Determinants of the Corporate Financing Structure in the Energy and Mining Sectors; A Comparative Analysis Based on the Example of Selected EU Countries for 2012–2020

Abstract

The main aim of the paper is to examine the interdependence of corporate financing structure on selected determinants in the energy and mining sectors of the economy. In addition, a comparison of the results between these economic sectors is made. In order to increase the representativeness of the sample, countries from both the “old” European Union (Germany, France, Great Britain, Spain, Italy, and Sweden) and newly admitted countries (Poland, the Czech Republic, Hungary, Romania, Slovakia, and Bulgaria) were included in the study. The research used data from the Orbis database for 2012–2020. A one-factor linear panel model was used to verify the hypotheses. The research partly confirmed the hypotheses. A negative relationship between the financing structure and profitability in both analysed sections was clearly established. The second determinant whose influence on the financing structure in both sections was found to be unambiguous was the ratio of current liabilities to current assets. However, this influence was positive. Another determinant whose influence on the financing structure was confirmed to be unequivocal and positive in nature was the size of the company. This occurred only in the energy sector. The research revealed that other determinants, such as asset structure, interest, and noninterest tax shields, also affect the financing structure of companies, but the statistical significance of these relationships is ambiguous.

1. Introduction

There can be little doubt that energy plays a key role in any country’s economy. Without an adequate energy supply, not only could other economic sectors not function, but also development in other countries would be impossible.

As a result of steadily increasing global economic growth, the demand for energy also increases. On the other hand, hitherto energy generation technologies based to an exceptionally large extent on fossil fuels (fossil energy carriers) cause worry and progressive climate change through high CO₂ emissions. Undoubtedly, these changes, together with environmental degradation, pose a grave threat to the world. Hence, the energy sector is attracting increasing interest, and various initiatives and actions have been taken for many years to enable a profound energy transformation, the most important of which are identified below.

The first initiative in the fight against climate change was the signing of an international agreement in Rio de Janeiro in 1992 (United Nations Framework Convention on Climate Change–UNFCCC or FCCC), which set out the principles for international cooperation on reducing greenhouse gas emissions responsible for global warming. This was followed up by the United Nations Conferences on Climate Change, the so-called Conferences of the Parties (COP), which have been held annually since 1995. The most significant events include the signing of the Kyoto Protocol to the United Nations Framework Convention on Climate Change (COP 3) on 11 December 1997 and its subsequent ratification by 141 countries (COP 11) and entry into force on 16 February 2005. It expired on 31 December 2012; however, the European Commission proposed a new treaty in the form of an amendment dubbed the ‘Doha amendment’ to the Kyoto Treaty on 6 November 2013. Another particularly important event was the United Nations Framework Convention on Climate Change in Paris [1], where on 12 December 2015, 195 countries agreed to adopt a final global agreement known as the Paris Agreement, under which emissions were to be reduced as part of the way to reduce greenhouse gas emissions. In this way, a global compromise was reached for the first time in the framework of climate change mitigation. In the document, the parties agreed to reduce carbon dioxide emissions ‘as soon as possible’ and assured that they would make every effort to keep global warming significantly below the 2 °C threshold.

However, at the recent climate change conference in Glasgow [2], China and the US, the largest consumer and producer of fossil fuels, respectively, together with India, decided that, in the context of coal use, the term phase out should be changed to phase down.

In addition, the energy policy of EU countries that has been in place for several years has led to the creation and adoption, in 2020, of the European Green Deal [3], which calls for a more efficient use of resources through a transition to a clean, ‘closed-loop’ economy, the combating of biodiversity loss, and the reduction of pollution levels.

As part of the European Green Deal, the EU’s ‘Fit for 55’ climate regulation package was developed, aiming to reduce greenhouse gas emissions in Europe by 2030 by at least 55% compared to 1990 levels and achieve climate neutrality by 2050 [4]. The package comprises a set of proposals for 13 interlinked pieces of legislation, which became part of the new European legal order with the adoption of the EU Climate Law. It must be added that ‘Fit for 55’ is not yet binding law and refers to a transitional period leading ultimately to climate neutrality across the EU by 2050.

Based on a detailed analysis and assessment of documents and assumptions introducing climate change worldwide, it should be noted that EU countries have proposed to adopt the most rigorous action plan in this regard, consisting of a transition to ‘green energy’ in the shortest possible period.

The ‘Fit for 55’ package is a tool that could, however, lead to a huge impoverishment of EU citizens. The proposal contains very ‘aggressive directives and regulations’ that include various types of taxation, charges, penalties, injunctions, and prohibitions for nonfulfilment of the provisions adopted by individual EU countries. The proposed changes should therefore be re-examined with a view to somewhat softening the proposed changes and, in particular, lengthening the time taken to reach the targets set so that the economies of the individual countries are able to bear the burden of the energy transition in the first place.

Energy generation is based on different technologies using renewable and nonrenewable sources. The latter mainly include fossil fuels, such as coal and lignite, oil, and natural gas. Both previously and currently, they are the primary media for energy generation. It is estimated that approximately 37% of global electricity comes from coal alone. Taking this into account, the energy sector is highly dependent on the mining sector.

The current energy transition is bound to require major changes in the asset and liability structures and equity structure of energy companies and massive investments in each country. These changes will consequently affect companies’ choices of financing. The main objective of the choice should be to search for optimal solutions regarding sourcing short- and long-term financing in such a way as to ensure the stability of financing and minimise the cost of raising capital.

Decisions concerning the formation of the financing structure of enterprises constitute one of the most important areas of modern corporate management. They are the subject of much analysis and research. Both theoretical and empirical work recognise a wide range of different determinants of corporate financing. However, not all sectors are sufficiently studied. The energy and mining industries are two such understudied ones. There is also a lack of comparative research, both between sectors and countries.

This paper consists of four parts. The first presents a review of the relevant literature and research into capital structure theory as well as the determinants of corporate finance, with a strong focus on the energy and mining sectors. The second part describes the objectives, scope of the study, research hypotheses, and characteristics of the data. The third part presents the model’s construction and the research methodology. The last part presents the results of empirical research for the selected EU countries.

2. Overview of Main Theories of and Research into Capital Structure Formation in an Enterprise

One of the key issues in corporate financial management involves the shaping of the relationship between the sources of financing raised by the company. However, in empirical studies, instead of financing sources, authors more often use the concept of capital sources, which forms a capital structure corresponding to the sources. It should also be emphasised that these concepts are variously interpreted and, consequently, research can be conducted from different points of view.

Based on a literature review, several approaches to defining them can be distinguished. The broadest approach considers the entire structure of liabilities and shareholders’ equity on a company’s balance sheet, commonly referred to as sources of finance (sources of capital are then the same as sources of financing), as the sources of capital (financing) from which the capital structure directly derives. The leading proponent of this approach is Masulis [5], according to whom the capital structure includes not only financial instruments sold through public and private issues or bank loans, but also trade liabilities, leasing contracts, tax and pension liabilities, payroll liabilities, guarantees provided, and other liabilities [6]. Supporters of such an understanding of the sources (structure) of capital include Ross, Westerfield, and Jaffe [7] and Higgins [8].

According to a second approach, capital structure is viewed as the ratio of the value of long-term debt to equity [9,10,11]. According to them, the structure (sources) of capital should include only the so-called “permanent” financing, i.e., equity, preference capital, and long-term (interest-bearing) debt [6]. In light of this interpretation, the so-called “permanence” of financing should be the criterion used to determine the sources (structure) of capital.

In another approach, according to Brealey and Myers [12], the capital structure should include only securities issued by companies, consisting of debt securities (e.g., bills and bonds) and ownership securities (e.g., ordinary and preferred shares). It should be noted that this approach ignores liabilities arising from borrowing secured in the banking system [13]. Therefore, this approach to capital structure should be considered to be the narrowest.

The basis for the fourth and most common approach to capital structure is a proper understanding of debt and the main feature of capital, which is the payment of interest on the debt incurred [14]. Considering this, a firm’s capital should include accumulated equity and debt capital, which is composed of all interest-bearing liabilities incurred. In this approach, debt capital should include not merely noninterest-bearing liabilities, such as, e.g., trade payables, wages and salaries, duties, and taxes.

Of the above-mentioned approaches, this research adopts the broadest understanding of capital structure, which is based on the entire structure of a company’s liabilities. In justifying the above choice, it should be pointed out that there are certain problems in obtaining the relevant data, e.g., companies’ nondisclosure of certain items in financial statements or their aggregation, which would significantly limit the research sample.

It should be stated that capital structure theory is, on the one hand, of unique importance in financial theory, but on the other hand, it is very complex and covers numerous aspects of a company’s activities. Among the many works providing an overview of extensive empirical research into capital structure, one should point to the work by Harris and Raviv [15] and the work by Nehrebecka, Białek-Jaworska and Dzik-Walczak [16].

The study of capital structure has a long history. The first attempts to describe the relationship between financing structure and the cost of capital and goodwill were found in the early 1950s in the work of Durand, who developed the net operating income (NOI) theory, the net income (NI) theory and the trade-off theory.

A break-through for the later development of capital structure theory came with the studies conducted by Modigliani and Miller [14]. The first ones concerned a tax-free economy, while the second ones concerned an economy with taxes. Another model developed by Miller showed that capital structure may, but does not have to, affect the change in the market value of a company. It depends on the tax system in place and the relationship between the different tax rates. Another example of an extension of the Modigliani–Miller theory through the use of a noninterest tax shield is the model developed by DeAngelo and Masulis [17], who showed that an increase in a company’s debt is not necessarily accompanied by a reduction in the amount of income tax.

