Content uploaded by Sebastian Sattler
All content in this area was uploaded by Sebastian Sattler on Oct 08, 2015
Content may be subject to copyright.
This article has been published in: "Deviant Behavior"
To access the published version, please use the following link:
Please only cite the version published in "Deviant Behavior" and use the
Sattler, Sebastian; Graeff, Peter; Willen, Sebastian (2013):
Explaining the Decision to Plagiarize: An Empirical Test of the
Interplay Between Rationality, Norms, and Opportunity. Deviant
Behavior 34(6): 444-463. (doi:10.1080/01639625.2012.735909).
Explaining the Decision to Plagiarize: An Empirical
Test of the Interplay Between Rationality, Norms,
SEBASTIAN SATTLER* Bielefeld University, Bielefeld, Germany
PETER GRAEFF, Bundeswehr University, Munich, Germany
SEBASTIAN WILLEN, University of Duisburg-Essen, Essen, Germany
* Phone: ++49 (0)521 1063080, email@example.com
Plagiarism is a fraudulent behavior that infringes upon the rules of universities and intellectual
property rights. Rational choice theory provides a theoretical framework for explaining this
deviant behavior. Our study increases knowledge of the little analyzed interrelation among the
individual determinants of deviant behavior. We use panel data from a large-scale random
sample of university students (N=2,806). The expected utility of plagiarism, internalized social
norms, and opportunities to plagiarize can explain the frequency of plagiarism. The significant
interaction between utility and opportunity could be interpreted as a form of temptation.
Promising strategies to reduce plagiarism are also discussed.
Plagiarism is a major problem in academic work. It is a type of fraud and a socially undesirable
behavior. By ignoring the requirement for proper acknowledgment of sources, plagiarists not
only violate copyrights and disobey academic rules, they also infringe upon general social norms
(for example, “Thou shalt not steal.”).
Moreover, plagiarism raises numerous practical consequences and ethical challenges.
Students in particular engage in this behavior in order to save time and obtain higher grades by
unfair means. This may create an incentive for non-plagiarizing fellow students to cheat as well.
But plagiarists are also confronted with the risk of detection and punishment as well as being
exposed to opportunity costs, since plagiarism does not improve their academic skills.
. Transactional costs are incurred as instructors try to protect themselves against the violation of trust by their
students and students try to protect themselves against the suspicion of their instructors. Educational
institutions need to spend time and financial resources to stop the occurrence of plagiarism. The teaching and
learning atmosphere at universities is injured by the instructors’ constant suspicion (Weinstein and Dobkin
addition, if the public were to discover how much plagiarism takes place at universities, the
reputation of academic work in general could be damaged as well as the reputation of the
university and its instructors and graduates (Martinson et al. 2005).
Several studies have attempted to determine the rate of plagiarism, such as the experimental
study by Weinstein and Dobkin (2002), who found that every sixth paper submitted at U.S.
colleges contained plagiarism. Self-report studies have revealed different plagiarism frequencies
(review paper by Whitley 1998); a study by Cochran et al. (1999) reported that 19% of students
stated they had plagiarized at least once during the last twelve months. The lifetime prevalence
rate measured among graduates was about twice as high (Alam 2004). Since the advent of
ubiquitous access to the Internet, the number of observed and reported plagiarism cases has
sharply increased, while the efforts required to plagiarize have dropped dramatically (McCabe
2005). This as well as current debate about incidences of plagiarism by politicians (e.g. the
former vice president of the European Parliament, the Iranian science minister, the former
Austrian minister of science and education, the former Hungarian President, and the former
minister of defense in Germany) and scientists highlights the need for research, but until now
large-scale panel studies have been lacking.
To develop countermeasures against plagiarism, it is necessary to generate and test a
theoretical model of the occurrence of plagiarism. But a review of the literature reveals that the
number of plagiarism studies without any explicit theoretical framework (e.g. Alam 2004)
exceeds the number of theoretically-grounded studies (e g. Bunn et al. 1992). In order for the
theoretical propositions to correspond adequately to the empirical research, this paper draws
upon the framework of rational choice theory (RCT). Because RCT has the capacity to explain
2002). Moreover, when they later seek employment, students who ultimately received their university degree
on the basis of plagiarized papers appear on paper to possess greater skills and efficiency than they actually do
decisions in general, it exhibits a high predictive power in several research fields (Coleman
1988; Diekmann and Preisendörfer 2003; Iannaccone 1995 etc.). Since plagiarism presupposes
the selection of material for plagiarizing, it can be considered an intentional and deliberate, either
verbatim or paraphrased, adoption of ideas.
To decide between plagiarism and legitimate
practices, we assume, consistent with RCT, that students consider the specific features of
plagiarism. These features can be expressed in terms of costs and benefits (Cochran et al. 1999;
Kerkvliet and Sigmund 1999; Teixeira and Rocha 2008).
We scrutinize the predictive power of RCT with respect to plagiarism, deriving testable
hypotheses from the general theory of rational choice as well as related theories of deviant
behavior in order to answer the research question of why plagiarism occurs. In doing so, this
paper contributes to the ongoing debate about “which theory of action should be used”
(Kroneberg et al. 2010) by presenting an extended RCT and by responding to the need to test
competing versions empirically and systematically, and, especially, to broaden the analysis of
interactions between the predictors utility, norms, and opportunity. For such an analysis, often
only scarce longitudinal data are available. Thus, this paper is the first large-scale panel study to
deliver causal explanations and predict the individual frequency of plagiarism within a period of
six months (one term); it thus goes beyond both explanations of criminal intentions as well as
approaches that address the question under what circumstances plagiarism occurs or not. It thus
avoids ad-hoc hypotheses and examines the suitability of existing rational choice models to
The paper starts with a review of the RCT literature and the corresponding studies on
plagiarism, followed by a description of the research design and the results of our study.
. The unconscious appropriation of ideas and sentences, so-called cryptomnesia (Taylor 1965), will not be
addressed in this paper.
Estimates for different variants of RCT models are then provided, resulting in the replication and
enhancement of previous empirical results. Based on these results, we derive practical
implications for reducing plagiarism and discuss opportunities for future research.
MODELING THE DECISION-MAKING PROCESS
Many of studies using RCT do not explicitly specify the link between the general theoretical
framework and the specific research question. As a general theory, RCT posits that actors chose
the best alternative with regard to their preferences and restrictions. It also provides general
propositions about the parameters that actors consider when making decisions such as costs,
benefits, and probabilities (e.g. Voss and Abraham 2000; Elster 1985). Given these propositions,
a bundle of general hypotheses can be derived, such as that the increasing cost of a behavior
reduces its occurrence (ceteris paribus). These hypotheses are empty and empirically
unsubstantial (cf. Hechter 1994), because the theory does not imply these costs or any concrete
consequences of action in a specific situation (Opp 1998).
Therefore, a contextual specification
is needed. This will be achieved in our study by elaborating the explanatory models on the basis
of the much more specific criminological version of RCT. A “crime-specific focus” (Clarke and
Felson 1993) for explaining plagiarism is thus provided.
Adapting Rational Choice Theory and its Extensions for Research
on Deviant Behavior
. As a consequence, this theoretical approach is applicable to a variety of decisions, which could be considered a
strength of this theory.
