Individual corrupt behavior: An experimental analysis of the influence factors personality and gender

How to cite this paper: Ulrich, P., Behringer, S., Unruh, A., & Frank, V. (2022). Individual corrupt behavior: An experimental analysis of the influence factors personality and gender. Corporate Ownership & Control,


INTRODUCTION
Compliance and non-compliance have played a greater role in theory and practice in the recent past than before (Van Rooij & Sokol, 2021). The topic is therefore interesting for corporate practice since compliance practice to date has focused strongly on the -control‖ mechanism for enforcing rules in organizations. However, the disadvantage of control is that it can be applied easier ex-post than ex-ante. In any case, it does not cover all circumstances. From a theoretical, especially psychological perspective, compliance, therefore, tries to focus more on the mechanism of -justification‖ and cognitive dissonance (Girandola, 1997).
In the area of non-compliance, there is a presumption that, depending on personality and gender, individuals experience moral dilemmas (Sutinen & Kuperan, 1999) and cognitive dissonance in relation to specific situations of violent behavior. In this respect, the goal of compliance management is to record the personality of employees and integrate it in the elaboration of tools in the area of compliance management.
In the theoretical, conceptual, and empirical literature to date, there are few findings on the factors influencing norm deviant behavior in organizations. Furthermore, even independent of the context of compliance, the literature on the correlation of personality and gender is quite unclear. There are already findings that personality traits measured by the Big Five model and corruption frequency correlate positively with each other. In addition, some studies show that women as individuals are less prone to corruption and also tolerate corrupt behavior in others less. If women are represented on a management board, the propensity of the body (e.g., executive or supervisory board) to engage in corruption decreases overall.
The study can benefit both theory and business practice, that is a better understanding of individuals' propensity to corruption is expected. It means that mechanisms of corporate governance and compliance can then be better adapted to the individual circumstances of personality and gender so that compliance success increases.
From the current perspective, however, there are also no specific cultural findings on male and female corruption in Germany, which is the focus of this paper. Thus, our research questions are: RQ1: To what extent do the contextual factors of personality and gender influence the propensity to corrupt and, ultimately, different behavioral manifestations of norm-defying behavior in organizations?
RQ2: How do personality and/or gender affect the corruption propensity and corrupt behavior of individuals? 2 The rest of the paper is organized in the following way. Section 2 reviews the literature to better understand the context of the study. The methodology for the study is presented in Section 3. Section 4 presents the results of the empirical analyses. Based on the discussions of the findings in Section 5, a set of conclusions are drawn and the implications and limitations of the study are also discussed in Section 6.

LITERATURE REVIEW
For a better understanding of the context of this study, a literature review was carried out. The searching process was structured in the following three main areas: personality and gender, gender and corrupt behavior, personality and corrupt behavior. The final sample includes 51 papers, of which 14 papers were assigned to the personality and gender category; 25 to the category gender and corrupt behavior, and the last 12 papers belong to the personality and corrupt behavior category.

Personality and gender
In the past, many researchers have focused on the relationship between gender differences and personality traits, often in an entrepreneurial context. Various methods and scales such as the Big Five personality traits have been used.
The Big Five personality traits are neuroticism, agreeableness, openness, neuroticism, and extraversion. It can be generally stated that women are often more neurotic, agreeable, conscientious, and open than men. This is confirmed by the research of Costa, Terracciano, and McCrae (2001), Feingold (1994), as well as Vianello, Schnabel, Sriram, and Nosek (2013). Costa et al. (2001) investigated in secondary analysis gender differences in personality traits across cultures. The data from the NEO Personality Inventory (NEO-PI-R) from 26 cultures (N = 23,031) indicate that the differences are largely consistent with gender stereotypes. Women indicated that they have a higher degree of neuroticism, agreeableness, warmth, and openness to feelings, while men have a higher degree of assertiveness and openness to ideas. It was found that the extent of gender differences varies between cultures.
Research on gender differences in personality is not a new trend. As early as 1994, Feingold conducted four meta-analyses to investigate gender differences in personality in literature  and in normative data for known personality inventories . It was found that men were more assertive and had slightly higher self-esteem than women. Extraversion, fear, trust and above all tenderness (e.g., care) were higher in women than in men. There were no significant gender differences in terms of social anxiety, impulsiveness, activity, ideas (e.g., thoughtfulness), as well as a place of control and order. Consistent gender differences in personality traits could generally be found across age groups, years of data collection, educational levels, and nations. Vianello et al. (2013) investigated gender differences in implicit and explicit measurements of the Big Five personality traits. The authors replicated previous research showing that women report higher levels of agreeableness, conscientiousness, extraversion, and neuroticism. A slightly higher level of implicit neuroticism and agreeableness was observed in women. Men, however, showed slightly higher levels of implicit extraversion and openness. There was no gender difference in implicit conscientiousness. This can be explained by the fact that explicit self-concepts partly reflect social norms and self-expectations about gender roles. In contrast, implicit self-concepts mainly reflect selfrelated experiences.
Furthermore, there are studies of personality traits in different cultural contexts, as already pointed out by Costa et al. (2001). Gender-specific differences in personality were identified in American and European contexts, but there is a lack of African and especially South African research in this area, which Laher and Croxford (2013) wanted to fulfill. In their study, they investigated whether there were gender differences in personality. In particular, they focused on the relevance for the South African organizational context, where personality assessments are often used for decisionmaking. For example, significant gender differences were found in neuroticism, anxiety, vulnerability, depression, self-confidence, extraversion, etc. From the results, it can be concluded that the differences between men and women are systematic and largely innate. These gender differences need to be known when personality tests are used for decision-making. In addition, constructive use of personality tests can promote team building and diversity.
In addition, some researchers not only differentiated between genders but also categorized subjects into age groups to identify differences in personality traits in particular age categories.
The results of Lehmann, Denissen, Allemand, and Penke (2013) showed both age and gender differences in personality traits. Mean scores of neuroticism and extraversion were negatively associated with age, whereas agreeableness and conscientiousness were positively associated. A curvilinear association is shown by openness to experience with age. Taking into account gender differences, women have higher levels of neuroticism, extraversion, and agreeableness. Men, on the other hand, are more open to experience. In the case of conscientiousness, neither the main effect of gender nor the interactions between age and gender were significant.
Magan, Mehta, Sarvottam, Yadav, and Pandey (2014) found a positive correlation between conscientiousness and age in the entire study population. Significance remained only in males when analyzed by gender subgroups. Neuroticism in women aged 26-35 years is inversely correlated with age in the analysis of age and gender subgroups. A positive correlation with age in men aged 36-45 years was shown with neuroticism and extraversion. Neuroticism decreased in men aged 46-55 years. It is reasonable to assume that personality traits change with age and in a gender-dependent manner.
The aim of Vecchione, Alessandri, Barbaranelli, and Caprara (2012) was to study the gender differences of the Big Five in men and women during a critical period of development (16-20 years). The authors analyzed longitudinal self-report data from 192 men and 211 women using multi-group latency growth modeling. At the age of 16, women scored significantly better in agreeableness, conscientiousness, and openness. In contrast, men performed better than women in terms of emotional stability. From age 16 to 20, conscientiousness and openness increased linearly for both males and females, while energy/extraversion remained stable. Emotional stability increased slightly for men and remained stable for women. The agreeableness showed a quadratic trend in women, first increasing and then decreasing over time. In contrast, this increased linearly for men. Finally, women showed higher interindividual variability than men with regard to the course of conscientiousness and emotional stability.
Furthermore, there are also studies based on a different research method. For example, Drydakis, Sidiropoulou, Patnaik, Selmanovic, and Bozani (2017) utilized a correspondent test (experiment) to capture the way in which firms respond to women who exhibit masculine and feminine personality traits. This field experiment investigated the effects of male and female personality traits on the wage scale for career starters. Since female personality traits are stereotypically attributed to women and since these traits appear to lead to lower salaries, they could provide one of many plausible explanations why women have higher unemployment rates than men and receive lower incomes at the same time.
The study of the correlation between gender and personal traits has a long tradition in psychology. Most authors find a distribution of personal traits close to stereotypes that women have higher levels of agreeableness, while men show more assertiveness and have slightly higher self-esteem. Nevertheless, the results are not unequivocal.

