Universiteit van Amsterdam Amsterdam School of Economics Game Theory and Behavioral Economics Track
Master Thesis
Characteristics Of Corrupt
Administration
A study of the environmental settings that allow and promote
embezzlement
Submitted by Idan Cohen Student number 11084863
2
Thesis Summary
This paper aims to deal with embezzlement, a form of corruption that is rarely studied. A group decision making game called The Governor Game is introduced and used in order to put to the test a few commonly suggested strategies to reduce corruption.
Statement of Originality
This document is written by Student Idan Cohen who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of
3
Table of Contets
Introduction ... 4
Corruption and embezzlement ... 5
Is corruption good or bad? ... 6
Corruption as a form of taxing ... 6
Corruption as a competition amplifier ... 6
Cutting Red Tape ... 7
Changing officials incentives ... 7
Waste of resources ... 8
Relatedness over efficiency ... 8
Increase in Red Tape ... 9
Tackling corruption ... 10
Staff rotation... 10
Transparency and freedom of information ... 12
The Behavioral Ethics Approach ... 13
The Governor Game ... 15
Method ... 18
Statistical model ... 23
Results ... 24
Findings 1 & 2 – Percentage of tokens taken from the public account is not significantly lower in T2, but is significantly lower in T3. ... 26
Discussion findings 1&2 ... 27
Findings 3 & 4 – Wealth levels reached in T1 are significantly lower compared with T2 and T3 ... 27
Discussion findings 3&4 ... 28
Conclusion ... 28
References ... 30
4
Introduction
Watching the 8 O’clock news and hearing about dictators stealing their own people’s money in distant regions of the world might cause people to think that acts of embezzlement are only performed by ego-maniacs that fought their way to power in exotic countries, but is that really the case? Or maybe in each and every one of us lays an embezzler that might raise its head given the right environmental settings?
First let’s face the facts, corruption is everywhere. It does not matter how developed our country is, some extent of corruption is present (Transparency International, 2015). According to Transparency International, out of the 7.6 billion people living in the world today, more Than 6 billion people live in countries with a serious corruption problem. Low income countries lose US$1 trillion a year to corruption (Transparency International, 2015). To that we can add billions in funds invested to fight corruption and the end result is clear. Corruption is extremely common and costly to all.
So, does embezzlement still feel so uncommon and far away? It was those questions that led to the writing of this paper.
Embezzlement, a very common form of corruption, is mostly left untouched when it comes down to previous studies. Understanding that studying embezzlement in the real world is not going to be an easy task, I have decided to conduct an experiment that will allow me to
contribute to the literature in two ways. The first was to offer a new experimental procedure to study embezzlement. The second was to examine the efficiency of strategies suggested to fight bribery on reducing embezzlement. Therefore I came up with the following research question: Does staff rotation and transparency are environmental conditions that affect the extent of tokens taken from a public account in a Governor Game, and how the amounts taken from the public account change the public’s willingness to invest?
5
The data gathered during this study is showing yet again how effective transparency is in reducing corruption. Moreover, the data gives away some evidence to the importance of the behavioral ethics insights and how the likings of norms can affect the likelihood of an ordinary person to take unethical actions like embezzling from a public account in an experimental setting.
Throughout the paper I will discuss the models and studies I found most relevant to the field of corruption. I will present the view of different approaches and paradigms on that complex subject and will then present the experimental method I have used in my study. Concluding remarks and suggestions for further research will close this paper.
Corruption and embezzlement
The first step with dealing with corruption is to define the acts that fall under this set of actions. There are plenty of suggested definitions for corruption, and it seems there are two core
concepts that exist in all of them. The first is the low moral standard that the corrupt person demonstrate, and the second is the position of power or trust the corrupt person holds and misuses for his or relatives benefits. I choose to follow one particular definition that was given by J.S Nye (1967), which I feel deals with corruption in the best manner, and captures both of the core aspects mentioned: “corruption is behavior which deviates from the formal duties of a public role because of private regarding pecuniary or status gains… This includes such behaviors as bribery, nepotism and misappropriation (illegal appropriation of public resources for private regarding uses).”
Embezzlement defiantly falls inside those lines, but it is important to have a clear definition for that specific form of corruption in hand, in order to fully assess that behavior. Embezzlement as defined by the Federal Bureau of Investigation is “misappropriation or misapplication of money or property entrusted to one’s care, custody, or control”. The main difference distinguishing embezzlement from other forms of theft is the position of power or trust that has allowed for the act of stealing to take place, and therefor allows for dealing with embezzlement just like any other form of corruption.
6
Is corruption good or bad?
Although there is an agreement between the vast majorities of countries regarding the adverse effects of corruption on a given country, there is still somewhat of a debate in the academic world about the consequences that corruption may bring about (Da-Hsiang Donald Lien, 1990). I will start with presenting the arguments that see corruption as benefiting with a country and later will move on to try and dissolve them and claim that having a corrupt administration is not beneficial what so ever. While corruption can influence a country in so many aspects, I shall focus on the economic aspect and on economic growth in particular.
Corruption as a form of taxing
Based on J.S Nye’s article from 1967, corruption may be used in order to allow for capital formation. A country that faces low private capital levels, may find it hard and get criticized for openly taxing citizens. In such case corruption, which is a relative of taxing (Shleifer and Vishny, 1993), may solve that issue. The problem is obviously the question regarding the use of the raised capital. In the case of taxes the money is presumably spent to benefit with the public, while in the case of capital gained in corrupt methods the money is used mostly to benefit with the government officials.
Corruption as a competition amplifier
Corrupt behavior is beneficial to the economy as it forces competition between companies for some kind of governmental good (Leff 1964). If we assume that governmental goods are scarce resources, then companies that have to obtain such a good in order to operate will compete for that right. While in the absence of corruption we might see companies compete for the
governmental goods, in the presence of corruption those goods are more costly and therefor allows only for the more efficient companies to compete. The bigger more efficient company can pay an official more than the less efficient company, which will in turn result in one of two solutions. The first is the operation of the highly efficient and more profitable companies in the industry. The second is the efficiency enhancing process that low efficient companies will have to go through in order to stay in business. In either case, without discussion regarding the final distribution of wealth, it is assumed that a higher level of efficiency is gained since corruption
7
increases competition and drives out of the game the marginal efficient players (Lui, 1985 and Bardhan, 1997).
Cutting Red Tape
In some countries bureaucracy can significantly delay any new business or service, which in turn slows down economic growth in the Macro level. That is the reason money transferred to overcome those obstacles is also referred to as speed money. For example we can think of an entrepreneur looking to open a restaurant. It may be the case that first he have to rent and decorate the place, and only then start working towards obtaining the required permits from the fire department and the police. Let’s also assume the restaurant cannot host customers until the permits are obtained, meaning the entrepreneur is bearing the costs of rent with no ability to cover them through generating income. In such case, even though the restaurant may be beneficial and in high enough demand to be profitable, the entrepreneur may decide to stay out of the market if the period for obtaining the permits is too long, and even if he does enter the market, there is a substantial delay in economic growth. Since such situations are
apparently very common in some places, they have been given the generic term “Red Tape”. Corruption is assumed to be a very useful cure against these kind of situations (Shleifer and Vishny, 1993, J.S Nye, 1967).
