Faculty of Economics and Business
Master Thesis
Are some firms more corrupt than others?
A firm-level analysis of corruption in Turkey and Russia (2009)
2
Abstract
This thesis investigates which firm-level determinants characterize whether or not a firm will pay a bribe. Using a dataset for firms from the manufacturing industries of Turkey and Russia provided by the European Bank for Reconstruction and Development and the World Bank the analysis presents mixed results. Contrary to the scarce literature on firm-level innovation and corruption, findings suggest that the introduction of new products and services trigger bribe payments. Despite the overall model insignificance in the case of Russia, we find significant support that a highly competitive environment, in which the firms operate, drives them to pay bribes.
3
Table of Contents
1. Introduction ... 4
2. Literature review ... 6
2.1. What is corruption? ... 6
2.2. Bribery versus lobbying ... 7
2.3. The effects of bribery ... 7
2.4. Firm- versus country-level analysis ... 9
2.4.1. Firm-level studies ... 10
2.4.2. Country-level studies ... 11
2.4.3. Conclusions of the literature review ... 12
2.5. Hypotheses ... 13
2.5.1. Firm characteristics ... 14
2.5.2. Firm environment ... 15
2.5.3. Interaction effects ... 15
3. Research methods ... 16
3.1. Survey and sample ... 17
3.2. Dependent variable ... 18
3.3. Independent and control variables ... 19
3.4. Econometric model ... 21
3.5. Estimation method ... 21
3.6. Evaluation of method assumptions ... 22
4. Empirical results ... 26
4.1. Descriptive statistics ... 26
4.2. Regression results for Turkey ... 28
4.3. Regression results for Russia ... 29
4.4. Robustness tests ... 32
4.5. Additional analysis of firm-level bribery ... 33
5. Conclusions ... 34
5.1. Added value ... 34
5.2. Limitations and future research ... 35
6. References ... 37
4
1. Introduction
It has been acknowledged for centuries that corruption is one of the phenomena which adversely affect the development of countries and their societies. A common definition for corruption is the misuse of public office for private gain (Svensson, 2005). It is a mirror of a country’s socio-economic condition and may arise because of bad policies and institutions put in place in order to collect bribes from individuals or firms trying to ‘bend’ them (Djankov et al., 2003). Corruption also refers to the ‘abuse of a position of trust in order to gain an undue advantage’, whereas bribery is a ‘specific offence which concerns the practice of offering something, usually money, to gain an illicit advantage’12
. Both actions are illegal and constitute a crime. Bribery takes many different shapes and forms – unofficial payments in cash, gifts and treats, theft of funds for public programs, bribes paid to avoid safety and health regulations etc. – and the literature has made a clear point about its consequences: it significantly distorts market conditions (Rose-Ackerman, 1975; Ades and DiTella, 1999); becomes a serious constraint in doing business (Fisman and Svensson, 2007), and has a negative effect on growth in the long run (De Jong et al., 2010).
Despite the apparent negative effects of bribery, research in the field - specifically on a firm-level - is scarce and needs more attention (Svensson, 2005, 2007). Due to the variety of forms it takes, finding a proper way of measurement on a firm-level has become a challenge. This is why a substantial share of the current literature is on a macro-level, based on cross-country studies using data from perception indices, which in most cases is not the best mean (Svensson, 2005; Treisman, 2006). The disadvantages of the indices come from their ordinal form and their subjective manner. Primarily they are created for the uses of the private sector, particularly for foreign investors. Thus, they mainly measure bribery in connection with doing business, excluding any other forms it may take (Svensson, 2005). Therefore a micro-level analysis, like in this thesis, explains the phenomenon of bribery better. It makes an in-depth analysis of the mechanics of the phenomenon, unlike the indices it includes various forms of bribery, and moves beyond the available country-level studies. (De Jong et al., 2010).
1http://www.anticorruption.ie/en/ACJS/Pages/FQ08000018
5 From an economic point of view, bribery is a demand and supply transaction, where the public official represents the demand side, and the firm represents the supply side. The demand side of the phenomena has been extensively investigated by scholars (Chen et al., 2007; Tonoyan et al., 2010; Shleifer and Vishny, 1993; Bliss and Di Tella, 1997). However, the supply side needs more attention, since the firm is the one paying the bribe. Therefore, this thesis will focus on the supply side of corruption taking the firm as the unit of analysis. The aim of this research is to understand which firms bribe and which do not, using empirics and the foundations of anomie theory (Merton, 1964, 1968). One of the added values of this research is the use of firm-level datasets, which include information about direct expenditures on bribery and other forms of corruption such as gifts, presents, treats etc. so, it studies the supply side of the phenomena – the firm - taking a micro-level perspective. Second, this thesis uses interaction effects of different bribery determinants and shows which firms pay bribes and which do not. Lastly, carrying out the research in this form will further reduce the gap between macro- and micro-level evidence on corruption.
The general research question of this study is: Why do some firms bribe and others not? Firm characteristics will be divided to two different sets of determinants to reveal the impact of each one on corruption. To be precise, the sets include firm characteristics and firm context. Previous studies have identified many variables which have impact on the phenomenon, namely firm’s age and ownership, the competition in the market in which the firm operates, infrastructure accessibility, trade regulations etc. However, they have not provided us with any results of their possible interactions. To fill this gap, this thesis examines interaction effects of the aforementioned determinants sets as well.
Taking these arguments into considerations, the specific research questions would be: 1) Which firm characteristics differentiate bribers from non-bribers?
2) Which firm context characteristics differentiate bribers from non-bribers?
6
2. Literature review
2.1. What is corruption?
The extant literature is rich in definitions for corruption. Being difficult to test empirically (Blackburn et al., 2010), corruption is sometimes viewed as an illegal action which damages the well-being of citizens (De Jong et al., 2010) diverting a public function for a social or economic benefit (Agatiello, 2010). The World Bank calls it “the single greatest obstacle to economic and social development. It undermines development by distorting the rule of law and weakening the institutional foundation on which economic growth depends”. As mentioned earlier, another commonly used definition for corruption is the misuse of public office for private gain, when a public official takes advantage of his position and provides a secretive support to an individual or a firm and in exchange he fulfills his interests (Transparency International, 1995; Svennson, 2005). This is triggered by the information asymmetry between the agent and the firms (Li and Ouyang, 2007) or to put it another way, the agent possesses valuable resources required by the firm to operate successfully. An interesting comparison is made by Iwasaki and Suzuki (2007) who call it a necessary evil for the times of the communist era, and now, during times of capitalist market economy, they look upon it as a self-interest act, prevalent more in transition economies than in developed ones. Some even go further and call it “cancer” of society 3
(Bose et al., 2008), due to its long-lasting negative consequences. In the study of De Jong et al. (2010), bribes are considered to be cash payments by firms to officials with the intention to influence their actions. They conclude that corruption allows entrepreneurs to take advantage of available government resources avoiding red tape. Despite the advantages, at the end they obtain diminishing returns due to the bribe paid, which in the long-run reduces the entrepreneurial spirit. One can notice from the literature that corruption indeed has many definitions and may involve firms in their home or foreign countries, as well as local or foreign governments. Though, for the purposes of this thesis, we will focus on corruption as “private firms paying individual officials privately for government goods and services” (Shleifer and Vishny, 1993; Li and Ouyang, 2007).