Gender and corrupt behavior
The impact of gender differences and corrupt behavior is the subject of many fields and laboratory studies. Researchers are particularly interested in the strength of the effect of gender on the propensity to corrupt and its implications for business and political practice, as the following studies show.
Alatas, Cameron, Chaudhuri, Erkal, and Gangadharan (2006,2009) investigate gender differences in behavior when confronted with a common bribery problem. They collected experimental data in Australia, India, Indonesia, and Singapore. They show that while women in Australia are less tolerant of corruption than men in Australia, no significant gender differences are seen in India, Indonesia, and Singapore. The findings suggest that the gender differences reported in previous studies may not be as universal as stated and may be more culture-specific.
The paper by Betz, O'Connell, and Shepard (1989) explores possible connections between gender and the willingness to engage in unethical business behavior. The authors used two approaches to gender and ethics, namely: the structural approach and the socialization approach. The results show that men are more than twice as likely as women to take actions that are considered unethical. Fifty percent of men were prepared to buy shares with inside information. Overall, the results support the gender socialization approach, as very few would perform any of the actions considered unethical.
Ameen, Guffey, and McMillan (1996) explore possible connections between gender and the willingness to tolerate unethical academic behavior. In surveys on academic misconduct, women were found to be less tolerant than men. Statistically significant differences were found for 17 of 23 questionable activities. In addition, women are less cynical and less likely to engage in academic dishonesty. Betz et al. (1989) find that the gender socialization approach dominates the structural approach. This fact is supported by these results.
Branisa and Ziegler (2011) use a sample of developing countries and regress corruption on women's representation, democracy, and other control variables. The results show that corruption is higher in countries where social institutions deprive women of their freedom to participate in social life. This is the case even after accounting for democracy and representation of women in political and economic life, as well as other variables. From the results, it can be concluded that in a context where social values disadvantage women, it might not be enough to promote democratic reforms and increase women's participation to reduce corruption.
Based on the results of previous research showing that more women in government are associated with less perceived corruption and that women are individually less likely to condone or express willingness to engage in corruption. Moreover, women seem to be less inclined individually to tolerate corruption or express their willingness to engage in corruption. However, these relationships are context-sensitive in a way that suggests a deeper causal mechanism. Esarey and Schwindt-Bayer (2014) examined 78 democracies between 1990-2010 in terms of accountability, corruption, and women's representation. Their findings suggest that accountability is the key mechanism. If individuals in government are more likely to be held accountable for their corruption, women will disproportionately avoid corrupt activities. In the case when accountability for corruption is weak, men and women will be equally involved in corruption. This evidence is robust to different standards of accountability and corruption.
Capezio and Mavisakalyan (2016) examine the relationship between women's representation on corporate boards and fraud. They show that the increased proportion of women on corporate boards is associated with a lower probability of fraud. They demonstrate the consistency of this result across different robustness checks. Their findings are an argument for policymakers who are interested in improving corporate governance and monitoring to establish regulation for greater gender diversity in boards.
In corruption research, there is limited reliable microdata on corrupt behavior, and the available field data is difficult to interpret. For this reason, Frank, Lambsdorff, and Boehm (2011) have looked at laboratory experiments on corruption that provide information on gender-specific effects. The main preliminary result is that when women are involved in a potentially corrupt transaction, they are more likely to fail. This is because they are too honest and more opportunistic when given the chance to break an implicitly corrupt contract. Moreover, they are less likely to engage in retaliatory non-performance.
Findings in numerous behavioral studies suggest that women are more trustworthy and public-spirited than men. In terms of promoting honest government, according to these findings, women should be particularly effective. The findings of Dollar, Fisman, and Gatti (2001) are consistent with this hypothesis and show that the increased number of women in parliament reduces the extent of corruption. This relationship is found across a large cross-section of countries; the result is robust to a wide range of specifications.
Swamy, Knack, Lee, and Azfar (2001) investigate the relationship between gender and corruption by using several independent data sets. As the microdata show, women are less likely to be involved in bribery and less likely to tolerate the acceptance of bribes. According to cross-country data, in countries where women hold a larger share of parliamentary seats, leadership positions in the government bureaucracy, and make up a larger share of the labor force, corruption is less severe.
In several works with field data, differences in the corrupt activities of men and women were found. This research work has disadvantages that can be overcome in a laboratory experiment. The laboratory provides controlled variation of the environment, and thus of the factors influencing decisions. These laboratory experiments control decision situations in a way that is difficult to duplicate in natural settings. In this type of setting, the experimenter determines and controls, for example, the information provided to subjects; the order in which the various parties can act; and whether the play is repeated or unique. This allows testing of precise predictions derived from game theoretic models (Falk & Heckman, 2009).
Lambsdorff and Frank (2007) played a corruption game with their students, which was embedded into a variant of the ultimatum game. Different roles were allotted to the students. In the role of public servants, the students chose between whistleblowing, opportunism, and reciprocity by delivery (of a contract). The other acted like businesspeople, chose how to frame the game, and whether to blow the whistle. Another much more likely outcome is that businesspeople allocate resources to punishing public officials for non-performance, and show a preference for negative reciprocity. Female public servants tended to behave more opportunistically, while female businesspeople were less interested in negative reciprocity. This confirms the positive role of women in fighting corruption. Businesspeople who strongly favored a corrupt game and received a form with corrupt wording were more willing to punish non-extraditing officials.
Lambsdorff and Frank (2011) asked why women are considered more resistant to corruption. To do so, they replicated the corruption game with students who, in the role of public officials, received a bribe and could choose between reporting (whistleblowing), opportunism, and reciprocity. At the end of the game, those who acted like businesspeople decided whether to publish or not. Male businessmen were more likely to refrain from maximizing bribes and provide resources for punishing opportunistic public servants. Instead of acting opportunistically, some public servants tended to reciprocate or report. The authors find that female public servants are less inclined to retaliate. Their resistance to corruption is less related to their willingness to report.
Overall, it can be concluded that women are less prone to corrupt behavior. The willingness to behave corruptly does not only change due to gender differences but it is also influenced by situational (e.g., board of directors gender diversity, etc.) and culture-specific factors (e.g., women's representation, political systems, etc.).

