Content uploaded by Victor Polterovich
All content in this area was uploaded by Victor Polterovich
Content may be subject to copyright.
Tolerance of Cheating:
An Analysis Across Countries
Jan R. Magnus, Victor M. Polterovich,
Dmitri L. Danilov, and Alexei V. Savvateev
Abstract: Cheating is a serious problem in many countries. The cheater gets high-
er marks than deserved, thus reducing the efficiency of a country’s educational
system. In this study, the authors did not ask if and how often the student had
cheated, but rather what the student’s opinion was about a cheating situation.
They investigated whether attitudes differ among students in Russia, the Nether-
lands, Israel, and the United States and conclude that attitudes toward cheating
differ considerably between these countries. They offer various explanations of
this phenomenon. In addition, they find that the student’s attitude toward cheat-
ing depends on the student’s educational level (high school, undergraduate, post-
graduate). Finally, they show that the data from the sample can be aggregated in
a natural and elegant way, and they suggest a tolerance-of-cheating index for
Key words: cheating, international, students, tolerance
JEL codes: A13,A20, K42
Cheating is a serious problem in many countries. The cheater is a free rider and
therefore gets higher marks than he or she deserves. The efficiency of the coun-
try’s educational system is reduced, because cheating distorts competition,
diminishes the student’s incentive to study, and leads to inaccurate evaluation of
the student’s abilities. More information about the phenomenon of cheating is
needed, if only to design appropriate deterrence mechanisms.
Several previous authors have studied the frequency and reasons for cheating.
Bunn, Caudill, and Gropper (1992) interviewed U.S. economics undergraduates
and concluded that many students cheat, that the brighter the student, the less
likely it is that he or she has cheated, and that there is a higher probability
attached to having cheated once if the student believes others to be cheating.
Spring 2002 125
Jan R. Magnus is professor of econometrics at CentER,Tilburg University, the Netherlands (e-mail:
email@example.com). Victor M. Polterovich is head of the Mathematical Economics Laboratory at the
Central Economics and Mathematics Institute (CEMI) of the Russian Academy of Sciences and pro-
fessor of economics at the New Economic School (NES) in Moscow. Dmitri L. Danilov is a Ph.D. stu-
dent at Tilburg University. Alexei V. Savvateev is a Ph.D. student at CEMI. The authors are grateful
to professors Michael Alexeev, Itzhak Zilcha, and Tatiana Selezneva and to Dr. Serguei Kokovin; to
Ph.D. students Nina Baranchuk, Andrei Bremzen, Maxim Ivanov, Serguei Izmalkov, Inna Maltseva,
Mila Todorova, Alexander Tonis, and Natalya Volchkova and others for help in collecting the data;
to students in four countries for filling out the questionnaires; and to three referees for helpful and
Whereas Bunn et al. covered only the cheating-once case, Mixon (1996) was
interested in habitual cheating. His main conclusion was that the determinants of
habitual cheating are much the same as those that relate to having cheated once.
Both Bunn, Caudill, and Gropper (1992) and Mixon (1996) stressed the analogy
between cheating and crime (especially theft).
Kerkvliet (1994), also using U.S. data, concluded that about one-third of stu-
dents had cheated on at least one occasion. Nowell and Laufer (1997) found that
nontenure track faculty, large classes, poor performance in the class, and being
employed all lead to more cheating. Kadane (1999) assessed whether data over
11 examinations supported an accusation of copying multiple-choice answers.
Finally, Kerkvliet and Sigmund (1999) explored the determinants of source-spe-
cific cheating behavior, including student characteristics and deterrent measures.
They concluded that large alcohol consumption and low grade point average
(GPA) increase the probability of cheating. Interestingly, they found that the fur-
ther along a student was in his or her academic career, the more likely he or she
was to cheat. The most striking result was the difference in student cheating
between students who were taught by teaching assistants and those taught by fac-
ulty; students taught by teaching assistants were 32 percent more likely to cheat
than students taught by faculty.