Changing officials incentives
Assuming that government is essential for the stable growth of a country, we can also assume that it is the work of the governmental officials that enhances growth. Adding those together we can say that the more effort and work the governmental officials put in is beneficial to the country’s growth. Incentives obviously play a major role in the choice of effort each official choses, and corruption may be interpreted as a piece rate bonus to be added to the fixed wages officials earn (Mauro 1995).
The listed above are some of the more common explanations to the contribution of corruption to economic performance and as they all make sense, at least theoretically, it is important to try and dissolve them and discus the adverse effects of corruption and in particular to show
8
how the suggested benefits may hinder economic growth or direct it towards less socially desirable directions.
Waste of resources
Although corruption may be a form of taxing, it is barley socially beneficial and the costs to a country’s capital could be significant. For instance, it is estimated that in only five years four Latin American countries have lost a total of 1.15$ billions that was taken from the public treasury by the countries’ leaders who chose to store their capital in off-shore accounts (Lieuwen, 1960).
In another example Mobutu, the former president of Congo has allocated governmental funds to his own private accounts, which at the mid 80’s summed up to around 4$ billion (Burns et al, 1997). Clearly, not a lot of those funds were destined to upgrading roads and education
systems across Congo.
Relatedness over efficiency
Underlying the claim that competition is amplified by corruption, is the assumption that the company that pays more will be the one that earns the governmental good. Looking deeper in to that we can say that once a public servant is given the right to decide which company is awarded a certain contract or permit, he is just a likely to choose the company which he is more related to. For example we can think about two construction companies competing for the permits to buy a new building. The owner of company A is not related to any official working in the construction and planning committee. The owner of company B on the other hand is the brother in law of the chair of the committee, which is also the person who under different circumstances would have sold his vote to the highest bidder. It is easy to imagine that under those or similar conditions the chair would have preferred to award the related person with the permit, even if the unrelated person have made a higher offer.
moreover, this kind of competition over permits may be analyzed as an ‘all pay auction’ (Lien, 1990), which may be beneficial to the auctioneer, the governmental official in this case, but could result in deterring efficient companies from competing for the good.
9
Increase in Red Tape
The claim that corruption is useful in cutting Red Tape is almost uncontested, but it is important to understand what other issues may raise from the presence of corruption. First, one may suggest that it is in the bureaucrats interest to impose more Red Tape than socially beneficial in cases where corruption is vastly present(Soto, 1989). This process will keep on driving up both corruption and Red Tape. As for speed money, Gunnar Myrdal (1968), citing the 1964
Santhanam Committee on the Prevention of Corruption appointed by the Government of India, has argued that corrupt officials may, instead of speeding up, actually cause administrative delays in order to attract more bribes.
The second issue arrives from the assumption that bureaucracy is necessary to some extent. We do want the police and the fire department to make sure that the new restaurant that is about to open is properly equipped to deal with any emergency situation that may arise. If the job of the inspector is to gather information about the likelihood of damage to the public by the operation of the restaurant we do want him to follow the protocol and report back his true findings. With corruption and speed money in place I can assume that the official is less likely to report his true findings regarding the safety of a new business. This is also a process that can keep on growing to the point where all restaurants owners are paying the money they have intended to put in to safety measures, like fire detectors, to officials that are in charge for the fire safety permits.
It seems that better compensation offered to the public servants has no real damaging effect when it is formal and legal, and the opposite may even be the truth in that case, since better incentives will allow to attract and keep better public servants which will result in better public services. It is important though that the entire compensation will be paid by the government, since the officials are to serve the entire public as a whole and provide the same service to each member of a country without differences based on ones means.
10
Tackling corruption
Understanding that corruption does harm country’s economic growth and development, it is now important to discuss the suggested ways to be taken in order to be able to deal with corruption. Various organizations, such as the Federal Bureau of Investigation and the United Nations, have declared the fight against corruption as the most important mission they
currently face. Countries that formerly aided developing and crisis struck areas are now turning down requests for help from countries perceived as corrupt. The fight against corruption is taken to a higher level, but in order to succeed in that fight a proper strategy and proper means are to be taken. In the following paragraphs I will discuss a few of the more popular measures of fighting corruption. While ways to fight corruption vary significantly, ranging from efficient wages (Becker & Stigler, 1974) to indoctrination (Klitgaard, 1991), I will focus mainly on the methods that will appear and be examined using the experimental method.
Staff rotation
Staff rotation has been offered as one way that may help mitigate corruption through changes made in the organizational level. The German government have released a directive to
implement staff rotation in various areas. On a series of publications and research Abbink (1999, 2004) have created a theoretic ground accompanied with laboratory experiments to examine the effect staff rotation has on corruption. On the latter he found that staff rotation has significantly reduced transactions of bribes. The explanation offered by Abbnik is that a relationship between the giver of a bribe and the receiver is based upon trust and reciprocity, and by moving the official from his role you cancel any chance of a long term relationship. Adding to that, by changing the officials that serve at a given role one can expect a higher level of uncertainty regarding the reaction of an official to the bribe offered.
While there is a lot of sense in the ideas presented above, accompanied with an impressive experimental methodology and results, there is some shortcomings to Abbnik’s claims. First, the author targeted corruption but only took bribes into consideration in his experimental design. In this paper I wish to build upon Abbnick’s findings and examine the suggested strategy in reducing embezzlement. Another question raised is regarding the termination of
11
relationships due to rotation. As I see it both ends of a bribe transaction are bound to each other by their act, since each side holds powerful information about the other. Keeping that in mind, and adding to that the reasonable assumption that the official, even if rotated, will stay and serve in the same ministry and will maintain the same influence and power will give a pretty grim end result. The relationship between the sides will survive a rotation and the official will still be able to influence, even if to a lower extent, the decisions made in his workplace. Furthermore, the rotated official will be able to give information to the briber about the moral standards of the new official, and by that will mitigate the uncertainty offered by Abbnik. A field research about the correlation between staff rotation and corruption is yet to be found, but there is sense in looking into the relation between the bigger form of governance and corruption. First, democracy could be perceived as a form of staff rotation in the highest levels, while long lasting dictatorship could be described as the lack of rotation. Moreover a change in elected ministers will trickle down to medium and low levels staff and will bring about a greater rotation.
The relationship between form of governance and corruption is widely discussed in the literature. Studies on both theoretic end (Morrow et al. 2001, Olson, 1993, Przeworski et al., 1995) as well as the data based studies (Saha et al., 2009, Ades & Di Tella 1999) have pointed a negative correlation between democracy and corruption. An important fact is that though different indexes for corruption used in Saha et al. (CPI) and in Ades & Di Tella (ICRG) the results are suggesting the same relationship.
Probability of being convicted and severity of punishment
The main idea standing behind sever punishments and increased likelihood of being caught can be grasped using the following rather simplistic expected utility approach similar to the one offered by Becker (1968):
𝐸(𝑈) = (1 − 𝑝) ∗ 𝑈(𝑀) + 𝑝 ∗ 𝑈(𝐶)
Where p is the probability of being caught, U(M) is the utility gained from the money obtained through a corrupt act and U(C) is the utility lost once caught. Since U(C) is negative, any
12
a smaller U(C) or in words, a more severe punishment. On the other hand, as p approaches zero the expected utility gained from corruption is positive and the punishment itself has almost zero marginal effect on the decision.
