof 1989). Involvement has also been shown to moderate attitude-intention (Petty, Cacioppo and Schumann 1983) and intention-behavior correspondence (Pieters and Verplanken 1995). Research has also examined the moderating role of involvement in the subjective norm-intention relationship. Most studies show the subjective norm-intention link to be stronger when involvemet is low (e.g., Nederhof 1989).

On the basis of these findings, and following the predictions of the ELM of attitude strength, it is expected that involvement moderates the relative weight of predictor variables within the TRA and its extended versions. Individuals who are highly involved with the product are expected to hold strong attitudes i.e., attitudes based on careful consideration of possible positive or negative consequences of the behavior, and therefore to form their intentions on the basis of these attitudes. In contrast, less involved individuals are expected to hold weaker attitudes and therefore to be more susceptible to social influence, more likely to be prevented from forming strong behavioral intentions by anticipated problems of control and more likely to base their intentions on pre-established behavioral patterns. The following hypothesis is tested:

H3: Involvement moderates the relative contribution of attitudes, subjective norms, perceived control and past behavior to the prediction of intentions.

Involvement is also expected to moderate the relative weight of intentions and past behavior in the prediction of actual behavior. Following a rationale similar to that of the ELM of attitude strength (Petty et al, 1995), Pieters and Verplanken (1995) attribute the moderating effect of involvement on intention-behavior consistency to the motivational nature of the variable and its impact on cognitive processing during intention formation. It is reasonable to assume that when intentions are formed through extensive processing, the effect of past behavior is mediated by this construct, as the TRA and TPB assume. However, when involvement and elaboration are low, it is more likely that habit and past behavior have an automatic, direct effect on behavior, unmediated by intentions. The following hypothesis is tested:

H4: Involvement moderates the relative contribution of intentions and past behavior to the prediction of actual behavior.

METHOD

Respondents

Seventy-eight students at the University of London completed a questionnaire examining their intentions to buy a computer in the following nine months (31 male, mean age 24.5 years). A different sample of 69 students completed a questionnaire assessing their intentions to use a computer in the next two weeks (34 male, mean age 24.6). Two weeks after the completion of this initial questionnaire, respondents were contacted again and reported their actual product usage.

Questionnaire measures

Product involvement (PI) was assessed with the Personal Involvement Inventory (Zaichkowsky 1985) which consists of 20 7-point bipolar items (e.g., important/unimportant, interesting/boring). Responses to these items were summed to provide a measure of product involvement (alpha=.96 for both samples). Table 1 presents the operationalization of the remaining variables. It should be noted that the measure of usage intentions was based on a preliminary study with a similar sample.

RESULTS

As can be seen in Table 2, purchase intentions (BI_PB) were strongly correlated with attitudes (AT_PB), subjective norms (SN_PB) and perceived control (PC_PB) and weakly orrelated with past behavior (PB_PB). Involvement (PI) was weakly related to all these variables. Respondents held a slightly positive AT_PB, while mean SN_PB and PC_PB were slightly negative. Actual usage behavior (BE_UB) was strongly correlated with both usage intentions (BI_UB) and past behavior (PB_UB), while BI_UB was strongly correlated with attitudes (AT_UB), subjective norms (SN_UB) and past behavior (PB_UB) and weakly correlated with perceived control (PC_UB, see Table 3). Although PI was moderately to strongly correlated with most variables, it was not significantly correlated with BE_UB. It should be noted that PC_UB was very high, indicating that respondents did not perceive the behavior to pose strong problems of control.

In order to examine whether perceived control improves the prediction of intentions, purchase and usage intentions were regressed on perceived control after the effect of attitudes and subjective norms had been partialed out. Consistent with H1, the addition of PC_PB explained a further 23% of the variance in BI_PB, producing a significant increment in R² (Table 4). In fact, PC_PB was the strongest predictor of all, a finding indicating that intentions were influenced more by the perceived ease or difficulty of performing the behavior than by respondents’ evaluation of the behavior or by the perceived social pressure. However, the contribution of PC_UB in the prediction of BI_UB was not significant and did not increase significantly the amount of explained variance (Table 5). Similarly, PC_UB did not increase significantly the amount of explained variance in BE_UB (Table 6). These findings indicate that perceived control does not always improve the prediction of intentions and behavior and suggest that the predictive value of the variable depends on whether a target behavior is associated with problems of control.

