Why do some firms bribe more than others? The effects of age and ownership on bribery

Why do some firms bribe more than others?

The effects of age and ownership on bribery

Master Thesis International Business and Management

Faculty of Economics and Business

1

Abstract

This study extends the literature on firm level determinants of corruption by exploring the impact of firm age, ownership structures, and foreign ownership on bribery. The relationship between these firm characteristics and bribery was estimated using a Logit regression, with the data covering three Asian countries: Cambodia, Laos and Vietnam and a total of 1272 observations. The analysis of firm-level characteristics affecting bribe payments made in this study draws on anomie theory and agency theory. The findings provide support for the negative relationship between shareholder ownership and bribery. We find no relationship between firm age and bribery, nor for foreign ownership and bribery. Moreover, significant interactions do exist between foreign ownership and partnerships and between foreign ownership and single ownership. Our evidence suggests that a firms ownership structure in relation to foreign ownership can explain why some firms pay more bribes than others.

2

Table of Contents

2.3 Firm level bribery ... 8

2.4 Hypotheses ... 11 2.4.1 Firm age ... 11 2.4.2 Ownership ... 11 2.4.2.1 Foreign ownership ... 15 3. Methodology ... 16 3.1 Data source ... 16 3.2 Countries ... 17 3.3 Sample ... 18 3.4 Dependent variable ... 19 3.5 Independent variables ... 19 3.6 Control variables ... 20 3.7 Statistical model ... 22 3.8 Method assumptions ... 22 4. Empirical results ... 26 4.1 Descriptive statistics ... 26 4.2 Regression results ... 27 5. Robustness tests ... 31

5.1 Additional analysis of bribery ... 31

6. Conclusion ... 35

6.1 Added value ... 35

6.2 Limitations and further research ... 36

Appendix ... 37

Appendix 1 – Survey questions ... 39

Appendix 2 – Figures and tables ... 41

Appendix 3 – Model fit ... 43

Appendix 4 – Robustness test ... 44

3

1. Introduction

Many firms hide corrupt practices behind secret subsidiaries and partnerships while others try to influence decision making illicitly, or exploit tax laws (Transparency International, 2012). These corrupt practices are persistent in the developed and developing world. According to the Bribe Payers Index (2011) not even the largest economies are totally free of bribing. Corruption is a global problem, although there are vast differences between the levels of corruption within and between countries. Therefore addressing corruption remains important for encouraging moral and ethical behavior among firms.

Research into corruption has gained much attention in recent years. While the focus has been mainly cross-country based with the use of country-level data, now increasing numbers of studies focus on the individual firm or manager, while using firm-level data. Studies on a country level try to understand the consequences of bribery. It is usually assumed that corruption has prejudicial effects on economic growth. However, Leff in 1964 already questioned this assumption by examining whether the effects of corruption favor or hinder economic development. Leff (1964) finds that corruption might raise economic growth. Others only find negative associations between corruption and growth; Mauro (1995) and Shleifer & Vishny (1993) find that corruption hinders private investment and thereby reduces economic growth. Méon & Sekkat (2005) also find that a weak rule of law, an inefficient government and political violence, further increases the negative impact of corruption on investment. Earlier literature concludes that corruption is often seen as negative, while this is not always the case. The results are ambiguous since countries characterized as corrupt often have high economic growth and foreign direct investment rates. Thus, in this sense the high growth rates of the corrupt emerging markets and markets in transition remains a mystery.

4

on why some firms bribe more than others and what distinguishes bribing firms from non-bribing firms (e.g. Jeong & Weiner, 2012; de Jong et al. 2012; Martin et al., 2007; Svensson, 2003; Lee, Oh & Eden 2010; Wu, 2009; Swamy, Knack, Lee & Azfar, 2001). All of the former are studies at the firm level instead of the country level. This type of study is gaining more attention nowadays.

It is well researched that firms from countries in transition are more willing to bribe than firms in developed countries. Therefore, transition economies can be characterized as corrupt (Tonoyan et al, 2010). An explanation for this is the underdeveloped formal institutional framework in these countries, which also provide the 'rules of the game' (North, 1994). Transition economies are economies that change from a centrally planned economy toward a free market economy. In transition countries where corruption rates are high, the unofficial economy is expected to be larger. It can be seen that in this type of market, firms can avoid paying bribes to public officials or avoid playing by the rules of the game. Besides the presence of corruption, high tax rates and low trust in the legal system contribute to a larger activity in the unofficial economy (Johnson, Kaufmann, McMillan & Woodruff, 2000). This unofficial economy is the unregulated and illegal market where business sometimes might take place. Unofficial economies often play a considerable role in transition economies.

For the purposes of this thesis the focus will be to gain insight into how firm age and different ownership structures influence the incidence of bribery of firms operating in a transition economy. Firm age as an independent variable is not researched much in existing literature and is often used as a control variable. It might be argued that younger firms have fewer resources and thus may bribe more than older firms and therefore is a topic worthy of research. Another topic of interest is firm ownership, only a few studies address ownership types in relation to bribery. Since it is already acknowledged that privately owned firms bribe more than state-owned firms, the focus will be on different private ownership types and the differences in bribes paid by these firms. This will contribute to a better understanding of why some firms bribe more than others.

5

The following research questions in particular will be addressed in order to answer the research question:

 What is corruption and how can it be measured?

 What is the relationship between firm age and bribery?

 Do different ownership types influence the incidence of bribery? What is the relationship

between ownership and bribery?

 Does a relationship exist between foreign ownership and bribery? Does this relationship

have a moderating effect on the different ownership types?

The aim of this study is to contribute to the understanding of why there are differences in bribe rates within countries. A better understanding of this will provide insight for managers on how to deal with corruption and operate in a more ethical manner in transition economies. Since the main research question is of a general nature, the focus will be on the following two firm characteristics: firm age and ownership structure. For the types of ownership a distinction will be made between single ownership, shareholder ownership, and partnerships. Two existing theories, namely Anomie theory and Agency theory, will be used to supplement the development of the hypotheses. The last sub-question focuses on the interaction effect between foreign ownership and the three ownership structures. For the purposes of this thesis, the main focus will be on the relation between firm age and the ownership types independently on bribery. This relationship will be empirically tested with the use of data from the Business Environment and Enterprise Survey (BEEPS). After testing the hypotheses and interpreting the results, the interaction effect between the ownership structures and foreign ownership will be tested for as an extra robustness test.

