The influence of country-level corruption and firm-level determinants on firm corrupt behaviour

(1)

The influence of country-level corruption

and firm-level determinants on

firm corrupt behaviour

Master Thesis by

(2)

2

Abstract

Corruption is still pervasive throughout the world. Understanding the causes for corruption is vital to properly combat it. A substantial share of research regarding corruption has been done on the country-level; studies on the firm-level seem to be underrepresented. Thus far, there is no complete understanding about the causes or incentives for firms to engage in corruption. The purpose of this thesis is to shed some light on possible determinants for corruptive firm behaviour, thereby enhancing the literature on corruption and a better understanding of the underlying causes. During this study corruption is mainly perceived from the firms’ perspective, unlike most of the prior literature that look at the government side of corruption. Multiple determinants for corrupt firm behaviour are considered, including: country-level corruption, length of applications, gender of the owner, gender of the top manager, and ownership structure. A sample of 15,883 firms from the BEEPS data base is analyzed with the use of a logistic regression. It includes several independent variables and a moderating effect. The results display the relationship between the determinants and firm corrupt behaviour. Findings suggest that country-level corruption is a good indicator for the odds of corrupt firm behaviour. As for firm-level determinants, significant support was found for gender of the owner, length of applications, and ownership structure. No relation was found between the gender of the top manager and corrupt firm behaviour.

Keywords: Corruption, bribery, gender, ownership structure, CPI, BEEPS, length of

(3)

3

3.3.3 Control variables... 23 3.4 Statistical analysis ... 25 3.4.1 Multicollinearity ... 25 4. Results ... 27 4.1 Descriptive statistics ... 27 4.2 Model results ... 28 4.3 Robustness check ... 31 5. Discussion ... 32 5.1 Discussion results ... 32 5.2 Theoretical implications ... 34 5.3 Managerial implications ... 35 5.4 Policy implications ... 36

5.5 Limitations and future research ... 36

6. Conclusion ... 38

Literature ... 39

(4)

4

1. Introduction

In recent events came to light that a large Dutch building company ‘Ballast Nedam’ supposedly bribed king Fahd of Saudi Arabia and his successor king Abdullah. It is estimated by the FIOD (fiscal information and investigation service of the Netherlands) that the bribes exceeded over 300 million dollars. The bribes were paid in exchange for large building projects (van den Eerenbeemt, 2017). This is just an example of many. Corruption is still pervasive throughout the world; it exists in all countries, public sectors and other societies. The World Bank estimated that 5 percent of exports to developed countries go to corrupt officials (Moss, 1997). “One of the most sinister features of bribery is its corrosive effect on the public’s respect for the rule of law, and therefore, on the structure and stability of society” (Mauro, 1995). It is a persistent feature of human societies; even in countries with a strong institutional environment and a low level of corruption such as the Netherlands.

Corruption, understood as the abuse of public for private benefit, has a two-faced view. One view is corruption as greasing the wheel, seen from a positive perspective which argues that it can foster development. This idea was put forward in the 1960s by Huntington (1968), Leff (1964), and Leys (1965). They discussed the beneficial effects of corruption. While on the contrary is the negative view stating that corruption sands the wheels; it harms society and it constrains economic growth and development. The last view gains considerable more support in the literature (e.g. Aidt, 2009; Mauro, 1995; Meon et al., 2005; Murphy et al., 1993). Notwithstanding the wide viewed perspective of the negative effects of corruption; it is still present throughout the world. International organizations are still fighting against corruption. International initiatives emerged such as the OECD’s ‘Convention on combating bribery’ to reduce corruption, but yet it remains persistent. “The apparent incongruity between personal beliefs and the remarkable persistence of corruption contextualizes the ineffectiveness of many firm-level ethics initiatives” (Collins et al., 2009). A better understanding of corruption and why firms engage in it is needed. This leads to the question what possible reasons there are why corruption is still present globally?

(5)

5 opportunities. Reducing these opportunities will in return reduce corruption. Broadman (2001) found that a system of strong market institutions, good enforcement mechanisms and a healthy competitive environment reduces these rent-seeking opportunities. There is no full understanding yet about the causes and determinants for corruption, but there is generally agreed upon by researchers and policy-makers that corruption thrives in environments with institutional deficiencies and non-transparent regulations (World Bank, 1997). A weak institutional environment seems to go hand in hand with corruption.

There is a vast amount of literature about the causes and consequences of corruption. Over the past decade, a sizable amount of literature examining various determinants of corruption has emerged (Bardhan, 1997). A large share of that has been done at country-level (e.g. Ades et al., 1999; Husted, 1999; Lipset & Lenz, 2000; De Rosa, 2010; Treisman, 2000; Wei 2000). Cross-country analysis is a common feature of research investigating the determinants of corruption. But this does not explain the within-country variation of corruption. Firms who operate in the same environment and face the same institutions still behave differently. A macro-level approach is useful, but it provides little information on the effect of more micro-level determinants of corruption (Chen et al., 2008). There is no complete understanding why some firms engage in corruption and others do not. The firm-level research seems to be underexposed, possibly because it can be problematic to gain insight in corruptive practices within firms. Actors involved in corruption have incentives to hide their behaviour, it proves difficult to actually observe and measure corruption (Collins et al., 2009). Some research on corruption has covered the firm-level such as the studies of Lee et al. (2010), Mocan (2004), Swamy et al. (2001), and Svensson (2003). However, most of the firm-level research is viewed from the bribe-takers perspective (the demand or government side). There is little known yet about the perspective of the bribes suppliers (i.e. firms), in particularly a theoretical understanding of bribery (Martin et al., 2007).

(6)

6 payments to government officials. This suggests that an internationally oriented firm seems to be less vulnerable to bribery. Another firm-level characteristic found is state ownership; a higher state ownership reduces the bribe size, possibly because the government is already interfering and controlling the firm so it aligns with their objective. Both Herrera & Rodriguez (2003) and Lee et al. (2010) found bribery differences between industries, were manufacturing firms are less prone to bribe than service firms.