This simple model also gives away one of the more problematic issues in the fight against corruption in countries where corruption is deeply rooted. While there is an agreement among scholars that higher probability of getting caught reduces the undesired activity (Becker and Stigler, 1974, Goel and Rich, 1989), the chances that a corrupt official will be caught and prosecuted in developing and third world countries are very low (Shleifer and Vishny, 1993, Fantaye, 2004). It is important to understand that increasing this probability includes different norms in the law enforcement and prosecution systems, which are a lot of times the benefiters of the corrupt way of life.
Transparency and freedom of information
Before getting to the heart of the importance of transparency, it is useful to define what transparency really is. One possible definition given by the OECD (2002) suggests that a
transparent environment is one in which economic agents possess essential information about the environment in which they operate. A lot of other definitions are in place, but all of them share the concept of access to information as key element of transparency.
Traditionally, there are three types of anti-corruption approaches (Shim & Eom, 2009), and transparency is a powerful tool serving at least two of this approaches. The law enforcement approach is based upon similar logic as the previously explained model and the concept of transparency serves that approach be increasing the probability of being caught and prosecuted for malpractice.
Social empowerment is the second approach to benefit from transparency. This approach is said to be the most efficient one in countries where corruption is deeply rooted and the core idea of this approach is to give the public more power and the means to asses and supervise the acts of the government (Johnston, 1998). Regardless of the other steps that this approach is suggesting, transparency is the first one to be taken on the way for a change based on social empowerment.
13
While it seems like transparency is widely agreed to be a useful mean in reducing corruption in the theoretical end, the empirical evidence may give different results. Bellver and Kaufmann (2005) have looked into the effect political transparency have on reducing corruption as measured by the CC index, and found that the relationship is non-linear and negative. Their conclusion is that political transparency is most useful where political competition is present. Kolstad and Wiig (2009) are suggesting that “though transparency may affect corruption in several ways, it is insufficient in itself” and further add that “In certain cases, transparency can also increase corruption”.
The Behavioral Ethics Approach
As mentioned earlier, the classic perception claims that acts of corruption involve low moral standards, but is it really the case? Or maybe it is the situation and environment in which the decision is being made that make ethical people turn sour? The Behavioral Ethics approach is set to understand and describe the situational and social forces that influence the human behavior when facing an ethical situation, and trying to find ways to nudge people into making the more ethic decision (Bazerman & Gino, 2012).
Behavioral ethics deviates from the classic cost-benefit analysis presented earlier in at least one critical way, which is the assumption that there is some kind of inner non-monetary costs or benefits to the decisions we make.
That mechanism can be broken down to a simple principal. A person gains from feeling ethical, and unethical behavior comes with a cost. It is important to note that all of those inner costs and gains are subjective and hard to quantify, but the decision will be made by aggregating the utility gained through both the monetary and non-monetary gains.
One possible way of presenting an ethical dilemma is the choice between the following options: 𝐸(𝑈(𝑀, 𝐼) = (1 − 𝑝) ∗ 𝑈(𝑀𝑢𝑛𝑒𝑡ℎ𝑖𝑐𝑎𝑙) + 𝑝 ∗ 𝑈(𝐶) + 𝑈(𝐼𝑢𝑛𝑒𝑡ℎ𝑖𝑐𝑎𝑙)
𝑈(𝑀, 𝐼) = 𝑈(𝑀𝑒𝑡ℎ𝑖𝑐𝑎𝑙) + 𝑈(𝐼𝑒𝑡ℎ𝑖𝑐𝑎𝑙)
Note that 𝑈(𝐼) is not affected by the probability of getting caught. This means that in cases where p=0 (e.g. some lab experiments) the model is simplified to
14
𝑈(𝑀, 𝐼) = 𝑈(𝑀𝑢𝑛𝑒𝑡ℎ𝑖𝑐𝑎𝑙) + 𝑈(𝐼𝑢𝑛𝑒𝑡ℎ𝑖𝑐𝑎𝑙)
𝑈(𝑀, 𝐼) = 𝑈(𝑀𝑒𝑡ℎ𝑖𝑐𝑎𝑙) + 𝑈(𝐼𝑒𝑡ℎ𝑖𝑐𝑎𝑙)
Where:
p is the probability of being caught
U(𝑀𝑢𝑛𝑒𝑡ℎ𝑖𝑐𝑎𝑙) is the utility gained from monetary result of the unethical act
𝑈(𝑀𝑒𝑡ℎ𝑖𝑐𝑎𝑙) is the utility gained from monetary result of the ethical act U(C) is the utility lost once caught
𝐼 is a function grasping the difference between people’s actual behavior and their desire to maintain a positive moral self-image.
It is normally the case that a moral decision is in place when𝑈(𝑀𝑢𝑛𝑒𝑡ℎ𝑖𝑐𝑎𝑙) > 𝑈(𝑀𝑒𝑡ℎ𝑖𝑐𝑎𝑙), meaning that the utility gained from the monetary payoff associated with an unethical decision is bigger than the one associated with the ethical one. The inner mechanism is the one that therefor can turn the decision around and have a person follow his moral code and choose the moral act although a monetary loss in involved. The idea that an unethical decision has some cost in utility terms is based upon the ethical dissonance model (Barkan et al., 2010). On the other hand ideas of the likings of the warm glow effect offered by Anderoni (1990) are the base for the utility gained by moral decisions. This form of analysis could also be useful in explaining why people do chose to preform unethical acts in laboratory studies, but do not chose to completely exhaust that option (Mazar, Amir, and Ariely, 2008). This behavior could be explained by diminishing utility gained from money earned in an experiment and increasing marginal inner-cost of immoral behavior.
The behavior of others
There are at least three ways in which the behavior of others can affect the actions taken by an individual facing an ethical decision. In line with the previously presented cost benefit analysis, an individual can update the probability of being caught after he witness the outcome brought upon a corrupt peer. Second is the saliency of ethicality at the moment of the decision. A study examining this hypothesis found that reading the 10 commandments before facing an ethical
15
decision have reduced the presence of dishonest behavior (Mazar et al., 2008), so any other form of bringing moral issues up during an everyday chat could do the trick. The third way is simply getting a clearer view on the social norms regarding dishonest behavior. The Norm Hypothesis is one that been offered in a few different studies (Mazar et al., 2008, Shalvi et al., 2015) and it seems as one of the more common explanation behavioral ethics has to offer for the phenomena of ethically wrong doing.
The norm hypothesis has another very compelling aspect since corruption may sometimes be perceived as a non-criminal act, and that is one of the explanations offered by officials to why is it so ample in some regions of the world. By changing the norms in a society we actually change the moral benchmark people in that society have regarding some acts. Later in this paper I will try to examine the importance norms may have on the behavior of a group in an unfamiliar situation.
Moral Credit
Nisan (1991) offers the moral balance model and suggests that people try and balance their good and bad deeds. If one feels like she acted in a very moral manner lately, she is more likely to perform an immoral act since she got ‘moral credit’ to spare. An experiment conducted by Monin and Miller (2001) have provided evidence that Nisan’s model does in fact hold in reality. In a more recent study Jordan et al. (2011) have extended the domains of morality in which this kind of moral accountancy holds. Thinking about how this kind of model may look like using statistical tools had me thinking about some sort of a stationary time series. Some highly moral acts may increase the more credit and some immoral acts will lower it, but eventually those differences are due to cancel each other out and the ‘moral account’ is supposed to have the same average level of moral across long periods of time. This theory as well will be put to the test later on.
