Professional Documents
Culture Documents
Attitudes vs General Habit in Travel Mode Choice
Attitudes vs General Habit in Travel Mode Choice
Attitudes vs General Habit in Travel Mode Choice
Introduction
‘This study was supported by a grant from the Ministry of Transport and Public Works,
The Netherlands. The authors wish to thank Birgit van Delft, Benno Labree, Hanneke Karthaus,
Pieter Meima, Dries Vermeulen, and Miranda Wildenbeest for their contribution in collecting
the data, and Ola Svenson for his comments on the manuscript.
2Correspondence concerning this article should be addressed to Bas Verplanken, University
of Nijmegen, Department of Social Psychology, PO Box 9104, 6500 HE Nijmegen, The
Netherlands; e-mail: verplanken@psych.kun.nl.
When confronted with trips we are sufficiently familiar with, such experiences
may enable us to arrive at travel mode choices in a rather mindless, habitual
manner. In other words, on such occasions, travel mode choice is a matter of
habit, rather than reasoned action. Habits are relatively stable behavioral
patterns, which have been reinforced in the past. Habits are executed without
deliberate consideration, and result from automatic processes, as opposed to
controlled processes like consciously made decisions (cf. Shiffrin & Schneider,
1977).
The relationship among an attitudinal component (in the form of behav-
ioral intention), habit, and behavior was formalized by Triandis (1977). In
his model, the probability of an act (Pa)is a weighted function of habit (H)
and behavioral intention (I), multiplied by facilitating conditions Q, that is,
the presence or absence of conditions that facilitate the performance of the
act (e.g., the individual’s ability to perform the act). This relationship is
expressed as:
results of that study suggest that both the use and nonuse habit (nonredun-
dantly) predict behavior, next to attitude and intention. Moreover, the nonuse
habit interacted with intention in its prediction of behavior.
Like seat belt use, the repetitive nature of travel mode choice makes a
role of habit in this behavior domain very likely (cf. Banister, 1978). Apart
from attitudes, therefore, habit may be included in models predicting travel
mode choice. It may be predicted that (a) the more favorable the attitude to-
ward a particular travel mode (and the more unfavorable the attitude toward
alternative modes), the more likely is the choice of that mode and (b) the more
strongly the habit of using a particular mode, the more likely is the choice of
that mode. It is worth noting that, in addition to the independent effects of
attitude and habit hypothesized above, the argument of Triandis (1977) sug-
gests that there is also an interaction effect of attitude and habit on choice:
When habit is strong, the attitude-behavior relation will be weak and, con-
versely, when habit is weak, the attitude-behavior relation will be strong (i.e.,
habit serves as a moderator variable; cf. Baron & Kenney, 1986).
Antecedents of Habit
Operationalizations of Habit
a set of six options (i.e., bus, bicycle, cab, car, train, or walking). The assump-
tion behind this operationalization is that, when respondents are asked to
respond to global stimuli, which are presented with a minimum of detail, there
is little possibility engaging in making elaborate tradeoffs between the pros and
cons of various travel modes. Rather, we expect that respondents’ responses are
guided by preexisting schemas or scripts about mode choice in general (cf.
Kahle, 1984; Kahle et al., 1981). When habit with respect to car choice is
strong, the choice of the car as the travel mode will dominate such schemas,
which is supposed to be reflected in a relatively large number of car choices
across the ten presented journeys.
Method
Respondents
habit, and a scale for decisional involvement. The subjects also responded to
several demographic items such as age, gender, number of cars available in the
household, number of household members having a drivers license, and an
estimate of the annual mileage. The questionnaire contained a number of other
additional items, which are not relevant here.
For the analyses to be reported here only those respondents were selected
who had a drivers license, and had a car at their disposal ( N = 199). The ages
ranged from 19 to 65 ( M = 39.9). There were 92 males and 107 females. The
self-reported mean annual mileage is 9,956.
Measures
Car choice behavior was measured by asking the respondents how often
during the past 2 months they had traveled by car in order to go shopping in
either one of the two cities (cf. Mittal, 1988; Wittenbraker et al., 1983).
Responses were given on a 4-point scale ranging from never (1) to very
often (4).
Attitude toward traveling by car for shopping in either one of the two cities
was measured by averaging two bipolar 7-point items ranging from very bad to
very good and very unfavorable to very favorable respectively ( r = 0.58, p <
.OOl). The same items were used to measure attitude toward traveling by train
for shopping in either one of the two cities ( r = 0.64, p < .OOl).
General habit of choosing the car as the travel mode was measured by
presenting respondents with ten imaginary situations calling for a choice of
travel mode (e.g., “Going to the beach with some friends, Going for sporting
as leisure activity, Going to visit friends in the village, Going for shopping after
work”). The respondents were required to choose between six possible travel
modes for each particular journey. The six modes presented were bicycle, bus,
cab, car, train, and walking. The car choice habit measure was calculated by
counting the number of times the car was mentioned as the travel mode. Thus,
the measure could have values ranging from 0 to 10. The measure correlated
moderately, however significantly, with the respondents’ annual mileage ( r =
.37, p < .OOl). There is also a sizable correlation ( r = .55, p < .001) with one
additional item, which reflected some form of habit. It was worded “Whether I
go to work or go shopping, I almost always travel by car,” and it was accom-
panied by a 7-point disagree-agree scale.