7 2.2. Bribery versus lobbying.
Another feature of this thesis is that it sets itself apart from lobbying and focuses on the short term payments by individual firms to “get things done”. Often lobbying is understood as just another form of bribery, because their unit of transaction is usually the same – money. However there are apparent differences between the two. One of them is that lobbying is a collective act undertaken by many firms in order to make a long-lasting change in rules or policies, whereas bribery has a shorter life span, involves individual companies and is used only to get around existing rules (Svensson, 2005). Also it is argued that lobbying is more likely to appear in more developed countries with higher income, whereas bribery appears in developing or transition economies with lower income (Harstad and Svensson, 2011). Harstad and Svensson also claim that lobbying is a legal act, bribery is not and a possible change in the rules via lobbying, changes the status quo affecting entire industries, whereas bribery is firm-specific. One of the implications of their study is that big firms tend to lobby, small ones bribe. This is due to a bureaucrat’s tendency to ask for a higher bribe in the future (De Jong et al., 2010). When the firm grows, the bribes increase in volume and at the end lobbying becomes the more reasonable option.
2.3. The effects of bribery.
The next lines will make a short review of bribery’s negative effects. Afterwards, a summary of its positive effects will be provided too. At the end, a conclusion derived from the comparison will be conducted to make the theoretical analysis more clear.
8 2001; Aghion et al., 2004; Bardhan, 1997; Chen et al., 2007; Bose et al., 2008; Wei, 1999, 2009; Robertson and Watson, 2004; Uhlenbruck et al., 2006). In this way bribery retards development and causes a grouping effect dividing countries into fast- and slow-growing countries (Veracierto, 2008). Corruption can also lead to the so-called “prisoner’s dilemma” where if all firms do not cheat they are all better off, however if every firm knows that others will cheat, regardless the conditions, they all use bribes being worse off at the end (Li and Ouyang, 2007). Sometimes it is argued as a tax which do not end up in the government’s public finances in the form of income (Fisman and Svensson, 2005), and this eventually may cause an increase in the government costs (Hodge et al., 2011). It conflicts with public’s benefits, when a firm pays bribes to circumvent health and safety regulations, or public funds are diverted for other purposes (Svensson, 2005). Mean and Sekkat (2005) point out that corruption may actually “sand the wheels” and create administrative delays on purpose, so that more bribes can be extracted. Therefore the total costs are increased and more contortions added. Other undermining effects are ineffective allocation of talent into unproductive activities (Acemogly, 1995) caused by appointing particular individuals into positions which they are not well qualified for; suppressing capital investment (Krussel and Rios-Rull, 1996), due to unfavorable business environment; and loss of economic and social resources (Ng, 2006; Blackburn et al., 2010) due to theft, cheating or resource diversion. A good example is the Philippines which during the last 25 years is anticipated to have lost $48 billion because of corruption, exceeding its foreign debt of $40.6 billion4; in Uganda corrupt activities increase firm operating cost by 8 percent5; few years ago the United Nations reported that political corruption costs governments about $1.6 trillion6 every year.
The opposing view is that corruption can be beneficial, referring to the so called “petty corruption” (Leff, 1964). It is believed to allow firms to run their operations easily in an economy where excessive administration, complicated regulatory barriers and poor bureaucracy distorts processes to go smoothly (Fisman and Svensson, 2005; Kaufmann and Wei, 1999; Vial and Hanoteau, 2010). Reducing the time lost in administrative delays, it speeds up the procedure for obtaining a permit or a license. “Petty corruption” relates to the so-called “grease the wheels”
4
The information is fromPhilippines Government estimate, cited from Reuter Newswire (1997). ‘‘Philippines corruption a ‘Nightmare’ – Ramos’’,11 January. Therefore recent estimates will differ.
9 hypothesis: it relaxes the consequences of audits and inspections, and helps in finding legal ways of reducing regulatory burden (Bardhan, 1997; Meon and Sekkat, 2005). In addition, Harstad and Svensson (2011) state that if government authorities are able to monitor corrupt officials and retain a fraction of the bribes, eventually bribery could turn to be beneficial. They further their arguments by saying that the government can reduce bureaucrats’ wages, because of the bribes collected by the latter, thus getting an indirect benefit from corruption. More evidence also shows that when bureaucracy is kept under control, corruption and high growth are achievable (Hodge et al., 2011). This can be seen in the case of China (Li and Wu, 2007) and post-war South Korea (Kang, 2002). The results provided by Vial and Hanoteau (2010) seem to confirm this line of thought as well. They discovered that bribes have a significant positive effect on plant growth, thus finding support for the “grease” argument. Interestingly, when there are effective institutions, bribes have unfavorable effects for an economy, on the contrary when the institutions are ineffective bribes give rise to efficiencies (Meon and Weill, 2010).
Despite the existing positive effects the purpose of this paper is not to show how beneficial corruption is. On the contrary, the evidence reporting that corruption has bad consequences is robust and has not been analyzed fully (Emerson, 2006). This is why this thesis centralizes its arguments on the negative aspect of the phenomenon in order to shed more light onto it, and tries to eradicate it no matter when and where it exists.
2.4. Firm- versus country-level analysis
10 2.4.1. Firm-level studies
Stepping on previous studies and using features of bargaining theory put forward by Svensson (2003), Chen et al., (2007) make a model that includes both country- and firm-level data in a cross country context. This type of analysis is useful because it takes into account a large number of factors that influence bribery. The authors tested the relation of corruption with several firm-level variables, including current sales as a proxy for Svensson’s ability to pay. Dependence on public infrastructure, possibility of exporting and current sales were found to be statistically significant. Martin et al., (2007) with the help of anomie theory (see section 2.5 for more information), carry out a study combining firm- and country-level determinants using data from 38 countries and 3,769 firms. They find that their firm-level hypotheses are significant and positive. To be precise, firms pay higher bribes the higher is the perceived competition in the market and perceived financial constraints. Additionally, they distinguish an active and a passive type of firm-level bribery. Active is when the firm itself engages in corruption to secure some valuable information and government contracts or block other competitors to enter the market, passive is when the firm is approached by a public official to pay a bribe. However in this paper we will not rely on this division of firm-level bribery.
