The effect of corruption on firm performance

(1)

The effect of corruption on firm performance

Student number: S2647966

Name: Azra Zejnilovic

Study Program: MSc ED&G

Supervisor: A. Minasyan

Key Words: institutions, corruption, firm performance

Abstract

I use recent data of 55 countries from the World Bank “Enterprise Survey” to examine the

effect of corruption on firm performance. Using Pooled OLS models, I find that the effect of

(2)

2

1. Introduction

Corruption as a measure of institutional quality is an old obstacle which has been confronting policymakers and social researchers for ages (Bardhan, 1997). Corruption can be a result of inefficient institutions or bad policies (Djankov, LaPorta, Lopez-de-Silanes and Shleifer, 2003). Corruption is often labeled as an obstacle to development and growth of firms (Hanousek & Kochanova, 2016). Nearly, 1 out of 4 people worldwide had to pay a bribe for a public service (Transparency International, 2017). Thereby, United Nations Secretary-General António Guterres (2018) stated that international corruption has astonishing annual costs of $3.6 trillion in the shape of stolen money and bribes. The bribe payments mark up $1 trillion of these costs. In conformity with this view, corruption also misallocates entrepreneurial talent and drives back foreign direct investment. Mauro, Medas and Fournier (2019: p. 5) state that “if all countries were to reduce corruption in a similar way, they could gain $1 trillion in lost tax revenues, or 1.25 percent of global GDP”. Zooming in at the firm-level, the OECD (2017) classified corruption as an invisible tariff. They estimated that the firm sales are “taxed” by approximately 5 to 10 percent in environments wherein bribery is extensive. In other words, these invisible corruption tariffs decrease the firm sales noticeably.

These worrying numbers demonstrate an increased concern in the public policy and the academic world also seems to follow this pattern (Bahoo, Alon & Paltrinieri, 2020). An expanding amount of research focuses on macro and micro level of corruption (Bahoo et al., 2020). Van Vu, Tran, Van Nguyen & Lim (2018) underscore that the topic of corruption has evolved into one of the most broadly debated topics in research. Looking at the macro level, only a handful of empirical research has identified that corruption can reinforce growth. Egger and Winner (2005) find positive effects of corruption via foreign direct investments and other researchers find these enhancing effects in environments with less effective or weak institutions (Aidt, Dutta and Sena, 2008; Dutt & Traca, 2010; Krammer, 2019; Méon & Weill, 2010). Overall, in the macro economic sphere there is more profound evidence that corruption has a negative effect (Gupta, Davoodi and Alonso-Terme, 1998; Habib and Zurawicki, 2002; Mauro, 1995, 1996, 1998; Mo, 2001; Tanzi & Davoodi, 1997; Wei, 2000). Considering the micro and/or firm-level, an increasing amount of scientific research discussing corruption has been issued in the past years (Martins, Cerdeira & Teixeira, 2020). This indicates that corruption is receiving much attention and can be considered a subject of debate.

A specific topic that has received more attention is the relationship between corruption and firm performance (Fisman & Svensson, 2007; Hanousek & Kochanova, 2016; Martins et al., 2020; Sharma & Mitra, 2015; Van Vu et al., 2018; Williams & Kedir, 2016). The reason for this development in the literature is the fact that the World Bank “Enterprise Survey” started supplying firm-level data on this topic (Sharma & Mitra, 2015). Some of these papers find that corruption shows damaging effects (Fisman & Svensson, 2007; Hanousek & Kochanova, 2016; Martins et al., 2020). On the contrary, other papers find positive (Williams & Kedir, 2016) or mixed effects (Sharma & Mitra, 2015; Van Vu et al., 2018). Generally, these studies provide very diverse, unclear and sometimes conflicting conclusions (Ashyrov & Masso, 2020; De Rosa, Gooroochurn & Görg, 2015; Van Vu et al., 2018).

(4)

4

certain whether the costs of corruption cancel out the benefits or vice versa with regards to the firm performance (Van Vu et al., 2018). Researchers and especially policymakers may also experience difficulties in drawing valid conclusions about the effect of corruption on the firm performance.

The goal of this paper is to explore the effect of corruption on firm performance by analyzing contemporary firm-level data. I argue it is better to focus on the firm-level because making use of aggregated data at the macro level cannot control for the heterogeneity between firms, which can eventually affect the performance of firms (Kasahara & Rodrigue, 2008). My main research question is the following: what is the effect of corruption on firm

performance? Answering this question is important in order to gain a better understanding of

how corruption affects the performance of individual businesses. As firms form the backbone of the economy, the harming firm-level effect contributes to a wider even more severe country-level effect (Mauro, 1995, 1996, 1998; Mo, 2001; Tanzi & Davoodi, 1997, Wei, 1999). Therefore, policymakers need to become aware that many bureaucrats and firms still show to participate in corruption and that their behavior is the source of these damaging effects. Thereby, the fact that a sizeable amount of scientific research has dedicated (and still is devoted) to this topic shows that fighting corruption is a subject that is of great importance. Additionally, my analysis with a treatment on the firm-level will add to the macro level studies. It will provide an opportunity to estimate the effect of corruption on the firm economic performance for various types of businesses in a diverse set of countries (Yasar, Paul & Ward, 2011).

The method used in this paper follows previous research (De Rosa et al., 2015; Yasar et al., (2011), with some adaptations to the model. This is in terms of either the measurement of the dependent variable corruption or the types of controls added. In my model I added controls for the firm characteristics and the type of industry. In order to be able to answer the research question, I compiled a multi-country sample. I collected and analyzed very recent Enterprise Survey data from the World Bank, which covers 55 countries over a moderate range of three years from 2017 up to and including 2019.

(5)

5

The remainder of the paper is structured as follows. Firstly, the current literature is reviewed resulting in a hypothesis. Second, the data is presented as well as the methodology used in the research process. The next section presents the results of the Pooled OLS regression from which the essential results are interpreted. Also this part introduces an adapted estimation method. Subsequently, additional checks on the adapted estimation method are provided. Finally, the conclusion consists of a brief discussion as well as policy implications, limitations and ideas for future research.

2. Theoretical background

2.1 Corruption

One of the obstacles of studying corruption comes from defining the concept of corruption (Jain, 2001). There are different definitions of corruption and it is challenging to agree on one particular definition. Corruption is generally seen as being a volatile and versatile phenomenon (Nur-tegin & Jakee, 2020; Willaims & Kedir, 2016). A very typical definition of corruption is the mistreatment of the public office for private gain that conflicts with the „rules of the game‟ (Jain, 2001; Svensson, 2005). Thereby, corruption is defined as “an illegal activity (bribery, fraud, financial crime, abuse, falsification, favoritism, nepotism, manipulation, etc.) conducted through misuse of authority or power by public (government) or private (firms) officeholders for private gain and benefit, financial or otherwise” (Bahoo et al., 2020: p. 2). Looking at corruption from a perspective of the transaction costs theory in economics (TCE), it can be seen as the allocation of a service from the bribe giver to the bribe receiver (Husted, 1994; Williamson, 1979; 1985). Corruption therefore displays the necessity to make supplementary, irregular payments to achieve things (Kaufmann, Kraay & Mastruzzi, 2003). Another description of corruption is that it can be compared to a fee or tax (Shleifer &Vishny, 1993). Comparable to taxes, bribes also create a wedge between real and privately allocated marginal product of the capital (Shleifer & Vishny, 2003). Because bribe payments are characterized by silence and uncertainty, bribes include higher costs of transaction compared to taxes (Shleifer & Vishny, 1993; 2003). Overall, we can say that “corruption is a synonym that something has gone wrong in the management of the state” (Rose-Ackerman, 1999; p. 15).

