Economic analysis of corruption at the company level

Bąk, Paulina Article Economic analysis of corruption at the company level Central European Economic Journal (CEEJ) Provided in Cooperation with: Faculty of Economic Sciences, University of Warsaw Suggested Citation: Bąk, Paulina (2020) : Economic analysis of corruption at the company level, Central European Economic Journal (CEEJ), ISSN 2543-6821, Sciendo, Warsaw, Vol. 7, Iss. 54, pp. 186-204, https://doi.org/10.2478/ceej-2020-0015 This Version is available at: https://hdl.handle.net/10419/324526 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ ISSN: 2543-6821 (online) Journal homepage: http://ceej.wne.uw.edu.pl To cite this article Bąk, P. (2020). Economic analysis of corruption at the company level. Central European Economic Journal, 7(54), 186-204. DOI: 10.2478/ceej-2020-0015 To link to this article: https://doi.org/10.2478/ceej-2020-0015 Economic analysis of corruption at the company level Paulina Bąk Open Access. © 2020 P. Bąk, published by Sciendo. This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. Paulina Bąk Faculty of Economic Sciences, University of Warsaw, Warsaw, Poland, corresponding author: [email protected] Economic analysis of corruption at the company level Abstract Corruption is a factor that affects a company severely either directly or indirectly. It may have a positive or negative impact on the economic situation of the company. This article verifies the hypothesis about the corporate perception of corruption as an obstacle to business performance. It also identifies which factors do have a substantial effect on the perception of corruption by companies. The study was carried out using the logit model. The data used were obtained from the Business Environment and Enterprise Performance Survey (BEEPS) database for 2016. Keywords corruption | companies | extortion JEL Codes G34, G38 1 Introduction Corruption is a factor that affects a company severely either directly or indirectly. It may be a positive or negative impact on the economic situation of the company. According to the 2018 Global Economic Crime and Fraud Survey, 49% of respondents admit that their company has been the victim of bribery and this percentage was 36% in 2016. According to Didier Lavion, the problem is not the emergence of corruption, but the employees’ answer to the question whether they fear that corruption may be affecting their organisation. The largest percentage of companies facing corruption risks happens in Africa, where as many as 62% of respondents answered yes to the question whether their organisation has a bribery problem. This percentage is 47% for Eastern Europe, for Western Europe – 45% and 46% for Asia. Additionally, 52% of respondents indicate ‘internal actors’ as the main perpetrators of corruption and this figure is 40% for ‘external actors’. Incidences of corruption are also recorded in Poland. As many as 200 investigations were conducted according to a report by the Central Anticorruption Bureau (CBA) in 2016. When compared, there were 225 investigations in 2015 and 181 in 2014. In 2016, 281 trial preparations were initiated. One example of such a trial is the incident of a private medical company from Silesia, which tried to obtain a private, beneficial contract from the National Health Fund. To clinch the deal, the amount of PLN 515,000 as bribery was to be handed to a member of the Council of the National Health Fund. Finally, the CBA (Centralne Biuro Antykorupcyjne) detained four people involved in the bribery in this case. Later in the investigation, it was additionally found that the company’s representatives, in exchange for winning the tender, offered financial benefits to the director of the Independent Public Clinical Hospital No. 5 of the Medical University of Silesia in Katowice (CBA, 2016). Corruption is an important problem in the field of economic science that requires careful analysis. First, I will explore how corruption has changed between 2012 and 2016. Additionally, the chart of the CPI index, i.e. the Corruption Perception Index, will be presented. It is a global index of the perceived levels of public sector corruption, which uses a scale from 0 (highly corrupt) to 100 (not corrupt). The index is prepared by Transparency International (Figure 1). The above maps show how the CPI index changed between 2013 and 2016. Red represents the highest levels of corruption in