2. Literature review

6

paying large amounts of money to get things done or expedite something. Bribing to win public contracts, avoiding regulations or accelerating services is a persistent concern in the world (Transparency International, 2009). In a recent study by Transparency International the results have shown that 2 out of the 5 respondents have been asked to pay a bribe to get things done. Studies that focused on emerging and developing countries only, show that the occurrence of paying bribes is even higher (Transparency International, 2009). Jain (2001) indicates three types of corruption: The first type of corruption is referred to as “grand corruption’ also known as political corruption, which are the actions of the political elites by which they exploit their power to make or change economic policies. The second type is “bureaucratic corruption”, which is also known as petty corruption. This is probably the most common type of corruption since it involves the payment of bribes and gifts to receive for example a service or approval. The last form of corruption that Jain (2001) indicates is known as “legislative corruption” and refers to the extent to which one can influence the behavior of legislators. The concepts bribery and corruption are closely related phenomena (Weber & Getz, 2004) and are used together here. Bribery and corruption can be distinguished as follows: a bribe is a payment to persuades a person to act in a way that is contrary to his responsibilities; corruption is the misuse of power for private benefit, and can include a broader range of corrupt activities, such as money laundering, fraud, abuse insider information, gift giving etc. (Weber & Getz, 2004).

7

corrupt practice … has become institutionalized in a society, they are more likely to conform to those social expectations to obtain legitimacy” (Spencer & Gomez, 2011: 282). Thus, it depends on the context whether or not bribery is seen as harmful for a country e.g. in many European countries bribery is seen as a problem, while in many emerging markets bribing is a daily business practice. Therefore, to say whether or not bribery is wrong, one has to consider whether it is an individual act or a rule of behavior (Steidlmeier, 1999). In the case of an individual act, one cannot achieve legitimacy of the corrupt practices if it is not accepted as a rule of behavior.

8

bribe practices, and the clear organization of corruption, contribute to the high growth rates of some Asian countries.

An often used definition comes from the World Bank, which defines corruption as ‘the abuse of public office for private gain’; while according to Transparency International (2009) it could also be defined as “the abuse of entrusted power for private gain”. Both definitions can be applied to the public and private sectors. The definition of bribery from the World Bank will be used in this thesis, as it concurs with the measures of bribery that are used in the Business Environment and Enterprise Survey (BEEPS). In addition to this, it is used here as a measure of administrative corruption and measures both informal payments and gifts, thus covering an extensive as well as the most prevalent form of bribery.

2.2 Measuring corruption

By its nature, corruption is difficult to measure, since it is hidden and happens outside the scope of the public (Jain, 2001). However, many studies do measure corruption. This occurs from two perspectives: the perspective of the bribe-taker or bribe-payer; and at the macro-level or firm-level. According to Svensson (2003) research on the determinants of corruption have three common features: Firstly, it is based on cross-country analysis; secondly, it uses data on corruption derived from perception indices; and thirdly, it explains corruption as a function of countries’ policy-institutional environment. This type of data allows to measure corruption on a macro-level. As an example, Mauro (1995) found that corruption lowers private investment, and thereby reduces economic growth. Another study on the macro level shows that corruption can affect resource allocation and the distribution of income in economies (Blackburn, Bose & Haque, 2006). A disadvantage of measuring corruption on a macro level is that it does not explain the within country variation in corruption (Svensson, 2003).

9

formulated in an indirect manner to avoid associating the respondent of wrongdoing; questions on corruption are asked at the end of the survey to assure a necessary level of credibility and trust between the interviewer and manager. Additionally, a number of questions on corruption are also asked in different sections of the survey to enhance the reliability of the answers (Reinikka & Svensson, 2006). A difficulty that can arise while measuring corruption, both on the macro- and firm level, is that answers to the surveys can be subject to non-response or false response. This may occur in order to protect one selves from possible government reprisals and due to the fact that corruption is a sensitive topic (Jensen, Li & Rahman, 2010). This happened to the BEEPS data as well, a large number of firms indicate that they do not know whether or not there firm bribes or refuses to answer.

From the literature, two data sources can be identified that measure corruption and bribery on a regular basis. This data comes from the BEEPS, as mentioned above, and the Corruption Perception Index (CPI) from Transparency International. The BEEPS survey offers micro data and measures three distinct but related aspects of corruption, namely administrative corruption, state capture and the influence on both. The first, administrative corruption, measures “the extent to which firms make illicit and non-transparent private payments to public officials in order to alter the prescribed implementation of administrative regulations placed by the state on the firm’s activities” (Hellman, Jones, Kaufmann, 2000:5). State capture measures the extent to which firms make payments to public officials in order to influence the formation of laws, rules and regulations. Furthermore, the BEEPS measures the extent to which firms have an influence on the formation of laws, rules, and regulation without making private payments to public officials. The survey data allows to measure corruption on a firm-level, and thus takes intra country variation into account. The second institution that measures corruption is Transparency International, the CPI ranks 176 countries on a scale from 0 (very corrupt) to 100 (very clean), in two-third of the countries that are included in the index, corruption is a big problem (Transparency International, 2012). Nevertheless, a limitation of Transparency International is that the data only captures perceptions of the extent of corruption from the perspective of business people and country experts (Transparency International, 2012).

2.3 Firm level bribery

10

corruption literature. As mentioned in the introduction, many studies have focused on the supply side of bribery and take institutional differences (from a macro level) into account as an explanation of why firms bribe. In recent times a handful of studies have focused on why some firms bribe more than others and what distinguishes bribing firms from non-bribing firms (firm level studies). Of particular interest is to investigate the question of why some firms bribe more than other firms. This is a firm specific operation, it can also be measured at the individual level of both managers and public officials. Firms cannot behave and operate without individual people acting on its behalf (Li & Ouyang, 2007), therefore, firms operating in the same environment can face different pressures to engage in corrupt practices due to differences in the characteristics of the firm and in the individuals that manage the firm. For example, Swamy et al (2001) examine the relation between corruption and bribery and shows that women are less likely to participate in corrupt practices. Additionally the level of income of an individual can affect the amount and incidence of bribery. In the literature it is argued that the rich bribe more than the poor, mainly because they use more officials to get things done (Hunt & Laszlo, 2012). Nonetheless, Hunt and Laszlo (2012) also show that the lower income parties use a higher share of their income to pay bribes. Which means that bribes are unequally distributed and the poor are hit much harder than the rich are.