Modigliani and Miller’s theories, however, ignored the risk of insolvency and bankruptcy that accompanies an increase in debt, which has been the subject of much criticism and led to the development of the bankruptcy cost theory. The most prominent contributions involved works by Baxter [18], Stiglitz [19], Kraus and Litzenberger [20], Scott [21] and Kim [22]. Under this theory, the optimal relationship between equity and debt is determined by considering both the benefits and risks that arise from debt financing, while the substitution of equity with debt (or vice versa) occurs until the maximum market value of the company is achieved.

As stated by Kraus and Litzenberger [20], an increase in the expected cost of bankruptcy diminishes a company’s tax shield. The bankruptcy cost theory of capital structure assumes the existence of an optimal capital structure, which is the result of an appropriate proportion between equity and debt capital. Its determination should consider both the tax benefits and the costs of financial distress, i.e., indirect and direct bankruptcy costs associated with debt financing. In an extension by Jensen and Meckling [23], the substitution theory also includes agency costs.

According to the agency cost theory, the formation of a company’s capital structure is affected by agency costs resulting from conflicts of interest between different stakeholder groups (shareholders, creditors, management, employees, etc.). Research into this area was initiated by Jensen and Meckling [23]. The agency cost theory indicates that in shaping the capital structure, consideration should be given not only to the benefits of debt financing, such as the tax shield effect, but also to the opportunities to reduce agency costs.

In the late 1970s, a new strand of capital structure theory emerged that was based on assumptions of an imperfect market, considering the presence of information asymmetries. Within this strand, two main approaches can be distinguished: the Pecking Order Theory and the Signalling Theory.

The pecking order theory is a competing approach to the previously mentioned substitution theory. It assumes significantly different criteria when shaping the capital structure. According to this theory, internal sources of financing are preferred as a first priority, followed, when necessary, by external financing–debt issuance first and equity shares last. The signalling theory, on the other hand, assumes that changes in the capital structure carry specific information about the financial health of the company, which signals to investors’ insider information held by well-informed individuals. Investors tend to positively perceive management’s decisions to increase debt based on the belief that the company’s future financial standing will be favourable and will allow the company to timely repay borrowed capital.

The second half of the 1980s witnessed the development of capital structure models using features of corporate organisation theory. These include a theory based on product and factor (resource) production interactions and a theory of competition for control of the company. The theory based on interactions in the market for products and production resources can be divided into two research streams. The first exploits the relationship between a company’s capital structure and its strategy when it competes in the market for products. On the other hand, the second area of research highlights the relationship between a company’s capital structure and the characteristics of its products or production resources. By contrast, the theory of competition for the acquisition of control of a company focuses on the market value of the right to control the company, which depends on the acquisition of various equity instruments.

In the context of the above theories, it is worth mentioning some of the key findings of the research conducted by Frank and Goyal [24]. They found that profitable companies reporting high profitability ratios tend to have lower leverage ratios, which is consistent with the hierarchy of the theory of financing sources. In addition, large firms have higher debt ratios, which is consistent with the substitution theory, and firms with significant physical assets have higher leverage ratios, which is consistent with the substitution theory.

It should be noted that the main theories of capital structure were being developed prior to the late 1980s. On the other hand, after the publication of the works of Myers [25] and Myers and Majluf [26], the emphasis of further research was on explaining the reasons why companies behave in a certain way when choosing their sources of financing. The most extensive research involved studies looking for factors influencing capital (financing) structure, and this research continues to this day. Additionally, although numerous studies and models identify a large number of potential factors determining capital structure, as Harris and Raviv [15] note, only a relatively small number of so-called “general principles” have been identified. The empirical research conducted in this area points to the complexity of this problem arising from the multiplicity of potential factors influencing the level of corporate indebtedness and their interrelationships. Indeed, each financing structure is unique and reflects the unique combination of a company’s circumstances. In addition, the various factors influencing the structure have their own intensity, with some factors being more important in a particular industry, while others may be more important in a particular legal system or country. Corporate financing structures are also subject to change over time to reflect economic developments.

Despite existing problems, further empirical research should seek to best map and understand the economic reality in which decisions on the choice of specific sources of finance are made. In particular, such research should focus on those determinants that are relevant and examine their nature and significance in shaping the financing structure. On the other hand, company management, in the absence of an optimisation account, should, when making their financial decisions, determine and strive for a target financing structure that, from an economic point of view, would be the best solution for them.

On the basis of the literature review, it can be concluded that important factors of a microeconomic nature found in empirical studies include company size, its profitability and profit volatility, type (structure) of assets, noninterest tax shields, industry and product specificity, growth rate and development forecasts, cost and availability of capital and business risk. Macroeconomic factors most commonly featured in research include the tax system, inflation, capital market situation, interest rates, country-specific factors and legal solutions. Due to the multitude of determinants featured in the research, only those investigated in this paper will be discussed below.

Many studies show that the size of a company significantly affects the company’s debt level. Both in theory and in practise, large companies tend to have a higher debt ratio. One reason is believed to be that the ratio of bankruptcy costs (in the event of liquidation) to a company’s market value is higher for small entities than for large ones. In addition, large companies can more easily protect themselves against bankruptcy, e.g., by diversifying production. Thus, higher debt for large companies does not necessarily indicate a higher probability of bankruptcy, and even if it occurs, its costs will be relatively lower. Hence, the share of debt in the capital structure can be expected to increase with the size of the entity [6,27,28]. This view is also supported by Martin, Cox and McMinn [29] and Boquist and Moore [30], according to whom the use of leverage is dependent on the size of the company.

Empirical research shows that the size of the enterprise is, in most cases, strongly positively correlated with its leverage ratio [28]. Most studies confirm that large companies are more leveraged than small ones. A positive relationship between the debt level and company size has been shown by studies conducted by, e.g., Titman and Wessels [31], Rajan and Zingales [32], Fama and French [33], Booth et al. [34], Gonenc [35], Bauer [36], Jiraporn and Gleason [37], Berk [38], Antoniou et al. [39], Akhtar and Oliver [40], Psillaki and Daskalakis [41], Črnigoj and Mramor [42], Avarmaa et al. [43], Hernádi and Ormos [44], Kędzior [45], Jõeveer [46] and Jaworski and Czerwonka [47]. However, some studies have shown an opposite (negative) relationship between debt level and company size [24,48].

Another important factor shaping a company’s capital structure is the profitability of the company, usually measured in terms of the profitability of assets or sales. The impact of profitability on capital structure is confirmed, among others, by the pecking order theory. It argues that high profitability contributes to lower indebtedness, as the company is then in a better position to allocate its profit to financing new investment projects. Low-profitability companies, on the other hand, are forced to take out loans [49]. The negative impact of profitability on capital structure appears in the vast majority of studies. More profitable companies made less use of debt, suggesting that the main source of financing for their operations was their profit [31,32,35,36,37,38,39,40,41,42,43,44,45,47,50].

The opposite conclusion follows from the signalling theory, according to which companies boasting high profitability and good financial health have high levels of debt. A positive relationship between profitability and debt levels has been shown in studies conducted by Berkman et al. [51] and Koralun-Bereźnicka [52]. The above discrepancies should serve as a particular motivation for further research.

In terms of profitability, it is worth paying attention to the level of profits as well as their volatility. As highlighted by some research authors themselves, the latter is among the more important factors reflecting negatively on leverage [53]. Companies with high earnings volatility will curb their debt financing because of the financial risks associated with debt servicing [54]. Investors will also be less inclined to commit their funds to these companies, especially in countries where lender protection is weak [55]. Significant earnings or operating cash flow volatility also make it difficult to predict a company’s financial position, especially its cash flows [45]. Such entities should therefore be more cautious when drawing on external capital to reduce liquidity risk.

According to some theories of capital structure, in particular, the bankruptcy theory, agency costs and signalling theories, capital structure depends on the structure of assets available to the company. This thesis is justified on the grounds that different types of assets provide different degrees of protection for creditors, who, depending on this structure, are exposed to different degrees of loss in the event of a possible bankruptcy of the debtor. Long-term assets, compared to current assets, generally provide better security for liabilities and are less exposed to impairment. The ability to incur debt therefore depends to a large extent on the values of those assets that can function as collateral. According to the bankruptcy theory and agency costs theory, entities with a better ability to secure debt will report a relatively higher proportion of debt in their capital structure [6]. The possibility of increasing a company’s indebtedness, depending on the increase in the share of long-term assets in total assets, is confirmed, inter alia, by Dietrich [56] and De Miguel and Pindado [57]. Similar conclusions have also been drawn by: Rajan and Zingales [32], Gonenc [35], Frank and Goyal [58], Berk [38], Gaud et al. [59], Antoniou et al. [39], Akhtar and Oliver [40], Ghani and Bukhari [60], Avarmaa et al. [43], Hernádi and Ormos [44], Kędzior [61] and Berkman et al. [51].

However, in theories that focus on information asymmetry, i.e., the pecking order theory or the signalling theory, the effect of asset type (structure) on capital structure is different. From the point of view of these theories, a higher share of long-term assets in the asset structure should be accompanied by a relatively lower level of debt, as evidenced by studies conducted, inter alia, by Bauer [36], Psillaki and Daskalakis [41], Črnigoj and Mramor [42], Jõeveer [46], Jaworski and Czerwonka [47] and Koralun-Bereźnicka [52].