Our study was different from those cited in several respects. We did not ask if
and how often the student had cheated but rather what the student’s opinion was
about a cheating situation. Thus, we tried to analyze the student’s attitude toward
cheating. All previous studies—with the exception of Davis, Noble, Zak, and Drey-
er (1994)—have been devoted to only one country. In contrast, we attempted to
compare attitudes across countries. The results of our survey and subsequent sta-
tistical analysis for the United States, the Netherlands, Israel, and Russia showed
that attitudes toward cheating differed considerably across those countries. We
offer three possible explanations of this phenomenon: cultural factors, design of the
educational system, and the possible occurrence of multiple equilibria.
In addition, we found that the student’s attitude toward cheating depended on
the student’s educational level (high school, undergraduate, postgraduate). Final-
ly, we show that the data from the sample can be aggregated in a natural and ele-
gant way, and we suggest a tolerance-of-cheating index for each country.
EXPERIMENTAL DESIGN AND SUMMARY OF THE DATA
In 1997, we conducted a small survey in four countries at three different lev-
els of education. Our design was very simple. We asked each respondent to con-
sider the following situation: Student C reports to the departmental office that
student A, while taking an exam, copied answers from student B’s paper with the
consent of student B. The questionnaire then asked the respondent to character-
ize his or her attitude toward each of A, B, and C on a 5-point scale: strongly neg-
ative (–2), negative (–1), neutral (0), positive (+1), or strongly positive (+2).
Thus, each respondent in our sample provided three answers. Of course, all
answers were anonymous. Because the questions were simple and quick to
answer, the response rate was close to 100 percent.
126 JOURNAL OF ECONOMIC EDUCATION
Our sample contained 885 students from four countries: 92 high school stu-
dents, 554 university undergraduates (mostly from economics departments), and
239 economics postgraduates. The majority of the interviewed students was from
Russia, 322 from Moscow, and 184 from provincial Russia (Ekaterinburg, Per-
vouralsk, Voronezh, and Novosibirsk). In the United States, we interviewed 112
students, in the Netherlands, 247. We also had a small sample from Israel con-
sisting of one class of 20 undergraduates.
We kept the students from Moscow and provincial Russia separate because
there was no a priori reason to believe that the behavior in the capital and the
province would be the same. For the purposes of our study, we considered
provincial Russia as a fifth country. (We checked that the responses from the stu-
dents in the cities outside Moscow were sufficiently homogeneous to be aggre-
gated.) We thus considered five countries (Moscow, provincial Russia, Israel, the
Netherlands, the United States) and three educational levels (high school, under-
graduate, postgraduate). For each of the 15 possible combinations of five coun-
tries and three educational levels, we could calculate the average attitude toward
students A, B, and C in our sample. Because three cells were empty, we provide
12 sets of cell averages and all totals, together with the number of observations
(N) in each cell in Table 1.
Spring 2002 127
Data Summary Statistics
Country High school Undergraduate Postgraduate Total
Russia A–0.18 –0.02 –0.43 –0.24
(Moscow) B0.78 0.85 0.19 0.52
C–1.88 –1.78 –1.76 –1.78
N40 124 158 322
Russia A–0.45 –0.08 0.00 –0.14
(Province) B0.33 0.65 0.43 0.58
C–1.27 –1.72 –1.57 –1.64
N33 144 7 184
Israel A— –0.50 — –0.50
B— 0.25 — 0.25
C— –1.15 — –1.15
Netherlands A–0.16 –0.78 –1.37 –0.83
B0.63 –0.06 –0.32 –0.05
C–1.53 –1.52 –0.51 –1.36
N19 187 41 247
United States A— –1.27 –1.55 –1.35
B— –0.87 –0.88 –0.88
C— –0.34 –0.03 –0.25
Total A–0.27 –0.48 –0.73 –0.53
B0.59 0.22 –0.04 0.19
C–1.59 –1.45 –1.30 –1.42
N92 554 239 885
Note: A is the student who cheated; Bis the student who allowed Ato copy his or her answers; Cis the student who
reported the cheating; and Nis the number of observations.