Personality and corrupt behavior
In the context of research on corrupt behavior, it is important to know whether it is possible to draw conclusions from a person's personality traits to an individual's propensity to corrupt. For this purpose, the researchers investigated the influence of personality traits on their norm-deviant behaviors in different settings.
The study by Stead, Worrell, Spalding, and Stead (1987) deals with situation-related and individual variables that influence behavior and the decision-making process. The authors examined the relationship between unethical decision-making behavior, decision history (a social learning variable), various personality, and demographic variables. The history of decision-making had an influence of unethical decision-making behavior. Situationrelated, social learning variables overshadowed the attempt to explain unethical decision-making behavior through the role of personality and demographic variables.
Limited research addressed psychological antecedents of national corruption-particularly personality, which is measured at an aggregate level. The authors Connelly and Ones (2008) examined the independent, combined, and unique effects of Hofstede's national personality and culture dimensions on perceived national corruption on the basis of 54 countries (Hofstede, 1984). Corruption tended to be lower in countries with a low score on neuroticism and a high score on extraversion. The relationship between conscientiousness and corruption was explained by wealth. From the results, it can be concluded that personality at the national level has a significant impact on the corruption of nations and that more detailed investigation is needed by researchers and policymakers alike.
The study by Kozako, Safin, and Rahim (2013) examined the influence of personality traits of the Big Five on counterproductive work behavior (CWB), focusing specifically on organizational (CWB-O) and individual (CWB-I). At the organizational level, there is a positive relationship between CWB-O and employees with high neuroticism and openness to experience. In contrast, a negative relationship was associated with comfort. At the individual level, a positive relationship was found between counterproductive and employees with neuroticism and openness to experience, while a negative relationship was evident with extraversion and kindness.
Corrupt behavior depends on situational and personal factors, as well as their interaction. Litzcke, Linssen, Schön, and Heber (2014) present the results of corruption studies focused on situational factors (situational risk factors). It has been found that situational risk factors of corruption are less important than is generally postulated. The implementation of predominantly situationaloriented anti-corruption measures, which has been practiced in many companies and organizations so far, produces a dangerous false sense of security.
A vignette-based 2 × 2-scenario experiment by Schön, Litzcke, Linssen, and Schilling (2011) was conducted, in which the situation factors, probability of discovery, advantage value, as well as the person factors organizational cynicism, the five-factor model of personality and intelligence were included. The results show that corruption is seen as widespread behavior, but is considered reprehensible. While the situation factors included had no significant effect on the willingness to engage in corruption, some of the personal factors led to significant results. Conscientiousness is particularly important here as a protective factor against corrupt practices.
The research by Hajhoseiny, Fathi, and Shafiei (2019) investigates the relationship between dark personality traits (i.e., narcissism, Machiavellianism, and psychopathy), the so-called dark triad (Paulhus & Williams, 2002), and corruption intention through the mediation of anxiety. Their hypothesis was that individuals with higher dark personality traits are more likely to be corrupt when they feel anxious. The dark triad was formed by high narcissism, Machiavellianism, and low psychopathy in the formative structural equation modeling (SEM) model. There is a significant indirect positive relationship of the dark triad with corruption intention through anxiety (a full mediator).
The study thus provides evidence that anxiety may contribute to corruption among people with dark triad characteristics.
The study by Zhao, Zhang, and Xu (2016) examines the association between the dark triad of personality (i.e., Machiavellianism, narcissism, and psychopathy) and corruption through a mediatorbelief in good luck. Based on Ajzen's theory of planned behavior (Ajzen, 1991), the authors assumed that individuals with dark triad would be more likely to engage in corruption because of belief in good luck. Taken together, this study provides evidence that belief in good luck may be one of the reasons explaining why people with the dark triad are more likely to engage in corruption regardless of the potential outcomes.
The listed studies show very heterogeneous results regarding the correlation between personality traits and corrupt behavior. There is a tendency for people with high levels of neuroticism and extraversion to be more prone to corrupt behavior than people with lower levels of these personality traits. The results vary due to different operationalizations of the personality traits.