A company’s capital structure can also be shaped by interest and noninterest tax shields. Noninterest tax shields are ‘substitutes’ for interest tax shields and arise from the occurrence of elements other than interest on borrowed credit that reduce the tax base. The most significant elements are depreciation and investment allowances.

Tax theories of capital structure argue that the formation of capital structure is significantly influenced by taxes that result from the specific tax system in place in a country. In general, the incidence of corporate income taxes in the economy increases the profitability of debt financing and, at the same time, increases a company’s market value. The higher the tax shield, the greater its benefits should be. However, this is not always obvious when considering not only corporate income taxes, but also personal taxes and noninterest tax shields.

Considering noninterest tax shields and their impact on debt levels, the results of existing studies vary the most in relation to other variables. Only a few studies confirm a positive relationship (e.g., Jaworski and Czerwonka [47], Koralun-Bereźnicka [52] or a negative one (e.g., Bauer [36]). In most studies, the nature of the impact of this variable on the debt level is indeterminate and ambiguous (e.g., Gaud et al. [59], Mokhova and Zinecker [62]).

Liquidity may also be a determinant of capital structure. On the one hand, companies with higher liquidity may maintain a higher level of debt due to their ability to meet their current obligations, in which case the relationship between liquidity and debt will be positive. On the other hand, companies with high cash reserves may use them to finance investments, including growth activities. In this case, the relationship will be negative [63,64].

However, large companies boasting high liquidity and profits can maintain a conservative capital structure and avoid debt. In turn, when the share of net profit in equity is high, companies can decide to pay dividends. The dividend pay-out rate decreases when most of the equity comes from the issue of shares [65].

According to numerous authors, classical capital structure theories can be a good point of reference for analysing the capital structure of energy companies [66,67,68,69]. The risk of excessive indebtedness for energy companies seems to be lower than for other business entities [70]. It can also be concluded that, in this respect, companies belonging to the mining and quarrying section follow suit. Firstly, the value of long-term assets is relatively high in the above groups of companies, which provides a good form of loan collateral. Secondly, the state is the dominant investor in them, which effectively limits financial risk. In addition, the global demand for electricity continues to steadily grow.

In the world literature on the subject, research on the capital structure of enterprises and its determinants has been conducted for many years. However, the energy sector, and in particular the mining sector, has received little attention from researchers so far. Among these studies, the following works should be mentioned: [51,66,67,68,69,70,71,72,73,74,75,76,77,78,79].

3. Aims and Scope of the Study, Research Hypotheses and Data Characteristics

The main aim of this paper is to analyse and evaluate corporate financing structures in the energy and mining sectors in selected EU countries. The achievement of this aim is served by subobjectives consisting of the identification and study of the impact of selected microeconomic determinants of corporate financing structures. In particular, the relevance, strength and direction of the impact of these factors will be examined.

Twelve EU countries were selected for the study and subsequently divided into two study groups. The first group includes six countries, which are some of the strongest economies in the old EU (Germany, France, the UK, Spain, Italy and Sweden). In turn, the second group consists of the six most significant economies from the new EU countries (Poland, Czech Republic, Hungary, Romania, Slovakia and Bulgaria). The individual countries were selected in such a way that it was also possible to assess whether the conditions of a given economic system affect the formation of companies’ financing structures.

In order to meet the objectives of the research and draw on the existing capital structure theories in pertinent literature, the following research hypotheses were formulated:

• H1. 

  The financing structure of the companies studied, as determined by the ratio of total liabilities to total assets, depends significantly on the structure of such assets themselves, the ratio of current liabilities to current assets, the size of the company, its profitability and the interest and noninterest tax shield.
• H2. 

  The share of total liabilities in total assets is positively correlated with the ratio of current liabilities to current assets, with company size, and with the interest and noninterest tax shield.
• H3. 

  The share of total liabilities in total assets is negatively correlated with the share of long-term assets in total assets and with profitability.

Although research into the capital structure of companies and its determinants has been ongoing for many years now, no unified theory has been developed in this area to date. Compared to other sectors, this research in the energy and mining section looks rather modest and therefore there is a need to expand and continue it. Therefore, it should be emphasised that the analysis and assessment of the impact of specific factors on the formation of the financing structure of energy and mining companies for a selected group of EU countries carried out in the paper broaden this body of knowledge about this section. In particular, the verification of the research hypotheses provided evidence on the relevance, strength and direction of the impact of microeconomic determinants on corporate financing structures in selected EU countries.

Furthermore, what is missing in the area of financing structures is a relevant comparative analysis. The added value of the research undertaken in the paper includes, inter alia, the comparative analysis in the area of financing structures and the determinants shaping these structures, which was carried out between selected countries and sectors.

The research covered two economically related sections—electricity, gas, steam and air conditioning supply (Subsequently called electricity supply) (classified as heading 04-D under NACE Rev. 2) and mining and quarrying (02-B). To conduct the research, relevant data were retrieved from the Orbis database for 2012–2020.

During stage one, all those companies for which the value of total assets reported in 2012 in each country totaled at least 100,000 euros were isolated for the research. In turn, the second stage involved an analysis of the retrieved data, which resulted in the removal of atypical (outlier) observations. These outlier observations accounted for approximately 5% of the individual variables. In addition, in some countries, a number of companies were eliminated due to a lack of comprehensive data.

The research relied on financial statement data expressed in book values, mainly because of the problems associated with valuing many financial categories, especially the market value of debt, at market values. The use of book values is in many cases preferred by managers deciding on the target capital structure.

4. Model Construction and Research Methodology

After an extensive analysis of relevant literature, the following panel model was used to statistically describe the relationship between companies’ financing structures and selected determinants (variables) [80]. (Panel data models allow the modelling of economic processes on the basis of complex data (e.g., on companies) in cross-sectional and temporal terms. They make it possible to account for diverse objects by introducing individual or temporal effects when the market environment changes. They also provide more profound information, e.g., by comparison with cross-sectional models, they usually result in reduced collinearity between explanatory variables, provide more ‘degrees of freedom’ and allow the introduction of additional variables—artificial or random.)

$$y_{it}={\alpha }_i+x_{it}\times \beta +u_{it};\mathrm{for}i=1,\dots ,\mathrm{N};t=1,\dots ,\mathrm{T}$$

(1)

where:

• t—period (time interval—year),
• i—number of the company,
• x_it—k-element vector of explanatory variables or of their known functions (logarithm-type transformation, compounding, lagging by one or two periods, etc.),
• β—vector of structural parameters,
• α_i—parameter reflecting the unobserved and unaccounted for effect of belonging to the i-th group in the model, the so-called individual effect,
• u_it—random component with a normal distribution and unknown variance.

Referring to the theory of capital structures, the following determinants (independent variables) of financing structure were used in the study: asset structure (X1), relative measure of working capital (X2), company size (X3), profitability (X4), noninterest tax shield (X5) and interest tax shield (X6).

The ratio of total liabilities to total assets (Y = TL/TA) was assumed as the dependent variable. In turn, the individual explanatory variables are defined as follows:

• X1—ratio of fixed assets to total assets (FA/TA),
• X2—ratio of current liabilities to current assets (CL/CA),
• X3—natural logarithm of total assets (lnTA),
• X4—rate of return on total assets (ROA),
• X5—share of depreciation and amortization in total assets (DA/TA),
• X6—ratio of income tax to gross profit/loss (before tax) (TAX/GPL).

Based on the variables defined above, the preliminary model that was assumed for estimation and further verification was is follows:

$${\left(\frac{TL}{TA}\right)}_{it}={{\alpha }_i+\left(\frac{FA}{TA}\right)}_{it}\times {\beta }_1+{\left(\frac{CL}{CA}\right)}_{it}\times {\beta }_2+{lnTA}_{it}\times {\beta }_3+{ROA}_{it}\times {\beta }_4+{\left(\frac{DA}{TA}\right)}_{it}\times {\beta }_5+\left(\frac{TAX}{GPL}\right)\times {\beta }_6+u_{it}.$$

In panel models, due to the presence of so-called individual and temporal effects, two different types of models are considered: models with fixed effects and models with random effects. In the former case, α_i is treated as parameters, i.e., unknown constants subject to estimation. In the latter case, on the other hand, α_i is random variable with a normal distribution, zero expected value and unknown variance. It should be noted that the fixed effects model is a special case of the random effects model. This means that the assumption of random effects is weaker than the assumption that there are certain unknown constants.

In the first case, the model’s structural parameters are usually estimated using the least squares method (LSM) for fixed effects (the within-group estimator). The random effects model (second case) has the structure of a generalised regression model, so in this case, the generalised LSM is the effective estimator of random effects. The first model’s inefficient parameterization and the need for an exceptionally large number of observations over time for the LSM estimator for α_i effects to be consistent are the model’s disadvantages. By contrast, it has the advantage of allowing for a nonzero correlation between x_it and α_i. Moreover, the within-group estimator for β is then consistent and unbiased, unlike the estimator of random effects in the second model. The advantage of the random effects model, on the other hand, is that it can allow incorporation of the influence of exogenous variables, e.g., the level of initial capital, whose values for all objects are constant over time. Additionally, when the number of periods is small, it is suggested that the random effects estimator be used. However, the presence of a strong correlation between x_it and α_i precludes the use of the random effects model [81].

