MASARYK U N I V E R S I T Y F A C U L T Y O F E C O N O M I C S A N D A D M I N I S T R A T I O N Externalities and corruption: survey experiment Master's thesis BOLOTOV ARTEM Supervisor: doc. Ing. Rostislav Stanek, Ph.D. Department: Department of Economics Programme Economics Brno 2022 MUNI E C O N Bibliographic record Author: BolotovArtem Faculty of Economics and Administration Masaryk University Department of Economics Title of Thesis: Externalities and corruption: survey experiment Degree Programme: ESF N-EKON Economics, Master's degree programme ESF NEKON01 Economics doc. Ing. Rostislav Stanek, Ph.D. 2022 Field of Study: Supervisor: Year: Number of Pages: 60-8071 Keywords: Perception of corruption, survey experiment, negative externalities, level of tolerance, framing effect, factorial design. Abstract The aim of the submitted thesis "Externalities and corruption: survey experiment" is to estimate the effect of negative externalities on level of tolerance towards corruption acts. The theoretical part is dedicated to the review of corruption issue, methods of studying corruption and literature review on previous corruption experiments. The second part of the thesis elaborates on the survey design and analyses gathered empirical evidence. The factorial design and multiple linear regression model were chosen to analyze the results of the survey. According to the findings, negative externalities have a statistically significant effect on level of tolerance towards corruption acts. Declaration I certify that I have written the Master's Thesis "Externalities and corruption: survey experiment" by myself under the supervision of doc. Ing. Rostislav Stanek, Ph.D. and I have listed all the literature and other sources in accordance with legal regulations, Masaryk University internal regulations, and the internal procedural deeds of Masaryk University and the Faculty of Economics and Administration. Brno, 12t h of May 2022 Bolotov Artem Acknowledgement Here I would like to thank doc. Ing. Rostislav Stanek, Ph.D. for his deep involvement into the experiment process, patience, and valuable advice on data processing. I would also like to thank Mgr. Martin Guzi, Ph.D. for his help with gathering the data required for the study. Table of Content List of Figures 16 List of Tables 14 List of Terms and Abbreviations 15 Introduction 17 1 Corruption and its studying by economists 19 1.1 The problem of corruption in modern world 19 1.2 Experimental approach and its efficiency in corruption analysis 23 1.3 Survey experiment, its strengths, and weaknesses 26 1.4 The perception of a corruption and its determinants 28 2 Survey design 34 3 Data description 40 4 Results 47 4.1 Respondents" reaction 47 4.2 2x2 factorial design 50 4.3 Regression model 53 5 Discussion 57 Conclusion 60 References 63 List of Figures Figure 1: Participants' distribution by sex 42 Figure 2: Participants' distribution by age 42 Figure 3: Participants' distribution by occupation 43 Figure 4: Participants' distribution by monthly income 43 Figure 5: Respondents' reaction to "Bribe" and "Injury" treatments 48 Figure 6: Respondents' reaction to "No bribe" and "No injury" treatments 48 Figure 7: Respondents' reaction to "Bribe" and "No injury" treatments 49 Figure 8: Respondents' reaction to "No bribe" and "Injury" treatments 49 Figure 9: Distribution of independent and dependent variables 50 Figure 10: Means of the dependable variable in different treatments 51 Figure 11: Graphical representation of means of the dependent variable 52 List of Tables Table 1: Distribution of externalities in the cases 35 Table 2: General information about the participants 41 Table 3: p-values for every category 44 Table 4: Outcome of regression model, calculated in R 54 List of Terms and Abbreviations Corruption - Dishonest or fraudulent conduct by those in power, typically involving bribery (Oxford Dictionary) Externality - an outward feature, or all outward features considered together; also, a side effect of some process or activity (Dictionary.com) Framing effect - a cognitive bias where people decide on options based on whether the options are presented with positive or negative connotations (Pious, Scott (1993). The psychology of judgment and decision mak- ing) EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT Introduction Economists and other social scientists have shown a long and ongoing research interest in corruption. The negative effects of corruption are widespread and uncountable. Only few dares to argue that corruption brings inefficiency to the world's economy. It has been among the biggest barriers to the world's economic growth in recent years. The corruption is drawing enormous amount of attention towards itself in media space, nevertheless, the fight against it is still ongoing and economists are not even close to winning this fight. Recently started diversification in methods of studying corruption has shown decent progress in understanding the corruption's nature. Economic experiments succeeded greatly where traditional studies stuck. Experimental approach turned out to be incredibly useful in testing hypotheses and anti-corruption policies and has the potential to contribute fur- ther. The basis for this paper was formed during the survey experiment, that has become key method of collecting the empirical data about different economic issues. The method was quite effective and relatively easy for implementing in current conditions. The research itself is dedicated to externalities that occurs during corruption acts. In the corruption terms, externalities are circumstances that do not influence the corruption act itself but are inseparable part of corruption activity. Gender of participants, tenacity of corruption link and immediate social and/or monetary benefit from the corruption activity are clear examples of such externalities. The main goal of the paper is to analyze the effect of negative externalities comparing to the positive ones on the perception of corruption by the respondents. Framing effect and conditions, that preceded corruption, were chosen as studied externalities. 17 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT The paper is divided into five chapters. The first chapter is dedicated to the literature review of the previous studies and corruption experiments. The second chapter is elaborating on the structure of survey that was used for the experiment. Third chapter includes the description of participants' data and statistical analysis of a representative sample. The results and main findings are presented in fourth chapter, so as econometrical methods that were used in data processing. Last chapter is presenting the discussion and the potential for the further studies. While conducting the study, all the information was taken from the previous economic studies on corruption and already conducted experiments. The experiment holds an explanatory approach, i.e., the study is conducted without predetermined hypothesis, but is aimed at elaborating the results. All the data was proceeded via Google Forms, Microsoft Excel, and R programming language. 18 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT 1 Corruption and its studying by economists 1.1 The problem of corruption in modern world The corruption has been a desired topic for economic studies for decades. Numerous studies have been conducted and plentiful papers have been written over the pastyears. Causes of corruption, corruption's effect on the economies of different states, anti-corruption policies and its implementation and efficiency can be found in the titles of most studies about the disturbing issue of corruption. Despite some studies showing that corruption can serve as so called "lubricant" to improve the efficiency of bureaucratic institutions (Lui, 1985), the harmfulness of corruption is undeniable. It causes income inequality and poverty growth (Gyimah-Brempong 2002, Gupta, Davoodi, Alonso-Terme, 1998), hinders new investments in different industries (Wei, 2000), and entails irrational distribution of government expenses (Mauro, 1997). In some industries and even countries corruption is one of the main reasons for low economic growth (Meon, 2005). John T. Noonan, Jr. in his book "Bribes" (1984) brings evidence on corruption's presence in political and economic societies since the dawn of civilization. However, the degree of attention currently paid to the corruption is unprecedented and nothing short of extraordinary. Vito Tanzi (1998) links the growing attention to the issue of corruption to the spreading of democratic ideologies in XXI century. Active media of democratic countries has created an environment in which discussion of corruption is no longer a taboo. Then, the increasing degree of globalization and a growing role of nongovernmental organizations, such as Transparency International, have greatly contributed to awareness of the problem. Lastly, the pursuit of productive efficiency has acquired greater importance in modern economy and different distortions attributed to corruption attract more attention than ever before. 