11

countries at both a micro- and macro-level. Their results suggest that the volume of exports, firm size and industry competition affect the probability of paying bribes as well. However it are not the firm characteristics solely that affect the decision to bribe. The business environment and institutional context are an important component as well in the choice to bribe or not (Chen et al, 2008; Martin et al, 2007).

In conclusion, most existing literature on the determinants of bribery at the firm-level are explained by considering firm size, competition, corporate governance, or characteristics of individual managers. Other characteristics of firms that could influence the extent of bribery are those such as firm age and ownership structures, which receive insufficient attention. These characteristics will be further examined in the following section.

2.4 Hypotheses 2.4.1 Firm age

A difference may exist between the age of a firm and bribe payments which can be explained through anomie theory, a socio-economic theory. Anomie theory gives an explanation of how cultural and social structures of a society exert pressure upon persons and firms to engage in alternative, deviant forms of behavior (Merton, 1938; 1968). When facing pressure, individuals might show a response in which cultural norms are rejected although one continues to abide to institutional norms. It also might be possible that individuals reject institutional practices and hold on to cultural goals, or reject both of them (Merton, 1968). The deviant forms of behavior that individuals or groups engage in are not necessarily dysfunctional (Merton, 1968), thus goals might still be reached although in a normless way. An extension on the classical anomie theory is the concept of institutional-anomie theory. This is influenced by pressures from several major institutions such as the economy, the family, and the state and account for sources of deviant behavior (Messner & Rosenfeld, 1997). It is argued that cultural and institutional elements limit the range of possible actions required to achieve the desired ends. The relationship between anomie theory and corruption is present in the literature. Martin et al (2007) argue that: “Culturally induced performance pressures … elevate rates of bribery among firms. Firms

12

acceptability or legitimacy. For these firms, bribery can be a path to the achievement of important goals” (p.1403/1404). Thus, anomie theory creates pressure for firms to deviate from

the acceptable norms of behavior and achieve their goals in an unethical way.

When relating anomie theory to bribery, the relevance of the age of firms can be made. According to Thornhill & Amit (2003) younger firms are attributable to inadequate resources and capabilities, thus experience more difficulties in generating a competitive advantage. Beside this, they argue that younger firms have more problems in generating appropriate amounts of revenue than older firms have. Thus, younger firms face a liability of newness which can create pressure to achieve goals regardless of whether or not it is ethical. Younger firms must develop relations with outside individuals and organizations, here trust may be lacking and additionally, these young firms may lack stable ties with customers and others (Stinchcombe, 1965 in Thornhill & Amit, 2003). Therefore, it is argued that younger firms are found to bribe more than older firms since they face more external pressures to deviate from the normal forms of behavior. On the other hand, Lee et al (2010) have shown that older firms have a tendency to pay more bribes. Older firms can have a lower incentive to avoid corrupt practices. Others, such as Kuncoro (2004), argue that the bribe rate is negatively correlated with the age of the firm. Thus, contradicting results are present in the theory.

Here it is argued that younger firms face a high pressure to achieve their goals and might more easily deviate from acceptable behavior to achieve this, and therefore will have a high bribe rate to overcome liabilities such as newness. This bribe rate becomes lower when firms become more mature and have acquired the necessary resources, capabilities, and skills to achieve their goals in a way that is consistent with the existing norms and values in a particular business environment, and will finally arise again when firms become ‘old’. Bribery might arise again when firms reach a certain post mature age because firms can lose their incentive to play by the rules of the game, lose their competitive advantage or lack the ability to match their business with the changing environment, and thus face more pressure to deviate from accepted behavior to achieve their business goals. The arguments above result in the following hypothesis:

13

2.4.2 Ownership

The relationship between entry modes of foreign direct investment and corruption has been largely discussed in the corruption literature (Uhlenbruck, Rodriguez, Doh & Eden, 2006; Karhunen & Ledyaeva, 2012; Tekin-Koru, 2006). This stream of research has its focus mainly on the choice between joint-ventures and wholly-owned subsidiaries to overcome the hazards of corruption. However, only a few studies look at the relationship between ownership structures of domestic firms and bribery (Lee et al, 2010; Clark & Xu, 2004; Jeong & Weiner, 2012). Lee et al (2010) examines three characteristics of firms that are important predictors of the vulnerability to corruption. These characteristics are foreign ownership, export orientation, and state ownership. Results show that firms with high levels of state and foreign ownership are less likely to bribe and pay smaller bribes. Clark and Xu (2004) have studied the incidence of bribery in private-owned and state-private-owned firms. In a survey of 2000 enterprises across 21 countries in Eastern Europe and Asia, their results indicate that firms with private ownership pay higher bribes than owned firms. Clark and Xu (2004) argue that private firms might bribe more than state-owned firms because private firms have less political power and influence. On the other hand, Jeong and Weiner (2012) focus on more diverse ownership structures with the use of data from Iraq; namely state-owned enterprises, privately held, publicly traded, and partially state-owned firms. Consistent with the results of the authors mentioned above, Jeong and Weiner (2012) find that owners in private firms have the highest bribe rates. Furthermore, the results show that managers in listed firms, bribe less than managers of fully state-owned firms. An explanation for this might be that listed firms must be more transparent to achieve legitimacy, which reduces the possibility to bribe.

14

state owned firms. For the purposes of this thesis, the following ownership structures will be examined: partnerships, shareholder ownership, and single ownership. The lack of existing research on this topic and the availability of firm-level data allows examination of these ownership structures in relation to the incidence of bribery.

Agency theory can be used to explain the differences in corruption between different ownership types. The agency theory of the firm argues that an agency problem occurs when two parties, the principal and the agent, enter into a relationship where the agent acts on behalf of the principal (Jensen & Meckling, 1976; Ross, 1973). The agent is often identified as the manager and the principal as shareholder. If the interests of both parties are aligned then no problem exists. “Agency costs arise when the interests of the firm’s managers are not aligned with those of the firm’s owners” (Ang, 2000: 83). The agent does not always act in the best interest of the principal which causes the entering of monitoring and bonding costs to have some divergence in the interest of the principal and agent (Jensen & Mecking, 1976). The theory tries to resolve two problems that might occur in agency relationships (Eisenhardt, 1989): the first is the problem of a goal conflict between the principal and agent where it is hard to verify what the agent is doing; and second, the principal and agent share the same risks while both deal with this differently. According to Ang (2000) firms of which managers are the only owners, and thus own 100 percent of the firm, have no agency cost by definition. The reason for this is that no principal-agent relationship exists. On the other extreme end are firms whose managers are employees without an equity stake in the firm. Here a principal-agent relationship does exist and thus agency costs arise (Ang, 2000). This might be a shareholder firm in which the manager is just a paid employee.