From a theoretical perspective, it is difficult to evaluate or predict the influence of corruption on firm performance from one single existing theory (Van Vu et al., 2018). That is why it is important to consider various theoretical views. Evaluated from the context of TCE, Rose-Ackerman (2008) explains that corruption is recognized from a cost and benefit point of view. This implies that corruption will hinder (facilitate) companies in a situation wherein the costs of the possible bribe deal surpass (lag behind) the benefits gained. Following the logic of the TCE, corruption can either be “grease” or “sand” in the wheels of businesses (Martins et al., 2020).

(6)

6

transaction costs related with the installment of new products (Krammer, 2019). Adding to this it can help firms to get over the disproportionate regulations (Méon & Weill, 2010) or it can save entrepreneurs a lot of valued time by assisting them to overcome long bureaucratic procedures (Lui, 1985). Accordingly, firms that are making use of bribes would face lower bureaucracy and lower costs which eventually may promote growth, enlarge the firm productivity and performance (De Rosa et al., 2015; Vial & Hanoteau, 2010).

On the contrary, corruption often is not restrained to the welfare increasing areas (Cuervo-Cazurra, 2006). Due to corruption firms are confronted with “hidden tariffs” (OECD, 2017). For example, firms have to carry supplementary costs related to contract risks (Shleifer & Vishny, 1993). Generally, payments of bribes do not assure that promised freight will be delivered. As bribery is an illegal activity, investors cannot seek for a solution in the court. This implies there are no legal means to claim for delivery and therefore firms face supplementary contract-related risks (Egger & Winner, 2005). Even if in a situation the bribe is paid and the freight is delivered, businesses will encounter higher costs because the bribe receiver will try to maximize the rents obtained from the bribe (Shleifer and Vishny, 1993). Adding to this, civil servants can retain permissions of certain permits until monetary incentives (bribes) are paid, again resulting in more processing time and a cost increase for the business (De Soto, 1989).

In a condition wherein corruption is present, it is possible that public officials will deliberately postpone the bureaucratic processes and transactions in order to maximize the chances to obtain rents (Myrdal, 1968). The delay in processes could create new obstacles for firms. Kauffman & Wei (2000) explain that firms face higher costs as a result of bribery and that they misuse their available productive time to cope with state officials. Another aspect of corruption is that public officials have power over the decision which firm receives subsidized loans at which rate, therefore this could result in higher capital costs for the businesses (Kauffman & Wei, 2000). These additional costs could decrease the overall performance of firms (Martins et al., 2020).

The additional costs stemming from corruption can result a deterioration of key resources such as the culture and reputation of the business (Hung, 2008; Van Vu et al., 2018). This may decrease or chase profits away from the enterprise and as a consequence can lower the firm performance (Van Vu et al., 2018). The effect of corruption also results in a non-economical way of utilizing resources dedicated to corruption (Cuervo-Cazurra, 2006). The high costs of corruptive payments may restrict the number of available resources that could otherwise be used to expand the firm productive capability (Demsetz, 1967). With this restriction of available resources it is difficult for a business to expand its capabilities. This restriction would therefore not result in an increase of firm productivity and would not result in for example a reduction of the costs per unit or in lower input prices. This lack of expansion could be costly on the long term because of a loss of possible (economic) advantages over competitors (Sirmon & Hitt, 2009) which could result in diminishing firm performance.

(7)

7

et al, 1991, 1993). Due to the additional costs stemming from corruption it is also possible that the firm‟s motivation to innovate can be affected (Hung, 2008; Van Vu et al., 2018). This is a result of changes in the entrepreneurs‟ motivation as they experience more uncertainty and should expect to gain less from their work (Murphy, Shleifer & Vishny, 1991). Overall it results in a situation wherein firms do not adequately appreciate talent, technology and innovation (Van Vu et al., 2018). The uncertainty that is tied to corruption generally sets hurdles for organizations with strong growth commitments to further develop their businesses (Bowen & Clerq, 2008). These difficulties reduce the incentive of companies to participate in more productive activities and hinder the possibility for development processes within the business (Bowen & Clerq, 2008; Fisman & Svensson, 2007). Accordingly, the investment levels change and firms feel discouraged to enhance productivity and invest for growth (Murphy et al., 1993). This causes businesses to participate less in firm development activities and possibly deteriorates the firm performance.

Corruption can also stimulate the entrepreneurs to assign their resources to sectors wherein the detrimental activities are excessively present compared to the productive activities (Aeeni, Motavaseli, Sakhdari & Dehkordi, 2019). In a situation where resources such as talent, capital and technology would be guided towards the more productive activities, the firm investments and innovation rate would probably increase (Acemoglu & Verdier, 1998; Murphy et al, 1991, 1993). This would result in positive effect on the productivity (Acemoglu & Verdier, 1998; Murphy et al, 1991, 1993) and eventually could positively affect the firm performance. Corruption disturbs this process and decreases the productivity of firms

(Lambsdorff, 2003) which would result in a decrease of the firm performance

Corruption leads to a supply of low quality and ineffective public services (Mauro et al., 2019; Rose-Ackerman, 1998) and intensifies the uncertainty businesses have to face (De Rosa et al., 2015; Shleifer & Vishny, 1993). Bardhan (1997) explains that the essential uncertainty of corruption generates wrong firm incentives. This is because corruption enlarges the returns from rent-seeking (and nonproductive) activities compared to the regular (productive) activities (Baumol, 1990). Adding to this, less productive general capital, compared to more productive specific capital, is more flexible and easier to relocate (Henisz, 2000). As a consequence, businesses will decide to invest in this general capital less productive capital (Henisz, 2000). In other words, corruption influences the quality of investments which result in a deterioration of efficiency (Hanousek, Shamshur & Tresl, 2019). As efficiency shows to affect the firm performance (Baik, Chae, Choi & Farber, 2013),

a decrease in firm efficiency could lead to a decrease in firm performance.

(8)

8

and growth of organizations (Myrdal, 1968; Rose-Ackerman, 1996). This indicates that corruption can be identified as a serious matter (Rodriguez, Siegel, Hillman & Eden, 2006) and that it negatively affects the firm performance. Therefore, the hypothesis that will be empirically tested is the following:

Hypothesis 1: The presence of corruption results in worse firm performance, controlled for other industry-level and firm-level characteristics that potentially affect the firm performance.

