based on this assumption of a context-independent general habit. Car use is a prototypical example. .... travel mode would you most likely use? Please respond ...
Transportation 30: 97–108, 2003 2003 Kluwer Academic Publishers. Printed in the Netherlands.
Does habitual car use not lead to more resistance to change of travel mode? SEBASTIAN BAMBERG1, DANIEL RÖLLE2 & CHRISTOPH WEBER2 1
Department of Social Psychology, University of Dresden, D-01062, Germany; 2 Institute of Energy Economics and the Rational Use of Energy, D-70550 Stuttgart, Germany
Key words: habit, past behavior, reasoned action Abstract. An experiment examined the effects of an intervention (combination of information and a free public transport ticket) in a changed decision context (moving to a new residence) on travel mode choice by car users. If past frequency of car use has resulted in an automatic response to goal-related cues, one should expect resistant to change of travel mode. However, the results failed to show this. Neither past behavior or a direct habit measure predicted future travel behavior. Instead, the intervention influenced attitude, subjective norm, and perceived behavioral control, and consistent with Ajzen’s theory of planned behavior, these were the main causes of the change of travel mode.
1. Introduction It has been argued that past behavior is the best predictor of future behavior. However, this could reflect a statistical association. Past behavior may simply be a proxy for the causal effects of other psychological factors. For example, Ajzen’s (1991) theory of planned behavior (TPB) postulates that behavior is guided by beliefs about the likely consequences of the behavior (attitude), beliefs about the normative expectations of others (subjective norm), and beliefs about the presence of factors that may facilitate or hinder performance of the behavior (perceived behavioral control). In combination these three factors lead to the formation of a behavioral intention which is the immediate determinant of behavior. According to the theory, behavior is thus assumed to be reasoned, controlled, or planned. The TPB has been challenged by the argument that human behavior is habitual or automatic (e.g. Aarts & Dijksterhuis 2000; Aarts et al. 1998; Bagozzi 1981; Fazio 1990; Ouellette & Wood 1998; Ronis et al. 1989; Triandis 1977). Measures of past behavior have played an important role in attempts to test the validity of this argument. If behavior is always reasoned, then frequency of prior behavior should only have an indirect link to later behavior since its effect would be mediated by intention and perceived behavioral control. However, in regression analyses past behavior is typically found to
98 improve the prediction of later behavior over and above intention and perceived behavioral control (see Ouellette & Wood 1998, for a meta-analysis). Findings of this kind are generally taken to mean that the behavior is habitual so that it is, at least in part, under the direct control of the stimulus situation. Frequency of past behavior is thus an indicator of habit strength that can be used as an independent predictor of later behavior. Developing a habit has far-reaching consequences for cognitive functioning, for instance, the ways situations are perceived and information processed. Verplanken and Aarts (1999) provide empirical evidence that habit attenuates the processing of information about the context in which choices are made as well as information about choice options, thus information that is used for judgments necessary to make informed choices. Habits seem to go together with a cognitive orientation that makes an individual attend less to new information and new courses of action, and is characterized by a preference for relatively simple and heuristic-based choices. Stable situational contexts are usually assumed to be a necessary prerequisite for individuals to develop habits. In contrast, Verplanken et al. (1994) and Verplanken and Aarts (1999) propose that once developed, habits can generalize to many different situations. For such general habits, not even situational invariance should be necessary. General habits are controlled by goal-related cues that appear in many different situations. The responsefrequency measure of habit developed by Verplanken et al. (1994) is directly based on this assumption of a context-independent general habit. Car use is a prototypical example. As shown by Verplanken et al. (1997), a strong car use habit makes travel mode choice script-based, so that minimal information is needed to make it. Verplanken and Aarts (1999) view habits as effective ways to reach frequently aspired goals with rewarding consequences. However, they also note “the dark side of a habitual mind set” (p. 125): Habits may turn into suboptimal behaviors when new situations are encountered, in changing environments, when goals change, or when new goals are adopted that are incompatible with current habits. If this view is correct, it would have important implications for attempts to change frequently performed, habitual behaviors. Similarly, current models of persuasion and attitude change assume that persuasive communication is only likely to result in attitudes that are related to subsequent behavior if the individual is highly motivated and able to process the information actively (e.g. Chaiken 1987; Petty & Cacioppo 1986). Thus in the case of strong habits, providing persuasive information would be an ill-fated strategy. For example, people with a strong car-use habit should have low motivation to attend to and process information about public transport. Even when persuasive communication changes attitudes and intentions, in the case of individuals with a strong habit, these