To examine the contribution of PB_PB in the prediction of BI_PB, PB_PB was entered into the equations after AT_PB, SN_PB and PC_PB. As expected PB_PB did not produce a significant increase in R² and did not reach significance levels (Table 4). However, PB_UB explained an additional 35% of the variance in BI_UB and, in fact, when the variable was entered into the equation all other predictors became insignificant (Table 5). This finding supports H2 and indicates not only that the effect of PB_UB was not fully covered by AT_UB, SN_UB and PC_UB, but also that PB_UB was a superior predictor than these constructs. However, PB_UB was not a significant predictor of BE_UB and did increase significantly R² (Table 6). Overall, these findings suggest that, at least in the case of frequently performed behaviors, the inclusion of past behavior improves prediction over the components of the TRA and the TPB. However, past behavior does not necessarily have a direct influence on behavior. Instead the effect of the variable can be indirect, mediated by intentions.

A series of moderated regression analyses were performed, in order to examine whether involvement moderates the weight of predictor variables. All interaction terms were entered into the equations after the effect of involvement and all predictor variables had been partialed out. As can be seen in Table 7, none of the interaction terms contributed significantly to the prediction of BI_PB. However, in the case of BI_UB, PI significantly moderated the effect of AT_UB (Table 8). Although the interpretation of this effect is obscured by the high PI-AT_PB correlation, the positive sign of the beta weight might indicate that high involvement is associated with an increased attitudinal influence on intentions. However, contrary to H3, PI did not moderate the effect of the remaining predictors. Although the beta coefficients of interaction terms, in both the purchase and usage intention equations, followed a similar pattern, indicative of an increased attitudinal influence and a decresed influence of other predictors, none of these moderating effects was reliable.

Table 9 presents the results of the moderated regression analyses for actual product usage. These results provide mixed support for H4. PI moderated significantly the effect of PB_UB on BE_UB in a way such that high PI was associated with a decrease in the impact of PB_UB. However, PI did not interact significantly with BI_UB, although it should be noted that this moderating effect was only marginally insignificant.

DISCUSSION

The TRA has dominated attitude theory and research for over two decades and is expected to remain an influential approach for years to come (Tesser and Shaffer 1990). Despite the extensive attention that the model has received both in social psychological and in consumer research, several aspects of the model require further refinement. One notable issue surrounding the TRA concerns the predictive value of the model and its individual components across situations. The objective of the present study was twofold: First, to examine whether the nature of a target behavior determines the incremental predictive contribution of perceived control and past behavior. Second, to examine the role of involvement as a potential moderator of the relative importance of the model’s components.

More specifically, the study examined the contribution of perceived control and past behavior to the prediction of product purchase and usage, two behaviors which differ in frequency of performance and degree of volitional control. Overall the findings support the assumption that the predictive value of these variables depends on the nature of the behavior. Perceived control was a strong predictor of purchase intentions and the components of the TPB afforded the optimal prediction. In the prediction of usage intentions however, the addition of perceived control did not increase the amount of explained variance above that afforded by the TRA. Although using a computer depends on certain requisite factors (e.g., technical skills/advice), it seems that respondents on the whole perceived themselves having good control over these factors and therefore formed their intentions on the basis of their attitudes and subjective norms. In contrast, control factors associated with the purchase behavior (e.g., money) are not easily controlled by the specific population. Individuals who believed they lacked the opportunity to perform the behavior were less likely to form strong intentions, even if their attitudes and subjective norms were favorable. Under these circumstances, intentions were determined by all three predictor variables, and in fact, perceived control was the strongest predictor. These results suggest that perceived control needs to be included in the TRA only when the target behavior is perceived to depend on the provision of certain resources and skills.