the country, while yellow – the lowest CEEJ • 7(54) • 2020 • pp. 186-204 • ISSN 2543-6821 • DOI: 10.2478/ceej-2020-0015 188 levels. When analysing the maps, it is observed that the red colour definitely dominates. The most corrupt regions are Africa, South America and Asia. Slight changes are observed in the fight against corruption in China and India. In these two countries, the level of the CPI index had been rising but fell considerably in 2016. Despite this drop, China still occupies a high place in global corruption rankings. Additionally, China openly admits to the problem of corruption within its borders and implements anti-corruption programs to help mitigate the risk of bribery. Despite such severe corruption issues, China is still attractive to entrepreneurs, and the Chinese trade market is one of the most dynamically growing markets globally. Russia is another important leader of corruption rankings. According to the Russian INDEM report, the inhabitants of Russia spend almost USD 12 billion annually on bribes, while companies pay USD 33 billion in kickbacks every year. According to expert analysis, corruption should be lower in regions where the inhabitants are richer and where government officials earn more. But, this theory does not hold true for Russia. The lowest level of corruption is noted in the north-western area, which cannot be classified as wealthy; on the contrary, it is associated with towns without prospects. The most corrupt areas are the Caucasus and rich Siberia, which are territories with 1 www.transparency.org (access 04/05/2019). the main deposits of oil, gas and other raw materials and the extraction is extremely profitable. Worldwide, government leaders are aware of the presence of corruption risk and declare that they will take suitable action to address it. Unfortunately, these are empty promises and such declarations are made to meet the needs of the public and the banks. Figure 2 shows the relationship between corruption and economic growth. The CPI index (horizontal axis) was used for the analysis, where 1 is indicative of high corruption, and 10 – low corruption. The vertical axis shows the indicator of social welfare, where the lowest values mean the least developed economy, and the higher values – a developed country. A positive correlation can be observed between the two indicators. Higher economic growth translates into a higher CPI, i.e. a lower level of corruption in the country. The figure reveals that Africa, which is the least developed, remains the world’s most corrupt region. A similar situation is observed in Asia, but slightly better. Among the countries without a corruption problem are Germany, Japan, Singapore, Norway, i.e. the top-ranking countries in the category of highly developed countries in the world. Based up the literature, we can say that the corruption has a negative impact on the effectiveness of the economic system and social welfare. The corruption tax differs from other taxes paid to the CPI 2013 CPI 2014 CPI 2015 CPI 2016 Fig. 1. CPI index between 2013 and 2016. Source: Transparency International.1 CEEJ • 7(54) • 2020 • pp. 186-204 • ISSN 2543-6821 • DOI: 10.2478/ceej-2020-0015 189 state. In the event of bribery, the payer cannot be sure that the service they pay for will be delivered. The corruption contract is oral; therefore the person accepting the payment may refuse to perform the service after taking the money. It may also happen that the performance of the service will not be within their capabilities resulting in inferior services. Moreover, in the worst-case scenario, the contractor may demand further payments for the service not yet performed, claiming that they are necessary to complete the job (The Economics 2011) Officials may differentiate the amount of corporate corruption tax. One of the methods used is to divide companies into the categories namely that are more inclined to pay bribes and that less inclined to do so. The willingness and ability to pay the tax depends on the cash resources at the disposal of companies. Some of them do not face barriers in paying the required amount, while others may not be in such a good financial position. Another method is the division of the industries based on their operations and their effectiveness. Therefore, businesses may view corruption as an obstacle to doing business in varying degrees. The objective of this article is to investigate