From a formal point of view, in the context of the sample we have, the choice between the two models depends on the results of an appropriate statistical test. The choice of the type of effects, i.e., of the form of the model, can be made, for example, on the basis of the classic Hausman test. The basis for the application of this universal test is the observation (property) that, regardless of the strength of the correlation between x_it and α_i, the estimator of fixed effects is consistent and unconstrained (when at least N → ∞). At the same time, the random effects estimator loses the property of consistency when this correlation is present. If the correlation under analysis is null (the null hypothesis of the Hausman test), the estimator of random effects is a more efficient estimator than the estimator of fixed effects, and the differences between the values of these estimators are statistically insignificant [81].

In order to determine the appropriate form of the model describing the postulated relationships, the deductive concept of model construction was used by applying the general-to-specific modelling procedure, which was used, for example, by Polasik and Marzec for the logit model [82]. First, models were created for fixed and variable effects, and then their parameters were estimated for the full set of explanatory variables. The next step involved a check of the significance of the model’s parameter estimates. If any variable proved to be statistically insignificant in both models, it was removed, and the model parameters were re-estimated with the number of regressors reduced by one. If there were no longer insignificant variables in both models at the same time, the Hausman test was performed to select the correct model for further analysis. The final model was obtained when the effect of each explanatory variable was statistically significant at a significance level of 0.1 (or less).

It is also important to mention some research limitations, mainly arising from the model adopted and the assumptions involved, as well as the availability of data. The main research limitations associated with the use of panel models include limited access to panel data, the problem of homogeneity of the objects under study and random sampling, short time series (long time series pose a problem with stationarity) and the problem of endogeneity of the explanatory variables.

Despite some of their limitations, panel models also have many advantages. First of all; they make it possible to model economic processes on the basis of complex data, e.g., on a company from a cross-sectional temporal perspective; they allow the introduction of individual or time effects; they provide richer information, e.g., relative to cross-sectional models; they usually allow for the reduction of collinearity between explanatory variables; they give more degrees of freedom (they increase the efficiency of estimators) and they allow for the introduction of additional variables—artificial or random.

For the individual empirical panels in each country, statistical verification was performed on:

−

Correlations between independent variables,

−

Collinearity between independent variables,

−

Cross-sectional dependence (Brausch–Pagan/LM test),

−

Autocorrelation of the random component (Breusch–Godfrey/Wooldridge test),

−

The presence of a stochastic trend for the dependent variable (Dickey–Fuller test),

−

Homogeneity of the random component (Brausch–Pagan test),

−

Normality of the distribution of the random component (based on the quantile-quantile plot, the Shapiro–Wilk test and the chi-square test for normality).

5. The Results of Empirical Research in Selected EU Countries

As previously mentioned, the analysis and evaluation of companies’ financing structures, and in particular the study of their dependence (relevance, strength and direction) on selected determinants, was conducted using the example of two sections, i.e., electricity supply and mining and quarrying, from twelve EU countries. In the former section, 2419 companies were analysed based on available data, from which outlier observations were rejected, while in the latter section, 616 companies were used. The numbers of companies covered by the study, with a breakdown by country, are shown in

Table 1. Results of verification of models explaining the TL/TA variable for 2012–2020 panel data in the Energy section of 12 selected EU countries.

Country	N(T × N)	Estimator Type LSDV R2	F Test	Hausman Test	Wald Test
Bulgaria	185(1665)	0.779	F(6.15) = 42.54 ***	χ2 (6) = 32.12 (fixed effects)	χ2 (185) = 112,103
Czech Republic	93(837)	0.830	F(6.74) = 116.39 ***	χ2 (6) = 57.36 (fixed effects)	χ2 (93) = 192,386
France	438(3942)	0.884	F(6.35) = 404.94 ***	χ2 (6) = 353.19 (fixed effects)	χ2 (438) = 1,051,100
Germany	81(729)	0.907	F(6.64) = 29.22 ***	χ2 (6) = 24.77 (fixed effects)	χ2 (81) = 27,211
Hungary	17(153)	0.893	F(6.13) = 16.59 ***	χ2 (6) = 34.63 (fixed effects)	χ2 (17) = 48,474
Italy	660(5940)	0.827	F(6.53) = 216.45 ***	χ2 (6) = 202.42 (fixed effects)	χ2 (660) = 1,054,250
Poland	71(639)	0.884	F(6.56) = 104.72 ***	χ2 (6) = 152.50 (fixed effects)	χ2 (71) = 7862
Romania	51(459)	0.857	F(6.40) = 44.90 ***	χ2 (6) = 19.98 (fixed effects)	χ2 (51) = 68,975
Slovakia	73(657)	0.905	F(6.58) = 32.03 ***	χ2 (6) = 26.69 (fixed effects)	χ2 (73) = 130,734
Spain	664(5976)	0.904	F(6.53) = 413.55 ***	χ2 (6) = 635.80 (fixed effects)	χ2 (664) = 1,528,660
Sweden	40(360)	0.887	F(6.31) = 11.99 ***	χ2 (6) = 51.55 (fixed effects)	χ2 (40) = 16,085
United Kingdom	46(414)	0.881	F(6.36) = 11.86 ***	χ2 (6) = 13.54 (fixed effects)	χ2 (46) = 14,752

Notes: The following significance levels were used: *—significance level of 0.1; **—significance level of 0.05; ***—significance level of 0.01. N—number of enterprises, T—number of periods (9 years). Source: author’s own calculations.

and

Table 2. Results of verification of models explaining the TL/TA variable for 2012–2020 panel data for the mining and quarrying section of 12 selected EU countries.

Country	N(T × N)	Estimator Type LSDV R2	F Test/Wald χ2	Hausman Test	Wald Test
Bulgaria	45(405)	0.901	F(6.35) = 18.92 ***	χ2 (6) = 20.78 (fixed effects)	χ2 (45) = 22,929
Czech Republic	13(117)	0.310	χ2 (6) = 49.22 ***	χ2 (6) = 6.89 (random effects)	x
France	111(999)	0.909	F(6.88) = 108.93 ***	χ2 (6) = 48.38 (fixed effects)	χ2 (111) = 40,672
Hungary	24(216)	0.898	F(6.19) = 60.43 ***	χ2 (6) = 67.72 (fixed effects)	χ2 (24) = 4552
Italy	66(594)	0.913	F(6.52) = 21.59 ***	χ2 (6) = 41.55 (fixed effects)	χ2 (66) = 105,763
Poland	24(216)	0.211	χ2 (6) = 192.34 ***	χ2 (6) = 9.76 (random effects)	x
Romania	82(738)	0.834	F(6.65) = 49.82 ***	χ2 (6) = 28.55 (fixed effects)	χ2 (82) = 56,553
Slovakia	10(90)	0.613	χ2 (6) = 139.54 ***	χ2 (6) = 9.96 (random effects)	x
Spain	187(1683)	0.899	F(6.149) = 70.65 ***	χ2 (6) = 131.82 (fixed effects)	χ2 (187) = 1,706,550
Sweden	31(279)	0.504	χ2 (6) = 126.35 ***	χ2 (6) = 4.73 (random effects)	x
United Kingdom	23(207)	0.575	χ2 (6) = 221.83 ***	χ2 (6) = 8.52 (random effects)	x

Notes: The following significance levels were used: *—significance level of 0.1; **—significance level of 0.05; ***—significance level of 0.01. N—number of enterprises, T—number of periods (9 years). Germany was omitted from the study due to insufficient data. Source: author’s own calculations.

. The study employed linear panel models and covered the 2012–2020 period.

The accomplishment of the main research objective was preceded by a calculation of the main statistical characteristics of the variables used in the models, i.e., the mean, median, lower and upper quartiles, standard deviation, coefficient of variation and skewness. The values of these characteristics are presented in Appendix A of the paper in

Table A1. Descriptive statistics for the TL/TA dependent variable in the energy section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	1665	0.4830	0.527	0.200	0.732	0.301	62.26	−0.154
Czech Republic	837	0.5188	0.535	0.310	0.732	0.262	50.59	−0.159
France	3942	0.6549	0.742	0.483	0.875	0.271	41.39	−0.871
Germany	729	0.4674	0.491	0.352	0.577	0.158	33.87	−0.244
Hungary	153	0.4190	0.371	0.235	0.554	0.253	60.28	0.581
Italy	5940	0.5838	0.649	0.370	0.817	0.280	47.92	−0.509
Poland	639	0.3643	0.346	0.244	0.483	0.160	43.82	0.351
Romania	459	0.5284	0.496	0.308	0.769	0.262	49.52	0.023
Slovakia	657	0.5991	0.631	0.421	0.807	0.245	40.88	−0.509
Spain	5976	0.4901	0.499	0.193	0.776	0.309	63.09	0.005
Sweden	360	0.5142	0.497	0.329	0.714	0.228	44.41	0.029
United Kingdom	414	0.5318	0.606	0.298	0.728	0.260	48.95	−0.387

Source: author’s own calculations.