The Governor Game
As can be understood form the discussion above, there is still a need to learn the effects those methods have not only on bribery, but on other forms of corruption as well. My intention, as
16
stated before, is to examine how some of those methods will affect the amounts of money embezzled from a public treasury. In order to do that I introduce the Governor Game. The game outline is as follows. A group of N human participants is formed and those are randomly assigned to one of two roles, the governor role and the public role. In each group one player is chosen to play the governor and the other N-1 are chosen to play the public.
Each public player is endowed with 10 game tokens and asked how he would like to divide them between two investment options. The first option is to invest in a private account. This account belongs to the player and whatever invested in that account is his to keep. Every token invested in that account is worth one token in the end of the round. The second options is to invest in a public account. This account belongs to all N-1 public players. Whatever invested in this account by the public players is multiplied by some factor.
The multiplication factor stands for the rapidly diminishing marginal cost of some commodities or services once purchased by an increasing amount of people (Cornes & Sandler,1996).
Another important motivation for the multiplication factor was to incentives the participants to invest in the public treasury, since the formulation of a public treasury is essential to study embezzlement.
After the public players have made their decision, it is the turn of the governor to play. At the beginning of each round the governor is endowed with some tokens. The amount of tokens endowed to the Governor may be higher equal or lower to the 10 tokens awarded to the public, depending on the goals of a given research.
The decision the governor faces is how much money to take from the public account. He may take any integer between 0 and the entire amount collected in the public treasury. Whatever is left in the public treasury after the governor has played is equally divided between the N-1 public players. None of the players is aware to the amounts contributed by other players to the public account. Each public player know about his own contribution and the governor only knows about the sum of tokens accumulated in the treasury. In the Governor Game a public investor may be exploited twice. Once by another public player that free rides and once by the governor who decide to take money off the public account, and in the basic game a public
17
player has no capability to distinguish between the two. The amount of information given to the players may also serve as an interesting variable and I shall use one possible manipulation in my research as will be presented in the method chapter.
Formally writing the payoffs of the players yields the following:
𝑃𝑎𝑦𝑜𝑓𝑓 𝑓𝑜𝑟 𝑃𝑢𝑏𝑙𝑖𝑐 𝑝𝑙𝑎𝑦𝑒𝑟𝑖 = 10 − 𝑃𝑢𝑏𝑖 + (∑ 𝑝𝑢𝑏𝑖∗ 𝛾 𝑁−1 𝑖=1 − 𝐺𝑜𝑣) ∗ 1 𝑁 − 1 𝑃𝑎𝑦𝑜𝑓𝑓 𝑜𝑓 𝐺𝑜𝑣𝑒𝑛𝑜𝑟 𝑝𝑙𝑎𝑦𝑒𝑟 = 𝑇𝑜𝑘𝑒𝑛𝑠 𝐸𝑛𝑜𝑑𝑜𝑤𝑒𝑑 + 𝐺𝑜𝑣
Where 𝑃𝑢𝑏𝑖is the amount public player i have invested in the public account, 𝐺𝑜𝑣 is the
amount taken by the governor from the public account and 𝛾 is the multiplication factor. Analyzing the game for the one shot case gives that a fully rational governor will take away the entire amount collected in the public treasury. This yields ∑𝑁−1𝑖=1 𝑝𝑢𝑏𝑖 ∗ 𝛾= 𝐺𝑜𝑣 and will result in the following public player payoff:
𝑃𝑎𝑦𝑜𝑓𝑓 𝑓𝑜𝑟 𝑃𝑢𝑏𝑙𝑖𝑐 𝑝𝑙𝑎𝑦𝑒𝑟𝑖 = 10 − 𝑃𝑢𝑏𝑖
Where it is easy to see that the optimal action of a rational public player is to invest all 10 tokens in the private account, keeping 𝑃𝑢𝑏𝑖 = 0.
The theoretical predictions in the infinitely repeated form of the game are much more
complicated. My assumption is that a public player that will receive less tokens than invested in the public account, will invest less in the public account in the following round. On the other hand, a public player that received back more tokens than invested in the public account will invest more in the public account in the following round. The strategic thing for a governor to do is to make sure that she give the public enough money back in order to maintain a flow of tokens into the public account. It is important to note that this is only the case where the game is infinitely repeated. If the game has a known end point, backward induction analysis will give the same result as the one shot game case. One way to prevent an ‘end effect’ is to tell
participants that the number of rounds to be played is randomly determined. In that case a discount factor may be introduced into the expected utility function of the players. The
18
interpretation of the discount factor will be the believe a player holds regarding the probability of continuation of the game. The expected utility to be maximized will then be of the following:
𝐸(𝑈) = 𝛿0∗ 𝑝𝑎𝑦𝑜𝑓𝑓0 + 𝛿1∗ 𝑝𝑎𝑦𝑜𝑓𝑓1+ ⋯ + 𝛿𝑛∗ 𝑝𝑎𝑦𝑜𝑓𝑓𝑛
Where 𝑝𝑎𝑦𝑜𝑓𝑓𝑛 is the payoff of a given player in round n and δ is the player’s objective discount factor.
The governor game is grasping the core issues of the act of embezzlement, since it allows for the creation and accumulation of public funds, and for those funds to be taken in order to benefit with a person who is not one of the funds benefiters. The game also allow for
manipulations that may isolate variables and point out effectiveness of methods in reducing the amounts taken from public accounts. To add to that, it is also possible to measure the public’s willingness to invest under different situations and to compare the effect of embezzlement on investment in public goods in an experimental setting. The maximum efficiency will be gained if the entire public will invest the given endowment in the public account while minimum
efficiency is the other extreme case where none of the public players is willing to invest even one token in the public account.
The introduced “Governor Game” can be seen as a hybrid of the Trust Game and the Public Goods Game, in the sense that the first stage of the game is exactly identical to the Public Goods Game and the money collected is then being entrusted with the Governor, in hope that it will pay back.
Method
64 participants were recruited using Facebook to play the Governor Game under three different treatments to which they were randomly assigned. Participants were not paid. Instructions given to players are presented in appendix 1.
First presented are the parameters that were identical across all three treatments:
The game is played in a group of four human participants. Three players play the public and one player plays as governor.
19
At the start of every round, the public type players receive 10 tokens and the governor receives 8 tokens.
A total of 8 rounds are played in each game. Participants were told that the number of rounds will be determined randomly.
In all rounds, the public type players are the ones to make their decisions first. Each public type player is asked how he would like to split the 10 endowed tokens. There are two options for the usage of tokens.
o Private account – every token invested in the private account is multiplied by 1, and whatever is invested in the private account belongs to the player.
o Public account – every token invested in the public account is multiplied by 4 (𝛾 = 4), and whatever is accumulated and left in the public account is shared with all other public type players equally.
After all the public type players made up their choices, it is the turn of the governor type player to play. He is aware to the amount in the public account and he needs to decide how many tokens to take from the public account. He does not know how much each player invested.