Still, these factors are not able to predict bribery by firms, there remains ambiguity why some firms act in a corrupt manner and others do not. Which factors drive firms to engage in corruption and bribery? We try to extend the literature on possible determinants for firm corrupt behaviour, as extension to previously found factors. During this study we look at the possibility of the length of applications for public services or utilities (for example the application for certain permits) as cause for bribery. Firms could be more inclined to bribe when the wait times for these applications are relatively high, thereby trying to circumvent the bureaucratic hassle and speeding things up. No research has been so far on the length of applications. Another firm characteristic that might be influencing bribery is the gender of the owners and top managers. In previous studies they found that women are less prone to act in a corrupt manner and are less likely to condone taking bribes (Svensson, 2003). Dollar et al. (2001) found similar results and base their findings on the argument that men are more selfish than women. Most of the studies regarding gender and corruption have been on the demand side of bribery; from the public official’s perspective. No research has been done so far on gender and the supply side of bribery. We investigate whether the gender of owners or top managers matters for firm corrupt behaviour. The last firm characteristic we propose to be related to firm corrupt behaviour is ownership structure, in particular the amount of owners. Firms owned by multiple owners might be less likely to bribe since more people are involved and aware of each other’s actions, by bribing they could possibly create hazards for themselves. A single owner might get away with bribing more easily and is less likely to face the negative consequences of it. A more in depth view of these factors will be given during the literature review.

(7)

7 The data for this study comes from two sources. The data on country-level corruption comes from the Corruption Perceptions Index (CPI). It is an aggregate corruption indicator compiled from multiple sources. Data on firm-level determinants is derived from the EBRD-World Bank Business Environment and Enterprise Performance Survey (BEEPS). The BEEPS is a firm-level survey of a representative sample of an economy’s private sector. It contains items involving corruption, including the prevalence of bribery. The survey is based on face-to-face interviews with business owners and top managers. With the use of these data sources we can compare firm-level characteristics across different countries and institutional environments, therefore being able to make broader statements about the findings. The data will be analyzed by computing a logistic regression. The results will show the relationship between the variables and firm corrupt behaviour.

The results of this thesis have theoretical, policy and managerial implications. This thesis aims to contribute to the literature by extending the work on corruption and its determinants, primarily on firm-level determinants as they are underexposed in the literature. Another key thing is that this study will give insight on the supply side of bribery, in contrast to the more prevalent literature on the demand side of bribery. It particularly aims to shed some light on the possible circumstances and firm characteristics where it is more likely for firms to behave corruptive. The knowledge gained from this research can help with a better control of corruptive behaviour such as bribery. This can both be applied by managers and policy or regulation makers. Understanding the possible factors associated with corruption and bribery is crucial when trying to reduce it to a minimum.

(8)

8

2. Literature

2.1 Corruption

The main theme of this research surrounds corruption. A broad definition of corruption is “the abuse of authority by bureaucratic officials who exploit their powers of discretion, delegated to them by the government, to further their own interest by engaging in illegal, or unauthorised, rent-seeking activities” (Blackburn et al, 2005). A much used definition is that of Schleifer & Vishny (1993): “the sale by government officials of government property for personal gain.” During this thesis a simplified version will be used, defining corruption as the abuse of public power for private benefit. Examples of abuse are government officials granting licenses or permits in exchange for bribes. All actions that public officials undertake in exchange for private benefits can be considered as corruption. This can be seen as a principal-agent problem, the government being the principal and the government officials as agent. The agent can pursue different goals than the government, in this case pursuing for their private benefit which in return harms the government. A common form of corruption is bribery, bribes being informal payments to public officials. Government officials have incentives for corruption when they control distribution of something that is desired by the private sector or something that the private sector wants to avoid (Rose-Ackerman, 1999).

There are however different views on the consequences of corruption and bribery as mentioned earlier. The positive ‘grease for the wheels’ view which started with Huntington (1968) and Leff (1964) who suggested that corruption might be good for economic growth. They argue that corrupt practices such as bribery or informal payments would enable individuals to avoid bureaucratic hassle and delay, thereby speeding up the process of development. Furthermore, Lui (1985) said that investors sometimes bribe to reduce their time queuing for something; in this sense bribing can be efficient. The negative ‘sand in the wheels’ view is more supported; hampering economic growth and development (Kaufmann, 1997; Mauro, 1995; Schleifer and Vishny, 1993; Wei, 2000). For instance, it distorts incentives for firms to invest in that specific country, resulting in less FDI which is a crucial factor for growth and development. Corruption is pervasive throughout the world; it is present in both developed and developing parts of the world. This is remarkably, since Noonan (1984) stated that “there is not a country in the world which does not treat bribery as criminal on its law books”.

(9)

9 firm’s probability of encountering corruption in a country. This pervasiveness reflects the expectations of firms being confronted with corruptive practices in a country and the need to approach and comply with these issues. The other type entails the degree of uncertainty of corruption and is described as ‘arbitrariness’. It is associated with the ambiguity of corruption. The unknown risk is about the enduring uncertainty regarding the size and number of bribes necessary to appropriate the initial gift of their exchange. Bribes can be ineffective in inducing the desired effect. Expectations of outcomes are hard to determine in the case of arbitrariness corruption.

Corruption is a complex concept. The exact causes for corruption are hard to determine, since it is dynamic and involves numerous factors. A variety of characteristics of a country play a role in this, such as the economic, political, and social systems (Treisman, 2000). A low risk of getting caught and punished creates incentives for corruption; this depends on the effectiveness of the legal system in a certain country. La Porta et al. (1999) found that a common law system instead of a civil law system generally reduces corruption more. This is hypothesized because a common law system has better protection of property against the state. However, Treisman (2000) found that it is not necessarily the legal system itself but that the influence comes from the ‘legal culture’. Moreover, Serra (2006) argued that the ‘legal culture’ origins in the past, countries with a British heritage seem to have lower levels of corruption, this in return is closely related to the arguments of La Porta et al. (1999) since countries with a British heritage often adopted the common law system.

Husted (1999) found that collectivistic cultures and those with more of an emphasis on

materiality ranked higher in corruption. Lipset & Lenz (2000) showed that countries with social systems where loyalty and obligation are highly valued personal traits result into a higher level of corruption. “Corruption may be endemic and linked to deep-rooted cultural or institutional features of a society, which are not easily overturned by specific policy measures” (De Rosa, 2010). Schleifer & Vishny (1993) and Serra (2006) found that weak governments with unstable political institutions have a difficult time preventing bribery, richer countries have in general less corruption, and countries where citizens are mostly protestant seem to have less corruption. Getz et al. (2001) and Husted (1999) argued that bribery within a country is also related to culture and socioeconomic factors.

(10)

10 country-level corruption, we instead investigate the influence of country-level corruption on the main focus of this study; namely firm-level corruption.

