Institutional environment and bribery involvement
impacts on firm performance:
Central and Eastern European firm-level evidence
International Business and Management 2014-2015 · Master Thesis (EBM719A20) Faculty of Economics and Business · University of Groningen
By C.E. Croes
S2378884 · Folkingestraat 38A · 9711JX Groningen 06 106 555 39 · c.e.croes@student.rug.nl
Institutional environment and bribery involvement impacts on firm performance: Central and Eastern European firm-level evidence
By C.E. Croes, Faculty of Economics and Business, University of Groningen January 2015
Abstract
Prior research has predominantly focused on the macro-level of corruption. Yet, the link between country- and firm-level corruption remains unclear. Additionally, not much is known about whether briberies actually pay off. For that reason, the institutional theory was applied in examining the influence of the institutional environment and firm bribery involvement on firm performance. A logistic regression was conducted on 1796 firms spread over 13 Central and Eastern European countries. Findings indicate that participation in bribery increased the likelihood of a higher firm performance. Furthermore, initial practical significance point toward positive interactions between power distance and individualism in relation with bribery involvement on firm performance. The opposite holds for masculinity and uncertainty avoidance.
TABLE OF CONTENTS
1. INTRODUCTION 5
2. THEORETICAL BACKGROUND 7
2.1.BRIBERY AS A MANIFESTATION OF CORRUPTION:A MACRO-PERSPECTIVE 7
2.1.1. The presence of corruption 7
2.1.2. The role of institutions in corruption 8
2.2.BRIBERY AS A MANIFESTATION OF CORRUPTION:A MICRO-PERSPECTIVE 9
2.2.1. The need for bribing 9
2.2.2. The antecedents and consequences of bribery 11
3. HYPOTHESES 14
3.1.INSTITUTIONAL PRESSURES, BRIBERY, AND FIRM PERFORMANCE 14
3.1.1. Regulative and normative pressures 15
3.1.2. Cultural-cognitive pressures 17
3.2.CONCEPTUAL MODEL 19
4. DATA AND METHOD 20
4.1.DATA 20 4.2.SAMPLE 21 4.3.VARIABLES 22 4.3.1. Dependent variable 22 4.3.2. Independent variables 23 4.3.3. Control variables 24 4.4.METHOD OF ANALYSIS 24 4.4.1. Preliminary analysis 25 4.4.2 Assumptions 27 5. EMPIRICAL RESULTS 29 5.1.DESCRIPTIVE STATISTICS 29 5.2.BASELINE RESULTS 30
5.2.1. Control variables effects 31
5.2.2. Independent variables effects 33
5.2.3. Interaction effects 33
5.3.ROBUSTNESS TESTS AND EXTENSIONS 35
5.3.1. Results with whole sample 35
5.3.2. Results with alternative bribery involvement measures 35
5.3.3. Results with alternative dependent variables 36
7. LIMITATIONS 42 8. CONCLUSION 44 APPENDIX 46 A. BEEPS2002 46 B. SAMPLE CHARACTERISTICS 47 C. FACTOR ANALYSIS 49
D. ROBUSTNESS TESTS AND EXTENSIONS 51
LIST OF TABLES
Table 1 Countries in sample. ... 22
Table 2 List of variables. ... 25
Table 3 Correlations, means, and SDs. ... 27
Table 4 Breusch-Pagan test for heteroscedasticity. ... 28
Table 5 Collinearity statistics. ... 28
Table 6 Descriptive statistics of variables. ... 30
Table 7 Logistic regression results. ... 32
Table 8 BEEPS 2002 relevant questions. ... 46
Table 9 Macro indicators per country. ... 47
Table 10 Frequencies of categorical variables. ... 47
Table 11 Descriptive statistics bribery questions BEEPS 2002. ... 48
Table 12 Component Loadings. ... 50
Table 13 Alternative logistic regression results: whole sample. ... 51
Table 14 Alternative logistic regression results: model AIV1. ... 52
Table 15 Alternative logistic regression results: model AIV2. ... 53
Table 16 Alternative logistic regression results: model AIV3. ... 54
Table 17 Alternative logistic regression results: model AIV4. ... 55
Table 18 Alternative logistic regression results: model AIV5. ... 56
Table 19 Alternative logistic regression results: model ADV1. ... 57
1. INTRODUCTION
Corruption is both pervasive and significant around the world (Shleifer & Vishny, 1993). A vast and growing body of literature on corruption reflects the complexity, multidimensionality, and importance of the issue (Sahakyan & Stiegert, 2012). This body of literature relies predominantly on macro data (Collins, Uhlenbruck, & Rodriguez, 2009; Jensen, Li, & Rahman, 2010; Mellahi, Demirbag, & Wood, 2012), and these typically reflect the consensus that corruption is harmful for a society (e.g. Cuervo-Cazurra, 2006; Ertimi & Saeh, 2013; Mauro, 1995; Zhao, Kim, & Du, 2003). It is generally condemned as a social evil (Sanyal & Samanta, 2004; Zhao et al., 2003), and seen as legally wrong, morally wrong and economically indecent (Ertimi & Saeh, 2013).
Conversely, some studies suggest that corruption can have postive effects (e.g. Cuervo-Cazurra, 2006; Leff, 1964; Méon & Weill, 2010), or that it is in some cases even necessary (Méon & Weill, 2010; Tanzi, 1998). For instance, the growth of international trade and business has created many situations in which the payment of bribes may be highly beneficial to the companies that pay them by giving them access to profitable contracts over competitors (Tanzi, 1998; Zhao et al., 2003), or to help the process in the presence of slow bureaucracy (Mauro, 1998). Micro-level studies, however, fail to give an unequivocal answer as to how corruption affects the growth and development of firms (Sahakyan & Stiegert, 2012). While the determinants of bribery have begun to be uncovered (Lee & Weng, 2013), little is known about whether briberies actually pay-off and help firms grow (Zhou & Peng, 2012) and, if so, under what conditions.
The purpose of this thesis is to further advance the understanding of this phenomenon and its strategic implications at the firm-level, as was appealed by Lee & Weng (2013) and Seleim & Bontis (2009). This study also extends outlooks set forth by Méon & Sekkat (2005) and Zhou & Peng (2012). It furthermore addresses the willingness of individuals and organizations to give bribes (Davis & Ruhe, 2003). As such, the formal research question is formulated as; to what extent do the institutional contexts of firms and their bribery involvement influence their performance?
operations, firms might better evaluate their domestic markets in view of cultural and institutional forces that may impede efforts to conduct business ethically. This is important as managers decisions are affected by influences that precede from the macro-environment, form firm and industry norms, as well as from their personal relationships and experiences (Collins et al., 2009). Additionally, recognizing for what reasons and where bribery is committed might assist the fight against corruption on an organizational level. Knowing the true reason behind such actions might help formulate better tactics to overcome it.
Data as provided by a survey conducted by the World Bank will help ascertain the propositions set forth. This firm-level data covers several characteristics and microeconomic performance of a large sample of firms operational in several Central and Eastern European countries. This will further be supported by a World Governance Indicator and Hofstede‟s cultural dimensions.