11 pressure towards bribery among firms when corruption levels are high. Johan and Najar (2010) in a paper, based on a sample of 123 venture capital and private equity funds, state that if a country has better legislation then fixed fees for investment fund managers are lower and performance fees are higher. They prove that the phenomenon of corruption is a more significant variable in determining fees in financial firms, rather than market environment or managerial features. Lastly, Ng (2006) focuses on the role of corruption in financial markets. His results reveal that bribery relates to higher borrowing costs for firms and low stock valuation along with worsening corporate governance. Conditions like these may force firms to search for informal ways for financing (Li and Ferreira, 2011).
2.4.2. Country-level studies
Studies like the ones above provide us with consistent evidence investigating the causes of corruption from a different perspective. However, despite the advantages of a firm-level approach, research on corruption has been dominated by country-level studies due to data availability7, and more needs to be done on a firm-level (Svensson, 2005; Wu, 2005; Chen et al., 2007; Li and Ouyang, 2007; Martin et al., 2007; Vial and Hanoteau, 2010).
One of the reasons for this research gap is that the firm-level data for that kind of research is difficult to get, because firm representatives tend to hide corrupt practices when surveyed. The macro-level approach looks at corruption as a function of several state level factors. Many scholars have unveiled important country-level determinants and consequences of bribery. A pioneering model for country-level determinants of corruption was carried out by Rose-Ackermann (1975). He assumed that corruption issues can be fitted in a principal-agent type of problem. Specifically, the bureaucrats are the “agents” of the society. However, the contract between the two parts is not easily enforced, leaving room for corruption. A different line of national-level studies state the importance of economic and structural policies as well as the role of institutions in determining the level of corruption (Svensson, 2005). They argue that economic development define the quality of the public service received and therefore the level of corruption and vis-à-vis (Bose et al., 2005). Other views state human capital as a determinant for corruption since educated labor force is required for efficient operation of the institutions
12 (Glaeser et al., 2004). In a country-level study by Treisman (2000), he argues that Protestant nations are more likely to monitor actions of bribery carried out by public officials, than Catholic and Muslim. He adds that in countries where common law legal system is in use, the level of corruption is lower. In a more recent study, he found that a decentralized political system leads to increase in corruption levels due to complex jurisdictions and bureaucratic impediments, referring to bribes in cash, gifts, unofficial payments reported by the companies from each respective country (Treisman, 2006). As a result, centralized administration may become the backbone for bribery activities, which is likely to be seen in transition economies as a part of their path-dependency from socialism to capitalism. The stronger path-dependence there is, the bigger corruption problems we have (Iwasaki and Suzuki, 2012). Interestingly, in developing countries where corruption is high, people are likely to start their own business ventures instead of employment in firms, whereas it is the opposite for developed countries where corruption is low (Mitchell and Campbell, 2009). In yet another research, Ali and Isse (2003) posit that unrestricted bureaucracy, rule of law and political legitimacy boost up country-level corruption. To sum up, previous scholars show that a country’s level of corruption is utilized in its cultural, religious and historical past as well as it depends on the economic development in the country, including its government and legal system (Ng, 2006). The findings stated above help us realize the country specific determinants of corruption. They also help us create appropriate policies against bribery, thereby serving their main purpose.
2.4.3. Conclusions of the literature review
Despite achievements the aforementioned country-level studies have several disadvantages. First, they focus only on the demand side of the phenomenon, namely the public official who asks or receives the bribe, as a result neither of these studies provides evidence on the characteristics of the companies which bribe the civil servants nor they say how much they pay. This statement is supported in the study of Cuervo-Cazzura (2006) where the act of bribery has a demand and a supply side, including institutions from public and private sectors. Second, they cannot tell us whether or not the firm gets an added value from paying a bribe. Lastly, even though we know that bribery is firm-specific8, country-level studies cannot tell us the conditions under which the firms will be willing to bribe. Answering such dilemmas is important in order to
13 understand the phenomenon of corruption better. This is why it is crucial to look at “both sides of the medal”, meaning that a firm-level approach is also required, because the firm is the unit that decides to bribe and plays a role in determining the magnitude of corruption (De Jong et al., 2010; Chen et al., 2007). This approach also makes a more concrete analysis using various firm encounters involving corruption. The use of this very technique, will allow us to lower the gap between country- and firm-level studies and see how corruption deviates under different conditions (Svensson, 2005). Only if we examine both sides of the phenomenon, we will be able to understand it better and build efficient policies to fight against it (Bardhan, 2006).
2.5. Hypotheses
To shed light into firm-level bribery and to investigate firm-level characteristics as its drivers, this thesis will rely on one of the most profound organizational and sociological theoretical foundations to define firms’ bribing behavior. In our argumentations the firm is the actor of bribery. For that purpose we use theoretical foundations to explain its behavior, since we take that the firm is not a person and “cannot “behave” without individual people acting on its behalf (Li and Ouyang, 2007).
14 hypotheses below relate to firm characteristics. Particularly, they are firm-level innovation and firm age. Chen et al., (2007) found that the environment in which the firm operates also affects firm’s bribery decision. Thus, hypothesis three concerns the firm environment. In particular, it is based on the competition level. Last but not least, resting again on the theoretical foundations of anomie theory, we state the interaction effects of some of the bribery determinants as our last two hypotheses four and five. On the other hand, these three particular main effects were chosen because of their practical relevance. Nowadays the number small and medium enterprises are increasing. Most of them are newly created ventures, which makes the firm age an interesting variable to detect whether or not those ventures are involved in any bribing activities when they are set up. In addition, in times of fast development when novelty and competitiveness determine whether you survive in the market or not, firms get challenged constantly. They have to stay up-to-date with the latest trends in their field otherwise they could go out of business. In these difficult conditions when competition is fierce and technology is evolving as fast as never before, successful firms have to keep up with the high pace. Those who are unable to do so, could find themselves in a situation when they have to decide to pay a bribe to get things done or face obstacles and lose valuable time and money. Due to such cases we consider competition and innovation levels to be suitable variables to check for incidence of bribery activities.
2.5.1. Firm characteristics
In their study De Jong et al., (2010) state that younger firms will be more prone to pay bribes in order to create the important government relationships they need at the beginning of their business. Those relationships will allow them to reach important resources and information which are vital when you start up a venture. However, this determinant turns out to be a less relevant feature and is insignificant in their study. Despite this fact, a new test of this hypothesis using a different dataset may prove otherwise. Therefore we have our first hypothesis:
Hypothesis 1 (H1): Firm age is negatively related to bribery.
15 namely an innovator, a producer and a corrupt government official he finds an inverse relationship between bribery and the rate of product innovation. Furthermore, when more resources are invested in the development of new products and services, the amount of funds which can potentially go to pay bribes get less, making the firm less attractive to public officials for bribe payments since the firm becomes unable to pay for any. From these studies we extract our second hypothesis.
