Académique Documents
Professionnel Documents
Culture Documents
The prediction of dishonest actions was studied in the context of the theory of
planned behavior. College students completed a questionnaire that assessed at-
titudes, subjective norms, perceptions of behavioral control, intentions, and per-
ceptions of moral obligations, as well as self-reports of behavior with respect to
cheating on a test, shoplifting, and lying to get out of assignments. A subsample
of respondents returned several months later for a second administration of the
questionnaire. Multiple regression analyses showed that the theory of planned
behavior predicted intentions with a high degree of accuracy, and that it was
moderately successful in the prediction of actual behavior. Addition of perceived
moral obligations to the prediction equation improved prediction of reported lying
behavior, but did not help to account for much variance in cheating and shoplifting,
Self-reports of past dishonesty were used to evaluate the sufficiency of the theory
of planned behavior. 0 1991 Academic Press, Inc.
This article is based on the first author’s master’s thesis that was submitted to the University
of Massachusetts. We thank Robert Feldman and Ervin Staub for their input. Address
correspondence and reprint requests to Lisa Beck, Department of Psychology, University
of Massachusetts, Amherst, MA 01003-0034.
285
0092-6566/91 $3.00
Copyright Q 1991 by Academic Ress, Inc.
All rights of reproduction in any form reserved.
286 BECK AND AJZEN
skills, cooperation of others, etc.; see Ajzen, 1985, for a review). Col-
lectively, these factors represent people’s actual control over the behavior.
To the extent that people have the required opportunities and resources,
and intend to perform the behavior, they should succeed in doing so.
The theory of planned behavior, however, deals with perceived, rather
than actual, behavior control. In many situations perceived behavioral
control may not be particularly realistic. This is likely to be the case when
the individual has relatively little information about the behavior, when
requirements or available resources have changed, or when new and un-
familiar elements have entered into the situation. Under those conditions,
a measure of perceived behavioral control may add little to accuracy of
behavioral prediction. A direct path from perceived behavioral control
to behavior is therefore expected to emerge only when there is some
agreement between perceptions of control and the person’s actual control
over the behavior.’ A structural representation of this model is shown in
Fig. 1.
’ The theory of planned behavior also deals with the antecedents of attitudes, subjective
norms, and perceived behavioral control. These antecedents have to do with various beliefs
288 BECK AND AJZEN
For ease of presentation, the model described in Fig. 1 does not show
possible feedback effects of behavior on the antecedent variables. Note
that the predictors in the theory of planned behavior are assumed to be
sufficient to account for intentions and actions, but that they are not all
necessary in any given application. The relative importance of attitude,
subjective norm, and perceived behavioral control in the prediction of
intention, and the relative importance of intention and perceived behav-
ioral control in the prediction of behavior, are expected to vary across
behaviors and populations. Thus, in some applications it may be found
that only attitudes have a significant impact on intentions, in others that
attitudes and perceived behavioral control are sufficient to account for
intentions, and in still others that all three predictors make independent
contributions. Similarly, to predict behavior it may sometimes be sufficient
to consider only intentions while in other instances intentions as well as
perceptions of behavioral control may be needed.
The original derivation of the theory of planned behavior (Ajzen, 1985)
differed in two major respects from the present model. First, it defined
attitude, subjective norm, perception of control, and intention in terms
of trying to perform a given behavior rather than in relation to actual
performance. However, early work with the model showed strong cor-
relations between measures of the model’s variables that asked about
trying to perform a given behavior and measures that dealt with actual
performance of the behavior (Schifter & Ajzen, 1985; Ajzen & Madden,
1986). Since the latter measures are less cumbersome, they have been
used in subsequent research, and the variables are now defined more
simply in relation to behavioral performance.
Second, the original formulation of the theory postulated interactions
between perceived behavioral control and intention, and between per-
ceived behavioral control and attitude. Research conduced to date, how-
ever, has revealed only main effects of intentions, attitudes, and perceived
behavioral control. This has led to the reformulation of the model shown
in Fig. 1.’