whether companies perceive corruption as an obstacle to business. I will attempt to analyse what factors have a significant impact on the negative perception of 2 https://www.economist.com/graphic-detail/2011/12/02/ corrosive-corruption (accessed 07/16/2020). corruption by companies. The study was carried out using the logit model. The data were obtained from the Business Environment and Enterprise Performance Survey (BEEPS) database for 2016. The structure of the article has been constructed to meet the purpose of this study. The article consists of three sections. The first section presents the definition of the term ‘corruption’ and a statistical analysis of the problem along with examples of bribery cases around the world. The second section offers a review of the literature devoted to research on corruption at the microeconomic and macroeconomic scales. The third section describes the empirical study and its results, as well as the conclusions. 2 The definition of corruption The Dictionary of the Polish Language provides one of the simplest definitions of corruption, which states that it is ‘accepting bribes by government officials or officers’ (Akademia Języka Polskiego, 2007). A slightly broader description can be found in the Encyclopaedia, which defines that ‘corruption is bribery, accepting or demanding a financial or personal gain by a person who performs a public function in exchange for fulfilling a duty or violating the law’ (Bieńko, 2007). A similar view is expressed by the Italian journalist Carlo Alberto Brioschi, who believes that corruption is the behaviour of a person who holds Fig. 2. Correlation between corruption and economic growth. Source: The Economist based on Transparency International, UN.2 CEEJ • 7(54) • 2020 • pp. 186-204 • ISSN 2543-6821 • DOI: 10.2478/ceej-2020-0015 190 a public office, consisting in accepting money, gifts and other tangible and intangible benefits in return for rendering a favour or making a decision in favour of the beneficiary (Brioschi, 2017). The definition is given in the journalist’s book in which he outlines the history of corruption and concludes that the problem of bribery has existed since the existence of public offices. Therefore, the question arises whether state institutions should be liquidated or their influence on the activities of companies should be strongly limited. The word corruption has a broad sense that is not just related to bribery. The author of the book Prawne kryminologiczne i kryminalistyczne aspekty łapówkarstwa [Legal criminological and forensic aspects of bribery], Tadeusz Chrustkowski, argues for a broad definition of corruption including, e.g. nepotism and blackmail (Chrustowski, 1985). The corruption problem is not a new phenomenon and has existed for thousands of years. Analysing the etymology of the word ‘corruption’, it is learnt that it comes from the Latin term corruptio which means corruption, bribery. Certain corruption crimes have always plagued societies, such as bribery in customs offices during the ancient Roman era. Nowadays, there are many other manifestations and forms of corruption. The definition of corruption differs across the disciplines of economics and psychology. According to Andvig, corruption is a violation of the principle of fair play in the society in order to achieve a personal gain (Andvig, 2006). The concept defined in this way is most often used in psychological or social sciences. In the field of economics, the most frequently used definition of the World Bank is that it is an abuse of public office to pursue private needs. Additionally, the World Bank has complemented the definition by a special division of corruption into administrative and political fraud (Pradhan 2000) Political corruption happens at the stage of drafting new laws itself. Government officials, as well as those in power, who have been authorised to work on new laws, engage in this practice to meet their own needs or the needs of businesses in return for bribes. Unlike political corruption, administrative corruption occurs when legal regulations already exist and must be violated to achieve one’s goals. In this case, companies are willing to pay the corruption tax to increase their prosperity, but, this is a waste at the level of society. 