3. Data and descriptive statistics

This paper uses a dataset sourced from the World Bank “Enterprise Survey”, which is publicly available. The Enterprise Survey is a firm-level survey that is based on a representative sample of the private sector of the economy (The World Bank, 2020a). It provides a wide range of information on the business environment, firm characteristics and performance. Using the existing individual country Enterprise Survey datasets of the World Bank, I compiled a multi-country dataset. My dataset is thus a small extract of the larger pool of existing datasets of the World Bank. I solely used this unbalanced cross-sectional dataset as a sample for this study and it consists of 31,550 observations of 55 countries in four continents. Appendix A1 provides a more detailed description of the countries that are included in the analysis. The period of coverage takes into account three years namely 2017, 2018 and 2019. These years were chosen because previous research has not yet analyzed the data from these specific years. Thereby, three years provide an opportunity to be able take into account the differences within the years. There were no countries for which surveys were conducted in two years. That is the reason why in the multi-country dataset I included all the available survey data for all countries throughout these three years.

The Enterprise Survey includes a wide range of business environment subjects containing crime, access to finance, performance measures, corruption, competition and infrastructure (The World Bank, 2020a). The Enterprise Survey is conducted by private contractors in the form of face-to-face interviews. These interviews are answered by top manager and business owners. In order to be able to answer specific questions regarding labor and sales, the interviewee was sometimes assisted by human resource and accountants (The World Bank, 2020a). The dataset includes information on the companies‟ sectors. This data was also included in the “Enterprise Survey” and is classified following a United Nations industry classification system, namely the four digit level International Standard Industrial Classification of All Economic Activities (ISIC) revision 3.1.

Appendix A2 and A3 provide an overview and frequency of the companies‟ industries, foreign interest (import and export orientation), ownership structure and size that are included in the sample. Taking into account this information it shows that the sample is quite diverse as it covers companies operating in various industries, with different foreign interests (import and export orientation), ownership structures and sizes.

(9)

9

3.1 Firm performance

In this analysis the dependent variable is specified as the natural logarithm (ln) of firm performance. I took the natural logarithm of this variable because it was not normally distributed. After the natural logarithm, firm performance shows to follow a more normal distribution and will confidently provide more valid statistical results. Figure 1 and 2 in appendix A6 display the distribution before and after I took the natural logarithm. Firm

performance (ln) takes a numerical positive value1 and is recorded in the natural logarithm of

gross profit in United Stated Dollars (USD), per company and in the last fiscal year available. Firm performance can be defined as the gross profit and displays the company‟s accounting performance in the last fiscal year (Yasar et al., 2011). The gross profit is calculated as the total revenue minus the costs of goods sold, following Yasar et al. (2011). The costs of goods sold were not available for every company. In order to be able to calculate the gross profit (firm performance) for these companies, I subtracted an alternative measure namely summed costs of goods sold from the total revenue. Appendix A7 gives a more detailed description of the financial variables used for the calculation of gross profit. After I calculated the gross profit for each firm, it was not yet possible to compare the monetary values. In order to make the data comparable across countries, I converted the monetary values for the variable „gross profit‟ in USD. For this conversion I used the official exchange rates (Local Currency United (LCU) per USD, period average) from the World Bank for the years 2015 up to and including 2019. Appendix A8 shows the countries included in the dataset, the LCU of these countries and the official exchange rates I used per fiscal year. I made use of the official exchange rate in the fiscal year in which the total annual sales and costs of goods sold were recorded, therefore there are different official exchange rates reported per LCU. These varying official exchange rates are also visible in appendix A8.

3.2 Explanatory and control variables

The institutional variable corruption is the explanatory variable. This variable is a ratio variable and represents the share of gifts/informal payments to public officials from the total annual sales or estimated total annual value in the interview year. This variable is based on the survey question “It is said that establishments are sometimes required to make gifts or informal payments to public officials to “get things done” with regard to customs, taxes, licenses, regulations, services etc. On average, what percentage of total annual sales, or estimated total annual value, do establishments like this one pay in informal payments or gifts to public officials for this purpose?”. This measure follows Rand and Tarp (2012) and Van Vu et al. (2018).

Firm characteristics, resources and features are the essential components that are influencing the firm performance (Alvarez & Busenitz, 2001; Misanghyi, Elms, Greckhamer & Lepine, 2006). The resource-based view argues that the crucial internal bundle of assets and capacities stimulate the firm performance (Wernerfelt, 1984). Misangyi et al. (2006) found that the differences at the firm-level account for between 31.7 and 44.2 percent of the

(10)

10

variance in the firm performance. Therefore, I included various control variables to control for the heterogeneity in firms and performance factors that could potentially influence the firm performance.

First, I included the numerical variable firm age as a control variable. It can be defined as the number of years since the firm has been established in the given interview year (Van Vu et al., 2018). Stinchcombe (1965) explains that older firms have more experience and therefore have gained more advantages from learning. As a result, older firms, compared to younger firms are not attached to the liability of newness that includes the risk of failure. Accordingly, the older firms benefit from higher performance (Stinchcombe, 1965). Based on the theory of Stinchcombe (1965) firm age is included in the model. Second, I added the numerical variable firm size, which reflects the total amount of full-time employees in the firm. This measurement follows previous literature that commonly uses the total amount of employees as a proxy for the firm size (Meyer, 2001; Van Vu et al. (2018). The motivation for adding this control variable lies in the fact that bigger firms have diversified capacities, formalization of procedures and the possibility to exploit economies of scale and scope (Majumdar, 1997). This makes the execution of procedures more efficient and grants the larger firms to achieve higher performance compared to smaller firms (Penrose, 1959). Third, I incorporated the dichotomous variable exporter in the model. Exporter takes value 1 if more than zero percent of the firm sales were direct exports and 0 if 0 percent of the sales were direct exports in the last fiscal year. Alvarez and López (2005) illustrate that there are firm-level productivity effects from exporting. Bernard, Eaton, Jensen and Kortum (2003) find that exporting firms perform better compared to non-exporting firms. Adding to this, Atkin, Khandelwal and Osman (2017) found that exporting firms enjoy higher profits. Previous literature shows that it is important to control for firm performance effects of international transmission and therefore it is included in the model. Fourth, I took the dichotomous variable

importer into account. It represents whether any of the material inputs, supplies or finished

(11)

11

the model. Lastly, I added the numerical variable skilled workers. Skilled workers indicate the share of highly skilled production workers of the total permanent full-time workers in the firm at the end of last fiscal year. Den Hartog and Verburg (2004) provide evidence that employee skills are positively related to firm performance and for that reason it is needed to control for the differences in skilled labor between the companies (Yasar et al., 2011). Figure 3 and figure 5 in appendix A6 provide the histograms of firm age and firm size. These show that both variables did not follow a normal distribution and implies the data is skewed. Therefore, I took the natural logarithm of both variables. Figure 4 and figure 6 in appendix A6 provides an insight of the distribution of these variables after the natural logarithm.