The basic principles and implications of RCT become more precise and understandable in
conjunction with ideas about deviant behavior. As mentioned above, plagiarism can be seen as a
kind of fraud or deviant behavior (Collins et al. 2007). So approaches to explain deviant behavior
provide a suitable explanatory framework (see Bunn et al. 1992; Cochran et al. 1999). Offences
can be seen as courses of action selected under uncertainty (Becker 1968, cf. Eide 1997; 2000;
Hastie and Dawes 2010) and analyzed using an economic approach by linking them with costs
and benefits. Actors have expectations concerning the costs and benefits of both legal and illegal
alternatives. Those expectations are based on available information, meaning that subjective
assessments are usually linked to objective opportunities (Ehrlich 1996; Fattah 1993). From a
criminological perspective, RCT would imply that offences can be considered under the
assumption of maximizing the utility of the actors.
The pioneering work of transferring the ideas of RCT to social science and to the explanation
of criminal actions in particular was done by Gary Becker (1968). He provided a parsimonious
model, also called the strict RCT version, to explain criminal choices: actors choose a criminal
action if the expected benefits of an offence (B) exceed the product of the expected level of a
penalty (C) and the expected probability of being caught (p). Becker (1968, p. 177) states that an
increase in either the expected probability (p) or formal costs (C) “… would reduce the utility
expected from an offence and thus would tend to reduce the number of offences because either
the probability of ‘paying’ the higher ‘price’ or the ‘price’ itself would increase” (cf. Ehrlich
1996; Piliavin et al. 1986).
. “The assumption of rationality does not imply that muggers (or economic professors) calculate costs and
benefits of available alternatives to seventeen decimal places – merely that they tend to choose the one that best
achieves their objectives.” (Friedman 1995, p. 43). For criticism see De Mesquita and Cohen (1995).
. Becker and Mehlkop (2006; cf. Nagin and Paternoster, 1993) found support for Becker’s theory in their
empirical study on shoplifting and tax evasion. Eide (2000) provides empirical evidence (which is however not
coherent) for the model in a variety of different surveys (see e.g. Piliavin et al. 1986). Kroneberg et al. (2010)
list several studies that approve RCT empirically, but also many which show mixed results.
The Becker model has been extended several times in the attempt to overcome some of its
limitations. One crucial extension refers to an actor’s assessment of his or her ability to carry out
the offence, referred to as the probability of success (Becker and Mehlkop 2006). In Becker’s
economic view (1968) the probability of success is the opposite of the probability of being
caught (1-p). It has been argued, however, that actors have expectations about the probability of
success that are independent of their expectations about the probability of getting caught (for
instance, Becker and Mehlkop 2006). In this understanding, the probability of success (q) is
associated with knowledge about accomplishing the offense and controlling expectations as well
as control beliefs (Mehlkop and Graeff 2010) and also corresponds to the ideas by Clarke and
Felson (1993) on the modeling of personal experiences. These four variables reflect the expected
utility (U) of conducting a deviant activity by subtracting the product of p * C from the product
of q * B (e.g. Tittle et al. 2010;; Mehlkop and Graeff 2010).
Another extension takes account of the social conditions of decision making by considering
norms (N). Because in any given situation norms have an impact on social behavior, they
encapsulate individual learning processes manifested during a decision-making process (Tittle et
al. 2010; Williams 1968). Norms and morality have been found to be crucial, strong, and robust
determinants for the explanation of (criminal) actions or intentions as well (Beck and Ajzen
1991; Cochran et al. 1999; Grasmick and Bursick 1990; Kroneberg et al. 2010; Nagin and
Paternoster 1993; Wikström and Svensson 2010). Classic economic models of deviant behavior
do not take them into consideration and assume that they play a minor role (cf. Becker 1976), but
ignoring them is a “major shortcoming” of these approaches (Kroneberg et al. 2010). Tittle et al.
(2010) emphasize that in criminological research norms have been emphasized less than other
variables. In the case of more recent developments of RCT, Opp (2012) has pointed out that
there are at least two positions concerning the influence of norms during a decision-making
process. The first camp of RCT adherents dismisses the role of norms in rational decision-
making processes (March and Olson 1989). The implicit assumption of these theorists is that
norms are features of social life that must be considered “irrational,” just like similar phenomena
such as trust (Williamson 1993). The associated behavior is not “rational,” because norms are
subject to dynamics other than utility, which is inherently linked to the principle of optimization
A different position is adopted by RCT theorists who assume that behaving in accordance
with or in opposition to a norm can indeed be considered from the theoretical perspective of
rational decision-making (Coleman 1990; Opp 2012). Choosing one action among several others
when a norm is involved follows the principle of utility maximization (Coleman 1990; Elster
1989; Tittle et al. 2010). Those acting in accordance with norms may experience intrinsic
rewards, while those opposing them suffer a bad conscience or feelings of guilt: so-called
internal negative sanctions or psychological costs (e.g. Chochran et al. 1999; Posner and
These costs arise because of personal ethics and preferences about the (non-)
violation of a norm (Coleman 1988; Posner and Rasmusen 1999) and incorporate non-tangible
factors into an extended version of RCT (Opp 2012; Tittle et al. 2010). Norms are learned but
actors may vary in their responsiveness to norms due to different individual learning histories or
even biogenetic characteristics (Tresler 1993).
From a theoretical perspective, the inclusion of more situational features can be seen as
valuable (Kroneberg et al. 2010). The most obvious feature is the existence of opportunity to
behave in a deviant manner. RCT-related theories such as the routine activity approach (Clarke
. Even the other RCT camp can interpret norms as part of a person’s moral system, which precedes the actual
decision-making process (Becker and Mehlkop 2006; Lindenberg 1983; Opp 2012).
and Felson 1993) also consider this feature by assuming that crime is also a function of the
opportunity to engage in criminal acts. In order to explain someone’s decision to commit a crime
within a given period of time (within a given set of internalized norms, preferences, abilities, and
evaluations), opportunity must be taken into account. Opportunities for deviant behavior have
been discussed a great deal in criminological literature about crime prevention (Clarke and
Felson 1993; Cornish and Clarke 2003; Fattah 1993; Wortley 1998). It has been assumed that
individuals use an opportunity if they are offered a suitable one (Bennett and Wright 1984), as
expressed by the saying “opportunity makes the thief.” So, in the words of crime opportunity
theory, crime is also a function of the number of potential targets (Fattah 1993).
A formal model that takes the previous arguments and concepts into account needs to include
expected utility designated by U, norms by N, and opportunities by O. Model 1 is capable of
serving as a starting point for evaluating the additional single explanation contributions by norms
and opportunities (Mehlkop and Graeff 2010).
Prob[Deviant behavior] = U – N + O (1)
Usually, in predicting a criminal decision it is assumed that utility and norm contribute
independently. But it is conceivable that these elements mutually influence one another during
the decision-making process. The connection that can be assumed to exist between norms and
utility is taken into account in equation 2. The degree to which norms are internalized reflects the
degree to which individuals are influenced by the general rules of a certain society, e.g. the
proscription of deviant behavior. There is evidence that the specific level of internalization
influences the effects of the assessment of instrumental incentives (Kroneberg et. al 2010;
Mehlkop and Graeff 2010; Tittle et al. 2010). The underlying assumption of this effect is not
taken into account by a strict RCT approach and its extended version (as mentioned above) only
considers the violation of norms in terms of internal negative sanctions or psychological cost.