99 changes in attitudes and intentions should have little if any behavioral effect because the habit is not under intentional control but automatically activated by the situation. Consequently, Verplanken and Aarts (1999) argue that habitual behaviors are very difficult to change. The best way of doing this would be to block or punish their execution (e.g. by physical or financial measures). The focus of the present study is the hypothesis that it is more difficult to change frequent car use. To test this hypothesis empirically, data from a project are used in which the effectiveness of an intervention (combination of information and a free public transport ticket) on travel mode choice was evaluated shortly after participants changed their place of residence. The theoretically interesting feature of this data set is that it allows an analysis of how a change of the decision context alone affects the travel mode choice of frequent and less frequent car users and whether they differ in how they are affected in this new decision context by an intervention promoting the use of public transport. If the assumption is correct that frequent car use is strongly determined by a general habit that is automatically activated by goal-related cues, then not even the change of the decision context should have an effect on the travel mode choice by frequent car users although it may have on infrequent car users. Furthermore, the additional intervention should have no or a smaller effect. An alternative hypothesis can be derived from the theory of planned behavior as a model of reasoned action. According to this theory, even in the case of a behavior that has become routine with practice, the behavior is regulated at some level of awareness such that the relevance of new information is noticed and taken into consideration. Thus, if the new information provided by the new decision context and the intervention is perceived as personally relevant and persuasive, it may change the cognitive basis of intentions and behaviors. As a result, it would be expected that frequency of past behavior will lose some of its predictive validity. In fact, with sufficient change in attitudes, subjective norms, or perceived behavioral control, there should be no effect of a prior behavior on later behavior.
2. Method 2.1. Participants and study design The participants of the study consisted of a sample of people planning to move within the next 6 months to Stuttgart, Germany. Stuttgart is the capital of Baden-Württemberg with about 600,000 residents and a high-quality public transport service. The participants were contacted prior to their move by
100 using addresses and telephone numbers obtained from rent advertisements appearing in the local newspapers. A lottery ticket with attractive money prizes was used as an incentive to participate. Of 600 who were contacted, 241 sent back the first questionnaire assessing actual travel mode choice, car use habit, and items measuring the TPB variables concerning car, bicycle, and public-transport use at their old place of residence. The mean age of the participants was 28.6 years (ranging from 17 to 58), 53% were male, 38% reported that they had a university degree, 98% had a driving license, and 74% reported that they could always use a car. The respondents were randomly assigned to a control (n = 123) and an experimental group (n = 118). They were, however, not informed that they were participating in an experimental intervention but believed that they took part in a university research project aimed at analyzing the impact of residential moves on daily travel. After six months, 191 (99 in the control and 92 in the experimental group) participants actually moved and were contacted again at their new place of residence. Only participants in the experimental group received via mail the intervention “Personal Public Transport-Test-Parcel.” This intervention was sent by the local transport company that did not make reference to the research project. The intervention consisted of an official welcome letter with a short presentation of the company and its services and the invitation to test these services. For this purpose the parcel contained a free ticket valid one day for all public transport services in Stuttgart. Furthermore, the parcel contained all the information needed for using these services including a map of the subject’s dwelling district where all public transport routes and stops were marked, a time table, a small brochure with examples of connections from the dwelling district to frequently used shopping, leisure and cultural facilities, information about the tariff and ticket sales system, and information about a telephone hot line where people can access further advice and information. Six weeks after the experimental group received the intervention, all 191 participants who had moved received a second mail questionnaire. Of these 191 participants, 169 completed the second questionnaire (90 in the control and 79 in the experimental group). 2.2. Questionnaire Most items included in the questionnaire were designed to assess the constructs entailed by the theory of planned behavior. The travel mode alternatives considered were driving a car, using public transport, and riding a bicycle. With respect to each alternative, respondents answered items designed to measure attitude, subjective norm, perceived behavioral control, and intention.
