Comparing Leading Theoretical Models of Behavioral Predictions and Post-Behavior Evaluations
Note: Ini hanya sebuah catatan, jangan dijadikan rujukan. Silahkan merujuk ke sumber aslinya
Comparing Leading Theoretical Models of Behavioral Predictions and Post-Behavior Evaluations
Comparing Leading Theoretical Models of Behavioral Predictions and Post-Behavior Evaluations
Juliette Richetin and Marco Perugini
University of Milan-Bicocca, Milan, Italy
Iqbal Adjali and Robert Hurling
Unilever Corporate Research, Bedford, U.K.
This study aimed at comparing the predictive power of the Theory of Planned Behavior (TPB), the Model of Goal-Directed Behavior (MGB), and the Extended Model of Goal-Directed Behavior (EMGB) for observed and self-reported behaviors concerning consumer nondurables. More specifically, the three models were compared in terms of their predictive power for intention and for behavioral desire (only MGB and EMGB). Additionally, the validity of four different models for predicting post-behavior evaluations was examined.
Results showed that the EMGB is the most powerful in predicting both intention and behavioral desire. Moreover, results revealed that, as expected, all three models showed a better predictive power for SRB than for observed behavior. Finally, results demonstrated that post-behavior evaluations are both online and memory-based.
© 2008 Wiley Periodicals, Inc.
Psychology & Marketing, Vol. 25(12): 1131–1150 (December 2008)
Published online in Wiley InterScience (www.interscience.wiley.com)
© 2008 Wiley Periodicals, Inc. DOI: 10.1002/mar.20257
RICHETIN, PERUGINI, ADJALI, AND HURLING
Psychology & Marketing DOI: 10.1002/mar
The central aim of this paper is to investigate the relationship between prevolitional processes, behaviors, and post-behavior evaluations in a consumer domain. First, literature will be reviewed concerning three main attitudinal models, the Theory of Planned Behavior (TPB) (Ajzen & Madden, 1986; Ajzen, 1991), the Model of Goal-Directed Behavior (MGB) (Perugini & Bagozzi, 2001), and its extension, the Extended Model of Goal-Directed Behavior (EMGB) (Perugini & Conner, 2000; Perugini & Bagozzi, 2004a, 2004b). The models will then be compared in the study in terms of their fit with the data and their predictive power. Second, the focus will be on what happens after behavior (post-behavior evaluations). Different theoretical perspectives will be reviewed concerning the role of attitudinal, memory-based, versus experiential, online-based factors in influencing relevant post-behavior evaluations. Subsequent tests will seek to determine which of these perspectives can best accommodate the empirical results. In a related manner, the impact of past behavior on post-behavior evaluations will be also considered.
TPB, MGB, AND EMGB
Among the theoretical contributions aimed at modeling the influence of attitudes on behavior, the Theory of Reasoned Action (TRA) (Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975) and its extension, the Theory of Planned Behavior (TPB) (Ajzen & Madden, 1986; Ajzen, 1988, 1991, 2002a), are the most known and widely adopted. According to the TPB, people act in accordance with their intentions and perceptions of control over the behavior, while intentions, in turn, are influenced by attitudes toward the behavior, subjective norms, and perceived behavioral control. The predictive power of the TPB has been shown for numerous behaviors. For example, two meta-analyses (Armitage & Conner, 2001; Godin & Kok, 1996) of 185 and 76 studies, respectively, found an average of 39% and 41% of variance explained in intention and 29% and 34% of variance in behavior.
However, the sufficiency of TPB and its predecessor TRA has often been questioned. For example, the inclusion of constructs such as self-identity (Sparks & Shepherd, 1992; Armitage & Conner, 1999a, 1999b); personal, descriptive, or moral norms (Beck & Ajzen, 1991; Rivis & Sheeran, 2004; Harland, Staats, & Wilke, 1999; Trafimow & Finlay, 1996); personality traits (Courneya, Bobick, & Schinke, 1999); level of trying (Mathur, 1998); anticipated regret (Sheeran & Orbell, 1999); or past behavior (Bagozzi, 1981) have been shown to improve the prediction of intention or behavior.
Starting from the observation of a certain lack of theoretical sufficiency of the TPB in terms of modeling the pre-volitional processes, Perugini and colleagues (Perugini & Bagozzi, 2001, 2004a, 2004b; Perugini & Conner, 2000) proposed the MGB and its extension, the EMGB. The main aim was to expand and deepen the TPB by incorporating constructs from three new theoretical areas (affective, motivational, and automatic processes) and by hypothesizing a different theoretical flow. In the MGB, intention to perform a behavior is primarily motivated by the desire to perform the behavior, and this behavioral desire is assumed to reflect the effects of attitude, subjective norms, perceived behavioral control, and anticipated emotions, and to mediate their influence on intention. In the EMGB, goal desire is added as a new construct. This inclusion is based on the assumption that the influence of a desire to achieve a certain goal will influence the desire to perform a certain behavior that is subjectively felt to be instrumental for goal attainment. As a result, the behavioral desire will be the most proximal determinant of the intention to perform the behavior in question, and goal desire will have an indirect effect on intention through behavioral desire (Perugini & Bagozzi, 2004a). This inclusion mainly reflects the distinction between means and ends in decision making, whereby behaviors are considered explicitly as functional to achieve a certain underlying goal that is subjectively desired (Bagozzi, 1993). In the TPB, the implicit assumption is that whatever the distal determinants (e.g., goal desire), they are fully mediated by the proximal determinants and therefore do not add to the prediction of intention and behavior (cf. Ajzen, 1991). Both the MGB and the EMGB also include frequency and recency of past behavior so as to the incorporate the influence of automatic and habitual aspects in decision making not reflected by the variables of the TPB.1 The use of the MGB and the EMGB depends on the kind of behaviors and situations investigated (for considerations on using one rather than the other, see Perugini & Bagozzi, 2004b).
