A Comparison of Three Behavioral Intention Models: the Case of Valentine's Day Gift-Giving

ABSTRACT - Three models of behavioral intention, Ajzen and Fishbein's (1980) theory of reasoned action (TORA), Ajzen's (1985) theory of planned behavior (TOPB), and the Miniard and Cohen (1983) model (MCM) were compared in terms of predictive ability and their ability to effectively separate normative form personal influence. For predicting gift giving intention, TOPB performed better than did TORA and MCM. The results also suggest that MCM exhibited greater ability to separate its components than did TORA or TOPB. For predicting gift giving behavior, results indicate that intention represents the direct antecedent to behavior over that of the perceived behavioral control component of TOPB.

Citation:

Richard G. Netemeyer, J. Craig Andrews, and Srinivas Durvasula (1993) ,"A Comparison of Three Behavioral Intention Models: the Case of Valentine's Day Gift-Giving", in NA - Advances in Consumer Research Volume 20, eds. Leigh McAlister and Michael L. Rothschild, Provo, UT : Association for Consumer Research, Pages: 135-141.

Advances in Consumer Research Volume 20, 1993      Pages 135-141

A COMPARISON OF THREE BEHAVIORAL INTENTION MODELS: THE CASE OF VALENTINE'S DAY GIFT-GIVING

Richard G. Netemeyer, Louisiana State University

J. Craig Andrews, Marquette University

Srinivas Durvasula, Marquette University

ABSTRACT -

Three models of behavioral intention, Ajzen and Fishbein's (1980) theory of reasoned action (TORA), Ajzen's (1985) theory of planned behavior (TOPB), and the Miniard and Cohen (1983) model (MCM) were compared in terms of predictive ability and their ability to effectively separate normative form personal influence. For predicting gift giving intention, TOPB performed better than did TORA and MCM. The results also suggest that MCM exhibited greater ability to separate its components than did TORA or TOPB. For predicting gift giving behavior, results indicate that intention represents the direct antecedent to behavior over that of the perceived behavioral control component of TOPB.

The theory of reasoned action (Ajzen & Fishbein, 1980) has been widely used across the social sciences. Though results support the model's predictive ability (see Farley, Lehman, & Ryan, 1981 and Sheppard, Hartwick, & Warshaw, 1988 for meta-analytic reviews), questions remain regarding one of its' boundary conditions, behavioral control, and its' ability to reflect the separate effects of attitudinal versus normative influence for explaining intentions. Recently, alternatives to the theory of reasoned action (TORA) have emerged. The theory of planned behavior (TOPB) is an extension of TORA that includes perceived behavioral control as a variable for predicting intentions and behavior (Ajzen, 1985), and the Miniard and Cohen (1983) model (MCM) assesses the separate contribution of personal and normative influence for the prediction and explanation of intentions. Though these two models show promise, TOPB has yet to be examined across a wide domain of behavior, and few published tests of MCM exist (e.g., Bearden & Rose, 1990; Miniard & Cohen, 1983). Furthermore, the three models (TORA, TOPB, and MCM) have yet to be compared across the same behavioral domain.

The primary goals of this article are to 1) compare TORA, TOPB, and MCM in terms of their ability to predict BI, and 2) compare the models with regard to their ability to separate normative from personal (attitudinal) influence for explaining BI. As a secondary goal, this study will also examine if the perceived behavioral control component of TOPB enhances behavioral prediction. As such, the study is congruent with the need to examine models where variables related to control may impact intention prediction and behavioral achievement (Ajzen & Madden, 1986; Sheppard et al., 1988), and the call to assess the contribution of normative and personal (attitudinal) influence for explaining behavioral intention across models (Miniard & Cohen, 1983). Also, since the behavioral domain of this research is gift giving, this study examines the personal vs. interpersonal motivations of gift giving. That is, it assesses the relative importance of personal attitudes and the perceived influence of relevant others (i.e., the recipient) for the prediction of gift giving - an issue receiving only limited empirical attention in the consumer behavior literature.

