How the Pervasiveness and Arbitrariness of Corruption Impacts
the performance of firms
University of Groningen Faculty of Economics and Business
Master Thesis International Economics and Business
Abstract: Existing research tends to examine the economic impact of corruption at a country-level. This paper investigates the impact of corruption through its two characteristics (pervasiveness and arbitrariness), on firm performance at the firm-level. A distinction is made between foreign and domestic firms in order to determine if a Liability of Foreignness effect occurs, as it is suspected that foreign firms will underperform compared to their domestic counterparts.
Key words: Corruption, Pervasiveness, Arbitrariness, Firm Performance, Liability of Foreignness
Name Student: CHRISTIAN DONNGES Student ID number: S2091224
Student email: c.r.a.donnges@student.rug.nl Date Thesis: 14 JUNE 2016
Page 2 of 35
Contents
1. Introduction ... 3
2. Literature Review ... 5
Corruption ... 5
Pervasiveness and Arbitrariness of Corruption ... 6
Liability of Foreignness... 8
Corruption and firm performance ... 9
Management and arbitrariness ... 11
Liability of Foreignness and arbitrariness ... 12
3. Data and Analysis ... 13
Page 3 of 35
1. Introduction
Currently many firms conduct their activities both domestically and internationally. Among the reasons for investing and locating certain activities abroad include achieving cost advantages or gaining access to cross-border knowledge and foreign knowledge resources (Jensen & Pedersen, 2012), with the aim of obtaining competitive advantages over their rivals. However, going abroad also bears its costs and firms are often unable to fully take these into account due to a combination of negligence and unawareness. This often results in firms underperforming in foreign markets. Larsen et. al. (2012) find that, on average, the savings realised by companies who expand their operations abroad to foreign markets are 6.7% less than anticipated. In foreign markets, MNEs operate at a disadvantage relative to domestic firms due to information asymmetries. Acquiring this information may also prove to be costly for them (Hymer, 1976). This is just one example of Liability of Foreignness
(LOF), which entails the costs of doing business abroad (CDBA) that local firms do not incur in the same market (Eden & Miller, 2004). Firms must learn to deal with external and internal elements (e.g. employees of different culture) which contribute to LOF. Meanwhile, domestic firms have more knowledge of these things hence do not incur these costs (Calhoun, 2002). There are numerous ways in which firms try to overcome LOF. One method of doing so would be engaging in corruption, defined as “the abuse or misuse of positions or resources of public officials for private gains” (Lee & Oh, 2007). Corruption often runs rampant in
emerging markets, due to a combination of lower institutional quality (Sharma & Mitra, 2015). Furthermore, Melgar et. al (2010) consider corruption and its perception to be a cultural phenomenon as they depend on a society’s understandings of rules and actions which constitute a deviation from these rules.
In existing research, scholars tend to treat corruption as a single variable. Rodriguez et. al. (2005) break corruption down into the two characteristics of pervasiveness and arbitrariness, an approach since used by a few other papers. These characteristics reflect the extent to which corruption is prevalent and ambiguous respectively (Pillay & Dorasamy, 2010). It is found that pervasiveness can lead to an improvement in firm performance whereas
arbitrariness a reduction in firm performance.
Page 4 of 35
economies in Central and Eastern Europe). This paper will differentiate itself by investigating the effect of corruption at the firm level. No attempt will be made to focus on a specific geographical region as this is of no concern.
The aim of this paper is to answer the following research question: How does the impact of
pervasiveness and arbitrariness of corruption on firm performance differ between domestic and foreign firms? This paper will contribute to existing research by examining whether the
effect of corruption’s two dimensions exhibit a LOF effect on foreign firms. As per Rodriguez et. al. (2005), corruption is broken down into the two characteristics of
pervasiveness and arbitrariness. MNEs are bound to come across various levels of both types in any location or institutional setting.
Splitting up corruption into pervasiveness and arbitrariness allows one to obtain a deeper understanding of its effects on firm performance. This is especially because it is expected that pervasiveness and arbitrariness have different effects on firm performance. Lee & Oh (2007) find that pervasiveness encourages firms to engage in bribes whereas arbitrariness has the opposite effect. The rationale behind this is that corruption may be considered as an extra tax that foreign firms pay in order to receive some sort of service in return. If the returns realised by firms exceeds their initial costs, then they will consider corruption a worthy investment opportunity. An environment with high pervasiveness signals that corruption is a worthwhile activity to engage in for firms, which leads to mimetic isomorphism as other firms would also want to replicate these successful strategies. Arbitrariness on the other hand would be a deterrent to corruption, as firms would prefer to avoid uncertainty.
The data analysed in this paper comes from the World Business Enterprise Survey (WBES) 1999-2000 will be used. From the available data it will be possible to analyse the effect of corruption’s two characteristics on firm performance, as well as distinguish the effects faced between domestic and foreign firms. The reason behind this is because it is suspected that foreign firms are more likely to be prone to corruption. According to Eden & Miller (2004), corruption distance (difference of corruption level between home and foreign country) can affect Liability of Foreignness (LOF), which comprises the social costs of doing business abroad.
In the following section, a literature review will examine the current research on corruption and its effect on economic performance. Based on this, a number of hypotheses are
Page 5 of 35
paper will conclude with a discussion of the findings as well as limitations and suggestions for future research.
2. Literature Review
Corruption
Corruption has been a big topic for decades, both inside and outside academia. A BBC survey from 2010 found that corruption was the most discussed problem, and a Gallup International survey from 2014 found corruption to be the number one problem identified by people. In an effort to counter corruption, research has been conducted into this field with the intention of gaining an understanding of its causes and consequences.
This paper defines corruption as: “the abuse or misuse of positions or resources of public officials for private gains” (Lee & Oh (2007). Among activities that constitute corruption are bribery, extortion, patronage, influence buying, favouritism, nepotism, fraud and
embezzlement (Ufere et. al., 2012).
In academia it is argued that corruption hinders economic performance (Mauro, 1995). According to Zelekha and Sharabi (2012), corruption hampers economic activity through the following mechanisms: distortion in resource allocation, increased uncertainty, degradation of legal mechanisms, loss of leadership, reduced marginal productivity of capital, increased inequality in income distribution and effect on small business sector. Rose-Ackerman (1999) identifies that “institutionalised trust” determines the extent of which entrepreneurship and innovation can flourish. Tzafrir (2005) finds a positive relation between trust and firm performance. At the firm level, Avnimelech et. al. (2014) provides evidence that corruption exhibits a directly negative effect on entrepreneurship. The effect was more profound in developed countries, which they attribute to institutional differences.