Decisional involvement was measured with eight 7-point Likert-type agree-
disagree items. Because decisional involvement is supposed to be situation-
specific, the subjects were presented with the stem, “When I go shopping in
City A or City B, .” The items were, respectively: “I want to expend
effort to find out which travel method is the best, I think about which travel
292 VERPLANKEN ET AL.
method I want to use, there is no doubt in my mind about how I will travel, I
want to know in detail which pros and cons various travel methods have, I think
it is useless to expend time and energy to find out which travel method is most
suitable, I want to prepare the journey well before, I don’t need to deliberate
about how I want to travel because I already know, and I use the first method
of transportation that comes to mind.” The eight items form a reliable scale
(coefficient alpha = 0.83). After reversing the third, fifth, seventh, and eighth
item, the items were averaged into a score representing decisional involvement.
Car availability was measured by the ratio of the number of cars available
in the household (usually one) and the number of household members having
a drivers license. This measure is widely used in travel behavior research to
reflect the amount of competition in a household with respect to car use.
Results
Path Analysis
Mean, Standard Deviation, and Intercorrelations of Car Choice Behavior, Car Attitude, Train
Attitude, Car Habit, Decisional Involvement, and Car Availability
~
attitude toward
train-choice
decisional general
involvement ' -.36 car-choice habit
y p availability
Figure I . Path analysis for antecedents of car choice behavior (the numbers are
standardized path coefficients; all coefficients are statistically significant at
.001-level, except the path between train attitude and car choice behavior,
which is significant at .05-level)
In order to test the hypothesized tradeoff between car attitude and car
choice habit in their prediction of car choice behavior, a stepwise hierarchical
multiple regression analysis was performed, in which behavior is predicted by
attitude, habit, and the Attitude x Habit interaction term. All variables were
first standardized. The prediction of car choice behavior by car attitude (R2 =
.33) was significantly improved by including car choice habit (R2 = .47;
F-change = 47.16, p < .001). Inclusion of the interaction term after attitude and
habit yielded a modest, however significant, improvement of the prediction (R2
= .49; F-change = 8.32, p < .005). In Table 2, the nature of the interaction is
shown. The sample was split as close as possible at the 33rd and 67th percen-
tiles of the distribution of the habit measure. Then, the correlation between car
Table 2
attitude and behavior was calculated for the three subgroups. This correlation
was stronger among the subgroup with relatively weak habit compared with
those who indicated a relatively strong habit. The difference between attitude-
behavior correlations of the strong versus weak habit subgroups was signifi-
cant, z = 2.67, p < .01. In other words, when habit was weak the attitude and
behavior measures shared 43% of their variance, whereas when habit was
strong, the attitude and behavior measures only shared 8% of their ~ a r i a n c e . ~
Discussion
The results of the present study suggest that habit is an important determi-
nant of travel mode choice. Car choice behavior was satisfactorily predicted
from attitudes toward choosing available options, on the one hand, and car
choice habit, on the other hand. Evidently, predicting mode choice behavior
from attitudes alone is not optimal because car choice habit also explains 14%
of variance. In this respect, mode choice resembles other highly repetitive
behavior, for which the predictive role of habit has been demonstrated (e.g.,
Kahle et al., 1981; Landis et al., 1978; Mittal, 1988; Wittenbraker et al., 1983).
It is worth emphasizing that the habit measure reported here is of a general
nature. That is, rather than measuring a situation-specific habit (i.e., the habit
of choosing the car for this particular shopping trip), our measure attempted to
tap a generalized habit of choosing the car as the travel mode. Consequently,
the habit component is more powerful because it applies to a large variety of
mode choice contexts. Together with Mittal’s (1988) identification of use
versus nonuse habit, the present distinction between specific and general habit
may constitute a useful qualification of the construct of habit for future re-
search .
The regression results with respect to the interaction between attitude and
habit in the prediction of behavior suggest that a tradeoff between attitude and
habit exists (cf. Bagozzi, 1981; Triandis, 1977). When habit is weak, the
attitude-behavior link is stronger than when habit is strong. On the other hand,
among the strong habit subgroup, a small but significant correlation of .28
between attitude and behavior indicated that attitudes still guide these indi-
viduals’ behavior to some extent, even if the behavior was repetitively ex-
hibited. When the behavior is frequently repeated, it is, of course, quite
conceivable that the reasons that originally resulted into the onset of the
behavior still retain much of their validity. When respondents are asked to
report their attitudes, these reasons become salient once more. Ronis et al.
(1989) discussed other mechanisms that may contribute to an attitude-behavior
correlation when behavior is repetitive. For example, when behavior is repeat-
edly exhibited, attitudes may be inferred from self-perceptions of that behavior
(Bem, 1967). However, the present data suggest that, although some correla-
tion remains when habit is strong, attitudes reflect behavior to a much greater
extent when habit is weak, which makes a self-perception explanation less
likely.