2.2 Firm corrupt behaviour

In the literature they mention different types of bribery and corruption exerted by firms. Bribery can be seen as the act of giving something to a recipient with a public function with the intention of influencing the recipient to the benefit of the giver. In this study it is the firm who pays the bribes to government officials to get certain things accomplished. Wu (2005) distinguishes between two types of bribery by firms: passive and active bribery. Passive bribery is when the firm feels the need to bribe to avoid being punished. Active bribery occurs when firms initiate the bribes themselves in order to benefit from it. During this study no distinction between the types of bribes will be made.

Hellman et al. (2000) discussed three forms of corruption exerted by firms. ‘State capture’ concerns payments to government officials to influence and shape the formulation of the rules of the game. The corruption form ‘influence’ tries to accomplish the same, but this form is without the payments but merely by means of the firm’s influence. The last form mentioned is ‘administrative corruption’, which entails petty forms of bribery in connection with the implementation of laws rules, and regulations. During this study we will focus solely on bribery as firm corruptive behaviour, so the corruption type ‘influence’ is disregarded. Another view on bribery is that of Martin et al. (2007) who try to grasp the concept of bribery with the anomie theory: “cultural and social drivers result in conditions in which pressure for goal achievement through any means – legitimate or not – displace normative control mechanisms.” It upholds that a large enough pressure for goal achievement can result in corruption if it is necessary for achieving that particular goal. They further build upon this theory and propose that national cultural factors and social institutions are explaining bribery by firms. Local firm conditions can create situations where the firms justify the need for bribes.

(11)

11 Corruption is not one sided, both the government officials and the opposing actor play a role in this. In this case we focus solely on firms as the opposing actor. Not only government officials have incentives for corruption. Firms that offer bribes or accept paying a bribe can obtain benefits or advantages that would otherwise not be accessible, such as being granted a license, securing contracts, bypass laws and regulations (Boddewyn, 1988; Martin et al., 2007; Wu, 2005). Moreover, Hellman et al. (2000) mentioned as well that firms influence government officials to extract advantages; with some firms being able to shape the rules of the game.

However, paying bribes can also increase uncertainty, because it is unsure whether the government officials keep their promises (Cuervo-Cazurra, 2008). Complying with bribery can induce a chain of bribes, where the government officials are pushing to gain the most benefits out of it. When bribes are paid it creates additional incentives for government officials to impose more bureaucratic controls and regulations. Some firms are able to create advantages with bribery, but firms would on average perform better when corruption would be minimized. On the surface bribery seems to be cost effective for firms, but the seemingly justifiably bribery practices have hidden costs (Wu, 2005). Bribery can result in legal and financial risks in the future. Fines and jail sentences are not uncommon. Reputation loss can have a significant influence on the firm’s value.

2.3 Corruption and bribery

(12)

12 obvious, but a country can have a high level of corruption while the bribery by firms is relatively low compared to other countries, due to other forms of corruption being predominantly present there. Moreover, the extent to which the country-level corruption actually influences corrupt firm behaviour remains yet to be tested as well. Nonetheless, a relation between country-level corruption and corrupt firm behaviour appears to be logical. Considering the literature background of the concepts corruption and bribery, the following hypothesis is developed:

H1: Higher country-level corruption increases the probability of firm corrupt behaviour. 2.4 Firm level determinants

Upcoming is a review of four possible firm-level determinants for corrupt firm behaviour. First the length of applications, after that the gender of the owner and top manager, and lastly the ownership structure.

2.4.1 Length of applications

(13)

13 theory of bribes where bribes act as speed money (Huntington, 1968, Leff, 1964, Lui, 1985). Moreover, Clarke & Xu (2003) did research on corruption and how it is affected by certain characteristics of the infrastructure sector who receive bribes, they argued that firms who benefit from a utility service are more likely to offer bribes to speed things up and reduce wait times for these utilities.

In conclusion, the complexity and lengthiness might be a reason for firms to try and speed things up. The approach during this study is somewhat different, since we propose that the wait times of applications for public services, permits or licenses could be of influence on firm corrupt behaviour. A firm might be more likely to engage in corruption when the wait times for certain applications are relatively long. Although these services have been linked to corruption and bribery, the lengthiness or wait times for these services and the influence on corruption has not been researched yet at firm-level. When there are short wait times for certain services or utilities there is no need to bribe to speed things up or circumvent bureaucratic hassle. However, even with short wait times it is possible that bribes are demanded by government officials or offered by firms. Nonetheless we theorize that on average bribes are more likely to be involved when there are longer wait times of applications for services or utilities. Therefore, the following hypothesis is formed:

H2: A longer length of applications for public services or utilities increases the probability of firm corrupt behaviour.

Continuing, Clarke & Xu (2003) argued that bribes in the utility sector might be higher in countries where other forms of corruption are also more common. Coupled with this, one could argue that a higher level of corruption would increase the bribes paid for public services or utilities. When corruption is low and bribing is not common, the long wait times of applications for services would be less of an incentive to bribe, considering that it is not normal to do so in such an environment. The opposite goes for countries with high levels of corruption, where long wait times of applications might be more of an incentive to bribe and try to speed things up, as bribing is more of the norm there. Considering the above, not only could the wait times of applications have a direct effect on the probability that a firm will bribe, it could also yield a moderating effect on the level of corruption in a country and the likelihood of firms bribing.

(14)

14

2.4.2 Gender

Not all individuals respond the same to corruption. A controversial topic is the investigation of differences in the behaviour of men and women. Swamy et al. (2001) found that “women are less involved in bribery, and are less likely to condone bribe-taking.” Moreover, the research from Dollar et al. (2001), Melnykovska & Michaoilova (2009), and Swamy et al. (2001) found that a greater representation of women in the parliament, at senior positions in the government, or country’s legislative body is related to a lower level of corruption in that particular country. The samples of Dollar et al. (2001) and Melnykovska & Michaoilova (2009) were transition countries, so the results might not be generalizable for other countries. These studies used cross-country data, specifically looking at the government side, during this study we will look more closely to the supply (firm) side.

The disparity between men and women could be because of both biological and social differences (Alatas et al., 2009). The main argument of Dollar et al (2001) is that men are more individually (selfish) oriented than women. They base their argument on several studies where women behave more altruistic and act for the greater good instead of their personal gain (Goertzel, 1983; Eagly & Crowley, 1986; Glover et al., 1997; Eckel & Grossman, 1998; Ones & Viswesvaran, 1998; Reiss & Mitra, 1998). The exact cause of these differences between men and women cannot easily be identified; the claims about these differences can easily be misinterpreted, it is out of the scope of this research to further investigate it into detail. Nonetheless, it does not take away the fact that these differences exist. Therefore, both women and men could have different attitudes toward corruptive malpractices.