101 Attitudes were assessed by means of the following two items: “For me, to take public transport (use my car/bicycle) for daily travel from my current place of residence would overall be good – bad” and “pleasant – unpleasant”. Responses were obtained on 10-point graphic scales. Items measuring subjective norm were formulated as follows: “Most people who are important to me would support my using public transport (car/bicycle) for daily travel from my current place of residence” and “Most people who are important to me think that I should use public transport (car/bicycle) for daily travel from my current place of residence.” Each of these items was followed by a 10-point graphic scale with endpoints labeled likely and unlikely. To assess perceived behavioral control, respondents answered the following two items: “For me to use public transport (car/bicycle) for daily travel from my current place of residence would be easy – difficult” and “My freedom to use public transport (car/bicycle) for daily travel from my current place of residence is high – low”. They used 10-point graphic scales. Participants also responded to the following two intention items by checking 10-point graphic scales: “My intention to use public transport (car/bicycle) for daily travel from my current place of residence is strong – weak” and “I intend to use public transport (car/bicycle) for daily travel from my current place of residence is likely – unlikely”. Finally, actual behavior1 was assessed by a one-day “mobility diary.” This mobility diary consisted of a booklet with one page for each trip on a prior specified day (Social Data 1993). For each trip respondents registered time and starting location, purpose (work, shopping, or leisure), travel mode (car, bike, walk, or public transport), destination, time of arrival, and estimated distance. In addition to assessing the predictors in the theory of planned behavior, the questionnaire also enquired into the respondents’ past behavior by asking them to indicate how often they had used alternative travel modes for daily travel from their current place of residence. Three travel modes were listed, car, public transport, and bicycle. The response alternatives were always, often, occasionally, seldom, and never. The response-frequency measure of habit (Verplanken et al. 1994) was also administered. Respondents were given the following instructions: “Listed below are a few leisure-time activities that you may often perform. Assume that you would like to spontaneously engage in one of these activities. Which travel mode would you most likely use? Please respond quickly without much deliberation.” The travel mode choices offered were car, public transport, bicycle, and walking. Participants were asked to indicate which one they would choose for the following 10 destinations or purposes: Summer excursion with friends to a lake, visit a friend, visit your parents, engage in sports, stroll through the city; evening visit to a bar, a trip on a nice day, routine grocery
102 shopping, eat in a restaurant, and go to the movies. To create two indices of habit, the 10 situations were divided at random into two groups of five, and within each group the number of times respondents chose the car was counted.
3. Results 3.1. Effectiveness of the intervention Analysis of the one-day mobility diary used to measure the actual travel mode choice showed that for the 169 participants who completed both questionnaires, public transport accounted for 12.8 percent of all trips reported at the old place of residence. At the new place of residence, this proportion increased significantly to 29.3 percent (p < 0.001). In contrast, the proportion who drove car declined from 55.5 percent to 41.8 percent (p < 0.001). The proportion of subjects who biked also changed significantly from 12.7 percent to 5.8 percent. Only the proportion of trips for which respondents reported that they walked remained unchanged (17.8 vs. 22 percent, p = 0.13). As can be seen in Table 1, before the move/intervention there is no significant difference in public transport use between the control and experimental groups (p = 0.99). In contrast, after the move/intervention in the control group there is only a small and insignificant increase of public transport use (from 18.9% to 24.2%, p = 0.25), whereas in the experimental group public transport use increased from 19.0% before to 46.8% (p < 0.001) after the move. The difference in public transport use between control- and experimental group is now statistically significant (p < 0.001). Table 2 presents the means of the difference-scores (public transport – ((car + bicycle)/2]) of the items measuring the TPB-constructs before and after the move for the control and experimental groups. As can be seen, after the move in the control group there is no change in the difference scores, whereas in the experimental group all difference scores change drastically in favor of public transport.
