363 Organizacija, V olume 57 Issue 4, November 2024 Research Papers 1 Received: 20th December 2023; Accepted: 22nd August 2024 Government Effectiveness in the Petroleum Sector: Two-step Analysis Combining Linear Regression and Artificial Neural Networks Barbara DORIĆ 1 , Dinko PRIMORAC 2 , Mirjana PEJIĆ BACH 3 1 University of Ljubljana, Faculty of Economics and Business, Ljubljana, Slovenia, barbaradoric@yahoo.com 2 University North, Croatia, Koprivnica, Croatia, dprimorac@unin.hr 3 University of Zagreb, Faculty of Economics & Business, Zagreb, Croatia, mpejic@net.efzg.hr Background and Purpose: To encourage petroleum industry development, a country needs to set up a regulatory framework that standardizes investment conditions. The objective of the research was to investigate the determi- nants of government effectiveness in the petroleum sector. Design/Methodology/Approach: Multiple regression analysis was conducted to investigate if government effec - tiveness in the petroleum sector is influenced by the country’s political stability, regulatory quality, the intensity of petroleum exploration and production activities, government take, and type of contract used. Artificial neural network analysis was additionally conducted to identify the importance of independent variables. Results: Political stability , regulatory quality , government take attractiveness, and the intensity of petroleum ac - tivities positively influence government effectiveness. A more attractive government take enhances effectiveness, while the type of contract for awarding petroleum rights did not significantly impact effectiveness. Artificial neural network analysis revealed that the most important variables were regulatory quality and political stability. Conclusion: The research concluded that political stability, regulatory quality, and the intensity of petroleum activ - ities are key factors in enhancing government effectiveness in the petroleum sector. These findings have practical implications, as they emphasize the importance of stable and well-regulated environments for achieving higher government effectiveness in the petroleum industry. This equips policymakers and industry professionals with ac- tionable insights for improving the sector’s performance. Keywords: Energy policy, Government effectiveness, Petroleum sector performance, Petroleum resources manage- ment, Industry development DOI: 10.2478/orga-2024-0026 1 Introduction As petroleum production expanded in the United States, disputes arose regarding land ownership and shares in profit (Hammerson, 2011). American state courts estab- lished a legal practice regarding the rights of oil leases and the management of revenue from its production. The 1889 Pennsylvania Supreme Court decision equated the produc- tion of oil and gas to that of other minerals, concluding that land ownership does not necessarily entail ownership 364 Organizacija, V olume 57 Issue 4, November 2024 Research Papers of minerals (Hammerson, 2011). Texas applied the offset rule for neighboring wells and the concept of ownership in place, which defined the ownership principles in petroleum production as either freehold ownership of the land, which included the right to minerals, or partial ownership, which did not include the right to minerals (Thurman, 2022). Capital investments in petroleum exploration and produc- tion and the return on investment in this activity could be compared to the riskiest investments in speculative trends on the capital market (Simkins & Simkins, 2013). Despite this, the possibility of exceptionally high profits in the case of a positive petroleum discovery motivated oil companies to take such risks. The United States legal system is based on the prin- ciples of Anglo-Saxon law and precedents, and court decisions have also established the legal practice for re- lationships among participants in petroleum exploration and production activities. The starting point is the freehold ownership category, which, along with the land ownership, entails the right to minerals, i.e., oil and gas. In contrast, European countries implemented different forms of feudal and royal limitations regarding mining rights. The owner- ship of petroleum in European countries is considered a public good and is regulated by provisions governing state ownership (Thurman, 2022). The principal dissimilarity between the petroleum exploration and production business in the United States and the rest of the world stems from the definition of min- eral ownership (Seba, 2008). In countries applying An- glo-Saxon law, oil leases are based on freehold ownership, which includes the right to minerals, whereby the lease includes compensation for the land and part of the value of the produced oil and gas. In most countries worldwide, where state ownership of minerals prevails, oil compa- nies acquire the right to minerals from the government. At the same time, the lease for the use of the land is agreed upon with landowners based on local laws and regulations (Simkins & Simkins, 2013). The relationships between the company acquiring the rights to minerals and the previous, i.e., original, owner (freehold owner in the Anglo-Saxon law or the state in continental law) are governed by a con- tract defining the terms and compensation for rights to pe- troleum. This is a specific compensation, income, or yield obtained by the landowner (state) and represents a cost for the petroleum lessee, different than all other taxes or ex- penses. This yield is known as a royalty, i.e., the fee for recovered quantities of petroleum. In the United States, it traditionally amounts to 1/8 (12.5%) of the market value of the produced petroleum (Johnston, 1994). In the 20th century, contractual relationships in pe- troleum exploration and production developed due to the rise in petroleum production and exploration and rising oil prices. Oil and gas became essential primary sources of energy, accounting for over two-thirds of primary energy consumption. The expansion of transport led to oil becom- ing one of the most important primary resources, and the use of energy became important in contemporary industri- al infrastructure. This increased value influenced the codi- fication and regulation of relationships among participants in petroleum exploration and production. In countries with a free market economy, petroleum companies conducting petroleum exploration and produc- tion activities were state-owned, forming part of a planned and targeted economy. Following the disintegration of a non-market and planned socio-economic system, free cap- ital ownership has become a global universal principle of relationships. Ownership and contractual relationships in the area of exploration and production of valuable natural resources with high capital intensity and value, such as pe- troleum, have become a matter of special attention for all government instances. Laws and legal regulations regarding petroleum pro- duction were once part of mining legislation. However, since petroleum is present in the Earth’s crust in varying physical and geological forms, exploration is performed using a range of technical means, and the production tech- nology differs from that in the production of solid mineral raw materials. Petroleum legislation sets out conditions for investments in petroleum exploration and production, le- gal prerequisites for development, and competitive terms regarding petroleum exploration and production (John- ston, 1994). It places significant emphasis on optimization during mineral raw material management processes