It is possible that the included control variables will not be able to take into account all the factors that could potentially influence the relationship between corruption and firm performance (ln). With omitted variable bias, there is an explanatory variable missing that is correlated with other explanatory variables and might drive the dependent variable. This omitted variable can be time-invariant and time-variant. When this is excluded from the model, the estimates are not reflecting the true values. The estimation of the explanatory variable can be underestimated or overestimated and as a result could be biased (Wooldridge, 2002). A way to control for the time-variant variables that might affect all the firms included in the sample, is to use year dummies. Wooldridge (2002) underscores that year dummies should be included in the model, to control for the amass changes over time. These year dummies will take into account the amass trends such as certain economic shocks that have happened over the years or other events that are universal across all firms in the sample. A way to control for unrecognized time-invariant variables is to make use of additional fixed effects. Accordingly, it is also needed to include fixed dummies that will take into account other unobserved differences as for example location, culture and industry characteristics.

(12)

12

Table 1: Descriptive statistics

Variables Obs. Mean

Standard

deviation Minimum Maximum

Dependent variable

Firm performance 31,812 5,786,970 109,000,000 0 14,700,000,000

Independent variable

Corruption 24,210 0.846 5.577 0 100

Firm-level control variables

Firm age 31,415 20.497 16.385 0 205 Firm size 31,550 140.974 9430.238 1 1,673,000 Exporter 31,312 0.221 0.415 0 1 Importer 19,084 0.515 0.500 0 1 Foreign 31,372 0.070 0.255 0 1 Capacity utilization 15,152 76.729 22.128 0 100 Skilled workers 16,573 25.755 25.910 0 100

Table 1 shows the descriptive statistics of each of the variable used in this research. The dataset consists of an unbalanced panel of 55 countries, with varying observations per country. Each country has a representative sample of the private sector of the economy (The World Bank, 2020a). The period of analysis is from 2017 up to and inlcuding 2019. Appendix A1 gives more information about the total observations per country and per year for the dependent variable firm performance (ln).

Table 1 shows that especially firm performance and firm size display large variety in the data. This is indicated by the large standard deviation, compared to their mean. Adding to this for these two variables and firm age, there is a large difference between the minimum and maximum value. This large variation in the data is confirmed by appendix A6. Appendix A6 shows the histograms for firm performance, firm age and firm size. Figure 1, 3 and 5 indicate that all three variables have a skewed distribution of the data and therefore do not follow a normal distribution. Especially the histogram for frim performance (figure 1) shows the histogram has a very long tail, which indicates the data is spread out. For the firm performance this variety indicates that some companies are performing very well whereas others are barely surviving. Something that could be related to this difference is the fact that the data includes firms from developed and developing countries. Firms from developed countries might have superior means to operate better and more efficient which results in a higher firm performance. Therefore, there are large performance differences present in de data. It is also possible that this differece is amplfied by the variety in firm age and firm size. It might be possible that older and larger firms perform better compared to younger and smaller firms, due to the experience and expertise gained over the years.

(13)

13

is recorded as higher than 0 percent. On the other hand there are 22,457 cases whererin the business recorded it pays 0 percent of gifts/informal payments. This high number of records of 0 decreases the average significantly. Overall, the inter-firm differences are quite big. A possible reason for the moderate amount of corruption is that the interviewees might have been afraid to truly speak out about the informal payments they had to make.

Figure 1: Frequency of corruption

(14)

14

Table 2: Correlation matrix

Variable (1) (2) (3) (4) (5) (6) (7) (8) (9) (1) Firm performance (ln) 1.000 (2) Corruption -0.045*** 1.000 (3) Firm age (ln) 0.243*** -0.020*** 1.000 (4) Firm size (ln) 0.596*** -0.005 0.261*** 1.000 (5) Exporter 0.236*** -0.012* 0.145*** 0.315*** 1.000 (6) Importer 0.245*** -0.001 0.126*** 0.249*** 0.296*** 1.000 (7) Foreign 0.152*** -0.002 -0.009 0.208*** 0.186*** 0.171*** 1.000 (8) Capacity utilization 0.134*** -0.067*** 0.010 0.128*** 0.068*** 0.057*** 0.029*** 1.000 (9) Skilled workers -0.085*** -0.016* -0.019** -0.116*** -0.077*** -0.083*** -0.049*** 0.005 1.000 *p < 0.10, **p < 0.05, ***p < 0.01

I composed a correlation matrix to examine which variables show to have high correlations and also to check whether there are any problems with multicollinearity between the explanatory and control variables. Table 2 shows this correlation matrix using pairwise deletion. The correlction matrix shows that firm size, compared to the other variables, has the highest correlation with firm performance (ln). This suggest that the bigger a firm is, the more likely that the firm performance (ln) is higher. The only explanatory variable corruption actually shows to have the smallest correlation with the firm performance. This would indicate that corruption might not have a high explanatory power for the firm performance (ln). The correlation matrix displays that all the other variables show to likely affect the firm performance (ln). All values are lower than the approximate treshold of 0.7 (Dormann et al., 2012). Therefore it can be concluded that there are no quastionably high correlations. Adding to this, I conducted a multicolliarity test based on variance inflation factors (VIFs). All values showed to vary between 1.02 and 1.38. These values are lower than 10, which is recognized as the treshold for detecting multicollinearity (Neter, Wasserman, and Kutner, 1996). From the correlations and the VIFs I can conclude that multicollinearity does not show to be an issue.

(15)

15

Figure 3: Leverage-versus-squared-residual plot – outlier detection

(16)

16

Figure 4: Linear prediction plot

The main variables of interest are firm performance (ln) and corruption. Figure 4 shows a linear prediction plot overlaid with a scatterplot including both variables with the 95% confidence interval. The scatterplot shows that there is not a clear relationship between corruption and firm performance (ln). However, taking a look at the linear predicted line, it seems there is a slight negative relationsip between corruption and firm performance (ln). A problem that is often encountered with cross-sectional data is heterokadasticity (Hill, Griffiths and Lim., 2011). This has effects on the linear regression model. Namely, as the size of the monetary unit increases, the higher the uncertainty that is associated with the outcome of the dependent variable (in this case firm performance (ln)) (Hill et al., 2011). This increased amount of uncertainty is presented by a larger error variance, with a larger size of the economic unit (Hill et al., 2011). In other words, the bigger the firm for example, the more difficult it will be to explain the variation in the firm performance (ln) by the variation in the corruption. To check for the presence heteroskadasticity, I performed a Breusch-Pagan test (p=0.000). The test was rejected, indicating there is indeed hetereoskadasticity. To control for heterokadasticity, robust standard errors are used.

4. Empirical specification and methods

4.1 Estimation method

(17)

17

time (the year of interviewing). This is also the main limitation of the data because it is not possible to explore the changes in firm performance over a period of time. Wooldridge (2002) explains that random sampling is sensible for cross-sectional data. In such an arrangement the explanatory variable is handled as random effects along with the dependent variable data (Wooldridge, 2002). In order to be able to conduct an analysis I pooled all the individual firm level data together, which resulted in a so called pooled cross-section. Wooldridge (2002) underscores that also in the pooling of cross-sectional data there is no duplication over a period of time. It is important to take into account the corrections for normality and heterokadasticity, as would have been done in purely cross sectional analysis (Wooldridge, 2002).