An example of a round played:
Public player 1 sends 5 tokens to the public account and 5 token to the private account Public player 2 sends 7 tokens to the public account and 3 token to the private account Public player 3 sends 3 tokens to the public account and 7 token to the private account Total amount of tokens invested in the public account is therefor 5 + 3 + 7 = 15. Now, after the decisions are made the tokens are to be multiplied by 4, so we get that the public account actually have 15 ∗ 4 = 60 tokens in it.
The governor is informed that there are 60 tokens in the public account. He chooses to take 30. After the governor played we are left with 60 − 30 = 30 tokens in the public account, which will then be equally divided between the three public players, meaning 303 = 10 for each one of the public type players.
20 The final amounts of this particular round are
Public player 1: 5𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑝𝑟𝑖𝑣𝑎𝑡𝑒 𝑎𝑐𝑐𝑜𝑢𝑛𝑡+ 10𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑝𝑢𝑏𝑙𝑖𝑐 𝑎𝑐𝑐𝑜𝑢𝑛𝑡 = 15 Public player 2: 3𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑝𝑟𝑖𝑣𝑎𝑡𝑒 𝑎𝑐𝑐𝑜𝑢𝑛𝑡+ 10𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑝𝑢𝑏𝑙𝑖𝑐 𝑎𝑐𝑐𝑜𝑢𝑛𝑡 = 13
Public player 3: 7𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑝𝑟𝑖𝑣𝑎𝑡𝑒 𝑎𝑐𝑐𝑜𝑢𝑛𝑡+ 10𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑝𝑢𝑏𝑙𝑖𝑐 𝑎𝑐𝑐𝑜𝑢𝑛𝑡 = 17
Governor player: 8Initial payment + 30𝑡𝑎𝑘𝑒𝑛 𝑓𝑟𝑜𝑚 𝑡ℎ𝑒 𝑝𝑢𝑏𝑙𝑖𝑐 𝑎𝑐𝑐𝑜𝑢𝑛𝑡 = 38
In order to describe the differences between the treatments, I will first describe the baseline treatment, which will be called T1 (Treatment 1).
In this treatment the governor type player is randomly chosen at the beginning of each round. This means that each one of the four group members have an equal probability of 14 to be chosen to play as governor at the beginning of a given round. Moreover, none of the players is aware to any of the other players decision. Using the previous example, player 2 know that he sent 7 tokens to the public account and he is being told that he got 10 back. He does not how much money player 3 invested in the public account and neither does he know how much money the governor took from the public account.
In T2, everything is the same except the fact that the governor is chosen at the beginning of the game, for the rest of the game. This means that the same group member is chosen to play the governor for all 8 rounds of the game. Besides that difference, everything else is exactly the same. My intention in creating T2 was to compare a situation with (T1) and without (T2) staff rotation.
The difference between T3 and T1 is that participants receive more information about the amounts collected in the public account. Looking to the example, each public player will be informed that 60 tokens were collected in the public account and that 30 tokens were kept by the governor. The rest, including the process of choosing the governor is identical to T1, meaning that a governor is randomly chosen at the beginning of each round. My intention in creating T3 was to examine the effect transparency does or does not have on embezzlement.
21
The goal of the experiment was to put to the test transparency and staff rotation as
anti-embezzlement methods, and the different treatments were designed in order to try and isolate the effects those methods have on the extent of the act.
The extent of the act was directly measured as the amount and percentage taken from the public treasury in each round and across all rounds, and the consequences of the acts of the governors may be measured using the wealth level reached in each group. Multiplication factor across all treatments was 4 (𝛾 = 4). In order to incentives public investment a multiplication factor greater than 3 was needed, to allow for Marginal Per Capita Return (MPCR) to be greater than 1. In this specific case MPCR was 1.33. The maximum wealth level reachable under those specifications is 120, and it will only be reached if all 3 public players will contribute their entire endowment to the public account.
The reason the governor is endowed with less tokens than the public to begin with is an attempt to grasp the disadvantageous wage compensation governmental officials face compared with their private sector counterparts.
Hypothesis 1:
Percentage of tokens taken from the public account in T3 will be lower than in the T1. This hypothesis is based upon findings from Kanagaretnam et al. (2010), where they have studied differences in the amount of money sent by the trustor and sent back by the trustee in the investment game, manipulating the amount of information given to the participants. Hypothesis 2:
Percentage of tokens taken from the public account in T2 will be lower comparing with T1. This hypothesis is the more interesting one in my opinion. There is a common way of thinking that claims more money is embezzled by dictators around the world comparing with
democratic leaders, but maybe the case is not as simple as it seems. Maybe dictators do have more money embezzled in total, but it is only because they are ‘playing more rounds’ as governors. Put another way, democratic leaders maybe embezzling more each round but because they are playing less rounds as governors they end up with less public money in their
22
bank accounts. In that case dictatorship is more beneficial in eliminating embezzlement, which is again, counterintuitive to the thought of many people when they think about forms of ruling. To better understand this idea, think of a governor that plays the game under T2. If she does not take any tokens at all from the public account, the final payoff she faces is
8𝑖𝑛𝑖𝑡𝑖𝑎𝑙 𝑒𝑛𝑑𝑜𝑤𝑚𝑒𝑛𝑡∗ 8𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑟𝑜𝑢𝑛𝑑𝑠= 64.
Since I assume participants are rational individuals, their goal will be to collect as many tokens as possible during the game. The only way for a governor to increase her payoff in T2 is to take money from the public treasury, but she will wish to do so in a way that will still encourage the public to invest in the following rounds.
Recalling previous notations and using specific numbers for the changeable parameters yields the following payoffs for a single round:
𝑃𝑎𝑦𝑜𝑓𝑓 𝑓𝑜𝑟 𝑃𝑢𝑏𝑙𝑖𝑐 𝑝𝑙𝑎𝑦𝑒𝑟𝑖 = 10 − 𝑃𝑢𝑏𝑖 + (∑ 𝑝𝑢𝑏𝑖∗ 4 3 𝑖=1 − 𝐺𝑜𝑣) ∗1 3 𝑃𝑎𝑦𝑜𝑓𝑓 𝑜𝑓 𝐺𝑜𝑣𝑒𝑛𝑜𝑟 𝑝𝑙𝑎𝑦𝑒𝑟 = 8 + 𝐺𝑜𝑣
Also recalling the assumptions regarding the strategic behavior of the governor, we can describe two goals for the governor:
1. ∑3𝑖=1𝑝𝑢𝑏𝑖,𝑡+1 ≥ ∑3𝑖=1𝑝𝑢𝑏𝑖,𝑡 where t is the current round played and t+1 is the next
round to be played. In words we can describe that goal as having the public account at least as big in the next round as it is in the current round. If this condition does not hold it shows of a decreasing willingness to invest of the public in the public account, a process that if continued will keep on deflating the public account to the point where no tokens are invested there, and therefor will leave the governor with no way to increase her payoff
2. Maximize own payoff
The first goal could be regarded as a mean to achieve the second goal, and in order to promise that the first goal is achieved the governor has to make sure it is beneficial for the public to
23
keep and invest in the public account. This will be achieved by promising that the return on investment of the public players will not drop.
Hypothesis 3:
Higher levels of wealth will be reached in T3 comparing with the T1. Hypothesis 4:
Higher levels of wealth will be reached in T2 comparing with the T1.