THREE BEHAVIORAL INTENTION MODELS

The Theory of Reasoned Action (TORA)

TORA posits that behavioral intention (BI) is the direct antecedent of behavior. BI, in turn, is determined by an individual's attitude toward performing the behavior (Aact) and the individual's perception of what relevant others think of the behavior, i.e., subjective norm (SN). The central equations of the theory are as follows:

Behavior _~ BI(w₁) (1)

BI = Aact(w₁) + SN(w₂) (2)

where w_is are empirically determined regression weights. BI is expected to accurately predict behavior if three boundary conditions hold: 1) the intention and behavior measures correspond in terms of specificity of target, context, action, and time frame; 2) intention does not change in the interval between BI and B assessment; and 3) the behavior in question is under the actor's volitional control, i.e., the actor can decide at will to perform or not perform the behavior (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975). Within these boundary conditions, support for the validity of TORA to predict BI and behavior is extensive (Ajzen & Fishbein, 1980; Farley et al., 1981; Sheppard et al., 1988). However, when boundary conditions are not met, explained variance estimates in BI and B are attenuated (Ajzen & Fishbein, 1980; Sheppard et al., 1988). The present study examines one of the boundary conditions, volitional control.

Another important function of BI models involves the degree to which a model accurately reflects the separate contribution of its' components to explain intention. In this regard, TORA has been questioned, as it has often been the case that Aact and/or SN correlate significantly with BI, but receive nonsignificant regression weights (Miniard & Cohen, 1979, 1983). This has largely been attributed to the view that Aact partially taps normative influence and SN partially taps attitudinal (personal) influence, which results in a high correlation between these two components. Thus, even though SN may be posited as a determinant of BI (but less so than Aact), SN may be nonsignificant in a regression equation due to its' overlap with Aact, reducing TORA's overall level of diagnostic utility (Miniard & Cohen, 1979, 1983). This study investigates this issue by comparing the normative-personal (attitudinal) overlap of TORA and TOPB to MCM.

The Theory of Planned Behavior (TOPB)

TOPB is an extension of TORA that incorporates nonvolitional elements for predicting behavior. As in TORA, the key antecedent to behavior prediction is BI and the relationships among Aact, SN and BI are identical for the two models. By adding a control variable though, TOPB expands the boundary conditions of TORA to goal-directed behavior - behavior not completely under an actor's volition. It has been suggested that even the most mundane of behaviors are sometimes subject to the influence of factors beyond one's control (Ajzen & Madden, 1986; Sarver, 1983). From this perspective, TOPB posits that most intended behaviors are those whose attainment is subject to some degree of uncertainty, and the chance of success a person will have of performing a behavior not only relies on intention, but factors that may interfere with behavioral control (e.g., resources such as money, time, opportunity and the cooperation of others). To enhance the prediction of intention and behavior, TOPB proposes measuring perceived behavioral control (PBC) - the person's belief as to how easy or difficult performing a behavior will be (Ajzen, 1985). Support for PBC as a predictor of BI and behavior has been demonstrated (Ajzen & Madden, 1986; Schifter & Ajzen, 1985).

As with previous studies, two versions of TOPB will be tested. The first version assumes that PBC is a predictor of BI, and that BI is the antecedent of behavior. The central equations of the first version are as follows:

Behavior ~ BI(w₁) (1)

BI = Aact(w1) + SN(w2) + PBC(w3) (2)

The second version considers the possibility of a direct effect from PBC to behavior as well as an effect via BI:

Behavior ~ BI(w₁) + PBC(w₂) (1)

BI = Aact(w1) + SN(w2) + PBC(w3)(2)

An effect from PBC to B is expected if the behavior of study is not under the individual's complete volitional control. This study, then, explores the possibility of PBC as an additional predictor of behavior and examines the volitional property of gift-giving.

The Miniard and Cohen Model (MCM)

The Miniard and Cohen (1983) model suggests that informational influence should be reflected only in one's personal attitudes and should be unrelated to normative beliefs about behavior. Their conceptualization is based upon earlier work in social psychology (Deutsch & Gerard, 1955; Kelman, 1961) and states that the opinions of others often serve as an important source of information about one's environment. The acceptance of the information depends on the source's credibility, and behavior based upon this information is independent of its' visibility or knowledge to a referent. Conversely, the normative aspects of MCM are restricted solely to a referent's normative power. Behaviors motivated by normative reasons are directly linked to an individual's desire to attain a reward or avoid some sanction from referents to whom the behavior would be known or visible. Miniard and Cohen (1983) indicate that these two influences are conceptually and empirically distinct in their model.