2. THEORETICAL BACKGROUND
This section sets forth an impression of extant literature of corruption, both from the macro-perspective and micro-macro-perspective. The former deals with corruption on the country level and emphasizes its relation with institutions. The latter provides a firm-level view of the role of bribery and its known antecedents and consequences.
2.1. Bribery as a manifestation of corruption: A macro-perspective
2.1.1. The presence of corruption
What is perhaps most puzzling about corruption is that it persists and flourishes even where it is universally decried (Collins et al., 2009). International corruption has continuously existed in many societies (Ertimi & Saeh, 2013), and has existed in multiple forms for several years (Beets, 2005). Such forms include within or across the hierarchical structure of government agencies, between government agencies and private sector, and across national borders (Sanyal & Samanta, 2004). As such, the presence of corruption varies from nation to nation (Davis & Ruhe, 2003). There are many participants in corruption; all cases of corruption in the public sector include some combination of public officials, individuals, firms, or industries on one side and public officials on the other (Eicher, 2009).
As a cultural, political and economic phenomenon it has attracted a lot of public attention around the world in recent years (Husted, 1999, Zhao et al., 2003). There are several explanatory reasons for the increased degree of attention in the past decades as put forward by Tanzi (1998). For instance, this is due to the growing role of non-governmental organizations in publicizing the problems of corruption and in trying to create anticorruption movements in many countries (Cuervo-Cazurra, 2006; Eicher, 2009; Tanzi, 1998). Moreover, this new public awareness of corruption can also be attributed to globalization; it has created visible, open clashes among private actors, public officials, individuals, and organized groups at an unprecedented level and pace (Eicher, 2009).
However, fighting corruption is very difficult because it is a multifaceted social phenomenon that penetrates horizontally and vertically in many aspects of societies (Seleim & Bontis, 2009). One of the key issues of this complex phenomenon is that it is difficult to measure (Mauro, 1998; Svensson, 2005), due to the different faces and its secretive nature (Ertimi & Saeh, 2013; Sahakyan & Stiegert, 2012). Many have also argued that another issue in combating corruption lies in the struggle with defining it (Davis & Ruhe, 2003; Eicher, 2009; Ertimi & Saeh, 2013). There are numerous definitions of corruption in academic literature and among donor agencies; most are quite broad and in some cases vague (Blahojevic & Damijan, 2013). Notwithstanding, the most influential strand of literature often defines corruption as the misuse of public office for private gain (Cuervo-Cazurra, 2006; Jensen, Li, & Rahman, 2010; Sahakyan & Stiegert, 2012; Sanyal & Samanta, 2004; Svensson, 2005; Tanzi, 1998; Lambsdorff, 2007; Treisman, 2007). Grouped within this phenomenon is a wide variety of unethical behaviors, such as bribery, fraud, kickbacks, economic espionage, extortion, and embezzlement to name a few (see Eicher, 2009; Jensen et al., 2010; Lambsdorff, 2007; Svensson, 2005).
2.1.2. The role of institutions in corruption
Several studies have been conducted on the influences of the environment – institutions as determinants of corruption (see Beets, 2005; Collins et al., 2009; Eicher, 2009; Jensen et al, 2010; Mauro, 1995; Méon & Sekkat, 2005; Méon & Weill, 2010; Svensson, 2005; Tanzi, 1998). These studies all advocate in general that the environment of the firm plays an important role in this phenomenon.
It has been stated that a society‟s strengths or roots of corruption, is dependent upon the level of development and how the following factors operate within the society; (1) political networks and practices; (2) economic incentives and rules governing the market; (3) state legitimacy; (4) effectiveness of governance; (5) rule of law; (6) social capital, and (7) cultural, moral, and ethical orientation (Eicher, 2009). Thus, it is when the above mentioned factors are poorly governed – when the institutions are weak – that a corrupt society persists.
create a fertile ground for corruption. In other words, poorer countries may have limited financial resources to enforce laws, including those related to corruption (Beets, 2005).
A cultural profile of a corrupt country has been described as a country in which there is high uncertainty avoidance, high masculinity, and high power distance (Husted, 1999). Similarly, Davis & Ruhe (2003), studies the relationship between Hofstede‟s four major variables, the perceptions of corruption, and several economic performance ratings. As such, it is determined that corruption seems to be predictable based on high levels of power distance, masculinity, and collectivism. Seleim & Bontis (2009) also conclude that individual collectivism practices and humane oriented practices encourage corrupted practices, and that uncertainty avoidance values increase levels of corruption. Thus, in certain cases, participants in acts of corruption may be influenced by cultures and cultural value systems that may endorse, rather than condiment such activities (Beets, 2005).
Studies such as that of Mellahi, Demirbag, & Wood, (2012), have found evidence to suggest that regulations in highly corrupt countries not only mitigate corruption but also create an environment that may cause and increase in corruption. It is further suggested that particular regulations create fertile ground for more interference by the government and hence more corruption. This reflects Mauro (1998) observation in that the ultimate source of rent-seeking behavior is the availability of rents; corruption is likely to occur where restrictions and government intervention lead to the presence of such excessive profits. Another interesting find, is that of Méon & Weill (2010). These authors find that corruption is less deterimental in countries where the rest of the instutional framework is weaker. Based on such findings they entertain a possible policy implication that countries plagued with ineffcient institutional framework may benefit from letting corruption grow.
2.2. Bribery as a manifestation of corruption: A micro-perspective
2.2.1. The need for bribing
should either provide for free or not provide at all – that is envisioned as the quintessential corrupt transaction (Lambsdorff, 2007; Treisman, 2007).
A corrupt transaction typically involves a supply side, i.e. the payer of the bribe, and a demand side, i.e. the recipient of the bribe (Beets, 2005), thus, bribery always involves two parties (Lee & Weng, 2013). The supply-side of bribes is demonstrated by the manipulation of business functions such as obtaining contracts, garnering favorable regulatory decisions, and other government or policy determinations (Martin et al. 2007). This illustrates that there can be many actors involved in corrupt transactions and that such may involve various combinations of firms, household and public officials (Blahojevic & Damijan, 2013; Eicher, 2009). Bribes act as a wedge between the price that suppliers are requesting and the price that customers are willing to pay (Lambsdorff, 2007). As such public services are therefore disproportionally allocated between bribers and non-bribers (Lee & Weng, 2013).
This manifestation of corruption (Sanyal & Samanta, 2004) is not new to most managers; kickback, red envelopes, and grease money are constantly reported by the press (Lee & Weng, 2013). It has been reasoned that these may allow the company to operate more profitably, at least in the short run (Eicher, 2009). This is evident in the claim of proponents of “efficient corruption” in that bribery may allow firms to get things done in an economy plagued by bureaucratic holdups (Fisman & Svenson, 2007; Mauro, 1995).
substantial costs to broader society (Collins et al., 2009). Another reason why corruption is costly is the distortions entailed by the necessary secrecy of corruption (Shleifer & Vishny, 1993). Those costs and the high costs of enforcing rules, relative to the available resources, as well as the reluctance of people in power to prosecute corruption acts (being corrupt themselves), allows such behaviors to persist (Sanyal & Samanta, 2004). In short, bribery collection by public officials and illegal payments by private businesses (Gaviria, 2002), is likely to persevere as public resources are crucial for firms (Lee & Weng, 2013).