For the analysis of the data I chose a Pooled OLS approach. Pooled OLS combines all firm-level observations together in one cross-section and fits a line of best fit to the observations. In my research I am interested in the firm regression results over all countries and not per country. Therefore, the firm-level observations of all countries were pooled which resulted in the application of a pooled model (Hill et al., 2011). Accordingly, I chose the Pooled OLS model as the empirical method for the data analysis. This approach makes is possible to explore the potential effect of corruption on firm performance at the firm-level. Previous research with comparable pooled datasets also made use of the Pooled OLS approach for the analysis of the data (Yasar et al., 2011; Athanasouli, Goujard & Sklias, 2012). Even though the Pooled OLS model does not take into account industry, time and country factors, it is still important to add fixed effects in order to prevent the results from being underestimated or overestimated. Hence, the Pooled OLS regression equation is:

Firm performance (ln)it =β0 + β1 (Corruptionit) +β2 (Firm age (ln)it) + β3 (Firm size (ln)it) +

β4 (Exporterit) + β5 (Importerit) + β6 (Foreignit) +

β7 (Capacity utilizationit) + β8 (Skilled workersit)+

Industryi + Yeart + Countryc + εit (1)

In equation (1) firm performance (ln) takes a numerical value and is recorded in log gross profit in USD, per firm and in the last fiscal year. In hypotheses 1 it is predicted that corruption will negatively affect firm performance. β0 is the constant term. β1 is corruption,

which is measured using the share of gifts/informal payments to public officials from the total

annual sales or estimated total annual value in the interview year. β2 is a control variable for

the age of the firm. β3 is a control variable for the size of the firm. β4 is a control variable for

the potential performance effects of exporting and β5 controls for the potential performance

effects of importing. β6 is a control variable that takes into account the foreign ownership of a

firm. β7 controls for the firm‟s capacity utilization. β8 is a control variable of the performance

(18)

18

4.2 Endogeneity concerns

In the given model there are concerns of endogeneity that need to be taken into consideration. Endogeneity is an issue that might result in biased estimated. My main concern of endogeneity is that there is omitted variable bias. Rosa et al. (2015) point out that a model could be influenced by unrecognized institutional characteristics at the firm-level. Other endogeneity issues that may appear are the bias for firms to misreport the unauthorized activities and the self-selection of firms into bribery as an alternative business efficiency strategy (Ezebilo, Odhuno & Kavan, 2019; Seck, 2020; Williams & Kedir, 2016; Zhou & Peng, 2012). Due to the cross-sectional nature of the data there might be difficult to interpret the estimated results as being causal. Therefore, it is also possible the model can suffer from reverse causality. De Rosa et al. (2015) give the example that this could imply that firms with a high productivity may have a superior capacity to engage in corruption or might be favored targets for extraction of bribes. Reflecting this to my model, firm performance could also potentially influence the share of gifts/informal payments to public officials from the total annual (estimated) sales/value (corruption). The government officials have knowledge about the performance of companies. Accordingly, companies with high performance might be a main target for engagement in corruption. In other words, the firm performance may affect the size of informal payments (corruption). Overall, it is not likely that the variable of corruption is exogenous in this estimation method.

One manner to overcome some of these endogeneity concerns is to use an instrumental variable. A precondition is that the instrument has a high correlation with corruption, low correlation with the regression error term and should not have a direct effect on firm performance (Hill et al., 2011). However, finding a good instrument for corruption that does not directly influence firm performance is, to my awareness difficult to find. Adding to this it is challenging to control for the misreporting of unauthorized activities and the self-selection of firms into bribery. The reasoning for this is that these problems happen in the process of data collection. Potentially these problems of misreporting and self-selection could be solved by improving the survey design or quality of the surveys (Bundi, Varone, Gava & Widmer, 2018). Unfortunately this is not something I can address in this research as this process took place before the data was collected.

(19)

19

Bank2. R&D expenditure is a numerical variable and takes into account the total annual cost

of research and development activities. This estimate provides an opportunity to incorporate an additional firm characteristic that could affect the firm performance and thus may distort the estimates. Rosa et al. (2015) argue that political instability and quality of courts are of some helpfulness to address the omitted institutional characteristics at the firm level. Previous research has shown that political instability negative affects firm performance (Ouédraogo, Ouédraogo & Lompo, 2020). Adding to this evidence shows that better judicial quality (courts) is a significant cause of higher firm performance, which implies a worse judicial quality could deteriorate the firm performance (Chakraborty, 2016). The variables political

instability and quality of courts are based on the answers from a question in the Enterprise

Survey of the World Bank3. Political instability is a dichotomous variable takes to 1 if the firm responds that the political instability is a very severe obstacle or major obstacle to the current operating of a business and 0 if political instability is moderate obstacle, minor obstacle or no obstacle to the current operating of a business. The quality of courts is defined in the same way. These are two estimates of the firm perception of the quality of the institutional environment. These estimates grant the opportunity to include some of the features of the institutional quality that could influence firm performance and thus could potentially bias the estimates. These variables are considered to be potential omitted variables and are included in adapted model that is presented in the result section.

Additionally, this panel takes the firm performance of the last fiscal year which is the year before the interview was held. This might control for reverse causality as this is comparable to taking a lagged value of time variables in panel data. Within my cross-sectional data panel each firm is only observed once at a certain point in time, therefore it is not possible to truly control for reverse causality using a time lagged variable. However, it is possible to check for the potential presence by reverse causality by changing the position of firm performance (ln) and corruption. This will demonstrate whether corruption also is influencing firm performance (ln). The results will be displayed and discussed in the section with the additonal checks.

5. Results

In this part empirical findings of the analysis are presented. Table 3 presents the Pooled OLS regression results and some remarks will be made. Because of endogeneity concerns, I adapted the estimation model by adding three additional control variables. The results are displayed in table 4. The end of this section discusses the potential effects of outliers and provides a check for the potential presence of reverse causality. Table 5 and table 6, respectively shows the results of these tests.

2_{This variable is based on the answer to the question: “During last fiscal year, how much did this establishment} spend on research and development activities, either in-house or contracted with other companies?”.

3

(20)

20

5.1 Pooled OLS regression

Table 3: Hypothesis 1: The effect of corruption on firm performance. Panel Pooled OLS result

Variable (1) (2) (3) Corruption -0.017*** (0.003) -0.013** (0.006) 0.004 (0.004) Firm age (ln) 0.251*** (0.032) 0.097*** (0.030) Firm size (ln) 0.962*** (0.020) 0.963*** (0.019) Exporter 0.317*** (0.054) 0.187*** (0.052) Importer 0.198*** (0.052) 0.391*** (0.049) Foreign 0.073 (0.078) 0.311*** (0.072) Capacity utilization 0.007*** (0.001) 0.004*** (0.001) Skilled workers 0.000 (0.001) 0.001 (0.001)

Industry fixed effects Yes Yes Yes

Year fixed effects Yes Yes Yes

Country fixed effects No No Yes

Intercept 13.655*** (0.151) 8.344*** (0.213) 7.787*** (0.349) Observations 20,658 5,726 5,726 R2 0.111 0.498 0.607 Adjusted R2 0.103 0.485 0.593 *p < 0.10, **p < 0.05, ***p < 0.01 Robust standard errors in parentheses

Table 3, presents the effect of corruption on firm performance (ln). The models presented are all either linear or log models. In order to be able to interpret the log-linear model the coefficient is multiplied by 100. Model 1 shows a negative at the 1 percent significant effect of corruption on the firm performance. For a 1 percent increase in the share of gifts/informal payments from the total annual (estimated) sales there is a 1.7 percentage decrease in firm performance.