19 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT The World Bank defines corruption as an abuse of public power for private benefit. The private benefit from engaging in corruption activities can be seen in material or social forms. Corruption is committed by the economic subjects in different forms of violation of existing laws and is present in different forms. Stephen D. Morris (1991) proposes such corruption classification: • "Political corruption" versus "bureaucratic corruption". • "Bribery" versus "extortion". • "Grand corruption" versus "petty corruption". Such classification implies that anti-corruption policies are ought to be versatile to have a chance in the fight versus different forms of corruption. There is a firm conviction, that the corruption needs to be fought. Promise to fight the corruption has become key point in almost every political program among different countries. United Nations (2005), the largest international organization, has dedicated a whole convention to corruption and how to fight it. Low corruption level in country is key demand in Copenhagen criteria, the rules that define whether a country is eligible to join the European Union. Nevertheless, every year we hear about new allegations against high-level officials in different countries and international organizations. Modern anti-corruption policies, developed by economists, are effective up to some point, but there is always a room for improvements. Gary Becker's (1968) "Crime and Punishment: An Economic Approach" is one of the first papers using economics to address the questions of crime and law enforcement. Becker's ideas have been applied to corruption of law enforcers such as the police. One proposed solution is to implement a controlling agent (e.g., an anti-corruption unit) who monitors the law enforcement agency. But if the 20 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT unit will be allowed to collect the fines from corrupt law enforcement agents, it may cause involving the anti-corruption in corruption behavior. Since the start of corruption studies, scientists have dedicated numerous papers aimed at studying the effect of corruption in order to develop more effective anti-corruption policies. Anderson H. James, and Cheryl W. Gray (2006) from the World Bank studied the correlation between an international cooperation and the corruption level in Eastern Europeans countries. Baltaci Mustafa and Serdar Yilmaz (2006) outlined in their study that implementing the anti-corruption policies on the national level is often combined with neglecting of control after these policies on the local governmental level. Klemencic, Goran and Janez Stusek (2008) in their OECD report provided a comparative overview of the anticorruption institutions based on their general model structure: multi-purpose agencies, law enforcement departments and units and policy development institutions. As far as there are enough studies, proving negative effects of corruption, economic and political scientific societies have all the data they need to analyze further influence of corruption. Understanding the nature of corruption may be the next priority in this field. Deep analysis of corruption's nature can shed more light on its causes and structure and help us to understand how people perceive it, why they are being engaged in corruption activities and why other parties tolerate such behavior. Nevertheless, the analysis of corruption is still a quite difficult challenge to economists. The main barrier on the way to corruption analysis is lack of empirical data. It is important, because empirical research allows testing variables that influence corrupt behavior and may thus give advice for political measures aiming to reduce corruption in a society. Since corruption activity is perceived as socially undesirable behavior, immediate participants of corruption activities tend to cover and lie about real situation in this field. Moreover, subjects who are not part of the corruption activity, but are aware of it, often do not reveal 21 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT full information aboutthe corruption process if they became the witnesses of it. The reason of it is that due to criminal nature of corruption it may be dangerous to testify against corruption elements. Nevertheless, households and companies filling up the questionnaires remain the main source of empirical data for conducting such studies. The subject of corruption is versatile and requires multiple different approaches for its understanding. Empirical data for corruption studies can be gathered in traditional or experimental way. There have been different methods of evidence gathering in traditional approaches of corruption analysis. The first and the main one is investigations upon request. Some economic subjects are highly motivated to report the corruption cases if they will benefit from such actions. Companies on competitive market will increase their market share and profits, if their rivals will be caught in some corruption activities, due to reputational and financial losses of corrupted company. Individuals that suffer from inequal income distribution will be more likely to reveal any corruption activities they witness. Such testimonies are still the main sources of numerous investigations that are being conducted by anti-corruption authorities. An alternative source of data for investigations and studies are "case studies". This approach is being used by scientists, when they analyze certain type of corruption in small area. Reinikka and Svensson (2004) captured evidence from a Central Government Transfer Program in Uganda, while analyzing the embezzlement level among the public servants in educational institutions. Olken and Barron (2007) studied the "bribery market" that includes road police and truck drivers in Indonesia. Such detailing helps to gather very accurate empirical data and to develop effective anti-corruption policies in these areas quickly enough. However, it is difficult to implement such results largely in other industries or ar- eas. 22 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT "Indirect signs studies" is a statistical method of anti-corruption investigations. This method implies that corruption may be the reason of different statistical discrepancies in economic data. Fisman and Wei (2004) quantified the effects of tax rates on tax evasion by examining the relationship in China between the tariff schedule and the "evasion gap", which they defined as the difference between Hong Kong's reported exports to China at the product level and China's reported imports from Hong Kong. The results allowed them to claim the presence of corruption on China-Hong Kong custom, and to calculate the estimated percentage of "evasion gap" there. 