Page 6 of 35
In such environments, there are strong incentives to earn income through rent-seeking
activities (e.g. bribery, lobbying) rather than through value-creating activities (Murphy et. al., 1993). Furthermore, corruption may be relatively more efficient than the weak institutions hence it may actually grease the wheels of economic activity. This can often be found in cases considering bureaucratic requirements, which serve as a burden to firms (Carr & Outhwaite, 2009) and are worsened in an environment characterised with weak and
inefficient bureaucracy. Consider the example where a firm must obtain some sort of license (e.g. an export license). Rather than undergo the proper but cumbersome process, the firm decides to pay a bribe to the officials tasked with issuing the specific license. The officials take the bribe in secret and decide to provide the license to the firm in a more convenient and efficient manner. They could provide it to them directly rather than making the firm wait for a certain amount of time. Both benefit from engaging in corrupt activities. This is because firms achieve their goals more efficiently, whereas the party taking the bribe enjoys the gift it was awarded by the firm.
The problem with corruption is that only those who engage in it benefit from corruption, whereas it exhibits negative externalities on third parties. For example, firms who are more productive but less willing to engage in corruption would lose out. A consequence of this is a resource misallocation towards firms which engage rent-seeking activities (Kurer, 1993), which is economically undesirable as it may result in more productive firms being pushed out of the market.
Pervasiveness and Arbitrariness of Corruption
Many researchers examine corruption through a single general variable rather than looking at its specific characteristics. A commonly used variable is the Corruption Perceptions Index, published by Transparency International on an annual basis. Countries are awarded a certain score and ranked against each other. A lower score signifies a higher level of corruption in the country.
Page 7 of 35
corruption, this paper opts to make use of the same approach in order to construct its measure for corruption.
Both variables are important to take into account when examining corruption as they
determine whether corruption can be beneficial or costly to firms. This is especially the case as pervasiveness and arbitrariness have the opposite effect on firm performance. For example, if arbitrariness is low, then firms have more certainty regarding the costs of corruption and can easily take this into account. Essentially, corruption would be akin to paying an extra cost or tax (Rodriguez et. al., 2005) with the intention of obtaining some sort of benefit. Through a cost-benefit analysis, firms can then decide whether the payoff of engaging in corruption would be worthwhile.
Conversely, if there is high arbitrariness then it is a lot more difficult for firms to properly account for the full cost of corruption. It leads to a situation where the terms of a corrupt arrangement can change suddenly and without warning and nothing can be done about it. For example, the party receiving the bribe may decide not to deliver the service they initially agreed upon or suddenly demand a higher bribe for the same service previously agreed upon. Due to corruption’s illegal and unethical nature a system of contractual enforcement is absent and involving local authorities in such matters may actually make land the bribe giver into legal issues. Consequently, firms will no longer be able to determine with ease whether engaging in corrupt activities is beneficial or harmful to their performance.
Regarding pervasiveness, this entails the extent of corruption and may signal whether it is a worthwhile or necessary activity to engage in. A high pervasiveness may indicate so, and has the potential encourage corruption for new entrants. Oliver (1991) finds that firms are likely to comply with pervasive corruption. A possible explanatory mechanism is mimetic
isomorphism, which is when firms will imitate the practices of the more successful firms in the markets even if they are corrupt (Ufere et. al., 2012). This may be especially true when there is a high level of competition, as this may exert more pressure on firms and their ability to survive on the market. Hence other firms may be forced to engage in corrupt activities just so they can survive in the market. Furthermore, Andvig & Moene, 1990 find that if
corruption is more pervasive then it is much easier for firms to find corrupt officials and get away with it. Hence in environments with a high pervasiveness of corruption create
Page 8 of 35
Uhlenbruck et. al. (2006) investigate the arbitrariness and pervasiveness of corruption on the entry strategies of telecommunication projects in emerging economies. They were unable to find a main effect of pervasiveness on its own, but when coupled with arbitrariness it was found to worsen the effect of the latter.
Liability of Foreignness
Hymer (1976) identifies certain types of disadvantages foreign firms face in a host country market not faced by their domestic counterparts. The first one is information asymmetry, as foreign firms would have less information and obtaining it could be costly. The second is that foreign firms could face differential and inferior treatment from the domestic government, buyers and suppliers of the host country which could persist through time. This suggests that firms would suffer LOF effects on a long-term basis.
Zaheer (1995) broadly defines the concept of Liability of Foreignness (LOF) as the additional costs a firm operating in a foreign market incurs that a local firm would not incur. While certain scholars have often used CDBA and LOF synonymously, Eden & Miller (2004) differentiates between the two concepts (figure 2.1) as they identify LOF as being one portion of CDBA, the other being Activity-Based Costs (ABC) which all firms incur.
Figure 2.1: Components of Costs of Doing Business Abroad (Eden & Miller, 2004)
Unfamiliarity hazards relate to asymmetric information, as it entails the costs arising from the lack of knowledge or experience in the host country.
Relational hazards examine those at both the intra- and inter-organisational level. An
Page 9 of 35
Discrimination hazards would be linked to situations when the firm is subject to less favourable treatment due to its foreignness. For example, ethnocentric consumers feel that consumption of foreign goods harms their home country. This causes them to avoid these even if they are superior and cheaper relative to those of domestic firms (Balabanis et. al., 2001).
Eden & Miller (2004) identified three types of institutional distances that foreign firms face. The first is regulatory distance, which is covers the setting, monitoring and enforcing of the rules (Xu & Shenkar, 2002). Asymmetrical information may arise due to a gap in the understanding of institutions between foreign and domestic firms (Calhoun, 2002) In
environments characterised with weaker institutions, these aspects would either be weaker or absent. This may lead to a higher level of arbitrariness. For example, a lack of contractual enforcement could result in deals not being honoured. It may also result in intellectual theft and counterfeiting activities where property rights and enforcements are weak
The second, normative distance, deals with the social norms, values, beliefs and assumptions about human nature and behaviour that are socially shared and carried by individuals
(Kostova, 1997). This deals with how things should be done according to the values of society. This is important to consider as the perception of corruption is dependent on what is accepted within the society. In cultures characterised with collectivism, hierarchy and relationships have a much more important role in conducting business activities. This may lead to nepotism, which can hinder firm performance (Bertrand & Schoar, 2006).