Internalized norms (which proscribe the deviant behavior) would counteract a positive net utility
from carrying out the deviant behavior because they reflect benefits for deliberately following
the rules about breaking them. Such assumptions about decisional options (such as filtering, the
framing of alternative actions, the downplaying or dismissing of potentially positive aspects of a
certain action, and the removal of specific sorts of behavior from one’s agenda) are described by
Becker and Mehlkop (2006; Kroneberg et al. 2010; Mehlkop and Graeff 2010; Trasler 1993;
Tittle et al. 2010). These assumptions imply that committing a crime is not an option – or at least
a less valuable option – for people with deeply internalized norms, leading to reduced utility
(Kroneberg et al. 2010). Therefore, the interaction between norm and utility must be considered
to be negative:
Prob[Deviant behavior] = U - N - U * N + O (2)
Applied to our topic, the literature about crime opportunities and rationality (e.g. Clarke and
Felson 1993; Wortley 1998) implies that the utility of an actor is related to his opportunities. A
crime becomes more likely if an actor expects a high utility outcome when an opportunity exists.
In turn, more opportunities give rise to realizing utility and could increase its importance,
accordingly. If few opportunities exist, especially those actors with high utility will be more
likely to take advantage of one, while those with lower utility will more cautiously select one
opportunity among many. Thus, as mentioned above, the influence of utility upon the probability
to exhibit a certain deviant behavior may vary when opportunities change (and vice versa).
Applied to RCT, this influence could be interpreted as some sort of temptation if the utility of a
certain action varies with the occurrence of opportunities (Hastie and Dawes 2010; cf. Gul and
Pesendorfer 2001). Here, temptation refers to an increase in the tendency to show deviant
behavior if the utility is increased by adding or seeking out more opportunities to realize this
behavior (cf. Fattah 1993). Temptations can suspend long-term preferences; whereas long-term
decisions would be more rational, immediate temptations can result in visceral and opportunistic
reactions (Trasler 1993). An equivalent model for this suggestion is:
Prob[Deviant behavior] = U - N + O + O * U (3)
Our final model describing the decision-making process of plagiarizing incorporates both
proposed interaction terms:
Prob[Deviant behavior] = U - N - U * N + O + O * U (4)
Explaining Plagiarism Using an Extended Rational Choice Theory
This section will apply these models’ predictions to the issue of plagiarism, beginning with an
overview of the most important empirical results concerning plagiarism.
Some authors (Cochran
et al. 1999; Kerkvliet and Sigmund 1999; Teixeira and Rocha 2008) provide theoretical
grounding for the presumption that a decision for or against plagiarism is connected to a process
. As far as we know, there is as yet no empirical test of the interaction between norms and opportunities. Since
our empirical testing is derived directly from theoretical propositions and because there were no propositions
available for this interaction, we refrained from including this term in our models. In an explorative fashion, we
tested whether there is a considerable statistical relation between opportunity and norms, but no model even
exhibited a significant interaction between them. In order to apply parsimonious modelling, we did not include
this insignificant interaction term. For the sake of saving space, we did not report upon the models here. Nor is
there a compelling theoretical argument for including a three-way interaction between utility, norms, and
opportunity. For explorative reasons, however, we did test such a three-way interaction. It does not become
significant in our models and its results are also left unreported due to space limitations.
. Also due to spatial limitation, we focus on empirical findings that could fit to the theoretical propositions of the
model presented above. Plagiarism has been a research topic in several scientific disciplines, e.g. in
Psychology. Empirical research in this area focuses on procrastination or motivation (Jordan 2001; Rettinger et
al. 2004; Roig and DeTommaso 1995).
of weighing costs and benefits. The authors assume that the expectation of punishment along
with a subjectively presumed probability of detection can be regarded as expected costs (see also
Tackett et al. 2010). The faculty’s rejection of a paper, or the student’s failing a course or even
being suspended from the university usually count as punishments. Teixeira and Rocha (2006)
provide evidence for the deterrent effect of sanctions, while Cochran et al. (1999; see also Bunn
et al. 1992) have found, in contrast, that formal sanctions often do not significantly affect the
plagiarism activity. The authors posit that the probability of formal sanction is too low to prevent
One reason for student´s belief in a low probability of detection is that they
overestimate the faculty’s costs of detection (Weinstein and Dobkin 2002). Additionally, they are
perhaps unaware of the university’s usage of software in order to detect plagiarism, as were two-
thirds of the students questioned in a study by Crisp (2004). This ignorance could indeed lead to
underestimating the probability of detection. In contrast to this suggestion is the finding by Bunn
et al. (1992), which revealed that 50% of the respondents had observed plagiarists being caught
red-handed. Michaels and Miethe (1989) found that the amount of formal punishment weighted
by its probability of occurrence had a deterrent effect on cheating behavior in the past (on papers
and homework). This effect could not be replicated in a study by Sattler (2007). Other studies
have reported results about the benefits of plagiarizing. One advantage of plagiarism is the
potential improvement in the student’s academic situation (e.g. better grades). Increasing
competition among students (due to the growing number of students) as well as high
expectations on the part of parents creates pressure to achieve good grades (McCabe 2001).
Michaels and Miethe (1989) have found that students expect that fraud will result in better
grades. Plagiarists also cited time saving as an expected profit (Alam 2004). Referring to the
. Possibly, the probability does not cross a threshold necessary to unfold effects. Perhaps students do not get
enough information about the whole subject of plagiarism.
theoretical model, the higher the utility of plagiarism (which is the difference between the
weighted expected benefits of plagiarism and its weighted expected costs), the higher the
individual frequency of plagiarism (Hypothesis 1).
Results on the assessment of social norms regarding plagiarism are contradictory. In a study
of American high school students (McCabe 2005), three of four students perceived cheating by
plagiarizing as not posing a serious problem. In a study of 363 British participants, five out of six
disapproved of plagiarism; in another study four of five described it as moderately serious fraud
(Duggan et al. 2004). It is interesting to note that plagiarism appears only slightly more
acceptable to plagiarists than to students who do not plagiarize (Alam 2004). Michaels and
Miethe (1989; Sattler 2007) found in their study, however, that students who disapprove more
strongly of cheating behavior are also less likely to plagiarize. This result could not be replicated
in the study by Bunn et al. (1992). Nor do Cochran et al. (1999) offer any evidence about the
effect of the degree of an individual’s moral condemnation of plagiarism but they do provide
some evidence that feelings of shame and embarrassment reduce the incidence of plagiarizing
behavior. These results give rise to the assumption that internalized norms reduce the individual
frequency of plagiarism (Hypothesis 2).
To our knowledge, the influence of varying opportunities has not yet been tested with regard
to plagiarism. But the assumption mentioned above can be applied here as well: the more
opportunities students have to plagiarize, the higher the individual frequency of plagiarism is
likely to be (Hypothesis 3).
Nor, as far as we know, has the interaction between norms and the utility of plagiarism yet
been tested. But, as mentioned above, it can be assumed that an internalization of norms
counteracts the illegitimate use of ideas or texts or leads to a decrease in utility, so the individual
frequency of plagiarism will be reduced (Hypothesis 4).
Moreover, it is assumed that there is a positive interaction between the utility of plagiarism
and the number of opportunities to commit it. This form of temptation increases the individual
frequency of plagiarism (Hypothesis 5). Figure 1 shows our theoretically assumed predictors.
[FIGURE 1 ABOUT HERE]
MEASURES AND DATA
The data for this analysis result from a research project funded by third-party resources (see
acknowledgements). The study investigated socially desired and undesired strategies in
achieving a university degree, study conditions, abilities etc. Three random sampling stages were
used in the project: first, four German universities were chosen, then the academic disciplines to
be included were decided upon; finally 11,000 students were sampled in a two-wave study. As
promised in a pre-notification letter and via e-mail invitation, the students took part in the study
voluntarily and anonymously (with monetary pre-paid incentives of 5 Euro ≈ 7 Dollar in t
). A special data confidentiality agreement was provided to all participants.