Empirical research over the past 15 years has provided evidence in
support of the theory of reasoned action in a variety of experimental and
naturalistic settings (see, e.g., Ajzen & Fishbein, 1980; Ajzen, Timko, &
White, 1982; Bentler & Speckart, 1979; 1981; Fredricks & Dossett, 1983;
Manstead, Proffitt, & Smart, 1983; Smetana & Adler, 1980). The be-
haviors involved have ranged from very simple strategy choices in labo-
about the behavior that constitute the informational foundation on which intentions and
actions are assumed to rest. Although also assessed in the present investigation, these beliefs
are of no interest for the purposes of this article.
’ The results of the present study again revealed no significant interaction effects.
PREDICTING DISHONEST ACTIONS 289
3 Some investigators (e.g., Bentler & Speckart, 1979; Fredricks & Dossett, 1983) have
suggested that past behavior be included as a predictor of later behavior, equivalent to the
other independent variables in the model. It must be realized, however, that although past
behavior may well reflect the impact of factors that influence later behavior, it can usually
not be considered a causal factor in its own right (see Ajzen, 1987).
PREDICTING DISHONEST ACTIONS 291
PILOT STUDY
The design for the main study required that the experimenter be able
to contact participants for a follow-up questionnaire and match each par-
ticipant’s responses from both sessions. The need to provide identifying
information for the follow-up may appear to participants to be a threat
to anonymity, and therefore may interfere with honest responding. Sixty
participants in the pilot study completed a questionnaire that assessed
how many times in the past 12 months they had cheated on a test or
exam, taken something from a store without paying, and used a false
excuse to get out of taking a test or turning in an assignment on time.4
The pilot study provided information on the participants’ willingness
to admit to the three behaviors under two conditions of threat to ano-
nymity. Each participant in the pilot study was asked to write his or her
name, address and phone number on a form attached to the questionnaire
indicating willingness to participate in a continuation of the study within
the next few weeks. One group of participants was asked for name,
address, and phone number on the initial consent form, and the other
group found a separate page at the end of the questionnaire explaining
the need for a second session and requesting name, address, and phone
number.
If participants were concerned about the lack of anonymity, there should
be fewer admissions to the behaviors among participants who gave iden-
tifying information before filling out the questionnaire. A 2 (item order)
by 2 (identification condition) by 3 (behavior) analysis of variance per-
4 The pilot questionnaire also elicited salient behavioral, normative, and control beliefs
with respect to each of the three behaviors. However, this aspect of the study is of no
relevance for our present purposes and will not be further considered.
292 BECK AND AJZEN
MAIN STUDY
Methods
Participants. A total of 146 college students (28 males, 118 females) were recruited through
psychology courses to participate in a questionnaire study near the end of the fall semester.
The participants were between the ages of 17 and 30 (M = 19.6, 4 participants did not
report age).
During the following (spring) semester, an attempt was made to recontact all participants
who had completed the first questionnaire and had indicated that they would be willing to
take part in a follow-up study. Some participants had moved and could not be located while
many of those contacted were not currently enrolled in psychology classes and declined the
invitation to return for the follow-up. Of the 146 participants who had completed the first
questionnaire, 34 (23%) returned for the second phase. An additional 46 participants who
had not completed the first questionnaire were also recruited.
Phase 1
The first questionnaire assessed the components of the theory of planned behavior (in-
tentions, attitudes, subjective norms, perceived behavioral control) as well as moral obli-
gation with regard to each behavior. In addition, participants were asked how many times
they had performed each of the behaviors in the previous 12 months.
The three behaviors of interest were defined as follows: cheafirrg, cheating on a test or
exam, shoplifting, taking something from a store without paying, and lying, using a false
excuse to get out of taking a test or turning in an assignment on time. Behavioral self-
reports were obtained by asking participants to indicate how often they had performed each
behavior by checking one of seven response alternatives; never, once, twice, 3 times, 4 times,
5 times, more than 5 times. The average correlation among the three behaviors was SO.