1.2 Experimental approach and its efficiency in corruption analysis Although traditional approaches were dominating from the first corruption studies till nowadays, experimental approach is becoming more popular these days. The first attempts of conducting an economic experiment belong to the scientists of XIX century, but experimental economics as a part of economics itself has been formed in post-war period, starting with Maurice Allais paradox in 1953 and economic experiments of Vernon Smith in 1962. While initially experimental economics was about finding a solution to casual economic problems like theory of choice under risk influence or convergence to market equilibrium, eventually the problematics of economic experiments shifted towards the studies on the phenomena of social interaction. Axelrod (1984) studied the possibility of cooperative strategies in a competitive environment, repeated games with incomplete information have formed a basis for Aumann's and Maschler's (1996) paper, and interaction in large networks in the context of multilateral conflicts of interest was the problematics of Rappoport (1986) study. The experimental approach has the potential to become the most effective method of corruption studying and anti-corruption policies testing. Results of numerous economic ex- 23 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT periments have already proven an important thesis about the cooperation between economics and sociology: the real-life behavior models from other social sciences can be a key factor to elaborating the behavior of economic subjects, which cannot be explained through basic economic models. These sociological models demonstrate that behavior of economic subjects can be not only ultrasocial, i.e., egoistic and profit-maximizing but also prosocial, which can be interpreted only as irrational by economists. Moreover, the economic experiments have formed a basis for a new branch of traditional economics - a behavioral economics. The popularity of experimental approach can be explained by its efficiency in gathering the vital empirical data for economic studies. First, they involve the creation of a controlled environment. The experiment is conducted in a laboratory, which means that the researcher can control the information that the participants have at the decision-making stage. In turn, this circumstance allows to clearly allocate the factors, the influence of which is studied, and almost unambiguously test the predetermined hypotheses. Besides, such experiments are often connected with a specific economic theory, in accordance with which adequate methods of staging the experiment and processing of the results are cho- sen. Second, unlike most widely used sociological methods, such experiments allow the researcher to create a mechanism that weakens or even removes the effect of social control. Even though the institution created in the laboratory is artificial (specially created by the researcher), it has quite distinct features of reality, being in this sense performative. Third, according to the results of economic experiments, participants earn real money depending on the decisions they make in the experiment, as well as the decisions of the other participants. Each participant receives a different amount of money at the end of the experiment. This is an important feature of the "philosophy" of economic experiments, since 24 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT the payoffs are related to the basic proposition of economic theory that people have preferences and optimize benefits according to given conditions. In total, experimental approach can increase the trustworthiness of gathered data by putting the participants in games that simulate specific conditions needed for studying specific corruption issue. Moreover, these conditions allow scientists to examine narrowed problematics and test predetermined hypotheses. But the main advantage of corruption experiment is the possibility to generate enough empirical data required for corruption studies aimed not only on small market segments, but on general understanding the nature of corruption and developing anti-corruption policies aimed for proper reduction of corruption's presence in the economy. The corruption experiments differ in their structure respectively. In natural experiments scientists do not actually create any economic conditions and just observe how changes in economic situation affect the behavior of economic subjects. Campbell and Ross (1968) in their natural experiment analyzed the changes of drivers' behavior in Connecticut after the change of speed limit in the state. Field experiment is conducted in real life conditions but is controlled by the scientists. Armantier and Boly (2011) conducted a field experiment in Burkina-Faso by bribing public officials during the school elections. In laboratory and laboratory-field experiments subjects are placed in artificial conditions: they are invited into a classroom and asked to play a game under the control of the experimenter and according to the rules set by the experimenter. In order to provide respondents with incentives to make optimal decisions, they are rewarded, as a rule, in proportion to the quality of their decisions. The difference between laboratory and laboratory-field experiments is that in the first ones the general theory is tested, and the researcher does not really care to what extent his proposed game simulates the specific conditions in which 25 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT the interaction corresponding to this theory occurs in real reality. In the meantime, laboratory-field experiments are not conducted in a university classroom, but in real life, that is, in the same environment where subjects make decisions similar to those that are the subject of the experiment. Thus, Habyarimanaetal (2007) conducted the laboratory-field experiment in Uganda's villages, where residents made decisions about local budgets, and their cooperativeness is strongly influenced by the degree of mono-ethnicity of a given community. Clearly, laboratory-field experiments are more difficult to implement than laboratory experiments, but they are useful when the environmental validity of conclusions is on the agenda. 1.3 Survey experiment, its strengths, and weaknesses The survey experiment, which formed the basis for this paper, is a laboratory experiment with its unique features. It is embedded within a survey that can be held online, via paper or phone. In the basis of survey experiment lies the hypothesis that needs an appropriate testing. However, the survey experiment can have an explanatory approach, which means conducting it without a strong hypothesis, but only with idea that will be explained with gathered data. In survey experiments, subjects do not make any real decision, but rather tell what decision they will make, i.e., they report their preferences under described conditions. The disadvantage of this method is the low motivation (lack of incentives) of respondents in telling the truth (which is not fraught with loss of validity only in the situation when they have no incentives to lie). At the same time, such experiments can be conducted in the form of a questionnaire, which allows them to be used on large samples (including representative ones). 26 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT Constructive validity, internal and external validities, and confirmatory testing are the key features of a survey experiment. Constructive validity allows the scientists to create the independent variable, to manipulate it and to create certain dependent measures. Internal validity allows to test the causality in gathered data. External validity allows to create the certain samples of participants. And the confirmatory testing allows you to conduct the survey either with or without the strong predetermined hypothesis. Survey experiments are considered the best in means of gathering the new data. Regardless of research's area, economists do not face the problem of gathering the required amount of data. Representative sampling allows the experimenter to sort out the needed participants or to seek the correlation between behavior of different groups of economic subjects and their defined features. Thus, Daniel W. Gingerich, Virginia Oliveros, Ana Corbacho and Mauricio Ruiz-Vega (2015) conducted a massive survey experiment in Costa Rica, interviewing 4,200 residents and creating large pool of respondents, required for the study. Meanwhile, Aksel Sundstrom (2016) has gathered only 201 respondents' answers for his research about small-scale corruption in government administrations that govern natural resources in South Africa. The idea behind Gingerich's experiment was to test the statement that corruption behaves as a self-fulfilling prophecy or, in the plain words, "the corruption corrupts". To test this hypothesis, the large sample of respondents was needed. Since the study was conducted in Costa Rica, the population of Grande Area Metropolitan, that includes 60% of Costa Ricans in the country, was chosen. The survey sample could be chosen in any other country as well, since the predetermined hypothesis was general and could be linked to a broad sample as well. At the meantime, the representative sample of Sundstrom's survey consisted of local South African fishermen, which was the key