The data collection processes were observed by a data protection officer. About 53.5% (N =
5,882) of the students responded to our invitation at t
, and 69.1%
(N = 3,486) six months later
. These rates can be considered high for an online survey (Shih and Fan 2008). When all
. Participants could choose cash incentives, vouchers for an online store or donations for two well-known charity
organizations – UNICEF and Amnesty International (5 Euro for each option).
. Only participants who completed the questionnaire in t
got an invitation to t
variables with missing cases were removed, the sample comprised 2,806 students. To test the
impact of missing values on our estimates, we reproduced all models using the multiple
imputation procedure by chained equations (ice, Royston 2005). Results did not change
therefore listwise deletion was used (cf. Kroneberg et al. 2010).
Frequency of Plagiarism: The dependent variable is the frequency of plagiarism. It was
measured in t
, asking whether students consciously presented ideas or quotations from others as
their own work in academic assignments for their major within the last six months.
ranged from never (value 0) to more than 10 times (11). Almost every fifth (17.78%) student in
our sample reported plagiarism behavior at least once. Students reported 0.377 cases on average
(see Table 1). Using this panel measure avoids causal order problems that would occur if past
behavior had been the variable of interest and measured in t
as well (Grasmick and Bursik
Expected Utility (U): In order to operationalize a variable indicating the subjectively expected
utility (U) of the deviant behavior investigated here, we first computed the product terms for the
expected weighted costs (p * C) and the expected weighted benefits (q * B) of plagiarism. This
was done in accordance with the theoretical arguments given in the literature (Cochran et al.
1999; Tibbets 1997). The subjective expected utility of plagiarism (U = q * B - p * C) was then
. Results are available upon request.
. In many studies researchers ask about plagiarism directly (e. g. Cochran et al. 1999 and others). Using the
signalling word “plagiarism,” an underestimation of this variable is more likely because of social desirability.
We did not use this word and checked, additionally, for a social desirability bias by using the German short
scale of the Balanced Inventory of Desirable Responding (BIDR; Winkler et al. 2006, cf. Paulhus 1991). This
scale consists of two subscales: self-deceptive enhancement (SDE) and impression management (IM). There
was no significant or substantive positive effect of both scales. The coefficient of the variables did not change.
Due to space limitations, results are not reported here.
computed (Tittle et al. 2010; Becker and Mehlkop 2006). For the measurement of the
independent rational choice variables in t
(see Table 1), we assume that “... actors are able to
make judgments about utility and probabilities either in numerical quantity or at least in verbal
categories” (Friedrichs et al. 1993, p. 4, our translation). Thus, the participants were asked to
estimate these variables. The advantage of this method is that it is able to test the core of the
rational choice model directly. The assessment of the probability of success (q) is similar to self-
efficacy measures (Ajzen 2002). Students estimated their ability to obtain benefits by means of
plagiarizing a paper. We used two items
with a five-point scale labeled very unlikely, unlikely,
somewhat likely, likely, and very likely (cf. Michaels and Miethe 1989). The category very
unlikely got the value 0.166, followed by 0.333; 0.5; 0.666 to 0.833 for very likely (Becker and
Mehlkop 2006). The two-item composite measure has a Cronbach’s α of 0.71. Benefits of
plagiarism (B) were assessed with a nine-point scale ranging from profiting nothing (value 1) to
profiting very much (9) (cf. Tittle et al. 2010). We measured costs of plagiarism (C) in a similar
way. Students were asked to judge whether they would be faced with no punishment at all (1) to
very severe punishment (9) if their teachers discovered them plagiarizing (cf. McCabe et al.
2002). The probability of detection (p) was based on the question about the likelihood of
detection if ideas and quotes were used without acknowledgment. Students were directed to use
percentage values between 0% (no chance of getting caught) and 100% (certainty of getting
caught) (cf. Beuer-Krüssel and Krumpal 2009). Values were divided by 100 according to the
scale of probability of success.
. Items: “I know how to engage in this behaviour to my advantage.” and “It is not difficult for me to engage in
Norms (N): Students reported their personal normative evaluation of plagiarism using three
based upon a five-point scale from 1 strongly disagree to 5 strongly agree (Beck and
Ajzen 1991; Tibbets 1997). Responses were averaged to generate a norm index (Cronbach’s α =
Opportunity (O): The opportunity to plagiarize (O) or, in other words, the number of potential
targets, was measured by the number of academic papers a student wrote during the past six
An interaction term between the utility of plagiarism (U) and the subjective norm (N) was
calculated as well as an interaction term between U and O. As recommended by Aiken and West
(1991), we centered all variables before computing these terms, which resulted in variables with
a mean value of zero and a standard deviation of 1. This was done to avoid multicollinearity of
lower and higher order terms in the models.
To ensure the quality of measurements and applicability of our method, we conducted several
stages of cognitive and quantitative pre-tests in advance.
Because our dependent variable is a classical count outcome indicating how many times such
discrete events are reported, count regression models will be used for testing the contribution of
RCT in explaining the rates of occurrence of plagiarism. For this class of variables small values
occur much more often than large ones. In such cases, standard ordinary least-squares (OLS)
approaches could produce inefficient, inconsistent, and biased estimates (Long and Freese 2001).
In our sample, the distribution is highly skewed to the right (skewness = 5.407) and likelihood-
. Items: “It gives me a bad conscience.”; “It is against my moral beliefs.”; “For me, such behaviour is
ratio tests for all models (p < 0.001) provide evidence for overdispersion (Hilbe 2009), which is a
variance greater than the mean (Var = 1.346; x
= 0.377; min = 0; max = 11). Therefore, the
Poisson-based negative binomial regression models are an appropriate class of models here and
possess significant advantages over Poisson models that assume equidispersion (Long and Freese
2001). The latter do not take unobserved heterogeneity among observations into account and
have downward-biased standard errors. A negative binomial distribution comprises a Poisson
distribution for the expected mean of the event counts and a gamma distribution for variation
(MacDonald and Lattimore 2010). Here, a residual variance parameter α (sometimes called D²) is
added. Due to the theoretical model, several interaction terms were modeled (see Hilbe 2011).
[TABLE 1 ABOUT HERE]
To find out about the independent contribution of expected utility, norm obedience, and
opportunity in the explanation of plagiarism, each variable was first considered alone in bivariate
regression models (see Models 1 to 3 in Table 2). All coefficients in these models differ
significantly from zero and exhibit expected signs. The results indicate that an increase of one
standard deviation in expected utility leads to a 96.2 percent (or 1.962 times) higher expected
individual frequency of plagiarism within the six-month period of investigation (see Model 1).
An increase of one standard deviation in norm internalization decreases this frequency by 44.9
percent (Model 2). An increase of one standard deviation in opportunity increases the expected
number of incidences of plagiarism by 25.2 percent (Model 3).
To derive insights about the explanatory power of these three models and the subsequent
model specifications, the reduction of the Bayesian Information Criterion (BIC
) score is used.
For negative binomial regression models, BIC
is a suitable goodness-of-fit measure derived
from the number of model parameters, the deviance statistic, and the degrees of freedom (Hilbe
2011). Lower values of BIC
indicate a more fitting model. A comparison of the bivariate models
reveals that Model 1, which uses utility as predictor, has the lowest BIC
. This model has to be
preferred to Model 2, which investigates the norm variable. The latter outperforms Model 3,
which uses opportunity as a predictor.
