Economics Network IREE Virtual Edition

The Determinants of Cheating by High School Economics Students: A Comparative Study of Academic Dishonesty in the Transitional Economies

Paul W. Grimes and Jon P. Rezek
International Review of Economics Education, volume 4, issue 2 (2005), pp. 23-45

Up: HomeLecturer ResourcesIREEVolume 4 Issue 2

Abstract

Secondary school students in six transitional economies, Belarus, Croatia, Kyrgyzstan, Lithuania, Russia and Ukraine, along with students in the USA, were surveyed about academic cheating. Regardless of geographic location, a substantial majority of all students reported that they had cheated on an exam or course assignment. In general, however, the percentages of students who reported that they had cheated and that they would assist others to cheat were higher in the transitional economies than in the USA. A bivariate probit regression model was estimated to determine the factors that contribute to the probability of cheating. The results indicated that the most consistently significant determinants were personal beliefs about the ethics and social acceptability of cheating and various attributes of the classroom environment. With the exceptions of Lithuania and Ukraine, students in each transitional economy had a higher probability of cheating relative to students in the USA, ceteris paribus. The relative differences ranged from 8.9% for Belarus up to 17.1% for Croatia. For Russia, the difference was a relatively high 15.4%. These and other results suggest that researchers must be extremely careful in making cross-national comparisons of student outcomes in the transitional zone if cheating trends are ignored.

JEL Classification: A22

Introduction

During the past several years, a seemingly unprecedented number of financial scandals involving corporate leaders and public officials have captured headlines around the world. Most of these incidents involved overt acts of dishonest behaviour, falsification of records, and lying to regulators and the general public. Some US critics argue that such incidents are a direct result of a serious deterioration in the ethical standards to which students are held in the nation's schools (Zelizer, 2002). This position is supported by a growing body of evidence which reveals that academic cheating is a widespread and common classroom activity. For example, a series of surveys revealed that the number of US high school students who reported cheating on an exam during the previous academic year increased from 61% in 1992 to 74% in 2002 (Josephson Institute of Ethics, 2002). These contemporary numbers are significantly higher than those recorded for previous generations of students: in a study spanning three decades, Schab (1991) reported a cheating rate of only 33.8% for US high school students in 1969. Furthermore, it is clear that academic dishonesty is not restricted to students in the USA, but rather is a universal phenomenon in secondary school classrooms throughout the world. Recent empirical studies show that cheating by secondary school students is prevalent in all four corners of the globe, including such far-flung nations such as Australia, Germany, Costa Rica, Austria (Waugh et al., 1995), Japan, South Africa (Burns et al., 1998), and Morocco (Benmansour, 2000).

The prevalence and extent of cheating has a number of implications for comparative education researchers. Perhaps the most important are the questions that can be raised about the validity of test scores in environments where cheating is commonplace. Education researchers routinely rely on standardised instruments administered in native classroom settings to measure cognitive outcomes, yet almost no attention is paid to the high probability that some students cheat when taking teacher-administrated exams. If the distribution and success of cheating are evenly distributed across treatment and control classrooms, the resulting test scores will overstate the true cognitive performance of students, but the net effect of cheating on measured outcomes would not bias comparisons across groups. However, if cheating is more (less) prevalent and successful in one group of students relative to another group, then comparisons between the groups are biased and, therefore, potentially unreliable. Logically, the probability that the incidence of cheating varies between aggregated samples of students is greatest when the samples are drawn from different institutional contexts and cultural settings. Thus, for researchers in the comparative education field who use classroom data drawn from different nations, it is important to recognise and understand how academic dishonesty may influence measured outcomes.

Since the mid-1990s, the National Council on Economic Education (NCEE), a US non-profit organisation, has been actively involved in producing and delivering economic literacy programmes in eastern Europe and the former Soviet republics of central Asia as a complement to the market reforms which are taking place in these transitional economies. The NCEE's programmes, funded through major grants from the US Departments of State and Education, include teacher training initiatives designed to improve the quality and extent of economic instruction throughout the elementary and secondary school curricula in the region. Empirical evaluations have been undertaken to measure the effect of these efforts on student understanding of market economics (Walstad, 1997; Spiro, 1998; Grimes and Millea, 2001) and, naturally, cross-national comparisons of the results have been made (Walstad and Rebeck, 2001; Watts and Walstad, 2002). Casual empiricism by economic educators who deliver teacher training in the transitional economies suggests that the historical and cultural societal emphases on collective welfare in the transitional zone result in a greater tendency for students to cheat at school relative to students in the USA.(note 1) These observations are corroborated by a recent empirical study of personal attitudes by Magnus et al. (2002), which indicates that a higher degree of personal tolerance of academic dishonesty exists in Moscow and provincial Russia relative to the USA. If such observations and empirical findings are correct, the international comparisons of student achievement, using test scores, across the transitional economies must be interpreted with great caution.

To understand better how academic dishonesty can vary across recently defined national boundaries, this study examines the incidence and determinants of academic dishonesty by secondary school students in several transitional economies where the NCEE has been actively engaged. High school students in the USA serve as the base comparison group. Specifically, an empirical model is estimated to reveal those factors that are associated with the probability that secondary school students cheat in their economics class. The procedures used to collect the data and construct the investigative sample are discussed in the next section. This is followed by a presentation of the empirical model and an overview of the results. The concluding section summarises the major findings and offers recommendations for future research.

The investigative sample

The transitional economies

During the summer of 2000, the NCEE sponsored a workshop at Indiana University which brought together a number of economic educators from several transitional economies in eastern Europe and central Asia with economic education researchers from the USA. Educators from six of the transitional economies represented at the workshop agreed to participate in this investigation by translating and administering survey instruments to secondary school students in their home nations. Thus, students from Belarus, Croatia, Kyrgyzstan, Lithuania, Russia and Ukraine were included in this study. Data were also collected from high school students in the USA by US high school economics teachers who had participated in an NCEE-sponsored study tour of Russia.(note 2) All data were collected during the 2001 and 2002 academic years.