while primarily protecting national interests and providing pe- troleum companies (investors) with security and stability as they carry out their investments and business activities (Green & Smith, 2023; Thurman, 2022). The regulatory framework in every country is based on the nation’s constitution, which grants taxing and legisla- tive authority that governs petroleum legislation and out- lines authority boundary conditions for relationships with foreign companies. The function of government is to pro- vide an adequate regulatory infrastructure for companies to work economically productive units and ensure they do not swindle the public, exploit workers, pollute their surroundings, prosecute unethically, or engage in morally or socially reprehensible practices (Parra, 2004). Hence, establishing a regulatory framework that standardizes in- vestment conditions is an essential step in promoting the growth of the petroleum sector. The study aims to examine the factors that influence government efficiency in the petroleum industry. This study employs multiple regression analysis (MLA) to ex- amine the potential impact of political stability, regulatory quality, the intensity of petroleum exploration and produc- tion operations, government take, and type of contract on government performance in the petroleum industry. Fur- thermore, a study of artificial neural networks (ANN) was performed to determine the significance of independent factors. 365 Organizacija, V olume 57 Issue 4, November 2024 Research Papers 2 Theoretical background Petroleum legislation sets out conditions for invest- ments, legal prerequisites for development, and com- petitive terms regarding petroleum exploration and pro- duction. It emphasizes optimization during mineral raw material management processes and provides security and stability for petroleum companies as they carry out their investments and business activities. The petroleum regulatory framework in a country is based on its constitution, which grants taxing and legis- lative authority for petroleum legislation and outlines au- thority boundary conditions for relationships with foreign companies. The constitution also includes specific petro- leum legislation that authorizes the national oil compa- ny or responsible ministry to negotiate certain aspects of agreements between the state and foreign companies. Tax liabilities are usually included in the agreement signed be- tween the parties (government and petroleum company) and regulated by separate laws. Governments provide an adequate regulatory infra- structure for companies to work economically productive units and ensure they do not swindle the public, exploit workers, pollute the environment, prosecute unethically, or defraud shareholders. Changes in the economic environ- ment and the increasing interest of foreign investors have indicated the need to regulate petroleum exploration and production in a manner defined and accepted within global practice. Petroleum legislation needs to create conditions for large investments, determine the legal prerequisites for energy development, and protect national interests in the petroleum sector. Petroleum lease contracts are more similar to financial contracts than typical land concessions or mining conces- sions due to the uncertainty of petroleum prices and the in- creased strategic role of petroleum. Both parties have indi- vidual interests, with the oil company minimizing risk and the government increasing its share in profit distribution. This results in direct increases in fiscal revenue through royalties, taxes, and indirect contributions. The fiscal regime, or petroleum taxation model, is a financial structure that oil companies must pay to countries for petroleum exploration and production activities. It is often represented as a government take versus an oil com- pany take, with the government taking the percentage of profit that goes directly to the state budget and the oil com- pany taking the percentage remaining with the company (Johnston, 2003). There are 145 countries worldwide with specific fiscal and contractual terms for engaging with oil companies for petroleum exploration and production oper- ations (IHS Energy, 2016). These regimes can be divided into two main categories: the concessionary system (based on royalty and tax payments) and the production-sharing system (based on petroleum production sharing) (Green & Smith, 2023). The fiscal regime, if balanced and regulated proper- ly, can attract significant investments in exploration and production activities and create wealth for the nation. The higher the government take, the greater the probability of creating wealth for the nation. To determine the attractive- ness of the fiscal regime, the government take is combined with other measures of profitability, including fiscal sys- tem flexibility, revenue risk, and fiscal stability (Johnston, 2003). The most common fiscal regime terms used world- wide are bonuses, fees, state participation, royalty, produc- tion sharing, cost recovery, and taxes (Simkins & Simkins, 2013). The terms of the fiscal regime differ among coun- tries, and not all are included within one particular regime. Production sharing is a fiscal regime that allows reve- nue from petroleum production to be shared between the domicile country and the oil company, allowing the com- pany to recover costs and make a return on investment (Johnston, 2003). The three main elements of production sharing are cost recovery, excess cost recovery, and profit share. Taxes are common to both fiscal regime systems, including corporate income tax, additional profit taxes de- fined only for petroleum operation companies operating in the domicile country, and dividend withholding taxes. The ideal fiscal regime should ensure a stable business environment, minimize sovereign risk, discourage undue speculation, provide the potential for a fair return, balance risk and reward, avoid complexity, limit administrative burden, allow flexibility, and promote healthy competition and market efficiency (Johnston, 2003). The most com- mon petroleum industry-recognized fiscal regimes fall broadly into two categories: the concession system, which includes special fees and taxes payable in money to the country where it is operating, and the contractual (produc- tion sharing) system, which includes production sharing arrangements where petroleum is usually shared in kind between oil company and domicile country (Simkins & Simkins, 2013). In the concessionary system, oil companies have the right to perform petroleum exploration and production at their own cost, assuming the overall risk of discovery and production risks. The royalty goes directly to the country as one part of the country’s petroleum profit, and all tax- es are payable on profit before income tax (Seba, 2008). Figure 1 shows the typical revenue distribution under the concession system and illustrates the hierarchy of royal- ties, deductions, and examples of possible taxation layers. Of the total revenues collected from petroleum produced, the royalty (percentage of total petroleum value) goes di- rectly to the country as one part of the country’s petroleum profit (state budget revenue). Before-tax calculations, roy- alties are deducted together with all capital and operating expenditures (CAPEX and OPEX) from total revenues to give the oil company profit before income tax. All taxes (income tax, petroleum special tax, and any other taxes) 366 Organizacija, V olume 57 Issue 4, November 2024 Research Papers are payable on profit before income