In essence, MCM was developed to reflect the separate effects of normative and personal influence for explaining intention, and to predict BI. The central equations of MCM are as follows.

Behavior ~ BI(w₁) (1)

BI = IPCPE(w1) + NPCNE(w2) (2)

IPCPE is the global evaluation of behavior based solely on personal reasons and NPCNE is the global evaluation of behavior based solely on normative reasons. Behavior, BI and w_is are the same in MCM as they are in TORA. Tests of MCM have shown a high level of validity for BI prediction. In fact, it was shown that MCM predicted BI as well as TORA (Miniard & Cohen, 1983). Studies also demonstrated MCM's ability to register the relative influence of personal and normative variables for explaining BI (Bearden & Rose, 1990; Miniard & Cohen, 1983). These results support MCM's predictive and diagnostic validity.

The preceding review of the three models suggests that TORA and MCM should exhibit comparable validity for the prediction of BI. Thus, explained variance estimates for TORA and MCM should be relatively equal for BI prediction. The addition of PBC though, should result in greater predictive ability, and it is therefore expected that TOPB will explain more variance in BI than TORA or MCM for the same behavioral domain. In terms of effectively separating its' components, the preceding review suggests that MCM will more accurately register the separate effects of its components than will TORA or TOPB. Thus, the IPCPE-NPCNE correlation of MCM should be significantly lower than the Aact-SN correlation of TORA and TOPB.

BEHAVIORAL DOMAIN

Gift giving was used as the behavior of study for a number of reasons. What follows is a brief review of the gift giving literature and a rationale for using gift giving as the behavioral domain of the present study.

Gift giving has been studied from several perspectives including the functions it serves (Belk, 1979), from an interactive paradigm (Banks, 1979), and from an anthropological viewpoint (Sherry, 1983). Belk (1979) posits that gift giving serves the four functions of communication, social exchange, economic exchange and socialization, while the interactive paradigm states that gift giving is a four stage process including a purchase stage, an interaction-exchange stage, a consumption stage and a communication-feedback stage (Banks, 1979). In a thorough review, Sherry (1983) has identified three dimensions of gift giving (i.e., social, personal and economic) that encapsulate the functions identified by Belk (1979) and the stages posited by Banks (1979). These dimensions are briefly discussed below.

From a social dimension, research suggests that gift giving is a form of social obligation or political maneuvering where a gift is given because it is expected or it will yield some favorable normative result (Schieffelin, 1980). Gift giving also serves to define the closeness of a relationship (Banks, 1979; Belk, 1979), and the role expectations in a relationship (Csikszentmihalyi & Rochberg-Halton, 1981). The major premise here is that the more intimate the relationship, the greater the gift in terms of monetary and time expenditures (Banks, 1979). It is also widely felt that the intimacy of a relation moderates gift type and number (Belk, 1979), and that special occasions (Christmas, weddings, birthdays) result in the giving of more expensive gifts. From a personal perspective, gift giving has been found to be a reflection of one's self-concept (Banks, 1979). For example, research suggests that the ideal self-concept of the giver may be more strongly related to gift choice than either the giver's self-concept or the perceptions of the gift by the recipient (Belk, 1979). From an economic dimension (i.e, the conferring of material benefit on a recipient) research shows that individuals attempt to maximize the equality of gift exchange, and that gift purchasers are more likely to begin search with an a priori specified price range than those buying the same items for personal use.

The above literature has offered great insight with regard to the functions, stages and dimensions of gift giving. However, gift giving has yet to be interpreted satisfactorily by social scientists and little attempt has been made to examine its structural and motivational components (Belk, 1979; Lutz, 1979; Sherry, 1983). Many scholars have encouraged research focusing on explanatory constructs as a means of examining motivational components, and it has been suggested that expectancy value models should capture the underlying influences that determine gift giving (Lutz, 1979; Sherry, 1983). Behavioral intention models provide the structure needed for examining some of these influences. For example, personal influences toward gift giving should be captured by the Aact component of TORA and the IPCPE component of MCM. Similarly, the influence of relevant others toward the purchase of various gifts should be captured by the SN and NPCNE measures of the two models. Possible control factors affecting gift purchase, such as the availability of the gift or the cost of the gift, should be reflected in the PBC component of TOPB. In addition to capturing these influences, BI models should allow for the estimation of the importance of each influence type. From this viewpoint, gift giving represents a behavior amenable to the goals of this study in that BI models can examine the motivational influences behind gift giving.