All of the transitional economies included in this study are relatively young national entities. Five of the six nations represented are former republics of the Soviet Union which gained their independence in late 1991: Belarus, Kyrgyzstan, Lithuania, Russia and Ukraine. Likewise, Croatia is a former state of the recently dissolved Yugoslavia, a socialist nation politically allied with the Soviet Union since the end of the Second World War. Given this shared history, today the nations within the sample have relatively similar educational institutional structures. Most follow the Russian model where 9 or 10 years of compulsory basic education begins at age 6. Basic education is followed by 2 or 3 years of secondary education which may consistent of general studies (in regular schools or gymnasiums) or a more specialised curriculum (in professional schools or technicums). Students complete their secondary education at about 17 or 18 years of age and are then eligible for advanced higher education at several different types of institutes, colleges and universities.

Each of the independent nations that emerged from the fall of the Soviet Union has experienced its own unique hardships in dismantling a command and control economy and replacing it with one based on market principles. Of the nations in the sample, Lithuania has made the most progress in terms of economic reform. Most Lithuanian industries have already been privatised and the nation, along with its Baltic neighbours, Estonia and Latvia, will be become a member of the European Union in the next round of expansion. Kyrgyzstan has also made significant progress in carrying out economic reform, but due to its geographic location and limited natural endowments, it remains a poor country with an economy dependent upon the production of basic agricultural and mineral commodities. Both Belarus and Ukraine have significant potential for economic development; however, in both nations, internal political divisions have severely limited the extent and effectiveness of structural reforms. Consequently, Belarus and Ukraine both remain dependent on Russia for energy and other essential needs. As for Russia, rich in natural resources, it continues to struggle with extensive market reforms which are hampered by an outdated capital infrastructure and a slowly evolving legal system. Russia continues to rely on the export of oil, gas, timber and other basic commodities, as its primary source of hard currency in the global marketplace. The only non-former Soviet Republic in the sample, Croatia, is a heavily industrialised nation whose reforms have been hindered by a decade of ethnic division and armed conflict throughout the Balkans region. As these conflicts subside, the economic outlook for Croatia is relatively bright given its close proximity to western European markets.

In addition to the shared history of socialist rule and the current struggle for economic reform, the nations in the sample also share the problem of overt public corruption. The social and political upheaval of the past decade naturally created opportunities for those in positions of authority to exploit their public positions for personal gain. Corruption and crime are major stumbling blocks to lasting structural change.(note 3) The educational systems in the transitional economies have not been immune to these problems. The western press has reported on numerous incidents involving the selling of grades and even diplomas and degrees by teachers and administrators (e.g. MacWilliams, 2001). Although each country in the survey has unique features and faces unique issues, clearly these common environmental factors, along with the inherent societal emphases on collective action, create great burdens for instructors to maintain ethical standards concerning individual performance in their classrooms.

The survey

The survey instrument was a slightly modified version of the one employed by Grimes (2003) in a comparative study of college students' personal perceptions of dishonesty in business and academics. Specifically, the survey instrument asked the respondents to answer a series of questions concerning their attitudes and experiences with cheating while in secondary school. The students were also asked to supply a number of personal demographic characteristics about themselves and their families. Given the nature of the questions, all students were guaranteed confidentiality and anonymity through a voluntary student release and consent agreement which was presented and read to each subject prior to completing the survey.(note 4) Each of the international research partners translated the agreement statement and survey instrument into their native language and local dialect, and each oversaw the administration and collection of the surveys. All of the students who completed the survey were enrolled on an economics course or on a closely related course taught by an economics instructor during the term in which the survey was given. The completed survey instruments were returned to the author and the responses compiled into a database for analysis.

The sample consisted of 1,097 students: 723 from the transitional economies and 374 from the USA. Table 1 provides a brief demographic profile of the students included in this investigative sample.

Table 1 Profile of student sample by nation

Nation (N) Mean age Percentage
male
Parents'
educationa
Percentage
employedb
Religiosityc
Transitional
economies (723)
17.27 32.78 54.64 15.63 24.62
Belarus (52) 16.50 15.38 88.46 26.92 13.46
Croatia (371) 17.87 23.45 40.70 11.59 38.54
Kyrgyzstan (28) 19.18 60.71 10.71 46.43 14.29
Lithuania (47) 17.55 61.70 55.31 6.38 8.51
Russia (47) 16.77 42.55 57.44 19.15 8.51
Ukraine (178) 16.01 42.70 79.87 17.42 8.99
USA (374) 16.35 48.13 57.48 48.40 52.94
Full sample (1097) 16.96 38.01 55.61 26.80 34.28

a Percentage reporting that one or both parents have at least a first college degree (bachelor's).
b Working full time or part time.
c Percentage reporting that they regularly attend organised religious services.

Several interesting differences between students in the transitional economy sub-samples and the US sub-sample can be seen in Table 1. First, the surveyed secondary school students in the transitional economies, on average, were almost 1 year older than their US cohorts. Second, a significant variation in the gender mix of students across the transitional economies was revealed in the collected data. The number of male students in the sub-samples ranged from 15.38% in Belarus to 61.70% in Kyrgyzstan, while the US sub-sample contained a more evenly balanced 48.13% males. These differences in the age and gender distributions were apparently artefacts of the different institutional arrangements prevailing in each country with respect to the schools that were surveyed.(note 5)

More importantly, significant differences were also revealed with respect to parental education and personal employment. In four of the six transitional economies, a smaller number of students, relative to US students, reported that at least one parent held a college degree. However, in Belarus and Ukraine a significantly greater percentage of student respondents lived in homes where at least one parent was a college graduate. With respect to work experience, students in the transitional economies uniformly reported a lower rate of employment than their US cohorts. Overall, only 15.63% of surveyed secondary school students in the transitional economies reported holding a job compared to the 48.40% of US high school students who so reported.