[TABLE 2 ABOUT HERE]
In Model 4 (equivalent to equation 1), the main effects of all predictors are considered
together. Even when changes in the coefficients can be detected, all variables remain highly
significant and behave as the theory would predict. The performance of this model exceeds the
prior ones in terms of BIC
In Model 5 (corresponding to equation 2), the theoretically postulated interaction term
between the utility and the norm term was tested. It reveals an insignificant effect, implying that
norm and utility do not interact with each other when their influence on the mean number of
plagiarism cases per person is considered. The insertion of this insignificant interaction term
decreases the overall model performance, as indicated by a lower BIC
A further step analyzed the interaction between utility and opportunity (Model 6,
corresponding to equation 3). It turns out that there is a significant, positive interaction between
utility and opportunity, which could be interpreted as the increased influence of utility on the
expected frequency of plagiarism when more opportunities to plagiarize exist (or vice versa).
This amounts to a clear hint about the tempting effects of increased opportunities, illustrated in
Figure 2 (e.g. Frazier et al. 2004).
[FIGURE 2 ABOUT HERE]
As shown in Figure 2, the different slopes of the lines indicate that there is an increasing
influence of utility, while the ascending slope of the line indicates the positive effect of the
number of opportunities. Utility and opportunities are positively related to each other in their
influence on the frequency of plagiarism: The more opportunities exist, the higher is the slope
increment, resulting in a stronger influence of utility (and vice versa). Especially for those
students expecting high utility (indicated by the dark gray dashed line), increased opportunities
lead to a much higher mean number of plagiarism cases per student. On the other hand, if utility
is low (indicated by the light gray dashed line), students care little about the number of
opportunities. In terms of BIC
, the performance of Model 6 is slightly better than Model 5.
Model 6 is slightly worse than Model 4 but contains one additional significant interaction effect.
In Model 7 (corresponding to equation 4), all variables are considered simultaneously. It is
worth noting that the coefficients of previous models do not change. In terms of model
), this multivariate model is worse than other multivariate models (M4 to
M6) and offers no additional significant effects. Given these results, one can conclude that, in
conjunction with one other, the utility of plagiarism and the opportunity to plagiarize increase the
expected frequency of this deviant behavior significantly. Norms counteract this effect but when
they do, there is no interplay with utility.
CONCLUSION AND RECOMMENDATIONS
This paper analyzes whether RCT is a suitable theoretical approach for predicting the individual
frequency of plagiarism using data from the first large-scale two-wave panel study in Germany.
Our theoretical framework also includes predisposing factors as well as situational factors. To
explain plagiarism, a refined RCT version is used, which considers utility, norms, and
opportunities as well as theoretically derived interactions between utility and norms, and
between utility and opportunities. Our empirical results largely correspond to our hypotheses.
Hypothesis 1 is supported by our data: the higher the expected utility derived from plagiarism,
the more often actors plagiarize. Psychological costs accrued by violating internalized norms
reduce the number of incidents of plagiarism, which is consistent with hypothesis 2. A greater
number of opportunities to plagiarize increase the expected incidence, as implied by hypothesis
3. The non-significant interaction between utility and norms falsifies hypothesis 4 and may
indicate that students strongly deliberate about how often to plagiarize. They accordingly behave
in a very rational manner. In such a situation, there is no framing of alternative actions or
filtering. Nor is there a downplaying or dismissal of the positive aspects of plagiarism. Tittle et
al. (2010), for example, provided evidence that an interaction between utility and norms does not
occur in every context. But according to hypothesis 5 utility and opportunity interact positively.
A link thus exists between the preference for a certain action and its chance of realization. In
other words, the effect of utility emerges regardless of normative predispositions but might be
altered by chances for cheating behavior (and vice versa). Temptation, as an interpretation of this
significant interaction effect, might pose some theoretical challenges to RCT (Posner 1997). But
temptation does not necessarily result in a change in preferences (Gul and Pesendofer 2001).
Even when the tempting conditions are not under the actor’s control, according to Cornish and
Clarke (1987, p. 942) “… choice-structuring properties may often play a more active role in
generating offending.” With respect to previous studies with similar research problems, our
entire results are largely consistent with their empirical findings, for instance, those of Becker
and Mehlkop (2006) or Mehlkop and Graeff (2010) (with the exception of hypothesis 3).
In contrast, however, our study challenges the predictive power of the strict RCT model by
testing whether norms and opportunity have an effect on the incidence of plagiarizing. It could
have been the case that these variables rendered the RCT measure insignificant—if so, it would
have become evident that opportunity or obedience to norms (or both) supersede utility-based
decision-making. But the RCT variable passes this test. We thus favor model 6 because 1) it
provides the second lowest BIC
value, indicating a relative good performance and 2) it contains
the postulated significant interaction between U and O. This effect is not unveiled in Model 4,
which has a marginally lower BIC
value. The interaction effect reflects the importance of costs
and benefits for plagiarism in conjunction with opportunities and it also shows that plagiarism
should be seen as a strategic deviant behavior, differing from spontaneous deviant acts. But the
force of norms and opportunity was demonstrated as well, highlighting the importance of a broad
version of rationality.
Our findings show that an extended RCT version offers hints about how to affect students’
decision to plagiarize. Recommendations for curbing plagiarism can be derived from this. This is
extremely important because of the negative externalities of plagiarism described above (see
Introduction). Due to the huge impact of utility, prevention and deterrence means should
explicitly address the reduction of utility by diminishing the benefits of plagiarism and its
occurrence as well as increasing costs and the probability of detection. Highly skilled students,
both with expertise in academic writing and time management, expect lower benefits from
committing plagiarism (Sattler 2007, p. 174). With regard to our findings, this would imply that
plagiarism can be fought by enhancing these skills at the very beginning of a student’s academic
career. Furthermore, if the values of non-cheating behavior could be communicated, a normative
basis would also counteract plagiarizing tendencies. Additionally, Sattler’s study provides
evidence that students experiencing great pressure from their parents perceive higher benefits
from plagiarism. Defeating plagiarism is thus not only a task for universities but also for the
students’ social network. But it is up to the universities to increase the benefits associated with
non-plagiarized papers (such as providing detailed feedback and giving students the feeling of
personal achievement through their own skills). More thorough feedback about the progress of a
student’s abilities and knowledge would enhance the learning process and provide increased
benefit to students.
In general, universities should publicly disseminate information about exposed fraud;
otherwise the deterrent effect of any potential punishment becomes negligible (Bunn et al. 1992).
Checking papers by means of search engines or software enhances the objective probability of
detection and the subsequent subjective presumption of it (Sattler 2007, p. 163). Tackett et al.
(2010) found that a combination of software for detecting plagiarism and severe punishment
were very effective in decreasing the incidence of plagiarism. If students still choose to
plagiarize they must take higher risks into account, otherwise they need to either be skilled in
concealing their plagiarism or expend greater effort in producing plagiarism that is hard to
detect. In this way, the time-saving benefits of plagiarism are reduced (Collins et al. 2007). Not
everyone is ready to invest the effort needed to create the deterrent effect of advanced plagiarism
checks, which place heavy demands on time and personnel.