2.2.2. The antecedents and consequences of bribery
Although micro-level data on corruption remains scarce (Seker & Yang, 2014), prior empirical research have found a wide variation in the determinants of bribery at the firm-level (e.g. Blahojevic & Damijan, 2013; Clarke & Xu, 2004; Martin et al., 2007; Sahakyan & Stiegert, 2012; Sanyal & Samanta, 2004), as well as, a wide variation in its effects (e.g Collins et al., 2009; Fisman & Svenson, 2007; Gaviria, 2002; Martin et al., 2007;Şeker & Yang, 2014; Wang & You, 2012).
It has been determined that firms whom are involved in corruption vary significantly from one another, based on factors such as their size, age, and number of competitors to name a few (e.g. Martin et al., 2007; Sahakyan & Stiegert, 2012; Sanyal & Samanta, 2004). Likewise, Martin et al. (2007) determine that perceived competitive intensity significantly and positively impacted the likelihood of a firm‟s engaging in bribery (Martin et al., 2007). It was also found that as firms aged or if the firm manager had a university degree, there was a lower likelihood that corruption would be viewed favorably (Sanyal & Samanta, 2004).
Literature such as that of Blahojevic & Damijan (2013), Clarke & Xu (2004), Martin et al. (2007), and Zhou & Peng (2012) has also set forth on determining how the environment might influence the bribe payments by firms. Blahojevic & Damijan (2013) base their research on the notion that the quality of the instutional system and corruption affect general economic development through microeconomic performance and that these effects are aggrevated when the institutional system interacts with different ownership structure of firms. Consequently, they find that private firms (domestic and foreign owned) are more involved in both informal payments and state capture.
Xu, 2004). In the same strand of literature, Zhou & Peng (2012) put forth that firms endogenously choose their level of bribery according to their environments and that the benefits and costs might differ for different types of bribes. As such, they find that the growth of large firms is less likely to be hurt than those of small firms. In addition, arguing that there are country level drivers as well as firm level drivers of bribery, Martin et al. (2007), find support for the influence on bribery activity of cultural values, including an achievement orientation, collectivism, and a humane orientation, and the complexities of their interactions with polities. On the firm level it is argued that performance and profitability goals can create significant pressures for firms and, when legitimate means to attain these important goals fail, normative structures may decay, leading forms to find deviant alternative (Martin et al., 2007).
Based on results of prior research, it appears that bribery in international business occurs as a result of the interaction of three key variables – economic conditions, cultural factors, and social changes (Sanyal & Samanta, 2004). This is noticeable as well in context-specific studies such as that of Fisman & Svensson (2007), Gaviria (2002), Sahakyan & Stiegert (2012), Şeker & Yang (2014), Wang & You (2012). These show that bribery‟s effect vary even if the firms are located in comparable countries based on their high corruption levels: namely, Uganda, Latin America, Armenia the Caribbean region and China. For instance, Sahakyan & Stiegert (2012), based on their Armenian firm sample, determined that large firms and firms with few competitors (monopoly, duopoly, triopoly) were statistically more likely to see corruption as favorable for firm performance. However, as firms aged or if the firm manager had a university degree, there was a lower likelihood that corruption would be viewed favorably. Wang & You (2012) demonstrated that corruption committed by firms in China indeed enhances the growth of their sales income. However, it is also determined that is not always the case: as institutional improvements continue, the benefits of corruption will be reduced and eventually exhausted, thus the growth-enhancing effect is transitory (Wang & You, 2012).
sector or type of business. Most importantly, the study also finds that corruption (and crime) substantially reduce sales growth.
Thus, it appears as though the majority studies seem to support the notion that corruption is a serious constraint in doing business (Fisman & Svenson, 2007), and that corruption could bring some benefits during transition, but will utlimately be destructive unless anti-corruption policy is put into place (Wang & You, 2012).
Other studies that have studied the consequences of bribery include Lee & Weng (2013). Lee & Weng (2013)‟s cross-national research focused on the consequences of bribery by emphasizing how this might affect firm exports as a widely used international strategy. This study concludes in that when firms obtain greater preferential treatment from home country government officials via bribes, they tend to have greater interest in the home country and less interest in foreign countries (Lee & Weng, 2013). This resonates with the line of thinking of Gaviria (2002) in that exporting firms will be less competitive in a country where a port official charges hefty bribes to complete pre-shipment inspection – thus, bribery will raise operational costs and create uncertainty (Gaviria, 2002).
3. HYPOTHESES
The premise of the present study is that exogenous stimuli and a firm‟s bribery involvement are likely to influence firm performance. Accordingly, the hypotheses proposed are concerned with the specific conditions under which bribery may have positive or negative effects on firm performance. These hypotheses are based on the institutional theory (DiMaggio & Powell, 1983; Scott, 2008), and bribery is considered as a response to institutional pressures to which a firm is subjected to.
3.1. Institutional pressures, bribery, and firm performance
The concept of institutions has many varying meanings and usage. North (1991) stated that institutions are the humanly devised constraints that structure political, economic and social interaction. Similarly, the conception of institutions, as per Scott (2008), is that institutions are comprised of regulative, normative and cultural-cognitive elements that, together with associated activities and resources, provide stability and meaning to social life. In other words, all social systems – hence, all organizations – exist in an institutional environment that defines and delimits social reality (Scott, 1987). This institutional environment is in turn supported by both formal and informal institutions which are each represented by formal constraints (sanctions, taboos, customs, traditions, and codes of conduct), and formal constraints (constitutions, laws, property rights) (North, 1991).
As such, the institutional arena contains a number of exogenous pressures that influence the structure and behavior of organizations (Dacin, 1997). Theories about the determinants of corruption emphasize the role of economic and structural policies and also the role of institutions (Svensson, 2005). For instance, the institutional framework is primarily concerned with an organization‟s relationship or fit with the institutional environment, the effects of social expectations (prescriptions) on an organization, and the incorporation of these expectations as reflected in organization characteristics (Dacin, 1997).
Mauro, 1995; Seleim & Bontis, 2009; Svensson, 2005), the link between bribery involvement and a firm‟s performance will be ascertained through the vantage point of each pillar.
3.1.1. Regulative and normative pressures
It is likely that one of the reasons why corruption flourishes is because the “regulative pillar” within the society is affected. The regulative pillar is associated with the capacity to establish rules, inspect other‟s conformity to them, and, as necessary, manipulate sanctions – rewards or punishments – in an attempt to influence future behavior (Scott, 2008). A mechanism of this is considered to be DiMaggio & Powel‟s (1983) coercive isomorphism. They reasoned that organizations result because of pressures exerted by political influence and the problem of legitimacy. In such sense, it is these environmental agents that are sufficiently powerful to impose structural forms and/or practices on subordinate organizational units (Scott, 1987).