Finally, significant differences in religious behaviour were found across the nations in the sample. The percentage of US students who reported regular attendance at organised religious services was more than twice that reported by students in the transitional economies when considered as a group. However, the percentage of students in Lithuania, Russia and Ukraine who reported regular religious attendance was only about one-sixth of the 52.94% found for the US sub-sample. This tremendous difference is clearly a lingering result of the Soviet era of secular state rule and overt religious persecution.

In addition to the demographic variations across the national sub-samples, the survey also revealed a number of differences in the prevalence and perception of academic dishonesty in the classroom. Table 2 reports various measures of the incidence of cheating by the surveyed students and their perceptions of this type of dishonest behaviour. Again, a number of interesting cross-national differences were apparent in the data.

Table 2 Self-reported incidence and perceptions of academic dishonesty by nationa

Nation Have
cheated
Asked
to cheat
Would
assist
Fear of
punishment
Ethically
wrong
Socially
acceptable
Transitional
economies
87.41 88.93 83.81 72.06 49.79 54.77
Belarus 63.46 38.46 42.31 69.23 71.15 53.85
Croatia 92.99 92.99 90.03 67.39 54.72 50.94
Kyrgyzstan 96.43 92.86 89.29 82.14 71.43 39.29
Lithuania 76.60 91.49 70.21 87.23 48.94 38.30
Russia 93.62 87.23 80.85 61.70 34.04 70.21
Ukraine 84.27 94.38 86.52 79.78 34.27 65.73
USA 69.52 79.14 44.65 77.00 68.72 46.52
Full sample 81.31 85.60 70.46 73.75 56.24 51.96

a Percentage reporting in the affirmative.

The second column of Table 2 reports the percentage of students in each relevant sub-sample that responded in the affirmative to the question, 'Have you ever cheated on an exam or course assignment?' The results for this question suggest that cheating is a universal, and apparently common, behaviour, as a majority of students in each nation surveyed admitted to having cheated previously.(note 6) The nearly 70% of US students who reported cheating is consistent with the recent rates revealed by other researchers and previous US surveys (see citations noted earlier). With the exception of Belarus, the percentage of high school students self-reporting that they had cheated was substantially higher in each of the transitional economies when compared to the USA. And, in three nations, Croatia, Kyrgyzstan and Russia, the percentage of students self-reporting that they had engaged in cheating exceeded a startling 92%.

In most of the nations surveyed, cheating appears to be a communal act. This is seen in the relatively high percentages of students who revealed that they had been asked to cheat by classmates, as reported in the third column of Table 2. Again, with the exception of Belarus, a vast majority of students in each nation, including the USA, reported that they had been approached in the past to cheat by others. Furthermore, many students responded that they would assist a fellow student in cheating if asked (refer to the fourth column). Only in Belarus and the USA did less than a majority of students indicate that they would help others to cheat.

Survey respondents were asked several questions concerning their perceptions of academic cheating. Apparently, high school students universally recognise cheating as a breaking of the rules of proper academic conduct, given that a majority of respondents across all of the national sub-samples responded that they feared the punishment of being caught. Even though a majority of students in each nation surveyed feared the punishment associated with cheating, in most cases a substantially smaller number of students believed that cheating was ethically wrong. This is seen by comparing the fifth and sixth columns of Table 2. Furthermore, when asked if cheating was a socially acceptable behaviour, a majority of students in four of the six transitional economies said yes it was. However, in Kyrgyzstan, Lithuania and the USA, less than a majority of students so indicated.

Taken together, the data reported in Tables 1 and 2 suggest that cheating is widespread and common throughout all of the nations included in the overall investigative sample and that, even though students fear the punishment of being caught, cheating is not consistently viewed as ethically wrong but is often seen as socially acceptable behaviour. This gross evidence also indicates that cheating may be more pronounced in the secondary schools of the transitional economies, when viewed as a group, relative to the USA. However, the data also show that systematic variations in the incidence and perceptions of academic cheating by students do exist between nations. To sort out the patterns of these differences, a more sophisticated analysis was undertaken using the individual student as the unit of observation.

Bivariate probit model and empirical results

The model

Many previous studies of student cheating behaviour rely only on comparisons of means between groups (e.g. Schab, 1980) or simple correlation techniques (e.g. Barnett and Dalton, 1981; Burns et al., 1998). However, recent researchers have used individual-level data to estimate behavioural models of the choice that students make when facing the decision to cheat. For example, Benmansour (2000) employed hierarchically ordered regression analysis to model the effect of pre-existing student beliefs on the decision to cheat, and Murdock, Hale and Weber (2001) estimated a standard logit equation to predict whether middle school students were cheaters or non-cheaters based on a vector of social and academic student characteristics. Likewise, at the college level, Bunn, Caudill and Gropper (1992) and Mixon (1996) used logit analysis to estimate the effect of various student characteristics and experiences on the decision of economics students to cheat in introductory courses.(note 7)

For this analysis, consider Kerkvliet's (1994) adaptation of Becker's model of crime to academic cheating. In the model, rational actors consider all costs and benefits of their options before making a utility-maximising decision whether to cheat or abstain from cheating. Kerkvliet shows that dishonest behaviour can be the rational result of an individual's internal calculus if the expected utility of cheating exceeds the utility of not cheating.

Let the utility derived from academic honesty be a function of the individual's expected benefits and personal characteristics:

Uinc = Ui(B˜ | Z)
(1)

where Ui is student i's utility without cheating, B˜ is a vector of the expected benefits without cheating, and Z is a vector of personal characteristics. Similarly, let the expected utility of cheating be a function of the individual's expected benefits and personal characteristics, but let it also be a function of their subjective judgement regarding the probability of being detected and the expected costs of cheating – in terms of both sanctions and psychic costs. That is,

Ui = (1 – pi)Ui(B | Z) + piU(B,C | Z)
(2)

where Ui is student i's utility with cheating, pi is the probability of detection, B is a vector of the benefits from cheating, and C is a vector of costs associated with detection. A student cheats if:

Uic > Uinc
(3)

Let CHEAT be a dichotomous variable representing the student's truthful response to a direct inquiry as to whether they have ever engaged in cheating. If equation 3 has held at some point in their academic career, then the student has cheated, and hence CHEAT equals 1. If equation 3 has never held, then the student has never cheated and CHEAT equals 0. Given that a student's behaviour is related to their personal characteristics, the costs and benefits of their decision, and their perception of the risk of detection, then the probability of cheating can be generalised as:

PROB(CHEAT = 1) = PROB(Uic > Uinc | x) = F(x′β)
(4)

where F(.) is a cumulative density function, x′ is a vector with elements B, C, Z and p, and β is a vector of parameters to be estimated.