3 A review of the literature on corruption research 3.1 Microeconomic research on corruption Numerous studies on corruption can be found in the literature. But, in most cases, the analysis focuses on the effects of corruption rather than the factors that significantly influence the emergence of the problem. Understanding the mechanism of corruption, as well as the circumstances in which such offences are committed, can help in establishing an effective anticorruption and state development policy. The first paper to investigate the determinism of corruption at the company level was the Hellman and Schankerman study in the year 2000. Here, empirical data were obtained from the BEEPS database and then estimated using an econometric model. The authors conclude that in order for the market reform to be successful, state capture must be limited (Hellman et al., 2000). Another study by Hellman and Kaufman from 2004 shows that there is an inverse relationship between trust in the state and the level of corruption. Company directors are more likely to resort to paying bribes when they have less faith in the national government’s decisions. Such firms avoid the courts and pay no taxes, except for the corruption tax (Hellman and Kaufmann, 2004). It can be definitively concluded that corruption is treated as a tax by the eyes of companies. Such a financial claim is associated with the need for secrecy as well as with uncertainty as to whether one of the parties is going to fulfil the terms of the contract. A study by Shleifer and Vishny from 1993 identified a new type of corruption, namely corruption coupled with theft. Their article discusses a study on the assessment of incidences of corruption determined by contractual compliance. The first correlation that they investigate is the mutual fulfilment of the agreement, i.e. when one party pays the bribe and the other party meets the stipulated requirements. The second correlation is the failure of bribe acceptor to implement the contract terms. Moreover, they may demand further payments for the rendered services in this situation. For discussions and analyses, two groups of countries, the West and the former USSR, were juxtaposed. In Western countries, both sides of the corruption show benefit at the price of social welfare. In the states of the former USSR, only the CEEJ • 7(54) • 2020 • pp. 186-204 • ISSN 2543-6821 • DOI: 10.2478/ceej-2020-0015 191 government official gains profit from the contract at the expense of the prosperity of society and the company (Shleifer and Vishny, 1993). Corruption has negative effects on business investment. A 2007 study by Fisman and Svensson proved that an increase in the corruption tax translates into a reduction of firm growth. Surprisingly, the tax increase results in a lesser slowdown of development (difference of three percentage points). Additionally, they observed that the role of corruption along with a decrease in the importance of taxes when new control variables were taken into account (Fisman and Svensson, 2007). Hellman and Johnson show the relationship between secure property rights and the level of investment. More secure laws result in a higher level of investment, which means less corruption in corporate policy. Furthermore, the authors demonstrated that corruption negatively affects secure relations between states and businesses. The bribery problem causes a stream of monetary extortion and is marked by high uncertainty (Hellman, 2003). Research done by Braguinsky and Mityakov showed the importance of foreign capital share for a firm’s levels of transparency. They describe this issue with the example of the revenues and market value of company vehicles in a Russian company with foreign capital. According to their findings, foreign-capital companies are more transparent than companies with only domestic capital (Braguinsky and Mityakov, 2015). A significant factor contributing to the emergence of corruption is also poor education and high inequality in the country. This was proved by the research conducted among the population by Hunt and Laszlo in 2012 (Hunt, 2012). Further, another study by Hunt from 2007 supplemented these results with other determinants such as misfortune in life, loneliness and poor health. Individuals with these characteristics are more prone to bribery. Research shows that officials may differentiate corruption tax prices depending on the taxpayer’s qualities (Hunt, 2007). On the company level, they are also diversified allowing bribe-acceptors to differentiate prices. This is corroborated