tax. Income tax goes to the country as the second part of the country’s petroleum profit (state budget revenue). The remainder after taxes is the oil company’s petroleum profit. Production sharing contracts (PSC) are a newcomer to the petroleum industry, starting in 1966 in Indonesia (Markus, 2014). These contracts involve a contractual relationship between the state and the oil company, with the state owning petroleum rights and the oil company en- suring the execution, technical, and financial realization of petroleum exploration, development, and production (Seba, 2008). The aim is to maximize income and initi- ate economic activities connected to petroleum explora- tion and production. Production is split between the host government and the contractor, with the government maintaining ownership of the produced petroleum. Stabi- lization clauses are essential to ensure the preservation of the tax system and fiscal proportions throughout the con- tract. Figure 2 shows typical revenue distribution under the production-sharing system and illustrates the calculation of revenues and costs that would be experienced in a full cycle. From total revenues, collected from the petroleum produced, cost recovery is deducted first. Cost recovery includes all capital and operating expenditures (CAPEX and OPEX) borne by the oil company in producing the petroleum, and this is the oil company’s revenue. What re- mains after cost recovery is the profit share. Profit share is then split between the country and oil company based on the contracted percentages. Profit share represents one part of the country’s petroleum profit (state budget revenue), while for the oil company, it is one part of the oil com- pany revenue. From oil company revenue, comprised of cost recovery and the company’s part of the profit share, all capital and operating expenditures (CAPEX and OPEX) are deducted. What remains after deductions is subject to taxation. Income tax is paid to the country, representing the second part of the country’s petroleum profit (state budget revenue), while the remainder after taxes is the oil company’s petroleum profit. Fiscal regimes are often categorized as hybrids, com- bining elements of both classifications, such as royalties and taxes. These hybrid systems aim to ensure petroleum profit from the start of production, with most countries using the concessionary system (royalty/tax contract) and production sharing system (production sharing contract). As shown in Table 1, most countries worldwide use these two systems: concessionary system (royalty/tax contract) and production sharing system (production sharing con- tract). Figure 1: Concession system - Typical cash flow diagram Source: Doric B. (2017) 367 Organizacija, V olume 57 Issue 4, November 2024 Research Papers Figure 2: Production sharing system - Typical cash flow diagram Source: Doric B. (2017) Type of contract Number of countries Royalty/Tax 78 Production Sharing Agreement 52 Mixed / Various 15 Table 1: Types of contracts (agreements) used worldwide Source: IHS Energy (2016a) Contract terms for oil exploration and production contracts include fiscal terms, work commitments, insur- ance, and local content. Work commitment is crucial for the host country to ensure the oil company commits to as much work as possible, enabling quicker development of potential new production and revenue generation. Insur- ance is essential to ensure high-quality work commitments and cover potential losses. Contracts often require the oil company to buy goods and services locally to boost local industry development and employ local labor. The recog- nition of exploration and production costs is sensitive, as the investment of funds and risk lies with the oil company. The country must establish a fiscal regime that maximizes revenues and provides investors with incentives to explore and develop petroleum efficiently. The license round pro- cedure for petroleum exploration and production is indus- try-standardized and consists of several steps. Concessionary and production-sharing systems have advantages and disadvantages, but the choice of system may not be as critical from an economic, accounting, and financial perspective (Johnston, 2003). The fundamental difference between the two systems lies in the ownership of the petroleum produced. To encourage petroleum industry development, a country needs to set up a regulatory framework that will standardize the conditions for investments in petroleum exploration and production and lay down the legislative prerequisites for the development and competitive condi- tions in this activity. The function of government is crucial in organizing and managing the petroleum sector since an adequate regulatory framework will ensure that explora- tion and production activities are conducted in a way that 368 Organizacija, V olume 57 Issue 4, November 2024 Research Papers will create wealth for the nation, protect the environment, and enable companies to work in a stable and competi- tive environment (Parra, 2004). Many authors (Falola & Genova, 2005; Shaffer, 2011; Thurber et al., 2011; Holden, 2013; Kemal, 2016) have argued that political stability is an important prerequisite for petroleum industry develop- ment and enables economic growth and generates wealth for the nation. Most of the theories to date have been based on historical information and statistical observations on the development of the petroleum industry. Thurber et al. (2011) went a step further in their empirical research, indicating that some countries implementing the Norwe- gian Model failed due to a lack of institutional quality and political stability, which influenced government effective- ness in the petroleum sector. This argument was further elaborated by Kemal (2016), who stated that the economic impact due to changes in petroleum governance might de- pend on political conditions. The literature has indicated that political stability is an essential factor that directly influences government effectiveness and thus shows the ability of the government to create a stable environment when it comes to investments in the petroleum sector. The theory discussed demonstrated that petroleum leg- islation is the main factor in regulating the complex rela- tionship between governments and oil companies in petro- leum exploration and production activities. It lays down the conditions for investments, the legal prerequisites for development, and competitive terms and conditions. As political stability directly dictates the government’s ability to implement adequate policies and regulations for indus- try development, regulatory quality is another important factor that can also impact government effectiveness. Exploration activities are crucial for discoveries and petroleum production. The more an oil company invests in exploration activities, the greater the probability of new petroleum discoveries and developing new petroleum pro- duction. Petroleum exploration intensity depends on geo- logical probability, which also depends on the investor’s (oil company) investment in exploration and production activities and the owner’s (government) ability to attract investments and enable exploration and production activi- ties. Therefore, the intensity of petroleum exploration and production activities is another factor that may influence government effectiveness. Moreover, suppose the relationship between the prom- ising petroleum potential and the requirements set in the fiscal regime is unfavorable to oil companies in advance. In that case, they will not proceed with the business and investments. Thus, governments must design an optimal fiscal regime to ensure a favorable balance of mutual rela- tions. The government take (share of the petroleum profit) is used as a measure to compare the fiscal regimes of dif- ferent countries in terms of petroleum profit going directly into the state budget and the fiscal regime attractiveness for petroleum sector investments. Accordingly, the theory indicates that government take attractiveness should also influence government effectiveness. The theory outlined in this chapter suggests that nei- ther the type of fiscal regime system nor the correspond- ing type of contract is better or worse. From the economic perspective, the same objectives can be achieved through both concessionary and production-sharing contracts. Therefore, it could be concluded that the type of contract has no influence on government effectiveness in the petro- leum sector. 