Conceptual framework
The conceptual framework of this study is presented in Figure 1.

Hypotheses
Against the discussed background on personality and gender, the following hypotheses are derived: H1: Women are more likely to refuse to carry out a corrupt instruction and more likely to report such an instruction internally.
H2a: Individuals who exhibit high levels of conscientiousness are less likely to follow instructions to perform corrupt acts and less likely to report such instructions.
H2b: Individuals who exhibit high levels of extraversion are more likely to refuse to carry out instructions to act corruptly and are more likely to report such instructions internally.
H2c: Individuals who exhibit higher levels of agreeableness are more likely to refuse to carry out instructions to act corruptly and are more likely to report such instructions internally.

Experimental design and procedure
A large part of behavioral research consists of functional analysis, for which prediction and control of behavior is the main research subject. In this process, researchers seek to find the causes of behavior within a given environment. To do this, they use experimental behavior analysis, which examines the basic laws and principles that govern behavior (Igaki, Romanowich, & Sakagami, 2019).
In a laboratory experiment, a researcher can completely control the situation. This leaves comparatively little room for other factors to distort the result but allows clear statements to be made about cause-and-effect relationships between the factors under investigation. Thus, a high degree of internal validity can be achieved in a laboratory experiment (Sawada & Aida, 2019).
In the following experimental setup, not all factors can be fully controlled, because the experiment was conducted with groups of students. Therefore, this is a quasi-experiment. The independent variables are completely controlled, but not all relevant confounding variables can be controlled. For example, quasiexperiments often lack randomization of the sample, i.e., subjects cannot be randomly assigned to a condition. As a result, the assignment of experimental conditions at the subject level is not completely determined by the experimenter (Gniewosz, 2011).
Due to the clear control of the experimental conditions and a realistic setting in a quasiexperiment, the effect direction from the independent to the dependent variable should be clearly demonstrated without the effect of third variables. Accordingly, the experiment is divided into two parts. In the first part, general information and characteristics of the test persons are collected (Questionnaires 1-3), while the second part records the behavior of the test persons in a specific, given situation (Questionnaires 4-6). Questionnaire 1 asks for general information about the respondents, such as gender, age, or course of study. Questionnaire 2 is the Big Five Inventory 10 (BFI-10) according to Rammstedt, Kemper, Klein, Beierlein, and Kovaleva (2012), which is used to measure the five personality dimensions. Questionnaire 3 uses the Hannover Corruption Scale (HKS 38) according to Litzcke et al. (2014) to measure attitudes towards corruption.
In the second part, a fictitious company case forms the core of the experiment. In our cases, the test persons are asked to put themselves in the role of a purchasing employee, who was offered a prompt promotion to purchasing manager because of his excellent work in recent years. This purchasing employee is asked either by the family patriarch or the chief executive officer to personally deliver a suitcase with a large amount of money to the main Polish supplier on his behalf. The fictitious company cases varied in the -type of company‖ and the -company size‖. The first influencing factor differed in -family company‖ and -non-family company‖. The influencing factor -company size‖ has the characteristics -small‖ (200 employees) and -large‖ (10,000 employees).
In questionnaire 4, the test person must indicate the company case variant assigned to him or her. In questionnaire 5, the respondents are asked to choose one of several possible options for action based on the given enterprise case. The decision had to be made whether to fulfill the task or not. In addition, the participants could decide whether to remain silent about the incident, report it internally, or report it to an external law enforcement agency (Bocchiaro, Zimbardo, & Van Lange, 2012). If a respondent decides to report the case to an external criminal prosecution authority, a fictitious reporting form is used to check whether the intention is actually accompanied by a corresponding action. Questionnaire 6 is used to record the social desirability of the participant's response behavior using the short scale Social Desirability-Gamma (CFE-G) according to Kemper, Beierlein, Bensch, Kovaleva, and .
The exact procedure of the experiment is illustrated in the following Figure 2. The experiment was carried out during different university lectures between January and March 2020. The test persons were students from several courses from different universities in Germany. The total number of participants was 134. The students received the instructions for the experiment from their lecturer and were provided with a link to access an online survey created with the Unipark survey tool. To access the online survey, they used their laptops or smartphones. The survey started with the initial instructions for the experiment, a declaration of consent, and the questionnaires. After an explanation of the instructions of the experiment by the researcher, the test persons were given 10-15 minutes to answer the first three questionnaires. After all these questionnaires (1-3) had been completed, they all started processing the company cases at the same time. The company cases A to D were each assigned to the test persons in equal proportions. The persons sitting next to each other did not get the same company case. The respondents were given another 10 minutes to read through the company case and answer the remaining questionnaires four to six. In order not to jeopardize the experimental setup, the researchers ensured that all external influences (e.g., conversations between the test persons) were suppressed during the entire time. Extraversion describes the strength of the tendency of outward attention. Extraverted people, for example, are more communicative, tend to go out of themselves and make social contacts. Compatibility describes the tendency to accommodate others, avoid confrontation, adapt, and strive for conformity. Conscientiousness describes the extent to which someone feels committed to his or her tasks and goals. Neuroticism says something about the strength and frequency of the stimuli needed to be impressed by one's feelings. For example, some people are more emotionally sensitive than others and react more sensitively to stimuli. Openness refers to mental agility, creativity and curiosity (Fehr, 2006).
There are numerous procedures for capturing the Big Five factors, such as the NEO Personality Inventory (NEO-PI-R) by Costa and McCrae (1992), or the NEO Five-Factors Inventory (NEO-FFI) (McCrae & Costa, 1989;Borkenau & Ostendorf, 1993). For an economic assessment of the five dimensions of personality according to the five-factor model, the Big Five Inventory 10 (BFI-10) by Rammstedt and John (2007), which was specially developed for surveys subject to time restrictions, was used within the framework of the company's own experiment.
The BFI-10 scale determines two items, one positive and one negative, per personality dimension and thus comprises a total of ten items. The answers to the corresponding ten questions are determined on the basis of a five-level rating scale from -does not apply at all‖ to -applies completely (Rammstedt et al., 2012).