Table 1. Effects of the change of residence and intervention on public transport use (N = 169).
Control-group (n = 90) Experimental-group (n = 79)
Old place of residence
New place of residence (after intervention)
18.9% 19.0%
24.4% 46.8%
103 Table 2. Difference scores of the TPB constructs before and after the move for the experimental and control groups. Difference scores
Attitude (1) Attitude (2) Subjective norm (1) Subjective norm (2) Perceived behavioral control (1) Perceived behavioral control (2) Intention (1) Intention (2)
Experimental-group n = 79
Control-group n = 90
Before
After
Before
After
–3.6 –3.6 –3.0 –2.7 –3.5 –3.5 –2.4 –2.8
–0.0*** –1.2*** –0.1*** –0.6*** –0.2*** –0.4*** –1.3*** –1.2***
–1.9 –3.0 –1.6 –1.5 –2.7 –2.6 –2.2 –2.1
–0.9 –2.6 –0.6 –0.6 –1.5* –1.3** –1.4 –1.0
Note: The means of the indicators can theoretically range from –10 to +10. Negative values mean that public transport is evaluated more negatively than the other alternatives. * p < 0.05; ** p < 0.01; *** p < 0.001.
3.2. Prediction of intention and behavior The data were submitted to structural equation analysis using the LISREL 8 computer program (Jöreskog & Sörbom 1993). Figure 1 displays the structural model parameters for public transport use as well as the amount of explained variance in intention and behavior at the old place of residence. Goodness-of-fit indices reached satisfactory levels (χ2(n = 241, df = 120) = 246.12, p = .000, RMSEA = 0.06, NNFI = 0.94, CFI = 0.95). As can be seen, the components of the TPB accurately predict intentions and actual public transport use. Attitude, subjective norm, and perceived behavioral control accounted for 67% of the variance in intention, and intention and behavioral control both have strong paths to public transport use (66% explained variance). Past car use has strong direct negative effects on attitude, subjective norm, perceived behavioral control, and intention but no additional direct effect on behavior. In the second step the response frequency measure has negative effects on attitude, subjective norm, and perceived behavioral control but no direct effect on behavior. Furthermore, past car use has a strong effect on the response frequency measure but this measure only partly mediates the effect of past car use frequency. Figure 2 displays the structural model parameters for public transport use as well as the amount of explained variance in intentions and behavior, simultaneously estimated for the data before and after the move (n = 169). To cross-validate the TPB model, the parameters of the measurement and structural model after the move were constrained equal to those before the move.
104
Figure 1. Structural model with standardized path coefficients and explained variance in intentions and behavior. Dependent variable is public transport use. The subscripts 1 refer to wave 1 before the move. A = attitude; SN = subjective norm; PBC = perceived behavioral control; I = intention; B = behavior. The latent constructs A, SN, PBC, and I were estimated by using the difference-scores (public transport – [(car + bicycle)/2]) of the respective two indicator items.
Figure 2. Structural model with standardized path coefficients and explained variance in intentions and behavior. Dependent variable is public transport use. The subscripts 1 and 2 refer to wave 1, before the move, and wave 2, after the move. A = attitude; SN = subjective norm; PBC = perceived behavioral control; I = intention; B = behavior. The latent constructs A, SN, PBC, and I were estimated by using the difference-scores (public transport – [(car + bicycle)/2]) of the respective two indicator items. * not significant.