(15)

15 the perspective of the demand side (government) of corruption. There is less known about gender and corruption on the supply side (firms).

Owner’s characteristics influence their behaviour and the way they manage their business (Blackburn et al., 2013; Dyke et al., 1992; Elizabeth & Baines, 1998; Fairlie & Robb 2009). The owner and their characteristics have an impact on the behaviour of the firm, possible also influencing their behaviour towards corruption. The owner is in a position where they have influence whether the firm engages in corruption or not. Taking this all in consideration the following hypothesis is proposed:

H4: Firms with female owners decreases the probability of firm corrupt behaviour.

Not only the owner has a position where they have influence on whether the firm engages in corruption or not. Zahra et al. (2005) suggested that top managers are often drivers for firm corrupt behaviour. Similarly, Collins et al. (2009) said “firm’s involvement in corruption is fundamentally driven by executives’ decisions.” They also linked characteristics of executives such as social ties and the rationalization of corruption to involvement in corruption. Likewise, Kostova & Roth (2002) found a link between characteristics of executives and managerial decision. Conclusively, it seems that persons in top positions within the firm also have influence on corrupt firm behaviour. Therefore, a second hypothesis containing gender is formed:

H5: Firms with female top managers decreases the probability of firm corrupt behaviour.

2.4.3 Ownership structure

Ownership structure refers to the distribution of equity by the equity owners. Several studies have been investigating the relationship between ownership structures and corruption. Clarke & Xu (2002) found that bribes are lower when the utility company is privately owned, the company being the bribe taker. Bribes were higher in the case of state ownership. In contrast to that study we will look at the bribe givers’ ownership structure.

(16)

16 controlled by families are more likely to pay bribes than firms under other forms of governance structure. Wu (2009) also mentioned that individual firms could prevent corruption by broadening the basis of ownership. During this study we will further investigate whether the amount of owners has a linkage to corruption. With different owners the stakes and decisions vary and create a different organizational environment.

A possible theoretical explanation of the amount of owners and the linkage to corruption is the agency theory. The agency theory has a principal-agent perspective; it refers to the relationship between the owner and the manager. Jensen & Meckling (1976) define an agency relationship as “a contract under which one or more persons to perform some service on their behalf which involves delegating some decision making authority to the agent.” Conflicts can arise when there is separation of ownership and control, meaning that the firm is controlled by managers instead of the owner. These conflicts are caused by the different interests of the owner and managers. The aim of the manager might be increasing their own private benefits, these goals might not be aligned with the objective of the owner. Bribery is a means to acquire certain privileges, which consequently, increases the firm’s value (Ramdani & van Witteloostuijn, 2012). This objective is in favour of the owner, and not necessarily that of the manager. The managers who gain no private benefits from bribery are less inclined to do so. A possible explanation is that they do not want to have their good name spoiled for something that is not in favour for themselves. Ramdani & van Witteloostuijn (2012) also found that when there is separation of ownership and control they are less likely to engage in bribery. We assume that in general single owners are also in control of the company, thus it seems plausible that proprietary ownership will increase the likelihood of firm corrupt behaviour. While a firm with different owners, such as a partnership or shareholders, will be less likely to engage in firm corrupt behaviour. One could also argue that there needs to be consensus between multiple owners about the decision to bribe, this could be more difficult to achieve for instance because of the accountability and responsibility of the owners. With multiple owners more persons are aware of the corruptive malpractices, thereby increasing the risk that this comes out to the public which in return could shed the reputation of the owners. All things considered, it seems that the ownership structure and the amount of owners are associated with corruption. Therefore the following hypothesis is formed:

(17)

17 2.5 Conceptual model

(18)

18

3. Methodology

This section presents the methods that are used to empirically test the hypotheses. First the data source BEEPS and sample size will be discussed. Following this are the measurements of the variables, this includes the independent, dependent, moderating, and control variables. After that comes the methodology of the analysis and some preliminary tests.

3.1 Data source

The data for this thesis comes from the EBRD-World Bank Business Environment and Enterprise Performance Survey (BEEPS). The BEEPS is a firm-level survey including a representative sample of an economy’s private sector. It contains items involving corruption; including the prevalence of bribery. The survey is based on face-to-face interviews with business owners and top managers. For this study the fifth round of BEEPS in 2012-2014 will be used; it covers 15,883 enterprises in 30 countries.

(19)

19 3.2 Sample size

The sample of BEEPS covers 15,883 enterprises in 30 countries of Eastern Europe and Central Asia. A list of the sample size can be found below in table 1. Formal registered companies with 5 or more employees are targeted for an interview. The BEEPS uses two instruments: a Manufacturing Questionnaire and a Services Questionnaire. During this thesis both questionnaires will be used since the questions that are needed overlap between those two. One of the goals of BEEPS is to generate a representative dataset that could be used for cross-country comparison. To try to achieve this, BEEPS employed a stratified random sampling method. The survey sample was designed to be broadly representative of the population of firms (Fries et al., 2003).

After removing the cases with missing values a sample of 12945 cases was left. The cases that had missing values seem to be random, no pattern was found so the sample appears to be valid. Peduzzi et al. (1996) did research on logistic regressions analysis, especially on the matter of the necessary sample size to prevent major problems with the results of the model. They found that for a logistic regression (see also the analysis part further in this section) a sample size of 10 events per variable would ensure reliable results and causes no problems. The sample size of this study meets this criterion.

Table 1 Sample size and country-level corruption

Country Sample size BEEPS Sample size after correction CPI

(20)

20 Lithuania 270 221 54 Moldova 360 290 36 Mongolia 360 336 36 Montenegro 150 121 41 Poland 542 422 58 Romania 540 465 44 Russia 4220 3290 28 Serbia 360 292 39 Slovak Republic 268 213 46 Slovenia 270 248 61 Tajikistan 359 273 22 Turkey 1344 1178 49 Ukraine 1002 737 26 Uzbekistan 390 355 17 Total 15883 12945 3.3 Variables 3.3.1 Dependent variable

The dependent variable firm corrupt behaviour is operationalized by means of bribery. Bribery is the act of giving something to a recipient with a public function with the intention of influencing the recipient to the benefit of the giver. This measurement is closely related to two types of bribery by Hellman et al. (2000) namely ‘state capture’ and ‘administrative corruption’. These types involve payments and bribes to public officials with the intention to alter the public officials’ behaviour with respect to certain laws, rules, regulations etc.