Hannover Corruption Scale
The Hannover Corruption Scale (HKS 38) was developed within the framework of research cooperation on the topic -Corruption -Risk Factors of the Person and the Situation‖. Despite the negative role of corruption for economic and social development (Kaufmann, 2005), there was no specific measuring instrument for recording attitudes towards corruption prior to the development of HKS 38 (Litzcke et al., 2014).
Other available instruments such as the Inventory of Occupational Attitudes and Self-Descriptions (IBES) (Marcus, 2007), the Personality Inventory for Integrity Assessment (PIA) and the Psychological Integrity Test (PIT) are integrity tests. These tests capture the construct of integrity, which is more comprehensive than corruption, and are used to assess the probability of a person showing counterproductive behavior. Counterproductive behavior is defined as actions that are harmful to organizations or individuals (Marcus, 2007).
The Hannover Corruption Scale (HKS 38) by Litzcke et al. (2014) was used to determine the individual's propensity for corruption and should provide a standardized measurement tool for further research on corruption. The authors developed the HKS 38 as a specific and economically applicable measuring instrument for recording attitudes towards corruption. In comparison to IBES, PIA and PIT, the HKS 38 is thus more specific (corruption instead of integrity) and focuses on attitudes instead of personality dimensions.
The scale comprises a total of 38 items. The item development was based on the general attitude model of Eagly and Chaiken (1993). The questionnaire of the HKS 38 was used without modifications for our experiment. To answer the 38 questions, the test persons were given a five-level rating scale from -I strongly disapprove‖ to -I strongly agree‖ (Litzcke et al., 2014).

Measuring reliability with Cronbach's Alpha
A common method for determining the reliability of scales is the analysis of internal consistency with Cronbach Alpha (Cronbach, 1951). The term internal consistency here refers to the degree to which the items of a scale represent each other and the scale that is often equated with homogeneity in the sense of one-dimensionality. As a rule, scales with a Cronbach Alpha from 0.705 are considered internally consistent or reliable (Cortina, 1993).
The HKS 38 has a very high internal consistency. Cronbach's alpha was 0.928 for this scale and 0.929 for the standardized Cronbach's. The minimal deviation indicates that the variances of the items do not differ significantly from one another. The basis for the development of the BFI-10 was the 44-part BFI by John et al. (1991). The items of the BFI have been incorporated into the German translation (for a detailed description see Danner et al, 1997). Of the 44 items, 10 were selected. The item selection procedure was described by Rammstedt and John (2007) in detail. They selected 2 BFI items for each dimension of the Big Five according to four criteria: 1) both the high and the low pole of each factor was displayed, so that each BFI-10 scale would consist of one correctly and one incorrectly rated item; 2) each scale should cover as wide a range as possible by selecting two items that measure both core aspects of the Big Five dimension, but are not highly redundant in content; 3) identical versions in English and German were constructed, so that the resulting instrument could be used for cross-cultural research and the use of chance was minimized; 4) as far as item selections still had to be made, the items were selected on the basis of two empirical criteria, namely their corrected item total correlations with the full BFI scales (and thus preferred central rather than peripheral item contents), and the simply structured pattern of their loading in the factor analyses of all 44 items (and thus preferred items that clearly related to one factor rather than the other four factors) .
Due to the extremely low number of items per scale, and the intended heterogeneity of the two items of a scale, the internal consistency was not suitable as a good estimator for the reliability of the five scales. Instead, Rammstedt and John (2007) and Rammstedt et al. (2012) determined the retest reliability over an interval of six to eight weeks. In the retest design, the mode of data collection was partly varied from computer assisted personal interview (CAPI) to paper pencil. In different samples, they were able to demonstrate satisfactory reliability and validity values for the BFI scale.