Even without model or statistical analysis, these data summaries provided two
preliminary conclusions. First, all Russian students hated informers (C = –1.73,
on average). The Russian saying: “First whip to the informer” appeared to pre-
vail. Students from Israel and the Netherlands were not keen on informers either,
but in the United States, students seemed to have a different attitude (C= –0.25).
Second, in each country, except provincial Russia, high school students were less
tolerant of the informer C than undergraduate students who, in turn, were less tol-
erant than postgraduates. One would, therefore, expect that the higher the level
of education, the less tolerant students were of A, the person who cheated. This
was indeed the case in the United States and the Netherlands but not in Russia.
Hence, we needed to allow for the possibility that the dependence on education-
al level is different in Russia than in the other countries.
DIFFERENCES IN ATTITUDES TOWARD CHEATING
We had five countries in our sample, and we used five dummy variables, x1,
. . . , x5, one for each country. For example, x1 = 1, if the respondent came from
Moscow and zero otherwise. In addition, we had three educational levels, so we
added two (not three) further dummy variables, x6 and x7. The dummy x6= 1, if
the respondent was an undergraduate and zero otherwise, and x7 = 1, if the
respondent was a postgraduate and zero otherwise. Adding another dummy for
high school students would have led to an identification problem. To the seven
main effects dummies, we added one interaction dummy x8, which took the value
1, if the student was a high school student from provincial Russia. This dummy
allowed for the possibly different dependence on educational level in Russia than
in other countries.
Each respondent i (i= 1, . . . N) produced three answers, his or her attitude to
A, B, and C, respectively. We let yi denote the answer to the first question (A) on
a 5-point scale (–2, –1, 0, 1, 2). Because we had five ordered categories, we for-
mulated a simple ordered-response model to analyze the data (see Maddaia 1983,
section 2.13): In the ordered-response model, we defined a latent variable y*
iβ+ εi,i= 1, . . . , N,
where xiwas an 8 × 1 vector of the dummy variables defined above, and β was
an 8 × 1 vector of parameters to be estimated. The errors εiwere assumed inde-
pendent and identically distributed as N(0, σ2).2 We did not observe y*, but rather
y, which took on five discrete values according to the following rule:
128 JOURNAL OF ECONOMIC EDUCATION
where τ1, τ2, and τ3 denote the threshold parameters. For purposes of identifica-
tion and without loss of generality, we set τ0 = 0. Even then, only the ratios τi/σ
and β/σ were identified. We followed the usual convention and normalized σ to
equal 1. We then estimated the three equations separately by maximum likeli-
hood. The resulting estimates and standard errors of the 11 parameters (8 βs and
3 τs) are presented in Table 2.
The estimated coefficients of an ordered probit model must be interpreted with
care. The value of ^
βk denoted the effect of a change in the kth dummy variable on
the expectation of the latent variable y*, and hence indirectly on the expectation
of the observed y. For example, we saw that the higher the level of education, the
more negative the student was toward A and B, and the more positive toward C.
Also, Russian students were most positive (and U.S. students most negative)
toward A and B, whereas the attitude toward C was precisely the opposite, as
compared to students in other countries.