15

shareholders, and thus managers do not bribe or bribe only at a very low rate. Besides this, shareholder owned firms have higher levels of transparency often because they have to comply with corporate governance codes, creating less opportunities to engage in opportunistic behavior and engage in corrupt behaviour. However, when a firms ownership structure is a partnership, it might experience intermediate levels of corruption. A partnership is an agreement between two or more persons in which business is done to derive a common goal where multiple owners have a personal liability for the actions of their partner (Greenwood & Empsen, 2003). In partnerships the risk exists that partners engage in self-seeking opportunistic behavior at the cost of the other partner and the question that arises here is how this can be monitored (Child, Faulkner & Tallman, 2005). The relationship between partners that are the managing-owners as well, which is often the case in emerging markets, can be characterized as a principal-principal relationships. This relationship is a problematic one (Child et al, 2005), as the owners have different goals and incentives. In relation to corruption, it can be expected that partnerships are more difficult to monitor if goals are not aligned, and therefore partners might engage easily in corrupt behavior to expedite matters as compared to shareholder ownerships. This leads to the following hypothesis:

H2: Single owned firms pay the most bribes to government officials, partnership owned firms

next, and publicly traded shareholding firms pay the least bribes.

2.4.2.1 Foreign ownership

16

get things done. The findings of Gueorguiev & Malesky (2012) demonstrate that foreign investment enterprises are less likely to pay bribes compared to domestic firms, although both type of firms do pay bribes when registering their operations and when securing government contracts.

To summarize, foreign firms pay bribes just as domestic owned firms, however it is expected that foreign owned firms bribe less compared to domestic owned firms. While differences might exist in bribe payments per ownership type we will first test for the relationship between foreign ownership and the act of bribery, the interaction in chapter 5 will elaborate more on this matter. The above leads to the following hypothesis:

H3: Foreign ownership is negatively related to bribery

3. Methodology

Data from The Business Environment and Enterprise Performance Surveys (BEEPS) will be used to test the hypotheses1. The BEEPS is a jointly conducted initiative by the World Bank and the European Bank for Reconstruction and Development, also known as the Enterprise Surveys. It provides company-level data of emerging and developing economies and has been conducted by the World Bank since 2002. One of the key questions on bribery in this survey is: “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 percent 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?“ (Enterprise

Surveys, Questionnaire Codebook, 2012 ).

The survey questions are answered by business owners and managers. The number of surveys conducted depends usually on the size of the economy where a distinction is made between large, medium-sized and small economies. Only firms that are formally registered and have five or more employees are able to participate. Furthermore, firms that are fully owned by the state are not targeted for the survey. The Enterprise survey consist of a manufacturing and service module, which differentiates the survey by having questions only applicable for

17

manufacturing (or service) firms. However, the majority part of the questions are the same for the different industries. Besides measuring corruption, the survey offers data on trade, crime, taxation, innovation, technology, performance, finance, sales etc. Some parts of the questions asked to firm owners or managers are highly sensitive, therefore, private contractors instead of government workers are used to collect the data and to assure that participants give more reliable and honest answers. (Enterprise Surveys Methodology, World Bank)

The methodology used to collect the data is stratified random sampling, which means that the population unit can be grouped or divided within homogeneous groups and simple random samples can be taken within each group (Enterprise Surveys, World Bank; Blumberg, Schindler & Cooper 2011). This is different from simple random sampling, as members of the population have the same probability of being selected and there is no distinction made between different groups that are overly present in a population. An advantage of stratified random sampling is that the estimations may be more precise. The sample must be large enough in order to be representative, and it must bear some proportional relationship to the size of the population from which it is drawn (Blumberg et al, 2011), the survey does this by conducting more surveys in large economies than in smaller ones. The sample size for each country is stratified by industry, firm size, and geographic region which ranges from 150 – 1320 per industry. Business owners and top managers from manufacturing and service sectors are participating in the survey. A disadvantage of the Enterprise Surveys is that the methodology used is not consistent across all countries.

3.2 Countries

The Enterprise Surveys offers data on a total of 135 countries. For this study, the choice is made to focus on Cambodia, Vietnam and Lao People’s Democratic Republic (PDR), better known as Laos. These three countries are classified as being corrupt and the following scores are given by the CPI to Cambodia, Laos and Vietnam respectively: 22, 21 and 31, which are high corruption rates. The countries are located in one of the fastest growing regions of the economy, and therefore studying the effects of corruption in those countries makes it even more interesting. Cambodia, Laos and Vietnam are members of the Association of Southeast Asian Nations (ASEAN)2 consisting of ten countries with the goal of fostering economic, cultural and political collaborations. The member countries have high corruption levels, except for Singapore.

2

18

Furthermore, the countries experienced high growth rates in the past, and continue to expect economic growth. Cambodia’s economy have had growth rates of 10% per year between 1998 – 2008. The economic crisis have had its impact here as well, economic growth declined heavily after 2008. After 2010, the Cambodian economy started to grow again with 6% per year, which is expected to continue in the next years (World bank, 2013a). On average, the Laos economy expanded by 7.1% per year from the period 2001 – 2010, this growth rate is expected to be higher in the future (approximately 7.6% per year) (World bank, 2013b). Vietnam, on the other hand, is experiencing relatively stable macroeconomic conditions, and its growth rate is somewhat lower than that of Laos. The Vietnamese economy is expected to grow with 5.5% in 2013 (World bank, 2013c).

Laos and Vietnam are surveyed by the BEEPS in 2009, Cambodia in 2007. A total of 502 owners and managers were surveyed in Cambodia, 360 in Laos, and 1,053 were surveyed in Vietnam (Enterprise Surveys, The World Bank). This result in a total of 1915 firms/managers that participate in the survey.

3.3 Sample

19

3.4 Dependent variable

The Enterprise survey measures bribery on several questions. The most accurate question to measure corruption is the following 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?” Respondents have the opportunity to

refuse to answer the question or might not know the answer. Therefore, non-response is excluded from the estimation. Data derived from this question is transformed into a binary scale indicating whether a firm bribes or bribes not. Assuming that not all firms are open on the actual percentage of annual sales paid in informal payments or gifts, this transformation results in the most answers on the bribery variable.