The dominant strand of literature asserts corruption as undermining the quality of market and political institutions, distorting investment decisions, reducing firm productivity and national economic growth, and affording undue political influence to those engaged in corruption (Jensen et al., 2010). In many countries, and especially in developing countries, the role of the state is often carried out through the use of numerous rules or regulation (Tanzi, 1998). Licenses, permits, passports, and visas are needed to comply with laws and regulations that restrict private activity (Shleifer & Vishny, 1993). With private firms, these control rights stem from the existing regulatory system and the discretion public officials have in implementing, executing, and enforcing rules and benefits that affect firms, such as business regulations, licensing requirements, permissions, taxes, exemptions, and public-good provision (Svensson, 2003). It is here that the danger lies. As Shleifer & Vishny (1993) put forward, insofar as government officials have discretion over the provision of these goods, they can collect bribes from private agents. Or as Gaviria (2002) posited, regulation is mainly a mechanism to create rents for bureaucrats. Thus, the existence of these regulations and authorizations give a kind of monopoly power to the officials who must authorize or inspect the activities (Tanzi, 1998). If rules and norms are to be effective, they must be backed with sanctioning power (Scott, 2008).
perceived by business people and their citizens to be less corrupt if they are highly developed, long-established liberal democracies, with a free and widely read press, a high share of women in government, and a long record of openness to international trade (Treisman, 2007). A freer and more influential press will also reduce the scope of corruption as will the existence of well-functioning institutional checks and balances (Gaviria, 2002). These are arguably the case in more developed countries. In such countries, one can assume that honest and effective supervisors, good auditing offices, and clear rules on ethical behaviors should be able to discourage or discover corrupt activities (Tanzi, 1998).
On the contrary, the effects of bribery are expected to be different in weaker regulatory environments. It has been attested by many theorists that a weak institutional environment enforces corruption (see Eicher, 2009; Mauro, 1998). In such an environment, refusing to participate in the corrupt system may be disadvantageous in the short term because bureaucratic process may move much slower and deals may be lost to another who is willing to be „a player‟ (Eicher, 2009). Thus, corruption may be expected to be more widespread in countries where red tape slows down bureaucratic procedures (Mauro, 1995); here bribes are likely to be more common (Cuervo-Cazurra, 2006). In such cases, corruption is said to be an extra-legal institution used by individuals or groups to gain influence over the actions of the bureaucracy (Leff, 1964). It is here, unsurprisingly, that the inefficiency of bureaucracy has often been considered the most prominent inefficiency that corruption can grease (Méon & Weill, 2010).
The previous arguments therefore maintain that a firm is more likely to engage in bribery involvement, if its environment encourages that behavior. If this is the norm, than likely bribery and corruption are necessary for conducting business. On the other hand, firms that operate in environments that have stronger sanctioning power are less likely to be corrupt themselves. Thus, the hypothesis is formulated as:
Hypothesis 1: The quality of the regulatory environment in which a firm operates, and
firm bribery involvement influences its subsequent firm performance.
3.1.2. Cultural-cognitive pressures
Formal and informal institutions develop trough historical experiences and factors that shape culture (Eicher, 2009). Culture is a set of beliefs and values about what is desirable and undesirable in a community of people, and a set of formal or informal practices to support the values (Javidan & House, 2001). In other words, cultural variables result from unique sets of shared values among different groups of people (Deresky, 2006). The cognitive-cultural pillar recognizes that “internal” interpretive processes are shaped by “external” cultural frameworks (Scott, 2008). A cultural-cognitive conception of institutions stresses the central role played by the socially mediated construction of a common framework of meaning (Scott, 2008). Additionally, in some cases firms are likely to encounter uncertainty in their business environment, and as such exhibit corruption activities as a form of mimetic isomorphism (DiMaggio & Powell, 1983).
Power Distance
Davis & Ruhe (2003) argued that societies high on power distance have norms, values, and beliefs such as: inequality is fundamentally good; most people are dependent on a leader; the powerful are entitled to privileges; the powerful should not hide their power, and there should be a hierarchy of power. Both the studies of Husted (1999) and Davis & Ruhe (2003) hypothesized that the higher the power distance in a country, the higher corruption would be in a country. Both studies proved that this is indeed the case. As such, it can be expected that the higher the power distance in a country, the higher the bribery involvement of firms would be. As corruption is arguably widespread in such countries this should have less of a negative effect on the performance. Thus, the hypothesis is formulated as:
Hypothesis 2: A firm’s bribery involvement is more likely to result in a positive effect on
Individualism / collectivism
This cultural dimension reflects the extent to which people prefer to take care of themselves and their immediate families, remaining emotionally independent from groups, organizations, and other collectivities (Schneider & Barsoux, 2003). Societies high on individualism have norms, values, and beliefs such as: individual achievement is ideal; people are responsible for themselves; people need not be emotionally dependent on organizations or groups (Davis & Ruhe, 2003). Contrariwise, countries with a collectivist orientation would have a preference for group as opposed to individual decision-making (Schneider & Barsoux, 2003). Such societies are likely to valorize harmony and saving face, whereas individualistic cultures generally emphasize self-respect, autonomy and independence (Deresky, 2006).
The impact of this dimension on micro corruption can be argued both ways. This is echoed in extant literature as well (see Davis & Ruhe, 2003 and Husted 1999). On the one hand, one could argue that in individualistic countries, people would be more likely to participate in bribery if this would help their individual performance. On the other hand, because of the need to save face, collectivist countries would not engage in bribery. Though, it could also be that within a collectivistic nation members would rely on their relationships to excel.
Past macro literature has found mixed results on the link between corruption and the individualism/collectivism cultural dimension. Davis & Ruhe (2003), for example, found that high collectivism leads to a higher corruption level, whereas, Husted (1999) find an insignificant relationship. As such, our hypothesis is formulated as:
Hypothesis 3: A firm’s bribery involvement and its result on its performance will be
influenced by the degree of individualism/collectivism characterizing the country.
Masculinity / Femininity
Bribery involvement Firm Performance Institutional environment Regulative and normative pressures Cultural-cognitive pressures x
is perceived to be, the more advantage is likely to be taken in order to achieve higher personal gain. Therefore, the hypothesis is:
Hypothesis 4: A firm’s bribery involvement is more likely to result in a positive effect on
performance the higher the masculinity degree in the country.
Uncertainty avoidance
Uncertainty avoidance as per Hofstede (1997) is the extent to which members of a culture feel threatened by uncertainty or unknown situations, and whom therefore try to avoid ambiguous situations by providing greater certainty and predictability (Davis & Ruhe, 2003). Thus, in other words, in environments where there is high uncertainty avoidance, corruption can be viewed as a mechanism to reduce the uncertainty (Husted, 1999). As such, the hypothesis therefore becomes:
Hypothesis 5: A firm’s bribery involvement is more likely to result in a positive effect on
performance the higher the uncertainty avoidance in the country.
3.2. Conceptual model
The conceptual model below provides a simplistic visualization of the corruption literature review provided. It also provides support to the research question and the hypotheses.
4. DATA AND METHOD
This section concerns the data and methodology applied in this particular study.
4.1. Data
The multilevel dataset of firm-level and country-level for this cross-sectional study is drawn from three types of datasets from two well-established sources. Namely, the World Bank, and Hofstede‟s cultural dimensions. All these datasets are published data.