3 Hypothesis development According to the theoretical findings outlined in the previous chapter, the objective of this empirical study is to examine the influence of political stability, regulatory quality, the intensity of petroleum exploration and produc- tion activities, government take (fiscal regime) attractive- ness, and the type of contract on government effectiveness. Based on the theory discussed, the following hypotheses were developed. The government effectiveness variable represents the quality of public service, the quality of civil service and its degree of independence from political pressures, the quality of policy formulation and implementation, and the credibility of the government’s commitment to such policies (World Bank, 2016). Many authors focused on government effectiveness when analyzing the petroleum governance model among various oil-producing countries together with oil-sector performance. Brunnschweiler and Bulte (2008) showed that there is a significant difference in government effectiveness in various oil-producing countries. Their empirical research suggested that better government effectiveness led to less resource dependence and higher direct investments, which in turn positively af- fects GDP. Like many other authors, for the government effectiveness variable, they used World Bank data, argu- ing that these data have the advantage of extensive cov- erage and objectiveness due to a large survey base, mak- ing them particularly attractive for econometric analysis (Brunnschweiler & Bulte, 2008). Kaufman et al. (2004) argued that the key advantage of the World Bank World- wide Governance Indicators (WGI) is that despite the mar- gins of error, these indicators are sufficiently informative that many cross-country comparisons result in statistically significant differences in estimated governance. The WGI measures six dimensions of governance, which are gov- ernment effectiveness, political stability, regulatory quali- ty, rule of law, control of corruption, and voice of account- ability. Government effectiveness data published by the World Bank has also been used by other authors (Heller & Marcel, 2012; Thurber et al., 2011), aiming to compare government effectiveness in the petroleum sector among various oil-producing countries. 369 Organizacija, V olume 57 Issue 4, November 2024 Research Papers The political stability variable represents the level of political stability and perceptions of the likelihood of po- litical instability and politically motivated violence (World Bank, 2016). Many authors (Falola & Genova, 2005; Shaffer, 2011; Thurber et al., 2011; Holden, 2013; Kemal, 2016) used political stability in their empirical research and showed that the level of political stability would pos- itively influence petroleum sector governance and conse- quently government effectiveness in the petroleum sector. The data used in the model was pulled from WGI data pub- lished by the World Bank in 2016. Based on this, the first hypothesis was developed as follows: • H1: Political stability positively influences gov- ernment effectiveness in the petroleum sector. The regulatory quality variable represents the per- ception of the ability of the government to formulate and implement comprehensive policies and regulations that permit and promote private sector development (World Bank, 2016). As elaborated in the previous chapter, pe- troleum legislation is very specific, complex, and indus- try standardized. Moreover, the ability of the country to implement industry-standardized petroleum legislation is very important in terms of the development of the petrole- um sector. Thus, regulatory quality is another variable that, in addition to political stability, should strongly influence government effectiveness. Some of the previous research related to the petroleum governance model and respective- ly government effectiveness in the petroleum sector used the regulatory quality indicator (Heller & Marcel, 2012; Thurber et al., 2011). The second hypothesis is developed as follows: • H2: Regulatory quality positively influences gov- ernment effectiveness in the petroleum sector. The intensity of petroleum exploration and production activities variable represents the ranking of the selected sample countries in terms of the intensity of petroleum ex- ploration and production activities in the five years from 2010 to 2016. The ability of the country to attract invest- ments in petroleum exploration and production activities shows the country’s (government) effectiveness in maxi- mizing its potential and revenues from petroleum (John- ston, 1994). This ability can be measured in the intensity of petroleum exploration and production activities. To rank oil-sector performance and measure effectiveness, Thurber et al. (2011) evaluated the ability of the country to develop and produce petroleum, which can only be done through investments in petroleum operations and showed that this ability positively influences petroleum sector effective- ness. The above indicates the third hypothesis: • H3: The intensity of petroleum exploration and production activities positively influences govern- ment effectiveness in the petroleum sector. The government takes the attractiveness variable, which represents the ranking among selected sample countries related to the attractiveness of the fiscal regime applied by the country in terms of economic stability and balanced share of profits between the country and the oil company for 2015. Many authors (Johnston, 2000 and 2003; Seba, 2008; Thurber et al., 2011; Holden, 2013; Ke- mal, 2016) have used government take in their empirical research and showed that government take attractiveness will positively influence government effectiveness in the petroleum sector. This argumentation leads to the fourth hypothesis: • H4: Government take attractiveness positively influences government effectiveness in the petro- leum sector. The type of contract represents the contract used among selected sample countries when concluding deals with oil companies, i.e., the production sharing contract or concession (royalty and tax) contract. Some authors, such as energy economist Daniel Johnston, have argued that the type of contract does not influence the ability of the coun- try to maximize petroleum profit and its effectiveness in the petroleum sector since both types can achieve the same objectives. The data used in the model were pulled from the Petroleum Economics and Policy Solutions (PEPS) data published by IHS Energy in 2016. The above indi- cates the fifth hypothesis: • H5: The type of contract used for awarding petro- leum rights does not