3.5 Independent variables

Operationalization of the independent variables is necessary in order to measure the influence of firm age and ownership structures on the bribes paid by firms. The different ownership structures are single owned firms, partnerships, and shareholder owned firms (publicly traded). It is already acknowledged in previous research that private owned firms are more corrupt than government/state owned firms, and therefore firms that are fully owned by the government/state are taken out of the analysis for all ownership types. The reference group for the ownership types are all other categories including business permits, privately traded shareholding firms and other firms such as cooperation’s and FDI firms.

Firm age (H1): Firm age is measured with the question ‘In what year did the establishment begin

operations?’ The current age of a firm is calculated as the year 2009 (or 2007 in the case of Cambodia) minus the year of establishment. Here, no attention is being paid on whether or not the establishment is formally registered at their start since this does not influence the period of business operations.

Ownership (H2): Single ownership: This type is defined as sole proprietorship and single

20

Partnership: The second ownership type includes partnerships and limited partnerships. In a

partnership two or more individuals share profits and liabilities, the individual does not necessarily have to be a natural person (Enterprise Surveys, The World Bank, 2013).

Shareholder ownership: This variable includes shareholding companies with shares traded in the

stock market. Privately traded shareholding companies are owned by shareholders for whom their claims over the firm are not publicly traded (Enterprise Surveys, The World Bank, 2013). Privately traded companies might have opaque governance structures, have low transparency and disclosure requirements. Therefore, this ownership type cannot be taken together with publicly traded firms. Shareholding companies have in most cases controlling owners (>10% of the shares), which is as expected since the dominant ownership structure of large firms in emerging markets are families or the state. The concept of the widely held firm is less common in emerging economies.(La Porta, Lopez-de-Silanes & Shleifer, 1999). The corporate governance structure of the Asian economy is in line with the propositions of La Porta et al (1999) as the sample contains only firms with controlling shareholders (according to the 10% rule of La Porta et al (1999)).

Foreign ownership (H3): Foreign ownership is measured as the percentage of the firm owned by

private foreign individuals, companies and organizations.

3.6 Control variables

21

variable is included to control for corruption as the biggest obstacle. And finally, two country dummies are included to control for country differences across Cambodia, Laos and Vietnam. The three countries used in the analysis are not all from the same year, the Cambodian data is collected in 2007 while the other two questionnaires are conducted in 2009. Since the data covers only three countries a country dummy for Cambodia is included which controls for year and country effects. An overview of all variables, definition and measures can be found in table 1.

Table 1 Variables, Definitions and Measures

Type Variable3 Definition Measure

Dependent Bribery Whether a firm uses informal payments or gifts to get things done.

Binary variable, equals 1 if a firm bribes.

Independent Firm age 2009 or 2007 – year of establishment Continuous

Independent Single ownership A firm owned and operated by one individual Dummy: 1 if it equals single ownerships, 0 otherwise

Independent Partnership Two or more individuals sharing the same profits and liabilities

Dummy: 1 if it equals a partnership, 0 otherwise

Independent Shareholder ownership

Shareholding companies with publicly traded shares

Dummy:1 if it equals shareholder ownership, 0 otherwise

Independent Foreign % of foreign ownership Continuous

Control Export Average % of direct exports Continuous

Control Firm size The number of fulltime employees per firm Continuous

Control Business sector The firms major business sector Service = 1, non-service = 0

Control Contact with

officials

Average % of time spent per week on dealing with regulations imposed by public officials

Continuous

Control Obstacle Corruption indicated as the biggest obstacle Dummy (1 = Corruption, 0 = other)

Control Country Vietnam Whether the sample includes Vietnam Dummy ( 1 = Vietnam, 0 = other)

Control Country Cambodia4 Whether the sample includes Cambodia Dummy (1 = Cambodia, 0 = other)

3 _{Appendix 1 gives an overview of the survey questions.}

4 _{The country dummy for Cambodia is also a dummy to control for year effects. Since the sample consist of only}

22

3.7 Statistical model

The following model will be used to test the hypotheses:

Logit Pr(y=1) = β0 + β1X1 + β2D2 + β3D3 + β4D4 + β5X5 +βcXc + ε

Where, y indicates whether a firm bribes; β0 is the intercept; X1 measures firm age; D2 measures

single ownership; D3 measures partnerships; D4 measures shareholder ownership;; X5 is a

measure for foreign ownership. XC stands for the control variables including: export (XC1), firm

size (XC2), business sector (XC3), contact with public officials (XC4), a dummy indicating

corruption as the biggest obstacle (XC5),a country dummy Vietnam (XC6), and country dummy

Cambodia (XC6); and ε is the error term.

The dependent variable is a non-metric binominal variable i.e. a firm bribes or not. This makes a Logit regression an appropriate method to test the hypotheses. The regression will be performed in three steps: first, only control variables will be included; second, the control and independent variables will be included; and finally, a squared term is added to the equation to test for the non-linearity in the relationship between firm age and bribery.

3.8 Method assumptions

Ordinary Least Square analysis has several assumptions that must be satisfied in order to have the best linear unbiased estimates. Although a Logit analysis is employed in this study, the method assumptions that apply for ordinary least square regressions still apply for Logit regressions. The assumptions that are tested for are multicollinearity, normality, endogeneity and homoscedasticity. A violation of these assumption might create problems with the reliability of the results.

Homoscedasticity

Homoscedasticity assumes that the variance of the error term is constant and the same for all observations (i) (var(ei) = σ2). If this assumption is violated and the error variance for all

23

lead to biased coefficients but makes the estimator no longer the best with the smallest variance (Hill, Griffiths & Lim, 2009). The problem with heteroscedasticity 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 heteroscedasticity is detected, there are methodological ways to correct for the weight of larger error variances to get estimates with the smallest sum of squared errors. In the sample, heteroscedasticity could be expected in the firm size and age. Smaller firms might less easily resist bribe request than larger firms do, the same might be expected for firm age. Younger firms might have less resources to resist bribes compared to older firms. This may lead to differences in error variances. Figure 1A and 1B in Appendix 2 plots the residuals, which estimates the errors, against firm age and firm size. From the graph it can be concluded that heteroscedasticity does not seem to be a problem as the variance is approximately similar for smaller and larger firms, and for young and older firms. There are more observations for smaller and younger firms which might make the graphs appear slightly heteroscedastic, however the residual variance is similar for small- and large and young- and older firms.

A Breusch-Pagan / Cook – Weisberg test for homoscedasticity is performed to test whether or not heteroscedasticity is a problem. From table 2 and 3 can be concluded that heteroscedasticity is not an issue. Since the sample covers only one time period (2007 or 2009), it is unfeasible to test for autocorrelation.