,

Table A2. Descriptive statistics for the FA/TA explanatory variable in the energy section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	1665	0.735	0.827	0.628	0.911	0.246	33.52	−1.381
Czech Republic	837	0.667	0.784	0.508	0.868	0.281	42.06	−1.097
France	3942	0.737	0.795	0.674	0.865	0.197	26.75	−1.714
Germany	729	0.702	0.752	0.632	0.825	0.181	25.74	−1.678
Hungary	153	0.506	0.551	0.317	0.743	0.267	52.76	−0.517
Italy	5940	0.583	0.711	0.310	0.837	0.312	53.49	−0.724
Poland	639	0.711	0.743	0.640	0.797	0.133	18.69	−1.286
Romania	459	0.568	0.609	0.398	0.774	0.268	47.27	−0.573
Slovakia	657	0.625	0.741	0.462	0.831	0.274	43.93	−0.916
Spain	5976	0.713	0.802	0.586	0.902	0.243	34.11	−1.133
Sweden	360	0.634	0.768	0.476	0.884	0.323	50.86	−1.003
United Kingdom	414	0.646	0.777	0.471	0.888	0.306	47.29	−0.894

Source: author’s own calculations.

,

Table A3. Descriptive statistics for the CL/CA explanatory variable in the energy section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	1665	0.842	0.562	0.173	1.122	0.924	109.70	1.861
Czech Republic	837	0.645	0.596	0.247	0.894	0.540	83.71	2.152
France	3942	0.674	0.411	0.123	0.916	0.780	115.69	2.045
Germany	729	0.795	0.727	0.384	1.046	0.524	65.96	1.201
Hungary	153	0.809	0.744	0.397	1.005	0.576	71.16	1.672
Italy	5940	0.793	0.568	0.240	0.994	0.806	101.71	1.968
Poland	639	0.857	0.682	0.426	1.095	0.632	73.82	1.874
Romania	459	0.915	0.740	0.451	1.179	0.725	79.25	1.987
Slovakia	657	1.057	0.813	0.475	1.354	0.869	82.27	1.554
Spain	5976	0.745	0.492	0.192	0.970	0.811	108.73	2.075
Sweden	360	0.861	0.674	0.405	1.088	0.673	78.08	1.936
United Kingdom	414	0.828	0.575	0.320	0.976	0.796	96.17	1.952

Source: author’s own calculations.

,

Table A4. Descriptive statistics for the lnTA explanatory variable in the energy section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	1665	7.625	7.579	6.294	8.464	1.779	23.33	0.852
Czech Republic	1665	7.625	7.579	6.294	8.464	1.779	23.33	0.852
France	3942	7.411	6.957	6.033	8.666	1.850	24.96	1.220
Germany	729	11.316	11.077	10.168	12.083	1.593	14.08	0.693
Hungary	153	9.080	9.083	7.552	10.208	1.865	20.54	0.884
Italy	5940	8.166	7.967	7.007	8.997	1.792	21.95	1.093
Poland	5940	8.166	7.967	7.007	8.997	1.792	21.95	1.093
Romania	459	8.669	8.490	7.281	9.668	2.083	24.03	0.904
Slovakia	657	8.619	8.532	7.770	9.300	1.305	15.14	0.916
Spain	5976	7.467	6.787	6.154	8.385	1.930	25.85	1.746
Sweden	360	9.823	9.810	9.119	11.024	1.548	15.76	−0.799
United Kingdom	414	12.208	12.163	10.605	14.550	2.364	19.37	−0.067

Source: author’s own calculations.

,

Table A5. Descriptive statistics for the ROA (in %) explanatory variable in the energy section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	1665	5.93	4.51	1.25	8.59	7.85	132.23	1.15
Czech Republic	837	6.59	6.06	2.82	9.49	5.85	88.74	−0.29
France	3942	5.58	4.58	1.81	8.36	6.53	116.97	1.31
Germany	729	6.77	5.69	3.18	9.03	4.82	71.16	1.68
Hungary	153	7.71	4.85	1.89	9.60	8.67	112.52	2.35
Italy	153	7.71	4.85	1.89	9.60	8.67	112.52	2.35
Poland	639	4.75	3.93	1.99	6.40	5.69	119.87	0.43
Romania	639	4.75	3.93	1.99	6.40	5.69	119.87	0.43
Slovakia	657	6.75	5.77	2.70	8.57	5.71	84.69	1.81
Spain	5976	4.42	2.81	0.63	6.91	6.47	146.32	1.14
Sweden	360	6.39	5.47	3.22	8.91	4.48	69.98	0.83
United Kingdom	414	9.39	7.72	4.78	12.50	6.37	67.80	1.42

Source: author’s own calculations.

,

Table A6. Descriptive statistics for the DA/TA explanatory variable in the energy section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	1665	0.054	0.047	0.031	0.066	0.038	70.50	1.866
Czech Republic	837	0.052	0.051	0.037	0.066	0.028	54.39	0.930
France	3942	0.060	0.057	0.045	0.073	0.030	50.20	1.417
Germany	729	0.048	0.049	0.035	0.059	0.019	40.22	0.802
Hungary	153	0.040	0.040	0.020	0.059	0.026	63.23	0.328
Italy	5940	0.046	0.043	0.021	0.061	0.034	74.58	1.275
Poland	639	0.065	0.063	0.050	0.077	0.028	43.21	1.354
Romania	459	0.046	0.042	0.021	0.062	0.033	71.68	1.315
Slovakia	657	0.069	0.068	0.045	0.089	0.041	59.57	0.863
Spain	5976	0.065	0.056	0.035	0.087	0.046	69.90	1.351
Sweden	360	0.046	0.049	0.038	0.061	0.024	52.51	−0.341
United Kingdom	414	0.041	0.032	0.021	0.058	0.027	65.52	0.867

Source: author’s own calculations.

,

Table A7. Descriptive statistics for the TAX/GPL explanatory variable in the energy section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	1665	0.082	0.099	0.074	0.102	0.049	59.83	0.036
Czech Republic	837	0.189	0.190	0.183	0.200	0.063	33.02	0.641
France	837	0.189	0.190	0.183	0.200	0.063	33.02	0.641
Germany	729	0.246	0.273	0.156	0.319	0.105	42.71	−0.337
Hungary	153	0.156	0.114	0.065	0.221	0.123	78.82	1.059
Italy	5940	0.269	0.291	0.243	0.331	0.110	40.77	−0.910
Poland	639	0.201	0.197	0.187	0.214	0.052	25.67	0.699
Romania	459	0.120	0.140	0.021	0.169	0.089	74.37	0.405
Slovakia	657	0.201	0.217	0.184	0.234	0.089	44.42	−0.537
Spain	5976	0.242	0.250	0.238	0.250	0.048	19.82	−0.946
Sweden	360	0.137	0.157	0.031	0.219	0.101	73.35	0.146
United Kingdom	414	0.189	0.194	0.169	0.214	0.059	31.16	−0.225

Source: author’s own calculations.

,

Table A8. Descriptive statistics for the TL/TA dependent variable in the mining and quarrying section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	405	0.3381	0.2909	0.1354	0.4761	0.2437	72.0823	0.7853
Czech Republic	117	0.2129	0.1628	0.0720	0.3338	0.1713	80.4607	0.9030
France	999	0.5147	0.5092	0.3457	0.6982	0.2296	44.6151	−0.0212
Germany	54	0.5627	0.5278	0.4090	0.6178	0.2193	38.9698	0.9354
Hungary	216	0.2717	0.1898	0.1135	0.3736	0.2174	80.0287	1.3311
Italy	594	0.4329	0.4203	0.1819	0.6627	0.2696	62.2854	0.1640
Poland	216	0.2494	0.2286	0.1517	0.3239	0.1372	55.0094	1.0752
Romania	738	0.4966	0.5009	0.2743	0.7141	0.2584	52.0401	0.0241
Slovakia	90	0.2579	0.2287	0.1164	0.3641	0.1587	61.5304	0.6605
Spain	1683	0.3245	0.2714	0.1375	0.4704	0.2367	72.9580	0.7250
Sweden	279	0.3660	0.3502	0.2255	0.5151	0.2088	57.0392	0.4012
United Kingdom	207	0.3178	0.2969	0.1813	0.4237	0.1957	61.6003	0.6588

Source: author’s own calculations.

,

Table A9. Descriptive statistics for the FA/TA explanatory variable in the mining and quarrying section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	405	0.5176	0.4990	0.3176	0.7333	0.2539	49.0515	−0.0149
Czech Republic	117	0.5544	0.5354	0.4339	0.6780	0.1543	27.8258	0.3701
France	999	0.3742	0.3602	0.2277	0.5028	0.1979	52.8982	0.3817
Germany	54	0.4379	0.3931	0.3126	0.6158	0.2191	50.0399	−0.3500
Hungary	216	0.5084	0.4937	0.3866	0.6435	0.1821	35.8235	−0.1385
Italy	594	0.4104	0.4092	0.2123	0.5919	0.2347	57.1845	0.0360
Poland	216	0.5661	0.5766	0.4418	0.6940	0.2031	35.8806	−0.5434
Romania	738	0.5169	0.5208	0.4007	0.6652	0.2045	39.5637	−0.3325
Slovakia	90	0.5416	0.6031	0.3237	0.7325	0.2073	38.2876	−0.3339
Spain	1683	0.4833	0.4813	0.3180	0.6494	0.2286	47.3025	0.0558
Sweden	279	0.5318	0.5529	0.3416	0.7166	0.2142	40.2737	−0.2916
United Kingdom	207	0.4596	0.4766	0.2908	0.6579	0.2536	55.1751	−0.1201

Source: author’s own calculations.