The remaining items used a seven-point semantic differential-type response format. They
are illustrated here with respect to cheating; the same questions were asked with respect
to shoplifting and lying. A measure of attitude toward each behavior was obtained by means
of the following five evaluative semantic differential scales: good-bad, plearant-unpleasant,
foolish-wise, useful-useless, unattractive-uttrucfive. The average interitem correlations were
SO for cheating, .57 for shoplifiting, and .53 for lying.
Three items were used to assess subjective norms with respect to each behavior: (1) “If
I cheated on a test or exam, most of the people who are important to me would not care-
disupprove.” (2) “No one who is important to me thinks it is OK to cheat on a test or
PBEDICIING DISHONEST ACTIONS 293
exam. ugree-disagree.” (3) Most people who are important to me will look down on me if
I cheat on a test or exam. likely-unlikely.” Average interitem correlations were .61, .47,
and .62 for cheating, shoplifting, and lying, respectively.
Perceived behavioral control with respect to each behavior was measured by four items:
(1) “For me to cheat on a test or exam is eusy-di#icult.” (2) “If I want to, I can cheat on
a test or exam. true-false.” (3) “1 can imagine times when I might cheat on a test or exam
even if I hadn’t planned to. likely-unlikely.” (4) “Even if I had a good reason, I could not
bring myself to cheat on a test or exam. likely-unlikely.” Average interitem correlations
for the three behaviors were 34, .48, and 34.
The measure of intention was based on responses to the following three items: (1) “If I
had the opportunity, I would cheat on a test or exam. likely-unlikely.” (2) “I would never
cheat on a test or exam. true-false.” (3) “I may cheat on a test or exam in the future. true-
false.” The interitem correlations for intentions were .70, .75, and .66.
Finally, moral obligation was also measured by three items: (1) “I would not feel guilty
if I cheated on a test or exam. true-false.” (2) “Cheating on a test or exam goes against
my principles. likely-unlikely.” (3) “It would be morally wrong for me to cheat on a test
or exam. likely-unlikely.” Interitem correlations were .45 for cheating, 54 for shoplifting,
and 55 for lying.
All items for a particular behavior were presented together, and the order in which the
behaviors were presented was counterbalanced. Within each behavior section, items assessing
the same variable were separated from each other except for the five evaluative attitude
scales5 The behavioral self-report item was presented before the other items for half of the
participants and after the other items for the other half, although the pilot study indicated
no differences in rate of admissions to the behaviors with regard to placement of the self-
report item. For all participants, the last part of the questionnaire contained the Marlowe-
Crowne Social Desirability Scale (SDS; Crowne & Marlowe, 1964).6
Participants completed the questionnaires in groups of 20 to 50, near the end of the fall
semester. Before filling out the questionnaire, each participant completed a consent form
which was coded with a three-digit number. The consent form, which was kept separate
from the questionnaire, also requested the participant’s address and phone number as an
indication of interest in participating in a continuation of the study which was to be conducted
in the spring. Each participant returned the consent form before beginning the questionnaire,
which was coded with the same three-digit number as the consent form. Participants were
instructed not to put their names on the questionnaire itself, and they were assured that
their responses would be kept anonymous.
Phase 2
The 34 returning respondents as well as 46 new control participants completed a ques-
tionnaire in Phase 2. The follow-up questionnaire consisted of the same measures of in-
tention, attitude, subjective norm, perceived behavioral control, and moral obligation as
on the initial questionnaire. The behavioral self-report items assessed behaviors performed
“in the past six months” in order to limit the participants’ reports to performance since the
initial questionnaire.
As in the fall questionnaire, item order was counterbalanced with regard to presentation
’ In addition, the measures of intention, attitude, subjective norm, and perceived control
were interspersed among sets of items assessing specific beliefs about each behavior (see
Footnote 3).
’ Examination of means and standard deviations revealed no apparent ceiling or floor
effects nor evidence for skewness in responses to the various items.