significantly influence gov- ernment effectiveness in the petroleum sector. The following chapter outlines the methodology, in- cluding data and analytical approaches, including MLA and ANN. 4 Methodology 4.1 Data After the variables to be used in the model were identi- fied, descriptive statistics were applied, and the variables’ descriptions were presented in Table 2. Government effectiveness data was pulled from WGI data published by the World Bank (2016). The data repre- sents an estimation of government effectiveness in each selected country for 2015. Countries were evaluated in the range from -2.5 to 2.5, where -2.5 indicates weak effec- tiveness and 2.5 indicates strong effectiveness. In order to avoid negative values, the range was adjusted by 2.5, to a range from 0 to 5, where 0 means weak and 5 means strong effectiveness. In 2015, countries were evaluated for political stability, ranging from -2.5 to 2.5, where -2.5 means weak stability and 2.5 means strong stability. These ranges were adjusted by 2.5 to a range of 0 to 5, where 0 means weak, and 5 means strong stability. Regulatory quality data were also pulled from the WGI data published by the World Bank (2016). The data repre- 370 Organizacija, V olume 57 Issue 4, November 2024 Research Papers sents an estimation of regulatory quality in 2015 in each selected country. Countries are evaluated in the range from -2.5 to 2.5, where -2.5 means weak quality and 2.5 means strong quality. The range was adjusted by 2.5 to avoid neg- ative values, and thus, a range from 0 to 5 was used, where 0 means weak, and 5 means strong quality. The intensity of petroleum exploration and production activities was pulled from the Petroleum Economics and Policy Solutions data published by IHS Energy (2016). Countries were evaluated on a scale of 1 to 10, where 1 means the lowest and 10 means the highest intensity of petroleum exploration and production activities. The same source was used to extract the data to meas- ure government take attractiveness in 2015. Countries were evaluated on a scale of 1 to 10, where 1 means low- est, and 10 means highest value. The type of contract was measured as a dummy var- iable. In the model, countries with production-sharing contracts were denoted with the number 0, while countries with a concession (royalty and tax) contract were denoted with the number 1. 4.2 Analysis To examine the influence of the defined independent variables on the dependent variable, a full MLA is run, in- cluding all independent variables that are considered pre- dictors of dependent variables. Since one of the variables appeared insignificant due to a low t ratio, that variable was dropped, and the reduced regression model was rerun (Azcel & Sounderpandian, 2009). One of several stepwise selection procedures is used. These techniques either select or eliminate variables, one at a time, in an effort to exclude those variables that either have no predictive ability or are highly correlated with other predictor variables (Kvanli et al., 2003). Stepwise procedures consist of forward regres- sion, backward regression, and stepwise regression, where stepwise regression is most commonly used. Stepwise re- gression can remove any variable whose partial F-value indicates that this variable does not contribute, given the present set of independent variables in the model (Kvanli et al., 2003). In the MLA, many problems may occur due to a large number of variables. The purpose of model diagnostics is to detect possible weaknesses of the model and, if neces- sary, to transform it. Typically, four problems (multicol- linearity, heteroscedasticity, autocorrelation of error terms, and the normality of error terms) need to be analyzed in order to prove the validity of the model (Šošić, 2004). If any of the four problems are detected, the basic model as- sumptions are not satisfied, and the validity of the model is questionable. To test the set hypothesis, it was necessary to examine the statistical dependence among variables, which is possi- ble using the MLA. The MLA shows the statistical depend- ence of one numerical variable (dependent variable) to two or more numerical variables (independent variables). To examine the influence of selected variables on government effectiveness, an MLA was used on a sample of 130 coun- tries worldwide. The dependent variable in the defined model is government effectiveness (GE), and the five in- dependent variables are political stability (PS), regulatory quality (RQ), the intensity of petroleum exploration and production activities (EPI), government take attractiveness (GTA), and type of contract (TC). Data were statistically analyzed using the programs SPSS 21 and EViews 7. The correlation matrix is used to check multicollineari- ty. The correlation matrix shows the correlation coefficients between the variables in the model. A serious multicollin- earity problem exists if the Pearson coefficient between the variables is 0.9 or greater (Belsey et al., 2004). Other multicollinearity problem indicators are variance inflation factor (VIF) and tolerance indicator (TOL), where VIF>10 or TOL<0.1 (Hair et al., 1995; Tabachnick & Fidell, 2001; Kvanli et al., 2003; O’Brien, 2007). Some authors have Table 2: Description of the variables Source: Authors’ work Variable Code Measurement Mean St. Dev Government Effectiveness GE 0-5 (0-weak, 5-strong) 2.41 0.956 Political stability PS 0-5 (0-weak, 5-strong) 2.21 0.968 Regulatory quality RQ 0-5 (0-weak, 5-strong) 2.41 0.966 Intensity of exploration and production activities EPI 1-10 (1-low, 10-high) 3.19 2.739 Government take attractiveness GTA 1-10 (1-low, 10-high) 5.98 1.504 Type of contract TC 0-production sharing contract, 1-concession (royalty and tax based) contract 0.40 0.492 371 Organizacija, V olume 57 Issue 4, November 2024 Research Papers argued that there is a possibility of moderate multicollin- earity if VIF>5 or TOL<0.2 (Bahovec & Erjavec, 2009). Since each of these indicators has certain advantages and disadvantages, they should both be examined to conclude whether multicollinearity exists. Multicollinearity often appears in empirical research, especially in regression models. Although there is no exact solution for multicollinearity, independent variables that contribute to it may be excluded from the model (Kvan- li et al., 2003). It is important to emphasize that VIF and TOL only indicate that the model is not ideal (Kvanli et al., 2003; O’Brien, 2007). A two-step approach for assessing the proposed study model has been established in previous research (Ster- nad Zabukovšek et al., 2019). To evaluate the relevance of the constructs in the proposed conceptual model, an importance-performance map analysis was employed. Furthermore, we investigated and verified the impact of independent factors on dependent variables using artificial neural networks (ANN), a computerized method used to estimate complex and non-linear features of interactions between variables. Research by Alhumaid et al. (2021) proposes that an ANN has three separate modalities: trans- fer function, network design, and learning rule. To be more precise, these modalities may be classified as feed-forward multilayer perceptron (MLP) networks, radian bases, and convolutional networks. A widely used approach is the Multilayer Perceptron (MLP) network, comprising layers of inputs and outputs linked by hidden nodes. The input layer of a neural network transfers unprocessed data to the lower layers, known as “synaptic weights.” The output of each layer is governed by the activation function em- ployed, and the most effective active function suggested is the sigmoidal function (Karlik & Olgac 2011). Therefore, this work employs the ANN to train and evaluate the the- oretical model, quantifying the importance of independent variables. 