A formal statistical analysis showed the following: First the significance of the
11 coefficients jointly was enormous: χ2(11) was above any reasonable level of
rejection. Second, if we tried to pool data for Moscow and provincial Russia,
while deleting the cross term x8, that is, if we tested the joint hypothesis β1= β2,
β8 = 0, then this was firmly rejected (p value is .002). However, if we kept the
cross-term, then Moscow and provincial Russia could be pooled (p= .12). We
used this fact later in estimating a model with corruption (Table 5). Third, we
Spring 2002 129
Russia (Moscow) 2.0361 2.8854 –1.3136
(0.1616) (0.1722) (0.2043)
Russia (Province) 2.0673 2.8563 –0.9319
(0.1943) (0.2009) (0.2352)
Israel 1.4095 2.3523 –0.0955
(0.2944) (0.2978) (0.3234)
Netherlands 1.0181 2.0223 –0.4955
(0.1637) (0.1729) (0.2075)
United States 0.3318 1.0690 0.5258
(0.1907) (0.1881) (0.2325)
Undergraduate –0.1735 –0.3409 0.1615
(0.1587) (0.1554) (0.2103)
Postgraduate –0.7483 –0.8379 0.5607
(0.1642) (0.1601) (0.2146)
Russia (Provincial) × high school –0.7888 –0.7332 0.7330
(0.2664) (0.2591) (0.3122)
τ11.0668 0.9082 0.7816
(0.0593) (0.0752) (0.0547)
τ22.9780 2.5007 1.3666
(0.0935) (0.0935) (0.0797)
τ33.4733 3.3087 1.7724
(0.1156) (0.1030) (0.1013)
χ2(11) 6,188 15,474 2,323
Note: Standard errors are in parentheses. See note to Table 1 for definitions of A,B, and C.
could test parameter restrictions across equations. These tests showed that stu-
dents of the same educational level in different countries had very different atti-
tudes toward A, B, and C. However, students within one country appeared to have
the same attitude toward A and B, independent of their educational level (p = .45
for undergraduates and .70 for postgraduates). Students within one country obvi-
ously did not have the same attitude toward A and C; if anything they had the
opposite attitude. If we tested A = –2C (and similarly B = –2C) across educa-
tional levels, then we could not reject this hypothesis. This suggested that aggre-
gation over educational levels may be possible, and that possibly a tolerance-of-
cheating index might be constructed. We return to this issue later.
First, collective and individualistic values differ between countries. In the
United States and Russia, two cultural differences appear to relate directly to
cheating. First, in the United States, in contrast to Russia, competition among
students is seen as an important intrinsic value of the educational system, a value
that affects interaction between students. Thus, cheating is condemned because
it is considered an unfair instrument of competition. Second, the attitude to the
law and to officials differ between the two countries. In the former USSR, the
judicial system served as an instrument of the party, and a common view was that
officials are enemies. This attitude existed toward policemen, civil servants, train
conductors, and also toward teachers, and may explain the strong negative atti-
tude toward informers among Russian students. It seems plausible that the same
cultural factors influence other behavior such as tax evasion or corruption. If so,
one may expect that cheating and corruption are closely correlated, and this
would be of interest because perceived corruption is much more difficult to mea-
sure than perceived cheating.
Design of the Educational System
One can argue against the cultural explanation by saying that many students in
the United States, the Netherlands, and Israel are actually foreigners who come
from many different cultures. Russian students in the United States probably do
not cheat.3Thus, the difference in tolerance of cheating might not depend on cul-
ture (or not only on culture) but on the design of the educational system: the grad-
ing system, selection procedures, severity of punishment, number of students in
classes, existence of study groups,4existence of code of honor, and so forth.5
Even if one could prove that young Russians do not cheat when studying in the
United States, this would not refute a cultural theory. To understand why, one
may use a game-theory approach that is widely applied in the theory of corrup-
tion (Tanzi 1997) and other types of deviant behavior. The approach follows
Becker’s (1968) economic analysis of the rationality of crime (relying on expect-
ed costs and benefits), where cheating is considered as a rational act where the
130 JOURNAL OF ECONOMIC EDUCATION
student balances expected utility of higher grades against expected costs (sever-
ity of punishment, probability of getting caught, prevailing attitude toward cheat-
ing). If many students in a collective have negative attitudes toward cheating,
then it is difficult to get help in cheating, and the probability is high that some-
body will inform the teacher. Moreover, a cheater and his or her assistants, if
detected, will get no sympathy from classmates, but informers are not con-
demned. Hence, the cost of cheating and assisting cheating is high, whereas the
cost of informing about cheating is low.