Participants
134 students took part in the experiment. Of these, 44% were male and 56% female. The age range is from 18 to 40 years. The majority of the test person was between 20 and 30 years old (85%). The students participating in the experiment were also asked about their respective courses of study. More than half of the test persons (55%) are studying for a bachelor's degree in International Business Administration. 13% are pursuing a bachelor's degree in Business Information Systems and 13% also indicate the MBA with a focus on Sustainability Management. 7% are doing a master's degree in Business Law and another 6% in Financial Management. The remaining 6% belong to various study programs.
Most of the participants (93%) have practical experience (internship or a vacation job). This leaves only 7% who have not yet gained any experience. Table 2 shows the descriptive statistics concerning gender in the study. The results show taht in general, there were more female participants than mail participants.   Table 3 shows the descriptive statistics of the independent and dependent variables. The participants of the sample have a high degree of conscientiousness, as the mean shows. The test persons have the least neurotic traits on average.

Descriptive statistics
On average, the sample shows a relatively high propensity for corruption.

Correlation analysis
To answer the question of whether the three variables gender, personality, corrupt behavior and propensity for corruption are related to each other, the correlation analysis by Pearson was used.

Gender × Personality (Big Five)
Firstly, we analyzed the correlations between gender and personality. The personality was operationalized by the Big Five, which consists of extraversion, compatibility, conscientiousness, neuroticism, and openness. The results showed that there is a significant correlation between gender and compatibility (r = 0.207, p = 0.016); gender and conscientiousness (r = 0.240, p = 0.005), and gender and neuroticism (r = 0.359, p = 0.000). These correlations can be interpreted as: a) There is a positive relationship between gender and compatibility. As a person is female, compatibility increases. b) There is a positive relationship between gender and conscientiousness. As a person is female, conscientiousness increases. c) There is a positive relationship between gender and neuroticism. As a person is female, neuroticism increases.

Corrupt behavior × Personality (Big Five)
Secondly, the relationship between personality traits and corrupt behavior will be analyzed. Corrupt behavior is operationalized by six different items. The participants could choose between fulfilling the task or not, as well as blowing the whistle externally, internally or not. Four significant correlations were found. The first one is a positive relationship between -doing not the task and report it internally‖ and -compatibility‖ (r = 0.229, p = 0.008). This can be interpreted to mean that a person who is more compatible will more quickly refuse to do the task and report internally. This result confirms the significant negative correlation between -fulfil the task and keep silence‖ and -compatibility‖ (r = 0.199, p = 0.021) because persons who are less compatible are more likely to complete the task and remain silent.
The third relationship could be found between -doing the task but reported externally‖ and -conscientiousness‖. This is a negative correlation, which means that a person who is more conscientious, would less perform the task and report externally.
The last correlation exists between -doing not the task and report it internally‖ and -extraversion‖. The person who is more extroverted refuses more often the task and reported internally. I would carry out the task assigned to me, but report the incident to an internal office. 0.131 0.021 -0.066 -0.015 0.105 I would carry out the task assigned to me, but report the incident to an external law enforcement agency.

Gender × Corrupt behavior
The analysis of the relationship between gender and corrupt behavior showed a positive significant correlation between -doing the task and report internally‖ and gender (r = 0.242, p = 0.005). If someone is a woman then the task will be fulfiled and reported internally more often.

Table 6. Pearson correlations Gender × Corrupt behavior
Pearson correlations Gender I would fulfil the task assigned to me and keep silence about the incident.
-0.038 I would carry out the task assigned to me, but report the incident to an internal office.
-0.030 I would carry out the task assigned to me, but report the incident to an external law enforcement agency.
-0.062 I would refuse to carry out the task, but keep silence about the incident.
-0.006 I would refuse to do the job and report the incident to an internal office. 0.242** I would refuse to perform the task and report the incident to an external law enforcement agency. 0.118 Notes: N = 134. **The correlation is significant at the 0.01 level (two-sided).

Gender × Propensity for corruption
As the variable gender is an ordinal scaled variable, the suspected correlation was checked with the Spearman-Rho correlation. The results show a negative significant correlation between gender and the propensity for corruption. This means that women show a lower propensity for corruption than men.

Propensity for corruption × Personality (Big Five)
The results of the Spearman-Rho correlation tests show a significant negative relationship between personality traits conscientiousness and the propensity for corruption. Persons who show a high degree of propensity for corruption are less conscientious.

Test of hypotheses
After certain correlations could be quantified with the help of correlation analysis, the cause-and-effect relationship is now also to be checked by means of regression analyses. Binary logistic regressions are used to test the derived hypotheses because some dependent variables are dichotomous scaled. This means that the variables can only take two forms (binary), which are coded with 0 and 1. In this experiment, gender was asked about three expressions, male, female, and diverse. Then none of the participants checked diverse, the gender is coded as a binary, first dependent variable to the expression's male (0) and female (1). The second dependent variable is corrupt behavior, which was operationalized with 6 items. These items could be answered with -No, I would not do that‖ (0) and -Yes, I would do that‖ (1).
For the logistic regression, the analysis of some independent variables is needed. Personality was measured with the BFI-10 scale, propensity for corruption with the HKS 38 scale. These both independent variables are coded metrically. Corrupt behavior also acts as an independent variable when checking some cause-and-effect relationships. In a logistic regression model, the independent variables should be coded metrically or in the case of categorical variables, as dummy variables. The latter concerns the variable corrupt behavior. Different logistic regression models were used to test the hypotheses.