105 In both waves the components of the TPB accurately predicted intentions and use of public transport. Attitudes, subjective norms, and perceived behavioral control accounted for 69% of the variance in intention in the first wave and 77% in the second wave, and in both waves intention and behavioral control had strong paths to public transport use. Figure 2 also displays the stability coefficients of the different measures over the two waves. These stabilities are low (that of behavior even insignificant), indicating that after the move respondents had strongly changed their perception of the situation as well as their behavior. For a direct empirical test of the intervention effects, membership to control vs. experimental group was introduced as a dummy variable in the model. As Figure 2 shows, this dummy variable has a direct significant effect on behavior as well as on intention, even after controlling the effect of the TPB-variables. As can be further seen in Figure 2, after controlling the effects of the TPB constructs and the intervention, past car use no longer exerts any significant effect on the intention to use public transport or the actual public transport use at the new place of residence. These results were replicated with the response frequency measure of car habit. The goodness-of-fit indices of the model displayed in Figure 3 are satisfactory (χ2(n = 169, df = 449) = 657.81, p = .000, RMSEA = 0.05, NNFI = 0.94, CFI = 0.94).
4. Discussion From a substantive point of view, an important result of the present study was that in new decision contexts former car users show a strong behavioral reaction even to small, relatively inexpensive interventions. Furthermore, it is the combination of the new decision context with an intervention that causes the change. Thus, a new decision context may influence behavior indirectly by creating a “sensible phase” in which people’s attention to new information and their motivation to process it more actively is increased. If in this situation people receive personally relevant and persuasive information, they seem to be more likely to use it for making new behavioral decisions. These results provide an interesting starting-point for the development of practical measures to promote the use of public transport. Such measures should try to target people who are confronted with changing decision contexts like changing the place of residence, changing from work to retirement or from education to work. But the existence of objectively high-quality public transport services as in Stuttgart is probably a necessary prerequisite for the success of such measures. From a theoretical point of view the most interesting result of our study is that we do not find the expected direct effect on future travel mode choice
106 of past car use or habit. At both time points the effect of past behavior and the habit measure is completely mediated by intention and behavioral control. Thus, the present data do not seem to support the hypothesis that the reaction of car users toward the intervention is influenced by a general car use habit, which is activated automatically by goal-related cues. The results are more in line with the alternative hypothesis that even when a behavior is routine, the introduction of new information by a new decision context and an intervention can change the cognitive foundation of intention which changes the intention that determines subsequent behavior. However, it would be premature to conclude that habit does not influence human behavior. A weakness of the present study consists in not comparing the reaction of car users in a stable decision context with that of car users in a new decision context. Thus one can argue that the new decision context may have “paved the way” for the intervention by breaking the prior existing car use habit. According to this view, only in stable decision contexts habits should influence car users reaction to an intervention. Yet, the results of another study (Bamberg et al. 2002) do not support this argument. That study analyzed students’ reactions to a newly introduced free-ticket within a stable decision context (same route from the apartment to university campus). Again, the intervention caused a drastic increase in public transport use. But also in this stable decision context neither past behavior nor the habit measure had a direct effect on public transport use after the intervention. Instead, the results of that study also confirmed the hypothesis that the intervention caused changes in the cognitive foundation of intention and perceived behavioral control and that these changes are responsible for the observed strong behavioral reaction. But more studies are required to clarify whether the notion of habit contribute to our understanding of what conditions determine the effectiveness of interventions trying to change frequent travel behaviors. A prerequisite for answering this question is a valid habit measure. One can question the validity of the response-frequency measure. The procedure asks respondents to indicate their intentions to perform a particular behavior in different hypothetical situations. Perhaps the resulting measure taps a generalized intention to perform the behavior in question than a habit. The justification for assuming that it may represent something other than a generalized intention is the instruction to participants to respond as quickly as possible. It is an empirical question whether time pressure has any effect on responses and, if so, whether the measure obtained under time pressure is in fact an indicator of habit strength.
107 Acknowledgement We thank Tommy Gärling for helpful suggestions to improve the manuscript.
Note 1. It is worth noting that the measure of behavior was obtained at the same time as the measures of intention. Strictly speaking, therefore, measurement of intention and performance of the intended behavior did not follow the logically required temporal sequence. This procedure was necessitated by limitations on the feasibility to recontact respondents. However, travel mode choice is quite stable over a short period of time (see Bamberg & Lüdemann 1996), and there is no reason to expect that a slightly delayed measure of behavior would have produced substantially different findings.