A number of studies have shown the predictive power of the MGB and the EMGB (e.g., Dijst, Farag, & Schwanen, 2008; Leone, Perugini, & Ercolani, 2004; Perugini & Bagozzi, 2001; Perugini & Conner, 2000; Taylor, 2007; Taylor, Bagozzi, & Gaither, 2001) for different behaviors such as weight control, studying, and traveling. Moreover, Perugini and Bagozzi (2004b) compared empirically the predictive power of the TPB, MGB, and EMGB. In a first set of seven studies, they found that the TPB accounted for 32% of variance for intention and 20% for behavior, whereas the MGB accounted for 58% of variance in behavioral desire, 58% in intention, and 26% in behavior. In a second set of four studies, they showed that the TPB, the MGB, and the EMGB accounted for 32%, 78%, and 79% of variance in intention, respectively. The MGB and EMGB accounted for 63% and 71% of variance in behavioral desire. It should be noted that the improvement of the MGB and EMGB is very strong for intention but not for behavior. This reflects the main focus of the MGB and EMGB on modeling pre- rather than post-volitional processes.
Self-Reported Behavior and Observed Behavior
Examining the TRA in relation to tax evasion, Hessing, Ellfers, and Weigel (1988) found that attitudes and subjective norms significantly correlated with the subjective but not with the objective behavioral measure. Consistent with these findings, some research suggested that behavioral self-reports were less valid compared to more objective behavioral measures (e.g., Armitage & Conner, 1999a, 1999b; Norwich & Rovoli, 1993). However, in their meta-analysis, Armitage and Conner (2001) found results that attenuate this assertion. They found that, even if the TPB accounted for a significant 11% more of the variance in selfreported behavior than did observed behavioral measures, the variance predicted in observed behavior was still substantial (20%).
Therefore, the results suggest that self-reported behavior is a reasonable proxy for objective behavior. However, while this might be true for aggregated repeated behavior, it is less so for incidental occasional behavior. The determinants of occasional behavior have been much less investigated within the tradition of the attitudinal model above, and there is a relative paucity of experimental studies investigating actual incidental behavior in laboratory (and therefore controlled) settings (e.g., Albarracín & Wyer, 2001). Given the distinction between incidental, occasional, and deliberative repeated behaviors (e.g., Wilson, Lindsey, & Schooler, 2000), it is not obvious and largely unexplored whether the attitudinal constructs gauged by models such as the TPB, MGB, and EMGB can be used to predict also an incidental, occasional behavior.
The TPB, MGB, and EMGB discussed above constitute three models to represent the deliberative decision-making processes underlying the execution of behavior and therefore to predict it. The models will be applied to a consumer nondurables domain, namely fizzy soft drinks. By defining and measuring the attitudinal constructs at the level of drinkingfizzy soft drinks (rather than of fizzy soft drinks), it was possible to test their predictive power both for repeated aggregated and incidental occasional drinking behaviors, as will be described later on.
Most studies in this attitudinal tradition typically stop at behavior as the final chain of events to be understood and predicted. However, equally interesting is what might determine post-behavior evaluations. The literature in the attitudinal field does not allow a clear-cut answer on this issue. However, theoretical elaborations, especially in the field of memory and judgment, but also in the attitude literature, can be used to build some meaningful hypotheses. Moreover, research in the marketing field has focused on issues concerning customer satisfaction or dissatisfaction that share several similarities to post-behavior evaluation (e.g., Oliver, 1993; Mooradian & Olver, 1997). For example, Mattila (2003) studied the role of memory in satisfaction judgments and showed that in a repurchase situation, consumers evaluated a product in a memory-based manner unless they were faced with a surprise event that led them to adjust their judgment online. In fact, according to the attitude literature, it is possible to articulate three main hypotheses, one holding that evaluations are only memorybased, another holding that evaluations are only online, and the last holding an additive pattern in which evaluations are both online and memory-based.
The online versus memory-based distinction has been examined especially in the field of judgment tasks. In fact, it appears that depending on details of the task, the judgment can be either memory-based or online (for a review, see Hastie & Park, 1986). On one side, research on availability or accessibility effects (e.g., Tversky & Kahneman, 1973) supports the traditional view, according to which judgment is memory-based (Allport, 1935): Individuals rely on the retrieval of concrete evidence from long-term memory in order to produce a judgment. Additionally, some research demonstrating that people’s stored evaluations are activated automatically and guide people’s interpretation of their environments (Houston & Fazio, 1989; Smith, Fazio, & Cejka, 1996) support the hypothesis of memory-based judgment. Another relevant issue concerning post-behavior evaluations, in line with this perspective, is to what extent repeatedly performing a certain behavior has an influence on subsequent evaluations. In fact, some research has shown the influence of behavior on attitudes (for a review, see Olson & Stone, 2005), which could be explained by different mechanisms such as cognitive dissonance, biased scanning, self-perception, and the use of heuristics (for a discussion, see Albarracín & Wyer, 2000). Regardless of these specific mechanisms, a straightforward reasoning is that by elementary principles of cognitive consistency, people tend to like what they freely decide to do repeatedly (Abelson et al., 1968). That is, the repeated voluntary execution of a behavior is a clear indicator of personal preference. Then, in turn, it should have a substantial influence on the evaluation, which is an expression of personal preferences, concerning that same behavior (see also the concept of cognitive inertia, Mattila, 2003).
On the other side, research on impression formation shows that information gathered directly from the object in the environment constitutes the basis of the judgment toward the object itself. This type of judgment is generally called online judgment. For example, spontaneous trait inference research shows that online evaluative processing of other people is the norm (Uleman, Newman, & Moskowitz, 1996). In line with this approach but more central to the attitude literature, the constructionist view suggests that people construct attitudes on the basis of information that happens to be accessible at the given point in time (for a review, see Wilson & Hodges, 1992). Attitudes are viewed as evaluative judgments that are constructed at the moment, as the current state of activation of a connectionist system, rather than as evaluation stored in memory (Smith, 1996). In line with this perspective, behaviors can have consequences after their execution. In particular, there are a series of important post-behavior evaluations that can be affected by action execution (e.g., Hurling & Martin, 2005; Hurling & Shepherd, 2003).