Hypothesis 1
H1 is aimed to show a cause-and-effect relationship between gender and different reporting behavior of corrupt behavior. For the overall model, the chi-square of the model's fit is significant (7.959, p = 0.005 < 0.05). The quality of the regression models (model fit) was determined using the three measures Maximum Likelihood (-2LL), Cox and Snell R² and Nagelkerkes R². The Nagelkerkes R² of 7.7% is acceptable, and the model performs well in distinguishing between -gender‖ and -fulfil task and report it internally‖. The classification performance shows that 61.9% of the participants could be correctly classified by the model with regard to the dependent variables. However, the predictive power of the overall model taking into account the independent variables has improved only slightly (by 5.9%) compared to the baseline model. A positive and significant regression coefficient (B = 0.999, p = 0.006) confirms H1. The logistic regression analysis of H1 shows that both the model as a whole (chi-square = 7.959, p = 0.005, n = 134) and the individual coefficients of the variables are significant. If a participant decided to refuse the task and report it internally, this independent variable increases by one unit each, the relative probability of being female increases by 171.4%. In other words, women are 2.7 times more likely to refuse the task and report this internally than men. The R-square that is output in regression analyses can be converted to an effect size according to Cohen (1992). For H1, the following effect size of 0.28 results (see Table 14), which is a medium effect according to Cohen (1992). This verified the hypothesis that women show less propensity of corruption and are less tolerant of corrupt behavior. Women report such norm deviant behavior more often than men.

Hypothesis 2a
H2a shows a cause-and-effect relationship between -conscientiousness‖ and -fulfil the task, but report it externally‖. For the overall model, the chi-square of the model's fit is significant (5.207, p = 0.022 < 0.05). The quality of the regression models (model fit) was also determined using the three measures Maximum Likelihood (-2LL), Cox and Snell R², and Nagelkerkes R². The Nagelkerkes R² of 11.3% is good, and the model performs well in distinguishing between -conscientiousness‖ and -fulfil task, but report it externally‖. 94.8% of the participants could be correctly classified by this model. Negative and significant regression coefficient (B = -1.042, p = 0.027) confirms H2a. The personality also has an influence on person's behavior. Three sub-hypotheses were examined with the aim of verifying a cause-andeffect relationship between personality traits and various manifestations of corrupt behavior. The first logistic regression model confirms a relationship between -conscientiousness‖ and -fulfil the task, but report it externally‖. This relation is negatively correlated, which means that a participant decided to fulfil the task, but report it externally, the relative probability of being conscientious decreases by 64.7%. The effect size of 0.35 can be interpreted as a medium effect (see Table 14).

Hypothesis 2b
H2 shows a cause-and-effect relationship between -extraversion‖ and -refuse the task and report it internally‖. For the overall model, the chi-square of the model's fit is significant (4.881, p = 0.027 < 0.05). The Nagelkerkes R² of 4.8 % is acceptable, and the model performs okay in distinguishing between -extraversion‖ and -refuse the task and report it internally‖. Only 56.7% of the participants could be correctly classified by this model. But the classification performance increased about 6.7% compared to the baseline model. The positive and significant regression coefficient (B = 0.473, p = 0.030) confirms H2b.
The second logistic regression model was positively correlated and showed that participants who refused the task and report it internally are more extraverted with a relative probability of 60.5%. The effect size of 0.22 corresponds to a small effect according to Cohen (1992).

Hypothesis 2c
H2c shows a cause-and-effect relationship between -compatibility‖ and -refuse the task and report it internally‖. For the overall model, the chi-square of the model's fit is significant (7.207, p = 0.007 < 0.05).
The Nagelkerkes R² of 7.0 % is acceptable, and the model performs well in distinguishing between -compatibility‖ and -refuse the task and report it internally‖. The classification performance of 62.7% is 12% higher compared to the baseline model. The positive and significant regression coefficient (B = 0.633, p = 0.009) confirms H2c-1. H2c-2 is also confirmed by the cause-and-effect relationships between -compatibility‖ and -fulfil the task and keep silence‖. For this overall model, the chi-square of the model's fit is significant (5.164, p = 0.023 < 0.05). The Nagelkerkes R² of 8.3 % is acceptable, and the model performs well in distinguishing between -compatibility‖ and -fulfil the task and keep silence‖. 91.0% of the participants could be correctly classified by this model. The negative and significant regression coefficient (B = -0.938, p = 0.026) confirms H2c-2. The last logistic regression model reviewed H2c that examined the relationship between compatibility and two types of corrupt behavior. Participants who decided to refuse the task and report it internally are more compatible with a relative probability of 30.2%. This corresponds to a medium effect size (0.27). The result is confirmed by a negative significant cause-and-effect relationship between -compatibility‖ and -fulfil the task and keep silence‖. If a participant decided to fulfil the task and keep silence about it, this independent variable increases by one unit each, the relative probability of being compatible decreases by 39.1%. The relationship has an effect size of 0.30, which can be interpreted as a medium effect.