To start with, the firm-level data is drawn from the comprehensive multinational study of the Business Environment and Enterprise Perfomance Survey (BEEPS) data of 2002, following Blahojevic & Damijan (2013), Lee & Weng (2013) and Şeker & Yang (2014). BEEPS is an initiative of the European Bank for Reconstruction and Development (ERBD) and the World Bank, to investigate the extent to which government policies and practices facilitate or impede business activity and investment in Central and Eastern Europe and the former Soviet Union (Blahojevic & Damijan, 2013).
The sample structure for BEEPS II was designed to be as representative (self-weighted) as possible to the population of firms within the industry and service sectors subject to the various minimum quotas for the total sample (MEMBRB Custom Research Worldwide, 2002). This approach ensured that there was sufficient weight in the tails of the distribution of firms by the various relevant controlled parameters of sector, size, location and ownership (MEMBRB Custom Research Worldwide, 2002). The survey is conducted uniformly in all the countries covered (Blahojevic & Damijan, 2013). It uses firm-level corruption indicators that measure firms‟ perceptions of and their actual experiences with corruption (Jensen et al., 2010); these are formulated as “unofficial payments/gifts to public officials”, and are asked later during the interview, after rapport has been built with the participant. In total, the 2002 survey conducted face-to-face interviews with 6667 prequalified representatives of firms from several industries over a span of 27 countries in Europe.
Secondly, for measures for the institutional context, the dataset of the Worldwide Governance Indicators (WGI) as compiled by Daniel Kaufman, Art Kraay and Massimo Mastruzzi at the World Bank (Kanyama, 2014) is used. Of this dataset, the “Control of corruption” indicator is selected as the proxy of for the regulatory context.
4.2. Sample
For this specific analysis and in order to compute the relevant constructs, the datasets available had to meet several criteria. Several questions of BEEPS 2002 were considered1. As the core of the present study is to determine how firm bribery involvement and the institutional environment affect firm performance, all questions regarding the former and latter were critical and had to be completed by the respondents. Therefore, firms were excluded based on the following criteria. Namely: (1) firms that missing values for questions with regards to firm performance and bribery questions, and (2) firms that indicated no knowledge about unofficial payments.2
The limited coverage by Hofstede‟s cultural dimensions was also taken into consideration. Cultural dimensions (power distance, uncertainty avoidance, individualism and masculinity) were available for only 13 countries of the 27 countries covered by the 2002 survey. The countries covered by Hofstede were Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland, Romania, Russian Federation, Slovakia, Slovenia, and Turkey. Consequently, firms from the other 14 countries for which Hofstede data was not available were excluded from this study3. Data for the WGI Control of Corruption indicator in 2002 was available for all countries.
Additionally, for statistical purposes, outliers and influential cases were also evaluated. Outliers are cases that differ substantially from the main trend of the data; in other words, 99% of the standardized residuals should lie between -3.29 and + 3.29. Therefore, any cases that do not lie in between these parameters are excluded, as they can affect the estimates of the regression coefficients. Influential cases were also assessed with the use of Cook‟s distance. This is a measure of the overall influence of a case on the model, with values greater than 1 representing potential issues. However, no influential cases were detected. All observations fall below the 1 value threshold. Effectively, this lead to a sample size of 1796 firms spread over 13 countries (table 1).
1
An overview can be found in Appendix A. 2
Such as respondents whom selected “don‟t know” in questions 54, 56 and 59.
Table 1 Countries in sample.
Country Firms Percentage
1 Bulgaria 122 6,8% 2 Croatia 88 4,9% 3 Czech Republic 145 8,1% 4 Estonia 70 3,9% 5 Hungary 143 8,0% 6 Latvia 73 4,1% 7 Lithuania 118 6,6% 8 Poland 259 14,4% 9 Romania 155 8,6% 10 Russian Federation 282 15,7% 11 Slovakia 73 4,1% 12 Slovenia 137 7,6% 13 Turkey 131 7,3% Total 1796 100% Source: BEEPS 2002. 4.3. Variables 4.3.1. Dependent variable
Several BEEPS 2002 questions were considered in the operationalization of the dependent variable, firm performance. A question with regards to sales price exceeding operating costs was not considered as a sufficient measurement of firm performance, as this question specifically stated for the respondent to link this with their main product line of services in the domestic market4. Therefore it was considered not representative for the firm performance as a whole. Another question focused only on exports and domestic sales. This was also not considered a proper measurement as firm performance in this context.
Alternatively, the question in BEEPS examining the relation between the levels of gross profits (expressed as a percent) in relation to the total sales for the year of 2001 was deemed appropriate. This was selected as the main measurement of firm performance for the purposes of this study5. The response of this question was a 7-level categorical variable, with a range of 7 scores. Namely; (1) negative, (2) 0%, (3) 1-10%, (4) 11-20%, (5) 21-30%, (6) 31-40% and (7) more than 40%. For analytical purposes, these are rescaled into binary variables. As the interest lies in determining how a firm‟s performance is affected by its environment and bribery environment, the main dependent dummy variable distinguishes between firms rankings below and above a gross profit ratio of 10%6.
4 Question 23 in BEEPS 2002. 5
Question 84a1 in BEEPS 2002.
4.3.2. Independent variables
Two types of independent variables can be distinguished in the context of this study. These are bribery involvement and the institutional environment. Bribery involvement, the firm-level independent variable is constructed using BEEPS data. Subsequently, the institutional environment is classified by two constructs; (1) the regulatory context and (2) the normative/cultural-cognitive context. The former is conceptualized as the control of corruption level, one of the six World Governance Indicators as measured by the World Bank. The latter, is conceptualized by dimensions of Hofstede.
Bribery involvement
As set forth by Olken & Ande (2012), the most direct way of measuring bribery is through surveys of bribe payers. In BEEPS 2002, five questions7 with regards to the topic are to be found. Therefore, to operationalize this construct all questions were taken into consideration following common practices with regards to survey measures (see Martin et al., 2009 and Zhou & Peng, 2012). As each is formulated differently, responses are also likely to vary. Two open questions are disqualified, as they are likely not representative of the actual bribery participation8.
A principal component analysis was conducted on twelve items of the survey that focused on “unofficial payments”.9
This was done in order to understand the structure of the set of variables. An oblique rotation (direct oblimin) as per the theoretical background of the questions was applied. Measures for the verification of adequate sampling size included the Kaiser-Meyer-Olkin (KMO) measure, and the KMO per individual item. Furthermore, to ensure validity and reliability, the criteria of eigenvalue greater than 1 was applied (Field, 2013; Hair, Anderson, Tatham, & Black, 1998). Two components scored higher than the criterion of the eigenvalue of 1, explaining jointly 59.5% of the variance. The bribery involvement construct was established by these two dimensions, with a Cronbach‟s α = 0.900. Similar measures have been used by Fisman & Svenson (2007) and Lee & Weng (2013), Zhou & Peng (2012).
Control of corruption
The control of corruption (CC) WGI indicator captures perceptions of the extent to which public power is exercised for private gain, including both petty and grand forms of corruption,
7 These are question 54, 55, 56, 57 and 59. 8
as well as “capture” of the state by elites and private interests (Kaufmann, Kraay, & Mastruzzi, 2010). This is constructed based on perceptions-based governance data sources. It ranges from a scale of minus 2.5 to 2.5, with the highest score indicating a better control of corruption level (less corruption). In other words, the lower the score, the weaker the regulatory environment is considered to be, as the control of corruption is lower. This indicator was selected as it provides a better overview of public power and corruption. Additionally, it also concurs with the common definition of corruption: the misuse of public office for private gain.