294 BECK AND AIZEN
TABLE 1
RELIABILITIES AND CORRELATIONS AMONG VARIABLES
I A SN PBC MO
Cheating
Intention (I) 37
Attitude (A) .67 .81
Subjective norm (SN) .34 46 .81
Perceived behavioral control (PBC) .79 .65 .37 .66
Moral obligation (MO) .69 .64 .50 .66 .n
Self-report of behavior .69 .53 .20 .61 .52
Shoplifting
Intention 38
Attitude .78 .84
Subjective norm .38 .50 .73
Perceived behavioral control .79 .76 .39 .78
Moral obligation .75 .71 .42 .61 .78
Self-report of behavior .74 ho .25 .67 .52
Lying
Intention .&Is
Attitude .53 .85
Subjective norm .37 .39 .a3
Perceived control .75 .52 .35 .67
Moral obligation .74 .67 .42 .64 39
Self-report of behavior .56 .33 .15 .48 .42
of the three behaviors, and the behavioral self-report item was located at the beginning of
each behavior section for half of the participants and at the end of each section for the
other half. Precautions for anonymity were again carried out, with returning participants
receiving questionnaires bearing the same three-digit number as on their first questionnaire,
and control subjects receiving questionnaires coded with a different series of digits. Partic-
ipants’ names did not appear on the questionnaires. At the end, participants in the control
group completed the Marlowe-Crowne Social Desirability Scale, but returning participants
did not complete the SDS a second time.
TABLE 2
HIERARCHICALREGREWON ANALYSES FOR INTENTIONS
r b Rr b Rr b R
TABLE 3
HERARCHICAL REGRESSION ANALYSES M)R BEHAVIOR
and 47%, in Phase 2, the corresponding percentages were 47, 29, and
21% in the same group and 48, 20, and 40% in the control group. This
can be explained by the fact that the Phase 2 questionnaire only asked
for self-reports of behaviors in the past 6 months, whereas the original
questionnaire had asked for behavior in the past year.
Overall, then, repeated questionnaire administration did not seem to
have a systematic impact on responses. This conclusion is confirmed by
the relatively high test-retest correlations over the 6 months of the study.
The test-retest reliability coefficients ranged from .53 for subjective norms
concerning lying to .83 for cheating intentions; the average test-retest
reliability across all measures was .71. Finally, hierarchical regressions for
the prediction of intentions in the subsample revealed results comparable
to those obtained in the total sample (Table 2).
Hierarchical regressions. Table 3 presents the results of hierarchical
regressions of behavior, as reported in Phase 2, on Phase 1 intentions
and perceived behavioral control, the predictors in the theory of planned
behavior, as well as on Phase 1 perceived moral obligation and past
behavior (self-reports of behavior in Phase 1). It can be seen that the
theory of planned behavior was moderately successful in predicting self-
reports of actual behavior; the multiple correlations ranged from .35 for
lying to .74 for cheating. Addition of perceived moral obligation raised
the multiple correlation appreciably (from .35 to .62) only in the case of
lying, while the further addition of past behavior improved prediction for
PREDICTING DISHONEST ACTIONS 299
lying as well as shoplifting, These results suggest that the theory of planned
behavior was not totally sufficient to account for performance of dishonest
activities. Although the variables specified in the theory did predict the
three behaviors quite well, and they accounted for most of the systematic
variance in cheating behavior, in the case of shoplifiting and lying they
left an appreciable proportion of the systematic variance unaccounted for.
CONCLUSIONS
Like much research in social and personality psychology, work with the
theory of planned behavior relies on self-reports to assess the theory’s
major constructs. This can involve problems of social desirability respond-
ing in the case of socially undesirable behaviors. The data of the present
study suggest, however, that there seemed to be no major systematic
distortions in responses concerning cheating, shoplifting, and lying. Par-
ticipants admitted to these behaviors in relatively large numbers, and
there was little evidence for strong effects due to measured social desir-
ability tendencies.