For this reason professional
. There are several other problems in addition to these expenditures (e.g. Collins et al. 2007; Crisp 2004):
plagiarism that remains undetected through sampled checking; texts inaccessible to searching software (like
detection units could be implemented inside universities (which is already the case in some
countries) or external services could be hired to take the burden off the teaching staff and to
check papers effectively (Sattler 2007). Teachers could then spend their time teaching rather than
spying and the teacher-student relationship would not be affected negatively. Presently, time
spent on dealing with plagiarism varies greatly among teachers as does their judgment of such
offences (Roig 2001). Standardizing methods of checking for plagiarism as well as penalties
would also make the situation fairer to all students. If teachers have to evaluate a large number of
student papers concurrently, this leads to perception that the probability of being caught as a
plagiarist is small (Collins et al. 2007). Another possible strategy would thus be to decrease the
number of participants in classes, making teaching more efficient and allowing the instructor to
know each student much better. Conspicuous changes in the student’s performance could be
discerned much more easily. Detection of possible plagiarism would become much more likely.
Our study also suggests that it is necessary to raise the cost of plagiarism by leveling
appreciable and severe punishment when it occurs. Rescinding credits already earned (Larkham
and Manns 2002) or suspending students from courses or from the university itself are significant
responses to plagiarism because they entail an increase in the price of attaining a degree. Fines or
the subsequent denial of degrees (after re-examining papers) may also deter students. Collins et
al. (2007) argue that universities can charge an extra fee for professional plagiarism detection
from students to recover the costs. Students who have not engaged in plagiarism would thus be
rewarded over time by a reduction in their fees. It is also indispensable to catch and severely
sanction repeated plagiarists. These offenders in particular should be confronted with higher
offline sources or online journals with access permissions); lack of detection of translations or re-written parts
If not, students who follow the norm may perceive themselves at a disadvantage and
likewise feel compelled to abandon the norm in order to compete on an equal basis.
As we have demonstrated in this paper, norms have a significant impact on the decision-
making process. One means of applying this finding are honor codes, which have the effect of
reducing fraud (Weinstein and Dobkin 2002). McCabe and Treviño (1993; McCabe et al. 2002)
explain the positive effect of honor codes by the fact that forbidden activities are clearly
explained. Sattler (2007) found some evidence for the fact that plagiarists sometimes simply do
not know what constitutes plagiarism and what does not. Nor are they aware of how universities
deal with plagiarism. For students who have to sign a pledge to follow academic rules, it is
harder to justify their fraud. Moreover, both the moral condemnation of plagiarism within a
student’s social networks as well as a personal moral condemnation of plagiarism reduce the
probability of success and increase the expected costs (Sattler 2007).
The empirical link between utility and opportunity would imply that even if all these
recommendations were implemented, plagiarism might still persist because the greater the
benefit of plagiarizing behavior, the greater the incentive to create or seek out opportunities for
plagiarizing. As long as the net utility is positive, a rational student would invest in creating or
advancing situations in which plagiarism is possible. The empirical results of this paper give rise
to this interesting point, but additional research would be needed to explain it.
Our study contains some elements that could be improved in forthcoming studies. First, we
used a self-report measure to assess the individual frequency of plagiarism; such direct response
measures are very common (Alam 2004; Cochran et al. 1999; Lim and See 2001). But since
plagiarism is about sensitive information, participants may refrain from telling the truth. We thus
checked for a social desirability bias but could not find any significant effect of two social
. Universities can register plagiarists on a blacklist and observe their papers with more vigilance.
desirability scales, self-deceptive enhancement (SDE) and impression management (IM).
Moreover, the item non-response was only 1.07 percent and efforts to provide objective and
subjective anonymity were very high. Future research could address this problem using special
question techniques like randomized response or crosswise technique (Yu et al. 2008).
Second, the longitudinal information could be improved. In our study, the individual
frequency of plagiarism was measured by the number of plagiarizing events over six months.
Plagiarism could also occur after this period. The opportunity measure captures the chances for
plagiarism, so the time interval is controlled. Ours is among the first studies able to analyze
longitudinal data. The results derived are encouraging, however, for future research that would
allow for a longer period and more than two time points.
In conclusion, it seems that the predictive power of an extended RCT is relatively high and
stable, which indicates that this approach is well-suited to explain plagiarism. Since it does not
account for all influences, however, additional explanations and interactions could be added as
well (such as self-control; Tittle et al. 2010). Further research is needed to explore the conditions
of plagiarism in particular with reference to possible extensions or refinements of RCT.
Aiken, Leona S. and Stephen G. 1991. “Multiple Regression: Testing and Interpreting Interactions.” Newbury Park,
Ajzen, Icek. 2002. “Perceived Behavioral Control, Self-Efficacy, Locus of Control, and the Theory of Planned
Behavior.” Journal of Applied Social Psychology 32:665–83.
Alam, Lubna Sheik. 2004. “Is plagiarism more prevalent in some form of assessment than others? Beyond the
comfort zone.” Proceedings of the 21
ASCILITE Conference, eds. Roger Atkinson, Clare McBeath, Diana
Jonas-Dwyer. Perth. Retrieved July 7, 2005 (http://www.ascilite.org.au/conferences/perth04/procs/pdf/alam.pdf)
Beck, Lisa and Icek Ajzen. 1991. “Predicting dishonest actions using the theory of planned behavior.” Journal of
Research in Personality 25:285–301.
Becker, Gary S. 1968. “Crime and Punishment: An Economic Approach.” Journal of Political Economy 76:169–
Becker, Gary S. 1976. “The Economic Approach to Human Behavior.” Chicago: University Of Chicago Press.
Becker, Rolf and Guido Mehlkop. 2006. “Social Class and Delinquency.” Rationality and Society 18:193–235.
Bennett, Trevor and Richard Wright. 1984. “Burglars on Burglary - Prevention and the Offender.” Brookfield:
Beuer-Krüssel, Mandy, and Ivar Krumpal. 2009. “The impact of frequency formats on the measurement of
subjective probability” [orig. title: Der Einfluss von Häufigkeitsformaten auf die Messung von subjektiven
Wahrscheinlichkeiten]. Methods, Data, Analysis 3:31–58.
Bunn, Douglas N., Steven B. Caudill, and Daniel M. Gropper. 1992. “Crime in the Classroom: An Economic
Analysis of Undergraduate Student Cheating Behavior.” Journal of Economic Education 23:197–207.
Clarke, Ronald V. and Marcus Felson 1993. “Introduction: Criminology, Routine Activity, and Rational Choice.”
Pp. 1-14 in Routine Activity and Rational Choice, eds. Ronald V. Clarke and Marcus Felson. New Brunswick,
USA: Transaction Publishers.
Cochran, John K., Mitchell B. Chamlin, Peter B. Wood, and Christine S. Sellers. 1999. “Shame, embarrassment, and
formal sanction threats: Extending the deterrence/ rational choice model to academic dishonesty.” Sociological
Coleman, James S. 1988. “Social Capital in the Creation of Human Capital.” American Journal of Sociology
Coleman, James S. 1990. “Foundations of Social Theory.” Cambridge: Harvard University Press.
Collins, Alan, Guy Judge, and Neil Rickman. 2007. On the economics of plagiarism.” European Journal of Law and
Cornish, Derek B. and Ronald V. Clarke. 1987. “Understanding crime displacement: an application of rational
choice theory.” Criminology 25:933–48.
Cornish, Derek B. and Ronald V. Clarke. 2003. “Opportunities, precipitators and criminal decisions: A reply to
Wortley´s critique of situational crime prevention.” Pp. 41-96 in Theory for Practice in Situational Crime
Prevention, eds. Martha J. Smith and Derek B. Cornish. Monsey, New York: Criminal Justice Press.