Cultural dimensions
Geert Hofstede developed a framework for understanding how basic national cultures influence behaviors in countries. This consisted of four dimensions, power distance, uncertainty avoidance, masculinity and individualism. Each of these dimensions is ranked per country, with 0 as the lowest score possible and 100 as the highest score possible. The cutoff middle is 50, where ranks below indicate a low score on the dimension, and ranks above 50 indicate a high score on the dimension.
4.3.3. Control variables
Four firm characteristics were controlled for. First, a variable of firm age was created based on the firm‟s founding year as reported by the BEEPS 2002. This is calculated by subtracting the year of establishment from the year of BEEPS. Secondly, an ownership dummy variable was created. This dummy variable distinguishes between domestic and foreign ownership with a cutoff percentage of 50% (equals 1). Thirdly, firm size is also accounted for. The firm size dummy variable differentiates between small firms (up to 49 employees and medium / large firms (up to 10.000 employees). The dummy variable categorizes small firms as 0, and otherwise as 1. And lastly, industry differences were accounted for with the measurement of an industry dummy variable which divides the sample in two main industries (1) manufacturing, and (2) services.
4.4. Method of analysis
are categorical, or a mix of continuous and categorical, and the dependent variable is categorical, logistic regression is necessary (Burns & Burns, 2009). This analysis technique will be conducted to determine how the independent variables (macro and micro) affect the outcome variable.
Table 2 List of variables.
Variable Description Measurement
Firm performance (↑10%) DV 10 % gross profit ratio over total sales Dummy (= 0 below, = 1 above) Firm performance (↑20%) DV 20 % gross profit ratio over total sales Dummy (= 0 below, = 1 above) Firm performance (↑30%) DV 30 % gross profit ratio over total sales Dummy (= 0 below, = 1 above) Control of corruption IV WGI indicator Continuous; 0.72 max.
Power distance IV Hofstede dimension Continuous; 100 max. Individualism IV Hofstede dimension Continuous; 100 max. Masculinity IV Hofstede dimension Continuous; 100 max. Uncertainty avoidance IV Hofstede dimension Continuous; 100 max. Bribery involvement IV Hofstede dimension Continuous; 100 max.
Firm age CV Firm age in 2002 Continuous
Firm size CV Small firm < 49 employees, large firms >
50 employees Dummy (= 0 small, = 1 large) Industry CV Services or manufacturing industry Dummy (= 0 services, = 1
manufacturing)
Ownership CV Foreign or domestic majority shareholder Dummy (= 0 domestic, = 1 foreign)
4.4.1. Preliminary analysis
Table 3 represents the correlations coefficients, means and standard deviation of the main variables. Additionally, this table also allows the interpretation of the effect sizes which is the measurement of the effect relative to errors in the sample (Field, 2013). These effects in general appear to be relatively low. In other words, the majority of the coefficients lie below r = .30, this implies a medium effect, which is likely to account for 9% of the variance, whereas, r = .10 represents a small effect likely to explain 1% of the variance. Correlation coefficients above r = .50 account for 25% of the variance.
power distance, individualism and masculinity. Therefore, this indicates that it is likely that bribery involvement is influenced by its environment.
Table 3 Correlations, means, and SDs. Mean SD 1 2 3 4 5 6 7 8 9 10 1 FM (↑10%) ,38 ,49 2 CC -,04 ,54 -,17** 3 PD 69,34 18,44 ,10** -,60** 4 IDV 48,07 16,19 -,10** ,43** -,65** 5 MAS 46,78 23,53 ,00 ,20** ,12** ,45** 6 UA 82,45 12,26 ,03 -,32** ,44** -,39** -,08** 7 BI 1,79 ,83 ,11** -,21** ,16** -,12** ,06* ,03 8 Firm age 13,40 13,51 -,10** ,04 ,03 -,04 ,06** ,05* -,08** 9 Firm size ,32 ,47 -,07** -,02 ,03 ,00 ,03 -,02 -,06* ,32** 10 Industry ,37 ,48 -,04 -,02 ,05* -,06* ,00 ,09** ,04 ,10** ,19** 11 Ownership ,14 ,35 ,04 ,03 -,05* ,05* ,02 -,04 ,00 -,07** ,18** ,03 Note: FM(↑10%)= Firm Performance at gross profit ratio of 10%, CC=Control of Corruption, PD = Power Distance, IDV = Individualism, MAS = Masculinity, UA = Uncertainty Avoidance, BI = Bribery Involvement.
** Correlation is significant at the 0.01 level (two-tailed). * Correlation is significant at the 0.05 level (two-tailed). SD, standard deviation.
4.4.2 Assumptions
Logistic regression does not assume a linear relationship between the dependent and independent variables (Burns & Burns, 2009) and the dependent variable should be dichotomous (Burns & Burns, 2009; Field, 2013). Other assumptions of this type of analysis include that the independent variables need not be interval, nor normally distributed, nor linearly related, nor of equal variance within each group (Burns & Burns, 2009). Field (2013) additionally points out that if the central limit theorem is taken into consideration, normality is likely not to be a concern in such a sample, as the larger the sample size is, the better the approximation of normality will be (Field, 2013). Another assumption is that the categories (groups) must be mutually exclusive and exhaustive; a case can only be in one group and every case must be a member of one of the groups (Burns & Burns, 2009). Finally, the sample size is advised with a minimum of 50 cases per predictor (Burns & Burns, 2009; Field, 2013). The sample at hand meets all the above criteria, with a relatively large sample size supporting the analysis; it consists of 1796 firms. Additional assumptions of homoscedasticity and multicollinearity are also considered. These are described as follows.
Homoscedasticity
the significance level which is greater than 0.5 indicates that heteroscedasticity is not present in the sample data.
Table 4 Breusch-Pagan test for heteroscedasticity.
Sample size Number of predictors² Breusch-Pagan test Significance level³
1796 10 16,293 ,0915
¹Dependent variable is firm performance (↑10%).
²Predictors are all independent variables: Firm age, firm size, industry, ownership, bribery involvement, control of corruption, power distance, individualism, masculinity and uncertainty avoidance.
³H0: homoscedasticity.
Multicollinearity
As there are several predictors in the analysis, multicollinearity is an additional concern. This exists when there is a strong correlation between two or more predictors (Field, 2013). At first glance (table 5), indicates that there are no multicollinearity signs, as there are no correlations above 0.8. However, to confirm this, the variance inflation factor (VIF) diagnostic as well as the tolerance statistic is calculated. The results are visible in table 5. These indicate whether a predictor has a strong linear relationship with the other predictors. As per the criteria discussed by Field (2013), it can be concluded that multicollinearity is not likely to be an issue with this data, as the tolerance is well above 0.1 and the VIF values are all below 10.
Table 5 Collinearity statistics.