DISCUSSION OF THE RESULTS
The experimental analysis shows that there are some cause-and-effect relationships between gender, corrupt behavior, and personality traits. Lots of various studies with different samples and operationalized variables showed that women are less corrupt, more risk-averse, more sensitive to social cues, less competitive, more altruistic, and more inclined to cooperate than men (Croson & Gneezy, 2009). On the other hand, they also find that women are generally less trusting and show more flexible ethical nurture. The results were confirmed by the laboratory experiment of Rivas (2013), in which she examined whether women and men who find themselves in the same situation behave differently or whether, on the contrary, they behave in the same way. The findings showed that women are actually less corrupt than men. However, in the research of Alatas et al. (2009), the relationship between gender and tolerance of corruption was not so clear. In India, Indonesia and Singapore were no significant gender differences to find. They conclude that the gender differences may not be as universal as stated and may be more culture-specific. This is in line with the findings of Esarey and Chirillo (2013). They found out that women disapprove of corruption more than men and are less likely to engage in corrupt practices, when they live in countries with democratic institutions. Nevertheless, the number of studies predominates, which confirmed that women are less corrupt than men. Also, our results support the view that women are less tolerant and reported such norm deviant behavior more often than men.
Gender differences can be caused by biological and social differences, such as differences in the social roles of men and women. A person's social role and presence in public can play an important role in the person's exposure to corruption. Thus, if women and men differ in their social roles, they can also be expected to differ in their attitudes to corruption. A higher level of susceptibility to corruption in daily life can promote tolerance and acceptance of corruption, which is reflected in norms of behavior. In addition, women in countries where they are less present in public life may be more vulnerable to corruption (and therefore less tolerant of corruption) (Alatas et al., 2006).
To develop a better understanding of why people act corruptly in companies, it is important to discuss the influence of personality traits on norm deviant behavior. The results of the logistic regression show significant relationships between conscientiousness, extraversion, compatibility, and different forms of corrupt behavior. Conscientiousness is a personality trait, which is defined as a -tendency to respond in certain ways under certain circumstances‖ (Roberts, Jackson, Fayard, Edmonds, & Meints, 2009), or, more generally speaking, the tendency to think, feel, and behave in a relatively enduring and consistent fashion across time in trait-affording situations (Roberts et al., 2009). In the concept of conscientiousness, there are several common variants that can be labeled as -responsibility‖ (moral or social), as some authors (Becker, 1998 People with a high level of extraversion enjoy being in social situations. They are characterized by warmth, positive effects, high energy, assertiveness, and open-minded nature. Some conceptualizations of extraversion (Costa & McCrae, 2008) also contain facets related to the pursuit of excitement, while others facets related to ambition (Watson & Clark, 1997). The excitement-seeking facet of extraversion provides the strongest justification for linking this characteristic to fraud. People with a high level of excitement are risk-takers who seek thrills and stimulating environments (De Bruin & Rudnick, 2007). They found a positive relationship between excitement-seeking (extraversion) and corruption propensity. As for extraversion, Eysenck (1997) was of the opinion that extraversion potentiates criminal behavior, which has been confirmed by several studies. In summary, the various studies produced mixed results on the relationship between extraversion and corrupt behavior.
The results of our study confirmed the negative related relationship between these two variables. People with a high degree of extraversion are such self-confident that they refused to do a task, which could be interpreted as a corrupt act and reported internally. Persons with a minor extraversion are less able to resist the instructions of the individual boss and would not report such corrupt tasks. These people would probably obey the instructions and keep quiet about it. A high degree of extraversion can be assessed positively and stands for higher resistance against corrupt behavior, as the results of the experiment confirmed.
Other factors that have been analyzed so far in the context of corruption are ownership structure, dividend policy and performance. For example, D'Amore, Lepore, Landriani, Paolone, and Pozzoli, (2019) show that the level of corruption perceived by investors is relevant to the relationship between ownership concentration and firm performance. These results suggest that institutional context variables should be taken into consideration by corporate governance scholars in their empirical investigations. There is also evidence that dividend policy is affected by the level of corruption, as more corruption requires more money for -unofficial payment‖ at the expense of official payments (Tran, 2020).

CONCLUSION
The discussed experiment was designed to see how people react in a situation where their boss requests to behave corruptly or unethically in the interest of the company. There are some findings and several concluding remarks to this experimental research. Firstly, women and men behave differently when they had to fulfill a corrupt task. Women refused to do such tasks and reported internally more often than men. Secondly, personality traits play a role in corrupt behavior. Are people more extraverted and compatible, than there is a higher probability that they behave less corruptly. In addition, they reported such misconduct more often to an internal unit. On the other hand, are people more conscientious and compatible, they would fulfil a corrupt task with which they were assigned by their boss. But more conscientious people would also report such misconduct to an external law enforcement agency in the opposite direction to compatible people who would keep silence. Thirdly, the first influencing factor, which differed in -family company‖ and -non-family company‖, had no significant impact on corrupt behavior or the propensity of corruption of the participants. Fourthly, the second influencing factor -company size‖ had also no impact on the unethical work behavior of the test persons. Fifthly, the test persons could not put themselves well into the case and had less imagination of how they would react in this case. Probably, the case or the context was too abstract for them, which can be a reason why these both influencing factors had no impact in this sample.
The experiment was done only with a German sample. This can have a limiting effect because the sample is not representative (results do not reflect the worldwide population). The students were socialized in the same culture. An individual's cultural participation influences how they behave toward others in the society, and their cultural participation influences how they treat them. Culture permeates social, economic, and political action (Stanley, 2006). Accordingly, an extension to other cultures would be necessary to achieve a better significance in the results of the experiment and exclude cultural effects.
It is often discussed whether students are the right choice as subjects for experimental studies. Are students not too -special‖ as a sample, i.e., not representative enough? Students as a subject for an experiment are a convenience sample for academics since they are easily accessible as they are represented in a large number of universities. Also, they are very time-flexible and can do an experiment at any time. In addition, students understand the rules and background of an experimental study very fast, because they have a deeper insight into this research approach (Weimann & Brosig-Koch, 2019). Friedman and Sunder (1994) summarized the advantages of undergraduates as subjects as follows: 1) they have plenty of free time, so are often available to participate; 2) they learn the rules for a new situation relatively quickly, and 3) they are cash poor, so incentives of a given size will seem relatively large to them. Often researchers pay a small cash amount for participation in the experiment. This is a welcome incentive for the students to be part of such studies.