METHOD

Elicitation procedures

Valentine's Day was chosen as the gift giving occasion of study for several reasons. First, it is viewed as an obligatory gift giving occasion (Belk, 1979) where the number and types of gifts are determined traditionally. Since it was important that a manageable number of gift types was considered, the limited types of gifts given on Valentine's Day made it suitable to the study. Second, the literature has called for a better understanding of gift giving by studying specific gift giving occasions, and Valentine's Day provides such an occasion (Sherry, 1983). Lastly, students represent a primary target of Valentine's Day gifts.

To ensure that a proper set of gift types would be considered and that possible control factors affecting the giving of each specific gift type included, elicitation procedures were conducted (Ajzen & Fishbein, 1980; Ajzen & Madden, 1986). Two female student samples were used; one for the gifts types (n=28) and one for the control factors (n=29). The gifts women most frequently and almost exclusively listed as giving their boyfriends for Valentine's day were clothing, flowers, dinner, greeting cards, and candy. After determining the relevant gift types, the elicitation procedure for control factors was conducted. Subjects frequently mentioned lack of money for clothing, dinner and flowers. Lack of time was also mentioned for clothing. The other gifts consistently elicited no response.

Subjects and Measures

Eighty-two female undergraduate students participated in the main study. Each subject was screened to determine that she was currently dating someone on a steady basis. In the first phase, one week before Valentine's day, all subjects received questionnaires identical in content with full instructions and practice questions familiarizing them with the scale format. Operational measures of the three models were completely counter-balanced within the questionnaire (across subjects) as was the order of presentation of each models' components. Scales were developed to reflect the components of each model and were all 7-point items. Subjects responded to all measures across all gifts and models.

TORA

Operationalization of TORA followed procedures set by Ajzen and Fishbein (1980). All items were scored on a -3 to +3 basis. Aact toward each gift was assessed via the average of three semantic differential items (good-bad, foolish-wise, beneficial-harmful) in response to the statement "The purchase of ____________ as a Valentine's day gift for my boyfriend would be. . .". Aact alpha estimates ranged from .84 to .92 across the gifts. SN was evaluated via a single item scale (approve-disapprove) in response to the statement "If I bought _________ as a Valentine's day gift for my boyfriend, most people who are important to me would. ". (An SN measure that was boyfriend specific was also collected. This measure produced results similar to those reported in Table 1.)

TOPB

With the exception of perceived behavioral control, TOPB was operationalized with the same measures as used for TORA. Two approaches to measure PBC were undertaken. First, for each behavioral alternative, two items were used following Ajzen and Madden's (1986) procedures. These items read "If I wanted to, I could easily buy ___________ as a Valentine's day gift for my boyfriend. . . (likely-unlikely) and "For me to buy _________ as a Valentine's day gift for my boyfriend is. . . (easy-difficult). In addition, for those alternatives where the elicitation procedures produced factors that might interfere with behavioral achievement (i.e., flowers, dinner, clothing), extra statements were developed. For example, "The lack of money might prevent me from taking my boyfriend out to dinner as a Valentine's day gift. . (strongly agree-strongly disagree). For each gift, PBC items were scored on a -3 to +3 basis, and then averaged to form an index. PBC alpha estimates ranged from .53 to .84.

MCM

The measures and procedures of Miniard and Cohen (1983) were used to operationalize MCM. The global constructs of IPC, PE, NPC and NE were measured by the following single item scales, respectively:

Suppose you were to buy __________ as a Valentine's day gift on the sole basis of personal considerations (e.g., your own private feelings about buying _________ as a Valentine's day gift). Given this, how favorable or unfavorable would you then feel about buying ___________ as a Valentine's day gift for your boyfriend? (extremely favorable-extremely unfavorable).