,

Table A10. Descriptive statistics for the CL/CA explanatory variable in the mining and quarrying section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	405	0.582	0.395	0.154	0.816	0.597	102.649	2.068
Czech Republic	117	0.368	0.304	0.151	0.477	0.337	91.493	3.141
France	999	0.599	0.451	0.298	0.763	0.509	84.954	2.688
Germany	54	0.465	0.326	0.250	0.449	0.421	90.577	1.765
Hungary	216	0.495	0.314	0.177	0.626	0.539	109.040	3.123
Italy	594	0.508	0.410	0.213	0.695	0.432	85.105	2.268
Poland	216	0.494	0.367	0.212	0.707	0.379	76.742	1.595
Romania	738	0.669	0.583	0.317	0.898	0.471	70.299	1.474
Slovakia	90	0.516	0.345	0.161	0.655	0.528	102.387	1.798
Spain	1683	0.523	0.361	0.168	0.668	0.558	106.694	2.600
Sweden	279	0.575	0.452	0.282	0.691	0.499	86.863	2.412
United Kingdom	207	0.468	0.410	0.253	0.617	0.332	70.930	1.662

Source: author’s own calculations.

,

Table A11. Descriptive statistics for the lnTA explanatory variable in the mining and quarrying section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	405	7.983	7.654	6.428	9.042	2.050	25.678	0.668
Czech Republic	117	9.960	9.578	8.859	10.992	1.672	16.786	0.765
France	999	8.745	8.636	7.888	9.593	1.308	14.961	0.196
Germany	54	11.129	10.244	9.277	12.702	2.252	20.239	0.900
Hungary	216	7.681	7.557	6.922	8.254	1.101	14.338	0.427
Italy	594	8.082	7.988	7.172	8.880	1.444	17.867	1.035
Poland	216	9.466	9.026	8.339	10.394	1.714	18.105	0.987
Romania	738	7.633	7.336	6.508	8.269	1.761	23.076	2.182
Slovakia	90	8.566	8.974	6.592	9.676	1.879	21.935	−0.236
Spain	1683	7.800	7.649	6.908	8.581	1.428	18.311	0.613
Sweden	279	8.634	8.732	7.343	9.490	1.829	21.186	1.724
United Kingdom	207	10.238	9.910	9.303	10.994	1.449	14.157	0.603

Source: author’s own calculations.

,

Table A12. Descriptive statistics for the ROA (in %) explanatory variable in the mining and quarrying section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	405	5.656	3.471	0.730	10.644	10.774	190.483	0.456
Czech Republic	117	9.402	8.090	5.407	12.870	6.379	67.846	0.775
France	999	6.822	6.013	2.585	10.963	7.948	116.506	−0.219
Germany	54	9.912	7.933	4.792	15.885	7.002	70.646	0.599
Hungary	216	11.924	10.501	4.445	16.709	8.875	74.434	1.197
Italy	594	5.226	4.063	0.935	9.414	8.801	168.404	0.293
Poland	216	10.748	9.168	5.218	13.637	9.160	85.219	1.828
Romania	738	8.929	6.103	1.366	15.104	10.850	121.518	0.730
Slovakia	90	6.778	5.493	2.137	9.777	6.811	100.489	1.229
Spain	1683	2.317	1.360	−0.491	5.026	7.532	325.130	0.466
Sweden	279	10.620	9.118	5.149	13.402	7.606	71.624	1.858
United Kingdom	207	7.399	7.472	4.032	12.252	11.406	154.156	−0.637

Source: author’s own calculations.

,

Table A13. Descriptive statistics for the DA/TA explanatory variable in the mining and quarrying section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	405	0.062	0.050	0.027	0.090	0.048	77.856	1.392
Czech Republic	117	0.052	0.046	0.031	0.060	0.037	70.103	2.201
France	999	0.056	0.049	0.029	0.076	0.037	66.375	1.154
Germany	54	0.061	0.060	0.033	0.085	0.037	60.003	0.618
Hungary	216	0.065	0.058	0.040	0.084	0.033	51.019	0.622
Italy	594	0.039	0.032	0.018	0.055	0.029	76.463	1.582
Poland	216	0.051	0.045	0.035	0.062	0.026	51.708	1.249
Romania	738	0.081	0.073	0.045	0.110	0.051	63.548	0.816
Slovakia	90	0.060	0.041	0.024	0.095	0.041	68.189	0.635
Spain	1683	0.035	0.030	0.014	0.046	0.030	85.055	2.393
Sweden	279	0.084	0.079	0.049	0.113	0.044	52.823	0.828
United Kingdom	207	0.055	0.050	0.024	0.073	0.045	80.284	1.439

Source: author’s own calculations.

and

Table A14. Descriptive statistics for the TAX/GPL explanatory variable in the mining and quarrying section.

Country	T × N	Mean	Median	Bottom Quantile	Top Quantile	Standard Deviation	Coefficient of Variation	Skewness
Bulgaria	405	0.074	0.099	0.000	0.104	0.062	83.513	0.900
Czech Republic	117	0.196	0.191	0.177	0.210	0.049	24.999	1.493
France	999	0.221	0.272	0.096	0.322	0.137	62.181	−0.589
Germany	54	0.249	0.267	0.198	0.299	0.067	26.901	−0.506
Hungary	216	0.081	0.077	0.050	0.098	0.045	55.422	1.359
Italy	594	0.242	0.277	0.164	0.334	0.132	54.618	−0.591
Poland	216	0.200	0.197	0.188	0.209	0.046	23.032	2.136
Romania	738	0.135	0.148	0.085	0.169	0.086	63.973	0.657
Slovakia	90	0.226	0.221	0.212	0.242	0.058	25.854	0.684
Spain	1683	0.239	0.250	0.235	0.251	0.056	23.503	−1.029
Sweden	279	0.180	0.192	0.124	0.230	0.091	50.437	−0.023
United Kingdom	207	0.184	0.201	0.147	0.236	0.102	55.151	−0.298

Source: author’s own calculations.

. Let us note the most important characteristics of each variable.

The overall debt ratio (TL/TA) is used as a measure of the structure of corporate financing (explanatory variable). Analysis of this ratio furnishes valuable information about the level of debt in the analysed sections and countries.

Within the group of analysed countries, in the electricity supply sector, the highest debt level in 2012–2020 was revealed in France, Slovakia and Italy, where total debt to total assets as measured by the mean stood at 65.5%, 59.9% and 58.4%, respectively, while the median in these countries amounted to 74.2%, 63.1% and 64.9%, respectively (see Table A1). In contrast, countries such as Poland and Hungary reveal the lowest debt, with a mean value of 36.4% and 41.9% of the total debt ratio and medians of 34.6% and 37.1%, respectively. Let it be added that the average value of this indicator for the surveyed countries in this section in the study period of 2012–2020 measured by the mean was 51.3%, while the same measured by the median stood at 53.3%.

The second researched section in the same countries and years was the mining and quarrying section. In this section, the highest levels of corporate debt are revealed in Germany, France and Romania. The mean values of total debt were 56.3%, 51.5% and 49.7%, and the median values were 52.8%, 50.9% and 50.1%, respectively (see Table A8). Countries with the lowest debt levels in this section include the Czech Republic, Poland, Slovakia and Hungary, for which the mean debt ratio was 21.3%, 24.9%, 25.8% and 27.2%, respectively, and the median stood at 16.3%, 22.9%, 19.0% and 22.9%. The entire mining and quarrying section covering the countries and years analysed has a mean value of 36.2% and a median value of 33.1%, which means a relatively low level of debt among the mining enterprises surveyed. The relatively low level of total debt for mining companies is also confirmed by the results of the research carried out by Sierpińska [79].

Comparing the two analysed sections in terms of overall debt, two major conclusions can be drawn. First, debt in the electricity supply section is at a relatively higher level compared to the mining and quarrying section. Given the survey results obtained, these differences should be considered significant. It should also be added that the higher indebtedness of companies in the electricity supply section compared to the mining and quarrying section was ascertained in all the countries analysed except Germany.

Secondly, relating the results to other sections, it should be concluded that debt in the analysed companies, especially in the mining and quarrying section, is at a relatively low level. In comparison, studies conducted for the same countries (excluding Sweden) showed that between 2009 and 2017, the average debt level (measured in terms of the mean) in the construction section was 62.1%, with a median of 61.9%, in trade 63.3% (63.2%), in information and communication 53.5% (49.2%) and in manufacturing 57.7% (56.0%) [83].

The study of capital-asset adequacy is also an important element in the evaluation of corporate financing, as reflected in capital structure theories. This is because asset structure is an important factor influencing the structure of a company’s liabilities, with capital structure theories explaining this influence in different ways.

Across the countries in the two sections analysed, the share of long-term assets in relation to total assets is at a fairly even level, with the share in the electricity supply section being at a much higher level. The average share of long-term assets in relation to total assets in the electricity supply section, as measured by the mean in the countries and years analysed, was 65.2%, with the median standing at 73.8%. By contrast, in the mining and quarrying section, both the mean and the median for this ratio are 49.2% each. It should also be noted that the higher share of long-term assets in total assets in the electricity supply section compared to the mining and quarrying section is also accompanied by a higher share of total liabilities in total assets (see Table A2 and Table A9).

Another important measure of proper corporate financing is the ratio of equity to fixed assets (E/FA). From an economic point of view, long-term assets should be fully financed out of equity. This principle is not easy to meet, which means that in most enterprises it is not respected.