by Svensson’s research in 1999, who concludes that civil servants act as monopolistic price discriminators based on quantitative research (Svensson, 1999). However, not all economists do consider that corruption is harmful for prosperity and productivity. On the contrary, it may turn out to have a positive impact on economic growth. For example, bribery may offer the possibility to bypass the hurdles of pointless regulations. According to these economists, in a situation where policies are inadequate, offences such as bribery can be beneficial and contribute to economic development. Corruption helps economic actors to deal with incompetent governments that do not care for the welfare of companies and consumers. The analysis carried out by Andrzej Cieślik and Łukasz Goczek focussed on perceived corruption in CORRUPTION, BRIBERY, PAID PROTECTIONISM, NEPOTISM, BLACKMAIL, MISAPPROPRIATION OFPUBLIC FUNDS, ACCEPTING A BRIBE, OFFERING A BRIBE, PARTICIPATION IN BRIBERY, INCITEMENT OF A BRIBE Fig. 3. Classification of types of corruption. Source: Chrustowski T, Prawne, kryminologiczne i kryminalistyczne aspekty łapówkarstwa [Legal, criminological and forensic aspects of bribery], Legal Publishing House, Warsaw 1985. CEEJ • 7(54) • 2020 • pp. 186-204 • ISSN 2543-6821 • DOI: 10.2478/ceej-2020-0015 192 companies based in post-communist countries. The authors showed that corruption is more pronounced in companies with domestic capital and those producing for the domestic market. Another facilitating factor is the time spent privately with government officials, as well as the number of state audits in the firm. The analysis covered the period from 1999 to 2010 (Cieślik and Goczek, 2015). The above-mentioned studies employed econometric models for the analysis of survey data. Some researchers choose the zero-one variable as the dependent variable, which is the answer to the question whether the respondent has encountered bribery. The selected econometric models are either logit or probit. The next step is to analyse the financial data using the least squares method (LSM). 3.2 Macroeconomic research on corruption Corruption has negative consequences for the economic development of a country. This hypothesis was supported by Łukasz Goczek and Andrzej Cieślik, who analysed the relationship between corruption and economic growth in transition countries. They used Mauro’s theoretical model of economic growth and the econometric model. Additionally, panel data from 29 countries in 1993–-2013 were used. As a result of the analysis, the authors concluded that there is a relationship between corruption and economic growth. The increase in bribery results in the inhibition of economic development (Cieślik and Goczek, 2016). On a macroeconomic scale, corruption also affects the level of investment in a country. A decline in investment leads to a lower level of GDP per capita, which means that the economy starts to grow more slowly or, in the worst case, stops growing. According to Te Velde, in more corrupt countries, investors refuse to enter the market. If we treat bribes as another tax, entrepreneurs do not want to pay additional fees as their goal is to maximise profit. Moreover, the important fact is that the corruption contract is extremely uncertain. Companies may pay bribes but may receive nothing in return. This is a risk that investors would prefer to minimise (Te Velde and Morrissey, 2001). Eric Ambukita conducted an analysis of the impact of corruption on the investments in a country. He concluded that higher levels of corruption discourage foreign investors. The study is related to African countries, which are the most corrupt countries in the world (Ambukita 2012). In a 1997 Tanzi and Davoodi came to different conclusions. The economists observed that corruption causes lower economic growth but contributes to higher public investment. This could also be due to slower economic growth, which also has an impact on public investment. Least squares estimation with the use of an econometric model was used for that analysis (Tanzi and Davoodi, 1998). 