The third is cognitive distance, which deals with how people notice, characterise, and interpret stimuli from their environments (Kostova, 1999). Consumer ethnocentrism may be linked to this issue, as the nature of their avoidance of foreign products is largely due to characterisation and interpretation.
Hypotheses formulation
Corruption and firm performance
When firms face corruption, there are two characteristics which it is affected by:
pervasiveness and arbitrariness. As varying levels of both are faced at one time, it is likely that an interaction effect is in play on top of their individual respective effects.
Page 10 of 35
characteristics of corruption, there are 4 possible scenarios for them. This is visually demonstrated in table 2.1:
Table 2.1: Possible outcomes of firms in terms of pervasiveness and arbitrariness of corruption
Level of Pervasiveness/Arbitrariness Low High
High A B
Low C D
Quadrant A (QA) depicts the situation of high pervasiveness and low arbitrariness. In other words, corruption is frequent but its effects are predictable. Consequently, firms can easily conduct a cost-benefit analysis and make a decision on whether or not corruption would be a worthwhile experience for them. Furthermore, its high pervasiveness signals that bribery is socially acceptable, which may also suggest that corruption may be beneficial or even necessary for business purposes. This will lead to a situation of mimetic isomorphism, as other firms will want to replicate the successful strategies regardless of whether they are corrupt or not. A low level of arbitrariness means firms are able to estimate the costs of engaging in corrupt activities to a high level of accuracy, and that they can count on the services they demand to be delivered accordingly.
Quadrant B (QB) depicts the situation where both pervasiveness and arbitrariness are high. This is a situation where corruption occurs frequently and bears a lot of uncertainty.
Corruption may be a socially accepted activity here, however the high arbitrariness would mean that firms here would not be able to easily determine whether it is worthwhile to engage in it. Overall it is unsure how beneficial it would be for firms to engage in corrupt activities in these types of environments.
Quadrant C (QC) depicts the situation of low pervasiveness and low arbitrariness. In this situation corruption is low-scale and its effects can be anticipated. In this environment firms are able to determine the costs of engaging in corruption, hence can make a decision on whether it is beneficial to engage in corrupt activities. As in quadrant B, it is unknown how beneficial it would be for firms to engage in corruption in this type of environment.
Page 11 of 35
suggest that in such an environment corruption is not a worthwhile activity for firms to engage in.
When this is all taken into account, the lower pervasiveness in QC and QD suggest that corruption is less beneficial here as firms are not engaging in it at a large scale. Meanwhile when looking QA and QB, the only factor differentiating these two is the arbitrariness. As arbitrariness concerns the uncertainty of corruption, it is expected to lower firm performance. Therefore QA is expected to represent the type of environment in which corruption is most beneficial for firms.
H1: Engaging in bribery improves firm performance the most when pervasiveness is high and arbitrariness is low
Management and arbitrariness
Existing research shows that management is important. Li & Zhou (2010) find that
managerial ties is one of the things which can help a foreign firm to achieve success. Bloom et. al (2011) find a strong positive relationship between management practices and firm performance.
One criticism of corruption is that it draws resources away from productive and economically efficient activities towards rent-seeking activities. An example of this is managers who spend time trying to create and foster relations with influential figures, such as government
officials. Faccio (2006) find that political connections are particularly common in countries with higher levels of corruption. Zheng et. al. (2014) provides evidence that political ties can buffer firms from threats to survival. The consequence of this is that firms who are more productive but not willing to engage in corruption would lose out in such situations.
This leads to a reallocation of resources towards firms that are relatively less productive but more willing to engage in corruption. In China party membership benefits an entrepreneur secure a loan from a bank (Li et. al., 2008). This example demonstrates how political
connections are conductive to business success. In such an environment, the more time a firm spends with government officials, the more likely it is to have a stronger relationship with them. Schramm (2006) finds that corruption benefits existing and better-connected firms over others.
Page 12 of 35
H2: The more time managers spend with government officials, the lower the firm performance in an environment with high arbitrariness
Liability of Foreignness and arbitrariness
Foreign firms are subject to extra costs in foreign markets, otherwise known as LOF which contributes to their CDBA (Eden & Miller, 2004) for reasons as outlined by Hymer (1976). Firms face unfamiliarity hazards due to asymmetric information. Unlike domestic firms, foreign firms lack knowledge of domestic customs when entering the market. Additionally, obtaining information may even prove to be costlier for foreign firms (Hymer, 1976). Moreover, firms may face discriminatory hazards due to differential treatment from customers up to the governments (Eden & Miller, 2004). Consumer ethnocentrism occurs when consumers are prejudiced against foreign products as they feel that their home country is harmed by consuming them (Balabanis et. al., 2001). Shimp & Sharma (1987) find that consumer ethnocentrism is a strong predictor of explaining purchasing behaviour, as it could lead to consumers preferring domestic over foreign products even if the latter were cheaper and of better quality. Good & Huddleston (1995) find that the predictive strength of
consumer ethnocentrism varies by country. Potential drivers of consumer ethnocentrism are nationalism and patriotism, dependent on the country (Balabanis et. al, 2001).
A corrupt environment may only serve to further exacerbate this problem as foreign firms make potential targets for public officials to exploit. For example, public officials may bend the laws and regulations to benefit themselves at the detriment of the foreign firm as the latter may not even be aware that they are doing so. However, this tactic would likely be less successful with local firms, as they tend to be more knowledgeable of the environment that they are operating in.
The aspect of corruption we are especially interested in is the arbitrariness. Namely, it is expected that the asymmetric information to exacerbate the arbitrariness linked with corruption. Hence the following is proposed as the third hypothesis:
Page 13 of 35
3. Data and Analysis
To answer the research question, the previously stated hypotheses are tested through an econometric model. The data used here is collected from the World Business Environment Survey (WBES) conducted in 1999-2000, consisting of face-to-face interviews with
managers and owners of companies located in numerous countries. The dataset along with the actual survey and variable explanations are both freely available on the internet from the World Bank. Created by Daniel Kaufmann and Andrew H. W. Stone, the survey was created with the aim of measuring corruption, judiciary, lobbying and quality of the business
environment. In total the survey contains 10,032 observations across 81 countries.