5 Results 5.1 Step 1: Multiple regression analysis Based on all the information and inputs above, the MLA has the following form: GE = 0.283 + 0.60 * TC + 0.033 * EPI + 0.115 * PS + 0.645 * RQ + 0.134 * GTA + ε (1) For the full model, all five variables were included in the model to suppose that they influence the government’s effectiveness. The MLA results are presented in Table 3. W for the model. In contrast, the type of contract (TC) variable was not shown as statistically significant since its p-value was greater than 0.05 (p-value = 0.376). The co- efficient of determination (R-square) is high (R2=0.869), indicating that the model fits the data well. This means that 86.9% of the variance of the dependent variable govern- ment effectiveness (GE) is explained by the inclusion of four independent variables (PS, RQ, GT, EPI, GTA, and TC). Table 3: Variables in the full model Note: ** statistically significant at 1%; * 5; Source: Authors’ work Variables Coefficients Standard errors t-values p-values Hypothesis Conclusion Constant 0.283 0.142 1.995 0.048* PS 0.115 0.054 2.107 0.037* H1  (+5%) RQ 0.645 0.063 10.239 <0.001** H2  (+1%) EPI 0.033 0.011 2.896 0.004** H3  (+1%) GTA 0.134 0.050 2.679 0.008** H4  (+1%) TC 0.060 0.068 0.887 0.376 H5 ∅ 372 Organizacija, V olume 57 Issue 4, November 2024 Research Papers Table 4: Correlation matrix Source: Authors’ work GE PS RQ EPI GTA TC GE Pearson’s r 1.000 PS Pearson’s r 0.718** 1.000 RQ Pearson’s r 0.919** 0.698** 1.000 EPI Pearson’s r 0.280** -0.011 0.226 1.000 GTA Pearson’s r 0.210* 0.281** 0.277 -0.109 1.000 TC Pearson’s r 0.426** 0.410** 0.462 0.248 0.443 1.000 Figure 3: Heatmap of the correlations between dependent and independent variables Source: Authors’ work Variables TOL VIF PS 0.466 2.146 RQ 0.437 2.290 EPI 0.799 1.251 GTA 0.740 1.351 TC 0.617 1.621 Table 5: Tolerance (TOL) and the variance inflation factor (VIF) Source: Authors’ work 373 Organizacija, V olume 57 Issue 4, November 2024 Research Papers Results indicate that an increase in regulatory quality within a particular country will directly increase govern- ment effectiveness, which is in line with part of Hypothesis 1, defining that regulatory quality has a positive impact on government effectiveness. Besides, an increase in govern- ment take attractiveness within a particular country will di- rectly increase government effectiveness, which is in line with the part of Hypothesis 2, defining that government take has a positive impact on government effectiveness. An increase in the intensity of petroleum exploration and pro- duction activities within a particular country will directly increase government effectiveness, which is in line with part of Hypothesis 3, defining that the intensity of petro- leum exploration and production activities has a positive impact on government effectiveness. Finally, an increase in political stability within a particular country will direct- ly increase government effectiveness, which is in line with part of Hypothesis 3, defining that political stability has a positive impact on government effectiveness. However, due to the fact that the type of contract (TC) variable did not enter into the reduced model due to its insignificance to the full model, the part of Hypothesis 5 defining that the type of contract used within the particular country does not influence government effectiveness was confirmed. The correlation matrix and associated parameters are presented in Table 4 to test for the possible presence of multicollinearity. Among independent variables, the max- imum linear correlation of 0.698 was observed between PS and RQ, while the coefficients of the linear correlations among other variables were lower. Since all Pearson’s co- efficients among independent variables are less than 0.9, we can conclude that there is no multicollinearity problem in the model. Figure 3 presents the heatmap of the correla- tions between dependent and independent variables. Multicollinearity was also tested with tolerance and the variance inflation factor (VIF). As shown in Table 5, tolerance was higher than 0.1, and VIF was lower than 10 for all variables included in the model. Based on the re- sults, it can be concluded that the model has no multicol- linearity problem. Additionally, the normality of residuals was test- ed using the Kolmogorov-Smirnov, Shapiro-Wilk, and Jarque-Bera tests. The results suggest that the empirical significance level for the Kolmogorov-Smirnov test is 0.200, the Shapiro-Wilk test is 0.786, and the Jarque-Bera test is 0.829. Accordingly, it can be concluded that at the significance level of 0.05, the null hypothesis stating that residuals are normally distributed cannot be rejected. The model diagnostics have shown that each prob- lem’s analysis fulfilled the basic model assumptions and proved that the initial assumptions were not undermined. The model diagnostics have shown that each prob- lem’s analysis fulfilled the basic model assumptions and proved that the initial assumptions were not undermined. 5.2 Step 2: Artificial neural network analysis The ANN was conducted using JASP 0.19. The same variables were used as in the MLA. However, only those variables that were significant in the MLA were used for training in the development of the ANN, indicating that TC was discarded. In the current study, the logistic sigmoid function aids in activating both output and hidden neurons with algo- rithm Rprop- (Resilient Propagation), which is a gradi- ent descent-based optimization algorithm primarily used Figure 4: Neural network training graphs Source: Authors’ work 374 Organizacija, V olume 57 Issue 4, November 2024 Research Papers for training ANN (Figure 4a). It is a variant of the Rprop algorithm that modifies the weight updates by adapting the step sizes based on the sign of the partial derivatives of the loss function. Unlike traditional gradient descent, Rprop- ignores the magnitude of the gradient, focusing instead on its sign to decide the direction of the update, making it effective for handling vanishing gradients and improving convergence speed in training deep ANN (Igel et al., 2005). In order to reduce overfitting in the ANN, we employed cross-validation techniques with a ratio of 90:10:10 for testing, training, and validating the collect- ed data. The ANN models exhibit relative errors of 0.200 and 0.344 for training data and testing data, respectively (Figure 4b). These results suggest that the optimum ideal number of layers for the models is 4. Based on the minimal rise in relative errors to testing from the training dataset, in conjunction with the use of ANN, it can be inferred that the suggested research models exhibit higher efficiency. The model summary for the Neural Network Regres- sion in Table 6 provides key metrics for evaluating the model’s performance. The network consists of 4 hidden layers with 18 nodes each, and the data is split into training (n=105), validation (n=12), and test (n=13) sets. The mod- el is optimized based on the validation set’s mean squared error (MSE), which is 0.200, while the test set MSE is slightly higher at 0.344. In the present study, the sigmoid function stimulates the activity of both output and hidden layers. The an- alytical method was employed to determine the optimal number of concealed layers, which was calculated to be 1 (Figure 