Hypothesis 2 (H2): Innovation is negatively related to bribery.
2.5.2. Firm environment
Svensson (2005) argues that if one increases the competitive pressure among firms, it would reduce the profits they can pay bribes with. Therefore one can make the logical conclusion that an increase in competition among firms reduces the level of corruption. Emerson (2006) in his study also investigates the effect of corruption on competitiveness. Using cross-country data and taking competition as a determinant of corruption, he finds that it is inversely related to competition, that is the higher the corruption level the less competitive industrial markets the economy has. Martin et al., (2007) find that perceived competitive intensity is significantly and positively related to bribery. They claim that as competitive pressure increases, firms will be more inclined to pay a bribe in order to maintain their position in the market. Similarly, Wu (2008) concludes that a market competition may cause an increase in bribery activities. Therefore testing these controversial results using new data will be of benefit and we have the third hypothesis.
Hypothesis 3 (H3): Competition level is positively related to bribery.
2.5.3. Interaction effects
16 young age, it still lacks the important connections with public or business figures. Also to utilize its new ideas in the market, it needs to solve these issues fast. Conditions like these induce “pressure” affecting the firm, forcing it to pay bribes. In addition to these arguments we use the results of Hansen (1992) and other scholars9 stating that there is an inverse relationship between firm age and its innovation level, meaning that young firms innovate more. Hence, taking the separate negative relations of both firm age and innovation to bribery into consideration, and the existing relation between firm age and innovation themselves, we conclude that the firm-level innovation variable, when added, will moderate the negative relationship between firm age and bribery and make it stronger and vis-à-vis:
Hypothesis 4 (H4): The negative relationship between firm age and bribery is stronger when there is firm-level innovation and weaker when there is no firm-level innovation.
Using the same line of thought, but this time adding the competition level as a moderator, we state that a young firm, that operates in a context of high competition and lacks vital connections and resources, will find itself under considerable amount of “pressure”. Unable to keep up with the other players in the market it will eventually cease work, losing the initially invested capital, unless it starts paying bribes and receives the assets necessary. Therefore, we conclude that the competition level variable, when added, will act as a moderator and make the relationship between bribery and firm age weaker, because we expect the competition variable and bribery to have a positive relationship. So the last hypothesis is:
Hypothesis 5 (H5): The negative relationship between firm age and bribery is weaker when there is high competition and stronger when there is no high competition.
3. Research methods
In this section first I will introduce the sample used to empirically test the hypotheses defined above. Following that I will describe the dependent, independent and control variables. Afterwards I will specify the econometric model and estimation method and make an empirical assessment of whether or not the methodological assumptions are satisfied.
17 3.1. Survey and sample
This analysis will use the data from the European Bank for Reconstruction and Development/World Bank Business Environment and Enterprise Performance Survey (BEEPS)10. It was initially undertaken in 1999 – 2000 and applied in approximately 4000 enterprises in 26 countries from Eastern Europe and Central Asia. The purpose of the survey is to gather feedback from enterprises operating in EBRD countries on the state of the private sector. It helps building a panel of enterprise data, which helps tracking any differences in the business environment over time. It uses many factors that influence the status of the business environment. These factors can be beneficial or not for firms and can play a pivotal role in whether or not a specific country will prosper. The last time the survey was conducted was in 2009, covering 29 countries and approximately 11,800 enterprises, using stratified random sampling11. Industry, establishment size and region stratification levels are used in all countries. The sample is stratified along manufacturing, retail trade and other services, the size stratification is defined by small (5 to 19 full-time employees), medium (20 to 99 full-time employees) and large (more than 99 full-time employees), whereas the region stratification level is defined by the geographical characteristics of each country. BEEPS is answered by business owners and top managers. Formal (registered) companies with 5 or more employees are targeted for face to face interviews. The survey includes manufacturing and service firms from the construction, retail, wholesale, hotels, restaurants, transport, storage, communications, chemicals and IT sectors. This survey provides quantitative evidence on corruption (Svensson, 2005) measuring it by different questions and obtaining feedback from enterprises in the EBRD countries. Moreover the BEEPS database has also been used by other scholars12 to investigate the issues concerning corruption.
For the purposes of this thesis observations from manufacturing industries from Russia and Turkey will be selected with data from 2009. The dataset for manufacturing firms of these two countries include 1463 completed interviews – 860 from Turkey and 603 from Russia. 44% of the establishments in Russia contacted for an interview were not eligible to complete the survey whereas this number for Turkey was 43%. In Russia the number of contacted establishments per completed interview was 6.14, for Turkey this was 5.6. The countries have been selected because
10 The data is freely available at http://www.enterprisesurveys.org/data 11http://www.enterprisesurveys.org/Methodology/
18 of their different social background and position in the Transparency International Corruption Index.13 According to Transparency International for 2010 Russia is in 154th place with a Corruption Perception Index score of only 2.1 (10 being maximum), whereas Turkey has a score of 4.4, taking 56th place. It is better to have a diverse sample that includes firms from a relatively less and a relatively more corrupt country, so that the end results will have a better base for generalization. Additionally Svensson (2005) claims that in countries with different religious backgrounds public officials will be constrained differently with regard to bribery. In Muslim and Catholic countries officials will be less constrained than in Protestant. In order to have complete information on all dependent and independent constructs the data was filled in using variable mean values where necessary, or assuming the spontaneous responses as negative answers. To be precise a spontaneous response refers to anwers such as „I don’t know”, „Refusal to answer”, „Do not apply” etc (see appendix for survey questions). For example in the cases of competition and innovation level variables spontaneous answers such as „I don’t know” or „Do not apply” were considered as negative. Despite these implemented techniques some observations were still dropped out of the sample, thus reducing the final sample size for Turkey to 757 observations and to 453 observations for Russia.
3.2. Dependent variable
The dependent variable used in the model will be corruption. The BEEPS survey measures this with answer to the question: “It is said that establishments are sometimes required to make gifts or informal payments to public officials to “get things done” with regard to customs, taxes, licenses, regulations, services etc. On average what percentage of total annual sales, or estimated total annual value, do establishments like this one pay in informal payments or gifts to public officials for this purpose?” Asking this question in such a way is useful, because the respondent will most likely give a response based on his/her own experience of corruption, providing a good measure of bribery. Furthermore the question asked in this indirect form makes it general and unspecific enough so that the respondent will give an answer. Lastly, this question is directed to all type of firms from the sample countries covering different industries, however for the purposes of this thesis we will take the answers given by firms from manufacturing industries. The respondents could answer either in value or percentage and it was difficult to combine both
19 type of responses into one variable. The solution was creating a dummy for the dependent variable that takes the value of 1 when the firm has paid a bribe (the value or percentage was larger than 0) and 0 when it has not (the value or percentage was 0).The respondents that refused to give an answer or stated that they do not know whether the firm has paid a bribe or not, were dropped out of the sample. 14
3.3. Independent and control variables
Firm age (Hypothesis 1)
The survey respondents are asked at which year the establishment is formally registered. Using this and the year of conducting the survey together, we can determine the first independent variable stated in the first hypothesis – firm age. Subtracting the year in which the company is founded from the year of the survey results in the years that the establishment has been in existence. As mentioned in Section 3.1 issues with some observations where spontaneous responses were given were tackled using the mean value of the newly constructed variable.