It should be acknowledged newer WBES data is available. However, the more recent dataset lacks consistency in terms of timing. For example, Argentina has firm data from the years 2006 and 2010 whereas Afghanistan has firm data from 2008 and 2014. It would not make sense to compare these firms against one another as one would have to control for time effects. Each country has different amounts of data entries and different years from which they come from. As a result, there is no consistency in terms of timing. On the other hand, the WBES 1999-2000 is more consistent in terms of timing, as firm data was collected within the period 1999-2000. While the time period is not all within the same year, time effects would be less pronounced in such a smaller period than it would be in one which spans 10 years. As corruption is a sensitive issue, the managers were guaranteed anonymity in an attempt to allow them to be more open without the fear of being punished. The problem with this however is that anonymity alone is not sufficient to ensure that managers answer as honest and openly as possible.
After selective elimination, the sample size used in this paper consists of 2,716 observations from 57 countries. The observations excluded were those with missing data from the
questions used to create the pervasiveness and arbitrariness variables (see table 3.1), as well as those which were considered to be influential outliers based on a Cook’s D test and a Studentised Residual test.
The remainder of the section is structured as follows. First a theoretical model is proposed and all the variables in it are explained. Second, the data will be analysed through descriptive statistics. Next the data will be subject to diagnostic tests to determine whether
Page 14 of 35
coefficient and significance of each explanatory variable. These will determine whether any of the proposed hypotheses will be accepted or rejected, subject to the 5% significance level.
Theoretical Model
First, a baseline model will be constructed which involves solely the control variables alongside sector and country dummies (SC and CC respectively) to control for their respective effects. For this, the following is proposed:
𝐹𝑆𝐴 = 𝛽1𝑆𝑍𝐹 + 𝛽2𝐴𝐺𝐸 + 𝛽3𝑆𝐶 + 𝛽4𝐶𝐶
In order to test out our hypotheses, the following model is proposed:
𝐹𝑆𝐴 = 𝛽1𝑞𝐴 + 𝛽2𝑞𝐵 + 𝛽3𝑞𝐷 + 𝛽4𝐹𝑂𝑊 + 𝛽5𝐵𝑆𝑃 + 𝛽6𝑆𝑍𝐹 + 𝛽7𝐴𝐺𝐸 + 𝛽8𝑀𝐺𝑇 + 𝛽9𝐵𝑆𝑃
∗ 𝑞𝐴 + 𝛽10𝐵𝑆𝑃 ∗ 𝑞𝐵 + 𝛽11𝐵𝑆𝑃 ∗ 𝑞𝐷 + 𝛽12𝑀𝐺𝑇 ∗ 𝑞𝐴 + 𝛽13𝑀𝐺𝑇 ∗ 𝑞𝐵 + 𝛽14𝑀𝐺𝑇 ∗ 𝑞𝐷 + 𝛽15𝐹𝑂𝑊 ∗ 𝑞𝐴 + 𝛽16𝐹𝑂𝑊 ∗ 𝑞𝐵 + 𝛽17𝐹𝑂𝑊 ∗ 𝑞𝐷
Regarding H1, coefficient β9 is examined as the variable of interest is the interaction between
bribery and quadrant A. H1 suggests that quadrant A is where corruption would pay off the most for firms, thus β9>0 is expected.
To test H2, it is examined whether β13<0. With both high pervasiveness and arbitrariness, the
effects of the latter will be felt the most. This is because in such an environment the effects of arbitrariness would lead to management not being able to achieve what it intends to through its relations with the government officials.
To test H3, it is examined whether β16<0. This is where both pervasiveness and arbitrariness
are high, in which a foreign firm is expected to have a lower performance compared to domestic firms. This is because the effects of arbitrary corruption can mostly be felt here, as the pervasiveness would exacerbate the effects of arbitrariness.
Dependent variable
Firm Sales (FSA): The annual total sales of a firm, expressed in dollars. This data was already available in the dataset, however it will be normalised through a natural log transformation. This variable serves as a proxy for firm performance.
Independent variables
Page 15 of 35
that 6 now indicated the highest level of pervasiveness. This was done to ensure the regression results obtained are more intuitive.
Arbitrariness of Corruption (ARB): The creation of this variable is based on 3 questions identified by Uhlenbruck et. al. (2006), who then created an average score out of it (see Table 3.1). The process is verbatim to that of creating PRV.
Quadrants (QA, QB, QC, QD): These quadrants refer to the 4 possible interactions between the pervasiveness and arbitrariness of corruption based on their intensity (low or high). In order to test out which quadrant is best, one dummy variable for each of them is created and set them equal to one if they fit the right conditions. This was first done by creating dummy variables for high levels of pervasiveness and arbitrariness, which equalled 1 if they were high (above average) and 0 otherwise. Then dummy variables were created for each quadrant based on the corresponding values of the dummy variables for pervasiveness and
arbitrariness. Namely:
QA = 1 if High Pervasiveness=1, High Arbitrariness =0. Otherwise =0 QB = 1 if High Pervasiveness, High Arbitrariness = 1. Otherwise =0 QC = 1 if High Pervasiveness, High Arbitrariness = 0. Otherwise =0 QD = 1 if High Pervasiveness = 0, High Arbitrariness = 1. Otherwise =0
In order to avoid collinearity, one of the quadrants is omitted from the actual regression. QC is chosen to be omitted, as it makes an ideal reference point because it depicts the situation of low pervasiveness and arbitrariness.
Foreign Ownership (FOW): This dummy variable reflects a firm’s foreignness based on whether it has any foreign ownership (=1) or not (=0). The inclusion of this variable is to determine whether there is an LOF effect.
Bribe Sales Percentage (BSP) represents the bribes firms pay on an annual basis relative to their sales of the corresponding year. It is measured on a scale from 1-7, with a higher number indicating a higher amount of bribery relative to sales. For the analysis it is rescaled to 0-6.
Page 16 of 35
Control variables
Firm Size (SZF): This categorical variable assigns the numbers of 1 (small), 2 (medium), 3 (large) to each firm. This data was already present in the dataset. Firm size is included as a control variable as larger firms are found to have better firm performance (Lee, 2009). AGE: This variable, expressed in years, represents how old the firm is. This is included as a control variable as its expected that older firms are more experienced and more established hence would have a higher amount of sales compared to newer firms.