Cheating and the attitude toward cheating are interconnected and self-sup-
porting. The larger the number of students in a collective that is cheating and tol-
erant toward cheating, the more often they cheat, the more tolerant they are, and
the less costly it is for every student to cheat and to be tolerant toward cheating.
This is the so-called coordination effect: the more consistently a norm is
observed in society, the greater the costs incurred by an individual deviating from
it. The coordination effect causes multiplicity of equilibria in socioeconomic sys-
tems (Arthur 1988; North 1997). Prevalence of cheating can be considered as a
stable inefficient equilibrium, a lock-in or institutional trap (Polterovich 2000).
This analysis can be converted into a formal model with cheating and free-of-
cheating equilibria. Cultural and organizational factors, as well as the history of
the system, define which of two equilibria prevails. If the system is free of cheat-
ing, then the cheating costs are high, and a newcomer may find it more benefi-
cial to observe the prevalent norm even if he or she is inclined to cheat. The influ-
ence of the educational level on cheating and on the attitude toward cheating is
ambiguous. On the one hand, learning effects (development of cheating tech-
niques), linkage effects (interdependence of cheating and friendship relations),
and cultural inertia (formation of cheating as a habit) decrease cheating costs
over time and fix cheating as a norm of behavior. On the other hand, the higher
the educational level, the more severe the punishment for cheating, and the larg-
er possible losses of accumulated investment in education: by being expelled, a
final-year student devalues a substantial part of the payments and efforts that
have been invested in his or her education.
If cheating prevails and cheating costs are low, then the norm-fixation process
is most important for the earlier stages of the education. For the advanced stages
and for a low cheating equilibrium, the threat of losing accumulated investment
seems to dominate. This can serve as a tentative explanation of nonmonotonic
dependence of the attitude to cheating in Russia in contrast to other countries
Comparisons of cheating behavior across countries may inter alia lead to prac-
tical conclusions about the effectiveness of different deterrent mechanisms. How-
ever, the comparisons and analysis would be simpler if we could characterize the
Spring 2002 131
cheating phenomenon by only a few indicators, preferably one. This would also
allow comparison with other social science indices, like those of corruption, eco-
nomic freedom, liberalization, and quality of institutions. Attitude toward cheat-
ing is a complex phenomenon that involves attitudes toward cheaters, those who
facilitate cheating, and informers. Our data were three-dimensional, because we
had three answers from each respondent. The question is whether we can aggre-
gate the answers, A, B, and Cand construct a one-dimensional tolerance-of-
cheating index (TCI). The obvious first guess about TCI would be based on (A+
B– C), because a person who is extremely negative on cheating would have A=
–2, B= –2, and C= 2 and would thus obtain a score of –6, whereas the opposite,
very tolerant, person would have a score of +6. We argue that of all linear com-
binations of A, B, and C, this particular choice was the optimal one.
From the data, we computed the following correlation matrix of the answers
to A, B, and C:
The largest eigenvalue of Ris 2.0004, and the associated eigenvector, called the
first principal component of R(see Anderson 1984), is υ = (0.5858, 0.6101,
–0.5336). The correlation between υ and the hypothesized vector (1, 1, –1) was
astonishingly high, namely 0.9985. We concluded, therefore, that the weighting
(1, 1, –1) was the best linear combination in the sense that it explained most of
the variation in the data.
Instead of (A+ B– C), we defined a linear function as
TCI = 5 – 5(A + B – C)/6.
Thus defined, the TCI is a number between 0 and 10, and the higher the number,
the lower was the tolerance to cheating. This is more intuitive and more in line
with other indices (such as the corruption index discussed later), because cheat-
ing (like corruption) is “bad,” and hence a high TCI is “good.” Given our defini-
tion of the TCI, we can calculate the “empirical” TCI directly from the cell aver-
ages in Table 1. These summary statistics are presented in Table 3.