requirement of the study, since it was conducted in 27 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT a local area on a specified issue. Sundstrom has used the explanatory approach, i.e., gathering the data without specific hypothesis to test. During the survey experiment, he investigated the effect of corruption, others' compliance behavior, and support for regulations on rule-violating intentions among South African resource users. Nevertheless, one of his conclusions was very similar to one in Gingerich's experiment - resource users are more likely to state rule-violating intentions when corruption among inspectors is widespread. Another example of survey experiment on corruption subject is Christoph Kogler's (2012) paper on determinants of tax compliance in four different European countries. The survey was designed to test the assumption of slippery slope framework, which again implies confirmatory approach. However, the present study tested this main assumption differing in terms of cultural and economic settings by presenting participants with different scenarios of trust and power. Thus, the respondents' pool was divided respectively to test the reaction of participants on different cases. In conclusion, the present results supported the assumptions of the slippery slope framework and confirmed the role of trust and power as important determinants of tax compliance. 1.4 The perception of a corruption and its determinants Serra (2006) conducted an important study on testing the empirical determinants of corruption. In his sensitive analysis he checked on correlation of 28 different economical, sociocultural, and political proxies with the spread of corruption in specific country. He states that different political and sociocultural institutions play a central role in the economic analysis of corrupt behavior. While these institutions differ from country to country, the behavioral norms they provide to people differ respectively. Moreover, corruption itself is an institution that can constitute a behavioral norm for members of a group. If individuals 28 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT implement these norms into their decisions, then the outcome may be an inefficient equilibrium. Therefore, the consideration of corresponding institutions is required to understand the individual behavior with respect to corruption (Bobková, Edberg, 2012). Nevertheless, empirical microeconomic studies also play a key role in analysis of corruption behavior because their results provide a basis for further analysis of mechanisms underlying corruption. Such studies are mostly focused on exploring one or several corruption determinants and slightly expand the field of understanding the corruption's nature. The subsequent chain of similar studies allows precise testing variables that influence corrupt behavior. The important feature of such studies is that they are mostly conducted in from of economic experiment, laboratory, or survey. While laboratory experiments allow to test precise predetermined hypothesis, survey experiments can have explanatory approach, which implies conducting the survey without any strong assumptions and explaining the gained results afterwards. Testing corruption empirically through survey experiments is a rather developing field. Surveys are used to gather the empirical evidence on corruption scale in some regions, like in Sundstróm (2016) or Gingerich (2015). The use of surveys as a method of economic experiment to analyze the behavior of people when stumbling upon corruption is mostly rare. Nevertheless, the researchers resort to using survey experiments in corruption studies more often than ever before. Despite this, some areas of corruption studies still have very few experimental studies conducted in form of surveys. This is quite an omission, considering the potential of survey experiments in these areas. Surveys are the useful in studies where the large amount of empirical evidence is needed. While the laboratory experiments still have some capacity to be reached, survey experiments allow testing corresponding assumptions on much bigger samples. 29 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT The perception of corruption is an area of corruption studies, where the large pool of respondents really matters. Fairly large representative sample plays a key role in understanding how people perceive the corruption and what factors affects their perception. While the laboratory experiments try to catch the exact behavior of participants in specific conditions, studies about corruption perception can rely on participants' reaction gained through the well-designed survey. The perception of corruption depends on vast numbers of its determinants. Some part of the determinants can be merged under the definition of corruption externalities. While the corruption is always perceived by people differently, the perception can differ with respect to set of different externalities. Gender of corruption elements, tenacity of corruption link and specific benefit from the corruption activity are assumed to change the people's perception of corruption. The influence of these externalities is important to be studied for better understanding the corruption's nature. While some of these externalities were studied before in a form of laboratory experiments, this study has been conducted in form of survey experiment. Abbink and Hennig-Schmidt (2006) run a corruption experiment that investigates the effect of in-context framed presentation of the experiment on the level of corruption compared to an abstract neutral terminology. Since the term 'corruption' usually has a negative connotation, Abbink and Hennig-Schmidt investigate whether a neutrally framed corruption experiment with abstract wording can catch the real-life reaction of participants adequately. In contrast to the hypothesis that framed instructions will provoke a negative attitude towards corruption and thus, lead to less corrupt behavior, there is no significant treatment effect through framing. A new strand of experiments has been developed by Barr and Serra (2009) who employ a modified one-shot ultimatum game and analyze the impact of framing and variations of 30 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT external costs. They find that bribe acceptance is comparatively low if negative externalities are comparatively high, hence individuals tend to abstain from corrupt behavior if the external costs are heightened. However, the effect may be due to inequity aversion of players. A different framing does not lead to a significant difference in bribe acceptance in their experiment. Lambsdorff and Frank (2010) test how the wording influences decisions in a corruption game. Proposers in an ultimatum game are in the role of a businessperson who can label her bribe as a 'bribe' or as a 'gift'. Results show that those businesspeople who prefer to call their bribe a 'bribe' are willing to punish opportunistic behavior of public servants harder than those businesspeople who label their bribe a 'gift'. All the studies mentioned above have found no strong empirical evidence on wording or other externalities influence over the behavior of participants. However, the hypothesis that different externalities, e.g., framing effect can affect the way corruption is perceived by the third parties, has been around in economic society for over 15 years. This paper will address the issue of negative externalities in corruption perception. The perception of corruption elaborates, how people see and understand the subject of corruption. Two externalities were used in a survey that formed the basis of empirical evidence. First externality studies how the consequences of actions of corruptive element affect the level of tolerance towards act of corruption. Second externality is already mentioned framing effect, which studies the influence of direct and indirect wording used while describing the corruption act, on level of tolerance towards the corruption. The survey experiment has been used as an appropriate method for gathering the empirical evidence. The structure of a survey is described in next section. The survey's design 31 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT follows a methodological guideline on the measurement of bribery and other forms of corruption through sample surveys, formed by United Nations Office on Drugs and Crime in Vienna (2018). The appropriate general information about participants was gathered according to these guidelines. They survey experiment was conducted in form of explanatory approach, i.e., without predetermined hypothesis about participants' behavior. Gained empirical evidence was econometrically proceeded and logically explained, then presented in this paper. 