As expected, past behavior did not improve the prediction of purchase intentions. Although usage intentions sufficed for the optimal prediction of usage behavior, the TRA performed less well in the prediction of these intentions. The relatively low amount of explained variance radically increased when past behavior was entered into the equation and once past behavior was taken into account attitudes and subjective norms no longer significantly predicted intentions. This finding suggests that the effect of past behavior is not always covered by attitudes, subjective norms and perceived control and supports the argument that, at least in the case of frequently performed behaviors, the inclusion of past behavior is essential.

TABLE 1

OPERATIONALIZATION OF VARIABLES

TABLE 2

MEANS, STANDARD DEVIATIONS AND CORRELATIONS OF SELECTED VARIABLES

TABLE 3

MEANS, STANDARD DEVIATIONS AND CORRELATIONS OF SELECTED VARIABLES

TABLE 4

REGRESSION OF BI ON AT_PB, SN_PB, PC_PB, AND PB_PB

TABLE 5

REGRESSION OF BI_UB ON AT_UB, SN_UB, PC_UB AND PB_UB

TABLE 6

REGRESSION OF BE_UB ON BI_UB, PC_UB AND PB_UB

TABLE 7

REGRESSION OF BI_PB ON AT_PB, SN_PB, PC_PB, PI_PB, AT_PB x PI, SN_PB x PI AND PC_PB x PI

TABLE 8

REGRESSION OF BI_UB ON AT_UB, SN_UB, PB_UB, AT_UB x PI, SN_UB x PI AND PB_UB x PI

TABLE 9

REGRESSION OF BE_UB ON BI_UB, PB_UB, BI_UB x PI AND PB_UB x PI

Together these findings suggest that the nature of a target behavior dictates what factors influence individuals’ behavioral decisions. For behaviors that are perceived to be under good volitional control the addition of a control variable appears to be redundant. Similarly, for discrete, non repetitive behaviors a measure of past behavior adds little to behavoral prediction. That seems to be particularly true in the case of purchase and usage of expensive, durable products. In any case, the findings suggest that researchers interested in the prediction of behavior should consider the nature of the behavior under study when selecting which model to apply.

The results concerning the moderating role of involvement provide mixed support for the hypothesized effects of the variable. No reliable effects were observed in the case of purchase intentions. However, product involvement was found to moderate the influence of attitudes on usage intentions and the effect of past behavior on actual usage behavior. Although the results did not consistently support the hypotheses, they seem to indicate that high involvement tends to increase the relative weight of attitudes and intentions and to decrease the weight of other predictors. The motivational nature of involvement is thought to account for these effects. According to the ELM of attitude strength (Petty et al. 1995), involved individuals scrutinize information related to the attitude object and the possible consequences of behaviors directed towards this object and, therefore, the relevant attitudes are strong and more influential than other factors. It is also possible that involvement has similar effects on the cognitive effort expended during intention formation. All the variables modeled in the TRA and the TPB are the end result of active information processing. When involvement is low, individuals might not expend sufficient effort in forming their intentions (i.e., when considering their attitudes, subjective norms and perceived control) and therefore intentions might not be well-formed (see Bagozzi and Yi 1989). In such cases, behavior might be directly influenced by pre-established behavioral patterns (i.e., past behavior) without the mediation of intentions.

Although the findings provide some support for the hypothesized moderating role of involvement and suggest that the predictive value of past behavior and perceived control depends on the nature of a target behavior, the present research is limited in several aspects. The most important limitation concerns the correlational nature of the data which restrains any causal interpretation of the results. Experimental designs in future research, involving more representative samples, are necessary before any conclusions can be drawn and generalized. Further, most of the explanations offered here for the observed effects are speculative and evidence is required before they can be accepted. The findings of the present research should therefore be viewed only as a starting point for future investigations.

REFERENCES

Ajzen, Icek (1985), "From Intentions to Actions: A Theory of Planned Behavior," In Action-control: From cognition to behavior, eds. Julius Kuhl and Jeurgen Beckmann, New York: Springer, 11-39.