Endogeneity

Here it is assumed that the error term is uncorrelated with the independent variables. If this assumption is violated such that xi is correlated with an unmeasured variables (ei), than the effect

Table 2

Breusch-Pagan / Cook-Weisberg test for heteroscedasticity

H0: Constant variance Variables: Firm size chi2(1) = 0.13 Prob > chi2 = 0.7177

Table 3

Breusch-Pagan / Cook-Weisberg test for heteroscedasticity

24

of xi on y will be consistently overestimated. Consequently least square estimates are

inconsistent and do not converge even in large samples. In addition, none of the hypothesis test procedures are valid (Hill et al., 2009: 272). There is no simple numerical or statistical test for this assumption; hence it has to be satisfied theoretically5. Two-stage least-squares (2SLS) regression with instrumental variables is one option to test for and overcome the potential problem of endogeneity, generalized least squares (GLS) is another alternative. From a theoretical perspective there is little reason to assume firm age and size to be correlated with unobserved variables in the error term as it can be easily estimated and observed by management. If endogeneity exists, then it is more likely that indicators that are based on respondent’s judgment are correlated with the error term and could thus be influenced by unobserved variables. However, such indicators are not present as independent variables in the sample. If concepts cannot be observed directly, instrumental variables are commonly used to measure those latent constructs. Since the common tests for endogeneity rely on instrumental variables, it is difficult to prepare such a test

in the present sample as such additional variables are not available. The exogenous nature of the independent variable has thus to be assumed.

Multicollinearity

Another assumption is that the independent variables are not perfectly correlated such that “the values of xik are not exact linear

functions of other explanatory variables.” (Hill et al., 2009: 154). If this assumption is violated, variables are said to be collinear, which makes it difficult to isolate the relationship between variables. In this case, relatively imprecise information about the

5 _{The sample residuals will always be uncorrelated to the independent variables, so using those to test for}

endogeneity will not be appropriate (Hill et al., 2009).

Table 4

Test for Multicollinearity

Variable VIF 1/VIF

25

unknown parameters will be obtained which is due to large standard errors. This causes difficulties in predicting the true parameters, an issue which might even become more problematic when there is little variation in the explanatory variables. This latter instance could in fact be present as some of the sample variables originate from categorical data. To test for the presence of multicollinearity the variance inflation factor is calculated (VIF) (Table 4) which is an index, estimating how much the variance of an estimated regression coefficient is inflated due to collinearity. The values indicate that multicollinearity is not a problem as all variables have values well below the cut-off value of 10 recommended by Neter, Wasserman & Nachtsheim (1985) (See also Table 7 in appendix 2 which provides the correlation values, showing that there is no problematic correlation among the independent variables).

Normality

The final test examines whether the normality assumption is satisfied. Normality assumes that the values of the error term are normally distributed about their mean. A violation of the normality assumption might lead to biased p-values and thus affects the test for statistical significance (Hill et al. 2009). Following a graphical analysis from Figure 2 (Appendix 2), which plots the residuals against the normal distribution (straight line), can be concluded that the residuals are approximately normal distributed. If the normality assumption has been violated, an alternative to solve the potential problem of non-normality was to transform the dependent variable.

26

4. Empirical results

Table 5 below gives an overview of the descriptive statistics. As can be seen from the table, the sample consist of several dummy variables which are less informative in this table. To get more insight in these dummy variables a frequency table can be found in Appendix 2 (table 6). From the total sample, 400 firms indicate that they pay bribes to government officials. Of these 1272 firms there are 127 firms who argue that corruption is the biggest obstacle of doing business in these countries. The independent variable firm age ranges from 0 to 56 years of doing business with an average of 10 years. The size of the firms in the sample varies from small firms with a minimum of 5 to large firms with 3994 fulltime employees6. The average firm size is 162

employees.

Foreign ownership is a continuous variable which ranges from 0 foreign ownership to a value of 100%, which means that a firm is fully owned by foreign individuals. Firms have on average 19% foreign ownership. Another control variable is the time spent per week with government officials, this value ranges to 90% which is quite high. An explanation for this high value is that many firms might do business with the government and therefore have to deal with their

6 _{Firms with less than five employees are excluded from participation in the Enterprise Survey.}

Mean Std. Dev. Min Max

Dependent variable Bribe 0.314 0.464 0.0 1 Independent variables Single ownership 0.575 0.495 0.0 1 Partnership 0.233 0.423 0.0 1 Shareholder ownership 0.017 0.130 0.0 1 Firm age 10.149 8.421 0.0 56 Foreign ownership 18.770 37.858 0.0 100 Control variables Firm size 162.297 370.243 5.0 3994 Official 4.590 11.411 0.0 90 Obstacle 0.100 0.300 0.0 1 Sector 0.193 0.395 0.0 1 Cambodia 0.316 0.465 0.0 1 Vietnam 0.436 0.496 0.0 1 Export 14.844 33.657 0.0 100

27

requirements a lot. On average firms are spending 5% of their time per week in dealing with officials. The percentage of direct exports is included as a control variable and is on average 15%. The absolute value ranges from firms that do not export, and thus sell their products in their home country, to firms that only export their products to other countries. From the independent ownership variables it can be seen that in total 57% of the firms from Cambodia, Laos and Vietnam have as their ownership structure a single ownership (100% ownership by one person), 23% of the firms are in a partnership, and 22 (2%) firms are shareholding firms with publicly traded shares. In total, 20% of the firms are active in the service sector while the other firms in the sample are doing business in the manufacturing sector. Correlations (table 7) between the independent variables are rather low with higher values for partnerships and single ownership. High correlations exist between the two country dummies for Cambodia and Vietnam. Furthermore, moderate positive correlations exist between Cambodia and bribe, and Cambodia and corruption as the biggest obstacle in doing business. The correlations are not causing severe problems in the data.

4.2 Regression results

28

from Laos in their probability to bribe, and while these countries are all neighbor countries with high growth rates they cannot be seen as one entirety.

Model 2 includes all independent variables except the squared firm age variable. Adding the independent variables to the regression model causes the Pseudo-R squared and Chi² to increase, indicating a good fit of the model. In Model 3 the variable firm age squared is added to the equation to test for the non-linear relationship between firm age and bribery. Here, the Pseudo-R squared and Chi² do not change, implicating that the model does not change significantly by adding a squared term to one of the variables. From the results can be seen that the coefficient of Firm age² is positive, which means that firm age and bribery tend towards a u-shaped relationship. However, the results are insignificant (β = 0.00888; n.s.), therefore we reject hypothesis 1 that there is a u-shaped relationship between firm age and bribery. Furthermore, the negative relationship between firm age and bribery that is shown to be significant in other studies (e.g. Kasuga, 2013; Kuncoro, 2004) is not seen in the current sample either. Model 2 even shows a positive insignificant relationship between age and corruption (β = 0.000727; n.s.). This insignificant result might be caused by limitations in the data.