In this application, the cost vector is modelled as containing four elements derived from students' responses to the survey instrument. These four dichotomous variables (FEAR, WRONG, ACCEPTABLE and POLICY) are detailed in Table 3. Students who express a fear of the sanction or who believe cheating is ethically wrong are assumed to incur larger psychic costs than other students and are less likely to engage in the dishonest behaviour.The coefficients corresponding to the FEAR and WRONG dummy variables are, therefore, expected to carry negative signs. Conversely, students who believe cheating is socially acceptable are assumed to be less likely to suffer from the psychic costs associated with cheating and are thus more likely to engage in academic dishonesty. The coefficient on the ACCEPTABLE dummy variable is therefore expected to carry a positive sign. Finally, students who are familiar with their institution's policies regarding cheating are assumed to have more information about the negative consequences; therefore, the POLICY dummy variable is expected to carry a negative coefficient.

The sole benefit to cheating in this framework is the improvement in academic performance. Students with a high GPA have achieved a significant amount of academic success; the marginal gain from cheating for these students is expected to be minimal. The potential benefit of cheating is larger for students of lower academic standing and hence the probability of dishonesty is higher within this group. The GPA coefficient is therefore expected to carry a negative sign.

The Becker model highlights the importance of the agent's perceptions regarding the probability of detection in the decision-making process. It is assumed that students'perceptions of the probability of detection are related to three elements derived from the questionnaire. These three dichotomous variables (CAUGHT, WITNESS and ASKED) are detailed in Table 3. Students who have observed others being caught cheating are likely to inflate the probability of detection and be less likely to cheat themselves. Catching perpetrators presumably serves as a deterrent to other potential cheaters; therefore, the coefficient on the CAUGHT dummy is expected to carry a negative sign. Conversely, students who have witnessed others cheating or have been asked to cheat themselves are likely to discount the probability of detection, be more familiar with the practice and be more likely to engage in the behaviour themselves. The coefficient corresponding to the WITNESS and ASKED dummy variables are therefore expected to be positive.

Table 3 Specification and descriptive statistics of variables for full sample (N = 1097)

Variable label Definition Mean Standard
deviation
Dependent variables
CHEATED Student response to the question:'Have you ever cheated on an exam or course assignment?' 1 = Yes; 0 = No0.813 0.390
ASSIST Student response to the question:'If you were asked to help someone cheat on an exam or course assignment, would you assist them?' 1 = Yes; 0 = No0.705 0.456
Independent variablesa
Personal characteristics (Z)
AGE [–] Student's age in years 16.957 1.169
GENDER [+] Student's sex; 1 = Male; 0 = Female 0.380 0.486
PARENTS' ED [–] Parental education; 1= One or both parents attained bachelor's degree or higher; 0 = No college degree0.556 0.497
RELIGIOSITY [–] 1 = Student regularly attends religious services; 0 = Otherwise0.343 0.475
WORK [+] 1 = Student works full- or part-time; 0 = Otherwise 0.268 0.443
Benefits of cheating (B)
GPA [–] Student's overall college grade point average converted to 4-point scale (A = 4 to F = 0)3.221 0.552
Costs of cheating (C)
FEAR [–] Student response to the question: 'Do you have a fear of punishment if caught cheating on an exam or course assignment?' 1 = Yes; 0 = No0.738 0.440
WRONG [–] Student response to the question: 'Do you consider cheating to be ethically or morally wrong?' 1 = Yes; 0 = No 0.562 0.496
ACCEPTABLE [+] Student response to the question: 'Is cheating socially acceptable?' 1 = Yes; 0 = No0.520 0.502
POLICY [–] 1 = Student reports familiarity with school's policy on cheating; 0 = Otherwise0.724 0.447
Probability of detection (p)
CAUGHT [–] 1 = Student reports observing others being caught cheating; 0 = Otherwise0.798 0.402
WITNESS [+] 1 = Student reports witnessing cheating by others; 0 = Otherwise 0.926 0.262
ASKED [+] 1 = Student reports being asked to help cheat on an exam or course by another student; 0 = Otherwise0.856 0.351

[ ] Expected sign of probit coefficient.

a The probit equation also contains categorical variables for national location.

Finally, personal characteristics may also affect a student's likelihood of cheating. From the survey instrument we posit a personal characteristic vector (Z) consisting of the five elements described in Table 3: age, gender, religiosity, employment status and parents' educational attainment. While research suggests that males have a greater propensity to cheat than females, the expected signs on the other coefficients are somewhat ambiguous. For instance, the greater maturity that comes with age may generate a lower propensity to cheat for older students. However, the dichotomous dependent variable CHEAT is the response to a question that is cumulative in nature, making it more likely for older students to answer in the affirmative simply because of the larger number of opportunities to cheat. Also, the moral and ethical convictions often associated with religious students would suggest an inverse relationship between religiosity and cheating. However, it is feasible that religious students are more honest in their responses, generating an apparent direct relationship between cheating and religiosity.

In this application, the behavioural model of a student's decision to cheat is posited with the following reduced-form relational regression equation:

(5) CHEAT = α + β1AGE + β2GENDER + β3GPA + β4PARENTS' ED + β5RELIGIOUSITY + β6WORK + β7FEAR + β8WRONG + β9ACCEPTABLE + β10POLICY + β11ASKED + β12WITNESS + β13CAUGHT + β14TRANSITIONAL ECONOMIES + ε

The empirical specification for each variable used to estimate the equation is presented in Table 3 along with its mean and standard deviation for the full sample. Based upon the empirical literature noted above, the expected sign for each independent variable is also reported in Table 3. The model was estimated employing two different specifications for the TRANSITIONAL ECONOMIES: once with a single dummy variable for these six nations as a group, and then using a vector of dummy variables, one for each nation. In both cases, the omitted reference group was the USA.