In making your decision concerning the buying of ___________ as a Valentine's day gift for your boyfriend, how much importance will you place on your on personal considerations (e.g., your own private feelings about buying ___________ as a Valentines day gift)? (absolutely no importance-the greatest importance)

Suppose you were to buy __________ as a Valentine's day gift on the sole basis of interpersonal considerations (e.g., how important others like your boyfriend might react to your purchase). Given this, how favorable or unfavorable would you then feel about buying ___________ as a Valentine's day gift for your boyfriend? (extremely favorable-extremely unfavorable).

In making your decision concerning the buying of ___________ as a Valentine's day gift for your boyfriend, how much importance will you place on interpersonal considerations (e.g., how others like your boyfriend might react to your purchase)? (absolutely no importance-the greatest importance)

The PE and NE measures were scored from +3 to -3 and the IPC and NPC measures were scored from 0 to 6. To obtain the overall IPCPE and NPCNE constructs, IPC was multiplied by PE, and NPC was multiplied by NE. All models used the same single item intention measure for each of the five gifts: "I intend to buy ________ as a Valentine's day gift for my boyfriend (likely-unlikely). Also, as with other tests of MCM, instructions explaining the differences between personal and normative influence were included for the MCM portion of the questionnaire.

In the second phase of the study, three days after Valentine's day, subjects responded to behavioral measures toward the gifts: "Which of the following did you purchase as a Valentine's day gift for your boyfriend? Please check the appropriate space(s)." Spaces checked were coded "1" and spaces not checked were coded "0". Twelve respondents either did not give a gift or gave a gift other than the five examined, and thus, were excluded from all analyses.

TABLE 1

BI PREDICTION

RESULTS

BI Prediction

Table 1 presents a comparison of the three models for BI prediction. For each gift, the correlation of each predictor variable with BI (r), the respective standardized regression coefficient (b), the multiple correlation (R), and R² are reported.

First, we will compare the predictive validity of TORA and TOPB. Since TOPB is an extension of TORA, the comparisons were made using hierarchical regression where Aact and SN are entered on the first step and PBC on the second step. The difference in explained variance from step one to step two serves as a comparison of the two models predictive utility (Ajzen & Madden, 1986). For BI-flowers, TORA produced an R² of .24, but when PBC was added to the model, the R² rose to .32 (F-change=8.50, p < .01). For BI-clothing, the R² for TORA was .26 and the R² for TOPB was .29, but this difference was not significant (F-change=2.26, p < .12). For BI- dinner the R² for TORA was .28 and the R² for TOPB was .32 (F-change=6.19, p <. 01). Lastly, TOPB and TORA yielded identical results for the prediction of BI-candy and BI-card with R²s of .37 and .45, respectively (F-changes=.00 and .65, respectively, ns).

Overall, it would seem that only modest support for TOPB as a better predictor of BI than TORA exists. However, since we looked at multiple dependent variables in comparing the two models, we also calculated an overall effect across the five gifts. To do this, Rosenthal and Rubin's (1986, p. 403) equation for combining nonindependent effects was applied to the data. Essentially, this equation calculates a t-test by combining statistics (such as ts, zs or Fs) in a manner that depends upon the degree of intercorrelation among the dependent variables considered. The overall effect for comparing TOPB to TORA for predicting BI was t=5.80 (p < .01, df=66), and suggests that TOPB was better than TORA for BI prediction.

To compare the predictive validity of TORA and MCM, and TOPB and MCM, we used a procedure suggested by Tabachnick and Fidell (1983, pp. 114-115). In essence, this procedure calculates a z-test between multiple Rs and compares explained variance estimates for two sets of predictors on the same dependent variable. First, we will compare TORA to MCM. Recall that is was hypothesized that TORA and MCM should exhibit equal predictive validity. Across gifts, TORA's R was higher than MCM's R, but the differences were not statistically significant. For BI-flowers, the R for TORA was .48 and the R for MCM was .46 (z=.17, ns). For BI-clothing, the Rs for TORA and MCM were .52 and .36, respectively (z=1.33, ns). For BI-dinner, the R for TORA was .51 and the R for MCM was .28 (z=1.80, p < .10), and for BI-candy, the R for TORA was .61 and the R for MCM was .48 (z=1.69, p < .10). Lastly, TORA's R for BI-card was .67 and MCM's R for BI-card was .59 (z=1.23, ns).