A summary and some other details regarding all the variables involved are found in table 3.1:
Table 3.1: Summary of Variables
Variable
Abbreviation Variable Name Explanation
FSA Firm Sales This variable takes the natural log of firm sales QA, QB, QC, QD Quadrant A, B, C, D See table 2.1 BSP Bribery Sales Percentage
Annual bribes expressed as a percentage of sales. The variable is measured through the following scale:
1: None, 2: 1-10%, 3: 11-20%, 4: 21-30%, 5: 31-40%, 6: 41-50%, 7: >50%
FOW Foreign
Ownership
Dummy variable, =1 if firm has any foreign ownership. Otherwise =0
MTG
Management Time spent with
Government
The amount of time managers spend with government officials, expressed in the following scale:
1: <1%; 2: 1-5% 3:6-10% 4:11-25% 5:26-50% 6:
>50%
SZF Size of the firm Firm size, expressed in a scale of 1-3 (low, medium, high), expressed as whole numbers only
PRV Pervasiveness of
Corruption
Variable is created by averaging the scores of the following questions from WBES 1999-2000:
Do firms like yours typically need to make extra, unofficial payments to public officials for any of the
following?
to get connected to public services
electricity, telephone)
to get licenses and permits to deal with taxes and tax collection
to gain government contracts when dealing with customs / imports
when dealing with courts
Page 17 of 35
1: Always, 2: Mostly, 3: Frequently, 4: Sometimes, 5:
Seldom, 6: Never
In this paper the scale has been reversed in order to make the results of the statistical analysis more intuitive
ARB Arbitrariness of Corruption
Variable is created by averaging the scores of the following questions from WBES 1999-2000:
Firms in my line of business usually know in
advance about how much this ‘additional payment’ is
If a firm pays the required “additional
payment” the service is usually also delivered as agreed
If a government agent acts against the rules I
can usually go to another official or to his superior and get the
correct treatment without recourse to unofficial payments
The responses of the questions were on the following scale:
1: Always, 2: Mostly, 3: Frequently, 4: Sometimes, 5:
Seldom, 6: Never
In this paper the scale has been reversed in order to make the results of the statistical analysis more intuitive
Page 18 of 35
Descriptive statistics
Table 3.2: Summary of variables
Variable Mean Standard Deviation Minimum Maximum LNVSAL 7.2824 7.4115 0.6931 20.7233 PRV 2.9996 2.4137 0.0000 6.0000 ARB 3.1532 1.1268 0.0000 6.0000 BSP 4.1572 1.4177 0.0000 6.0000 FOW 0.1418 0.3489 0.0000 1.0000 SZF 1.7422 0.6862 1.0000 3.0000 AGE 16.6960 20.8789 0.0000 201.0000
From table 3.2, it is shown that the means of both PRV and ARB are close to 3. PRV is slightly less, suggesting that on average firms experience low pervasiveness. The standard deviation however proves to be quite large, and it is shown that the sample does involve firms which reach the extremes of 0 and 6
In terms of ARB, the mean is slightly above 3 which suggests that the average firm faces high arbitrariness. The standard deviation is shown to be much smaller compared to PRV,
suggesting that the deviation from the mean is less in terms of ARV.
Regarding BSP, it is shown that the average firm engages in a high level of bribing relative to their sales.
The mean of FOW suggests not many firms have any level of foreign ownership.
Page 19 of 35
Figure 3.1: Scatter plot of PRV and ARB
In figure 3.1, the plots represent each firm in terms of their levels of pervasive and arbitrary corruption. The two lines represent their respective averages, and separate the plot area to represent the four quadrants as shown in table 2.1. This is done in order to obtain a better understanding of where the observations fit in terms of both dimensions. No specific pattern is shown, suggesting that the two variables aren’t correlated with one another.
Diagnostic tests
Outlier analysis
First the data will be analysed for potential outliers. One way to do so would be to conduct a leverage-versus-residual-squared (LVR) plot (figure 3.2):
Page 20 of 35
Figure 3.2: Leverage-versus-residual-squared plot
The LVR plot (figure 3.2) is divided into four parts by lines which mark the averages of the leverage and normalised residual squared. Points above the horizontal line are those which have a leverage larger than the average of the sample, while those to the left of the vertical line exhibit a normalised residual squared larger than the average of the sample. The leverage represents how far an independent variable deviates from its mean.
Hence any observations located in the northwest quadrant are those whose leverage and normalised residual squared are larger than the average. Based on this graph some outliers are observed. Hence some tests will be conducted.
In order to determine what to do with these outliers, a Cook’s D test is conducted in order to identify the most influential variables. According to this test, if the calculated Cook’s D is found to be larger than 4/N, the observation is considered an outlier. The test identified 442 significant outliers.
Next the studentised residual is calculated. For this test, outliers are those where the absolute value of their studentised residual exceeds 2.5. For this test, 1 outlier in addition to those identified by the Cook’s D test was identified.
Any outliers identified by either of the two tests was removed from the final dataset. Tests for normality
0 .0 5 .1 .1 5 .2 L e v e ra g e 0 .005 .01 .015
Page 21 of 35
In order to test for normality, the residuals are plotted on a Histogram (figure 3.3) and a Kernel Density plot (figure 3.4).
Figure 3.3: Histogram of residuals
Figure 3.4: Kernel Density Estimate of residuals
Both graphs show that the data exhibits a leptokurtic distribution, suggesting the data is not normally distributed. To confirm this the residuals are subject to a Skewness and Kurtosis test as well as a Shapiro-Wilk test.
Table 3.3: Skewness and Kurtosis test for normality
Variable Observations Pr(Skewness) Pr(Kurtosis) adj chi2(2) Prob>chi2
R 2.2e+03 0.0290 0.0000 62.35 0.0000
Table 3.4: Shapiro-Wilk test for normality
Variable Observations W V Z Prob>z r 2169 0.98533 18.735 7.476 0.00000 0 .2 .4 .6 De n s it y -5 0 5 Residuals 0 .2 .4 .6 De n s it y -5 0 5 Residuals Kernel density estimate Normal density
kernel = epanechnikov, bandwidth = 0.1565
Page 22 of 35
For both the Skewness and Kurtosis test (Table 3.3) as well as the Shapiro-Wilk test (Table 3.4), it is observed that in both cases the p-values are below 0.05. Hence the null hypothesis is rejected and it is concluded that the data is not normally distributed. This means that the results obtained through an OLS regression and any conclusions for the hypotheses may be challenged.
Test for heteroscedasticity
To determine whether heteroscedasticity it is present, an IM test (table 3.3) and Breusch-Pagan test are conducted (table 3.4).
Table 3.3: IM test Source chi2 df p Heteroscedasticity 2132.89 338.00 0.0000 Skewness 746.15 62.00 0.0000 Kurtosis 329.99 1.00 0.0000 Total 3209.03 401.00 0.0000
Table 3.4: Breusch-Pagan test for heteroscedasticity
chi2(9) 677.48 Prob > chi2 0.0000
From both tests, it is observed that p-value = 0.0000 < 0.05. Hence the null hypothesis is not rejected and it is concluded that heteroscedasticity is present.