Ajzen, Icek (1991), "The theory of planned behavior," Organizational Behavior and Human Decision Processes, 50 (December), 179-211.

Ajzen, Icek and Martin Fishbein (1980), Understanding attitudes and predicting social behavior, Englewood Cliffs, NJ: Prentice-Hall.

Bagozzi, Richard P., Johann Baumgartner and Youjae Yi (1992), "State versus action orientation and the theory of reasoned action: An application to coupon usage," Journal of Consumer Research, 18 (March), 505-518.

Bagozzi, Richard P. and Susan K. Kimmel (1995), "A comparison of leading theories for the prediction of goal-directed behaviors," British Journal of Social Psychology, 34, 437-461.

Bagozzi, Richard P. and Youjae Yi (1989), "The degree of intention formation s a moderator of the attitude-behavior relationship," Social Psychology Quarterly, 52 (December), 266-279.

Farley, John. U., Donald R. Lehmann, and Michael J. Ryan (1981), "Generalizing from imperfect replication," Journal of Business, 54 (October), 597-610.

Fishbein, Martin and Mark Stasson (1990), "The role of desires, self-predictions, and perceived control in the prediction of training session Attendance," Journal of Applied Social Psychology, 20 (February), 173-198.

Houston, Michael J. and Michael L. Rothschild (1978) "Conceptual and methodological perspectives in involvement," In Research frontiers in marketing: Dialogues and directions, ed. S. Jain, Chicago, IL: American Marketing Association, 184-187

Kraus, Stephen J. (1995), "Attitudes and the prediction of behavior: A meta-analysis of the empirical literature," Personality and Social Psychology Bulletin, 21 (January), 58-75.

Krosnick, Jon A. and Richard E. Petty (1995), "Attitude strength: An overview," In Attitude Strength: Antecedents and Consequences, eds. Richard E. Petty and Jon A. Krosnick, Hillsdale, NJ: Erlbaum, 1-24.

Landis, Dan, Harry Triandis and John Adamopoulos (1978), "Habit and behavioral intentions as predictors of social behavior," Journal of Social Psychology, 106 (December), 227-237.

Madden, Thomas. J., Pamela S. Ellen and Icek Ajzen (1992), "A comparison of the theory of planned behavior and the theory of reasoned action," Personality and Social Psychology Bulletin, 18 (February), 3-9.

Nederhof, Anton J. (1989), "Self-involvement, intention certainty and attitude-intention consistency," British Journal of Social Psychology, 28 (June), 123-133.

Petty, Richard. E., John T. Cacioppo and David Schumann (1983), "Central and peripheral routes to advertising effectiveness: The moderating role of involvement," Journal of Consumer Research, 10 (September), 135-146.

Petty, Richard. E., Curtis P. Haugtvedt and Stephen M. Smith (1995), "Elaboration as a determinant of attitude strength: Creating attitudes that are persistent, resistant, and predictive of behavior. In Attitude Strength: Antecedents and Consequences, eds. Richard E. Petty and Jon A. Krosnick, Hillsdale, NJ: Erlbaum, 93-130.

Pieters, Ric G. and Bas Verplanken (1995), "Intention-behavior consistency: Effects of consideration set size, involvement and need for cognition. European Journal of Social Psychology, 25, 531-543.

Sheppard, Blair H., Jon Hartwick, J. and Paul R. Warshaw (1988), "The theory of reasoned action: Meta-analysis of past research with recommendations for modifications and future research," Journal of Consumer Research, 15 (December), 325-345.

Tesser, Abraham and David R. Shaffer (1990), "Attitudes and Attitude Change," Annual Review of Psychology, 41, 479-523.

Triandis, Harry. C. (1980), "Values, attitudes and interpersonal behavior," In Nebraska Symposium on Motivation, eds. H. E. Howe, Jr. and M. M. Page, Vol. 27, Lincoln: University of Nebraska Press, 195-259.

Zaichkowsky, Judith Lynne (1985), "Measuring the involvement construct," Journal of Consumer Research, 12 (December), 341-352.

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