However, the two processes (memory-based versus online) are not necessarily mutually exclusive. For example, according to the “anchoring-and-adjustment model” (Lopes, 1982), a judgment is made by retrieving constructs in memory and by subsequently adjusting it depending on the context of the judgment and the motivation. Therefore, a judgment can be both memory-based and online.
This perspective is congruent with dual-attitude models suggesting that an attitude is constructed online with the elements that are stored in memory accessible at the point in time (e.g., Wilson, Lindsey, & Schooler, 2000; Petty et al., 2000).
On the basis of this literature, four different possibilities can be formulated about post-behavior evaluations of a product as influenced by incidental consumption. If one considers that judgments are only memory-based, the evaluation should be predicted only by elements concerning the product that are stored in memory. In this sense, any influence of the just-executed behavior would not be expected because it would at most activate the attitude stored in memory (Fazio, 1990). Within this perspective, two different possibilities can be elaborated.
First, if the attitude toward the behavior does not have a direct effect on postbehavior evaluations but has an effect mediated by the repeated execution of behavior, the evaluation should be predicted directly by the repeated behavior, which should mediate fully the influence of the attitude toward the behavior. Second, if the attitude toward the behavior has a direct effect on post-behavior evaluations, both the attitude toward the behavior and the repeated behavior should predict the evaluations. In both cases, the underlying assumption is that the more people consume a certain product, the more they evaluate it positively. On the other side, if one considers that judgments are only onlinebased, the evaluation should be predicted only by the context of the evaluation, that is, by the behavioral experience people undergo (e.g., drink fizzy soft drinks) just before evaluating the object. Finally, if one considers that judgments are both memory- and online-based, the evaluation should be predicted by both the context of the evaluation and elements that are stored in memory. Therefore, it should be reflected in an additive pattern of prediction.
Aims of This Contribution
This contribution has two main aims. The first aim is to compare the TPB, MGB, and EMGB for the prediction of self-reported and observed behaviors concerning drinking fizzy soft drinks. More specifically, given that the three models do not differ in terms of predictors of behaviors, the focus will be on comparing them in terms of their predictive power of the pre-volitional processes. The second aim is to test whether post-behavior evaluations are both online - and memory-based (additive pattern).
Participants and Procedure
Seventy-five women and thirty-three men (Mage= 23.8, SD =5.97) from an English university participated in a three-session study at one-week intervals.
They were paid £12 for their participation. In the first session, each participant sat individually in a cubicle at a table with a desktop computer and completed measures of their attitude toward drinking fizzy soft drinks, subjective norm, perceived behavioral control, positive and negative anticipated emotions, behavioral desire, intention, and goal desire. 2
The study included some additional measures that will not be considered because they are not relevant for this paper. The measures were administered via computer in the order above. In the second session, participants completed a self-reported last-week consumption of fizzy soft drinks. The last session was presented as a taste test. Participants were asked not to drink in the previous three hours before their session started.
They evaluated a can of fizzy soft drink with instructions to drink as much as they needed in order to evaluate it. First, each participant was given a small quantity of water (50 ml) so as to create a similar baseline condition for everybody and thus minimize external effects on the evaluations. The choice of the soft drink for the taste test was based on the responses given by each participant to the average weekly consumption questionnaire filled during the first session. For each participant, the fizzy soft drink that they consumed most frequently was chosen. If a participant did not report any consumption, the choice was random. After the taste test, the participants completed for the second time a last-week self-reported fizzy soft drinks consumption grid. Finally, the participants were debriefed, thanked for their participation, and paid. The data from three participants were discarded because they did not attend the last session, leaving a total of 105.
Attitude. Participants were presented with the stem “I think that for me to drink fizzy soft drinks is” followed by six bipolar scales (bad–good, unpleasant–pleasant, negative–positive, unenjoyable–enjoyable, unhealthy–healthy, unsatisfying–satisfying) on a 7-step answer scale ranging from 1 (not at all) to 7 (extremely). The reliability was good (α = 0.89).
Subjective Norms. Subjective norms were assessed by the following three items: (1) “People who are important to me think I should drink fizzy soft drinks,” which was rated on a 7-point scale ranging from 1 (unlikely) to 7 (likely); (2) “People who are important to me would approve of my drinking fizzy soft drinks,” which was rated on a 7-point scale ranging from 1 (not at all) to 7 (completely); and (3) “People who are important to me would be very happy if I drink fizzy soft drinks,” which was rated on a 7-point scale from 1 (false) to 7 (true). The reliability was not very good (α = 0.57). The first item was the most problematic and was eliminated.3 The reliability of the new composite including only the last two items was slightly better (α = 0.62, r = 0.45).
Perceived Behavioral Control. Perceived behavioral control was assessed with the following three items: (1) “How much control do you have over drinking fizzy soft drinks,” which was rated on a 7-point scale from 1 (no control) to 7 (complete control); (2) “If I wanted to, it would be easy for me to drink fizzy soft drinks,” which was rated on a 7-point-scale from 1 (highly unlikely) to 7 (highly likely); and (3) “For me to drink fizzy soft drinks is:” followed by a 7-point scale ranging from 1 (difficult) to 7 (easy). The reliability was not satisfactory at all (α = 0.35). The first item was unrelated to the others and was eliminated. In fact, the reliability of the new composite including the last two items was much better (α = 0.70,r = 0.55).
Anticipated Emotions. Anticipated emotions were measured with 10 items on a 7-point scale ranging from 1 (not at all) to 7 (extremely). Participants indicated how (delighted, disappointed, embarrassed, gratified, guilty, happy, pleased, regretful, satisfied, worried) they would feel should they drink fizzy soft drinks.
Half of the adjectives referred to negative and half to positive anticipated emotions. 4
The reliabilities were good (α = 0.87 and 0.90, respectively).