60 1 00 48
41 48 1 00
132 JOURNAL OF ECONOMIC EDUCATION
Tolerance-to-Cheating Index, Obtained from Data Cell Averages
Country High school Undergraduate Postgraduate
Russia (Moscow) 2.94 2.82 3.73
Russia (Province) 4.04 3.09 3.33
Israel — 4.25 —
Netherlands 3.33 4.43 5.98
United States — 6.50 7.00
We confronted these empirical TCIs with the predicted values from our
ordered probit model (Table 4). Comparison of Tables 3 and 4 shows that our
model provided a reasonable, although by no means perfect, approximation to
the data. The standard errors in Table 4 are relatively small, showing a fair
amount of accuracy. The two preliminary conclusions mentioned earlier were
confirmed: first, Russian students were most tolerant of cheating, then Israeli and
Dutch students were, whereas students from the United States definitely did not
like cheaters; second, high school students were more tolerant of cheating than
undergraduates were, who in turn were more tolerant than postgraduates, with
the exception of high school students in provincial Russia.
CHEATING AND CORRUPTION
We applied the TCI concept to test the idea—mentioned earlier—that a link
existed between cheating and corruption, because both depend on similar cultur-
al factors. A widely used indicator of perceived corruption is the so-called Trans-
parency International Corruption Perception Index, annually updated for more
than 50 countries by Transparency International.6The rankings in 1997 for the
four countries in our study were Russia 2.27, United States 7.61, Israel 7.97, and
We re-estimated our model using the 1997 corruption index instead of the five
country dummies, together with three educational levels and the dummy for high
school students in provincial Russia. Because the rankings of countries using the
Spring 2002 133
Tolerance-to-Cheating Index Predictions
Country High school Undergraduate Postgraduate
Russia (Moscow) 2.61 (0.30) 2.92 (0.28) 3.74 (0.33)
Russia (Province) 3.98 (0.41) 3.08 (0.29) 3.92 (0.37)
Israel 3.91 (0.51) 4.28 (0.50) 5.28 (0.61)
Netherlands 4.07 (0.39) 4.45 (0.38) 5.47 (0.48)
United States 5.84 (0.55) 6.30 (0.51) 7.45 (0.54)
Note: Standard errors are in parentheses.
Tolerance-to-Cheating Index Predictions, Corruption Model
Country High school Undergraduate Postgraduate
Russia (Moscow) 2.53 (0.49) 3.07 (0.47) 3.92 (0.58)
Russia (Province) 4.00 (0.72) 3.07 (0.47) 3.92 (0.58)
Israel 4.02 (0.66) 4.69 (0.67) 5.79 (0.83)
Netherlands 4.34 (0.71) 5.04 (0.72) 6.19 (0.87)
United States 3.92 (0.64) 4.57 (0.65) 5.66 (0.81)
Note: Standard errors are in parentheses.
corruption index and using the TCI index were different, one should not expect
a very good fit. Nevertheless the fit was reasonable. The TCI predictions (Table
5) were comparable but certainly not the same as those in Table 4. The results did
not appear to contradict the cultural theory of cheating.
To what extent does the attitude to social behavior patterns vary among coun-
tries? This question is important in understanding institutional development and
reform design. However, not much research is devoted to this topic.7We have
tried to contribute to this literature by comparing the attitude of students toward
cheating in four countries.
Our study shows that students have a different attitude toward cheating
depending on where they live and that a student’s opinion also depends on his or
her level of education. We discussed several possible explanations of the results.
Our questionnaire characterized the attitude toward cheating by a three-dimen-
sional vector of attitudes toward cheaters, assistants, and informers. However, we
show that a scalar indicator (the TCI) is sufficient to capture the essence of tol-
erance of cheating. The index can be used, for example, to compare deterrence
mechanisms used in different countries.
Another hypothesis that was partially checked asserts a link between cheating
and corruption from common cultural roots. More work is needed to check this
and related hypotheses.
1. It would be of interest to replace the education dummies by a one-dimensional measure of edu-
cation, say years of schooling. This would be smoother and more informative, but the required
data were not available to us. Similarly, one could attempt to replace country dummies by rel-
evant descriptive statistics for the countries concerned, for example, per capita gross domestic
product (GDP) or the unemployment ratio. We did not do this, although we made one small
attempt in this direction by considering a corruption index.