Tolerance VIF Control of corruption ,50 2,01 Power distance ,22 4,49 Individualism ,27 3,76 Masculinity ,40 2,53 Uncertainty avoidance ,77 1,30 Bribery involvement ,92 1,09 Firm age ,86 1,16 Firm size ,83 1,21 Industry ,95 1,05 Ownership ,95 1,06
5. EMPIRICAL RESULTS
First, an overview of the descriptive statistics of the sample is provided. Subsequently, the results of the logistic regression are presented.
5.1. Descriptive statistics
Table 6 provides the descriptive statistics of the variables conducive to the context of this study. The distribution of frequencies of categorical variables can be found in Appendix B. Of the 1796 firms in the sample, 1114 fall in the category of below 10% profit ratio. Almost all (94.7 %) of the sample firms are characterized by a profit ratio of below 30%. Of those 7.0%, 126 firms, indicate that they have a negative profit ratio in relation to total sales, whereas 146 firms, 8.1% report a 0% profit ratio.
As for the independent variables, the firm-level construct of bribery involvement variable with a mean of 1.79 (from a scale from 1 to 6, with 1 representing never and 6 representing always10), indicate that bribery involvement is not widespread in the sample. In fact, 75% of the respondents appear to have indicated “seldom” (Q3 = 2.25) on the questions representing the construct.
The macro level independent variables represent the nature of the business environment of the sample countries11. Overall, the countries have a low control of corruption, with the average of -0.04, and a maximum of mere positive 0.72. The standard deviation (positive value), however, does differ from the mean (a negative value), therefore indicating that there is quite a variance in the rankings. Yet, it appears that at least 25% of the sample size has a negative control of corruption score (Q1 is -0.83) and the majority, 75% falls below a score of -0.36 (Q3 is 0.35).As the nature of the countries in the sample is transitional, such a low score is not surprising. In other words, the majority of countries in the sample have a weak control of corruption, implying a weaker institutional environment, and therefore concurring with extant literature.
As for the cultural indicators, it appears that the sample is characterized on average by a low individualism ranking, as well as a low masculinity ranking (both the average and median, quartile 2, rank below the 50 middle point), with the lowest score reported of a 9 for masculinity. On the other hand, the countries rank on average high on the uncertainty
10 1 = never, 2 = seldom, 3 = sometimes, 4, frequently, 5 = usually, 6 = always 11
avoidance indicator and medium high on the power distance indicator. Nevertheless, the standard deviations of the cultural dimensions differ in comparison to their averages, therefore indicating that there are variations in between the countries‟ cultural dimensions rankings.
The control variables show that the majority firms in the sample are relatively young with an average age of 13 years, and 75% of the firms raking below 13 years old (Q3 = 12). In other words, half of the sample firms are less than 10 years old (Q2 = 9). However, the range in between the youngest firm and mature firm is remarkably large, namely 69 years. As for the firm size, the majority (68.4%) appear to have less than 50 employees, namely 1228 firms. The rest, 568 firms, are represented by medium and large firms, exceeding up to 10.000 employees. Furthermore, the sample appears to be characterized by the service industry, with 1137 service firms and 659 manufacturing firms. Finally, the sample accounts for 1545 domestic firms and 251 foreign firms.
Table 6 Descriptive statistics of variables.
Mean SD Q1 Q2 Q3 Min. Max.
Dependent variables 1 Firm performance (↑10%)¹ ,38 ,485 - - - 0 1 Independent variables 2 Control of corruption² -,04 ,54 -,38 ,04 ,36 -,92 ,72 3 Power distance³ 69 18 57 68 90 40 104 4 Individualism³ 48 16 33 39 60 27 80 5 Masculinity³ 47 24 36 42 64 9 110 6 Uncertainty avoidance³ 82 12 74 85 93 51 95 7 Bribery involvement 1,79 ,83 1,08 1,50 2,25 1,00 6,00 Control variables 8 Firm age 13 14 6 9 12 3 72 9 Firm size¹ ,32 ,465 - - - 0 1 10 Industry¹ ,37 ,482 - - - 0 1 11 Ownership¹ ,14 ,347 - - - 0 1
¹ Distribution of frequencies of dummy variables can be found in Appendix B.
² Control of corruption ranges from -2.5 to a high of 2.5, with the highest number representing a better control of corruption. ³ Cultural indicators range from 0-100, with 50 representing the middle; below 50 is considered low, and above is considered high on the scale of the respective dimension.
Source: information derived from BEEPS 2002, the World Bank and Hofstede.
5.2. Baseline results
parameters were assessed. Although the value of Nagelkerke R² increases in each model, indicating better fit, they still remain each below 10%. This indicates a low explanation of the variance in the outcome variable, firm performance. On the other hand, prediction success was above 60% in each model, with the full model representing 64% (87.3% for profit below 10%, and 26.4% for profit above 10%).
Further details with regards to each model can be found in table 7. The focus will be on the regression coefficients, the significance levels but also on the practical significance level, which is represented by the odds ratio. The first model includes the control variables only. Subsequently, the micro independent variable, bribery involvement, is added in model two. The third model includes the addition of the other (macro) independent variables. Lastly, the fourth third model represents the interaction effects of the micro and macro variables.
5.2.1. Control variables effects
Table 7 Logistic regression results.
Model 1 Model 2 Model 3 Model 4
Control variables + micro independent variable + macro independent variable Interaction effects
β Exp(B) Sig. β Exp(B) Sig. β Exp(B) Sig. β Exp(B) Sig.
Constant -,249( ,080) ,779 ,002** -,736 (,136) ,479 ,000*** 1,617 (,652) 5,037 ,013* ,853 (1,573) 2,346 ,588 Firm age -,014 (.004) ,986 ,001*** -,013 (,004) ,987 ,002** -,014 (,004) ,986 ,001** -,014 (,004) ,986 ,001*** Firm size -,192 (,116) ,825 ,098* -,172 (,117) ,842 ,141 -,207 (,119) ,813 ,081* -,221 (,119) ,802 ,064* Industry -,103 (,104) ,902 ,322 -,128 (,105) ,879 ,220 -,138 (,107) ,871 ,198 -,136 (107) ,873 ,203 Ownership ,243 (,143) 1,275 ,089* ,241 (,143) 1,273 ,092* ,278 (,146) 1,320 ,057* ,293 (,147) 1,341 ,045* Bribery involvement ,263 (,059) 1,301 ,000*** ,149 (,062) 1,160 ,016* ,681 (,771) 1,975 ,377 Control of corruption -,818 (,134) ,441 ,000*** -,536 (,335) ,585 ,110 Power distance -,018 (,006) ,982 ,002** -,025 (,014) ,975 ,076* Individualism -,024 (,006) ,976 ,000*** -,035 (,015) ,965 ,018* Masculinity ,013 (,003) 1,013 ,000*** ,023 (,008) 1,023 ,006* Uncertainty avoidance -,005 (,005) ,995 ,302 ,012 (,011) 1,012 ,282 BI x CC -,180 (,177) ,835 ,309 BI x PD ,004 (,007) 1,004 ,581 BI x INDV ,006 (,007) 1,006 ,383 BI x MAS -,006 (,004) ,994 ,165 BI x UA -,010 (,006) ,990 ,070* Observations 1796 1796 1796 1796 Nagelkerke R² ,019 0,034 ,077 ,082 -2LL 2359,361 2339,517 2279,707 2273,157 Chi² 25,4856*** 45,330** 105,141*** 111,691*** H-L Goodness-of-fit 13,246 6,112 8,884 7,689
Note: Dependent variable is firm performance (↑10%). Standard errors in parenthesis.