Behavioral Desire. Behavioral desire was measured by the following three items: (1) “How strongly would you characterize your desire to drink fizzy soft drinks?” which was rated on a 6-point scale from 1 (no desire) to 6 (very strong desire); (2) “I desire to drink fizzy soft drinks,” which was rated on a 7-point scale ranging from 1 (unlikely) to 7 (likely); and (3) “Drinking fizzy soft drinks is something that I desire to do,” which was rated on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). The reliability was good (α = 0.94).
Intention. Intention was assessed by the three following items: (1) “I will drink fizzy soft drinks,” which was rated on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree); (2) “How likely is that you will drink fizzy soft drinks?” which was rated on a 7-point scale ranging from 1 (highly unlikely) to 7 (highly likely); and (3) “I intend to drink fizzy soft drinks,” which was rated on a 7-point scale ranging from 1 (unlikely) to 7 (likely). The reliability was good (α = 0.95).
Goal Desire. Participants were first asked to write down a reason for drinking fizzy soft drinks. Then the desire toward this goal was measured with the three following items (reason Y was replaced by the reason written by the participants): (1) “How strongly would you characterize your desire to REASON Y?” which was rated on a 6-point scale ranging from 1 (no desire) to 6 (very strong desire); (2) “I desire to REASON Y,” which was rated on a 7-point scale ranging from 1 (unlikely) to 7 (likely); and (3) “The intensity of my desire for REASON Y can be described as,” which was rated on a 7-point scale ranging from 1 (nil) to 7 (extreme). The reliability was good (α = 0.82).
Behaviors. Behavior was measured in two ways. First, two self-reported behaviors were measured by asking participants their last-week consumption (in the last two sessions) of a series of fizzy soft drinks (e.g., Coke, Pepsi, Sprite, lemonade), expressed in units (e.g., a can of 330 ml equals one unit). Participants were asked to report how many in each category they drank during the last week.
Given the high correlation between the two measures (cf. Table 1), one single index was obtained by aggregating them, hereinafter named Self-Reported Behavior (SRB). Second, similar to Hofmann, Rauch, and Gawronski (2007) for sweets consumption, an incidental occasional behavior (amount drunk) was measured by weighing the amount of the drink each participant drank (i.e., the difference between the weight at the beginning and the weight after) during a taste test. The cans were weighed with a professional precision balance (Ohaus Adventurer SL) with a measurement error of 0.01 grams. To reduce measurement error, each can was kept at a similar fresh temperature in the refrigerator and individually weighed before and immediately after the taste test.
Post-Behavior Evaluations.The post-behavior evaluations were measured in the last session by asking participants to evaluate a can of fizzy soft drink immediately after they tasted it. First, participants indicated to what extent the fizzy soft drink was tasty, insipid, refreshing, unappetizing, and energizing on an 11-step answer scale ranging from 0 (not at all) to 10 (extremely).
Additionally, they gave an overall evaluation of the fizzy soft drink by answering to what extent they like the drink on an 11-step answer scale from 0 (not at all) to 10 (extremely). Given the substantial correlations among the responses, they were aggregated in a single index (overall evaluation, α = 0.72). Finally, participants indicated to what extent they would recommend the fizzy soft drink to a friend on an 11-step answer scale from 0 (not at all) to 10 (extremely).
Three hypotheses were elaborated concerning the predictive power of the TPB, MGB, and EMGB. First, it was hypothesized that the three models would better predict self-reported than incidental behavior (H1). This hypothesis is straightforward and follows from considering that explicit attitudinal processes best predict deliberative repeated behaviors (Wilson, Lindsey, & Schooler, 2000).
Second, it was hypothesized that the MGB and EMGB would show a better predictive power for intention than the TPB (H2). Finally, it was expected that the EMGB would show a better predictive power than the MGB for behavioral desire (H3). These two hypotheses directly reflect findings in previous studies, as reviewed in the introduction. Concerning the prediction of post-behavior evaluations, a model based on the anchoring and adjustment models was tested: The postbehavior evaluations of fizzy soft drinks (i.e., overall evaluation and recommend to a friend) would be both online- and memory-based and thus predicted by both consumption behavior and attitude toward drinking fizzy soft drinks (H4).
Tests of the TPB, MGB, and EMGB for Predicting Behaviors The TPB, the MGB, and the EMGB were formally tested for the two behaviors with structural equation models using Lisrel 8.7 (Jöreskog & Sörbom, 2003).
Partial disaggregation models were used in order to reduce the number of observed variables, which is particularly useful with smaller sample sizes to reduce the likelihood of computational problems and to obtain smaller standard errors (Bagozzi & Heatherton, 1994). Therefore, two random parcels were calculated for the explicit attitude, containing three items each, and two random parcels for both positive and negative anticipated emotions, containing two and three items each. For the other constructs, the actual number of observed variables was considered (2 for subjective norms, 2 for perceived behavioral control, 3 for goal desire, 3 for behavioral desire, 3 for intention, 2 for SRB, and 1 for amount drunk). The correlation matrix is reported in Table 1. A full structural equation model was therefore used in order to investigate the goodness of fit for the TPB, MGB, and EMGB for the prediction of selfreported behavior and observed behavior (amount drunk). Goodness of fit was ascertained by examining the chi-square, which should be nonsignificant.
In addition, the comparative fit index (CFI), the non-normed fit index (NNFI), and the root mean square error of approximation (RMSEA) were also used as indicators of goodness of fit. Values above 0.95 for both CFI and NNFI and below 0.06 for the RMSEA can be considered as satisfactory (Hu & Bentler, 1999). The predictive power of the model was tested by examining the part of variance explained (R2 ) in the criteria (i.e., intention, SRB, and amount drunk for the TPB; behavioral desire, intention, and behavior for MGB and EMGB). The standardized parameter estimates are reported in Figures 1, 2, and 3 for the TPB,
Table 1. Correlations among Constructs.
Figure 1.Parameters for the TPB.