To further compare TORA and MCM's predictive ability, we calculated an overall effect size again using Rosenthal and Rubin's (1986) procedure (t=2.15, p < .05, df=67). Thus, though there were no significant differences between TORA and MCM when comparing them on an individual dependent variable basis, the combined effect shows that TORA predicted BI better than did MCM.

It was also hypothesized that TOPB would exhibit greater predictive ability than MCM. For BI-flowers, the R for TOPB was .57 and the R for MCM was .46. Though directionally supported, this difference was not statistically significant (z=1.21, ns). A similar result was found for BI-card as TOPB's R of .67 was greater than MCM's R of .59, but the difference was not statistically significant (z=1.22, ns). However, for the other three gifts, TOPB explained significantly more variance in BI than did MCM. For BI-clothing, BI-dinner, and BI-candy TOPB's R estimates were .55, .57, and .61. These estimates were greater than corresponding estimates produced by MCM of .36, .28, and .48 (z=1.91, 2.74, and 1.70, p < .05, respectively). Furthermore, the overall effect size was t=3.03 (p < .05, df=67). These results suggests that TOPB was superior to MCM for BI prediction, as hypothesized.

TABLE 2

BEHAVIOR PREDICTION

Separation of Components

It was predicted that MCM would separate its normative and attitudinal (personal) components better than TORA/TOPB. To test this prediction, we calculated a z-test for differences between correlations within the same sample (Tabachnick & Fidell, 1983). This test allows for a significance test when both criteria and predictors are different. Thus, we tested if the IPCPE-NPCNE correlation was less than the Aact-SN correlation, and across the five gifts, the IPCPE-NPCNE correlation was significantly less than the Aact-SN correlation. For flowers, the IPCPE-NPCNE correlation was .28 and this was significantly lower than the Aact-SN correlation of .65 (z=2.98, p < .01). For clothing, the IPCPE-NPCNE correlation of .22 was less than the Aact-SN correlation of .54 (z=2.30, p < .01). The correlations between IPCPE-NPCNE and Aact-SN for dinner were .24 and .52, respectively (z=1.91, p < .05), and the correlations between IPCPE-NPCNE and Aact-SN for candy were .23 and .65 (z=3.52, p < .01). The IPCPE-NPCNE correlation for card of .43 was less than the corresponding Aact-SN correlation of .64 (z=1.72, p < .05). Lastly, the overall effect was also significant (t=3.36, p < .01, df=67).

Behavioral Prediction

Table 2 presents the results pertaining to the prediction of behavior. Again, the correlation between predictor and criterion (r), the standardized regression coefficient (b), R, and R² are offered. Each model (TORA, TOPB and MCM) considers BI as the predictor of behavior. TOPB, however, suggests that PBC will also be a significant predictor of behavior under conditions of imperfect control. Consistent with the earlier TOPB studies (Ajzen & Madden, 1986; Schifter & Ajzen, 1985), PBC was added to the regression equation after BI in hierarchical fashion. Across gifts, BI was the only significant predictor of behavior. The overall effect, via Rosenthal and Rubin's (1986) procedure, was also not significant (t=.85, ns). Thus, though correlated with behavior for three of the five gifts, PBC did not enhance behavioral prediction beyond that of BI.

DISCUSSION

This study compared the theory of reasoned action, the theory of planned behavior, and the Miniard and Cohen model. It was predicted that TORA and MCM would show no significant differences for BI prediction, and on an individual dependent variable basis, this was the case. However, when combining the effects from all five comparisons, TORA showed an improvement in predictive validity over that of MCM. It was also hypothesized that TOPB would be a better predictor of BI than TORA and MCM. As expected, TOPB was significantly better than TORA for predicting BI for three of the five gifts, and the overall effect showed a significant improvement in BI prediction of TOPB over TORA as well. TOPB also explained more variance in BI than did MCM for three of the five gifts, and this was further emphasized by the significance of the overall effect for comparing TOPB to MCM. These results, coupled with previous tests of TOPB, TORA, and MCM (Ajzen & Madden, 1986; Bearden and Rose, 1990; Schifter & Ajzen, 1985), suggest that PBC should be considered as an additional predictor of BI in future studies involving these models.