4 Research methods The logit model has been used in this study. The explained binary variable identifies companies based on the answer to the question of whether corruption is an obstacle to doing business. The probability that a bribery problem will arise in a company is represented by one of two values: 0 – it is not an obstacle 1 – it is an obstacle The probability depends on many companyspecific characteristics and other explanatory variables. In the following analysis, the explained variable takes one of the two values 0 or 1, i.e. it is not measured continuously. Further, the use of simple regression would be incorrect in this case. Because of this reason, we decided to use regression for the binary dependent variable, the logit model. The dependent variable is discrete and takes only two values. Moreover, the variable takes only the value 0 or 1 in our study to indicate the absence or presence of obstacles. Thus, the model can be expressed as the following formula: ( ) ( ) i Pr y 1 F ,Xi= = θ (1) In other words, the probability of choosing a specific alternative is a function of the observed characteristics of Xi and the set of q parameters. In the beginning, it was assumed that Function 1 is linear, which produced a linear probability model with the following properties: ( ) ( ) i0 i0 Pr y 1 X Pr y 0 1 X = = β ==−β (2) CEEJ • 7(54) • 2020 • pp. 186-204 • ISSN 2543-6821 • DOI: 10.2478/ceej-2020-0015 193 The vector of b parameters reflects the influence of each variable on the likelihood of one of the two possible situations. The advantage is that it can be estimated using the LSM method. yi X0 i= β+ε (3) Unfortunately, it also has some drawbacks. As equation (3) does not guarantee that the fitted value of the X’b model, it is interpreted as the probability of the occurrence of the situation taking the value of 1, lies between zero and one. Second, because the response variable takes only two values 0 and 1, the random term has a binomial distribution and is equal to e = −X’b or e = 1 − X’b. Therefore, it can be shown that the conditional variance is: var (e|X) = X’b (1−X’b). Since X’b may have different values, it cannot be sure that the variance in the linear probability model will be non-negative. Moreover, the random term is heteroscedastic. Considering these undesirable features of the linear model, two other probability functions are used in econometric practice. The adoption of the normal distribution leads us to the probit model: ( 1) ( ) X i Pr y t dt X ′ −∞ ′ = = () =Φ ∫ β φβ (4) When expressed as the logistic distribution with the logit model: ( 1) ( ) 1 X iX e Pr y X e ′ ′′ = = = Λ + β β β (5) Since both functions are probability functions, their values are limited by 0 and 1. Both distributions are symmetrical about zero. What makes them different is the variance. For the normal distribution, it is 1, for logistic - p2/3. Both distributions are very similar to each other, the difference being that the logistic distribution has thicker tails. The probability function for a single observation is described by the equation: () 1 (|) 1() 0 ii ii ii x dla y Pr y x x dla y Λ=   =−Λ =   β β where: () 1 i i x ix e xe Λ= + β β β and the expected value is: (|)1 ( )0(1 ( )) ( ii i i i Ey x x x x= ⋅Λ + ⋅ −Λ =Λ ) β ββ (Mycielski 2010) 5 Statistical data The study used data on how companies perceive corruption. The source of information is the questionnaire Business Environment and Enterprise Performance Survey (BEEPS). The survey was done in 2016 and covered 16,566 companies from 32 countries. The data provides information on whether corruption is an obstacle in running a business and knowledge about the characteristics of the firm. The authenticity of the collected data is high as the data collection has been authorised and approved among others by the World Bank. However, it should be taken into account that when answering the question of whether corruption is an obstacle, the respondents may use understatement when they have any fears or concerns. Figure 4 shows the percentage of companies’ perception of corruption in the context of doing business. It can be observed that the vast majority of respondents take the position that the problem of bribery is an obstacle in and for a business. Furthermore, the answers were grouped into two categories, i.e. not an obstacle and an obstacle. The list of variables (explained and explanatory) with a description is shown in Table 1. Further, zero-one variables were also used in the model. The first group is related to the sector in which the company operates. The sectors are as follows: Food products, Textiles, Clothing, Chemicals, Plastics and rubber, Mineral products, Metals and raw materials, Metal products, Machines and