* p < 0.10; ** p < 0.05; *** p < 0.001.
5.2.2. Independent variables effects
The second and third models of the logistic regression represent the additions of the independent variables. The addition of the bribery involvement construct in the second model has a significant positive result (β = 0.263; p < 0.001). The exp(b) value of 1.301 indicates that as bribery involvement increases, the odds ratio will be 1.3 as large and therefore firms are 1.3 times to have profit ratio above 10% when other when other variables are controlled for. In this model, firm age and ownership also remain significant; firm age remains negatively associated (β = - 0.013; p < 0.05) and ownership remains positively associated (β = 0.241; p < 0.10).
In the third model, the rest of the independent variables are added. These are the macro-level determinants of control of corruption, power distance, and individualism, masculinity and uncertainty avoidance. The effects of these appear to differ from one another. Control of corruption, power distance, individualism, and uncertainty avoidance all have negative values, and are significant, with the exception of uncertainty avoidance (β = - 0.005; p > 0.10). As for the previous ones, the values are β = - 0.818; p < 0.001, β = - 0.018; p < 0.05, and β = - 0.024; p < 0.001, respectively. Masculinity, on the other hand is positively and significantly related (β = 0.013; p < 0.001). As for the control variables, firm age (β = - 0.014; p < 0.05) and firm size (β = - 0.207; p < 0.10) remain negatively and significantly related in model 3, whereas, ownership (β = 0.278; p < 0.10) and bribery involvement (β = 0.149; p < 0.05) remain with positive and significant values.
5.2.3. Interaction effects
Contrariwise, the interaction effect between bribery involvement and uncertainty avoidance does appear to be both negative and significant (β = -0.010; p < 0.10). Interestingly, in the full model, the previously significant control variables remain significant whereas the micro-level independent variable, bribery involvement, does not (β = 0.681; p > 0.10).
In terms of the hypotheses, hypothesis 1 stated that the quality of the regulatory environment would influence bribery involvement and firm performance. The regulatory environment is measured in this context with the control of corruption variable. The interaction between control of corruption and bribery involvement results in a negative β in relation to firm performance when firms bribe. This concurs with the proposition stated by hypotheses 1; the higher the control of corruption, the better the regulatory environment, and therefore the lower the firm performance. However, as this is not a significant result (p > 0.10) hypothesis 1 is rejected.
Hypothesis 2 conjectured that bribery involvement would be positive in relation to high power distance. In both model 3 and 4, power distance remained significant and negatively related to the outcome variable. This indicates that as power distance increases, the profit ratio is likely to be less than 10%. The interaction between power distance and bribery involvement, although positive, is not significant contribution to the prediction of a profit ratio of more than 10% (β = 0.004; p > 0.10). As such, hypothesis 2 is not supported.
Lastly, hypothesis 3 examined the relation between the degree of individualism and firm bribery involvement on its performance. This cultural dimension reported a significant and positive relation with the outcome variable in both model 3 and 4. This implies that the higher individualism, the higher the firm performance is likely to be. However, the interaction effect fails to provide a significant result, and as such hypothesis 3 is rejected.
Hypothesis 4 considered the effects of the masculinity dimension and bribery involvement on firm performance, and argued that the higher the masculinity the more positive the effect. This cultural dimension, masculinity, is positively related with the outcome variable in both model 3 and 4, indicating that the higher the masculinity, the higher the firm performance. The interaction effect of bribery involvement and masculinity provide a negative β = - 0.006, which implies that the higher this predictor, the less likely a firm is to obtain a gross profit ratio of above 10%. This defies expectations and as this is insignificant (p >0.1), hypothesis 4 is not supported.
was not significant in model 3 or 4, however, the interaction effect between uncertainty avoidance and bribery involvement does lead to a negative and significant effect on firm performance (β = - 0.010; p < 0.10). Thus, the higher this predictor, the less likely firm performance will be (gross profit ratio above 10%). Although this is significant, it does not concur with the hypothesis, which predicted the opposite. As such, hypothesis 5 is rejected.
5.3. Robustness tests and extensions
Additional analyses were performed to establish robustness. First, as selection bias due to sample composition (removal of outliers, influential cases and missing values) might have occurred, a logistic regression is re-estimated, without the removal of outliers and influential cases. Secondly, several logistic regressions are re-estimated based on different alternative measures for the bribery involvement construct. Finally, the alternative dependent dummy variables are used in a logistic regression based on the same initial sample. The results of these can be found Appendix D.
5.3.1. Results with whole sample
The logistic regression is re-estimated without the removal of any firms prior to the analysis. Thus, the regression is conducted with the observation of 6667 firms, the original database as provided by BEEPS 2002. However, missing cases upon conducting the regression are listed as 3348 firms. Therefore, the ultimate total of the actual firms included in the analysis are 3319 firms (49.8%). Of these, 2056 firms report a profit ratio below 10%, whereas 1263 firms have a higher profit. The regression results reported are approximately similar to the original one, no major differences are detected. Therefore it can be concluded that the conservative dataset applied in the main regression are not biased.
5.3.2. Results with alternative bribery involvement measures
The first alternative logistic regression (model AIV1, Appendix D) considered the perception of bribery, with statements that considered how common it was for firms to pay irregular additional payments/gifts, and if these firms usually knew in advance about how much this would be, as the main measurement for the bribery involvement construct. In general, the all variables relate in the same way as in the main logistic regression. The most remarkable about this regression, however, is that three interaction effects are found to be significant.
The first interaction, bribery perception x individualism, reports a positive and significant relation with the outcome variable (β = 0.008; p < 0.10). Thus, as bribery perception and individualism increases, the more likely a firm is to have a gross profit ratio above 10%. This evidently concurs with hypotheses 3.
The interaction of bribery perception x masculinity, on the other hand, is negative and significant (β = - 0.006; p < 0.10). Thus, as bribery perception and masculinity increases, the less likely firm performance is to be higher. Finally, just as the main regression, the interaction between bribery perception x uncertainty remains negative and significant.
Similar results are found in model AIV5. That is, this regression also finds additional significant results with regards to the interaction effects. This model assesses bribery as whether it had impact or not, which was a dummy variable extracted from one of the five questions within BEEPS 2002 focused on unofficial payments. As opposed to the first alternative model, however, the effects in between bribery x uncertainty avoidance is now positive and significant (β = 0.023; p < 0.05). This shows evidence for hypothesis 5. On the other hand, the interaction between bribery x individualism remains positive and significant, whereas, the interaction between bribery x masculinity remains negative.
Model AIV4, which employed government additional payments/gifts as the main bribery construct, contrariwise, reports a completely new significant interaction effect that is the interaction effect between bribery x power distance. This is positive and significantly related to the outcome (β = 0.005; p < 0.05). Thus, as power distance and bribery increases, the firm is likely to report a higher performance outcome. This shows support for hypothesis 2.
The second and third alternative logistic regressions had similar results as the main regression; no other significant relations were found.
5.3.3. Results with alternative dependent variables