Firm-level innovation (Hypothesis 2)
The next independent variable is firm-level innovation. Survey respondents are asked whether in the last three years, their establishment has introduced new products or services. This question is measured in the questionnaire on a cardinal scale taking the value of 1 for a positive answer and 2 for a negative response. However for the purposes of this thesis we will model it as a dummy variable taking the value of 1 for a positive answer and 0 otherwise. Observations consisting of spontaneous responses will again take the value of 0.
Competition level (Hypothesis 3)
To assess the competition level of the environment in which the firm operates, the respondents of the survey are asked whether the practices of formal competitors are an obstacle. In the survey this question is measured by a 5-point Likert-scale ranging from ‘No obstacle’ to ‘Very Severe Obstacle’. For the purposes of this thesis we will transform and measure it by a dummy taking the value of 0 for no obstacle and 1 if it is an obstacle ranging from minor to very
20 severe. We will assume that observations consisting of spontaneous answers do not represent any obstacles hence taking the value of 0.
Interaction terms (Hypothesis 4 and 5)
Lastly we will construct the interaction terms via multiplication of the previously defined independent variables. Specifically they would be firm age interacting with competition level and firm age interacting with firm-level innovation. Before building the interaction terms the firm age variable will be mean centered in order to increase the interpretability of the interactions and overcome possible problems with multicollinearity (Aiken and West, 1991).
Control variables
21 3.4. Econometric model
Based on the information provided in the previous paragraphs and to test the above stated hypotheses the following econometric model will be used in this thesis:
e OWN SIZE AGExCOMPET AGExINN COMPET INN AGE CORR 7 6 5 4 3 2 1 0 Table 1
Variables and measures
Type of Variable Description of Variable
Corr Dependent Dummy for bribe payments; 1 = Bribes paid; 0 = Bribes not paid Age Independent Firm age measured in years
Inn Independent Dummy for firm-level innovation; 1 = Innovation; 0 = Otherwise Compet Independent Dummy for competition level;
1 = Competition is an obstacle; 0 = Competition is not an obstacle AgeXInn Independent Interaction term between Firm Age and Firm-level Innovation AgeXCompet Independent Interaction term between Firm age and Competition level
Size Control Firm size measured by number of full-time employees
Own Control Dummy for ownership;
1 = Sole proprietorship; 0 = Otherwise
3.5. Estimation method
Following some of the pioneer studies in the field of bribery the estimation method which will be employed in this thesis is a Logistic regression.15 16 This model is appropriate to use because the available data can be measured on categorical scale showing whether or not the establishment has ever made informal payments engaging in bribery activities. Its main task is to explain the effects of the dependent variables on the response probability (Wooldridge, 2000), or on the probability that the firm will engage in bribe payments. As mentioned earlier, for this purpose the dependent variable will be transformed into a dummy taking the value of 1 referring that the firm has made bribe payments and 0 referring that the firm has not. The choice of a logit
15 See Svensson(2003), Wu(2008), De Jong et al., (2010).
22 model ensures that the values of the dependent variable stay strictly between zero and one for all of its observations (Wooldridge, 2000). This model also allows us to include both answers17 to the question used to measure the dependent variable, helping us to include more observations in our sample. Furthermore it helps us estimate the bribery activity from the perspective of the firm (Svensson, 2003).
3.6. Evaluation of method assumptions
According to Wooldridge (2000) the logit model can be derived from an underlying latent variable model which satisfies the classical linear model assumptions. Additionally, he states that those assumptions should not be strictly met in the case of logit models. Despite this in the following paragraphs we will test whether or not our datasets meet the crucial assumptions which need to be satisfied in order to have unbiased estimates, meaning that the variables must not contain any problems regarding heteroskedasticity, endogeneity, multicollinearity and normality.
Homoskedasticity
When homoskedasticity is present the variance of the error term is constant and same for all observations (i) (var(ei) = ϭ2
). If this is violated and the error variance for all observations is not identical we conclude that heteroskedasticity is present. It results in biased estimates of the standard errors which lead in turn to bias in the test statistics and thus to incorrect hypotheses tests, p-values and confidence intervals. Heteroskedasticity itself does not lead to biased coefficient estimates but makes the least square estimator to be no longer the best linear estimator (Hill et al., 2009). The problem with heteroskedasticity is that more weight is given to observations with potentially larger error terms. In that case observations furthest away from the true regression line provide us with the least information about the true regression line. In case heteroskedasticity is detected, there are methodological ways to correct for the weight of larger error variances. In the sample for Turkey and Russia heteroskedasticity does not seem to be a problem18.
17
Total annual sales paid as informal payment in percent and in values.
23 As mentioned earlier in Section 3.3 firm size can determine whether or not the firm will pay a bribe. Therefore heteroskedasticity could be an issue since smaller firms are under the influence of more factors during their operations, whereas larger firms are more resilient to such outside effects. This may lead to differences in error variances. Figure A1 and A2 in the appendix plot the residuals, estimating the errors, against the natural logarithm of firm size measured by number of full-time employees per firm. From these graphs we conclude that heteroskedasticity does not seem to be a problem in both samples since there are no obvious trends visible and the error variances are approximately similar. In addition Breusch-Pagan/ Cook-Weisberg tests were carried out (see Tables 2 and 3) which confirmed the results from the graphs.
Thus we can conclude that heteroskedasticity is not an issue. Since the samples for Turkey and Russia cover only one time period, it is not feasible to test for autocorrelation (correlation across time). Even if heteroskedasticity was present, we would have assumed that it does not appear to be a problem.19 Furthermore due to its flexible functional forms, the logit model tend to work well even if heteroskedasticity is present (Wooldridge, 2000).
Endogeneity
If the assumption that the error term is uncorrelated with the independent variables is violated, such that one dependent variable is correlated with an unmeasured variable, then we will overestimate the effect of the independent variables to the dependent and an endogeneity
24 problem will occur. Additionally simple numerical or statistical tests for this assumption are not available therefore it has to be satisfied theoretically20. Two-stage least-squares (2SLS) regression with instrumental variables is one option to test for and overcome the endogeneity problem and generalized least squares (GLS) is another. Theoretically speaking there is little reason to assume for firm age to be correlated with unobserved variables in the error term, because it can be easily observed and estimated. In case endogeneity exists, it is likely to appear in the two indicators for firm-level innovation and competition level since they both depend on the respondent’s perceptual judgment therefore could be influenced by unobserved variables. Using instrumental variables is a common practice if the indicators cannot be observed directly. However in our sample there are no such instrumental variables at hand. Since the common tests for endogeneity rely on instrumental variables, and such variables are not available in the current sample we will make the assumption that the nature of the independent variables is exogenous.