Figure 2.Parameter Estimates for the MGB.
Figure 3.Parameter Estimates for the EMGB.
MGB, and EMGB, respectively (with the exception of the correlation between the two behaviors, the correlations among predictors are omitted for the sake of simplicity; values in italics are significant).
Goodness of Fit
The TPB obtained a good fit x 2 44 (N=105) = 41.35, p = 0.59; high relative fit index (CFI = 1.00, NNFI = 1.00); and a negligible mean error of approximation forparameter estimates (RMSEA = 0.00). The TPB predictors accounted for 56% of the variance in intention to drink fizzy soft drinks and for 26% and 7% of the variance in SRB and amount drunk, respectively. The MGB also achieved a good fit x 2 131 (N=105) = 151.07, p = 0.11 (CFI = 0.99, NNFI = 0.98, RMSEA = 0.04). The MGB predictors accounted for 69% of the variance in the behavioral desire to drink fizzy soft drinks, for 67% of the variance in intention to drink fizzy soft drinks, and for 26% and 6% of the variance in SRB and amount drunk, respectively. The fit of the EMGB was also good x 2 182 (N= 105) = 213.17, p= 0.06 (CFI = 0.98, NNFI = 0.98, RMSEA = 0.04). The EMGB predictors accounted for 72% of the variance for behavioral desire, for 67% of the variance for intention, and for 27% and 6% of the variance in SRB and amount drunk, respectively. Therefore, the results showed that all three models achieve a good fit and can be considered as empirically supported. They also confirmed H1, namely that they predict self-reported repeated better than incidental behavior.
Comparison of the Predictive Power of the Three Models
The comparison between the three models in terms of predictive power consisted of two levels of comparison. First, the comparison between the TPB, the MGB, and the EMGB resulted in testing the differences concerning the part of variance explained for intention for TPB and MGB or EMGB (the last two models do not differ at this level, so they can be considered as a single model in the comparison). However, considering the difference in layers between the TPB on one side and the MGB and the EMGB on the other side, the comparison cannot be reduced to a formal comparison of the R2. In fact, in the TPB, intention is supposedly predicted by attitude, subjective norms, and perceived behavioral control, whereas in the MGB or the EMGB, intention is predicted by behavioral desire and perceived behavioral control. Thus, only a qualitative comparison is feasible: The MGB and EMGB predicted intention (R2 = 0.67) better than the TPB did (R2 = 0.56), therefore supporting H2.
Second, the comparison of predictive power between the MGB and the EMGB consisted of the comparison of the part of variance explained for behavioral desire. In the MGB, behavioral desire is predicted by attitude, positive and negative anticipated emotions, subjective norms, and perceived behavioral control; in the EMGB, intention is a function of attitude, positive and negative anticipated emotions, subjective norms, perceived behavioral control, and goal desire.
Therefore, the difference between the two R2 s is equal to the squared semipartial correlation associated with goal desire, which shares the same null hypothesis of the regression coefficient of goal desire. Thus, if goal desire is a significant predictor, the two R2s are significantly different. Indeed, in the EMGB goal desire was a significant predictor of behavioral desire (g = 0.19). The EMGB was therefore significantly better than the MGB in terms of predictive power for behavioral desire, as was hypothesized (H3).
Tests for the Prediction of Post-Behavior Evaluations
To investigate whether the post-behavior evaluations of a can of fizzy soft drink during the taste test (i.e., overall evaluation and recommend to a friend) were both memory-based and online, the model in which attitude, SRB, and amount drunk predicted the two evaluations was formally tested using structural equation models. 5
As previously, two random parcels for attitude were considered. Additionally, two random parcels were calculated for the overall evaluation, containing three items each. For recommend to a friend, SRB, and amount, the actual number of observed variables was considered (1, 2, and 1, respectively). Predictive power was tested by examining the part of variance explained (R2) in the criteria (i.e., overall evaluation and recommend to a friend).
The model (cf. Figure 4) obtained a good fit x 212 (N=105) = 10.33, p= 0.59 (CFI = 1.00, NNFI = 1.01, RMSEA = 0.00). The predictors accounted for 49% of the variance in the overall evaluation and for 43% of the variance in recommend to a friend. The variance explained in SRB and amount was 14% and 9%, respectively. Attitude significantly predicted SRB and amount drunk (b= 0.37 and 0.30, respectively) and both overall evaluation and recommend to a friend (b=0.49 and 0.50, respectively). SRB was a significant predictor of recommend to a friend but not of overall evaluation (g = 0.23 and 0.13, respectively). Conversely, amount drunk significantly predicted overall evaluation but not recommend to a friend (g = 0.30 and 0.09, respectively). The parameters therefore did not support either a mediated memory-based model (i.e., there were significant direct unmediated influences of attitude on post-behavioral evaluations) or an unmediated one (i.e., the behaviors influenced the post-behavioral evaluations). An online-based model was not supported either, given that attitude predicted post-behavioral evaluations. On the contrary, the results strongly suggested that both memory and online processes were responsible for postbehavioral evaluations and therefore supported the additive pattern model (H4).
Figure 4. Parameter Estimates for the Additive Pattern Predicting Post-behavior Evaluations.
The first aim of this study was to compare the TPB, MGB, and EMGB in terms of predictive power for intention on the one hand and the MGB and the EMGB for behavioral desire on the other. The results confirmed the hypotheses: The MGB and EMGB showed better predictive power for intention than the TPB, and the EMGB showed better predictive power than the MGB for behavioral desire.
Finally, the results demonstrated, as hypothesized, a better predictive power for repeated self-reported behavior over an incidental occasional behavior such as the amount of soft drink drunk in a laboratory test. The second aim of this study was to test an additive pattern hypothesizing attitude, self-reported behavior, and amount drunk as predictors of post-behavior evaluations. Results showed that this additive pattern was supported.