Tests for collinearity
Collinearity occurs when there are one or more exact linear relationships among the explanatory variables (Hill et. al., 2012). This creates a problem as it means that the
Page 23 of 35
Table 3.5: Collinearity Matrix
LNVSAL PRV ARB BSP FOW SZF AGE MGT
LNVSAL 1 PRV 0.8918 1 ARB 0.0098 0.0253 1 BSP 0.3229 0.2572 -0.0465 1 FOW 0.1864 0.1444 -0.0127 0.1187 1 SZF 0.2489 0.1698 0.0379 0.1654 0.2219 1 AGE 0.3311 0.2881 0.0153 0.1606 0.0565 0.3118 1 MGT -0.3354 -0.3109 0.0171 -0.2468 -0.0149 0.0366 -0.048 1
The correlation matrix (table 3.5) displays the pairwise correlation between all the variables. None of the correlations in this matrix are close to 1 or -1, hence it is demonstrated that none of the explanatory variables are correlated with one another.
Table 3.6: VIF test
VARIABLE VIF SQRTVIF TOLERANCE R2
PRV 1.25 1.12 0.7998 0.2002 ARB 1.01 1.01 0.9885 0.0115 BSP 1.14 1.07 0.8750 0.1250 FOW 1.09 1.04 0.9208 0.0792 SZF 1.23 1.11 0.8163 0.1837 AGE 1.19 1.09 0.8375 0.1625 MGT 1.13 1.06 0.8848 0.1152 MEAN 1.15 CONDITION NUMBER 14.9482
A Variance Inflation Factor (VIF) test is conducted as an additional test for collinearity. While Heiberger & Holland (2013) suggest a threshold of 10, this paper will use a threshold of 5 as proposed by Hair et. al (2006). The condition number is a measure of how stable the data is. From 10 onwards it is considered unstable. As it is observed to be 14.9842, it is concluded that the data here is unstable.
Page 24 of 35
As a result of these two tests, it is concluded that the data does not suffer from collinearity problems.
Results
Table 3.7: Results of models testing hypotheses
Page 25 of 35 MFO*QB -0.9846* (0.5123) MFO*QD -0.3357 (0.5350) PFO 0.0093 (0.0138) PFO*QA 0.0058 (0.0172) PFO*QB -0.0002 (0.0155) PFO*QD -0.0037 (0.0167) PRV 0.1525** (0.0653) ARB 0.0120 (0.0578) PRV*ARB -0.0026 (0.0146) CONSTANT 18.4164*** -0.4656 13.9786*** 13.9786*** 13.7554*** 12.1268*** 13.1492*** (1.0948) (0.7066) (0.5545) (0.6194) (0.7139) (3.1494) (0.5249) COUNTRY
EFFECTS YES NO YES YES YES YES YES
SECTOR
EFFECTS YES NO YES YES YES YES YES
R2_A 0.9025 0.8338 0.9599 0.9599 0.9587 0.9453 0.9587
RMSE 2.4281 2.9825 1.4647 1.4647 1.4864 1.7625 1.4861
AIC 9581.388 7774.0572 5636.1817 5636.1817 5681.6632 840.842 5672.334 BIC 9936.2293 7870.2156 6052.8680 6052.8680 6098.3494 1077.5973 6040.941
N 2064 1544 1544 1544 1544 198 1544
Note: Dependent variable is LNVSAL. Standard errors in parentheses. All figures rounded to 4 d.p. * p < 0.1, **p < 0.05, ***p < 0.01
Model 0 represents the baseline model, which solely involves the dependent and control variables. Model 1 tests out the hypotheses denoted in section 2. Model 2 up to Model 6 represent those used for robustness checks, which will be discussed later on in this section. In Model 1 it is observed that QA and QB have significant positive coefficients (β1, β2 >0),
Page 26 of 35
than that of pervasiveness hence firms still perform better than in an environment with low pervasiveness and arbitrariness.
When the three quadrants are interacted with BSP, it is shown that both BSP*QA and BSP*QB have significantly positive coefficients (β9, β10>0) and BSP*QD has a positive but
insignificant coefficient (β11<0). Furthermore, the coefficient of β9>β10 suggesting that
engaging in bribery pays off the most for firms in QA. In other words, corruption is most beneficial for firms which operate in an environment characterised with high pervasiveness and low arbitrariness. This is in line with H1, hence this hypothesis is confirmed.
Next the three quadrants are interacted with MGT. The coefficients for MGT*QA and MGT*QB are significantly negative (β12, β13 <0) whereas the one for MGT*QD is negative
(β14 <0)but insignificant . This suggests that the more time management spends with
government officials the lower the firm performance is when pervasiveness is high. In
addition to this, β12<β13 suggests it is less worthwhile to engage in corruption in environments
with higher pervasiveness but lower arbitrariness. A possible explanation is that management spends time with government officials in order to lower the arbitrariness of corruption, hence it is more likely to happen in QB than QA and more likely to deliver benefits. This is in line with expectations thus H2 is confirmed.
In order to test H3, the three quadrants were interacted with FOW. β15 and β17 exhibit a
positive but insignificant coefficient, whereas β16 is positive and significant. From this it can
be derived that foreign firms operating in an environment characterised with both high
pervasiveness and arbitrariness perform better than their domestic counterparts. This suggests that there is no LOF effect, as being foreign is apparently beneficial. This is the opposite of what is expected, and therefore H3 is rejected.
Robustness checks
To establish robustness, additional tests are performed. These involve either adding other explanatory variables or replacing existing ones. The results are also shown in table 3.7, represented by Model 2 up to Model 6.
Country and Sector Effects
Page 27 of 35
for each individual country and sector is created. Then each of them is included in the models previously tested, however their coefficients are not reported in table 3.7.
Model 2 is Model 1 with the addition of country and sector effects. When looking at the results, it is shown that Model 1 isn’t entirely robust to the country and sector effects. An interesting observation is that the coefficient of β1 loses its significance whereas β2 retains it
but the sign changes which suggests that firms in QB would perform worse than QC.
Additionally, the coefficients of β9 and β10 are robust however now it is observed that β9<β10.
This suggests that a higher arbitrariness leads to higher firm performance when coupled with high pervasiveness, which is not in line with H1 thus this hypothesis is rejected.
All coefficients of MGT are insignificant thus fail to be robust. β16 > 0 suggesting that foreign
firms still perform better than domestic firms when taking country effects into account.