In terms of the ability to separate normative from personal (attitudinal) influence, MCM was superior to TORA/TOPB. The differences between TORA/TOPB and MCM are due to the relationships between attitudinal (personal) and normative influences as specified by each model. MCM posits that personal and normative influences are separate from one another, while TORA and TOPB state that normative and attitudinal influences may affect each other through inferential processing (Fishbein & Ajzen, 1975; Ryan, 1982). Thus, one would expect a stronger correlation between Aact and SN, as compared to the IPCPE-NPCNE correlation, and this was the case in the present study. Consistent with Miniard and Cohen though, MCM does more accurately separate the effects of personal (attitudinal) and normative variables for explaining BI than does TORA. This result highlights MCM as a valuable research tool since one of the purposes behind the development of the model was to isolate normative from personal influence. In the design of behavioral change strategies, identifying the separate effect of normative and personal influence is often required.

For behavior prediction, BI was the only direct antecedent. PBC had no predictive efficacy for the purchase of the gifts considered beyond that of intention. This suggests that Valentine's Day gift giving is best viewed as a volitional behavior, and consistent with Ajzen (1985), PBC will only be a direct antecedent of behavior for goal-directed or nonvolitional behaviors.

Gift giving

Though the primary purpose of this study was to compare three models in terms of predictive validity and the ability to separate their normative and personal (attitudinal) components, the results offer some interesting insights into gift giving. First, since Valentine's Day is classified as an obligatory gift giving occasion (Belk, 1979; Sherry, 1983) and this study looked at subjects in a well defined relationship, one might expect normative influence (SN, NPCNE) to play the dominant role for explaining and predicting BI. Across gifts though, Aact was more highly correlated to BI than SN, and IPCPE was more strongly correlated to BI than NPCNE for four of the five gifts. This result is not inconsistent with extant research where it was found that for certain gifts (i.e., clothing, jewelry) the attitudinal variable was more strongly correlated to BI than the normative variable (Warshaw, 1980). This result is also consistent with Belk's (1979) contention that, even for a behavior like gift giving where the consideration of the receiver seems paramount, personal considerations are of great influence. However, it is possible that financial constraints, which should have been captured by the PBC construct for TOPB, may have also been reflected in subjects' responses to Aact and IPCPE, contributed to the higher Aact-BI and IPCPE-BI correlations.

A second finding of interest for gift giving pertains to its' volitional properties. Though PBC was shown to be a significant predictor of BI beyond that of Aact and SN, after accounting for the impact of BI, PBC had no effect on actual gift giving behavior. This suggests the gifts examined in this study are relatively free of control problems.

A third finding pertains to the number of each gift type actually given. Clothing was the most popular gift purchased (about 32% of the subjects gave it to their boyfriends as a gift for Valentine's day). This finding is consistent with Belk's (1979) finding where clothing was the gift most often given to others. Candy and dinner were both given by 19% of the subjects, but only about 7% of the respondents gave flowers. What at first seemed a bit surprising was the fact that a greeting card was the second most popular gift given. Approximately 22% of the subjects gave it to their boyfriends as their only gift. Given the intimacy of the relationship and the view that the closer the relationship the dearer the cost (Belk, 1979), one might expect that a card alone would be one of the least given gifts. Clearly, a card is the least expensive of the five gifts in monetary terms. However, a greeting card can be a token gift or the result of thoughtful hours of search and may be used to express or reconfirm some special aspect of a relationship (Sherry, 1983). In the present case then, the giving of a greeting card only may be the result of considerable time investment, and the more intimate the relationship, the more time spent in gift search (Belk, 1979; Sherry, 1983). It still must be recognized that the financial constraints endemic to college students may have contributed to the choice of gift(s) given.