The survey offered two measures of cheating behaviour (CHEAT): the first was each student's response to 'Have you ever cheated on an exam or course assignment?' and the second was each student's response to 'If you were asked to help someone cheat on an exam or course assignment, would you assist them?' In both cases, the decision to cheat was presented as a dichotomous choice: to cheat or not to cheat. Therefore, equation (1) can be estimated twice: first using the categorical dependent variable CHEATED, which reflects the answer to the first question, and then using the categorical dependent variable ASSIST, which reflects the response to the second question. Clearly, CHEATED and ASSIST are related to each other through the latent processes that determine a student's propensity to engage in academic misconduct. The two specifications could be estimated consistently by individual single equation probit methods; however, this would be inefficient in that it ignores the correlation which must exist between the error terms. Thus, both specifications of equation (1) were estimated jointly using the bivariate probit analysis technique, which takes advantage of the correlation between the disturbances of each specification 'in the same spirit as the seemingly unrelated regressions model' (Greene, 2003, p. 710). The model was estimated using the Limdep (Greene, 1999) econometrics software package, which generates the full information maximum likelihood estimates and calculates the standard and conditional marginal effects for each independent variable.

The results

The estimated full information maximum likelihood bivariate probit coefficients are reported in Table 4 and the corresponding direct marginal effects for each independent variable are presented in Table 5. The significant rs and corresponding Wald statistics indicated that the disturbances for the CHEATED and ASSIST specifications were correlated and, thus, that the appropriate estimation technique had been employed. Overall, many of the probit coefficients obtained their expected sign and were statistically significant at standard levels. Furthermore, the estimated number of correct predictions generated by the model exceeded 77%, indicating that the equation produced a good fit of the empirical data.(note 8)

Before turning to the results for the estimated relative effects of the TRANSITIONAL ECONOMIES on the probability of cheating, there are several other findings that warrant attention and consideration. First, neither AGE nor GENDER was found to have a statistically significant effect on the probability of cheating or assisting others in cheating. This implies, at least for this sample of students, that the likelihood of academic misconduct in secondary school does not vary according to a student's age or sex, holding all else constant. The results for RELIGIOSITY are mixed. As seen in Table 4, in two of the four estimated specifications, the RELIGIOSITY coefficient enters the model with a positive and significant sign and in only one case does the coefficient obtain a statistically significant negative sign.

Table 4 Full information maximum likelihood bivariate probit results

Transitional economies By nation
Variable CHEATED ASSIST CHEATED ASSIST
CONSTANT –0.344 –1.689 1.348 0.065
(0.407) (2.199)b (1.290) (0.072)
AGE 0.030 0.065 –0.077 –0.05
(0.650) (1.489) (1.321) (0.968)
GENDER 0.017 –0.030 0.078 0.023
(0.156) (0.299) (0.668) (0.218)
GPA –0.236a –0.159c –0.268a –0.115
(2.460) (1.798) (2.699) (1.226)
PARENTS'ED –0.165c –0.026 –0.087 0.074
(1.707) (0.270) (0.780) (0.718)
RELIGIOSITY 0.342a –0.053 0.223c –0.194c
(2.960) (0.509) (1.793) (1.772)
WORK –0.107 0.342 –0.068 0.079
(0.951) (0.309) (0.574) (0.691)
FEAR –0.003 –0.056 0.071 –0.012
(0.027) (0.502) (0.548) (0.104)
WRONG –0.394a –0.392a –0.443a –0.431a
(3.429) (3.750) (3.621) (3.814)
ACCEPTABLE 0.272b 0.312a 0.296a 0.337a
(2.657) (3.229) (2.812) (3.382)
POLICY 0.012 0.113 0.098 0.154
(0.104) (1.103) (0.844) (1.406)
ASKED 0.8562b 1.170a 0.906a 1.095a
(6.926) (9.240) (6.935) (7.782)
WITNESS 0.380b –0.019 0.410b 0.077
(2.062) (0.093) (2.171) (0.371)
CAUGHT 0.342a 0.232c 0.311a 0.102
(2.852) (1.909) (2.497) (0.794)
TRANSITIONAL ECONOMIES 0.549a 0.998a
(4.550) (8.742)
BELARUS 0.525b 0.450b
(2.416) (1.944)
CROATIA 1.001a 1.493a
(5.599) (9.015)
KYRGYZSTAN 0.593c 1.467a
(1.676) (3.793)
LITHUANIA 0.303 0.571a
(1.181) (2.446)
RUSSIA 0.908a 0.784a
(2.544) (3.183)
UKRAINE 0.107 0.829a
(0.629) (5.074)
r 0.474a 0.469a
(8.471) (8.089)
Wald statistic 71.610 65.420
Log-likelihood –892.560 –868.741
Percentage correct predictions 77.301 77.301

( ) Absolute value of t statistic.
a Statistically significant at the .01 level, two-tailed test.
b Statistically significant at the .05 level, two-tailed test.
c Statistically significant at the .10 level, two-tailed test.