The rationale behind hypotheses 3&4 is based upon an assumption that governors will
demonstrate a certain level of saturation, which might be explained by a diminishing marginal utility gained from tokens. Using an example to clarify that assumption think of two cases. In the first there are 120 tokens in the public account and in the second there are 40 tokens. Sending back 33 tokens will promise that the investment in the public account was worthwhile for all public players in both cases, but I tend to think it is more possible to witness such an allocation under the second case. This in turn will result in more tokens gathered by the public, which will drive them to increase, when possible, the amounts invested in the public account.
Statistical model
The data collected used the governor game has participants randomly placed in groups and treatments and then make sequential decision. The first important issue to notice is the fact that the insightful data is lying in the group level, since players may be rotated between roles. Moreover, sequential decisions made within each group are not independent.
A few different models were used along the statistical analysis, but notations and some
variables were included in all of them and therefor are important to discuss. The general model was of the following form
𝑌𝑖,𝑡 = 𝛽0+ 𝛽1𝑇2+ 𝛽2𝑇3+ 𝛽3𝑋2,𝑖,𝑡+ 𝑢𝑖,𝑡+ 𝜀𝑖,𝑡
Where the parameters are i – Entity (group) indicator t – Time (round) indicator
24 0.68 0.62 0.52 0 0.2 0.4 0.6 0.8 T1 T2 T3 Treatment
Figure 1 -Average Percentage
Taken
0.39 0.63 0.68 0 0.2 0.4 0.6 0.8 T1 T2 T3 TreatmentFigure 2 - Average Wealth
𝑌𝑖,𝑡 – Dependent variable in group i at round t
𝑇2 𝑎𝑛𝑑 𝑇3 – Binary variable that indicates the treatment the group was assigned to. time
index is omitted since it does not change across time o 𝑇2=1 if treatment is 2, and 𝑇2 = 0 otherwise
o 𝑇3=1 if treatment is 3, and 𝑇3 = 0 otherwise
𝑋2,𝑖,𝑡 – Independent variable in group I at round t. 𝑢𝑖,𝑡 - Between-entity error
𝜀𝑖,𝑡 - Within-entity error
For all regression models data was set as panel data with groups being the entities and rounds the time variables.
Results
To give the reader a better grasp at the data some descriptive statistics are presented. In Table 1 you can find the average percentage of tokens taken from the public account between the different treatments through all groups and rounds. Table 2 shows the average percentage of wealth reached between treatments across all groups and rounds.
Personal details were collected from participants, but were omitted from this discussion because they were found to be irrelevant to the results.
Another important aspect of the data collected lays in the dynamics of the relevant variables through rounds. Augmented Dickey-Fuller test was applied on each of the groups in order to test for the existence of a unit-root. This was done in order to determine whether previous
25
observations of the relevant variables (percentage taken and wealth) have influence over rounds. Percentage taken was found to be stationary in one group in T1. Wealth was found to be stationary in 3 groups (1 in each of the treatments). The fact that a unit root was found in all T2 groups allows me to conclude that participants in that treatment did not follow the moral balance model offered by Nisan (1991). Figure 3 shows the percentage taken in every round at each group in T2. If a participant was to follow a moral balance model it would have been expected that he will maintain the same average level of percentage taken over time, and a rise in percentage taken in some rounds would have been followed by a reduction on percentage taken in the preceding rounds. The fact that in T2 the same participant acted as governor across all rounds allowed me to monitor one’s decisions in eight different points in time, while under T2 and T3 the sample of decision made by the same person was much smaller. While there is no evidence of stationary in T2 it is important to take into consideration that even eight
observations are a small sample when it comes to time series analysis. Moreover, it is more than possible that even if Nisan’s model is correct, it might be the case that participant did not only consider their in game decisions while trying to balance their moral acts, but also took into consideration other actions took previously that day or even prior to that day.
0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1 2 3 4 5 6 7 8 Per ce n tage Take n Round
Figure 3 - Percentage taken over rounds in T2
groups
1 2 3 4 5 626
Findings 1 & 2 – Percentage of tokens taken from the public account is not
significantly lower in T2, but is significantly lower in T3.
The results (shown in table 1) allow me to reject hypotheses 2 and conclude from the data that the percentage taken from the public account is not affected by staff rotation (or the lack of it). Together with that the data allows me to approve hypothesis 1 and conclude that the
percentage taken from the public account is lower in the presence of transparency.
TABLE 1- REGRESSION RESULTS USED TO ASSES HYPOTHESES 1&2
Some explanation about the variables used in the regression model presented above.
Wealth – the wealth level reached in a given round. The wealth is the total amount of tokens earned by all group members divided by 120 (the maximum available)
First_T2 and First_T3 – Variables capturing the interaction between the amount taken in the first round in a given group and T2 and T3 respectively
o 𝐹𝑖𝑟𝑠𝑡_𝑇2 = 𝐹𝑖𝑟𝑠𝑡_𝑇𝑎𝑘𝑒𝑛 ∗ 𝑇2 o 𝐹𝑖𝑟𝑠𝑡𝑇3 = 𝐹𝑖𝑟𝑠𝑡𝑇𝑎𝑘𝑒𝑛∗ 𝑇3
𝑇2 = 1 𝑖𝑓 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 = 2 𝑎𝑛𝑑 𝑇2 = 0 𝑖𝑓 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 ≠ 2 𝑇3 = 1 𝑖𝑓 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 = 3 𝑎𝑛𝑑 𝑇3 = 0 𝑖𝑓 𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 ≠ 3
A few more findings are worth mentioning. First, Wealth is negatively and significantly (at the 5% level) affect the percentage taken from the public account. The meaning is that as wealth is growing bigger, the percentage taken out of that wealth is reduced.
Another finding that comes to attention is the significance (at the 5% level) and positive effect the amount first taken in T3 has on the percentage taken.
_cons .7927365 .0626577 12.65 0.000 .6699297 .9155433 First_T3 .5270289 .1746022 3.02 0.003 .1848149 .8692429 First_T2 .1822732 .1585676 1.15 0.250 -.1285136 .4930599 T3 -.3348589 .1036873 -3.23 0.001 -.5380824 -.1316355 T2 -.1180696 .1203531 -0.98 0.327 -.3539573 .1178181 Round .0007997 .0068033 0.12 0.906 -.0125346 .0141339 Wealth -.2731159 .0953667 -2.86 0.004 -.4600311 -.0862007 Percentage~n Coef. Std. Err. z P>|z| [95% Conf. Interval]
27
Discussion findings 1&2
The first and immediate result is that transparency significantly reduces the amount taken from a public account in the Governor Game. This finding is consistent with other studies showing that transparency is efficient in reducing others forms of unethical and self-serving behaviors (Dana, Weber & Kuang, 2007). The finding that wealth and the percentage taken are negatively correlated could be explained by some sort of diminishing marginal utility gained from taken tokens. This could also be explained by Fehr and Schimdt’s (1999) theory of fairness, which claims that people prefer to not be too advantageous compared with others. The third finding, where the amount first taken in T3 is positively correlated with the percentage taken, is the one I find the most interesting. The main difference between T2 and T3 is the amount of
information given to the players in the group, meaning that participants in T3 were actually learning about the acts of the governor, while in both the other treatments the acts of the governor remained unknown (A low return from the public account could be attributed to either the acts of the governor or to the acts of the other group members). The meaning is that in T3 the first participant to play as governor sets the norm within the group regarding what might be allowed to take from the public account. The fact that percentage taken is positively correlated with First_T3 stress out the importance of the Norm induced to the group. It seems that participants in T3 have learned the accepted amount to be taken and then allowed
themselves to increase the amount a bit more. Note that round does not significantly affect the amount taken, which means that it is not a gradual process.