5). Furthermore, we employed a cross-validation approach to assess and train the collected data to prevent overfitting in the ANN (Ahmed et al., 2021). Table 6: Model Summary: Neural Network Regression Note: The model is optimized with respect to the validation set mean squared error. Source: Authors’ work Hidden Layers Nodes n(Train) n(Validation) n(Test) Validation MSE Test MSE 4 18 105 12 13 0.200 0.344 Figure 5: Network Structure Plot Source: Authors’ work 375 Organizacija, V olume 57 Issue 4, November 2024 Research Papers Table 7 presents the feature importance metrics, rep- resented by the mean dropout loss, indicating the relative significance of each variable in the model. A lower mean dropout loss suggests that the variable is more important to the model’s predictive capability. In this table, regulatory quality (RQ) is of the highest importance, suggesting it is the most crucial variable. Political Stability (PS) follows while Government Take Attractiveness (GTA) has the low- est importance. Different levels of permutations revealed similar results, indicating the stability of the solution. In addition, we examined the prediction performance plot to validate the neural network’s computational effi- ciency and precision. The neural network model of the data produced a Root Mean Square value of 0.344 and a coefficient of determination of 0.891. These values outper- formed those obtained from MLS, suggesting that ANN has considerable potential for analyzing government ef- fectiveness in the petroleum sector. Figure 6 verifies a well-defined correspondence between the observed and anticipated values of both models. Table 7: Feature importance metrics Mean dropout loss Variable 25 permutations 50 permutations 100 permutations RQ 1.724 1.735 1.732 PS 0.631 0.635 0.641 EPI 0.609 0.612 0.613 GTA 0.559 0.558 0.555 Source: Authors’ work Figure 6: Predictive Performance Plot Source: Authors’ work 6 Conclusion The changing relationships in global petroleum mar- kets during the late 20th century and the increase and fluc- tuation of petroleum prices in the early 21st century have increased the economic importance of revenue and profit from their production. As a result, the legal relationships in petroleum exploration and production processes and the regulation of these relationships through state interven- tions have gradually achieved a universal value and be- come a crucial subject within the competence of legislative and executive government authority. The objective of the study is to analyze the determi- nants that impact the effectiveness of government oper- ations in the petroleum sector. The present study utilizes MLA to investigate the possible influence of political sta- bility, regulatory quality, the level of petroleum explora- tion and production activities, government take, and con- tract type on government performance in the petroleum 376 Organizacija, V olume 57 Issue 4, November 2024 Research Papers sector. Moreover, ANN was investigated to ascertain the importance of independent variables. The empirical research indicated that political stabil- ity influences government effectiveness in the petroleum sector and additionally stressed the importance of politi- cal stability in developing the national petroleum industry. Regulatory quality was shown to be another factor influ- encing government effectiveness in the petroleum sector, thus proving that petroleum legislation is the main factor in regulating the complex relationships between govern- ments and oil companies in petroleum exploration and pro- duction activities. Accordingly, the empirical research showed that the intensity of petroleum exploration and production activi- ties influences government effectiveness in the petroleum sector and thus demonstrated that countries with a higher intensity of petroleum exploration and production activi- ties also have greater government effectiveness in the pe- troleum sector. The correlation between government take in terms of attractiveness ranking and government effec- tiveness verified that countries with a more attractive gov- ernment take and the fiscal regime has better government effectiveness. However, empirical research has shown that the type of contract does not influence government effec- tiveness in the petroleum sector. Hypothesis H1-H4 was accepted since the model in- dicated that government effectiveness is positively in- fluenced by the country’s political stability, regulatory quality, intensity of petroleum exploration and production activities, and government take attractiveness. In contrast, hypothesis H5 was rejected due to the lack of a relationship between government effectiveness and type of contract. One of the most important examinations was the em- pirical confirmation that the type of contract used when awarding petroleum exploration and production rights to oil companies does not influence government effective- ness. The theory presented suggested that generalizations are often made about the superiority of a concessionary system over a production-sharing system from the oil com- pany’s point of view, despite the overwhelming similari- ties from the economic, accounting, and financial points of view, suggesting that the choice of system may not be such a critical issue. The theory suggested that neither type of contract, concession (royalty and tax based) contract or production sharing contract, is better nor worse, as, from the economic perspective, the same objectives can be achieved. The empirical research demonstrated that the type of contract did not influence government effective- ness in the petroleum sector, thus further supporting the presented theory. This is one of the most important the- oretical contributions since the theory to date has specu- lated based on various observations without quantification through empirical research. The neural network analysis provided valuable in- sights into the key determinants of government effective- ness in the petroleum sector. The neural network model, featuring four hidden layers and 18 nodes, demonstrated that regulatory quality and political stability were the most significant variables, with the highest predictive capabili- ty. The neural network’s performance, as indicated by the validation and test mean squared errors, suggests a robust model that complements the findings of the MLA. While the analysis provided significant insights, it is important to acknowledge certain limitations. The relative- ly small sample size may restrict the generalizability of the findings, and the complexity of the neural network model poses a risk of overfitting despite the use of cross-valida- tion techniques. Additionally, the study primarily focused on a specific set of variables, potentially overlooking other factors that might influence government effectiveness in the petroleum sector. Future research should consider expanding the da- taset to include a broader range of countries and varying economic contexts, which could enhance the model’s ro- bustness and applicability. Further exploration of alterna- tive machine learning techniques, such as deep learning or ensemble methods, could provide deeper insights and improve predictions’ accuracy. Additionally, incorporat- ing external factors, such as global oil market dynamics and technological advancements in petroleum extraction, could offer a more comprehensive understanding of gov- ernment effectiveness in the petroleum sector. Acknowledgement This paper is the result of the project “Entrepre- neurship and management in modern business” UNIN- DRUŠ-24-1-3 of the University North, Croatia. Literature Ahmed, S. R., Kumar, A. K., Prasad, M. S., & Keerthi- vasan, K. (2021). Smart IOT based short term forecasting of power generation systems and quality improvement using resilient back propagation neural network. Revista geintec-gestao inovacao e tecnolo- gias, 11(3), 1200-1211. Alhumaid, K., Habes, M., & Salloum, S. A. (2021). Examining the factors influencing the mobile learning usage during COVID-19 Pandemic: An Integrated SEM-ANN Method. IEEE Access, 9, 102567-102578. https://doi.org/10.1109/access.2021.3097753 Azcel, A., & Sounderpandian, J. (2009). Complete Busi- ness Statistics. New York: The McGraw-Hill. Bahovec, V ., & Erjavec N. (2009). Introduction to Econo- metric Analysis (in Croatian). Zagreb: Element. Belsey, D. A., Kuh, E., & Welsch, R. E. (2004). Regres- sion diagnostics: Identifying influential data and sources of collinearity. New York: Wiley. 