As was discussed in the introduction section, various operationalizations of
habit are reported in the literature. The instrument we use to measure habit does
not rely on respondents’ abilities to remember past behavior or to remember
psychological processes that are supposed to reflect habit. Rather, the measure
is based on current schematic thinking, which is supposed to be guided by the
strength of the habit component (cf. Kahle, 1984). In this respect, our measure
has an advantage compared to previously used measures. A point of criticism
might be raised concerning whether the measure reflects an attitude, rather than
a habit. Although we do not have external validation data concerning our
measure, we think the measure reflects habit, and not, or only partially, atti-
tude, for a number of reasons. First, the measure component behaves in the
model as was hypothesized on theoretical grounds. Second, its correlation with
attitude is moderate and can be expected to be higher if the measure reflects car
choice attitude. Moreover, the occurrence of the (predicted) interaction of habit
and attitude in their prediction of behavior would be unlikely if the habit
ATTITUDE VERSUS GENERAL HABIT 297
level of habit. The behavior of those with strong habit may be especially
difficult to change with traditional persuasive communications aimed at chang-
ing attitudes. As the present results suggest, in that case, attitudes are only
weakly related to behavior and, consequently, such attempts cannot be ex-
pected to succeed. The model tested in this study may lead to the consideration
of approaches that are primarily aimed at changing the habit component. This
might be attempted directly, for example, by employing technical measures
that reduce the attractiveness of car travel or by providing incentives for the
target behavior. Alternatively, increasing the level of decisional involvement
(cf. Geller et al., 1989) may constitute a promising strategy. Higher levels of
decisional involvement may result in consideration of alternative travel modes
and may thus decrease the strength of habit. Once habit strength is reduced,
persuasive communications aimed at changing attitudes may be more success-
ful in terms of behavioral changes. Especially if infrastructure has been im-
proved (e.g., new bus lines) but old habits have remained, strategies focusing
on increasing decisional involvement and decreasing habit may be promising
perspectives. The goal, then, is not to change notorious car drivers into rigid
users of public transport, but to increase the awareness that travel mode choice
may often involve some level of deliberation and decision-making. Such a
strategy may thus promote a more flexible travel choice behavior.
References
Fishbein, M., & Ajzen, I. (1975). Belief; attitude, intention, and behavior: An
introduction to theory and research. Reading: Addison-Wesley.
Fredricks, A. J., & Dossett, D. L. (1983). Attitude-behavior relations: A com-
parison of the Fishbein-Ajzen and the Bentler-Speckart models. Journal of
Personality and Social Psychology, 4 5 5 0 1-5 12.
Geller, E. S., Kalsher, M. J., Rudd, J. R., & Lehman, G . R. (1989). Promoting
safety belt use on a campus: An integration of commitment and incentive
strategies. Journal of Applied Social Psychology, 19,3-19.
Geller, E. S., Winett, R. A., & Everett, P. B. (1982). Preserving the environ-
ment: New strategies for behavior change. New York: Pergamon Press.
Gilbert, G., & Foerster, J. (1977). The importance of attitudes in the decision
to use mass transit. Transportation, 6, 321-332.
Hartgen, D. T. (1974). Attitudinal and situational variables influencing urban
mode choice: Some empirical findings. Transportation, 3, 377-392.
Houston, M. J., & Rothschild, M. L. (1978). Conceptual and methodological
perspectives on involvement. In S . Jain (Ed.), Research frontiers in mar-
keting: Dialogues and directions (pp. 184- 187). Chicago: American Mar-
keting Association.
Hull, C. L. (1 943). Principles of behavior: An introduction to behavior theory.
New York: Appleton-Century Crofts.
Kahle, L. R. (1984). Attitudes and social adaptation: A person-situation inter-
action approach. Oxford: Pergamon.
Kahle, L. R., Klingel, D. M., & Kulka, R. A. (1981). A longitudinal study of
adolescent attitude-behavior consistency. Public Opinion Quarterly, 45,
402-414.
Landis, D., Triandis, H. C., & Adamopoulos, J. (1 978). Habit and behavioral
intentions as predictors of social behavior. The Journal of Social Psychol-
ogy, 106,227-237.
Mittal, B. (1988). Achieving higher seat belt usage: The role of habit in
bridging the attitude-bahavior gap. Journal of Applied Social Psychology,
18, 993-1016.
Mittal, B. (1 989). Measuring purchase-decision involvement. Psychology and
Marketing, 6 , 147-162.
Mittal, B., & Lee, M. S. (1989). A causal model of consumer involvement.
Journal of Economic Psychology, 10,363-389.
Nisbett, R. E., & Wilson, T. (1977). Telling more than we can know: Verbal
reports on mental processes. Psychological Review, 84,23 1-259.
Recker, W. W., & Golob, T. F. (1976). An attitudinal modal choice model.
Transportation Research, 10,299-3 10.
Reibstein, D. J., Lovelock, C. H., & Dobson, R. de P. (1980). The direction of
causality between perceptions, affect, and behavior: An application to
300 VERPLANKEN ET AL.