In addition to deliberately denying socially undesirable attitudes or
behaviors, people may also be unaware of socially undesirable tendencies
and thus provide inaccurate self-reports. It was noted in the introduction
that, according to the theory of planned behavior, inaccurate perceptions
of control will tend to reduce the ability of the model to predict behavior,
although it should not affect the model’s ability to predict intentions. The
results of the present study are consistent with these considerations. As
expected, prediction of intentions was quite accurate, whereas a consid-
erable proportion of systematic variance in behavior remained unac-
counted for. Moreover, the model was more successful in predicting cheat-
ing than it was in predicting shoplifting or lying. This finding can also be
explained in terms of inaccuracies in the predictor variables. One would
expect that experience will increase the accuracy of perceived behavioral
control and that, therefore, prediction will be better for behaviors with
which respondents have had more experience. In fact, more participants
in the present study had cheated in the preceding 6 months (47%) than
had shoplifted (29%) or lied to get out of class assignments (21%). It
thus appears that lack of experience and lack of insight into one’s own
motivations and capabilities may result in reports of beliefs, attitudes, and
intentions that are not sufficiently reflective of a person’s true dispositions
to permit accurate prediction of later behavior. It is also possible that
shoplifting and lying occur more impulsively than does cheating.
The role of prior behavior is also of interest in this context. We have
argued that prior behavior has no direct effect on later behavior, that
people do not behave in a certain way today because they behaved that
way in the past. Rather, prior and later behavior are expected to correlate
300 BECK AND AJZEN
well with each other to the extent that the underlying determinants of
the behavior have not changed. The relation between prior and later
behavior is thus assumed to be mediated by the underlying determinants.
The results of the present study showed that there is considerable con-
sistency in dishonest behaviors over time, and that this consistency is not
fully reflected in the constructs of the theory of planned behavior. The
above discussion suggests that inaccuracies and misperceptions in self-
reports may provide an explanation for the failure of the theory’s con-
structs to mediate the relation between prior and later behavior. It is of
course also possible that dishonest behaviors are influenced by factors not
included in the model and not assessed in the present study. One additional
factor that was considered, perceived moral obligation, did aid in the
prediction of lying, but it had less of an effect on cheating or shoplifting,
and even with its inclusion in the prediction equation, systematic variance
in behavior was left unaccounted for.
Overall, however, the results of this investigation provide considerable
support for the theory of planned behavior. Intentions with respect to
cheating, shoplifting, and lying were strongly related to attitudes, sub-
jective norms, and perceptions of behavioral control, and self-reports of
behavior could be predicted significantly from intentions and perceived
behavioral control. At the same time, it appears that understanding the
determinants of dishonest behaviors can be more problematic than un-
derstanding performance of socially acceptable behaviors. Further re-
search is needed to identify factors not included in the theory of planned
behavior that may influence peformance of such dishonest actions as cheat-
ing, shoplifting, and lying.
REFERENCES
Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In J. Kuhl and
J. Beckmann (Eds.), Action-control: From cognition to behavior (pp. 11-39). Heidel-
berg: Springer.
Ajzen, I. (1987). Attitudes, traits, and actions: Dispositional prediction of behavior in
personality and social psychology. In L. Berkowitz (Ed.), Advances in experimental
social psychology (Vol. 20, pp. l-63). New York: Academic Press.
Ajzen, I. (1988). Attifudes, personality, and behavior. Chicago: Dorsey Press.
Ajzen, I. (in press). The theory of planned behavior. Organizationul Behnvior and Human
Decision Processes.
Ajzen, I., & Fishbein, M. (1977). Attitude-behavior relations: A theoretical analysis and
review of empirical research. Psychological Bulletin 84, 888-918.
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, inten-
tions, and perceived behavioral control. Journal of Experimental Social Psychology 22,
453-474.
Ajzen, I., Timko, C., & White, J. B. (1982). Self-monitoring and the attitude-behavior
relation. Journal of Personaliv and Social Psychology, 42, 426-435.
PREDICTING DISHONEST ACTIONS 301