Crisp, Geoffrey T. 2004. “Plagiarism and the Reputation of the University: How to Distribute Effort between
Educating Students on Attribution and Rigorous Detection of Cheating?” Pp. 53-56 in Proceedings of the
Australian Universities, eds. Rob Carmichael. Melbourne: Australian Universities Quality Agency.
De Mesquita, Bruce Bueno and Lawrence E. Cohen. 1995. “Self-interest, equity, and crime
control: A game-theoretic analysis of criminal decision making.” Criminology 33:483–518.
Diekmann, Andreas and Peter Preisendörfer. 2003. “Green and Greenback. The Behavioral Effects of
Environmental Attitudes in Low-Cost and High-Cost Situations.” Rationality and Society 15:441–72.
Duggan, Fiona, Nik Pollinger, and Annabel Kilner. 2004. “FreshMinds´ research on plagiarism.” Retrieved
September 10, 2005 (http://www.freshminds.co.uk/FreshMinds_plagiarism_survey.pdf)
Ehrlich, Isaac. 1996. “Crime, Punishment, and the Market for Offenses.” Journal of Economic Perspectives 10:43–
Eide, Erling. 1997. “The Economics of Crime: Main Problems and Some Solutions.” Nordic Journal of Political
Eide, Erling. 2000. “Economics of Criminal Behavior.” Pp. 345-389 in Encyclopedia of Law and Economics. The
Economics of Crime and Litigation, eds. Boudewijn Bouckaert and Gerrit De Geest. Cheltenham: Edward Elgar.
Elster, Jon. 1985. “The Nature and Scope of Rational Choice Explanation.” Pp. 60-72 in Essays on Actions and
Events, eds. Ernest LePore and Brian P. McLaughlin. Oxford & New York: Blackwell.
Elster, Jon. 1989. “Social Norms and Economic Theory.” Journal of Economic Perspectives 3:99–117.
Fattah, Ezzat A. 1993. “The Rational Choice/Opportunity Perspectives as a Vehicle for Integrating Criminological
and Victimological Theories.” Pp. 225-258 i Routine Activity and Rational Choice, eds. Ronald V. Clarke and
Marcus Felson, New Brunswick, USA: Transaction Publishers.
Frazier, Patricia A., Andre P. Tix, and Kennetz E. Barron. 2004. “Testing Moderator and Mediator Effects in
Counseling Psychology Research.” Journal of Counselling Psychology 51:115–34.
Friedman, David. 1995. “Rational criminals and profit-maximizing politics. The economic analysis of law and law
enforcement.” Pp. 43-58 in The new economics of behavior, eds. Mariano Tommasi and Kathryn Ternulli.
Cambridge: Cambridge UP.
Friedrichs, Jürgen, Martin Stolle, and Gudrun Engelbrecht. 1993. “Rational choice theory: Problems of
operationalization” [orig. title: Rational Choice Theorie: Probleme der Operationalisierung]. Zeitschrift für
Grasmick, Harold G. and Robert J. Bursik. 1990. “Conscience, Significant Others, and Rational Choice: Extending
the Deterrence Model.” Law & Society Review 24:837–61.
Gul, Faruk and Wolfgang Pesendorfer. 2001. “Temptation and self-control.” Econometrica 69:1403–35.
Hastie, Reid and Robyn M. Dawes. 2010. “Rational Choice in an Uncertain World: The Psychology of Judgment
and Decision Making.” London: Sage Publications.
Hechter, Michael. 1994. “The Role of Values in Rational Choice Theory.” Rationality and Society 6:318–33.
Hilbe, Joseph, 2009. “Logistic Regression Models.” Boca Raton, FL: Chapman & Hall/CRC.
Hilbe, Joseph, 2011. “Negative Binomial Regression.” Cambridge: Cambridge University Press.
Iannaccone, Laurence R. 1995. “Voodoo Economics? Reviewing the rational choice approach to religion.” Journal
for the Scientific Study of Religion 34:76–89.
Jordan, Augustus E. 2001. “College Student Cheating: The Role of Motivation, Perceived Norms, Attitudes, and
Knowledge of Institutional Policy.” Ethics & Behavior 11:233–47.
Kerkvliet, Joe and Charles L. Sigmund. 1999. “Can We Control Cheating in the Classroom?” Journal of Economic
Kroneberg, Clemens, Isolde Heintze, and Guido Mehlkop. 2010. “The Interplay of Moral Norms and Instrumental
Incentives in Crime Causation. Criminology 48: 259–94.
Larkham, Peter J. and Susan Manns. 2002. “Plagiarism and its Treatment in Higher Education.” Journal of Further
and Higher Education 26:339–49.
Lindenberg, Siegwart M. 1983. “Utility and Morality.” Kyklos 36:450–68.
Long, J. Scott and Jeremy Freese. 2001. “Regression models for categorical and limited dependent variables.”
Thousand Oaks, London and New Delhi: Sage Publications
MacDonald, John M. and Pamela K. Lattimore. 2010. “Count Models in Criminology.” Pp. 583-698 in Handbook of
Quantitative Criminology, eds. David Piquero and Alex R. Weisburd. New York: Springer.
March, James G. and Johan P. Olsen. 1989. “Rediscovering Institutions. The Organizational Basis of Politics.” New
York: Free Press.
Martinson, Brain C., Melissa S. Anderson, and Raymond de Vries. 2005. “Scientists behaving badly.” Nature
McCabe, Donald L. 2001. “Cheating: Why Students Do It and How We Can Help Them Stop.” American Educator
McCabe, Donald L. 2005. In New CAI Research Conducted. Retreieved August 31, 2005
McCabe, Donald L. and Linda Klebe Treviño. 1993. “Academic dishonesty: honor codes and other contextual
influences.” Journal of Higher Education 64:522–38.
McCabe, Donald L., Linda Klebe Treviño, and Kenneth D. Butterfield. 2002. “Honor Codes and
Other Contextual Influences on Academic Integrity.” Research in Higher Education 43:357–
Mehlkop, Guido and Peter Graeff. 2010. “Modelling a Rational Choice Theory of Criminal Action: Subjective
Expected Utilities, Norms, and Interactions.” Rationality & Society 22:189–222.
Michaels, James W. and Terance D. Miethe. 1989. “Applying Theories of Deviance to Academic Cheating.” Social
Science Quarterly 70:870–85.
Nagin, Daniel S. and Raymond Paternoster. 1993. “Enduring individual differences and rational choice theory of
crime.” Law & Society Review 27:467–69.
Opp, Karl-Dieter. 1998. “Can and should Rational Choice Theory be tested by survey research? The Example of
Explaining Collective Political Action.” Pp. 204-230 in Rational Choice Theory and Large-Scale Data Analysis,
eds. Hans-Peter Blossfeld and Gerald Prein. Boulder: Westview Press.
Opp, Karl-Dieter. 2012. “Norms and Rationality. Is Moral Behavior a Form of Rational Action?” Theory and
Paulhus, Delroy L. 1991. “Measurement and control of response bias.” Pp. 17–59 in Measures of personality and
social psychological attitudes, eds. John P. Robinson, Phillip R. Shaver and Lawrence S. Wrightman. New York:
Piliavin, Irving, Craig Thornton, Rosemary Gartner, and Ross L. Matsueda. 1986. “Crime, Deterrence and Rational
Choice.” American Sociological Review 51:101–19.
Posner, Richard A. 1997. “Rational Choice, Behavioral Economics, and the Law.” Stanford Law Review 50:1555–
Posner, Richard A. and Eric B. Rasmusen. 1999. “Creating and Enforcing Norms, with Special Reference to
Sanctions.” International Review of Law and Economics 19:369–82.