Findings 3 & 4 – Wealth levels reached in T1 are significantly lower compared with
T2 and T3
The same statistical model was used in order to assess the effect the independent variables had on the dependent variable Wealth. The results, presented in table 2 suggest that both T2 and T3 increase the level of wealth reached. This finding is significant in the one percent level.
28
Another finding that could be derived from looking at the output table is that the level of wealth reached is diminishing in rounds.
TABLE 2- REGRESSION RESULTS USED TO ASSES HYPOTHESES 3&4
All variables used in the described model turned out significant at the 5% level besides the percentage first taken in treatment 3 and percentage taken, which turned out to be significant in the 10% level.
Discussion findings 3&4
As expected, strategic behavior from governors in T2 and the lower percentage taken thanks to transparency in T3 have allowed significant higher levels of wealth in T2 and T3. Together with that, wealth is negatively correlated with the amounts taken in the first round of both
treatments. This may be explained by disappointment from the return on investment and loss of trust in both treatments, while the lack of trust may be pointed towards either the other public players or the governor. This point is made even more important since wealth is decreasing over rounds, showing that the disappointment and lack of trust is gradually increasing.
Conclusion
In a world heavily influenced by corruption it is important to fully understand the situations that cause people to take part in corrupt acts and the effect those corrupt acts have on public trust. _cons .5665762 .0696072 8.14 0.000 .4301486 .7030039 First_T3 -.3411213 .1967961 -1.73 0.083 -.7268345 .0445919 First_T2 -.4430969 .1714894 -2.58 0.010 -.77921 -.1069838 T3 .4292906 .109886 3.91 0.000 .213918 .6446631 T2 .5190212 .122716 4.23 0.000 .2785022 .7595402 Round -.0147107 .0057726 -2.55 0.011 -.0260249 -.0033966 Percentage_~n -.1475465 .0806635 -1.83 0.067 -.305644 .010551 Wealth Coef. Std. Err. z P>|z| [95% Conf. Interval]
29
The biggest contribution of this paper to the literature is the introduction of the “Governor Game” that allows to examine embezzlement in laboratory conditions. Since most findings of this paper are in line with the hypothesis presented a priori, there is sense in applying this experimental procedure using monetary incentives in attempt to put strategies designated to reduce corruption to the test.
The data collected has two main shortcomings. The first is the relatively small pool size that totaled in 64 participants that were divided into 4 groups in T1 and 6 groups in each of T2 and T3. The fact that participants have repeated the game for 8 rounds has allowed for a bigger data set than would have created using a one shot game and have increased the statistical power of this study. While useful in some sense, the repeated trails also have a tendency to repress some type of motives, best put by Loewenstein (1999):” How many times can a subject get angry about someone splitting a pie unevenly?”
The second shortcoming is the fact that participants were not paid at all. In order to conduct a more realistic and internal valid study a performance based incentivizing scheme needs to be used. This is even more important in light of the repeated trail method that was used.
One example of a future study would be to manipulate the amount of tokens endowed to governors in attempt to learn more about the efficiency of increasing public servant’s salary in reducing corruption. Another example could try and impose norms on governors, by informing them about the amounts taken by their preceding counterparts within the same group or in different groups in the same session.
The fight against corruption is growing stronger and one important contribution is that of non-governmental organizations trying to raise awareness to the magnitude of the problem and change the norms regarding corruption. Websites such as Transparencyinternational.org and IPaidaBribe.com are allowing people to shed light on corruption and expose the wrong doings caused by it without bearing the costs whistle blowers used to face. I hope that studies such as this one, reaffirming the importance of transparency and norm structuring, will help people understand the importance of those organization’s work.
30
References
Abbink, K. (2004). Staff rotation as an anti-corruption policy: an experimental study. European Journal of Political Economy, 20(4), 887-906.
Abbink, K. (1999). Staff rotation: A powerful weapon against corruption?. Rheinische Friedrich-Wilhelms-Universität Bonn
Ades, A., & Di Tella, R. (1999). Rents, competition, and corruption. The american economic review, 89(4), 982-993.
Bardhan, P. (1997). Corruption and development: a review of issues. Journal of economic literature, 35(3), 1320-1346
Becker, G. S., & Stigler, G. J. (1974). Law enforcement, malfeasance, and compensation of enforcers. The Journal of Legal Studies, 3(1), 1-18.
Becker, G.S. (1968). Crime and punishment: An economic approach. Journal of Political Economy, 76: 169-217.
Becker, G. S., & Stigler, G. J. (1974). Law enforcement, malfeasance, and compensation of enforcers. The Journal of Legal Studies, 3(1), 1-18.
Braman, S. (2006). Change of state: Information, policy, and power. Cambridge, MA:Massachusetts Institute of Technology Press.
Burns, J., & Huband, M. (1997). How Mobutu built up his $4 bn fortune. The Financial Times, 12.
Cornes, R., & Sandler, T. (1996). The theory of externalities, public goods, and club goods. Cambridge University Press.
Cressey, D. R. (1953). Other people's money; a study of the social psychology of embezzlement. Dana, J., Weber, R. A., & Kuang, J. X. (2007). Exploiting moral wiggle room: experiments
demonstrating an illusory preference for fairness. Economic Theory, 33(1), 67-80.
Fantaye, D. K. (2004). Fighting corruption and embezzlement in third world countries. The Journal of criminal law, 68(2), 170-176.
31
Fehr, E., & Schmidt, K. M. (1999). A theory of fairness, competition, and cooperation. Quarterly journal of Economics, 817-868.
Friedrich, Carl Joachim. Man and his government: An empirical theory of politics. McGraw-Hill, 1963.
Goel, R. K., & Rich, D. P. (1989). On the economic incentives for taking bribes. Public choice, 61(3), 269-275.
Johnston, M. (1998). Fighting systemic corruption: Social foundations for institutional reform. The European Journal of Development Research, 10(1), 85-104.
Kanagaretnam, K., Mestelman, S., Nainar, S. K., & Shehata, M. (2010). Trust and reciprocity with transparency and repeated interactions. Journal of Business Research, 63(3), 241-247.
Kaufmann, D., & Bellver, A. (2005). Transparenting transparency: Initial empirics and policy applications.
Kolstad, I., & Wiig, A. (2009). Is transparency the key to reducing corruption in resource-rich countries?. World Development, 37(3), 521-532.
Lambsdorff, J. G. (1999). The Transparency International corruption perceptions index 1999: Framework document. Transparency International, Berlin. www. transparency. de [13.12. 00]. Leff, N. H. (1964). Economic development through bureaucratic corruption. American
behavioral scientist, 8(3), 8-14.
Lien, D. H. D. (1990). Corruption and allocation efficiency. Journal of Development Economics, 33(1), 153-164.
Lieuwen, Edwin (1960). Arms and Politics in Latin America. P.149
Loewenstein, G. (1999). Experimental economics from the vantage‐point of behavioural economics. The Economic Journal, 109(453), 25-34.