devices, Electronics, Construction, Wholesale, Retail, Hotels and Restaurants, Transport, IT and other services. The next group is related to the country where the company is based. The list of the countries comprises Bulgaria, Albania, Croatia, Belarus, Georgia, Tajikistan, Turkey, Ukraine, Uzbekistan, Russia, Romania, Kazakhstan, Bosnia and Herzegovina, Azerbaijan, Macedonia, Armenia, Kyrgyzstan, Estonia, Czech Republic, Italy, Latvia, Lithuania, Slovakia, Slovenia, Serbia, Cyprus, Greece, Moldova, Mongolia, Montenegro, Poland and Kosovo. The next group describes the size of the town where the company is located: small location (town CEEJ • 7(54) • 2020 • pp. 186-204 • ISSN 2543-6821 • DOI: 10.2478/ceej-2020-0015 200 Appendix Tab. A1. Comparison of the estimation results of the logit and probit models Variable logit probit Years on the market 0.00876502 0.00530975 Employment −0.00008314 −0.00004948 Sta_capital −1.133988 −0.62454681 For_capital −0.197049 −0.1253618 State_prod 0.02541215 0.01607862 Time_officials 0.36951254 0.22890388 Audits 0.00907696 0.00543676 Efficiency 0.02150128 0.01288578 Thefts 3.4365243 1.8378906 Other 0.03748113 0.02313259 Food products −0.14439709 −0.0845678 Textiles 0.3521586 0.21495443 Clothing 0.11952786 0.07476893 Chemicals 0.52001581 0.31532122 Plastics and rubber −0.03725552 −0.01415368 Mineral products 0.32495071 0.19949212 Metals and raw materials −0.0577727 −0.02944644 Metal products 0.02482283 0.01903884 Machines and devices 0.13714882 0.08354676 Electronics 0.2596035 0.15581276 Construction 0.3544781 0.22319262 Wholesale 0.20622816 0.13128126 Retail −0.0664786 −0.03453299 Hotels and restaurants −0.1737417 −0.09251344 IT 0.33941652 0.2147104 small_loc −0.12165995 −0.07618389 medium_loc −0.33236935 −0.2025015 large_loc −0.33919617 −0.20681956 small_company −0.0014848 0.00030351 medium_company −0.02023952 −0.01255519 medium_company −0.02023952 −0.01255519 Bulgaria –1.1978914 −0.73944942 Albania −1.1592144 −0.71822055 Croatia −1.1413132 −0.70407362 Belarus −2.0420131 −1.2431616 Variable logit probit Georgia −2.6648174 −1.5931303 Tajikistan −1.4865773 −0.9180442 Turkey −1.2972634 −0.80528348 Ukraine −0.40252307 −0.24803484 Uzbekistan −3.5135959 −2.0308543 Russia −0.88202438 −0.54765707 Romania −0.13433538 −0.08237083 Kazakhstan −1.2451245 −0.76820525 Bosnia_and_Herzegovina_Ha −1.0461743 −0.6461674 Azerbaijan −3.0499892 −1.792391 Macedonia −1.8585547 −1.1448237 Armenia −1.8512206 −1.1417558 Kyrgyzstan 0.35124133 0.21087963 Estonia −3.0235201 −1.8171122 Czech Republic −1.1688309 −0.7194871 Italy −2.2520163 −1.3720082 Latvia −1.9601472 −1.2002027 Lithuania −1.7678234 −1.0833992 Slovakia −0.99406504 −0.6160503 Slovenia −2.1529345 −1.304369 Serbia −1.5845897 −0.97913306 Cyprus −1.6468074 −1.0145575 Greece 0.46229253 0.26145001 Moldova −1.0586318 −0.65745899 Mongolia −1.665327 −1.025903 Montenegro −2.827147 −1.6969201 Poland −1.5308906 −0.939425 _cons 0.54153984 0.33342044 Source: BEEPS database. The table reveals that the coefficients of the logit model are higher than those of the probit model. This is the rule for probabilities outside the ‘tails’ of the distribution. Unfortunately, when choosing a logit or probit model, there is no clear-cut test that would answer the question which model should be used. When making decisions, analysts are usually guided by the comfort of working with the given model. In the study, we decided to use the logit model. CEEJ • 7(54) • 2020 • pp. 186-204 • ISSN 2543-6821 • DOI: 10.2478/ceej-2020-0015 201 Tab. A2. Model estimation for the industry effect Corruption Coeff. zP>|z| (Std. Err.) Other 0.0385 0.270 0.785 (0.1409) Food products −0.1382 −0.960 0.337 (0.1439) Textiles 0.3672* 1.920 0.055 (0.1912) Clothing 0.1311 0.770 0.441 (0.1701) Chemicals 0.5306*** 2.700 0.007 (0.1964) Plastics and rubber −0.0306 −0.150 0.882 (0.2059) Mineral products 0.3376** 2.070 0.039 (0.1634) Metals and raw materials −0.0580 −0.180 0.860 (0.3297) Metal products 0.0288 0.170 0.863 (0.1669) Machines and devices 0.1393 0.940 0.348 (0.1485) Electronics 0.2530 1.070 0.284 (0.2361) Construction 0.3605*** 2.630 0.009 (0.1373) Wholesale 0.1995 1.580 0.114 (0.1263) Retail −0.0695 −0.580 0.563 (0.1201) Hotels and restaurants −0.1644 −1.010 0.311 (0.1623) IT 0.3336 1.450 0.147 (0.2302) The symbol * means the significance of the variable at the level of 0.1, ** significance of the variable at the level of 0.05 and *** significance of the variable at the