Robust Estimate of Variance
Model 2 has demonstrated that the results obtained in Model 1 aren’t robust. Hence it will now be subject to the sandwhich estimator of variance, including the country and sector effects. These results are shown in Model 3, which show that the coefficients are identical to Model 2 but that that some variables regain their significance. An interesting observation is that β11 obtains a significantly negative effect, which is in line with expectations as it
represents the opposite scenario of QA. However as β9 < β10 the results of H1 is rejected as
once again.
The MGT coefficients remain insignificant, and β16 is robust.
Majority firm ownership
In Model 4, the variable FOW is replaced with majority firm ownership (MFO), a dummy which is set equal to 1 if more than 50% of firm ownership is foreign. This is done to
distinguish between firms who are mostly domestic owned and firms who are mostly foreign owned.
The corresponding model is Model 4. Based on the results it is shown that the coefficients are insignificant for MFO and its corresponding interactions. This suggests that firms with a majority of foreign ownership do not perform significantly different. However the β9 < β10
result still holds, thus H1 is rejected.
Page 28 of 35
In Model 5, FOW is replaced with the percentage of foreign ownership (PFO). This is to check whether the actual level of foreign ownership matters.
Based on the results it is shown that the coefficients of PFO and its subsequent interaction variables are all insignificant. Furthermore, none of the explanatory variables are significant at all. This suggests that the model is likely not a suitable one. Furthermore the amount of observations of MFO was much lower compared to FOW.
Pervasiveness and arbitrariness
For model number 6 the quadrants were replaced with the overall variables for pervasiveness and arbitrariness respectively. Here it was found that pervasiveness had a significant positive effect, but arbitrariness had no significant effect. When coupled together, no significant effect was found.
4. Conclusion
In an era characterised by the second unbundling, firms are splitting up tasks and relocating them to different geographical locations. Motives range from cost reduction to gaining
expertise. However, when entering new countries firms find themselves in environments with different institutional settings, which may help or hinder their performance. Among the problems they can face is corruption.
Existing research in this area argues that corruption is detrimental to economic activity (Mauro, 1995). Most of the existing research examines the effects of corruption as a whole. More recent research has identified and broken corruption into 2 characteristics:
pervasiveness and arbitrariness. This paper takes another step by taking into account the “foreignness” of the firm, distinguishing them from local firms based on whether they had any foreign ownership.
The aim of this paper is to answer the following research question: How does the impact of
pervasiveness and arbitrariness of corruption on firm performance differ between domestic and foreign firms?. In order to do so, data was collected from the WBES 1999-2000 survey
and then subject to regression analysis.
Page 29 of 35
environments which exhibited a high level of pervasiveness. This is in line with the findings of previous research. Regarding H2, it was found that if managers spent more time with government officials then firm performance would drop hence this hypothesis is supported. However no LOF effect is found, as it was found that foreign firms would perform better than domestic firms in an environment characterised with high pervasiveness and arbitrariness. Hence H3 is rejected.
However when subjecting the results to country and sector effects, H1 and H2 failed to hold entirely. H1 failed because while corruption proved beneficial in QA, it proved more
beneficial in QB which suggests arbitrariness also has a positive effect on firm performance. This is not in line with expectations. The time managers spent with government officials proved to be insignificant in all robustness check models. The result of foreign firms
performing better than their domestic counterparts in QB is found to be robust throughout all models. The results regarding foreign firms performing better than domestic ones remains robust, hence H3 is still rejected.
Limitations
The WBES 1999-2000 consists of survey dataset, hence there are some problems that should be acknowledged from using this type of data. First of all, the problems of non-response is significant in this dataset as well as bias. Hence the reliability of the data should be
questioned.
Another limitation is that while newer WBES data is available, the anonymous and
confidential nature of the data renders a panel data analysis unfeasible at the firm level, as it is not known to which firm each entry represents. If this were possible, it would be interesting to investigate whether the effects of corruption on firms persists through time. This is because it is possible that firms learn to deal with corruption and implement strategies which dampen the adverse effects. A suggestion for future research would be to assign some sort of
identifier (e.g.) to a certain firm. This still obscures firm identity, however enables one to track its performance through time.
Page 30 of 35
Page 31 of 35
5. References
1. Ades, Alberto, and Rafael Di Tella. "Rents, Competition, and Corruption." American
Economic Review 89.4 (1999): 982-93.
2. Andvig, Jens Chr., and Karl Ove Moene. "How Corruption May Corrupt." Journal of
Economic Behavior & Organization 13.1 (1990): 63-76.
3. Avnimelech, Gil, Yaron Zelekha, and Eyal Sharabi. "The Effect of Corruption on Entrepreneurship in Developed vs Non-developed Countries." Int Jrnl of Ent Behav &
Res International Journal of Entrepreneurial Behavior & Research 20.3 (2014):
237-62.
4. Balabanis, George, Adamantios Diamantopoulos, Rene Dentiste Mueller, and T. C. Melewar. "The Impact of Nationalism, Patriotism and Internationalism on Consumer Ethnocentric Tendencies." Journal of International Business Studies J Int Bus Stud 32.1 (2001): 157-75.
5. Baumol, William J. "Entrepreneurship: Productive, Unproductive, and Destructive."
Journal of Business Venturing 11.1 (1996): 3-22.
6. Bertrand, Marianne, and Antoinette Schoar. "The Role of Family in Family Firms."
Journal of Economic Perspectives 20.2 (2006): 73-96.
7. Bloom, Nicholas, Benn Eifert, Aprajit Mahajan, David Mckenzie, and John Roberts. "Does Management Matter? Evidence from India." Working Paper No. 16658 (2011): n. pag. National Bureau of Economic Research.
8. Calhoun, Mikelle A. "Unpacking Liability of Foreignness: Identifying Culturally Driven External and Internal Sources of Liability for the Foreign Subsidiary." Journal
of International Management 8.3 (2002): 301-21.
9. Carr, Indira M., and Opi Outhwaite. "Investigating the Impact of Anti-Corruption Strategies on International Business: An Interim Report." (n.d.): n. pag.
10. Eden, Lorraine, and Stewart R. Miller. "Distance Matters: Liability Of Foreignness, Institutional Distance And Ownership Strategy." Advances in International
Management Theories of the Multinational Enterprise: Diversity, Complexity and Relevance 16 (2004): 187-221.