32 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT 33 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT 2 Survey design The corruption experiment took a form of a survey experiment. Since the study is aimed at determining the influence of negative externalities on corruption perception by economic subjects, two different externalities were chosen. The first one is aggravating circumstance that defines how lethal was the damage caused by the individual engaged in corruption activity. The second one is the choice of words while describing the act of corruption. The second externality also includes shifting of the responsibility as a minor feature. As the previous chapter concludes, the externalities are often expected to have some influence on participants' behavior. The presented survey has two sets of externalities, including positive and negative in each of sets. The assumption states that the negative externalities have similar effect on perception of corruption by respondents, nevertheless, the specific effect of each of them is to be studied. The changes in participants' reaction to corruption activities will be explained only with presence of different set of externalities, which is the only distinction between the survey parts. To analyze the changes of participants' perception of corruption, two surveys and four cases were created. Each survey consists of two cases, endowed with unique set of negative and positive externalities. The distribution of externalities in cases and surveys is presented in table below. 34 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT Table 1: Distribution of externalities in the cases Survey 1 Survey 2 Case 1 Negative; negative Positive; negative Case 2 Positive; positive Negative; positive The main goal of first survey was to check whether negative externalities have any significant impact on corruption perception. It was expected that participants will tolerate corruption act with positive externalities more than one with negative. Respectively the second survey was aimed at determining which one of the negative externalities has more impact on participants' level of tolerance. Also, the cross-check between surveys allows to reaffirm the changes in tolerance level with changes in externalities. To determine the impact of certain externality being negative or positive, the comparison of cases from different surveys was conducted. To reaffirm the validity of representative samples of both surveys' participants, the test on statistical significancy of representative samples was conducted. The data collection was organized through distribution of a survey in MUNI informational system and various author's social circles. The distribution of a survey was done through sharing the link that forward the potential respondent to the survey environment. Since two surveys were created for the study, the link randomly assigned participant to one of the two existing surveys. As sharing the data in research linked to a corruption (read: socially undesired behavior) may be sensitive, survey participants were granted with full anonymity and potential disclosure of personal data is not possible. 35 EXTERNALITIES A N D CORRUPTION: SURVEY EXPERIMENT In the active part of survey participants were presented to the two fictitious corruption cases. In survey number one the first case was described as below: "Jake is a son of a rich businessman who lives in Cleveland, Ohio. On his 16th birthday, he got brand new Ford GTas a presentfrom his parents. To test hisfirst car, he went on a ride with hisfather and exceeded the speed limit causing an accident, in which another driver got injured badly. To soften the punishment for his son, the businessman bribed the police and thejudge, andJake did notface any consequencesfor his actions. Furthermore, the businessman paid a visit to local newspaper and bribed its head director, so the story did notget any resonance in media." The externalities presented in this case are strongly negative. First, the consequences of a situation resulted in an injury of a human being. This should evoke strong negative feeling towards individuals involved in a corruption activity. Then, the description of a corruption act is quite forthright - verb "bribe" is used twice, an emphasis is made on fact, that the culprit was not punished, and the act of corruption is widespread among different public servants. This was also defined as a negative externality, that should lower the level of tolerance towards act of corruption. The second case of a first survey had some differences from the starting one: "Jake is a son of a rich businessman who lives in Cleveland, Ohio. On his 16th birthday, he got brand new Ford GT as a presentfrom his parents. To test hisfirst car, he went on a ride with hisfather and exceeded the speed limit causing an accident, in which another parked car with no one inside got damaged. When the police arrived, Jake'sfather explained to them that he wants to keep his son's record clean and asked to write a fine on his name, promising to make it up for police workers later. The police agreed on Jake'sfather's terms." 36 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT The negative externalities from the initial example were replaced by positive ones. Instead of injury of another driver, this time it was only an auto that was damaged. The word choice is also different: "promise to make it up" and "explained" replaced "bribe", and the corrupted agent has the motivation, that can be seen as a moral one by some participants. In addition, although the culprit was not punished once again, the responsibility for criminal actions was not completely lifted, but shifted to his relative. Such narrative should evoke understanding if not compassion to the individuals engaged in corruption. Cases from the survey number two were designed to combine the positive and negative externalities. Such design helps to determine the weight of certain externality on level of tolerance towards corruption. The initial case in this survey is described below: "Jake is a son of a rich businessman who lives in Cleveland, Ohio. On his 16th birthday, he got brand new Ford GTas a presentfrom his parents. To test his first car, he went on a ride with his father and exceeded the speed limit causing an accident, in which another parked car with no one inside got damaged. To soften the punishment for his son, the businessman bribed the police and the judge, and Jake did not face any consequencesfor his actions. Furthermore, the businessman paid a visit to local newspaper and bribed its head director, so the story did not get any resonance in media." The result of the culprit's actions is presented as positive externality. There is an emphasis on a fact that nobody was damaged in result of a culprit's actions. The second externality is negative, as in the first survey. Use of verb "bribe", multiple corruption acts and absence of punishment are expected to evoke negative feelings towards the individuals engaged in corruption activity. In the second case the externalities were presented in reverse order: 37 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT "Jake is a son of a rich businessman who lives in Cleveland, Ohio. On his 16th birthday, he got brand new Ford GTas a presentfrom his parents. To test his first car, he went on a ride with his father and exceeded the speed limit causing an accident, in which another driver got injured badly. When the police arrived, Jake'sfather explained to them that he wants to keep his son's record clean and asked to shift the responsibility for the accident on him, promising to make it up for police workers later. The police agreed on Jake'sfather's terms." Once again, factor of injured driver is expected to lower the acceptance of corruption. In the contrary, the second externality is positive, and is supposed to flatten the act of corruption in the participants' perception. To measure the level of tolerance towards the corruption activity, the participants were offered to evaluate their feelings after familiarization with each case. The participants were answering the following question by choosing one of the given options: "In your opinion, how acceptable is the behavior ofJake'sfather and the authorities from the law perspective in the previous situation? a) Totally acceptable b) Rather acceptable c) Rather unacceptable d) Totally unacceptable" The certain order of options was chosen to avoid the first-item bias, putting socially undesirable answer to be the first one. After the practical part of a survey, participants were asked to share some general information that does not reveal their identity. Participants were sharing their sex (M, F, 38 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT other), age (under 18,19-24, 25-40, over 40), source of income or field of activity (student or part-time employed person, employed person, unemployed or retiree) and approximate monthly income (under 500 EUR, 500-1000 EUR, over 1000 EUR). This information could be used to analyze the potential correlation between different groups of participants and their level of tolerance towards corruption, i.e., to test whether men tolerate corruption more than women, or whether people with high income are more tolerant towards corruption than people with lower income. 