(21)

21

positive effect of 19.8 percent on the firm performance. These results therefore confirm the positive effects of exporting and importing and are in accordance with previous research (Atkin et al., 2017; Bernard et al., 2003).

In model 3 I also added country fixed effects, next to the industry and year fixed effects. Excluding country fixed effects can have serious consequences, either by biasing the results or masking the effect of the individual observations (Kaufmann & Wei, 2000). For example it is possible that there is variety in the characteristics of the country (e.g. quality of regulation) that could be related with both firm performance and corruption (Kaufmann & Wei, 2000). After adding the country fixed effects, the results show to change significantly. An apparent effect of adding the country fixed effects is that corruption now shows to have a minor positive but insignificant effect on the firm performance. The sign of the coefficient has changed as well as the magnitude. Another noticeable change is the coefficient of firm age. This effect has been reduced significantly. In model 2, a 1 percent increase in firm age shows to increase the firm performance by approximately 0.3 percent. Whereas, looking at model 3, a 1 percent increase in firm age shows to increase the firm performance by approximately 0.1 percent. This finding supports theory by Stinchcombe (1965) explains that older firms have more experience and therefore have gained more advantages from learning and are, compared to younger firms, not suffering from the liability of newness. Also the variable foreign, shows significantly different results in model 3 compared to model 2. First, the coefficient of foreign changes in magnitude and also changes from insignificant to significant. For a firm for which the foreign ownership share goes from (less than) 50 percent to more than 50 percent, the firm performance increases with 31.1 percent. These large differences in coefficients could be driven by the addition of the country fixed effects. It is possible that the large variation in the firm age and foreign ownership share of companies is explained by the difference in country characteristics. In other words, it is probable that in model 2 the effect of the variable firm age was overestimated and that the effect of the variable foreign was masked by unincorporated country characteristics. Lastly, in both models 2 and 3 no significant effect is found for high skilled workers, this finding does not verify results Den Hartog and Verburg (2004), who find support for a positive correlation between employee skills and firm performance.

Adding more variables to the model increases the R2, therefore I also included the

adjusted R2. From the adjusted R2 I can conclude that explanatory power of the three models

differ a lot. Model 1 only includes the explanatory variable and shows that approximately 10.1 percent of the variance in the dependent variable is explained by the variance in the explanatory variables included in the model. Model 3 includes all the control variables and the fixed effects shows to be the strongest model with an explanatory power of 0.59.

(22)

22

firm performance and that corruption is not the main driving factor. Looking at the results of full model 3 that includes all the controls and fixed effects, I can say that there is no evidence that corruption affects the firm performance. In summary, this analysis finds no support for hypothesis 1. The presence of corruption does not result in worse firm performance, controlled for other industry-level and firm-level characteristics that potentially affect the firm performance.

5.2 Adapted estimation

As discussed before, in this research there are concerns of endogeneity that need to be taken into consideration. The previous estimation method does not extensively take into account these concerns. For that reason I introduced an adapted estimation method that re-estimates the models. With this adapted estimated I introduced additional variables in order to try to control for potential omitted variables bias. Three variables are added to the model namely, R&D expenditure, political instability and quality of courts. This is the only adaptation to the model, the rest of the variables remain as they are. The results of the adapted estimation can be found in the table 4.

Table 4: Hypothesis 1 - Potential omitted variables: The effect of corruption on firm performance. Panel Pooled OLS result

Variable (1) (2) (3) (4) (5) (6) Corruption -0.017*** (0.003) -0.013** (0.006) 0.004 (0.004) -0.035** (0.015) -0.032** (0.015) -0.031** (0.014) Firm age (ln) 0.251*** (0.032) 0.097*** (0.030) 0.023 (0.069) 0.023 (0.069) 0.035 (0.071) Firm size (ln) 0.962*** (0.020) 0.963*** (0.019) 0.865*** (0.053) 0.870*** (0.054) 0.878*** (0.055) Exporter 0.317*** (0.054) 0.187*** (0.052) 0.034 (0.115) -0.011 (0.116) -0.012 (0.118) Importer 0.198*** (0.052) 0.391*** (0.049) 0.126 (0.114) 0.124 (0.114) 0.092 (0.116) Foreign 0.073 (0.078) 0.311*** (0.072) 0.237 (0.155) 0.236 (0.157) 0.202 (0.158) Capacity utilization 0.007*** (0.001) 0.004*** (0.001) 0.006*** (0.002) 0.006*** (0.002) 0.006** (0.002) Skilled workers 0.000 (0.001) 0.001 (0.001) 0.000 (0.003) -0.000 (0.003) -0.000 (0.003) R&D expenditure (ln) 0.149*** (0.041) 0.148*** (0.043) 0.160*** (0.045) Political instability -0.032 (0.111) -0.040 (0.112) Courts -0.064 (0.116)

Industry fixed effects Yes Yes Yes Yes Yes Yes

Year fixed effects Yes Yes Yes Yes Yes Yes

(23)

23 Intercept 13.655*** (0.151) 8.344*** (0.213) 7.787*** (0.349) 9.711*** (0.535) 9.797*** (0.565) 9.752*** (0.570) Observations 20,658 5,726 5,726 1,086 1,061 1,031 R2 0.111 0.498 0.607 0.687 0.693 0.697 Adjusted R2 0.103 0.485 0.593 0.625 0.630 0.633 *p < 0.10, **p < 0.05, ***p < 0.01 Robust standard errors in parentheses

Table 4 shows the results from the Pooled OLS regression after adding potential omitted variables. Columns 1-3 in table 4 are identical to columns 1-3 in table 3, whereas column 4-6 are newly displayed results. The latter are re-estimations of column 3, but with added variables. Because I added one firm variable and two institutional variables to the regression, the number of observations decreased significantly.

The full model, as displayed in model 6, shows that the explanatory variable corruption is now negative at the 5 percent significant instead of slightly positive insignificant (as can be seen in model 3). This means that, after adding a selection of possible omitted variables, a 1 percent increase in the share of corruption decreases the firm performance by 3.1 percent. This is in line with previous research that did not find evidence for corruption as “grease the wheels” of business argument (Batra, Kaufmann & Stone, 2003; Kaufmann & Wei, 2000) and it thus supports the corruption as “sand in the wheels” of business argument (Fisman & Svensson, 2007; Hanousek & Kochanova, 2016; Martins et al., 2020). The added control for R&D expenditure has a positive at the 1 percent significant effect on the firm performance. Specifically, a 1 percent increase in R&D expenditure increases the firm performance by approximately 0.2 percent. This result complements the conclusions of previous research (Chan et al., 2001; Eberhart et al., 2004), in the sense that R&D has a positive effect on the firm performance. Looking at political instability and quality of courts, the variables show to have the expected negative sign, however, show insignificant results. These results do not verify the findings by Ouédraogo et al. (2020) and Chakraborty (2016), who find support for a negative correlation between on the one hand political instability and firm performance and on the other hand the judicial quality (courts) and firm performance. Regarding the control variables most of them keep the expected sign, however, do decrease in size and are not significant anymore.