Confirming previous studies, once again the MGB and the EMGB showed a better predictive power than the TPB for intention. The major reason that explains this superiority of the EMGB and MGB over the TPB is the inclusion of behavioral desire and, as a consequence, the distinction between intention and behavioral desire (Perugini & Bagozzi, 2004a). The role of desires cannot be underestimated and influences several aspects of the decision-making process (e.g., Dholakia, Gopinath, & Bagozzi, 2005; Dholakia et al., 2006). Moreover, the EMGB showed a stronger power in predicting behavioral desire than the MGB.
In other words, the inclusion of goal desire improved the prediction. Therefore, even if it might be tempting to assume that there is no need to have a strong goal behind the desire to drink fizzy soft drinks, nonetheless it has a statistically significant impact. Obviously, the contribution of goal desire might be greater for behaviors that are more problematic, such as exercising or studying, as the possession of an overarching strong goal can facilitate the desire to perform a behavior when to do so one has to overcome some obstacles. Yet the results reveal that not only does goal desire significantly add to the prediction of desiring to drink fizzy soft drinks, but that it has a more important role than more intuitively compelling and well-established constructs. Indeed, positive and negative anticipated emotions, perceived behavioral control, and subjective norms did not play a relevant role, whereas attitudes had the lion’s share in explaining desire, as would have been easy to predict.
Consistent with previous results concerning the TPB (e.g., Armitage & Conner, 1999a, 1999b; Norwich & Rovoli, 1993), the results also have shown that the three models have better predictive power for self-reported than for observed behavior. Armitage and Conner (2001) suggested that this difference in the variance explained in self-reported and observed behavior might be due to the distinction between subjective and objective measures. Indeed, the measures of the TPB, MGB, or EMGB constructs, as well as self-reported behaviors, are subjective, whereas the observed behavior such as amount drunk is objective. This difference might also reflect the fact that the subjective measure of behavior directly mapped onto the prior measure of intention, whereas the objective measure did not (Armitage & Conner, 2001). An alternative explanation could consist in considering that this difference reflects the distinction between automatic and controlled processes that lead to behavior. Indeed, an observed behavior performed contextually in a lab can be conceived as more incidental and more automatic than a self-reported behavior. The TPB, and the MGB and EMGB as well, adopt a more deliberative approach to behavior and do not fully integrate recent dual-process models (e.g., Gawronski & Bodenhausen, 2007; Strack & Deutsch, 2004; Wilson, Lindsey, & Schooler, 2000). These dual-process models consider incidental behavior as mainly influenced by automatic processes and deliberate behavior as mainly influenced by deliberate processes. This lack of focus on more automatic processes for predicting incidental behavior might explain the difference in variance explained for the two kinds of behavior.
Finally, the results demonstrated that post-behavior evaluations were predicted by both self-reported and observed consumptions of and attitude toward drinking fizzy soft drinks. Therefore, both the context of the evaluation and the elements stored in memory are predictors of the evaluations. In other words, contrary to research that posits that evaluations are exclusively memory-based or online, this study showed that the evaluations are both, as anchoring and adjustment models postulate. However, the results did not support a full additive pattern because amount drunk only predicted overall evaluation and SRB only predicted recommend to a friend. In other words, the amount one consumes on the spot strongly influences the subsequent evaluation of the product, whereas the amount one consumes regularly strongly influences to what extent people would recommend the product to others.
Some limitations of this study should be acknowledged. First, the first two items measuring intention are more measures of behavioral expectations than strictly of intentions. However, these behavioral expectations should be highly correlated with intentions and therefore the results should not be affected. Second, in light of some recent research, it could be argued that the EMGB could be expanded. Indeed, Bagozzi, Dholakia, and Basuroy (2003) showed that goal intention is a mediator of the relationship between goal desire and behavioral desire. In this present case, the inclusion of goal intention in the EMGB as a predictor of behavioral desire would probably have weakened or suppressed the observed effect of goal desire on behavioral desire. Finally, concerning the prediction of post-behavior evaluations, the attitude towards drinkingfizzy soft drinks has been used as a predictor of evaluations. It could be argued that a more precise measure would have involved an attitude toward fizzy soft drinks.
Therefore, it is possible that the results somehow may slightly underestimate the influence of stored attitudes on post-behavior evaluations. However, it is reasonable to expect that two such attitudinal measures should be very highly correlated, and it is quite unlikely that a different measure of attitude could have any substantial impact on the results, especially in terms of support for the additive pattern.
To conclude, this study showed that the TPB or TRA are not sufficient in modeling the pre-volitional processes that ultimately lead to behavior and that the distinction between behavioral desire and intention on one side and the consideration of goal desire on the other seem to be essential to improve the prediction of what people intend to do. Finally, this study supports the hypothesis that post-behavior evaluations are both online- and memory-based and contributes to a better understanding of the processes underlying post-consumption satisfaction, a variable that is known to be strongly linked to important phenomena such as customer loyalty and switching intentions (e.g., Anton, Camarero, & Carrero, 2007; Olsen, 2002).
Abelson, R. P., Aronson, E., McGuire, W. J., Newcomb, T. M., Rosenberg, M. J., & Tannenbaum, R. H. (1968). Theories of cognitive consistency: A sourcebook. Chicago, IL: Rand McNally.
Ajzen, I. (1988). Attitudes, personality, and behavior. Chicago, IL: Dorsey.
Ajzen, I. (1991). The theory of planned behavior. Organizational Behavior and Human Decision Processes, 50, 179–211.
Ajzen, I. (2002a). Perceived behavioral control, self-efficacy, locus of control, and the theory of planned behavior. Journal of Applied Social Psychology, 32, 665–683.
Ajzen, I. (2002b). Residual effects of past on later behavior: Habituation and reasoned action perspectives. Personality and Social Psychology Review, 6, 107–122.
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.
Albarracín, D., & Wyer, R. S., Jr. (2000). The cognitive impact of past behavior: Influences on beliefs, attitudes, and future behavioural decisions. Journal of Personality and Social Psychology, 79, 5–22.