39 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT 3 Data description The total number of participants was 185 respondents. The general information about the participants was gathered in the second part of the survey, where the respondents were asked to answer four questions. In these questions participants indicated their sex, age, occupation, and approximate monthly income. Gained data and distribution of mentioned categories across surveys can be seen in the table and graphs below. 40 EXTERNALITIES A N D CORRUPTION: SURVEY EXPERIMENT Table 2: General information about the participants Category Survey 1 Survey 2 Total amount Sex M 47 41 88 F 55 41 96 Other 1 0 1 Age Under 18 0 0 0 18-24 62 63 125 25-40 34 18 52 Over 40 7 1 8 Occupation Student/Part-time employed 70 66 136 Employed 33 14 47 Unemployed/Retiree 0 2 2 Monthly income Under 500 EUR 52 42 94 500-1000 EUR 32 35 67 Over 1000 EUR 19 5 24 41 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT Sex 1 ? n 1 Z U -i r\r\1 U U oU D U ,1 n ^ _ — ll iill ii 1u Survey 1 Survey 2 • M • F • Other Total Figure 1: Participants' distribution by sex Age 140 Survey 1 Survey 2 Total • under 18 «18-24 "25-40 over 40 Figure 2: Participants' distribution by age 42 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT Occupation 160 Survey 1 Survey 2 Total • Student/part-time employed • Employed • Unemployed/retiree Figure 3: Participants' distribution by occupation Income too Survey 1 Survey 2 Total • under 500 EUR • 500-1000 EUR • over 1000 EUR Figure 4: Participants' distribution by monthly income 43 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT To determine if there is a significant difference between two proportions, the Fisher's exact test was used for sex and occupation categories, and Pearson's chi square test was used to calculate the differences for age and income categories. The results of statistical tests can be seen in the table below. Table 3: p-values for every category Category p-value Sex 0.6568 Age 0.028166 Occupation 0.0276 Income 0.029127 The results of statistical tests indicate that respondents were distributed equally with no significant statistical difference only in sex category. At p < .05, the differences in results from tests in age, occupation and income categories are statistically significant. Nevertheless, it is early to make any conclusions about the validity of the representative sample only after the statistical tests. While male and female respondents were distributed without significant statistical difference, other categories should not be analyzed thoroughly. Young students with low and middle income are the main part of representative sample, since the survey was shared mostly in MUNI information system and was answered mostly by the MUNI students. Moreover, if we exclude category with income over 1000 EUR, the result of Fisher's exact test for the income category shows the p-value = 0.4237, which is not significant at p < .05. The effect of mentioned categories and their contrast to the mean value is analyzed in multiple linear regression model. The nationality of respondents was not included among general information questions. Since the estimated number of different nationalities could potentially exceed 20, the 44 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT randomization mechanism would most likely fail to distribute them without significant statistical difference. In conclusion, although the representative sample of respondents cannot be considered as ideal in statistical means, it roughly has any effect on experiment results. Given the content of respondents' pool, consisting mostly of students, the sex of respondents can be considered as only important indicator of statistical variance between surveys' propor- tions. 45 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT 46 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT 4 Results 4.1 Respondents" reaction Cases, presented to survey respondents, differ in two possible externalities. First, the consequences of an accident formed two externalities: negative, where another driver was badly injured, and positive, where damage was done only to a car. Second, the actions of corruption element (culprit's father) were used to study a framing effect, i.e., how appropriate wording can change the perception of corruption. In the cases with the positive externality individual tends to shift the responsibility on him and the corruption is described implicitly. In the cases with the negative externality explicit and direct wording was used to describe the corruption act, that occurs after the accident. In the further analysis, these externalities are denoted as "Injury" - "No injury" and "Bribe" - "No bribe". The first survey consisted of a case with "Injury" and "Bribe" externalities and a case with "No injury" and "No bribe" externalities. The distribution of respondents' reactions to the survey can be seen on graphs below. 47 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT Injury, Bribe 93 Totally acceptable Rather acceptable Rather unacceptable Totally unacceptable Figure 5: Respondents' reaction to "Bribe" and "Injury" treatments No injury, No bribe 45 30 23 5 Totally acceptable Rather acceptable Rather unacceptable Totally unacceptable Figure 6: Respondents' reaction to "No bribe" and "No injury" treatments Cases in the second order were presented in further way: "No injury" and "Bribe", then "Injury" and "No bribe". The respondents' reaction is shown on graphs below. 48 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT No injury, Bribe 57 17 7 1 Totally acceptable Rather acceptable Rather unacceptable Totally unacceptable Figure 7: Respondents' reaction to "Bribe" and "No injury" treatments Injury, No bribe 43 21 16 2 Totally acceptable Rather acceptable Rather unacceptable Totally unacceptable Figure 8: Respondents' reaction to "No bribe" and "Injury" treatments First impression can tell a lot about the reaction to the externalities. First, the level of tolerance towards the corruption act is higher, when at least one of the externalities is positive and even higher if both are positive. Second, the level of tolerance in cases with mixed 49 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT externalities is roughly the same. Last, for the complete understanding of gathered data advanced methods of statistical analysis are needed. 