Rettinger, David A., Augustus E. Jordan, and Francisco Peschiera. 2004. “Evaluating the motivation of other
students to cheat: A vignette experiment.” Research in Higher Education 45:873–90.
Roig, Miguel and Lauren DeTommaso. 1995. “Are College Cheating and Plagiarism Related to academic
procrastination?” Psychological Reports 77:691–98.
Roig, Miguel. 2001. “Plagiarism and Paraphrasing Criteria of College and University Professors.” Ethics &
Royston, Patrick. 2005. “Multiple Imputation of Missing Values. The Stata Journal 5:527–36.
Sattler, Sebastian. 2007. “Plagiarism in student papers: Explanation models using rational choice theory” [orig.
title: Plagiate in Hausarbeiten. Erklärungsmodelle mit Hilfe der Rational Choice Theorie]. Hamburg: Kovac.
Shih, Tse-Hua and Xitao Fan 2008. “Comparing Response Rates from Web and Mail Surveys: A Meta-Analysis.
“Field Methods 20:249–271.
Tackett, James, Gregory A. Claypool, Fran Wolf, and Joe Antenucci. 2010. “The impact of plagiarism detection
software on college plagiarism.” Journal of Business and Accounting 3:68–80.
Taylor, Frederick Kräupl. 1965. “Cryptomnesia and Plagiarism.” British Journal of Psychiatry 111:1111–18.
Teixeira, Aurora A. C. and Maria Fátima Rocha. 2006. “Academic Cheating in Austria, Portugal, Romania and
Spain: a comparative analysis.” Research in Comparative and International Education 1:198–209.
Teixeira, Aurora A. C. and Maria Fátima Rocha. 2008. “Academic cheating in Spain and Portugal: An empirical
explanation.” International Journal of Iberian Studies 21:3–22.
Tibbetts, Stephen G. 1997. “Gender differences in students’ rational decisions to cheat.” Deviant Behavior 18:393–
Tittle, Charles R., Olena Antonaccio, Ekaterina Botchkovar, and Maria Kranidioti. 2010. “Expected utility, self-
control, morality, and criminal probability.” Social Science Research 39:1029–46.
Trasler, Gordon. 1993. “Conscience, Opportunity, Rational Choice, and Crime.” Pp. 305–322 in Routine Activity
and Rational Choice, eds. Ronald V. Clarke and Marcus Felson, New Brunswick, USA: Transaction Publishers.
Voss, Thomas and Martin Abraham. 2000. “Rational Choice Theory in Sociology: A Survey.” Pp. 50–83 inThe
International Handbook of Sociology, eds. Stella R. Quah and Arnaud Sales. London: Sage Publications.
Weinstein, Jeffrey W. and Carlos E. Dobkin. 2002. “Plagiarism in U.S. Higher Education: Estimating Internet
Plagiarism Rates and Testing a Means of Deterrence.” Retrieved July 8, 2005
Whitley, Bernard E. 1998. “Factors associated with cheating among college students: A review.” Research in
Higher Education 29:235–74.
Winkler, Niels, Martin Kroh, and Martin Spiess. 2006. “Development of a German bidimensional short scale for
measuring social desirability” [orig. title: Entwicklung einer deutschen Kurzskala zur zweidimensionalen
Messung von sozialer Erwünschtheit]. DIW Berlin Discussion Papers. Retrieved May 10, 2010
Wikström, Per-Olof H. and Robert Svensson. 2010. “When does self-control matter? The interaction between
morality and self-control in crime causation.” European Journal of Criminology 7:395–410.
Williams, Robin M. 1968. “Values: The Concept of Values.” International Encyclopaedia of the Social Sciences
Williamson, Oliver E. 1993. “Calculativeness, Trust, and Economic Organization.” Journal of Law & Economics
Wortley, Richard. 1998. “A two-stage model of situational crime prevention.” Studies on Crime and Crime
Yu, Jun-Wu, Guo-Liang Tian, and Man-Lai Tang. 2008. “Two new models for survey sampling with sensitive
characteristic: design and analysis.” Metrika 67:251–63.
This research was funded by the German Federal Ministry of Education and Research (FMER;
01PH08024, headed by Sebastian Sattler and Martin Diewald). We would like to thank all those
who helped conduct the study, especially Nina Chudziak, Anja Göritz, Anatol-Fiete Näher,
Dominik Koch, Ines Meyer, Andrea Schulze, Floris van Veen, and Constantin Wiegel. We
would like to thank Donald L. McCabe, Guido Mehlkop, Robert Neumann, Deborah Weber-
Wulff, and an anonymous reviewer for critical comments. The FMER did not influence any
interpretations or compel or encourage the research team to produce any specific results. The
views expressed do not necessarily reflect the policies of the funder. We herewith declare that we
are fully liable for the integrity of the data and the correctness of the data analysis. All authors
contributed to, read, and approved the manuscript. Sattler (2007) has given a broad overview of
the research field of plagiarism. In order to save space, we draw directly from his book for some
suggestions and quotations.
SEBASTIAN SATTLER, M.A. is a Ph.D. candidate in Sociology at the Faculty of Sociology at
the Bielefeld University. He has recently completed the most extensive study on academic
misconduct in German-speaking countries funded by the German Federal Ministry of Education
and Research. His research interests include the explanation of behavior, pharmaceutical
cognitive enhancement, the measurement of sensitive behavior as well as misconduct and fraud
in universities (especially plagiarism). His master thesis and book on the explanation of
plagiarism won a prize given by the German Society of Sociology.
PETER GRAEFF, Ph.D., is affiliate professor for methodology and statistics at the Bundeswehr
University, Munich. His research interests are regression techniques for quantitative empirical
research, social capital, corruption and deviant behavior. He has published several books and
articles in the leading Journals in these fields such as European Sociological Review, Quality &
Quantity and the Journal of Mathematical Sociology.
SEBASTIAN WILLEN, Dipl. Soz. is currently a Ph.D. candidate in Social Science at the
Institute of Sociology at the University of Duisburg-Essen. At his time at the Bielefeld
University, he was a research assistant in a project on academic misconduct. His research
interests are intercultural opening, men's fertility and complex statistical methods.
FIGURES AND TABLES
Model to Explain Plagiarism
Estimated Frequency of Plagiarism as Related to Utility and
Opportunity (Based on Model 5)
NOTES: ‒‒ ‒‒ ‒‒ -SD
‒ ‒ ‒ ‒ +SD
; The effect of norms is set to its mean.
-SD Mean +SD
Estimated Frequency of Plagiarism
Descriptive Statistics (N = 2,806)
Probability of Success
Probability of Detection
Negative Binomial Regression of the Individual Frequency of
Plagiarism on Utility, Norm and Opportunity (N = 2,806)
[1.754 | 2.196]
[1.531 | 1.939]
[1.536 | 1.948]
[1.507 | 1.909]
[1.511 | 1.918]
[0.491 | 0.618]
[0.604 | 0.763]
[0.594 | 0.756]
[0.6404 | 0.762]
[0.594 | 0.756]
[1.123 | 1.395]
[1.174 | 1.443]
[1.175 | 1.445]
[1.144 | 1.415]
[1.143 | 1.415]
[0.941 | 1.175]
[0.937 | 1.168]
[1.013 | 1.263]
[1.011 | 1.261]
NOTES: Incidence rate ratios. 95% confidence intervals in brackets.
p < .05;
p < .001 (two-sided).