2. We ignored the fact that the errors may not have been independent between answers: If a
respondent had a very negative view on cheating, he or she would be negative on Aand B(the
cheaters), but positive on C(the informer). To account for this dependence would have required
estimation of a multivariate ordered probit model. Such an approach was beyond our purpose
in this article. The possible dependence did not affect the consistency of our estimates, although
it did affect their efficiency. However, estimates from a multivariate regression model showed
that the differences are very small.
3. This supposition was supported by interviewing a number of Russian students who were cur-
rently studying for a Ph.D. in the United States. Of course, the statement needs further proof.
4. In the former USSR, every student used to belong to a permanent group of about 30 students.
This group stayed together for several years, taking the same academic program with only small
variations. Today, most Russian universities still use this system. Solidarity between students in
the group is high, and someone who informs officials about cheating is strongly condemned by
5. See also Davis, Noble, Zak, and Dreyer (1994), who compare United States and Australian stu-
dents in terms of their learning-oriented and grade-oriented behavior.
6. The index is available on the Web site www.transparency.de; also see Bardhan (1997). The
higher the corruption index, the lower is the corruption level.
7. An exception is Shiller, Boycko, and Korobov (1991) who compared Moscow and New York
inhabitants in their attitudes toward a free market. Differences were found to be not very signif-
134 JOURNAL OF ECONOMIC EDUCATION
Anderson, T. W. 1984. An introduction to multivariate statistical analysis. 2nd ed. New York: John
Arthur, W. B. 1988. Self-reinforcing mechanisms in economics. In P. W. Anderson, K. Arrow, and D.
Pines, eds., The economy as an evolving complex system, 9–31. Santa Fe, N. M.: Addison-Wesley.
Bardhan, P. 1997. Corruption and development: A review of issues. Journal of Economic Literature
35 (September): 1320–46.
Becker, G. 1968. Crime and Punishment: An economic approach. Journal of Political Economy 76
Bunn, D. N., S. B. Caudill, and D. M. Gropper. 1992. Crime in the classroom: An economic analy-
sis of undergraduate student cheating behavior. Journal of Economic Education 23 (Summer):
Davis, S. F., L. M. Noble, E. N. Zak, and K. K. Dreyer. 1994. A comparison of cheating and learn-
ing/grade orientation in American and Australian college students. College Student Journal 28:
Kadane, J. B. 1999. An allegation of examination copying. Chance 12 (3): 32–36.
Kerkvliet, J. 1994. Cheating by economics students: A comparison of survey results. Journal of Eco-
nomic Education 25 (Spring): 121–33.
Kerkvliet, J., and C. L. Sigmund. 1999. Can we control cheating in the classroom? Journal of Eco-
nomic Education 30 (Fall): 331–43.
Maddala, G. S. 1983. Limited-dependent and qualitative variables in econometrics. Cambridge:
Cambridge University Press.
Mixon, F. G., Jr. 1996. Crime in the classroom: An extension. Journal of Economic Education 27
North, D. 1997. Institutions, institutional change and economic performance. New York: Cambridge
Nowell, C., and D. Laufer. 1997. Undergraduate student cheating in the fields of business and eco-
nomics. Journal of Economic Education 28 (Winter): 3–12.
Polterovich, V. 2000. Institutional traps. In L. R. Klein and M. Pomer, eds., The new Russia: Eco-
nomic transition reconsidered, chap. 6. Stanford, Calif.: Stanford University Press.
Shiller, R. J., M. Boycko, and V. Korobov. 1991. Popular attitudes toward free markets: The Soviet
Union and the United States compared. American Economic Review 81 (June): 385–400.
Tanzi, V. 1997. Corruption around the world: Causes, consequences, scope and cures. IMF Staff
papers 45 (December): 559–94.
Spring 2002 135