Table 5 Direct marginal effects derived from bivariate probit coefficients

Transitional economies By nation
Variable CHEATED ASSIST CHEATED ASSIST
CONSTANT –0.061 –0.079 0.229 0.003
AGE 0.005 0.003 –0.013 –0.002
GENDER 0.030 –0.001 0.013 0.001
GPA –0.042 –0.007 –0.045 –0.005
PARENTS'ED –0.029 –0.001 –0.015 0.003
RELIGIOSITY 0.060 –0.003 0.038 –0.009
WORK –0.019 0.002 –0.011 0.004
FEAR –0.001 –0.003 0.012 –0.001
WRONG –0.069 –0.018 –0.075 –0.019
ACCEPTABLE 0.048 0.015 0.050 0.015
POLICY 0.002 0.005 0.017 0.007
ASKED 0.152 0.055 0.154 0.048
WITNESS 0.067 –0.001 0.070 0.003
CAUGHT 0.060 0.011 0.053 0.005
TRANSITIONAL ECONOMIES 0.097 0.047
BELARUS 0.089 0.020
CROATIA 0.171 0.065
KYRGYZSTAN 0.101 0.064
LITHUANIA 0.051 0.025
RUSSIA 0.154 0.034
UKRAINE 0.018 0.036

From these results, it appears that regular attendance at religious services is positively associated with cheating, and negatively associated with assisting others to cheat, when all else is held constant. Interestingly, these results are not at odds with all previous empirical studies. Bruggeman and Hart (2001), in a study comparing students enrolled in private religious high schools to students in secular public schools, found that those in religious schools were 'more likely to yield to tempting circumstances and exhibited a higher incidence of cheating than the secular schoolchildren' (p. 341). Such results may be a manifestation of the self reported nature of survey data. Regardless, more detailed work is needed to clarify the role of religious beliefs and practices in the tendency to commit and report academic misconduct.

The results indicate that there is a significant negative relationship between a student's grade point average (GPA) and the probability of cheating. This suggests that those students who have already attained high grades are less compelled to undertake dishonest acts to maintain or improve their academic record. From Table 5, the estimated marginal effects indicate that a 1 unit improvement in a student's GPA is associated with about a –4% change in the probability that the student self-reported having cheated on an exam or class assignment. Likewise, a 1 unit increase in a student's GPA lowers the probability of assisting another student to cheat by something less than 1%.

Although a majority of students reported that they feared the punishment of being caught cheating, FEAR was not found to reduce significantly the probability of cheating or assisting others in cheating. For teachers and administrators this may be a disturbing finding when coupled with the result that a student's knowledge of the school's academic misconduct POLICY also had no significant effect on cheating behaviour. However, the personal belief that cheating is ethically wrong significantly reduced the probability of academic misconduct, while the personal belief that cheating is social acceptable significantly increased the probability. From Table 5, the results show that those who believed cheating is WRONG had about a 7% lower probability of cheating and about a 2% lower probability of assisting others in cheating. Likewise, those who believed cheating to be socially ACCEPTABLE had about a 5% higher probability of cheating and about 1.5% higher probability of assisting others to cheat.

The classroom environment and behaviour of other students was found to have an effect on student cheating behaviour. Specifically, if students had been ASKED in the past by classmates to cheat, they were more likely to cheat themselves and to help others cheat. Also, if a student was a WITNESS to cheating in class, he or she was more likely to cheat themselves, ceteris paribus. Interestingly, watching classmates being CAUGHT for cheating was not found to be a deterrent, but rather it positively affected the probability of academic misconduct. This suggests that the severity of the sanctions for cheating are generally not strong enough to deter misconduct and/or that the observed probability of being caught and punished are so low relative to the rewards that observing others cheat actually encourages more cheating.

For a teacher or school administrator desiring to reduce the probability of academic miscond1uct in the secondary school classroom, the results here suggest at least three possible avenues to explore: (1) convincing students to believe that cheating is ethically wrong; (2) convincing students to believe that cheating is not socially acceptable behaviour; and (3) creating a classroom environment where students do not collaborate with each other in academically dishonest behaviour. All three of these options are difficult and likely require learning experiences at home and in the community as well as in the classroom.

Turning to the results concerning the TRANSITIONAL ECONOMIES, the second and third columns of Tables 4 and 5 reveal that when the six transitional economies in the sample are considered as a group, secondary school students in those economies had a statistically significant greater probability of cheating and assisting others to cheat, relative to their cohorts in the USA, ceteris paribus. Overall, holding all else constant, students from the transitional economies region had a 9.7% greater probability of cheating and a 4.7% greater probability of assisting classmates to cheat than their US counterparts.

The last two columns of Tables 4 and 5 report the results for equation (1) when the transitional economies are disaggregated and enter the model individually. With the exceptions of Lithuania and Ukraine, students in each transitional economy had a higher probability of cheating than students in the USA. The relative differences ranged from 8.9% for Belarus up to 17.1% for Croatia. For Russia, the difference was a relatively high 15.4%. With respect to the relative probability of assisting classmates to cheat, all of the transitional economy categorical variables entered the model with positive and significant coefficients. The marginal effects ranged from a low of 2.0% in Belarus to a high of 6.4% in Kyrgyzstan.

As an additional test of the relative differences between nations in the probability of academic misconduct, the marginal effects for the ASSIST specifications of equation (1) were re-estimated conditioned upon CHEATED = 1. The results are found in Table 6. Specifically, these marginal effects reflect the impact of each independent variable on the probability that Student A will assist Student B to cheat given that Student A has cheated before. The pattern of results for the student characteristics variables are consistent with those discussed above.

Table 6 Marginal effects on ASSIST conditioned upon CHEATED = 1

ASSIST
Variable Transitional
economies
By nation
CONSTANT –0.481b –0.056
(2.143) (0.220)
AGE 0.018 –0.010
(1.392) (0.713)
GENDER –0.010 0.002
(0.341) (0.077)
GPA –0.033 –0.019
(1.292) (0.689)
PARENTS'ED 0.002 0.026
(0.071) (0.917)
RELIGIOSITY –0.036 –0.068b
(1.190) (2.219)
WORK 0.016 0.027
(0.515) (0.829)
FEAR –0.017 –0.007
(0.506) (0.221)
WRONG –0.093a –0.100a
(3.014) (3.093)
ACCEPTABLE 0.077a 0.081a
(2.741) (2.886)
POLICY 0.033 0.039
(1.125) (1.279)
ASKED 0.297a 0.266a
(7.718) (6.504)
WITNESS –0.027 –0.001
(0.491) (0.006)
CAUGHT 0.049 0.012
(1.412) (0.343)
TRANSITIONAL ECONOMIES 0.264a
(7.915)
BELARUS 0.100
(1.557)
CROATIA 0.376a
(8.260)
KYRGYZSTAN 0.391a
(3.606)
LITHUANIA 0.148b
(2.271)
RUSSIA 0.176a
(2.584)
UKRAINE 0.233a
(4.919)

( ) Absolute value of t statistic.
a Statistically significant at the .01 level, two-tailed test.
b Statistically significant at the .05 level, two-tailed test.
c Statistically significant at the .10 level, two-tailed test.