377 Organizacija, V olume 57 Issue 4, November 2024 Research Papers Brunnschweiler, C. N., & Bulte, E. H. (2008). The re- source curse revisited and revised: A tale of paradoxes and red herrings. Journal of Environmental Econom- ics and Management, 55(3), 248-264. https://doi. org/10.1016/j.jeem.2007.08.004 Doric, B. (2017). A sustainable administrative design for petroleum sector performance management. Doctoral dissertation. University of Ljubljana. Falola, T., & Genova, A. (2005). The Poli- tics of the Global Oil Industry. https://doi. org/10.5040/9798400698781 Green, J., & Smith, A. (2023). Energy Law in a Nutshell. West Academic Publishing. Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. C. (1995). Multivariate Data Analysis (3 rd ed). New York: Macmillan. Hammerson, M. (2011). Upstream Oil and Gas: Cases, Materials and Commentary (1 st ed.). London: Globe Law and Business. Heller, P. R. P., & Marcel V . (2012). Institutional De- sign in Low-Capacity Oil Hotspots. Retrieved from http://www.resourcegovernance.org/sites/default/ files/documents/institutional-design-in-low-capaci- ty-oil-hotspots.pdf Holden, S. (2013). Avoiding the resource curse the case Norway. Energy Policy, 63, 870-876. https://doi. org/10.1016/j.enpol.2013.09.01 Igel, C., Toussaint, M., & Weishui, W. (2005). Rprop using the natural gradient. Trends and Applications in Constructive Approximation, 259-272. https://doi. org/10.1007/3-7643-7356-3_19 IHS Energy. (2016). IHS Petroleum Economics and Policy Solutions - Complete Country E&P Informa- tion and Petroleum Sector Risk Analysis. Available at: https://cdn.ihs.com/www/pdf/PEPS-brochure.pdf IHS Energy (2016a). Petroleum Economics and Policy Solutions (PEPS). Retrieved from https://www.ihs. com/products/oil-gas-petroleum-news-peps.html Johnston, D. (1994). International Petroleum Fiscal Sys- tems Analysis (1 st ed.). Tulsa: PennWell Corporation. Johnston, D. (2000). Fiscal System Design: The Ideal System. Petroleum Accounting and Financial Man- agement Journal, 19(2), 121-127. Johnston, D. (2003). International Exploration Econom- ics, Risk, and Contracts Analysis. Tulsa: PennWell Corporation. Karlik, B., & Olgac, A. V . (2011). Performance analysis of various activation functions in generalized MLP architectures of neural networks. International Jour- nal of Artificial Intelligence and Expert Systems, 1(4), 111-122. Kaufman, R. K., Dees, S. & Sanchez, M. P. (2004). Does OPEC Matter? An Econometric Analysis of Oil Pric- es. The Energy Journal, 25 (4), 68. Kemal, M. (2016). Ownership Rights versus Access Rights Allocation to Critical Resources: An Empiri- cal Study of the Economic Impact of Changes in Oil Governance. Colorado School of Mines, Working Paper No. 2016-02. Retrieved from http://econbus. mines.edu/working-papers/wp201602.pdf Kvanli, A. H., Pavur R. J. & Keeling K. B. (2003). Introduction to Business Statistics. United States: Shepherd, Inc. Markus, U. (2014). Oil and Gas: The Business and Poli- tics of Energy. United States: Palgrave Macmillan. O’Brien, R. M. (2007). A Caution Regarding Rules of Thumb for Variance Inflation Factors. Quality and Quantity, 41 (5), 673–690. Parra, F. (2004). Oil Politics: A Modern History of Petro- leum. London: IB Tauris. Seba, R. D. (2008). Economics of worldwide petroleum production (3 rd ed.). Tulsa: PennWell. Shaffer, B. (2011). Energy Politics. Philadelphia: Univer- sity of Pennsylvania Press. Simkins, B., & Simkins, R. (2013). Energy Finance: Analysis and Valuation, Risk Management, and the Future of Energy. United States: Wiley. Šošić, I. (2004). Applied statistics (in Croatian). Zagreb: Školska Knjiga. Sternad Zabukovšek, S., Kalinic, Z., Bobek, S., & Tom- inc, P. (2019). SEM-ANN based research of factors’ impact on extended use of ERP systems. Central European Journal of Operations Research, 27(3), 703-735. https://doi.org/10.1007/s10100-018-0592-1 Tabachnick, B. G., & Fidell, L. S. (2001). Using Mul- tivariate Statistics (4 th ed.). Boston, MA: Allyn and Bacon. Thurber, M. C., Hults, D. R., & Heller, P. R. P. (2011). Exporting the “Norwegian Model”: The effect of ad- ministrative design on oil sector performance. Energy Policy, 39(9), 5366-5378. https://doi.org/10.1016/j. enpol.2011.05.027 Thurman, H. V . (2022). The Legal Landscape of Oil and Gas. Cambridge University Press. World Bank. (2016). Worldwide Governance Indicators. Retrieved from http://info.worldbank.org/governance/ wgi/index.aspx#home Barbara Dorić a scientist (PhD in economics) who worked for Croatian most established companies in the last 10 years. She received her PhD education at the University of Ljubljana, Faculty of Economics. She has published several scientific papers in the field of economics and has participated in several international conferences. In her career, she was Executive Director of the Centre for Monitoring of Energy Business and Investments, President of the Management Board of Croatian Hydrocarbon Agency, Managing Director at LNG Croatia, Member of the Management Board at INA Group, etc. 378 Organizacija, V olume 57 Issue 4, November 2024 Research Papers Dinko Primorac is a Croatian scientist, university professor and entrepreneur. He received his undergraduate education at Webster University, graduated from the Faculty of Economics in Zagreb, and received his PhD from Megatrend University. As a professor, he has taught at several Croatian and international universities. He has published numerous scientific and professional articles, as well as university textbooks. He has participated in numerous international scientific conferences, and as a reviewer, he gives his contribution to professional and scientific journals. As a scientist, he participates in several domestic and international economic scientific projects. His specialities are entrepreneurship and macroeconomics. He is a member of several supervisory boards of renowned Croatian companies. Mirjana Pejić Bach is a full professor at the Department of Informatics, Faculty of Economics in Zagreb. She holds a PhD in system dynamics modelling from the Faculty of Economics, University of Zagreb. She was trained at the MIT Sloan School of Management in system dynamics and OliviaGroup in data mining. Mirjana is the leader and collaborator of numerous projects in which she cooperates with Croatian companies and international organizations, mainly through European Union projects and the bilateral research framework. Her research areas are the strategic application of information technology in business, data science, simulation modelling, research methodology, qualitative and quantitative, especially multivariate statistics and modelling structural equations.