References Aarts H & Dijksterhuis A (2000) Habits as knowledge structures: Automaticity in goal-directed behavior. Journal of Personality and Social Psychology 78: 53–63. Aarts H, Verplanken B & van Knippenberg A (1997) Habit and information use in travel mode choices. Acta Psychologica 96: 1–14. Aarts H, Verplanken B & van Knippenberg A (1998). Predicting behavior from actions in the past: Repeated decision making or a matter of habit? Journal of Applied Social Psychology 28: 1355–1374. Ajzen I (1991) The theory of planned behavior. Organizational Behavior and Human Decision Processes 50: 179–211. Bagozzi RP (1981) Attitudes, intentions, and behavior: A test of some key hypotheses. Journal of Personality and Social Psychology 41: 607–627. Bamberg S, Ajzen I & Schmidt P (2002) Choice of travel mode in the theory of planned behavior: The roles of past behavior, habit and reasoned action. Basic and Applied Social Psychology. In press. Bamberg S & Lüdemann C (1996) Eine Überprüfung der Theorie des geplanten Verhaltens in zwei Wahlsituationen: Rad vs. Pkw und Container vs. Hausmüll. Zeitschrift für Sozialpsychologie 27: 32–46. Chaiken S (1987) The heuristic model of persuasion. In: Zanna MP, Olson JM & Herrman CP (eds) Social Influence: The Ontario Symposium (Vol. 5). Hillsdale, NJ: Erlbaum. Fazio RH (1990) Multiple processes by which attitudes guide behavior: The MODE model as an integrative framework. In Zanna MP (ed) Advances in Experimental Social Psychology (Vol. 23, pp 75–109). San Diego, CA: Academic Press. Jöreskog KC & Sörbom D (1993) LISREL 8: Structural Equation Modeling with the SIMPLIS Command Language. Hillsdale, NJ: Erlbaum. Ouellette JA & Wood W (1998) Habit and intention in everyday life: The multiple processes by which past behavior predicts future behavior. Psychological Bulletin 124: 54–74. Petty RE & Cacioppo JT (1986) Communication and Persuasion: Central and Peripheral Routes to Attitude Change. New York: Springer-Verlag. Ronis DL, Yates JF & Kirscht JP (1989) Attitudes, decisions, and habits as determinants of repeated behavior. In: Pratkanis AR, Breckler SJ & Greenwald AG (eds) Attitude Structure and Function (pp 213–239). Hillsdale, NJ: Erlbaum.
108 Triandis HC (1977) Interpersonal Behavior. Monterey, CA: Brooks/Cole. Verplanken B & Aarts H (1999) Habit, attitude, and planned behaviour: Is habit an empty construct or an interesting case of goal-directed automaticity? In: Stroebe W & Hewstone M (eds) European Review of Social Psychology (pp 101–134), Chichester, England: Wiley. Verplanken B, Aarts H & van Knippenberg A (1997) Habit, information acquisition, and the process of making travel mode choices. European Journal of Social Psychology 27: 539–560. Verplanken B, Aarts H, van Knippenberg A & van Knippenberg C (1994) Attitude versus general habit: Antecedents of travel mode choice. Journal of Applied Social Psychology 24: 285–300.
About the authors Sebastian Bamberg is visiting professor at the Institute of Social Psychology at the Technical University of Dresden. His main research interests are the development of psychological action theories and their application to the understanding of how interventions change ecologically related behaviors. Daniel Rölle is a social scientist and Ph.D. candidate. At the moment he works as a research associate in the Institute of Energy Economics and the Rational Use of Energy, University of Stuttgart. His research interest is on factors explaining travel behavior, especially on the link between the acceptability of some policy measures and the intention to change travel mode. Christoph Weber received his Ph.D. in economics from the University of Hohenheim (Stuttgart). He is now head of the department of Rational Use of Energy and Energy Demand (REN) at the Institute of Energy Economics and the Rational Use of Energy (University of Stuttgart). His main research interests include Energy Demand, Energy Management and consumer behavior.