The three different ownership variables are added in Model 2. Single ownership shows a positive relationship to the incidence of bribing, this is what was excepted from an agency perspective as well. Nevertheless, this relationship is not significant in both model 2 and 3. Shareholder ownership shows a negative relation to the probability of bribing, the coefficient is significant in model 2 and 3 (β = -1.952; P < 0.10). This is expected from an agency theory point of view since shareholding firms have a stronger principal-agent relationship and thus can be better monitored. However, as these shareholding firms are publicly traded, a reputation effect might influence these results as well. The third ownership variable, partnership, is negative but not significant in both model 2 and 3. The coefficient falls between single- and shareholder-owned firms. Hypothesis 2 is only partly supported, the only significant result is from shareholder ownership, while the other ownership structures are of the expected sign only.

29

discover whether a relationship between bribe payments and foreign ownership as a moderator of the three ownership structures exist.

A final look at the Average Marginal Effects (Appendix 3), which shows the predicted probability of bribing holding all other values constant, shows that the predicted probability to bribe in a single ownership firm is 1,5% greater compared to firms that are not in a single ownership. Having a shareholding owned company decreases the probability of bribing with 23.4% compared to the other ownership structures. Likewise, being in a partnership, decreases the probability of bribing by 2.10%. The continuous variables can be interpreted as following, a one unit change in foreign ownership increases the probability of bribing by 0.009%. Similarly, firm age increases the probability of bribing by 0.01%. However a squared term is used for firm age as we test for a u-shaped relationship between firm age and bribing, from the average marginal effects table we can see that firm age squared gives a very low value for the probability of bribing. Furthermore, the marginal effects show that firms from Cambodia bribe much more than firms from Laos and Vietnam.

30 Standard errors in parentheses

+

p < 0.10, * p < 0.05, **p < 0.01, *** p < 0.001

Table 8 Logit Regession

Model 1 Model 2 Model 3

31

5. Robustness tests

Robustness tests are necessary in order to make any valid causal inferences from the regression models. As a robustness check the same regression will be performed in a different model, which will be a Probit regression. Furthermore, in section 5.1 an additional robustness test is performed in which the variable foreign ownership is added as an interaction term of the second hypothesis. From a theoretical perspective it is valuable to test for this relationship, besides it will give more insight into ownership as a justification of bribe payments.

The Logit regression performed in the previous section is replaced by a Probit model, this Probit regression gives the same results as the Logit regression performed earlier. As can be seen from table 9 in appendix 4, shareholding ownership is now significant at a level of p < 0.05. The coefficients of the other independent variables stay more or less the same. Implementing the statistical model into a different method give similar regression results, implying that the results are robust. However, more insight in the relationship between ownership structures and the incidence of bribery is needed to make valid conclusions since only one of the hypotheses is partially supported, this is done in the following paragraph.

5.1 Additional analysis of bribery

An additional analysis is performed to test whether an interaction exist between the three different ownership types and foreign ownership. “An interaction effect is said to exist when the effect of an independent variable on a dependent variable differs depending on the value of a third variable, commonly called a “moderator variable”” (Jaccard, 2001:12). In this study, the moderator variable is ownership by foreigners. The interaction that is tested for is one between a dummy and a continuous variable. To successfully test the moderating effect, the continuous moderator variable is standardized and mean centered as suggested by Jaccard (2001). First, the reason to test for an interaction will be explained from a theoretical point of view. Secondly, the interaction results will be analyzed.

32

institutional isomorphism can be identified: “1) coercive isomorphism that stems from political influence and the problem of legitimacy; 2) mimetic isomorphism resulting from standard responses to uncertainty; and 3) normative isomorphism, associated with professionalization” (DiMaggio and Powell, 1983:150). Mimetic isomorphism results from uncertainty in the business environment. And firms investing in countries where corruption is the norm while doing business might mimic their competitors and adopt similar corrupt practices. It is not only mimetic isomorphism that influences foreign firms, coercive pressures might also play a large role in corrupt behavior among firms. Firms can face formal and informal pressures exerted by other organizations on which they are dependent (DiMaggio & Powell, 1983), the former results in coercive isomorphism. Normative isomorphism captures the influence of headquarters on local subsidiaries (Venard, 2009), as well as the intervention and influence exerted by government officials (Vernard & Hanafi, 2008). From the above we can conclude that all three isomorphic pressures influence the behavior of foreign owned single ownership firms. On the other hand, these isomorphic pressures are to a lesser degree present in case of a partnership or shareholding ownership structure. Here, foreign owned firms have a local partner that understands the local way of doing business and foreign owned firms can stick to their local norms and values and are not, or to a lesser extent, subject to mimetic and coercive pressures. Thus, the interaction effect between single and foreign ownership is expected to be positively related to bribe payments, the interaction between partnerships and foreign ownership is expected to be negatively related to bribe payments, the latter relationship is expected for shareholding firms as well.

33

foreign ownership is significant and negative (β = -0.0171; p < 0.001). Thus, foreign ownership when interacting with a partnership strengthens its negative relationship with bribe payments. Model 4b shows the interaction of single ownership and foreign ownership, this interaction is significant and positive (β = 0.0109; p < 0.01). This significant coefficient reveals that the relationship between single ownership and bribe payments becomes stronger when foreign firms are involved. The third interaction, between shareholder ownership and foreign ownership, is

shown in Model 4c. This interaction is positive but not significant (β = 0.221; n.s.), the variable shareholder ownership is insignificant in Model 4c as well. The coefficient itself and the standard error takes an anomalous value. An explanation for this is that the data sample is too small to make any valid conclusions for the interaction between shareholding ownership and foreign ownership. Figure 3 displays the two significant interactions on the probability of bribing.