Albarracín, D., & Wyer, R. S., Jr. (2001). Elaborative and nonelaborative processing of behavior-related communication. Personality and Social Psychology Bulletin, 27, 691–705.
Allport, G. W. (1935). Attitudes. In C. Murchison (Ed.), A handbook of social psychology (pp. 798–844). Worcester, MA: Clark University Press.
Anton, C., Camarero, C., & Carrero, M. (2007). The mediating effect of satisfaction on consumers’ switching intention. Psychology & Marketing, 24, 511–538.
Armitage, C. J., & Conner, M. (1999a). The theory of planned behaviour: Assessment of predictive validity and “perceived control.” British Journal of Social Psychology, 38, 35–54.
Armitage, C. J., & Conner, M. (1999b). Distinguishing perceptions of control from selfefficacy: Predicting consumption of low fat diet using the theory of planned behavior. Journal of Applied Social Psychology, 29, 72–90.
Armitage, C. J., & Conner, M. (2001). Efficacy of the theory of planned behaviour: A metaanalytic review. British Journal of Social Psychology, 40, 471–499.
Bagozzi, R. P. (1981). Attitudes, intentions and behavior: A test of some key hypothesis. Journal of Personality and Social Psychology, 41, 607–627.
Bagozzi, R. P. (1993). On the neglect of volition in consumer research: A critique and a proposal. Psychology & Marketing, 10, 215–237.
Bagozzi, R. P., & Heatherton, T. F. (1994). A general approach to representing multifaceted constructs personality constructs: Application to state self-esteem. Structural Equation Modeling, 1, 35–67.
Bagozzi, R. P., Dholakia, U.M., & Basuroy, S. (2003). How effortful decisions get enacted: The motivating role of decision processes, desires, and anticipated emotions. Journal of Behavioral Decision Making, 16, 273–295.
Bargh, J. A. (2006). What have we been priming all these years? On the development, mechanisms, and ecology of nonconscious social behavior. European Journal of Social Psychology, 36, 147–168.
Beck, L., & Ajzen, I. (1991). Predicting dishonest actions using the theory of planned behavior. Journal of Research in Personality, 25, 285–301.
Courneya, K. S., Bobick, T. M., & Schinke, R. J. (1999). Does the theory of planned behavior mediate the relationship between personality and exercise behavior? Basic and Applied Social Psychology, 1999, 21, 317–324.
Dholakia, U. M., Gopinath, M., & Bagozzi, R. P. (2005). The role of desires in sequential impulsive choices. Organizational Behavior and Human Decision Processes, 98, 179–194.
Dholakia, U. M., Gopinath, M., Bagozzi, R. P., & Nataraajan, R. (2006). The role of regulatory focus in the experience and self-control of desire for temptations. Journal of Consumer Psychology, 16, 163–175.
Dijst, M., Farag, S., & Schwanen, T. (2008). A comparative study of attitude theory and other theoretical models for understanding travel behavior. Environment and Planning A, 40, 831–847.
Fazio, R. H. (1990). Multiple processes by which attitudes guide behavior: The MODE model as an integrative framework. In M. P. Zanna (Ed.), Advances in experimental social Psychology, Vol. 23 (pp. 75–109). New York: Academic Press.
Fishbein, M., & Ajzen, I. (1975). Belief, attitude, intention, and behavior. Reading, MA: Addison-Wesley.
Gawronski, B., & Bodenhausen, G. V. (2007). Unraveling the processes underlying evaluation: Attitudes from the perspective of the APE model. Social Cognition, 25, 687–717.
Godin, G., & Kok, G. (1996). The theory of planned behavior: A review of its application to health-related behaviors. American Journal of Health Promotion, 11, 87–98.
Harland, P., Staats, H., & Wilke, H. A. M. (1999). Explaining proenvironmental intention and behavior by personal norms and the theory of planned behavior. Journal of Applied Social Psychology, 29, 2505–2528.
Hastie, R., & Park, B. (1986). The relationship between memory and judgment depends on whether the judgment task is memory-based or on-line. Psychological Review, 93, 258–268.
Hessing, D. J., Ellfers, H., & Weigel, R. H. (1988). Exploring the limits of self-reports and reasoned action: An investigation of the psychology of tax evasion behavior. Journal of Personality and Social Psychology, 54, 405–413.
Hofmann, W., Rauch, W., & Gawronski, B. (2007). And deplete us not into temptation: Automatic attitudes, dietary restraint, and self-regulatory resources as determinants of eating behavior. Journal of Experimental Social Psychology, 43, 497–504.
Houston, D. A., & Fazio, R. H. (1989). Biased processing as a function of attitude accessibility: Making objective judgments subjectively. Social Cognition, 7, 51–66.
Hu, L., & Bentler, P. M. (1999). Cut-off criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.
Hurling, R., & Martin, K. J. (2005). Perceived preservation format and food preference. International Journal of Consumer Studies, 29, 502–507.
Hurling, R., & Shepherd, R. (2003). Eating with your eyes: The effect of appearance on expectations of liking. Appetite, 41, 167–174.
Jöreskog, K. G., & Sörbom, D. (2003). LISREL 8 user’s reference guide. Chicago, IL: Scientific Software.
Leone, L., Perugini, M., & Ercolani, A.-P. (2004). Studying, practicing, and mastering: A test of the Model of Goal-Directed Behavior (MGB) in the software learning domain. Journal of Applied Social Psychology, 34, 1945–1973.
Lopes, L. L. (1982). Toward a procedural theory of judgment. Wisconsin Human Information Processing Program, Technical Report, 17, 1–49.
Mathur, A. (1998). Examining trying as a mediator and control as a moderator of intention–behavior relationship. Psychology & Marketing, 15, 241–259.
Mattila, A. S. (2003). The impact of cognitive inertia on postconsumption evaluation processes. Journal of the Academy of Marketing Science, 31, 287–294.
Mooradian, T. A., & Olver, J. M. (1997). “I can’t get no satisfaction”: The impact of personality and emotion on postpurchase processes. Psychology & Marketing, 14, 379–393.