(21)

21 WBES; their variable was a measurement of the amount of unofficial payments paid to utilities.

Another measurement for firm corrupt behaviour is the BEEPS question about the percentage of total annual sales paid as bribes. This measurement would be better as it is a continuous variable. The problem however is that the sample size would be reduced by 28,7% due to missing values. The other critical note is that 75,3% of the participants answered this question with ‘0’, implying that no bribes are paid. Which is remarkable, considering that nearly 50% of the participants responded that the firm has bribed before (see the descriptive statistics table 4 on page 27). Therefore, the initial proposed method by having a dichotomous dependent variable will be employed.

3.3.2 Independent & Moderating variables Country-level corruption

The first independent variable is country-level corruption, this variable is retrieved by measuring to what extent a country misuses public power for private benefit. In this case we will measure corruption by looking at the level or degree of corruption in a particular country. Both the ‘corruption perceptions index’ (CPI) compiled by Transparency International and the ‘worldwide governance indicators’ (WGI) compiled by the World Bank are common measurements for country-level corruption. These measurements are concepts which are somewhat related to the corruption type ‘pervasiveness’ as mentioned earlier (Rodriguez, 2005).

(22)

22 BEEPS sample. The index scores countries on a scale from 0 (highly corrupt) to 100 (very clean). In table 1 listed earlier is an overview of countries that are included in the sample size with their according corruption levels.

Length of applications

This variable acts both as an independent as moderating variable. This variable will be operationalized by looking at the wait times of applications for different services and utilities. For instance, we will look at the wait periods for electrical utilities, water utilities, construction-related permits, licenses for imports, and licenses for operating. In the BEEPS they ask whether they applied for such services or utilities, and in reference to that application they ask how long it approximately took in days to receive this since the day of admission. We will measure this variable by summing up the total amount of waiting days for these applications, thereby constructing the length of applications. The applications for these services and utilities have all been applied for in the last two years. With this measurement we try to indicate how much a firm was forced to wait for public services and utilities over the last two years.

Gender of owner

There are two different variables for gender. The first one is the gender of the owner(s) of the firm. This construct is measured with the BEEPS question in reference to question whether there is a female owner or not. This is a dichotomous variable since the answers only consists of two values.

Gender of top manager

Gender of top manager will be measured using a BEEPS question about the gender of the top manager. This question is a dichotomous variable as well since the question refers to whether the top manager is female or not.

Ownership structure

(23)

23 of the largest owner’s share it is hard to determine the actual amount of owners. For instance, a largest share of 50% ownership could result in two owners or an infinite amount of owners. Thus, it is transmuted into a dichotomous variable. This is line with the main argument that firm corrupt behaviour will be higher with a single owner than with multiple owners and should yield the best results for the hypothesis.

3.3.3 Control variables

To ensure more valid results we control for several variables on firm- and industry-level.

Foreign ownership

The first control variable is foreign ownership. Lee et al. (2010) found that firms with a high level of foreign ownership are less prone to bribe. They base their arguments on the residual control theory, arguing that firms with greater bargaining power pay fewer bribes to government officials as they are better able to withstand the pressure from those officials. A high level of foreign ownership increases the bargaining power. Therefore we expect to find a negative relationship between foreign ownership and firm corrupt behaviour. To control for this variable we measure the percentage of ownership by foreign individuals, companies or organizations, given as answer in the BEEPS survey.

Firm performance

Another control variable comes from the study of Clarke & Xu (2004), who found that firms paid more bribes when their ability to pay was higher. A variable linked to ability to pay is firm performance, as a better performance increases the ability to pay (Clarke & Xu, 2004; Svensson, 2003). Clarke & Xu (2004) measured performance by looking at sales growth over the previous 3 years. During this study we take a different approach because of the data source that is used. During the BEEPS the participants are asked to answer a question regarding the expected performance of the firm next year. We assume that a positive attitude towards this question implies that the firm’s performance is good. It will be measured as an ordinal scale from ‘(1)’ implying the best performance to ‘(3)’ being the worst.

Firm size

(24)

24 ability to refuse bribes. To measure this control variable we will look at the number of employees that worked at a particular establishment.

Industry

At last, the control variable industry. The study of Lee et al. (2010) showed that manufacturing firms pay on average fewer bribes than firms in the service industry, Herrera & Rodriguez (2003) found similar results. To control for this variable a distinction will be made between manufacturing and service firms. In the BEEPS survey it is indicated in which industry the firm operates.

Table 2 Summary of variables and measurements

Variables Data type Measurement Dependent

Firm corrupt behaviour Dichotomous Whether a firm bribes or not

Independent

Country-level corruption Interval Corruption Perception Index

Moderating

Length of applications Interval Total amount of waiting days for applications

Ownership structure Dichotomous Whether the largest ownership share size is 100% or not

Gender owner Dichotomous Female owner or not

Gender top manager Dichotomous Female top manager or not

Control variables

Firm size Interval Total amount of employees

Firm performance Ordinal Whether they expect next year’s

performance to increase, stay the same, or decrease

Industry Dichotomous Manufacturing industry or service industry

(25)

25 3.4 Statistical analysis

The method employed in this thesis is a logistic regression. This statistical method is used because the dependent variable bribery is dichotomous, meaning that either a firm bribes or not. The outcomes are presented by ‘2’ (bribing) and ‘1’ (no bribing). A logistic regression can handle different types of data. The results of the logistic regression will show the relationship between the explanatory variables and the binary outcome of bribery. It will show the probabilities or odds whether a firm will bribe or not depending on the predictors, in this case the independent, moderating variable and control variables. A probit analysis is not necessary as the logistic and normal distributions are quite similar, so that the choice of distribution is not that important (DeMaris, 1995). Conclusions reached based on results of a probit analysis or logistic regression should be identical, in this case a logistic regression will be used.

Homoscedasticity analysis is not needed for a logistic regression; the variances do not need to be heteroscedastic. An assumption that has to be met is that a logistic regression model should have minimal multicollinearity, which will be analyzed and presented in the next part (DeMaris, 1995). As earlier stated, Peduzzi et al. (1996) found that a logistic regression had no major problems with 10 events per variable. There is an overestimation of effects when the sample size is small to moderate. During this study we include nine variables in the logistic regression model for a sample size of 12945 cases so these criteria have been met.