The interaction results are confirmed by the theory as well, since single owned firms face isomorphic pressures to adapt to the local business environment in order to survive. Furthermore, in their ownership decisions, foreign owned firms make a choice between joint ownership or single ownership (Cui & Jiang, 2012). This choice depends on the strength of institutional pressures (i.e. corrupt practices). When government intervention, regulatory restrictions and normative pressures are high firms often choose a joint ownership structure (Cui & Jiang, 2012) e.g. a partnership in this study. The former implies that isomorphic pressures cause firms to adapt to local business practices and thus pay bribes when necessary. While on the other hand foreign

Figure 3 Graph interaction

.1 5 .2 .2 5 .3 Pr(B ri b e ) 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 Foreign ownership

Interaction Partnership - Foreign ownership

.3 2 .3 3 .3 4 .3 5 .3 6 .3 7 Pr(B ri b e ) 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 Foreign ownership

34

firms engaging in a partnership pay less bribes as they have a partner with knowledge of the local business practices which is in line with our interaction results.

Standard errors in parentheses

+

p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001 * Foreign ownership is mean centered in model 4 – 5

Note: Control variables are included in model 3 – 5

Model 3 Model 4a Model 4b Model 4c Model 5

Single 0.0882 0.0936 0.000606 0.0898 0.0459 Ownership (0.193) (0.197) (0.196) (0.193) (0.199) Shareholder -1.952+ -2.000+ -1.852+ -18.71 -17.90 Ownership (1.068) (1.083) (1.063) (1580.1) (1256.6) Partnership -0.123 -0.0652 -0.142 -0.120 -0.0907 (0.220) (0.223) (0.218) (0.219) (0.222) Firm age 0.000390 0.00368 0.00523 0.000312 0.00536 (0.0218) (0.0219) (0.0220) (0.0218) (0.0220) Firm age² 0.00000888 -0.000000336 -0.000113 0.0000204 -0.0000505 (0.000522) (0.000524) (0.000528) (0.000522) (0.000526) Foreign Ownership* 0.000536 0.00328 -0.00638* 0.000298 -0.00145 (0.00213) (0.00227) (0.00319) (0.00215) (0.00407) Interaction Partnership X -0.0171*** -0.0130* Foreign ownership (0.00486) (0.00571) Single ownership 0.0109** 0.00592 X Foreign ownership (0.00373) (0.00448) Shareholder ownership 0.221 0.213 X Foreign ownership (19.45) (15.47) Constant -1.818*** -1.824*** -1.728*** -1.814*** -1.784*** (0.311) (0.311) (0.309) (0.307) (0.312) Observations 1272 1272 1272 1272 1272 Pseudo R² 0.167 0.175 0.172 0.168 0.178 Log Likelihood -659.8 -653.1 -655.4 -658.8 -651.4 Chi2 264.4*** 277.8*** 273.1*** 266.4*** 281.2***

35

6. Conclusion

Prior literature emphasized both the negative and positive effects of corruption. Unfortunately corruption increases the costs of doing business for many firms and bribes remain common in many countries around the world: a total of 24.8% of all firms in the world surveyed by the World Bank are expected to give gifts to public officials to get things done (Enterprise Survey, 2013). This suggests that it is a severe problem around the world.

The objective of this study was to enhance the literature on the effect of firm level characteristics on bribery, especially the impact of firm age and different ownership structures on the incidence of bribery. To test the hypotheses a database consisting of Cambodia, Laos and Vietnam was used; three South East Asian counties that have a corrupt business environment and high economic growth rates.

36

looked at the relationship between foreign ownership and bribery. A negative relationship was expected but not found. When we compare this result with other studies in the corruption literature, it can be seen that there are many differences in the positive or negative effect of foreign ownership on corruption. Nonetheless, as the robustness test shows, foreign ownership is a significant moderator in the relationship between bribe payments and ownership structures. To conclude, the relationship between ownership and bribery is not as strong as expected in the current sample. Indicating that differences in bribe payments among firms might not be explained much by different ownership structures. However, the interaction of the different ownership structures with foreign ownership explains more differences in bribery.

6.1 Added value

The major contribution of this study is that it is one of the first studies at the firm-level that delves deeper into different ownership structures as an explanation of why some firms bribe more than others. Existing studies seem to focus on ownership as the difference in state, foreign and domestic ownership (e.g. Lee et al, 2010; Jeong & Weiner, 2012). The present study has its main focus on three different ownership structures i.e. single ownership, partnerships, and shareholder ownership. For that reason this is one of the first studies that examines the relationship between ownership structures and bribe payments. Unfortunately, only partial support for the ownership structure hypothesis (H2) was found. We did however find interesting results in the interaction between foreign ownership and the three ownership structures in relation to the act of bribing. Two out of the three interactions are significant and of the expected sign, implying that foreign firms behave significantly differently which might be due to differences in isomorphic pressures that arise in the different ownership structures.

37

The findings of this study have several strategic implications for managers and policymakers. Firstly, for managers who want to invest in foreign countries characterized as corrupt it provides an indication of bribe requests it might expect and thus managers are informed of what they can expect in advance. Managers can more easily overcome the hazards of bribing when engaging in a partnership. Firms that solely operate in a foreign country might pay more bribes as they do not have the knowledge nor the network to avoid corrupt behavior. Consequently, single owned firms face more pressure to engage in deviant forms of behavior and to mimic other firms. This study informs managers so they might be better able to anticipate corrupt behavior in their firms and bribe requests. Second, the results give managers an indication of which kind of firms pay bribes to get things done and that can expect more bribe requests from government officials. Another implication for managers or boards of directors is that agents must be monitored or controlled in order to keep bribe rates low. For managers of firms that want to invest in a corrupt country the results give an indication of which entry strategy to choose. As the results for foreign ownership in relation to partnership and single ownership are significant, firms that do not want to engage in corrupt behavior might be better off choosing a partnership when investing abroad. On the other hand, firms without a local partner pay more bribes and might less easily survive without adapting to local isomorphic pressures.

For policy makers two key implication exists. Firstly, governments that want to reduce bribery in their country have a better overview of which type of firms, based on their ownership structure, pay bribes. Secondly, the results give policy makers an indication of which type of firms face more problems with government officials. In combating corruption, policy makers can exert more control over these officials and firms.

6.2 Limitations and further research

38

but corruption rates might be different nowadays. Thirdly, the sample does not contains enough publicly traded shareholding companies which causes a large error in the interaction effects. Fourth, there are limitations in the use of data collected by the World Bank as the enterprise survey contains several sensitive questions that many firms might give a false response or non-response for. While the missing data was missing randomly, it decreased the number of data in the final dataset that was used to test the hypotheses heavily. Additionally using a Logit model limits the measurement of bribery to whether a firms bribes or not. No inferences can be made on the actual size or percentage of the bribe payments.