4.2 2x2 factorial design A 2x2 factorial design is a type of experimental design that allows researchers to understand the effects of two independent variables (each with two levels) on a single dependent variable. In this experiment will be studied the effect of corruption act consequences ("Injury" and "No injury" treatments) and framing effect ("Bribe" and "No bribe" treatments) on level of tolerance towards the corruption. Framing effect Bribe No bribe Actions effect Injury Level of tolerance Level of tolerance Actions effect No injury Level of tolerance Level of tolerance Figure 9: Distribution of independent and dependent variables As can be seen, there are two independent variables, each with two treatments: • Independent variable 1: Wording • Treatments: Bribe, No bribe • Independent variable 2: Consequences • Treatments: Injury, No injury And there is one dependent variable: Level of tolerance 50 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT A 2x2 factorial design allows to analyze two following effects: main effect, the effects that just one independent variable has on the dependent variable and interaction effect, that occurs when the effect that one independent variable has on the dependent variable depends on the level of the other independent variable. Respectively, this allows to estimate the effects of wording and consequences on level of tolerance independently and to find out if there is any interaction effect between them. One way to conduct such analysis is to visualize main and interaction effects on a graph using means of the dependable variable from the table. Since the data on level of tolerance was gathered in text format, there is need to transform it into the integer numbers. The answers "Totally unacceptable" were presented as 0, meaning lowest possible level of tolerance, while the answers "Totally acceptable" were presented as 3, marking the highest level of tolerance. The means were calculated for every cell of 2x2 factorial design table. Framing effect Bribe No bribe Actions effect Injury 0,1 0,7 Actions effect No injury 0,4 1,2 Figure 10: Means of the dependable variable in different treatments 51 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT The mean level of tolerance for the cases with "Bribe" and "Injury" treatments is about 0.1, for the treatments "Bribe" and "No injury" is 0.4, for the treatments "No bribe" and "Injury" is 0.7, and for the treatments "No bribe" and "No injury" is 1.2. The following graph represents main and interaction effects on the level of tolerance. Level of tolerance 1,4 1,2 1,0 0,8 0,6 0,4 0,2 0,0 Injury No injury ^ ^ " B r i b e No bribe Figure 11: Graphical representation of means of the dependent variable Since the two lines are not flat on the graph, the injury treatment should have a statistically significant effect on level of tolerance. Then, since the two lines are far apart, there is a noticeable difference between the "Bribe" and "No bribe" treatments. Therefore, the bribe treatment should also have a statistically significant effect on level of tolerance. Last, since the lines are mostly parallel, there should not be a statistically significant interaction effect between bribe and injury treatments, i.e., the effect that bribe treatments has on level of tolerance, is not dependent on injury treatment, and vice versa. 52 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT 4.3 Regression model Regression models are used to describe relationships between variables by fitting a line to the observed data. Regression allows to estimate how a dependent variable changes as the independent variable change. Since two independent variables are presented in this study multiple linear regression is an appropriate model for conducting the analysis. The formula of multiple linear regression for this model is: y = Po + + / ? 2 X 2 + # 3 * 1 * 2 + £ where: • y is the predicted value of the dependent variable, level of tolerance in the model. • Pq is the value of y when all the parameters are set to 0, respectively, level of tolerance in treatments "No bribe" and "No injury". • P1 and P2 are the regression coefficients of in dependent variables X1 and X2, wording and consequences treatments respectively. • /?3 is an interaction term, measuring the joint effect of independent variables' influence. • £ is a model error, i.e., the level of variation in an estimation of y. To find the best-fit line for each independent variable, multiple linear regression calculates three things: • The regression coefficients that lead to the smallest overall model error. • The t-statistic of the overall model. • The associated p-value (how likely it is that the t-statistic would have occurred by chance if the null hypothesis of no relationship between the independent and dependent variables was true). 53 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT It then calculates the t-statistic and p-value for each regression coefficient in the model. While it is possible to do multiple linear regression by hand, it is much more commonly done via statistical software. Programming environment RStudio that uses language of programming R was used during this study. The function takes the most important parameters from the linear model and puts them into a table. Outcome of the regression model is depicted in Table 4. Table 4: Outcome of regression model, calculated in R Coefficients Estimate Std. Error t value Pr(>|t|) (Intercept) 1.208137 0.107113 11.279 <2e-16 Bribe -0.755703 0.106664 -7.085 7.37e-12 Injury -0.450825 0.106664 -4.227 3.01e-05 Age25-40 0.009748 0.086591 0.113 0.910 Age over 40 -0.083125 0.186039 -0.447 0.655 Gender M -0.063520 0.075131 -0.845 0.398 Gender Other 0.332332 0.503836 0.660 0.510 Income less than 500 EUR -0.006489 0.083877 -0.077 0.938 Income more than 1000 EUR -0.063008 0.121112 -0.520 0.603 Bribe:Injury 0.138568 0.154107 0.899 0.369 54 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT The regression coefficients of the model are displayed after "Coefficients" label. Row 1 of the coefficients table is labeled "Intercept" - this is the y-intercept of the regression equation. It's helpful to know the estimated intercept in order to plug it into the regression equation and predict values of the dependent variable. Level of tolerance = 1.208137 + (-0.755703 * wording treatment) + (-0.450825 * consequences treatment) ± e The most important things to note in this output table is next row - the estimates for the independent variables. The "Estimate" column is the estimated effect, also called the regression coefficient or r 2 value. The estimates in the table tell that including the "Bribe" treatment in corruption case lowers the level of tolerance by 0.755703, and that including an "Injury" treatment lowers the level of tolerance by 0.450825. Due to the values being so low (p < 0.001 in both cases), the null hypothesis can be rejected in cases of both "Bribe" and "Injury" treatments as well as in case of "Intercept" as the effect is significant at p<0.05. Findings from multiple linear regression model analysis imply several important issues. First, there is strong influence of "Bribe" and "Injury" treatments on level of tolerance towards the corruption act. The p-values are being lower that 0.001 implies the statistically significant effect of both negative externalities on level of tolerance is presented in the gathered empirical data. The results of regression model basically support the findings from factorial design. 55 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT Second, the interception result stands with the 2x2 factorial design findings, i.e., the level of tolerance with treatments "No bribe" and "No injury" is equal to 1.2 in both cases. This level denotes level of tolerance towards set of positive externalities. Third, the "Bribe" treatment has bigger impact on level of tolerance, than "Injury" treatment. Since the estimated effect of "Bribe" treatment is lower, it means that the level of tolerance towards cases when this externality is presented as negative, is lower than towards cases with negative "Injury" externality. According to these findings, the perception of corruption is more dependable on wording than prior circumstances of corruption act. Fourth, there is no statistically significant effect of any categories from general information (sex, age, occupation, or income), neither of them is being in big contrast with the means from the model. None of the above-mentioned categories have any influence on level of tolerance towards corruption acts. Moreover, the interaction term of "Bribe" and "Injury" treatments also does not have any significant influence on the results. In conclusion, the regression model outlines three immediate effects: Interception effect with two positive externalities and strong influence of treatments "Bribe" and "Injury". 