3.4.1 Multicollinearity

(26)

26 Table 3 Multicollinearity Variable VIF Firm size 1,035 Performance 1,013 Foreign ownership 1,025 Industry 1,028 Ownership structure 1,101 Country-level corruption 1,018 Length of applications 1,008 Gender of owner 1,421

Gender of top manager 1,369

(27)

27

4. Results

4.1 Descriptive statistics

The descriptive statistics are presented in table 4. The table displays that nearly 50% (Mean 1,49) of the firms sometimes bribed. Continuing, the CPI ranges from 17 to 64, with a mean of 37. This implies that countries with relatively low levels of corruption are rather scarce. Interpretation and generalizing the results has to be done cautiously. The average application length in days is relatively low with an average of 23,24 days, an explanation for this is because 67,9% of the firms did not apply for services that we considered during this study, thus lowering the mean. In 35% of the firms there is a female amongst the owners; it is not a reflection of the percentage of female owners, this statistic would in reality probably be lower. Moreover, the variable foreign ownership has a low mean, this is due to only 6,8% of the companies having some degree of foreign ownership, this could influence the results of the control variable in the logistic regression.

Table 4 Descriptive statistics Variables Obs. Mean Std.

Deviation

Minimum Maximum

Dependent Variable

Bribery 12945 1,49 0,500 1 (No) 2 (Yes)

Independent & Moderating variables

CPI 12945 37,26 11,457 17 64

Gender Owner 12945 1,65 0,476 1 (Female owner) 2 (No female owner) Gender Top manager 12945 1,80 0,398 1 (Female) 2 (Male)

Length of applications 12945 23,24 86,275 0 (Days) 1485 (Days)

Ownership structure 12945 1,4506 0,49757 1 (Single owner) 2 (Multiple owners)

Control Variables

Industry 12945 1,59 0,492 1 (Manufacturing) 2 (Service)

Firm size 12945 64,74 274,397 1 11000

Performance 12945 1,57 0,713 1 (Increase) 3 (Decrease)

(28)

28 4.2 Model results

In this section the results of the logistic regression will be presented. Table 5 on page 30 provides a summary of the results. The pseudo-R² is not given, as we cannot draw conclusions based on those results, since there is no commonly accepted measure to determine how well the model fits the data in a logistic regression (DeMaris, 1995). Four different models are compiled; the results of them will be summarized one by one.

Model 1

Model 1 only examines the effect of the control variables as predictors for bribery by firms as these variables have been linked to the dependent variable in previous research. Neither the control variables firm size nor industry were significant (p>0,1). It is unexpected that industry has no significant relationship with bribery, since previous studies showed that firms in the manufacturing industry pay fewer bribes than firm in the service industry. Foreign ownership is moderately significant (p<0,1) but the impact on bribery is considerable small as the coefficient is -0,002. The last control variable performance has a high significant (p<0,01) relationship with bribery. It is a strong predictor since the coefficient is 0,92, firms expecting an increase in performance next year are less likely to pay bribes than the other firms. These results differ from previous research.

Model 2

In model 2 the independent variables country-level corruption and length of application (now as main effect) are added. Some changes occurred among the control variables. Foreign ownership is not significant in this model, while it was moderately significant in the last model (p<0,1). Performance remained highly significant, but the coefficient increased (from β = 0,92 to β = 0,158). The impact of performance on the probability of bribing is now more substantial.

(29)

29 odds are that when the value of length of applications increases by a single day the odds that the firm will bribe increases. These findings support hypothesis 2 at first glance, but the last model will be leading.

Model 3

In model 3 only the moderating variable is added, to properly see the effects of this variable. The control variables remained nearly the same, only the variable industry became moderately significant (p<0,01) with a coefficient of 0,037. The odds that service firms will bribe are lower than manufacturing firms. These findings are the opposite of previous studies. The moderating variable length of applications (CLC x LOA) which influences the relationship between country-level corruption and firm-corrupt behaviour is in this model highly significant (p<0,01), but the coefficient is 0,000. Despite the results being significant, the impact on the odds of firm bribery is nonexistent. At first glance, hypothesis 3 is rejected, but the last model including all the variables is leading.

Model 4

Model 4 includes all of the variables, and is therefore leading. The remaining variables ownership structure, gender of owner, and gender of top manager are now also included. Among the control variables are no significant changes when comparing to the previous model 3. The results of country-level corruption as well as length of applications are identical to the previous model; the findings give support for hypotheses 1 & 2. The moderating effect of length of applications is not significant in this model as opposed to model 3. This model includes all the variables and the results from this model are therefore leading. Hypothesis 3 is rejected.

The independent variable gender of owner is newly added in this model. This variable is significant (p<0,05) with a coefficient of 0,087. These results display that the odds of bribery are higher for firms without a female owner. This is in line with hypothesis 4. Moreover, the independent variable gender of top manager is newly added in this model and is not significantly related to bribery, resulting in hypothesis 5 being rejected. The last independent variable is ownership structure, it proves to be significantly related to bribery (p<0,05). With a coefficient of -0,079 it shows that proprietary ownership increases the probability of bribery. The odds that firms with a single owner will bribe are higher than firms with multiple owners. These results support hypothesis 6.

(30)

30

Table 5 Logistic regression results

Model 1 Model 2 Model 3 Model 4

Control Variables Firm Size 0,000 (0,000) [1] 0,000 (0,000) [1] 0,000 (0,000) [1] 0,000 (0,000) [1] Performance 0,92*** (0,025) [1,097] 0,158*** (0,026) [1,171] 0,158*** (0,026) [1,172] 0,157*** (0,026) [1,170] Foreign Ownership -0,002* (0,001) [0,998] 0,000 (0,000) [1] 0,000 (0,001) [1] 0,000 (0,001) [1] Industry -0,008 (0,036) [0,992] -0,058 (0,037) [0,943] -0,067* (0,037) [0,944] -0,066* (0,037) [0,936] Independent variables (Main effects)

Corruption Perception index (CPI) -0,041*** (0,002) [0,960] -0,041*** (0,002) [0,960] -0,041*** (0,002) [0,960]

Ownership structure (OS) -0,079**

(0,038) [0,924]

Gender owner (GO) 0,087**

(0,045) [1,091]

Gender top manager (GTM) -0,027

(0,053) [0,973] Length of applications (LOA) 0,001***

(0,000) [1,001] 0,002** (0,001) [1,002] 0,002*** (0,001) [1,002] Moderating variables CPI x LOA 0,000*** (0,000) [1] 0,000 (0,000) [1] Constant -0,176** (0,071) 1,301*** (0,094) 1,283*** (0,94) 1,308*** (0,147) Chi² 17,832*** 718,849*** 719,839*** 730,836*** -2 Log likelihood 17919,8 17218,8 17217,8 17206,8 Observations 12945 12945 12945 12945

Standard error in parentheses. Odds ratio in brackets.