11. Faccio, Mara. "Politically Connected Firms." American Economic Review 96.1 (2006): 369-86.
12. Good, Linda K., and Patricia Huddleston. "Ethnocentrism of Polish and Russian Consumers: Are Feelings and Intentions Related." International Marketing Review 12.5 (1995): 35-48.
13. Hair, Joseph F., Bill Black, Barry Babin, Rolph E. Anderson, and Ronald L. Tatham.
Multivariate Data Analysis. Upper Saddle River, NJ: Pearson Prentice Hall, 2006.
14. Heiberger, Richard M., and Burt Holland. Statistical Analysis and Data Display: An
Intermediate Course with Examples in S-plus, R, and SAS. New York: Springer
Science & Business Media, 2013.
15. Hill, R. Carter., William E. Griffiths, and Guay C. Lim. Principles of Econometrics. 4th ed. Hoboken, NJ: John Wiley & Sons, 2012.
16. Hymer, Stephen. "The International Operations of National Firms: A Study of Direct Foreign Investment." MIT Press (1976): n. pag.
17. Jensen, Peter D. Ørberg, and Torben Pedersen. "Offshoring and International
Competitiveness: Antecedents of Offshoring Advanced Tasks." J. of the Acad. Mark.
Page 32 of 35
18. Kimuyu, Peter. "Corruption, Firm Growth and Export Propensity in Kenya."
International Journal of Social Economics Int J of Social Economics 34.3 (2007):
197-217.
19. Kostova, T. "Country Institutional Profiles: Concept And Measurement." Academy of
Management Proceedings 1997.1 (1997): 180-84.
20. Kostova, Tatiana. "Transnational Transfer of Strategic Organizational Practices: A Contextual Perspective." The Academy of Management Review 24.2 (1999): 308-24. 21. Kurer, Oskar. "Clientelism, Corruption, and the Allocation of Resources." Public
Choice 77.2 (1993): 259-73.
22. Lee, Jim. "Does Size Matter in Firm Performance? Evidence from US Public Firms."
International Journal of the Economics of Business 16.2 (2009): 189-203.
23. Lee, Seung-Hyun, and Kyeungrae Kenny Oh. "Corruption in Asia: Pervasiveness and Arbitrariness." Asia Pacific J Manage Asia Pacific Journal of Management 24.1 (2007): 97-114.
24. Li, Hongbin, Lingsheng Meng, Qian Wang, and Li-An Zhou. "Political Connections, Financing and Firm Performance: Evidence from Chinese Private Firms." Journal of
Development Economics 87.2 (2008): 283-99.
25. Li, Julie Juan, and Kevin Zheng Zhou. "How Foreign Firms Achieve Competitive Advantage in the Chinese Emerging Economy: Managerial Ties and Market Orientation." Journal of Business Research 63.8 (2010): 856-62.
26. Mauro, Paolo. "Corruption and Growth." The Quarterly Journal of Economics 110.3 (1995): 681-712.
27. Melgar, Natalia, Maxim Rossi, and Tom W. Smith. "The Perception of Corruption."
Nternational Journal of Public Opinion Research 22.1 (2010): 120-31.
28. Murphy, Kevin M., Andrei Shleifer, and Robert W. Vishny. "Why Is Rent-seeking so Costly to Growth?" American Economic Review 83.2 (1993): 409-14.
29. Oliver, Christine. "Strategic Responses to Institutional Processes." The Academy of
Management Review 16.1 (1991): 145-79.
30. Pillay, S., and N. Dorasamy. "Linking Cultural Dimensions with the Nature of Corruption: An Institutional Theory Perspective." International Journal of Cross
Cultural Management 10.3 (2010): 363-78.
31. Rodriguez, P., K. Uhlenbruck, and L. Eden. "Government Corruption And The Entry Strategies Of Multinationals." Academy of Management Review 30.2 (2005): 383-96. 32. Rose-Ackerman, Susan. Corruption and Government: Causes, Consequences, and
Reform. New York: Cambridge UP, 1999.
33. Schramm, Carl J. "The High Price of Low Ethics: How Corruption Imperils American Entrepreneurship and Democracy." Journal of Markets and Morality 9.2 (2006): 277-91.
34. Sharma, Chandan, and Arup Mitra. "Corruption, Governance and Firm Performance: Evidence from Indian Enterprises." Journal of Policy Modeling 37.5 (2015): 835-51. 35. Shimp, Terence A., and Subhash Sharma. "Consumer Ethnocentrism: Construction
and Validation of the CETSCALE." Journal of Marketing Research 24.3 (1987): 280-89.
36. Smarzynska, Beata, and Shang-Jin Wei. "Corruption and Composition of Foreign Direct Investment: Firm-Level Evidence." NBER Working Paper No. 7969 (2000): n. pag. National Bureau of Economic Research.
Page 33 of 35
38. Ufere, Nnaoke, Sheri Perelli, Richard Boland, and Bo Carlsson. "Merchants of Corruption: How Entrepreneurs Manufacture and Supply Bribes." World
Development 40.12 (2012): 2440-453.
39. Uhlenbruck, Klaus, Peter Rodriguez, Jonathan Doh, and Lorraine Eden. "The Impact of Corruption on Entry Strategy: Evidence from Telecommunication Projects in Emerging Economies." Organization Science 17.3 (2006): 402-14.
40. Xu, Dean, and Oded Shenkar. "Institutional Distance and the Multinational Enterprise." Academy of Management Review 27.4 (2002): 608-18.
41. Zaheer, S. "Overcoming The Liability Of Foreignness." Academy of Management
Journal 38.2 (1995): 341-63.
42. Zelekha, Yaron, and Eyal Sharabi. "Corruption, Institutions and Trade." Economics of
Governance Econ Gov 13.2 (2012): 169-92.
43. Zheng, Weiting, Kulwant Singh, and Will Mitchell. "Buffering and Enabling: The Impact of Interlocking Political Ties on Firm Survival and Sales Growth." Strat.
Page 34 of 35
6. Appendix
Figure A1: Histogram of Value of Sales
Figure A2: Histogram of Natural log of Sales
Table A1: Skewness and Kurtosis test of the dependent and explanatory variables
0 1 .0 e -0 8 2 .0 e -0 8 3 .0 e -0 8 De n s it y 0.00 200000000.00400000000.00600000000.00800000000.001000000000.00 vsal 0 .1 .2 .3 .4 .5 De n s it y 0 5 10 15 20 lnvsal
Variable Obs Pr(Skewness) Pr(Kurtosis) adj
Page 35 of 35
Table A2: Shapiro-Wilk Test
Variable Obs W V z Prob>z