However, living in the transitional economy region increased the probability of a cheater assisting others to cheat by an extremely large 26.4%, relative to those in the USA and holding all else constant. Likewise, when the region is disaggregated by nation, the estimated marginal effects are quite large for five of the six countries, ranging from 17.6% in Russia to 39.1% in Kyrgyzstan. Interestingly, the conditional marginal effect for Belarus is not statistically significant. These results for the conditional marginal effects that address the relative likelihood of collusive behaviour between students suggest that in most transitional economies the cultural and societal emphases on collaboration and shared experiences may result in relatively greater tendencies for students to engage in academic misconduct when compared to the US experience. These clear patterns of cross-national difference should not be overlooked by researchers trying to make cross-national comparisons of student behaviour and cognitive outcomes.

Conclusions

Drawing upon a unique set of data collected from surveys of high school students in the USA and in six transitional economies of eastern Europe and central Asia, this study examined the relative incidence and determinants of academic misconduct. Tabulations of the survey results reveal that cheating in high school is a common behaviour and that a majority of students in each nation reported that they had cheated on an exam or class assignment. However, the percentage of high school students who had cheated was greater in five of the six transitional economies when compared to the USA. The results also showed that a majority of students in each nation feared the punishment of being caught cheating, but many students also believed that cheating was socially acceptable behaviour.

To reveal the factors that influence a student's decisions to cheat and to assist others to cheat, a bivariate probit model was estimated. Among the factors found to reduce the probability of academic misconduct most significantly were the students' GPA and the personal belief that cheating is ethically wrong. Factors that were found to affect the probability of academic misconduct positively included the personal belief that cheating was socially acceptable and having been asked by fellow students to cheat. The bivariate probit results also indicated that living in a transitional economy, in most cases, increased the probability of cheating and assisting others to cheat, relative to the USA and holding all else constant. However, this effect was found to vary between nations in the transitional zone, and when the marginal impact of assisting others to cheat was conditioned upon previous cheating, the relative differences became quite large.

For comparative education researchers and others interested in student behaviour, the implication of these findings is very clear – cheating does occur in secondary school classrooms around the world and the extent of this behaviour can vary dramatically between national settings. Thus, cross-national comparisons of student outcomes must only be made with a great deal of caution if cheating behaviour is not closely monitored. Future researchers will be interested in determining the root sources of the observed differences across cultural and institutional contexts and refining the measures used to calculate the differences. These researchers will need to explore how academic dishonesty is related to regionalised cultural value systems, personal respect for authority, institutionalised grading and promotion systems, severity of punishment schemes, drop-out behaviour, peer group formation, honour codes, student selection processes and a host of others factors that vary cross-nationally. Finally, additional research is needed to determine the empirical effects of variations in cheating behaviour on measured student outcomes across national borders. The results of such research may suggest procedures and safeguards that can be invoked to improve the collection of comparative student data.

References

Barnett, D. H. and Dalton, J. (1981) 'Why college students cheat', Journal of College Student Personnel, vol. 11, pp. 545–51.

Benmansour, N. (2000) 'Motivation, satisfaction, success attributions and cheating among high school students in Morocco', Mediterranean Journal of Educational Studies, 52, pp. 83–102.

Bruggeman, E. L. and Hart, K. J. (2001) 'Cheating, lying, and moral reasoning by religious and secular high school students', Journal of Educational Research, vol. 89, pp. 340–4.

Bunn, D. N., Caudill, S. B. and Gropper, D. M. (1992) 'Crime in the classroom: an economic analysis of undergraduate student cheating behavior', Journal of Economic Education, vol. 23, pp. 197–207.

Burns, S. R., Davis, S. F., Hosino, J. and Miller, R. L. (1998) 'Academic dishonesty: a delineation of cross-cultural patterns', College Student Journal, vol. 32, pp. 590–7.

Greene, W. (1999) Limdep: Version 7.0, Plainview, NY: Econometric Software.

Greene, W. (2003) Econometric Analysis, 5th edn, Upper Saddle River, NJ: Prentice Hall.

Grimes, P. W. (2003) 'Honesty in academics and business: a cross-cultural evaluation of student attitudes', Teaching Business Ethics, vol. 49, pp. 273–90.

Grimes, P. W. and Millea, M. J. (2001) An Outcomes Evaluation of the International Economic Exchange Program in Kazakhstan: Final Report, Mississippi State, MS: Center for Economic Education and Financial Literacy.

Josephson Institute of Ethics (2002) 2002 Report Card: The Ethics of American Youth, Merina del Rey, CA: Josephson Institute of Ethics Press.

Kerkvliet, J. (1994) 'Cheating by economics students: a comparison of survey results', Journal of Economic Education, vol. 25, pp. 121–33.

Kerkvliet, J. and Sigmund, C. L. (1999) 'Can we control cheating in the classroom?', Journal of Economic Education, vol. 30, pp. 331–43.

MacWilliams, B. (2001) 'Corruption, conflict, and budget cuts afflict academe in former Soviet republics', Chronicle of Higher Education, vol. 48, 14 December, pp. A43–A46.

Magnus, J. R., Polterovich, V. M., Danilov, D. L. and Savvateev, A. V. (2002) 'Tolerance of cheating: an analysis across countries', Journal of Economic Education, vol. 33, pp. 125–35.

Mixon, F. G. (1996) 'Crime in the classroom: an extension', Journal of Economic Education, vol. 27, pp. 195–200.