(31)

31 4.3 Robustness check

Initially the dependent variable was measured as dichotomous. As robustness we want to check if the determinants also indicate how often a firm bribes, as the initial measurement solely looks whether a firm bribes or not. The measurement is based on the answer involving the frequency of irregular additional payments or gifts to get things done. The question’s answers ranges from ‘(1) never’ to ‘(6) always’, this time without transmuting the answer into a dichotomous variable. Thus, an ordinal regression will be compiled. An ordinal regression has to meet certain assumptions. The independent variable performance is originally measured as ordinal; this variable will now be treated as categorical variable. Another assumption is multicollinearity; this one has been met as shown in table 3. Lastly, an ordinal regression should have proportional odds. However, since there is an interaction effect in the model the parallel line test cannot be executed with a full model. Instead, the test is run with the model excluding the interaction effect, to see whether the other variables meet this assumption. The results of the parallel test are displayed in table A3 in the appendix. It proves to be significant (p<0,01), so the validity of the model is uncertain, as the null hypothesis is rejected. However, several studies looked into this and found that the tests often reject the proportional odds assumption. Especially when the sample size is rather large, the model includes numerous explanatory variables, or when there are continuous explanatory variables (Brant, 1990; Allison, 1999; Clogg & Shihadeh, 1994).

The results of the model are listed in table A3 in the appendix, but hard conclusions cannot be drawn from this. The model is improved over the standard model by including the explanatory variables (-2 Log likelihood, p<0,01). The goodness of fit is not rejected either (insignificant Pearson test). Thus, it seems that the data fits the model well. Keep in mind that the proportional odds assumption is still violated.

(32)

32

5. Discussion

The objective of this thesis is to examine several determinants for firm corrupt behaviour. In this part we will discuss and analyze the results from the research. It will be divided into several parts. First we will discuss the results in general, followed by the theoretical-, managerial- and policy implications. At last the limitations and future research.

5.1 Discussion results

The results showed support for hypothesis 1 that country-level corruption is related to firm corrupt behaviour. Lower country-level corruption lowers the probability of firm corrupt behaviour. This is in line with previous studies of Wu (2009) and Martin et al. (2007). The results also show that CPI is good indicator for bribery by firms, although it does not explain the differences among the firms but shows that it is generalizable for the private sector as a whole. Country-level corruption is part of the institutional environment, and it shows that firms’ behaviour is affected by this; it is an extension of the institutional theory. The standards and norms present in a society influence the decision making of firms, also on the matter whether to bribe or not.

Moreover, the length of applications proved to be a predictor of bribery as well, since hypothesis 2 is supported. Longer wait periods for public services or utilities result into a greater probability of bribery. This is an addition to previous literature which mentioned that overregulation, an unstable and complex legal administrative framework, and inefficiency of government service delivery is related to corruption and bribery (Gavirira, 2002; Lambert-Mogiliansky 2001). The findings add value to those studies and show that cumbersome bureaucracy gives room for corruption. The results are also in line with those of Clarke & Xu (2003) who found that nearly a quarter of the enterprises in transition economies pay bribes to utility firms. Although the relation between the length of applications and bribery has been found, it remains to be seen if the initiative to bribe came from the demand or supply side. Both parties could have been responsible.

(33)

33 Furthermore, the influence of gender on bribery. This relationship is new in this context. It builds upon previous research that investigated how men and women make different decisions and uphold different attitudes towards corruption. Most of the prior research looked at government positions, the demand side. In this study we looked at the supply side of bribery, by looking at the positions within firms. The logistic regression results show that the gender of the owner is related to bribery of a firm. Without the presence of female owners the firms are more prone to bribe. This could possibly imply that males and females behave differently and have possible different attitudes towards bribery. Mixed results regarding gender were previously found, but these findings seem to be consistent with the studies of Swamy et al. (2001), Dollar et al. (2001) and Melnykovska & Michaoilova (2009). The underlying causes are difficult to determine, but the findings display that these differences exist. Moreover, the findings show that owner’s characteristics influence the behaviour of the firm; this is consistent with previous studies (e.g. Blackburn et al., 2013; Dyke et al., 1992; Elizabeth & Baines, 1998; Fairlie & Robb 2009).

On the contrary, hypothesis 5 is not supported, no differences are found between gender of the top manager and bribery of the firm. Several reasons could be at hand. One explanation is that female top managers behave differently than female owners. They each uphold a different attitude towards bribery. Men and women could behave the same when they are positioned at this function in a firm. Another reason could be that top managers are not necessarily making the decision for the firm whether to bribe or not, and that the owner has more influence on that matter. Then again, this would be inconsistent with the studies of Zahra et al. (2005) and Collins et al. (2009) who suggested that top managers and executives are often the drivers for firm corrupt behaviour.

The last independent variable, ownership structure, is significantly related to bribery by the firm as hypothesis 6 is supported. It proves that ownership structure with respect to the amount of owners is a predictor of corruptive firm behaviour. These findings are consistent with the statement of Wu (2009) who stated that firms could prevent corruption by broadening the ownership, as our findings show that proprietary ownership is related to an increased probability of bribing. Firms with multiple owners are less inclined to bribe. The results do not show if there is a linear relationship between the amount of owners and corrupt firm behaviour, but they do display that having more than one owner decreases the odds of bribery.

(34)

34 firms tend to pay fewer bribes than small firms to utilities; in our study no difference has been found. This could possibly imply that large firms tend to pay fewer bribes to utilities, but not in paying bribes in general.

Secondly, the control variable firm performance. In all models the relationship is highly significant with a positive coefficient. A firm expecting a decrease in performance increases the probability of bribery by that firm. This contradicts the suggestions of Clarke & Xu (2004) and Svensson (2003) that bribery depends on the firm’s ability to pay which partly depends upon the firm’s performance. The measurement of firm performance was different than the other studies, see appendix A3 under firm performance for further remarks.

No significant relationship was found in most of the models between foreign ownership and corrupt firm behaviour. Lee et al. (2010) found that a greater amount of foreign ownership increases the ability to withstand the pressure to the demand of bribes. It is based on the arguments that a greater percentage of foreign ownership results into greater bargaining power (Eden et al., 2005; Vernon, 1971). The outcomes are divergent and it is disputable whether or not foreign ownership matters, but the evidence of this study leans more towards a nonexistent relation between foreign ownership and bribery by firms. Keep in mind that only a small percentage of the firms are foreign owned.