In the electricity supply section, only Hungarian companies comply with the rule. By contrast, in the remaining countries, the ratio of equity to long-term assets is below 1. The average value of the E/FA ratio for the surveyed countries in this section in 2012–2022, measured by the mean, was 75.7%, with the median at 64.9%. In the mining and quarrying section, on the other hand, equity, with the exception of Romania, fully finances long-term assets. The average value of the E/FA ratio in this section, measured by the mean, was 130%, with the median at 136%.

In assessing the proper financing of companies, attention should also be paid to the degree to which current assets are financed by current liabilities (CL/CA). This financing in the mining and quarrying section is at a lower level compared to the electricity supply section. The average level of the CL/CA ratio for the mining and quarrying section, as measured by the mean, stands at 52.2%, with the median at 39.3%. For the electricity supply section, the values are 81.8% and 63.2%, respectively (cf. Table A3 and Table A10).

The largest surveyed entities in terms of the value of their total assets are companies in Germany and the United Kingdom, for which, after logarithmisation, these values are, respectively: for the electricity supply section, the mean value is 11.32, the median 11.08, respectively, and 12.21 (mean) and 12.16 (median), while for the mining and quarrying section, the mean value is 11.129, with the median at 10.244, the mean at 10.23 and the median at 9.91, respectively (see Table A4 and Table A11).

Another important variable derived from the theory of capital and used in the study is company profitability as measured by return on assets (ROA). Let it be noted that the mining and quarrying section has a slightly higher profitability compared to the electricity supply section. The average level of profitability measured in it is 7.97%, and the median is 6.56%. For the electricity supply section, these figures are 6.40% and 5.01%, respectively (see Table A5 and Table A12).

The ratio of depreciation and amortisation to the value of total assets (DA/TA) and the value of income tax related to the value of gross profit/loss (TAX/GPL) were taken as measures of the interest and noninterest tax shield. The average values in terms of both mean and median for both analysed sections are similar, with the DA/TA variable having values of approximately 5%, while the TAX/GPL variable stands at approximately 19% (see Table A6, Table A7, Table A13 and Table A14).

During the next stage of the analysis, the research models were verified using appropriate tests, and then the proposed hypotheses were verified.

The application of the Hausman test revealed that the fixed effects model was more appropriate for the electricity supply section in each country. However, in the mining and quarrying section, this model was only appropriate for certain countries (Bulgaria, France, Spain, Romania, Hungary and Italy), while in countries such as the Czech Republic, Poland, Slovakia, Sweden and the United Kingdom, the test indicated that the random effects model was more appropriate. It should be added that in the case of the mining and quarrying section, Germany was omitted from the study due to insufficient data for panel analysis (see Table 1 and Table 2).

The fit of the models to the data, as measured by the LSDV R² coefficient of determination in the electricity supply section, was at a fairly high level, ranging from 0.779 to 0.907, while the mining and quarrying section had more varied and lower values. The lowest fit was ascertained for Poland (0.211) and the Czech Republic (0.310), while for the remaining countries, it ranged between 0.504 and 0.913.

The assumption made in the regression model is that there is no correlation or only a weak correlation between the independent variables. The study found a weak correlation of this type. Evaluation of the collinearity of the independent variables showed only slight collinearity among the predictors, as evidenced by the low values of the variance inflation factors (VIF) test coefficients. In the electricity supply section, the values range from 1.02 to 3.67, and in the mining and quarrying section, from 1.01 to 2.23 (see

Table 3. VIF test results for 2012–2020 panel data in the Energy section of 12 selected EU countries.

Country	FA/TA	CL/CA	lnTA	ROA	DA/TA	TAX/GPL
Bulgaria	1.186	1.106	1.099	1.142	1.104	1.099
Czech Republic	1.704	1.147	1.167	1.069	1.782	1.046
France	1.206	1.083	1.168	1.204	1.124	1.186
Germany	1.759	1.476	1.142	1.109	1.293	1.025
Hungary	2.851	1.300	1.089	1.215	1.709	2.471
Italy	1.641	1.168	1.058	1.135	1.390	1.069
Poland	1.722	1.462	1.339	1.094	1.162	1.040
Romania	1.334	1.166	1.089	1.102	1.138	1.068
Slovakia	1.971	1.264	1.202	1.129	1.703	1.211
Spain	1.310	1.174	1.123	1.139	1.164	1.020
Sweden	3.666	1.211	1.097	1.153	2.567	1.614
United Kingdom	1.495	1.218	1.546	1.273	1.548	1.087

Source: author’s own calculations.

and

Table 4. VIF test results for 2012–2020 panel data for the mining and quarrying section of 12 selected EU countries.

Country	FA/TA	CL/CA	lnTA	ROA	DA/TA	TAX/GPL
Bulgaria	1.812	1.512	1.409	1.176	1.103	1.111
Czech Republic	1.470	1.164	1.249	1.211	1.153	1.210
France	1.599	1.471	1.317	1.212	1.121	1.474
Hungary	1.609	1.391	1.106	1.140	1.375	1.077
Italy	1.256	1.139	1.190	1.201	1.033	1.243
Poland	2.228	1.685	1.476	1.440	1.358	1.100
Romania	1.152	1.109	1.070	1.062	1.073	1.014
Slovakia	2.021	1.954	1.447	1.270	1.556	1.085
Spain	1.360	1.280	1.219	1.145	1.034	1.048
Sweden	1.999	1.595	1.228	1.601	1.383	1.429
United Kingdom	1.641	1.366	1.219	1.421	1.343	1.402

Source: author’s own calculations.

). Due to the presence of cross-sectional dependence, autocorrelation and heterogeneity in the variance of the random component, the final model was estimated using robust covariance matrix estimation.

The final step of verification involved a check of the normality of the random component distribution. Although statistical tests showed that the random components of the models do not have a normal distribution, they somewhat resemble its shape. They are characterised by quite high leptokurtosis in many cases.

After the panel models were obtained and the corresponding tests were conducted, the research hypotheses were verified. The first of these (H1) referred to the significance of the relationship between the financing structure of the studied companies and the structure of assets, the ratio of current liabilities to current assets, the size of the company, profitability and interest and noninterest tax shields.

The tests indicate a significant relationship between the financing structure and the adopted explanatory variables in most of the analysed cases, except that a much higher significance was revealed by the variables for enterprises belonging to the electricity supply section. In this section, out of 72 cases, 12 cases showed a lack of significant relationship, with the highest number (6) for the interest tax shield (Czech Republic, Germany, Poland, Romania, Hungary and the UK). The other cases of nonsignificance involved the FA/TA variable (Slovakia, Sweden and Italy), the CL/CA variable (Sweden), the lnTA variable (Bulgaria) and the DA/TA variable (Poland) (see

Table 5. Results of estimation of models explaining the TL/TA variable for 2012–2020 panel data in the Energy section of 12 selected EU countries.

Country	Explanatory Variables	FA/TA	CL/CA	lnTA	ROA	DA/TA	TAX/GPL
Bulgaria	Estimation	0.2497	0.0434	0.0166	−0.0055	−0.8373	−0.2513
	Error	0.0470	0.0057	0.0139	0.0007	0.1694	0.1050
	t-test	5.3190	7.5970	1.1940	−7.9850	−4.9430	−2.3930
	p-value	0.0000 ***	0.0000 ***	0.2326	0.0000 ***	0.0000 ***	0.0168 **
Czech Republic	Estimation	0.1898	0.0709	0.1458	−0.0144	−2.8841	−0.0140
	Error	0.0632	0.0116	0.0224	0.0010	0.3438	0.0808
	t-test	3.0040	6.1210	6.5160	−14.91	−8.390	−0.1735
	p-value	0.0028 ***	0.0000 ***	0.0000 ***	0.0000 ***	0.0000 ***	0.8623
France	Estimation	0.2307	0.0261	0.1697	−0.0066	−0.9673	−0.2590
	Error	0.0193	0.0031	0.0089	0.0004	0.1034	0.0256
	t-test	11.9600	8.3180	18.9700	−15.2100	−9.3510	−10.1400
	p-value	0.0000 ***	0.0000 ***	0.0000 ***	0.0000 ***	0.0000 ***	0.0000 ***
Germany	Estimation	−0.3376	0.0544	0.0835	−0.0054	0.4311	0.0170
	Error	0.0377	0.0082	0.0173	0.0009	0.2160	0.0359
	t-test	−8.9640	6.6400	4.8350	−5.9910	1.9960	0.4724
	p-value	0.0000 ***	0.0000 ***	0.0000 ***	0.0000 ***	0.0464 **	0.6368
Hungary	Estimation	−0.3362	0.1661	0.0701	−0.0042	−2.0559	−0.0021
	Error	0.1088	0.0197	0.0402	0.0015	0.8762	0.1246
	t-test	−3.0900	8.4500	1.7420	−2.8630	−2.3460	−0.0171
	p-value	0.0024 **	0.0000 ***	0.0839*	0.0049 ***	0.0205 **	0.9864
Italy	Estimation	0.0211	0.0619	0.0787	−0.0051	−1.1756	0.2315
	Error	0.0201	0.0033	0.0071	0.0004	0.0907	0.0220
	t-test	1.0520	18.8900	11.1000	−13.6200	−12.9700	10.5100
	p-value	0.2927	0.0000 ***	0.0000 ***	0.0000 ***	0.0000 ***	0.0000 ***
Poland	Estimation	−0.1041	0.1025	0.2011	−0.0063	−0.1774	−0.0504
	Error	0.0447	0.0084	0.0158	0.0006	0.1331	0.0522
	t-test	−2.3300	12.2500	12.7200	−10.5100	−1.3330	−0.9669
	p-value	0.0202 **	0.0000 ***	0.0000 ***	0.0000 ***	0.1829	0.3340
Romania	Estimation	−0.1190	0.0987	0.0581	−0.0069	−0.6271	−0.0505
	Error	0.0568	0.0101	0.0134	0.0007	0.2238	0.0690
	t-test	−2.0960	9.7750	4.3300	−9.5750	−2.802	−0.7316
	p-value	0.0367 **	0.0000 ***	0.0000 ***	0.0000 ***	0.0053 ***	0.4648
Slovakia	Estimation	0.0266	0.0395	0.0913	−0.0050	−0.4757	−0.1467
	Error	0.0430	0.0064	0.0186	0.0009	0.1797	0.0456
	t-test	0.6195	6.1640	4.8950	−5.4530	−2.6480	−3.2200
	p-value	0.5358	0.0000 ***	0.0000 ***	0.0000 ***	0.0083 ***	0.0014 ***
Spain	Estimation	0.1281	0.0444	0.1712	−0.0045	−0.4937	0.1184
	Error	0.0141	0.0027	0.0065	0.0003	0.0608	0.0365
	t-test	9.0760	16.1900	26.4000	−14.3700	−8.1160	3.2440
	p-value	0.0000 ***	0.0000 ***	0.0000 ***	0.0000 ***	0.0000 ***	0.0012 ***
Sweden	Estimation	0.0663	0.0026	0.1847	−0.0056	1.2739	−0.1790
	Error	0.0808	0.0105	0.0316	0.0021	0.4857	0.0842
	t-test	0.8202	0.2500	5.8510	−2.7220	2.6230	−2.1250
	p-value	0.4127	0.8028	0.0000 ***	0.0068 ***	0.0091 ***	0.0343 **
United Kingdom	Estimation	0.1520	0.0442	0.0597	−0.0036	1.0163	0.0870
	Error	0.0576	0.0091	0.0244	0.0013	0.4866	0.0960
	t-test	2.6380	4.8760	2.4450	−2.7090	2.0890	0.9063
	p-value	0.0087 ***	0.0000 ***	0.0150 **	0.0071 ***	0.0374 **	0.3654