Murdock, T. B., Hale, N. M. and Weber, M. J. (2001) 'Predictors of cheating among early adolescents: academic and social motivations', Contemporary Educational Psychology, vol. 26, pp. 96–115.

Nowell, C. and Laufer, D. (1997) 'Undergraduate student cheating in the fields of business and economics', Journal of Economic Education, vol. 28, pp. 3–12.

Schab, F. (1980) 'Cheating in high school: differences between the sexes (revisited)', Adolescence, vol. 15, pp. 959–65.

Schab, F. (1991) 'Schooling without learning: thirty years of cheating in high school', Adolescence, vol. 26, pp. 839–48.

Spiro, J. (1998) Sustained Effects of the International Economic Education Exchange Program in Kyrgyzstan: A Longitudinal Study, New York: Education Development Center.

Walstad, W. B. (1997) The Effects of the International Economic Education Exchange Program on Student Economic Understanding and Market Attitudes: A Research Report, Lincoln, NE: National Center for Research in Economic Education.

Walstad, W. B. and Rebeck, K. (2001) 'Teacher and student economic understanding in transition economies', Journal of Economic Education, vol. 32, pp. 58–67.

Watts, M. and Walstad, W. B. (eds) (2002) Reforming Economics and Economics Teaching in the Transition Economies: From Marx to Markets in the Classroom, Cheltenham, UK, Edward Elgar Publishing.

Waugh, R. F., Godfrey, J. R., Evans, E. D. and Craig, D. (1995) 'Measuring students' perceptions about cheating in six countries', Australian Journal of Psychology, vol. 47, pp. 73–80.

Zelizer, G. L. (2002) 'Break cheating pattern early', USA Today, 19 November 2002; online at http://www.usatoday.com/news/opinion/2002-11-19-oped-zelizer_x.htm .

Notes

[1] From the author's conversations and interviews with both the US and international participants in the Developing Skills in Evaluation Workshop held at Indiana University in July 2000. Many participants had extensive experience working with teachers and secondary school students in the transitional economies and the USA.

[2] The US sample includes students from rural, suburban and urban school districts. These data were collected in California, Minnesota and Wisconsin.

[3] Rankings of nations by the degree of corruption present in their political and economic structures routinely place Russia and the other former Soviet republics high on the list. See, for example, the survey results compiled by the Center for Corruption Research sponsored by the University of Passau and Transparency International: http://www.gwdg.de/~uwvw/icr.htm .

[4] A copy of the survey instrument can be found in Grimes (2003).The student consent and release statement and data collection procedures were developed to conform with the prevailing human subjects standards imposed by the US Federal Government. The survey instruments were reviewed and approved by the Mississippi State University Internal Review Board (IRB), which ruled that the same procedures and safeguards for anonymity had to be applied for all of the student respondents in the foreign countries participating in the survey. Interestingly, no student, either domestic or foreign, failed to provide a signed consent form. The consent statement and survey instrument were written to assure all students that their responses would in no way affect class grades. Thus, there was no cost to the students for revealing past transgressions. However, the authors acknowledge that this zero cost may also have potentially allowed some students to 'brag' about their illicit behaviour in the same manner that is observed in drinking or sexual behaviour surveys. All results should be viewed from the perspective that the respondents were juveniles and that immature behaviour may be reflected in their answers to the survey.

[5] As opposed to the general education offered by most high schools in the USA, some of the schools surveyed in the transitional economies offer highly specialised curricula. For example, in Croatia one school in the survey prepares students for careers in hospitality management while several surveyed schools in other nations offer scientific curricula. Thus, the gender mix of these schools probably reflects the local societal gender mix for each targeted career path.

[6] The wording of the survey left all student respondents with the flexibility to interpret 'cheating' according to their own cultural and institutional context. Note that the specific behaviours which actually constitute cheating may vary across institutions within each nation as well as between each nation.

[7] Several additional empirical studies that examine the cheating behaviour of college students in economics courses have appeared in the literature (e.g. Kerkvliet, 1994; Nowell and Laufer, 1997; Kerkvliet and Sigmund, 1999). However, the current study appears to be the first to explore cheating behaviour in the high school economics classroom.

[8] Various specification tests of the model were conducted, including alternative estimation procedures such as single-equation probit analysis. The primary results presented here are stable across alternative specifications of equation (1) and estimation techniques. Copies of these tests are available upon request.

Acknowledgements

Financial support received from the National Council on Economic Education through funding provided by the United States Department of Education. This project could not have been completed without the assistance of many dedicated educators and classroom teachers in the transitional economies and the USA.

Those who collected the original data include Lioubov Artamonova, Alexander Balkunov, Daira Baranova, Mikhail Chepikov, Tatiana Chilina, Thomas Eldridge, Efka Heder, Svetlana Kolova, Larysa Krasnikov, Scott Kromminga, Illia Kristo, Inga Lapina, Irina Lavruhina, Lana Mahoney, Danute Poskiene, Mark Schoenbohm and Anatoily Venger.

Special thanks are extended to Barbara DeVita and Mary Blanusa of the NCEE for their oversight and coordination of the administrative and international financial aspects of the project, and to Phillip Saunders and all of the participants of the Developing Skills in Evaluation Workshop held at Indiana University. Thanks also to Hunter Sharpe for data entry and database construction.

Helpful econometric advise was recevied from Kevin Rogers and anonymous referees. Editorial assitance was skilfully provided by Marybeth Grimes.

Contact details

Paul W. Grimes
Professor of Economics
College of Business and Industry
Mississippi State University
Mississippi State, MS 39762-9580
USA

Tel: (662) 325 1987
Email: pwg1@ra.msstate.edu

Jon P. Rezek
Assistant Professor of Economics
College of Business and Industry
Mississippi State University
Mississippi State, MS 39762-9580
USA

Tel: (662) 325 1970
Email: jrezek@cobilan.msstate.edu

Top | IREE Home | Economics Network | Share this page