The last control variable is industry, which was found to be moderately significant (p<0,1) in three of the four models, including the complete model 4. Firms in the service industry are less likely to pay bribes than firms in the manufacturing industry. Rodriguez (2003) and Lee et al. (2010) had opposing results as they found that manufacturing firms pay fewer bribes in general. As it is moderately significant the relationship remains uncertain and no hard conclusions can be drawn from this.

5.2 Theoretical implications

(35)

35 2004; Swamy et al., 2001; Svensson, 2003). This is accomplished by looking at factors that are previously not considered yet in this context. The findings display that there is more to it than has been found so far. For instance, the findings show that the characteristics of the owner such as gender matter in relation to corruption. This is in line with previous studies of for example Dollar et al (2001) and Swamy et al. (2001) who found a relation between corruption and gender, although in a different context (government positions). This also implies that the characteristics of the owners affect their behaviour towards business. The results increase the evidence that differences exist between the attitudes and behaviour of men and women with respect to corruption. Furthermore, the results of length of applications builds upon previous studies who linked bureaucracy, overregulation, quality of governmental services, and a complex legal administrative frame work to corruption (e.g. Chen et al., 2008; Friedman et al., 2000; Gaviria, 2002; Lambert-Mogiliansky, 2001). The variable length of application can be refined and used as a measurement, indicating the period a firm is forced to wait for matters related to the government. This measurement is new as it looks at the firm-level what the consequences are for individual firms. The last findings indicate that single owned firms are more likely to bribe than firms with multiple owners. This adds value to previous findings of ownership structure and corporate governance in relation to corruption, as the amount of owners seem to matter as well.

5.3 Managerial implications

The results are also relevant from a managerial perspective and entail several implications. Firms who are looking to expand to foreign countries have to be aware of the level of corruption in that country. As firms who operate in those foreign countries are influenced by that environment and need to be aware that bribing can be the norm in certain countries. Bribing might not be problematic in that foreign country, but they have to be aware of their home countries’ norms and regulations. They might face different consequences in their home country even when the corruptive malpractices took place in the foreign country. It could also shed the reputation of the firm in their home country. Even firms who do not have the intention to bribe could have difficulty refusing the demand of bribes, or sometimes they are practically forced to do so.

(36)

36 of ownership as Wu (2009) stated. The results also show that firms with female owners are less inclined to bribe. This can give managers an indication which firms have a higher probability of bribing, and whether or not they want to engage with such firms. The length of applications is also relevant for managers. Keeping in consideration that long wait periods of applications raises an increased probability of bribing. By acknowledging this they can try to counter that and avoid that government officials come in a position where they can demand bribes.

5.4 Policy implications

The results also provide some policy implications. So far research regarding bureaucracy and corruption share the same view that increased complexity gives room for corruption. During this study we looked at the wait periods of certain applications and found that longer wait periods resulted into more bribes. Policy makers should take this into account. By setting policies to reduce the complexity of bureaucracy and regulations they can prevent that government officials gain too much power where they can exert it to demand bribes. Shortening the wait periods of applications for government services reduces the room for corruption.

Policy makers could set certain standards regarding the ownership structure, and try to stimulate the broadening of the ownership base of firms, as firms with multiple owners are less inclined to bribe. Encouragement of firms with female owners might also prevent and reduce corruption. This has to be done cautiously since more research is needed before drawing hard conclusion on this matter as it is also rather controversial.

5.5 Limitations and future research

The results and methodology are subject to several limitations, one has to be aware of this. Some limitations will be first discussed, followed by suggestions for further research.

(37)

37 applications which are not included in this study. Another limitation regarding length of application is that no distinction was made between privately and government owned utilities, previous research has shown that government owned utilities are more prone to demand bribes and are less efficient. It was out of the scope of this study to take this into account. Furthermore, the gender of the owner. A side note here is that during this study we merely looked if one of the owners is female. It could be that there are numerous owners and only a small percentage of that is female. Moreover, in this study the measurement is based on whether firms paid bribes or not, but there was no measurement when these bribes were actually paid. Perhaps the bribes were paid when the characteristics of the firm were different and they overhauled the structure of the company, thus altering the relationship between the independent variables and bribery. Nonetheless, we assume that the participants based their answers on recent events as they come more easily to mind instead of events that happened in the past. Lastly, one needs to be aware that the dependent variable is still measured using subjective data, given the secrecy attached to bribing the data can still be unreliable, remarks about this have been made previously when discussing the data source. Also, mistakes can be made by either the interviewers or participants who misinterpret the questions. Further remarks about the limitations can be found under appendix A3.

(38)

38

6. Conclusion

(39)

39

Literature

Ades, Alberto, and Rafael Di Tella, 1999. “Rents, Competition, and Corruption.” American Economic Review 89 (4), 982-993.

Aidt, T. S. (2009). Corruption, institutions, and economic development. Oxford Review of Economic Policy, 25(2), 271-291.

Alatas, V., Cameron, L., Chaudhuri, A., Erkal, N., & Gangadharan, L. (2009). Gender, culture, and corruption: Insights from an experimental analysis. Southern Economic Journal, 663-680.

Alhassan‐Alolo, N. (2007). Gender and corruption: Testing the new consensus. Public Administration and Development, 27(3), 227-237.

Ali, A. M., & Isse, H. S. (2002). Determinants of economic corruption: a cross-country comparison. Cato J., 22, 449.

Allison, P. D. (1999). Logistic regression using the SAS system: Theory and application. Cary, NC.: SAS Institute.

Andvig, J.C., 1991. The economics of corruption: a survey. Studi Economici 46 (43), 57– 94. Bardhan, P., 1997. Corruption and development: a review of the issues. Journal of Economic

Literature 35 (3), 1320–1346.

Blackburn, K., Bose, N., & Haque, M. E. (2006). The incidence and persistence of corruption in economic development. Journal of Economic Dynamics and Control, 30(12), 2447-2467.

Blackburn, R. A., Hart, M., & Wainwright, T. (2013). Small business performance: business, strategy and owner-manager characteristics. Journal of small business and enterprise development, 20(1), 8-27.

Boddewyn, J.J., 1988. Political aspects of MNE theory. Journal of International Business Studies 19, 119–144.

Brant, R. (1990). Assessing proportionality in the proportional odds model for ordinal logistic regression. Biometrics, 46, 1171-1178.

Broadman, H. G., & Recanatini, F. (2001). Seeds of corruption–Do market institutions matter?. MOCT-MOST: Economic Policy in Transitional Economies, 11(4), 359-392. Brunetti, A., & Weder, B. (2003). A free press is bad news for corruption. Journal of Public

economics, 87(7), 1801-1824.

The influence of country-level corruption and firm-level determinants on firm corrupt behaviour