Multicollinearity
The next assumption is that the independent variables are not perfectly correlated such that”the values of the independent are not exact linear functions of other explanatory variables” (Hill et al., 2009). In case this assumption is violated, variables are collinear. That makes difficult to isolate the relationship between variables. The primary concern is that when the multicollinearity degree increases, the estimates of the regression model become unstable and the standard errors of the coefficients can get inflated. To test for multicollinearity we calculate the variance inflation factor (VIF) which represents an index, estimating how much the variance of an estimated regression coefficient is inflated due to collinearity. VIF values indicate that multicollinearity is not a problem, since all variables have values below the cut-off value of 10 suggested by Neter et al. (1985). However the variance inflation factor for the firm age variable was quite high relative to others in both datasets, therefore it was standardized using its mean value. The VIF values after that operation are presented in Tables 4 and 5.
25
Normality
The final test which will be carried out is the test for normality. It assumes that the values of the error term are normally distributed about their mean (Hill et al., 2009). When we plotted the Pearson’s residuals against the normal distribution, we discovered that the residuals are not normally distributed. This issue may be because of the categorical origin of the dependent variable. It takes only two values, which implies that the error term also takes two values, therefore the usual “bell-shaped” curve describing the error distribution does not hold. Furthermore the logit model could posses a non-normal distribution, though it still can be employed in statistical estimations (Wooldridge, 2000). Despite this setback the method employed here has an advantage because of its simplicity, and it has been found to provide good estimates of the marginal effects of changes in independent variables on the choice of probability (Hill et al., 2009). Alternatively this issue of non-normality can be overcome by transforming the dependent variable.
After running these tests we conclude that our model does not contain any problems with heteroskedasticity, endogeneity and multicollinearity. The only potential problem we face is from the test for normality. It is caused by the dichtonomous form of the dependent variable. As mentioned earlier according to Wooldridge (2000) the logit model is still consistent and can be employed even though it has a non-normal distribution. Hence in the next section we move forward and run consecutively the logistic regressions on the data samples for Turkey and Russia.
Table 4
Test for Multicollinearity (Turkey) Variable VIF 1/VIF Firm size(log) 3.00 0.333041 Ownership 1.05 0.949293 Firm age 4.74 0.210878 Innovation 1.76 0.568267 Competition level 2.69 0.371978 Firm age X Innov. 2.07 0.482654 Firm age X Comp. 3.67 0.272514 Mean VIF 2.71
Table 5
26
4. Empirical results
4.1. Descriptive statistics
Tables 6 and 7 below give an overview of the relevant descriptive statistics for Turkey and Russia. The parametric statistics are most informative for the variables regarding firm age and firm size, since both are measured on continuous scales. In order to have a more complete overview we included also the natural logarithm of firm size in the table. The other variables (interaction terms, control variables and binary independents) are also included, however their descriptive statistics are less informative.
Table 6
Descriptive Statistics (Turkey)
Variable Observations Mean Std.Dev. Min Max
Corruption 757 0.977 0.297 0 1 Firm age* 757 9.10 1 -1.55 5.95 Firm age 757 18.04 11.57 0 87 Innovation 757 0.449 0.497 0 1 Competition 757 0.701 0.457 0 1 Firm age X Innovation 757 0.388 8.274 -18.04 41.95 Firm age X Competition 757 0.129 9.872 -17.04 68.95 Firm size 757 175.540 821.726 2 20843 Firm size(log) 757 3.921 1.429 0.693 9.944 Ownership 757 0.071 0.257 0 1 *Standardized variable
27 Nearly 45% of them have introduced new products or services in the past three years and 70% of them argue that the competition from other firms represents an obstacle for their operation. Lastly, approximately 93% of all sample firms are not owned by one individual.
As for the case of Russia we have 453 observations for manufacturing firms in total (Table 7). In that dataset the youngest firms have existed for only 1 year whereas the oldest for 166 years giving an average age of 17 years per firm in the sample. On average manufacturing firms in Russia employed 211 full-time employees, the smallest firm employing only 1 person and the biggest 5075 persons. The Frequency Statistics Table for Russia (see Table 9 in the Appendix) tells us that 30% of the manufacturing firms in Russia have paid bribes, 73% had a new product or service introduced and 83% of them claim that other formal firms present an obstacle for their operation. Lastly, only 1% of the surveyed firms are owned by one individual.
Table 7
Descriptive Statistics (Russia)
Variable Observations Mean Std.Dev. Min Max
Corruption 453 0.291 0.454 0 1 Firm age* 453 -1.17 1 -0.86 8.21 Firm age 453 16.63 18.18 1 166 Innovation 453 0.730 0.444 0 1 Competition 453 0.830 0.376 0 1 Firm age X Innovation 453 0.028 15.989 -13.63 149.36 Firm age X Competition 453 -0.540 15.848 -15.63 149.36 Firm size 453 211.34 459.214 1 5075 Firm size(log) 453 4.294 1.405 0 8.532 Ownership 453 0.011 0.104 0 1 *Standardized variable
28 4.2. Regression results for Turkey
Table 12 below provides the logistic regressions for Turkey. The first model excludes the
29 case we have only four iterations leading us to believe that we do not have multicollinearity problems in the model.
In sum, our regression results for Turkey do not offer significant support for our hypotheses. However, we found a positive and significant relationship between firm-level innovation and bribe payments among manufacturing firms, or when a firm-level innovation among manufacturing firms is present they are likely to engage in bribe payments.
TABLE 12
Logit Model Regression Results for Manufacturing firms from Turkey. Dependent Variable is Corruption. (N=757)
Variables Model 1 (Controls) N=757 Model 2 (Main effects) N=757 Model 3 (Interaction Terms) N=757 Firm size(log) -.093(.0899) -.095(.0932) -.096(.0939) Ownership -1.204(.739) -1.078(.742) -1.049(.743) Firm age -.052(.128) -.307(.334) Innovation .758(.253)*** .779(.258)*** Competition .105(.277) .1193(.280)
Innovation X Firm age .0179(.024)
Competition X Firm age .0127(.025)
Constant -1.804(.369)*** -2.278(.460)*** -1.833(.025)***
Pseudo R2 0.0086 0.0282 0.0299
LR chi2 4.15 13.68 14.48
Prob > Chi2 0.1254 0.0178** 0.0432**
Note. Standard Errors are in parentheses. ***p<0.01, **p<0.05, *p<0.1
4.3. Regression results for Russia
30 fit of the data to the model is insignificant. In what follows, I present and discuss the findings as an exploratory study bearing the aforementioned concern in mind.