level of 0.01. Source: BEEPS database. Tab. A3. Model estimation for the country effect Corruption Coeff. zP>|z| (Std. Err.) Bulgaria −1.2071*** −5.160 0 (0.2338) Albania −1.1592*** −5.520 0 (0.2098) Croatia −1.1568*** −4.760 0 (0.2430) Belarus −2.0415*** −7.030 0 (0.2905) Georgia −2.6626*** −7.030 0 (0.3789) Tajikistan −1.4810*** −6.660 0 (0.2225) Turkey −1.2983*** −6.630 0 (0.1958) Ukraine −0.3955** −2.060 0.04 (0.1922) Uzbekistan −3.5079*** −8.670 0 (0.4045) Russia −0.8858*** −4.940 0 (0.1792) Romania −0.1506 −0.730 0.464 (0.2055) Kazakhstan −1.2396*** −5.000 0 (0.2477) Bosnia and Herzegovina −1.0536*** −4.920 0 (0.2142) Azerbaijan −3.0388*** −10.450 0 (0.2907) Macedonia −1.8757*** −8.560 0 (0.2192) Armenia −1.8567*** −8.300 0 (0.2236) CEEJ • 7(54) • 2020 • pp. 186-204 • ISSN 2543-6821 • DOI: 10.2478/ceej-2020-0015 202 Corruption Coeff. zP>|z| Kyrgyzstan 0.3629 1.580 0.114 (0.2298) Estonia −3.0317*** −6.020 0 (0.5035) Czech Republic −1.1714*** −4.790 0 (0.2448) Italy −2.2347*** −8.460 0 (0.2640) Latvia −1.9786*** −6.820 0 (0.2901) Lithuania −1.7852*** −6.230 0 (0.2866) Slovakia −1.0013*** −3.820 0 (0.2618) Slovenia −2.1769*** −5.180 0 (0.4199) Serbia −1.5975*** −6.970 0 (0.2292) Cyprus −1.6615*** −5.090 0 (0.3262) Greece 0.4579 1.440 0.149 (0.3170) Moldova −1.0548*** −4.940 0 (0.2134) Mongolia −1.6670*** −7.510 0 (0.2219) Montenegro −2.8361*** −8.060 0 (0.3518) Poland −1.5388*** −6.210 0 (0.2477) The symbol * means the significance of the variable at the level of 0.1, ** significance of the variable at the level of 0.05 and *** significance of the variable at the level of 0.01. Source: BEEPS database. Tab. A4. Descriptive statistics for the country variables Variable Number of observations Std. Dev. Min. Max. Bulgaria 16.566 0.1318 0 1 Albania 16.566 0.1458 0 1 Croatia 16.566 0.1458 0 1 Belarus 16.566 0.1458 0 1 Georgia 16.566 0.1458 0 1 Tajikistan 16.566 0.1456 0 1 Turkey 16.566 0.2730 0 1 Ukraine 16.566 0.2384 0 1 Uzbekistan 16.566 0.1516 0 1 Russia 16.566 0.4357 0 1 Romania 16.566 0.1776 0 1 Kazakhstan 16.566 0.1868 0 1 Bosnia and Herzegovina 16.566 0.1458 0 1 Azerbaijan 16.566 0.1516 0 1 Macedonia 16.566 0.1458 0 1 Armenia 16.566 0.1458 0 1 Kyrgyzstan 16.566 0.1266 0 1 Estonia 16.566 0.1273 0 1 Czech Republic 16.566 0.1229 0 1 Italy 16.566 0.1355 0 1 Latvia 16.566 0.1410 0 1 Lithuania 16.566 0.1266 0 1 Slovakia 16.566 0.1262 0 1 Slovenia 16.566 0.1266 0 1 Serbia 16.566 0.1458 0 1 Cyprus 16.566 0.1458 0 1 Greece 16.566 0.1383 0 1 Moldova 16.566 0.1458 0 1 Mongolia 16.566 0.1458 0 1 Montenegro 16.566 0.0947 0 1 Poland 16.566 0.1779 0 1 Kosovo 16.566 0.1098 0 1 Source: BEEPS database. ContinuedTab. A3. Model estimation for the country effect CEEJ • 7(54) • 2020 • pp. 186-204 • ISSN 2543-6821 • DOI: 10.2478/ceej-2020-0015 203 Tab. A5. Descriptive statistics for the Business sector variables Variable Number of observations Std. Dev. Min. Max. Other 16.566 0.2715 0 1 Food products 16.566 0.2587 0 1 Textiles 16.566 0.1567 0 1 Clothing 16.566 0.1876 0 1 Chemicals 16.566 0.1516 0 1 Plastics and rubber 16.566 0.1428 0 1 Mineral products 16.566 0.1966 0 1 Metals and raw materials 16.566 0.0739 0 1 Metal products 16.566 0.2001 0 1 Machines and devices 16.566 0.2360 0 1 Electronics 16.566 0.1165 0 1 Construction 16.566 0.2810 0 1 Wholesale 16.566 0.3493 0 1 Retail 16.566 0.4215 0 1 Hotels and restaurants 16.566 0.2001 0 1 IT 16.566 0.1351 0 1 Source: BEEPS database. Fig. A1. ROC curves. Source: BEEPS database. The accuracy of data classification depends on the degree by which the constructed model differentiates between successes and failures. The measure of accuracy is the area under the ROC curve. A field equal to 1 means that the model perfectly discriminates between success and failure and a value equal to 0.5 means no possibility of discrimination. Tab. A6. Test for the correctness of the functional form Source: BEEPS database. 0.00 0.25 0.50 0.75 1.00 Sensitivity 0.00 0.25 0.50 0.75 1.00 1 - Specificity Area under ROC curve = 0.5674 Fig. A2. ROC curve (only significant variables). Source: BEEPS database. Tab. A7. Test for the correctness of the functional form (only significant variables) Source: BEEPS database. CEEJ • 7(54) • 2020 • pp. 186-204 • ISSN 2543-6821 • DOI: 10.2478/ceej-2020-0015 204 Fig. A3. ROC curve – only macroeconomic variables. Source: Bąk 2020. Tab. A8. Test for the correctness of the functional form – only macroeconomic variables Source: Bąk 2020.