The adapted estimation shows that the explanatory power varies even more compared

to the previous estimation. The values of the adjusted R2 range from approximately 0.1 to 0.6.

The biggest change in adjusted R2 is from model 3 to 4. This implies that the added control

variable R&D expenditure shows to add most explanatory power to the model. Whereas, the added control variables political instability and courts show to increase the explanatory power of the models with only 0.005 and 0.003 respectively. This indicates that these two variables do not seem to add much to the models.

(24)

24

explanatory power to the model. As a result only R&D expenditure is helpful in partially addressing endogeneity concerns stemming from omitted variable bias. Model 6 is the prioritized model as it includes all control variables, fixed effects and potential omitted variables and thus gives less biased results. It indicates a negative effect of corruption on firm performance. Nevertheless, the coefficient should be interpreted with caution because the total amount of observations decreased significantly (5,726 versus 1,031). Appendix A9 gives an overview of the change of observations per country between the third and sixth model of this estimation. This variety in observations is sizeable due to the fact that there is a lot of missing data for the R&D expenditure. As a result the other control variables have lost their significance. In summary, I can say that there is evidence that corruption negatively affects the firm performance and that the coefficient should be interpreted with caution. As a result this analysis find finds support for hypothesis 1. The presence of corruption does result in worse firm performance, controlled for other industry-level and firm-level characteristics that potentially affect the firm performance.

5.3 Additional checks

In the previous section I used different methods in order to analyze whether corruption influences firm performance. In this section I will report two additional checks on this relationship. First, outliers will be excluded. As has been discussed in the descriptives section, potential outliers were identified with the help of the Cook‟s distance. These outliers could have potentially affected the findings and could have distorted the actual results. As a first test I excluded these outliers. A total amount of 311 cases had a higher cook‟s distance than the cut-off value and were therefore deleted from the dataset. After this deletion I performed a Pooled OLS regression for which the results are shown in table 5. Second, to check for the potential presence of reverse causality, I changed the position of firm performance and corruption in the regression. The results are displayed in table 6.

(25)

25 R&D expenditure (ln) 0.124*** (0.032) 0.119*** (0.032) 0.132*** (0.033) Political instability 0.009 (0.098) 0.022 (0.102) Courts -0.079 (0.107)

Country fixed effects No No Yes Yes Yes Yes

Intercept 13.610*** (0.148) 8.242*** (0.204) 7.734*** (0.295) 9.115*** (0.483) 9.143*** (0.504) 9.193*** (0.503) Observations 20,347 5,415 5,415 1,030 1,006 979 R2 0.114 0.557 0.672 0.741 0.749 0.749 Adjusted R2 0.105 0.546 0.660 0.689 0.696 0.695 *p < 0.10, **p < 0.05, ***p < 0.01 Robust standard errors in parentheses

(26)

26

Table 6: Reverse causality: The effect of corruption on firm performance. Panel Pooled OLS result Variable (1) (2) (3) (4) (5) (6) Firm performance (ln) -0.112*** (0.019) -0.081** (0.033) 0.027 (0.033) -0.104*** (0.040) -0.088** (0.039) -0.087** (0.039) Firm age (ln) -0.110 (0.079) 0.037 (0.079) 0.249* (0.143) 0.184 (0.142) 0.202 (0.143) Firm size (ln) 0.155*** (0.058) 0.023 (0.054) 0.053 (0.077) 0.053 (0.074) 0.059 (0.080) Exporter -0.181 (0.135) -0.102 (0.121) 0.006 (0.209) 0.029 (0.201) 0.011 (0.207) Importer 0.156 (0.130) 0.169 (0.146) 0.114 (0.216) 0.174 (0.215) 0.144 (0.203) Foreign -0.140 (0.155) -0.281* (0.161) 0.037 (0.192) -0.016 (0.185) 0.011 (0.187) Capacity utilization -0.013*** (0.003) -0.011*** (0.003) -0.004 (0.004) -0.007** (0.004) -0.006* (0.004) Skilled workers 0.002 (0.002) 0.000 (0.002) -0.003 (0.003) -0.003 (0.003) -0.002 (0.003) R&D expenditure (ln) 0.024 (0.057) -0.014 (0.048) -0.011 (0.053) Political instability 0.265 (0.171) 0.216 (0.232) Courts 0.372 (0.374)

Country fixed effects No No Yes Yes Yes Yes

Intercept 2.007*** (0.309) 2.063*** (0.482) 4.656* (2.372) -0.049 (0.771) 0.238 (0.737) -0.297 (0.818) Observations 20,658 5,726 5,726 1,086 1,061 1,031 R2 0.019 0.082 0.142 0.481 0.540 0.547 Adjusted R2 0.010 0.059 0.112 0.378 0.446 0.451 *p < 0.10, **p < 0.05, ***p < 0.01 Robust standard errors in parentheses

(27)

27

6. Discussion and conclusion

This research contributed to the increasing debate on the effect of corruption by providing an opportunity to estimate the effect of corruption on the firm economic performance for various types of businesses in a diverse set of countries (Yasar et al., 2011). Previous research has found very differing, ambiguous and sometimes contradictory conclusions (Ashyrov & Masso, 2020; De Rosa et al., 2015, Gooroochurn & Görg, 2015; Van Vu et al., 2018). In order to try to improve the understanding of corruption on firm performance, I investigated the research question: what is the effect of corruption on firm performance? For this purpose I analyzed firm-level data from the World Bank “Enterprise Survey”. In line with previous research (Fisman & Svensson, 2007; Hanousek & Kochanova, 2016; Martins et al., 2020), I tested whether corruption has a damaging effect on the firm performance.

Analyzing whether corruption increases or reduces firm performance, the above analysis provides the results of different Pooled OLS models. The introduced estimation is controlling for industry and year fixed effects and later also adds country fixed effect. The results are varying between the models. The model including only the industry and year fixed effects provide evidence that corruption negatively influences the firm performance of organizations, which would support hypothesis 1. On the other hand when the model includes industry, year and country fixed effects the coefficient sign changes and becomes insignificant. This implies that the model without the country fixed effects showed to exclude too many country differences that potentially could have influenced the difference in firm performance. This implies that the effect of corruption was probably inflated by unincorporated individual country differences. The country fixed effects therefore show to explain the biggest part of the variance in firm performance. The results indicate that there are other country characteristics that affect the firm performance and that corruption was not the main driving factor. Therefore, the findings of this estimation did not support hypothesis 1.

(28)

28

The empirical implication of this research is that it confirms that corruption has a deteriorating effect on firm performance and it could therefore act as “sand in the wheels” of business by disintegrating key resources and increasing costs (Fisman & Svensson, 2007; Van Vu et al., 2018). It is, however, necessary to recognize that is it very difficult to isolate the effect of corruption on firm performance, as there are surely many other factors that need to be considered. Therefore, it is crucial to keep in mind the broader environmental circumstances. Especially factors at the country and institutional level need to be taken into account. This is needed in order to gain a better understanding of how these circumstances interact with corruption. Adding to this, such factors play an important role in determining the effect of corruption on firm performance and will definitely contribute to gaining a more profound understanding of corruption and the consequences it can cause.