Future Research Involving BI Models

Future studies may want to consider several issues. First, TORA/TOPB typically uses multiple items to operationalize their constructs. In the present study, we used single items to measure SN and BI (i.e., potentially low reliability) which may have resulted in some variability in prediction. Future studies should employ multiple items for SN and BI measurement. Along this line are concerns related to the MCM measures. MCM also uses single items where low reliability may affect predictive results. Also, it is possible that the clearer separation of components for MCM may have benefitted from the MCM instructions explaining the differences between personal and normative influence. Future studies may want to devise multiple items for MCM constructs and test if MCM instructions affect the separation of personal from normative components. Lastly, classifying a behavior as volitional or goal-directed a priori should be helpful. By doing so, PBC measures that more accurately reflect all possible factors affecting behavioral achievement can be developed, enhancing behavioral prediction.

REFERENCES

Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In J. Kuhl & J. Beckman (eds.), Action-control: From cognition to behavior, Hillsdale, NJ: Erlbaum, 11-39.

Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice Hall.

Ajzen, I., & Madden, T.J. (1986). Prediction of goal-directed behavior: Attitudes, intentions and perceived behavioral control. Journal of Experimental Social Psychology, 22, 453-474.

Banks, S. (1979). Gift-giving: A review and an interactive paradigm. in Advances in Consumer Research, Vol. 6, W. Wilkie (ed.), Ann Arbor, MI: Association for Consumer Research, 319-324.

Bearden, W.O., & Rose, R.L. (1990). Attention to social comparison information: An individual difference factor affecting consumer conformity. Journal of Consumer Research, 16, 461-471.

Belk, R. (1979). Gift-Giving Behavior. Research in Marketing, Vol. 2, J. Sheth (ed.), Greenwich, CT: JAI Press, 95-126.

Csikszentmihalyi, M., & Rochberg-Halton, E. (1981). The meaning of things: Domestic symbols and the self. New York, NY: Cambridge University Press.

Deutsch, M., & Gerard, H.B. (1955). A study of normative and informational influence upon individual judgement. Journal of Abnormal Psychology, 51, 629-636.

Farley, J.U., Lehman, D.R., & Ryan, M.J. (1981). Generalizing from 'imperfect' replication. Journal of Business, 54, 597-610.

Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention, and behavior: An introduction to theory and research, Reading, MA: Addison-Wesley.

Kelman, H.C. (1961). Processes of opinion change. Public Opinion Quarterly, 25, 57-78.

Lutz, R.J. (1979). Consumer gift-giving: Opening the black box. in Advances in Consumer Research, Vol. 6, W. Wilkie (ed.), Ann Arbor MI: Association for Consumer Research, 329-331.

Miniard, P.W., & Cohen, J.B. (1979). Isolating attitudinal and normative influences in behavioral intention models. Journal of Marketing Research, 16, 102-110.

Miniard, P.W., & Cohen, J.B. (1981). An examination of the Fishbein-Ajzen behavioral intentions model's concepts and measures. Journal of Experimental Social Psychology, 17, 309-339.

Miniard, P.W., & Cohen, J.B. (1983). Modeling personal and normative influences on behavior. Journal of Consumer Research, 10, 169-180.

Ryan, M.J. (1982). Behavioral intention formation: The interdependency of attitudinal and social influence variables. Journal of Consumer Research, 9, 263-278.

Rosenthal, R., & Rubin, D.B. (1986). Meta-analytic procedures for combining studies with multiple effect sizes. Psychological Bulletin, 99, 400-406.

Sarver, V.T. (1983). Ajzen and Fishbein's "theory of reasoned action": A critical assessment. Journal for the Theory of Social Behavior, 13, 155-163.

Schieffelin E. (1980). Reciprocity and the construction of reality. Man, 15, 502-517.

Schifter, D.B., & Ajzen, I. (1985). "Intention, perceived behavioral control and weight loss: An application of the theory of planned behavior. Journal of Personality and Social Psychology, 19, 843-851.

Sheppard, B.H., Hartwick, J., & Warshaw, P.R. (1988). The theory of reasoned action: A meta-analysis of past research with recommendations for modifications and future research. Journal of Consumer Research, 15, 325-343.

Sherry, J.F. (1983). Gift giving in anthropological perspective. Journal of Consumer Research, 10, 157-168.

Tabachnick, B.G., & Fidell, L.S. (1983). Using multivariate statistics. New York, NY: Harper & Row.

Warshaw, P.R. (1980). Buying a gift: Product price moderation of social normative influences on gift purchase intention. Personality and Social Psychology Bulletin, 6, 143-148.

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