The first model again includes only the control variables for ownership type and the natural logarithm for firm size. Both controls and the overall model turn to be insignificant in this case (p>0.1). Adding the main effects into the model increases the Likelihood Ratio and the Pseudo R-squared of the model. Though the overall model still remains insignificant (p>0.1). Despite this insignificance we carry on with the interpretation of the signs and coefficients of all variables. Model 2 in the table shows that firm age has indeed a negative relationship with bribe payments as stated in our first hypothesis (β=-0.001). Same as in the case with Turkey though, this result is insignificant (p>0.1) leading us to reject our first hypothesis. Next, we find that there is a positive relation between firm-level innovation and bribery (β=0.158), among manufacturing firms in Russia, as it is in Turkey. This result contrasts our second hypothesis, which stated for a negative relationship. Despite this finding we are forced to reject our second hypothesis as well due to insignificance of the variable (p>0.1). Our third hypothesis had stated that competition level is positively related to bribery. Indeed our results confirm that statement showing a positive significant relationship (p<0.1). To be precise, when the competition from other formal competitors represents an obstacle for the firm it increases the probability of a bribe payment to occur by 0.60. This transformed into odds ratio means that bribe payment will occur 1.5 to 1 when competition is an obstacle. Thus we accept our third hypothesis.
31 the probability of the firm to pay a bribe by 0.59, which transformed to odds ratio is again 1.5 to 1 that a bribe will be paid when competition is an obstacle.
To sum up, despite their insignificance, our logit regression results for Russia’s sample provided us with interesting results. 21 In model 2 and 3 we confirmed our third hypothesis that there is a positive relationship between bribe payments and competition. Other than that our findings were in line with our fourth and fifth hypotheses, though insignificant. Lastly, as in the case with Turkey, we found a positive relation between firm-level innovation and bribe payments. Whereas our theoretical predictions for the first hypothesis were confirmed in Model 2, but rejected in Model 3.
TABLE 13
Logit Model Regression Results for Manufacturing firms from Russia. Dependent Variable is Corruption. (N=453)
Variables Model 1 (Controls) N=453 Model 2 (Main effects) N=453 Model 3 (Interaction Terms) N=453 Firm size(log) -.029(.0744) -.023(.082) -.024(.0821) Ownership .438(.927) .41(.930) .50(.934) Firm age -.001(.993) .16(.343) Innovation .158(.244) .165(.245) Competition .60(.310)* .586(.311)*
Innovation X Firm age -.0165(.013)
Competition X Firm age .004(.017)
Constant -0.766(.335)** -1.422(.475)*** -1.413(.475)***
Pseudo R2 0.0008 0.0093 0.0122
LR chi2 0.43 5.11 6.69
Prob > Chi2 0.8053 0.4027 0.4624
Note. Standard Errors are in parentheses. ***p<0.01, **p<0.05, *p<0.1
32 4.4. Robustness tests
To ensure that the inferences of the model are valid, robustness is a required characteristic. For example it is often used to examine how the regression coefficients might behave when the model specification is modified. Evidence that estimates are robust is when the coefficient estimates do not change their signs and magnitude significantly after such modifications. We speak of structural validity when we have plausible and robust coefficients (White and Lu, 2010).
We performed the following tests of robustness. First, we ran the model using a probit method and implementing it in Turkey’s sample. Second, we replaced the sample with data for Russia and ran the model again using probit method. The results are presented in Table 14 for Turkey and Table 15 for Russia in the Appendix. As we can see the outcome is not significantly different from what we had earlier for Turkey. The models with the main effects and interaction terms are again significant at the 0.05 level. Interestingly, this time the control variable for ownership in Model 1 is significant at the 0.1 level, however the overall model turns out to be insignificant (p>0.1). Firm age again proves to be negatively related to bribe payments (0.032 and β=-0.162) however it is insignificant in both models. As for the competition level variable, its positive relationship to bribery is again present (β=0.063 and β=0.073), but insignificant (p>0.1). In Models 2 and 3 we find again a significant positive relationship between firm-level innovation and bribe payments (p<0.01 and p<0.05; β=0.399), thus rejecting our second hypothesis. Lastly including the interaction terms in our last round does not give us any unexpected results for Turkey. Both interaction effects are insignificant (p>0.1). The interaction of firm age and innovation contradict to our fourth hypothesis (β=0.009), whereas competition level interacting with firm age support our theoretical predictions stated in our fifth hypothesis (β=0.007).
Firm-33 level innovation when interacting with firm age strengthens its negative relationship with bribe payments (β=-0.010) and competition level when interacting with firm age weakens it, causing it to turn positive (β=0.0031).
Implementing our specified model using a different method gave us similar results. From here we argue that our findings are robust and further continue to the next section where we will discuss the insights stemming from the results and make a note about some of the problems which appeared during the analysis.
4.5. Additional analysis of firm-level bribery
The results provided by the Logit regressions are rather interesting. First, for the sample of Turkey most results are in line with our hypotheses, though none was significant except one result. We discovered a positive relationship between introduction of new products and services and bribe payments among manufacturing firms. We received the same result for Russia’s sample, though it was insignificant. This was opposite to our predictions. Nevertheless this case is also possible since when a firm tries to deliver a new product to the market it needs to go through all the official bureaucratic procedures in order to gain the necessary permits, licenses, registrations etc. which are very likely to be under the control of public officials. A support for this finding is the study of Blackburn and Forgues-Puccio (2009). They find that when corruption is present, depending on the circumstances, particularly when it is organized, it lowers the costs of R&D. A lower cost encourages more firms to undertake research activity and hence introduce new products and services. They further claim that the relationship between corruption and innovation has yet to be analyzed by systematic empirical investigations and the corruption phenomenon with its different forms and shapes can have “as much to do with different cultural, social and political features as it have with economic circumstances”. Their analysis is not focused on the incidence of bribes but rather on the structure of corruption, whether it is organized or not, taking that corruption is present in both cases. So in short, under specific circumstances corruption may indeed lead to more innovation. This conclusion is quite general and further investigation is necessary as suggested by Blackburn and Forgues-Puccio (2009), which is not the topic of this thesis.
34 relationship between competition level and bribe payments. We assume that those results were caused by the smaller sample size of Russia (N=453) compared to Turkey (N=757). According to Hosmer and Lemeshow (2000) large sample sizes are required for logistic regressions to give enough numbers in both categories of the dependent variable. They argue that the more explanatory variables there are, the larger should be the sample, advising of samples greater than 400. Peduzzi et al., (1996) also suggest that the sample size in logistic regressions could be difficult problem. Although our sample for Russia is slightly above the 400 cut-off line we assume that this is the issue causing the insignificance of the overall model.