The policy implication of this research is that it shows that corruption is a serious issue that requires more attention. This research shows that corruption is harming firm performance. As firms form the backbone of the economy, the harming firm level effect contributes to a wider even more severe country level effect (Mauro, 1995, 1996, 1998; Mo, 2001; Tanzi & Davoodi, 1997, Wei, 1999). Therefore, policymakers need to become aware that many bureaucrats and firms still show to participate in corruption and that their behavior is the source of these damaging effects. Thereby, the fact that a sizeable amount of scientific research has dedicated (and still is devoted) to this topic shows that fighting corruption is a subject that is of great importance. Generally, the institutions should be re-arranged in order to discourage corruption and the introduction of new mechanisms should depress the opportunities for engagement in corrupt (Williams & Kedir, 2016). Thereby, policymakers should thoroughly think about the enforcement of policies that will actively fight corruption that incorporate the view of the diminishing effects of corruption. This could contribute to this policy becoming a mediator in the negative connection between corruption and firm performance. Overall, the necessity to continue the attempts to get rid of corruption still remains.

(29)

29

there are many inter-country differences. Some firms operating in a country report a high level of corruption, whereas, others do not report corruption at all. Therefore the effect on the pooled level could show biased results and could introduce difficulties with interpretation. Generally, it is demanding to control for all these differences, however, this research shows it is important to include some country-level variables in the analysis. Third, because of the cross-sectional nature of the data it is difficult to control for potential reverse causality. Thereby, the quality of the data also is not supporting the search for good instrumental variables that could help in controlling for reverse causality.

Future research could focus on examining the effect of corruption on the firm performance by also taking into account the numerous separate country differences. Controlling for these differences might provide less biased and more consistent results. Thereby, it would also be helpful if the quality of data regarding corruption would enhance. This would result in less missing observations, less misreporting and therefore would provide a better picture of the corruption that is present in a certain country. Future research could also take into account different measures of corruption and different measures of firm performance. In this research I only took into consideration one type and maybe some types of corruption are more/less damaging than others. Adding to this, these diverse types of corruption might have type-specific effects on different forms of firm performance. However, after all this research only discusses the effect of corruption and not the causes. Future research could look at what kind of formal institutional shortcomings actually result in the predominance of corruption in certain countries. This would provide new and perhaps more practical insights for governments. Nevertheless, this would need estimates to reshape the informal institutions (for example the beliefs, norms and values) as regards to the general acceptance of corruption (Williams & Kedir, 2016). Eventually, this should decrease the imbalance between formal and informal institutions (such as corruption) in countries.

(30)

30

7. References

Acemoglu, D., & Verdier, T. (1998). Property Rights, Corruption and the Allocation of Talent: A General Equilibrium Approach. The Economic Journal, 108(450), 1381–1403.

Aeeni, Z., Motavaseli, M., Sakhdari, K., & Dehkordi, A. M. (2019). Baumol‟s theory of entrepreneurial allocation: A systematic review and research agenda. European Research on

Management and Business Economics, 25(1), 30–37.

Aidt, T. S., Dutta, J., & Sena, V. (2008). Governance regimes, corruption and growth: theory and evidence. Journal of Comparative Economics, 36(2), 195–220.

Aitken, B., & Harrison, A. (1999). Do domestic firms benefit from foreign direct investment? Evidence from Venezuela. American Economic Review, 89(3), 605–618.

Alvarez, R., López, R. (2005). Exporting and performance: evidence from Chilean plants.

Canadian Journal of Economics, 38 (4), 1384–1400.

Alvarez, S. A., Busenitz, L. (2001). The entrepreneurship of resource-based theory. Journal of

Management, 27(6), 755–775.

Ashyrov, G., & Masso, J. (2020). Does corruption affect local and foreign-owned companies differently? Evidence from the BEEPS survey. Post-Communist Economies, 32(3), 306–329.

Athanasouli, D., Goujard, A., & Sklias, P. (2012). Corruption and firm performance: Evidence from Greek firms. International Journal of Economic Sciences and Applied

Research, 5(2), 43-67.

Atkin, D., Khandelwal, A. K., Osman, A. (2017). Exporting and firm performance: evidence from a randomized experiment. Quarterly Journal of Economics, 132(2), 551-615.

Bahoo, S., Alon, I., & Paltrinieri, A. (2020). Corruption in international business: A review and research agenda. International Business Review, 29(4), 101660.

Baik, B., Chae, J., Choi, S., & Farber, D. B. (2013). Changes in operational efficiency and firm performance: A frontier analysis approach. Contemporary Accounting Research, 30(3), 996-1026.

Bardhan, P. (1997). Corruption and development: a review of issues. Journal of Economic

Literature, 35(3), 1320–1346.

(31)

31

Baumol, W. J. (1990). Entrepreneurship: Productive, unproductive and destructive. Journal of

Political Economy, 98(5), 893–921.

Bernard, A. B., Eaton, J., Jensen, J. B., & Kortum, S. (2003). Plants and productivity in international trade. American Economic Review, 93(4), 1268–1290.

Bowen, H. P., & Clercq, D. D. (2008). Institutional context and the allocation of entrepreneurial effort. Journal of International Business Studies, 39(4), 747–767.

Bundi, P., Varone, F., Gava. R., & Widmer, T. (2018). Self-Selection and Misreporting in Legislative Surveys. Political Science Research and Methods, 6(4), 771–789.

Chakraborty, P. (2016). Judicial quality and regional firm performance: The case of Indian states. Journal of Comparative Economics, 44(4), 902-918.

Chan, L. K., Lakonishok, J., & Sougiannis, T. (2001). The stock market valuation of research and development expenditures. The Journal of Finance, 56(6), 2431-2456.

Corporate Finance Institute (20202). Capacity Utilization - Definition, Example, and

Economic Significance. Retrieved 7 October 2020, from

https://corporatefinanceinstitute.com/resources/knowledge/economics/capacity-utilization/#:~:text=Example%20of%20Capacity%20Utilization,100%25%2C%20indicating %20full%20capacity.

Cuervo-Cazurra, A. (2006). Who cares about corruption? Journal of International Business

Studies, 37(6), 807-822.

De Rosa, D., Gooroochurn, N., & Görg, H. (2015). Corruption and productivity: Firm-level evidence. Journal of Economics and Statistics (Jahrbuecher Fuer Nationaloekonomie Und

Statistik), 235(2), 115–138.

De Soto, H. (1989). The Other Path. New York: Harper & Row.

Demsetz, H. (1967) Forward a Theory of Property Rights. The American Economic Review, 57(2), 347-359.

Den Hartog, D. N. & Verburg, R. M. (2004). High performance work systems, organizational culture and firm performance. Human Resource Management Journal, 14(1), 55-78.

The effect of corruption on firm performance