Norwich, B., & Rovoli, I. (1993). Affective factors and learning behaviour in secondary school mathematics and English lessons for average and low attainers. British Journal of Educational Psychology, 63, 308–321.
Oliver, R. L. (1993). Cognitive, affective, and attribute bases of the satisfaction response. Journal of Consumer Research, 20, 418–430.
Olsen, S. (2002). Comparative evaluation and the relationship between quality, satisfaction and repurchase. Journal of the Academy of Marketing Science, 30, 240–249.
Olson, J. M., & Stone, J. (2005). The influence of behavior on attitudes. In D. Albarracin, B. T. Johnson, & M. P. Zanna (Eds.), The handbook of attitudes (pp. 223–271). Mahwah, NJ: Erlbaum.
Perugini, M., & Bagozzi, R. P. (2001). The role of desires and anticipated emotions in goal-directed behaviours: Broadening and deepening the theory of planned behaviour. British Journal of Social Psychology, 40, 79–98.
Perugini, M., & Bagozzi, R. P. (2004a). The distinction between desires and intentions. European of Journal of Social Psychology, 34, 69–84.
Perugini, M., & Bagozzi, R. P. (2004b). An alternative view of pre-volitional processes in decision making: Conceptual issues and empirical evidence. In G. Haddock & G. R. Maio (Eds.), Contemporary perspectives on the psychology of attitudes: The Cardiff symposium (pp. 169–201). Hove, UK: Psychology Press.
Perugini, M., & Conner, M. (2000). Predicting and understanding behavioral volitions: The interplay between goals and behaviors. European Journal of Social Psychology, 30, 705–731.
Petty, R. E., Tormala, Z. L., Briñol, B., & Jarvis, W. B. G. (2000). Implicit ambivalence from attitude change: An exploration of the PAST model. Journal of Personality and Social Psychology, 90, 21–41.
Rivis, A. J., & Sheeran, P. (2004). Descriptive norms as an additional predictor in the theory of planned behaviour: A meta-analysis. Current Psychology, 22, 264–280.
Sheeran, P., & Orbell, S. (1999). Augmenting the theory of planned behavior: Roles for anticipated regret and descriptive norms. Journal of Applied Social Psychology, 29, 2107–2142.
Smith, E. R. (1996). What do connectionism and social psychology offer each other? Journal of Personality and Social Psychology, 70, 893–912.
Smith, E. R., Fazio, R. H., & Cejka, M. A. (1996). Accessible attitudes influence categorization of multiply categorizable objects. Journal of Personality and Social Psychology, 71, 888–898.
Sparks, P., & Shepherd, R. (1992). Self-identity and the theory of planned behavior: Assessing the role of identification with green consumerism. Social Psychology Quarterly, 55, 388–399.
Strack, F., & Deutsch, R. (2004). Reflective and impulsive determinants of social behavior. Personality and Social Psychology Review, 8, 220–247.
Taylor, S. A. (2007). The addition of anticipated regret to attitudinally based, goal-directed models of information search behaviours under conditions of uncertainty and risk.
British Journal of Social Psychology, 46, 739–768.
Taylor, S. D., Bagozzi, R. P., & Gaither, C. A. (2001). Gender differences in the selfregulation of hypertension. Journal of Behavioral Medicine, 24, 469–487.
Trafimow, D., & Finlay, K. A. (1996). The importance of subjective norms for a minority of people: Between subjects and within-subjects analyses. Personality and Social
Psychology Bulletin, 22, 820–828.
Tversky, A., & Kahneman, D. (1973). Availability: A heuristic for judging frequency and probability. Cognitive Psychology, 5, 207–232.
Uleman, J. S., Newman, L. S., & Moskowitz, G. B. (1996). People as flexible interpreters: Evidence and issues for spontaneous trait inference. Advances in Experimental Social Psychology, 28, 211–279.
Wilson, T. D., & Hodges, S. D. (1992). Attitudes as temporary constructions. In L. L.
Martin & A. Tesser (Eds.), The construction of social judgments (pp. 37–65). Hillsdale, NJ: Lawrence Erlbaum Associates.
Wilson, T. D., Lindsey, S., & Schooler, T. (2000). A model of dual attitudes. Psychological Review, 107, 101–129.
This research was funded by Unilever (grant SPHERE CH-2003-1010). Correspondence regarding this article should be sent to: Juliette Richetin, Department of Psychology Milan-Bicocca, Viale dell’Innovazione, 10 (U9), 20126 Milano, Italy (email@example.com).
1 The role of past behavior will not be considered further as a predictor of behavior in this contribution for two main reasons. First, the theoretical status of past behavior as a predictor of behavior has been the subject of much controversy (cf. Ajzen, 2002b). Second, the recent literature on automaticity within social cognition provides a complementary angle that investigates in more depth the influence of automatic factors on action (e.g., Bargh, 2006). However, past behavior will be considered as a factor influencing evaluations.
2 It is common practice to assess goal variables before behavioral variables. However, given that the focus was on the behavioral level (soft drinks), goal desire was measured at the end of the measurement session.
3 Using either of the two composites did not significantly affect the results.
4 The standard format of measuring negative and positive anticipated emotions that asks separately negative vs. positive anticipated emotions was not used in the context of failing vs. succeeding in drinking soft drinks. The behavior does not have a strong component of trying and therefore it would have been awkward to use the standard stem.
5 An alternative strategy yielding the same results would have been to separately test the four possibilities with different models and compare the fit and parameters of the models. However, given that three possibilities (mediated memory-based, unmediated memory-based, online-based) are nested in a model testing the fourth possibility (both memory- and online-based) and, as a reviewer suggested, only this latter was tested and the relevant parameters inspected to verify which possibility was supported empirically.
Table 1. Correlations among Constructs.
Figure 1.Parameters for the TPB.
Figure 2.Parameter Estimates for the MGB.
Figure 3.Parameter Estimates for the EMGB.