5. Conclusions
5.1. Added value
The purpose of this thesis was to make a contribution to the firm-level literature investigating which firm-level characteristics determine bribery activities. The two research questions related to the circumstances under which firms engage in bribe payments have been answered by reviewing the literature. Based on the theory, we have stated five hypotheses regarding those specific circumstances and how they affect whether or not the firm will engage in bribery. Particularly they were firm age, firm-level innovation, competition level and their interactions. The specified hypotheses were tested using two sample countries – Turkey and Russia – with a total sample size of 1210 observations including only manufacturing firms.
35 providing other services by the public officials. This way bribery can easily be detected and required actions can be taken.
When we tested our hypothesis using Turkey’s sample we also had an interesting finding. We found a significant positive relationship between firm-level innovation and bribe payments. We argue that the introduction of new products or services could cause payments of bribes. This finding answers our first research question and shows that firm-level innovation is one of the factors characterizing bribing firms. We conclude that this is indeed an important finding especially considering the fact that the research on firm-level innovation and corruption is scarce. Furthermore the industries which innovate are the ones generating high added value for an economy. They create an environment which fosters research and development, sustainable growth and long-term investments. Such characteristics are always beneficial for economies and should be promoted, therefore our finding that in innovative sectors bribery could easily occur helps the respective authorities to re-focus their resources and look for ways to prevent that ‚evil’ from happening.
This thesis provides added value to the extant literature concerning the field of corruption. It investigates the appearance of bribe payments from the perspective of the firm, thus making a contribution to the expanding firm-level literature on the topic. It builds on a sample of manufacturing firms from Turkey and Russia, providing quantitative empirical support, contrasting to the qualitative case studies. It confirms some of the results of previously established research in this field, but most importantly it discovers a new interesting link between innovation and corruption activities – a track of research which is yet to be developed.
5.2. Limitations and future research
36 may bring more detailed information on the magnitude of the bribes paid by the firms. An interesting comparison would be if one takes countries opposing each other in terms of perception for being more or less corrupt and employs the current or newly build model. Lastly new firm-level characteristics which are found to be related to bribe payments can be included in the model and tested.
37
6. References
Acemoglu, D. (1995). “Reward structures and the allocation of talent”. European Economic
Review, Vol.39, No.1, 17–33.
Ades, A. and DiTella, R. (1999). “Rents, Competition and Corruption”. American Economic
Review, Vol.89, 982-993.
Agatiello, O. (2010). “Corruption not an end”. Management Decision, Vol.48, No.10, 1456-1468.
Aghion, P., Alesina, A., and Trebbi, F. (2004). “Endogenous political institutions”. Quarterly
Journal of Economics, Vol.119, No.2, 565–611.
Aiken, L. and West, S. (1991). “Multiple Regression: Testing and Interpreting Interactions”.
Sage
Ali, A. M., and Isse, H. S. (2003). “Determinants of economic corruption: A cross-country
comparison”. Cato Journal, Vol.22, 449–466.
Bardhan, P. (1997). “Corruption and development. A review of issues”. Journal of Economic
Literature, Vol.35, 1320–1346.
Bardhan, P. (2006). “The economist’s approach to the problem of corruption”. World
Development, Vol.34, No.2, 341-348.
Bennedsen, Morten, Sven E. Feldmann, and Lassen,D.D. (2009). “Strong Firms Lobby, Weak
Firms Bribe: A Survey-Based Analysis of the Demand for Influence and Corruption”. Mimeo, Melbourne Business School.
Blackburn, K. and Forgues-Puccio,G.F. (2009). „Why is corruption less harmful in some
countries than in others?”. Journal of Economic Behaviour & Organization, 797-810.
Blackburn, K., Bose, N., Haque, M. (2010). “Endogenous corruption in economic
development”. Journal of Economic Studies, Vol.37, No.1, 4-25.
Bliss, C. and DiTella, R. (1997). „Does competition kill corruption”. The Journal of Political
Economy, Vol.105, 1001-1023.
Bose,N., Capasso,S., Murshid, A.P. (2008). “Threshold Effects of Corruption: Theory and
Evidence”. World Development, Vol.36, No.7, 1173-1191.
Campos, N. and Giovannoni, F. (2007). “Lobbying, Corruption and Political Influence”.
38 Campos, N. and Giovannoni, F. (2008). “Lobbying, Corruption and Other Banes”. William
Davidson Institute, WP 930.
Chen, Y., Yasar, M. and Rejesus, R. (2007). „Factors influencing the incidence of bribery
payout by firms: a Cross-country analysis”. Journal of Business Ethics, Vol.77, 231-244.
Cuervo-Cazurra, A. (2006). “Who cares about corruption?”. Journal of International Business
Studies, Vol.37, 807–822.
Cullen, J. B., Parboteeah, K. P., and Hoegl, M. (2004). “Crossnational differences in
managers’ willingness to justify ethically suspect behaviors: A test of institutional anomie theory”. Academy of Management Journal, Vol.47, 411–421.
De Jong, G., Tu, P.A., and van Ees, H. (2010). “Which Entrepreneurs Bribe and What Do
they get from it? Exploratory Evidence from Vietnam”. Entrepreneurship Theory and Practice, 1-23.
Djankov, S., La Porta,R., Lopez-de-Silanes,F. and Shleifer,A. (2003). “The Regulation of
Entry”. Quarterly Journal of Economics, Vol. 227, 1–37.
Durkheim, E. (1897/1966). “Suicide: A study in sociology”. New York: Free Press.
Emerson, P. (2006). “Corruption, competition and democracy”. Journal of Development
Economics, Vol.81, 193-212.
Estache, A., Goicoechea, A. and Trujillo, L. (2006). “Utilities reform and corruption in
developing countries.”. Mimeo, World Bank, Washington DC.
Fama, E.F. and Jenssen, M.C. (1983). „Separation of ownership and control”. Journal of Law
and Economics, Vol.26, 301–325.
Fisman, R. and Svensson, J (2007). “Are corruption and taxation really harmful to growth?
Firm level evidence”. Journal of Development Economics, Vol.80, 63-75.
Glaeser, E., La Porta, R., Florencio Lopez-de-Silanes and Shleifer,A. (2004). “Do Institutions
Cause Growth?” Journal of Economic Growth, Vol.9, No.3, 271–303.
Harstad,B., Svensson, J. (2011). “Bribes, Lobbying and Development”. American Political
Science Review, Vol.105, No.1, 46-59.
Hill, R.C., Griffiths, W.E., and Lim, G.C. (2009). „Principles of Econometrics 4th Edition.” John Wiley & Sons Ltd.
Hodge, A., Shankar, S., Prasada Rao, D.S., Duhs, A. (2011). “Exploring the Links Between