56 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT 5 Discussion The findings from the previous chapters implies that there is statistically significant influence of negative externalities on the level of tolerance towards corruption acts. Unlike previous studies (Abbink, Hennig-Schmidt, 2006), the results from the regression model clearly show that smooth wording can lower the aversion to the corruption act by a significant percentage. Therefore, the public opinion about the corruption can be manipulated to certain extent. However, the appropriate wording is still to be determined by the competent authorities and implemented by media sources. For now, we can suppose that media sources that cover the corruption cases have some influence over the readers' perception of corruption. Another externality has weaker, but still significant effect on level of tolerance. The prior actions of individuals engaged in corruption activities have influence on immediate tolerance of their corruption activity. If the damage that was done by them is minor, the tolerance towards following corruption act is higher. But according to the previous experiments and their findings, the main source of corruption spreading is the corruption itself. In this context, it does not matter whether the corruption act was performed after minor accident or the huge one. The corruption act can start a chain reaction that will lead to higher corruption level in corresponding environment. This study questions the findings of previous experiments and provides a vast field for further studies. The corruption studies will benefit from expanding the representative sample, making it statistically independent with respect to different categories of respondents. Also, there could be noticeable differences across the countries and nations due to the cultural influence on perception of corruption. 57 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT In conclusion, the study can form a basis for further anti-corruption experiments that will prove or refute the study's findings in particular environment. Nevertheless, the externalities should get more attention in upcoming papers on corruption issues. 58 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT 59 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT Conclusion The study was aimed at analyzing the influence of negative externalities on the level of tolerance towards the corruption acts. The survey experiment was chosen as an appropriate method of gathering the required empirical data and conducting an econometrical analysis of it. The survey was divided into two different versions, each with certain given set of externalities in two cases. The survey was aimed at measuring the level of tolerance of respondents towards cases with corresponding set of externalities. Empirical evidence was gathered through the Google Forms surveys and extracted into the Microsoft Excel sheets, where the data was processed and represented in the paper. The statistical tests do not allow to consider the respondents' pool as variable enough because the basis of representative sample was formed by the students. However, the regression analysis has shown that none of the respondents' categories has any significant effect on the findings. The extracted data was analyzed in factorial design and in multiple regression model. The results of the analysis have shown the presence of the effect of negative externalities on level of tolerance of respondents. The level of tolerance was lower with the presence of at least one negative externality. Additionally, the framing effect seems to have more significant effect than the circumstances prior to the corruption act. These finding go in contrast to the previous studies. Among the possible reasons of it there are quality of representative pool and treatment definitions. Respectively, the further studies on this topic can shed more light on corruption's nature and can help to implement these findings into the working anti-corruption policies. 60 EXTERNALITIES AND CORRUPTION: SURVEY EXPERIMENT 61 REFERENCES References ABBINK, Klaus; HENNIG-SCHMIDT, Heike, 2006. Neutral versus loaded instructions in a bribery experiment. Experimental Economics. Vol. 9, issue 2, pp. 103-121. ANDERSON, James; GRAY, Cheryl W., 2006. Anticorruption in Transition 3: Who is Succeeding... and why? World Bank Publications. No. 7089. ARMANTIER, Olivier; BOLY, Amadou, 2011. A Controlled Field Experiment on Corruption. European Economic Review. Vol. 55(8), pp. 1072-1082. BARR, Abigail; SERRA, Danila, 2009. The effects of externalities and framing on bribery in a petty corruption experiment. Experimental Economic. Vol. 12, issue 4, pp. 488-503. BECKER, Gary, 1968. Crime and Punishment: An Economic Approach. Journal of Political Economy. Vol. 76, No. 2, pp. 169-217. BOBKOVA, Nina; EGBERT, Henrik, 2012. Corruption investigated in the lab: a survey of the experimental literature. University College London (UCL), Anhalt University of Applied Sciences Bernburg. MPRA Paper No. 38163. CAMPBELL, Donald T.; ROSS, H. Laurence, 1968. The Connecticut Crackdown on Speeding: Time-Series Data in Quasi-Experimental Analysis. Law & Society Review. Vol. 3, No. 1, pp. 33-54. CORBACHO, Ana; GINGERICH, Daniel W.; OLIVEROS, Virginia; RUIZ-VEGA, Mauricio, 2016. Corruption as a self-fulfilling prophecy: evidence from a survey experiment in Costa Rica. American Journal of Political Science. Vol. 60, No. 4, pp. 1077-1092. FISMAN, Raymond; WEI, Shang-Jin, 2004. Tax Rates and Tax Evasion: Evidence from "Missing Imports" in China. Journal of Political Economy. Vol. 112, No. 2, pp. 471-496. GUPTA, Sanjeev; DAVOODI, Hamid; ALONSO-TERME, Rosa, 2002. Does Corruption Affect Income Inequality and Poverty? Economics of Governance No. 3(1), pp. 23-45. GYIMAH-BREMPONG, Kwabena, 2002. Corruption, economic growth, and income inequality in Africa. Economics of Governance. No. 3, pp. 183-209. 63 REFERENCES HABYARIMANA, James, 2007. Why Does Ethnic Diversity Undermine Public Goods Provision. American Political Science Review. Vol. 101, issue 4, pp. 709-725. HART Sergiu, 2006. Robert Aumann's Game and Economic Theory. The Scandinavian Journal of Economics. Vol. 108, No. 2, pp. 185-211. KIRCHLER, Erich; MUELBACHER, Stephan; GANGL, Katharina; HOFMANN, Eva; KOGLER, Christoph; POLLAI, Maria; ALM, James, 2012. Combining Psychology and Economics in the Analysis of Compliance: From Enforcement to Cooperation. Working Papers from Tulane University, Department of Economics. No. 1212. KLEMENCIC, Goran; STUSEK, Janez, 2008. Specialized Anti-Corruption Institutions: Review of Models. Paris: Organization for Economic Cooperation and Development. LAMBSDORFF, Johann Graf; FRANK, Björn, 2010. Bribing versus gift-giving - An experiment. Journal of Economic Psychology. Vol. 31, issue 3, pp. 347-357. LUI, Francis T, 1985. An Equilibrium Queuing Model of Bribery. Journal of Political Economy. Vol. 93, No. 4, pp. 760-781. MAURO, Paolo, 1997. The Effects of Corruption on Growth, Investment, and Government Expenditure: A Cross-Country Analysis. IMF Working Papers. No. 96(98), p. 1. MEON, Pierre-Guillaume, 2005. Does Corruption Grease or Sand the Wheels of Growth? Public Choice. No. 122(1), pp. 69-97. MILGROM, Paul R., 1984. Review: Axelrod's "The Evolution of Cooperation". The RAND Journal of Economics. Vol. 15, No. 2, pp. 305-309. MORRIS, Stephen D., 1991. Corruption & politics in contemporary Mexico. NOONAN, John T. Jr., 1984. Bribes. OLKEN, Benjamin A ; BARRON, Patrick, 2009. The Simple Economics of Extortion: Evidence from Trucking in Aceh. Journal of Political Economy. University of Chicago Press, Vol. 117(3), pp. 417-452. RAPPAPORT, Alfred, 1986. Creating Shareholder Value: The New Standard for Business Performance. Simer and Schuster Publishing Group, New York. 64 REFERENCES REINIKKA, Ritva; SVENSS0N, Jakob, 2004. Local Capture: Evidence from a Central Government Transfer Program in Uganda. The Quarterly Journal of Economics. Vol. 119, No. 2, pp. 679-705. SERRA, Danila, 2006. Empirical determinants of corruption: A sensitivity analysis. Public Choice. Vol. 126, pp. 225-256. SUNDSTROM, Aksel, 2016. Corruption and violations of conservation rules: A survey experiment with resource users. World Development. Vol. 85, s. 73-83. TANZI, Vito, 1998. Corruption Around the World: Causes, Consequences, Scope, and Cures. IMF Staff Papers. Vol. 45(4), pp. 1-1. United Nations Office on Drugs and Crime, 2005. UNITED NATIONS CONVENTION AGAINST CORRUPTION. WEI, Shang-Jin, 2000. How Taxing is Corruption on International Investors? The Review of Economics and Statistics. Vol. 82, issue 1, pp. 1-11. YILMAZ, Serdar; BALTACI, Mustafa, 2006. Keeping an Eye on Subnational Governments: Internal Control and Audit at Local Levels. World Bank Publications. No. 37257. 65