Lui, F. T. (1985). An equilibrium queuing model of bribery. Journal of political economy, 93(4), 760-781.
32
Morrow, J. D., Smith, A., Bueno de Mesquita, B., & Siverson, R. M. (2001). Political competition and economic growth. Journal of Democracy, 12(1), 58-72.
Myrdal, G. (1968). Asian drama, an inquiry into the poverty of nations. Vol. II. New York: Random House, 1968.
Nye, J. S. (1967). Corruption and political development: A cost-benefit analysis. American
political science review, 61(02), 417-427.
OECD (2002), Foreign Direct Investment for Development- Maximising Benefits, Minimising Costs, OECD Secretariat.
Olson, M. (1993). Dictatorship, Democracy, and Development. American Political Science Review, 87(03), 567-576.
Przeworski, A., Limongi, F., & Giner, S. (1995). Political regimes and economic growth (pp. 3-27). Palgrave Macmillan UK.
Relly, J. E., & Sabharwal, M. (2009). Perceptions of transparency of government policymaking: A cross-national study. Government Information Quarterly, 26(1), 148-157.
Saha, S., Gounder, R., & Su, J. J. (2009). The interaction effect of economic freedom and democracy on corruption: A panel cross-country analysis. Economics Letters, 105(2), 173-176. Shim, D. C., & Eom, T. H. (2009). Anticorruption effects of information communication and technology (ICT) and social capital. International review of administrative sciences, 75(1), 99-116.
Shleifer, A., & Vishny, R. (1993). Corruption. The Quarterly Journal of Economics, 108(3), 599-617.
Soto, H. D. (1989). The other path. The Invisible Revolution in the Third World, New York. Svensson, J. (2005). Eight questions about corruption. The Journal of Economic
33
Welsh, D. T., Ordóñez, L. D., Snyder, D. G., & Christian, M. S. (2015). The slippery slope: How small ethical transgressions pave the way for larger future transgressions. Journal of Applied Psychology, 100(1), 114.
Appendix 1 – participant instructions
Uniform Instructions
Dear participants,You are about to take part in an experimental setting which includes decision making in a group. The procedure takes about 15 minutes.
As mentioned, the goal of the experiment you are about to participate in is to examine decisions made in a group setting, therefore in the next stage you will be joined with three other participants. Those participants will be your group members throughout the entire game. Within each group there are two different roles to be assigned to the members of the group. Role A – this role will be played by three members of the group at each round. As a Role A player you will be given 10 game tokens at the beginning of the round.
Role B – this role will be played by one member of the group at each round. As a Role B player you will be given 8 game tokens at the beginning of the round.
At the beginning of each round, the Role A players will be asked to divide their tokens between two accounts:
Private account – Each Role A player has a private account. Tokens invested in the private account will keep their value and will be paid to the owner of the account at the end of the round.
Public account – there is only one public account for the group. All tokens invested in the public account by all Role A player will be multiplied by four.
34
After the amounts have been multiplied, the Role B player will decide how much of the tokens will be sent back to the Role A players and how much will he keep to himself. The amount sent back to Role A players will be evenly divided between the three.
For example, if all three Role A players chose to invest 5 tokens in the public account, the public account have 15 tokens in it. This means that after multiplication the public account has 60 tokens in it. If out of the 60 tokens the Role B player decides to keep 6 token to himself, than the rest of the tokens, 54 to be exact, will be evenly divided among the Role A players, resulting in 18 tokens for each of them.
The number of rounds played is random but will be at most 15.
Please note that each game token is worth 1 cent (1 token = 1 cent). We will pay you the value of each token you have collected during the game as a bonus payment, in addition to the participating payment of 20 cents. For example, if you have collected 50 tokens during the game, you will receive 50 cents as a bonus payment, plus twenty cents as the basic
participation payment.
The number of rounds played in a game is random but will be at most 15. You have to complete all rounds of the game in order to receive your payment.
Following you have two questions that are designated to check you understand the game. You have to correctly answer both questions in order to proceed to the game.
Assuming that all type 1 players have invested a total of 10 tokens in the public account. How many tokens will be in the public account after multiplication?
Type a numeric answer here:
Assuming that after Type 2 player have made his decision 45 tokens are left in the public account. How many tokens will be paid to each type 1 player?
35
Treatment specific instructions – Treatment 1
The different roles will be randomly assigned to the group members at the beginning of each round and other members of the group will not learn about your decisions or about the role you play.
Moreover, none of the group members will be informed about the total amounts accumulated in the public account and the amounts sent back to Role A players.
Since there are 4 group members, every player has a 0.25 probability to be chosen to play Role B and 0.75 probability to be chosen to play Role A. the lottery determining the different roles will take place at the beginning of each round.
Assigned Role A
You have be chosen to play Role A for this round. You have received 10 tokens for this round.
How many tokens do you want to invest in the public account?
(PENDING) please wait until all other players have made their decisions.
You have invested X tokens in the private account and 10-X in the public account. You got Y tokens back from the public account.
Total tokens earned this round is X+Y Assigned Role B
You have be chosen to play Role B for this round. You received 8 tokens for this round.
36
The amount of tokens in the public account is X. How many of the tokens do you want to take for yourself?
Total tokens earned for this round is 8 + Gov
Treatment specific instructions – Treatment 2
The different roles will be randomly assigned to the group members at the beginning of the experiment and will remain fixed for the entire duration of it. Other members of the group will not learn about your decisions or about the role you play.
Since there are 4 group members, every player has a 0.25 probability to be chosen to play Role B and 0.75 probability to be chosen to play Role A. the lottery determining the different roles will take place at the beginning of the game and will only happen once.
Assigned Role A
You have be chosen to play Role A. You will be playing this role for the entire duration of the experiment.
You have received 10 tokens for this round.
How many tokens do you want to invest in the public account?
(PENDING) please wait until all other players have made their decisions.
You have invested X tokens in the private account and 10-X in the public account. You got Y tokens back from the public account.
Total tokens earned this round are X+Y Assigned Role B
You have be chosen to play Role B. You will be playing this role for the entire duration of the experiment.
37
(PENDING) please wait until all other players have made their decisions.
The amount of tokens in the public account is X. How many of the tokens do you want to take for yourself?
Total tokens earned for this round is 8 + Gov
Treatment specific instructions – Treatment 3
The different roles will be randomly assigned to the group members at the beginning of each round.
All group members will be fully informed about the total amounts accumulated in the public account and the amounts sent back to Role A players.
Since there are 4 group members, every player has a 0.25 probability to be chosen to play Role B and 0.75 probability to be chosen to play Role A. the lottery determining the different roles will take place at the beginning of each round.
Assigned Role A
You have be chosen to play Role A for this round. You have received 10 tokens for this round.
How many tokens do you want to invest in the public account?
(PENDING) please wait until all other players have made their decisions.
Total amount accumulated in the public account was J, and the Role B player took Gov (number) to himself.
You have invested X tokens in the private account and 10-X in the public account. You got Y tokens back from the public account.
Total tokens earned this round are X+Y. Assigned Role B
38 You have be chosen to play Role B for this round. You received 8 tokens for this round.
(PENDING) please wait until all other players have made their decisions.
The amount of tokens in the public account is X. How many of the tokens do you want to take for yourself?
Total tokens earned for this round is 8 + Gov
Personal Details questions Age:
