Cheating, Trust and Social Norms: Data from Germany, Vietnam, China, Taiwan, and Japan

Abstract

The data presented here contain information on cheating behavior from experiments and general self-reported attitudes related to honesty-related social norms and trust, together with individual-level demographic variables. Our sample included 493 university students in five countries, namely, Germany, Vietnam, Taiwan, China, and Japan. The experiment was monetarily incentivized based on the performance on a matrix task. The participants also answered a survey questionnaire. The dataset is valuable for academic researchers in sociology, psychology, and economics who are interested in honesty, norms, and cultural differences.

Dataset:https://data.mendeley.com/datasets/fpp33zgds7/1.

Dataset License: CC BY 4.0.

1. Summary

In daily life, it is not uncommon to encounter dishonesty and cheating: individuals and companies attempt to evade taxes, drivers fail to report the accidental damage they have caused to another parked vehicle, children lie to their parents, and parents lie to their children [1]. The current literature on economics pays close attention to the study of honesty, because honest behavior plays a crucial role in many economic transactions. A meta-analysis of dishonest behavior [2], which included 130 studies and nearly 565 experimental trials, also suggested that it can be promising to study real-world circumstances that give people an opportunity to behave dishonestly. In the same vein, the combination of 90 experimental studies in the work of [3] confirmed that people, in fact, lie very rarely, because they are driven by a preference for being honest. Despite the fact that there are numerous studies investigating the determinants of lying, which may include the economic background, social preferences, cognitive skills, and the beliefs of the individual, the influence of trust and social norms on lying has still not been properly studied.

The main reason that these five countries were chosen for this study is their differences in terms of political systems, cultural roots, and economic development. For example, German people are more likely to be honest in the democratic system [4]. In the same vein, different worldviews could have different associations with cheating behaviors [5]. Moreover, while the linkage between trust and honesty has been studied [6], the previous studies only carried out luck-based experiments (for example, dice-rolling and coin-flip paradigms). In contrast, this paper provides the dataset from the real-effort experiment of a matrix task game for participants. Apparently, the Confucian culture encourages people to work hard and diligently to obtain a better social status. Therefore, this study decided to choose different dimensions, for example, political systems (democratic countries—Germany, Japan, and Taiwan and socialistic countries—China and Vietnam), economic conditions (developed economies—Japan, Taiwan, and Germany and developing economies—Vietnam and China), and cultural roots (Confucianism—Vietnam, Taiwan, Japan, and China and Christianity—Germany [7]). These differences make our data unique, contributing to the expansion by replication. We hypothesized that those who were exposed to the Confucian tradition would cheat less than the German sample, because our experiment was mainly based on the real-effort task [7]. Therefore, our research question was: “What differences exist between the dishonest behaviors, social norms, and trust in Western and East Asian samples?”.

The preliminary findings from part of these data in [7] examined cheating behaviors. Accordingly, the study emphasized the differences in cheating behaviors across these countries, explained by the cultural differences. The paper analyzed mainly descriptive statistics and visualization, while there was no predictive power by any regression. In addition, there is potential to examine the data more deeply by exploring the differences in social norms and trust at the individual level. Therefore, this data paper aims to increase the understanding of the data from different perspectives. In particular, the experimental and questionnaire dataset can be used to understand the association between the cheating behavior in the experiment and the stated attitudes towards honesty, cultural dimensions, social norms, and trust amongst the university students. The data also offer insights about differences between countries in cheating behavior.

2. Data Description

The attached dataset (https://data.mendeley.com/datasets/fpp33zgds7, accessed on 15 August 2022) contains two parts, the experimental data and survey records from 493 participants concerning cheating behaviors and social attitudes, consisting of social norms and trust.

We relied on the matrix task, devised by [1]. Participants were asked to find two numbers in each matrix of twelve numbers, which added up to exactly 10. Each participant could solve up to 20 matrices within a fixed time. The participant then received an answer sheet to conduct a self-evaluation and reported the number of solved matrices; the participant obtained a fixed payment for each matrix. However, one could cheat while reporting the number and, in the extreme case, report 20 correct matrices. Our design followed [4,5,6], which allowed cheating behavior to be retrieved after the experiment at an individual level. Therefore, our experimental data contain a binary variable (whether the participants cheated or not) and a continuous variable (the number of over-reported correct answers, which captured the magnitude of the cheating behavior).

In the questionnaire, we elicited four main groups of variables: the Cognitive Reflection Test (CRT), attitudes towards cheating in exams as well as everyday life, in-group versus out-group trust, and socio-demographics. These data related to the questionnaire were not published in the previous study [7]. Accordingly, the data allow the readers to match the dishonest behaviors and other data dimensions (such as demographic factors, social norms, and in-group vs. out-group trust).

Therefore, our data consist of two sets of data: (i) experimental data and (ii) survey data. In the following subsection, we summarize the data acquisition and describe the data characteristics. We also define the coding and type of every single variable in the dataset.

2.1. Experimental Data

Table 1. Summary of the descriptive statistics of the experimental sessions.

Sessions	Language Code	Total	Cheaters	Average Over-Report	Average of Correct Answers
1. Germany session 1 (Hochschule Trier)	DE	35	15	10.93	6.13
2. Germany session 2 (University of Trier)	DE	10	1	3.00	5.30
3. Germany session 3 (University of Trier)	DE	18	0	0.00	7.44
4. Germany session 4 (University of Magdeburg)	DE	40	8	6.00	6.80
5. Vietnam session 1 (Banking University HCMC)	VN	20	4	1.50	5.65
6. Vietnam session 2 (Banking University HCMC)	VN	35	5	2.20	6.23
7. Vietnam session 3 (University of Economics HCMC)	VN	10	1	1.00	7.30
8. Vietnam session 4 (University of Economics HCMC)	VN	35	6	6.50	5.80
9. Vietnam session 5 (Other universities)	VN	14	1	1.00	5.93
10. Vietnam session 6 (University of Commerce)	VN	22	1	1.00	6.91
11. Vietnam session 7 (Foreign Trade University)	VN	33	3	1.33	6.03
12. China session 1 (University of Xiamen)	CN	10	0	0.00	8.80
13. China session 2 (University of Xiamen)	CN	6	1	2.00	5.83
14. China session 3 (University of Xiamen)	CN	20	0	0.00	8.60
15. China session 4 (Zhongnan Business School)	CN	24	1	1.00	6.92
16. China session 5 (Zhongnan Business School)	CN	25	4	1.25	6.68
17. Taiwan session 1 (Shih Chien University)	TW	20	0	0.00	5.75
18. Taiwan session 2 (Shih Chien University)	TW	31	1	1.00	6.06
19. Taiwan session 3 (National Chengchi University)	TW	15	1	3.00	6.40
20. Japan session 1 (Hiroshima City University)	JP	35	3	2.00	9.83
21. Japan session 2 (Hiroshima City University)	JP	35	2	3.00	8.37
Total/Average		493	58	5.21	6.86

summarizes the language code, the total participants in each session, the number of cheaters, and the mean of over-reported matrices across countries. In our dataset, we also generated the natural logarithm of the over-reported number of matrices (Magnor: magnitude normalized), while the binary variable (Y) represents cheating behavior (1) and non-cheating behavior (0). In addition, in our data file, the ‘sessionsize’ presents the number of participants in each session, and ‘CA’ means the number of correct answers. Finally, the variable ‘DIV’ denotes the name of each session.

As a robustness check, we also treated those who over-reported one correct answer as non-cheaters [2], as they could have simply miscounted their scores. Thus, in our data, we also have two new variables, generated for robustness check, which are entitled Yrobust (for binary values) and Magrob (for continuous values). These variables differ from the original dataset with the cheating behavior, including those who claimed their actual performance to be more than one correct answer higher.

Table 2. The explanation of the main variables.

Variables	Explanation	Value Labels
Y	Defines the participant as a cheater or a non-cheater	1—Cheater; 0—Non-cheater
Yrob	Defines the participant as a cheater or a non-cheater after excluding one over-reported answer	1—Cheater; 0—Non-cheater
Underreport	Defines the participants who underreported the correct answers	0—Correctly reported the number of matrices; 1—Underreported the correct answer
Magnor	The natural logarithm of the number of over-reported answers	Continuous values
Magrob	The natural logarithm of the number of over-reported answers after excluding one over-reported answer	Continuous values
sessionsize	The number of participants in each session	Continuous values
CA	The number of correct answers after excluding the cheating answers	Continuous values
DIV	The number representing the session	Ranging from 1 to 21 sessions

introduces how our main variables were defined in the dataset codebook.

We elicited the religions and worldviews of the participants and defined the following variables.

Table 3. The explanation of the dummy variables.

Variables	Explanation	Value Labels
Germany	Participant in Germany	1—True and 0—Otherwise
Vietnam	Participant in Vietnam	1—True and 0—Otherwise
China	Participant in China (P.R.)	1—True and 0—Otherwise
Taiwan	Participant in Taiwan	1—True and 0—Otherwise
Japan	Participant in Japan	1—True and 0—Otherwise
Confucian	Associate themselves with Confucianism	1—True and 0—Otherwise
Buddhist	Associate themselves with Buddhism	1—True and 0—Otherwise
Taoist	Associate themselves with Taoism	1—True and 0—Otherwise
Christian	Associate themselves with Christianity	1—True and 0—Otherwise
Muslim	Associate themselves with Islam	1—True and 0—Otherwise
Marxist	Associate themselves with Marxism	1—True and 0—Otherwise

summarizes the variables’ names as well as the value labels. The religion and worldviews were associated with the socioeconomic factors and cheating behaviors (age, sex, religion, household information, etc.) [8]. Therefore, we included these determinants to capture the individual characteristics. Recently, there has been evidence that cultural worldviews [5] and political systems [4] could shape the tendency toward honesty. Although there has been no direct study on Marxism and cheating behavior, those who experienced socialist East Germany were more likely to cheat [5].

To provide the basic information regarding our participants,

Table 4. Summary of other socioeconomic factors.

Variables	Mean	Std. Dev.	Min	Max	Skewness	Kurtosis
Age	20.87	2.64	17	34	1.91	7.48
Household members	3.66	1.71	0	11	0.11	3.85
Household income (€)	3974.24	6180.19	0	60,667	5.42	41.62
Major (1—Economics)	0.72	0.45	0	1	−0.96	1.93
Sex (1—Male)	0.44	0.45	0	1	0.53	1.93
Germany	0.21	0.41	0	1	1.43	3.05
Vietnam	0.34	0.48	0	1	0.66	1.43
China	0.17	0.38	0	1	1.73	4.00
Taiwan	0.13	0.34	0	1	2.15	5.62
Japan	0.14	0.35	0	1	2.05	5.20
Confucian	0.15	0.36	0	1	1.93	4.73
Buddhist	0.45	0.50	0	1	0.21	1.04
Taoist	0.06	0.24	0	1	3.66	14.46
Christian	0.15	0.35	0	1	1.99	4.99
Muslim	0.01	0.11	0	1	8.88	80.01
Marxist	0.09	0.28	0	1	2.92	9.53

Notes: ‘How old are you?’, ‘How many people are in your household?’, ‘How much money does your household have on average as total income (i.e., salary from jobs, income from a business, support from relatives or the state, etc.)?’, and ‘In what major did you graduate?’ were the respective five questions for the following variables. ‘Germany’, ‘Vietnam’, ‘China’, ‘Taiwan’, ‘Japan’, ‘Confucian’, ‘Buddhist’, ‘Taoist’, ‘Christian’, ‘Muslim’, and ‘Marxist’ are defined in Table 3. In terms of questions about income and monetary variables, we used purchasing power parity to convert the local currencies to the EUR to ensure consistency.

summarizes the descriptive statistics of the socioeconomic factors in terms of the participants.

2.2. Survey Data

These data were not mentioned and analyzed in the previously published work [7]. The questionnaire that participants answered consisted of the following parts:

• Attitudes towards cheating behaviors in daily life and examinations

We asked the respondents several questions regarding their attitudes towards cheating behavior. They first had to rank the following items in order of importance from very important to least important: fairness, politeness, honesty, courage, and responsibility, following the previous studies of [7,8,9,10], shown in Panel A in

Table 5. The explanation for variables regarding attitudes towards cheating behaviors.

Variables	Explanation	Value Labels
Panel A:
Fairness	Rank in order of importance from very important to least important.	1—The least important; 5—The most important
Politeness	Rank in order of importance from very important to least important.	1—The least important; 5—The most important
Honesty	Rank in order of importance from very important to least important.	1—The least important; 5—The most important
Courage	Rank in order of importance from very important to least important.	1—The least important; 5—The most important
Responsibility	Rank in order of importance from very important to least important.	1—The least important; 5—The most important
Panel B:
E51	Buying something that you know is stolen.	1—Never justified, 2—Rarely justified, 3—Sometimes justified, 4—Always justified, 5—I don’t know.
E52	Keeping money that you found in the street.	1—Never justified, 2—Rarely justified, 3—Sometimes justified, 4—Always justified, 5—I don’t know.
E53	Having an affair when you are married.	1—Never justified, 2—Rarely justified, 3—Sometimes justified, 4—Always justified, 5—I don’t know.
E54	Failing to report accidental damage you have caused to a parked vehicle.	1—Never justified, 2—Rarely justified, 3—Sometimes justified, 4—Always justified, 5—I don’t know.
E55	Avoiding a fare on public transport.	1—Never justified, 2—Rarely justified, 3—Sometimes justified, 4—Always justified, 5—I don’t know.
E56	Cheating on taxes.	1—Never justified, 2—Rarely justified, 3—Sometimes justified, 4—Always justified, 5—I don’t know.
E57	Someone accepting a bribe in the course of their duties.	1—Never justified, 2—Rarely justified, 3—Sometimes justified, 4—Always justified, 5—I don’t know.
E58	Making up things on a job application.	1—Never justified, 2—Rarely justified, 3—Sometimes justified, 4—Always justified, 5—I don’t know.
E59	Motivating a child by telling them there will be a reward.	1—Never justified, 2—Rarely justified, 3—Sometimes justified, 4—Always justified, 5—I don’t know.
E510	Lying to a little child that there is a dangerous wolf.	1—Never justified, 2—Rarely justified, 3—Sometimes justified, 4—Always justified, 5—I don’t know.
Panel C:
A	Student A—Student cheating	1—Strongly negative, 2—Negative, 3—Neutral, 4—Positive, 5—Strongly positive.
B	Student B—Student accepting the cheating behavior	1—Strongly negative, 2—Negative, 3—Neutral, 4—Positive, 5—Strongly positive.
C	Student C—Student reporting	1—Strongly negative, 2—Negative, 3—Neutral, 4—Positive, 5—Strongly positive.
Worry	Caught cheating during an examination in front of the whole class	1—Not at all and 5—Extremely.

.

Furthermore, we also asked the participants ten questions about their attitudes towards cheating behavior in daily life adapted from [9], shown in Panel B in Table 5. In the same vein, we also surveyed the participants’ attitudes towards cheating behavior on examinations with the following question forms, which are summarized in Panel C in Table 5:

1. ‘Student A and Student B take an exam. Student A copies from Student B’s answers. Student B allows that. Student C reports that to the departmental office. Please characterize your attitude toward each of A, B, and C on a 5-point scale from strongly negative (1) to strongly positive (5).

2. Imagine getting caught cheating during an examination in front of the whole class. On a 5-point scale, how concerned would you be about your reputation (1—not at all and 5—extremely)?

b.

Trust questions

Regarding how to construct the trust indices consisting of the in-group and out-group, we discussed in the study [11] non-reversed and reversed questions, representing in

Table 6. The in-group and out-group trust statements in the questionnaire.

Question	Statements	Non-Reversed or Reversed
Panel A: In-group trust
1	I trust them to keep their promises.	Non-reversed value
2	They will gossip about me when I am absent.	Reversed value
3	I am very happy to lend personal belongings (e.g., books, CDs) or money to them, when they are in need.	Non-reversed value
4	If one of them asks me to lend a larger amount of money and promises to return it as soon as possible, I would hesitate, because he or she might not pay me back.	Reversed value
Panel B: Out-group trust
1	I have faith in their promises	Non-reversed value
2	If they act in a friendly way towards me, they might be unfriendly behind my back.	Reversed value
3	I would like to help them when they are in trouble, such as lending my cellphone to them to make a phone call.	Non-reversed value
4	If I am alone and I see somebody who seems to be injured and in need of help, I would hesitate, because I would be afraid to be tricked.	Reversed value

Notes: The seven-point Likert scale was applied for these trust questions.

. Therefore, we also scaled these indices into a consistent form.

c.

Cognitive Reflection Tests (CRT)

Furthermore, the participants were asked six cognitive reflection questions, following [12,13,14], as shown in

Table 7. The Cognitive Reflection Test questions.

Question	The Detailed Questions
1	A pen and an eraser cost EUR 1.10 in total. The pen costs EUR 1.00 more than the eraser. How much does the eraser cost? _____ cents?
2	If it takes 5 machines 5 min to make 5 widgets, how long would it take 100 machines to make 100 widgets?
3	In a lake, there is a patch of lily pads. Every day, the patch doubles in size. If it takes 48 days for the patch to cover the entire lake, how long would it take for the patch to cover half of the lake?
4	If you’re running a race and you pass the person in second place, what place are you in?
5	A farmer had 15 sheep, and all but 8 died. How many are left?
6	Laura is 10 years old. Her father has three daughters. The first two are named April and May. What is the third daughter’s name?

.

The number of correct answers was denoted as the new variable ‘CRT’.

Table 8. The explanation for the variables regarding the trust and Cognitive Reflection Tests (CRT).

Variables	Explanation	Value Labels
In-group trust	The level of in-group trust from four questions	Continuous values from 1 to 7 (1—lowest trust and 7—highest trust)
Out-group trust	The level of out-group trust from four questions	Continuous values from 1 to 7 (1—lowest trust and 7—highest trust)
CRT	The number of correct answers on the Cognitive Reflection Tests.	Continuous values ranging from 0 to 6.

represents how the trust and CRT variables were defined.

3. Methods

All data were collected in paper-and-pencil form in the laboratory experiment. To investigate the lying behavior to earn a payoff, we replicated the matrix task devised by [1]: students received a sheet of paper with 20 matrices. Each matrix consisted of twelve three-digit numbers (for example, 4.16, see Figure 1 for an example). The students had to find a pair of numbers in each matrix, which summed up exactly to 10. Afterward, the experimenter provided the answer sheet, highlighting the position of two correct numbers in each matrix. Finally, the participants counted by themselves how many correct answers they had. Next, the participants reported the number of the correctly solved matrices on a payment slip to claim their payoffs. Students were requested to leave the experiment room by throwing all documents into the trash bin or leaving them on the table (supposedly, so that others who took the test later could not see them in advance). The amount of money was given to participants outside the room.

Unbeknownst to the participants, we used a hidden identifiable number among the matrices to match their answers in the survey to their actual performance in the experiment. Using this method allowed us to check how many correct answers a participants had and compare it with the number of self-claimed matrices. This defined the magnitude of a participant’s cheating behavior. The participants’ seats were randomized: we distributed the documents, representing where they should seat. By doing this, the experiment maintained the randomization as well as protected the anonymity of the participants, even though we were able to match the cheating behavior with the survey questionnaires.

Our experiments were conducted in five different countries. Hence, the instruction and the experimenters followed a detailed translated script to ensure our data were comparable across tests. Before commencing the experiments in these countries, we also carried out a pre-test session to test how many matrices the average participant could solve in a predefined time. We accordingly adjusted the time limit for the matrix task for different countries based on the average competence for each country: in general, it was four minutes; however, for China, it was only three minutes, because on average the Chinese students performed the task extremely fast.

Before the experiment, the participants were asked to answer the questionnaire with the socioeconomic factors, social norms, trust, and cognitive reflection tests. The experiment and survey data were collected via an anonymous self-administered questionnaire. The full questionnaire and the instructions for the experiment are included in the Appendix.

4. Some Highlights from the Data Analysis

This study employed the matrix-task paradigm to study differences in dishonest behaviors between China, Japan, Germany, Taiwan, and Vietnam. There were two main findings from these unique data based on the short letter [7]. First, participants from Germany cheated more than those who were from East Asia. Second, there was a slight difference in cheating behaviors within Vietnam (South–North) and within Germany (East–West); both countries were split into two different political regimes in the past. This was one of the first research outputs from these novel data [7], and further study could offer different analyses.

To summarize, this study shed new light on differences in dishonest behaviors from the angles of cultural, economic, and institutional determinants from five countries. The data are available for the readers to use, if they would like to analyze them. We emphasized the originality and novelty from our data by providing the comprehensive dataset from the human decision making based on the matrix-task paradigm, social norms, and trust, followed by the survey after the official experiments.