Notes: The following significance levels were used: *—significance level of 0.1; **—significance level of 0.05; ***—significance level of 0.01. Source: author’s own calculations.

).

The second section, mining and quarrying, is characterised by a higher number of nonsignificant variables. Out of the 66 cases analysed, 36 cases showed significant dependence of the explanatory variables on the dependent variable, while the rest revealed a lack of significance. The largest number of cases of nonsignificance in this section is found for the lnTA variables (seven countries), ROA (seven countries), DA/TA (six countries) and TAX/GPL (six countries) (see

Table 6. Results of estimation of models explaining the TL/TA variable for 2012–2020 panel data in the mining and quarrying section of 12 selected EU countries.

Country	Explanatory Variables	FA/TA	CL/CA	lnTA	ROA	DA/TA	TAX/GPL
Bulgaria	Estimation	0.0713	0.0932	0.0292	−0.0007	0.3175	−0.1954
	Error	0.0420	0.0115	0.0179	0.0005	0.1594	0.0987
	t-test	1.6950	8.0930	1.6340	−1.4250	1.9920	−1.9800
	p-value	0.0909 *	0.0000 ***	0.1033	0.1551	0.0471 **	0.0485 **
Czech Republic	Estimation	0.0607	0.1627	−0.0130	0.0017	−0.8160	0.3350
	Error	0.0898	0.0316	0.0199	0.0014	0.2620	0.1642
	Z test	0.6766	5.1530	−0.6835	1.2160	−3.1160	2.0400
	p-value	0.4987	0.0000 ***	0.4943	0.2238	0.0018 ***	0.0413 **
France	Estimation	0.1010	0.1335	0.0636	−0.0030	−0.1362	0.0413
	Error	0.0299	0.0089	0.0112	0.0005	0.1196	0.0302
	t-test	3.3820	15.0200	5.6750	−7.8580	−1.1390	1.3650
	p-value	0.0007 ***	0.0000 ***	0.0000 ***	0.0000 ***	0.2551	0.1727
Hungary	Estimation	−0.3430	0.2605	−0.0660	−0.0003	0.0705	−0.3531
	Error	0.0549	0.0167	0.0187	0.0008	0.2908	0.1430
	t-test	−6.2430	15.5600	−3.5250	−0.4141	0.2424	−2.4690
	p-value	0.0000 ***	0.0000 ***	0.0005 ***	0.6793	0.8087	0.0144 **
Italy	Estimation	−0.1354	0.1225	0.0003	−0.0023	0.5189	0.1574
	Error	0.0388	0.0151	0.0132	0.0007	0.2023	0.0447
	t-test	−3.4910	8.1330	0.0203	−3.4870	2.5650	3.5240
	p-value	0.0005 ***	0.0000 ***	0.9838	0.0005 ***	0.0106 **	0.0005 ***
Poland	Estimation	−0.0544	0.1871	−0.0067	−0.0024	−0.3973	0.0950
	Error	0.0474	0.0186	0.0116	0.0008	0.2412	0.0859
	Z test	−1.1480	10.0500	−0.5785	−3.2150	−1.6470	1.1060
	p-value	0.2508	0.0000 ***	0.5629	0.0013 ***	0.0996 *	0.2687
Romania	Estimation	−0.0123	0.1592	−0.0217	−0.0044	−0.0638	0.0374
	Error	0.0411	0.0140	0.0121	0.0006	0.1239	0.0591
	t-test	−0.2999	11.3500	−1.7890	−7.8160	−0.5147	0.6328
	p-value	0.7644	0.0000 ***	0.0740 *	0.0000 ***	0.6069	0.5271
Slovakia	Estimation	0.0459	0.2232	0.0091	0.0001	−0.4306	−0.0333
	Error	0.0612	0.0208	0.0144	0.0012	0.5123	0.1338
	Z test	0.7499	10.7500	0.6293	0.0656	−0.8407	−0.2485
	p-value	0.4533	0.0000 ***	0.5292	0.9477	0.4005	0.8037
Spain	Estimation	−0.1680	0.1408	0.0505	−0.0002	−0.0410	0.0067
	Error	0.0237	0.0073	0.0094	0.0004	0.1032	0.0436
	t-test	−7.1060	19.2600	5.3670	−0.6520	−0.4037	0.1531
	p-value	0.0000 ***	0.0000 ***	0.0000 ***	0.5144	0.6865	0.8784
Sweden	Estimation	−0.2460	0.2511	−0.0104	0.0000	0.3917	−0.2180
	Error	0.0565	0.0261	0.0116	0.0010	0.1845	0.0797
	Z test	−4.3500	9.6240	−0.8923	0.0505	2.1230	−2.7350
	p-value	0.0000 ***	0.0000 ***	0.3722	0.9597	0.0338 **	0.0062 ***
United Kingdom	Estimation	−0.2988	0.3818	0.0109	−0.0009	−0.0123	−0.0097
	Error	0.0573	0.0270	0.0117	0.0008	0.2003	0.0800
	Z test	−5.2160	14.1300	0.9340	−1.1180	−0.0614	−0.1212
	p-value	0.0000 ***	0.0000 ***	0.3503	0.2637	0.9510	0.9036

Notes: The following significance levels were used: *—significance level of 0.1; **—significance level of 0.05; ***—significance level of 0.01. Germany was omitted from the study due to insufficient data. Source: author’s own calculations.

).

Another hypothesis (H2) tested the positive relationship between a company’s financing structure and the ratio of its current liabilities to current assets, the size of the company and the tax and nontax shields.

In the electricity supply section, the results of the study confirmed the correctness of this hypothesis in most cases. Among the 48 cases covered by the study (4 variables, 12 countries), 27 cases confirmed the positive impact of these variables on the dependent variable, with the majority of 22 cases involving the CL/CA and lnTA variables. Both variables were significant in 11 countries, except the CL/CA variable, which was insignificant in Sweden and the lnTA variable in Bulgaria.

In the case of the DA/TA variable, a positive impact occurred in three countries (Germany, Sweden and the UK), in eight it turned out to be negative, and in one country (Poland) it was insignificant. The least conclusive results were obtained for the TAX/GFR variable. In this case, a positive relationship was observed only in Spain and Italy and a negative one in four countries (Bulgaria, France, Slovakia and Sweden), while it was insignificant in the remaining six countries (see Table 5).

In the second section analysed, mining and quarrying, the most unequivocal positive effect on the shaping of the financing structure was observed only for the CL/CA variable. This variable was significant in each country covered in the study.

For the other variables in this section, the number of cases (countries) that were statistically significant was already much smaller and more diverse. For the lnTA variable, a positive relationship was revealed only in French and Spanish companies. In two countries (Romania and Hungary), the relationship turned out to be negative, and in the remaining countries, it was insignificant.

For the noninterest tax shield, a positive impact on debt levels was ascertained in Bulgaria, Sweden and Italy. In two countries (the Czech Republic and Poland), the impact was negative, and in the remaining countries, no significant relationship was found to exist.

In contrast, in the case of the interest tax shield, a positive impact on the size of the debt was found to exist in two countries (the Czech Republic and Italy) and a negative one in three countries (Bulgaria, Sweden and Hungary). In the other countries analysed, the relationship was found to be insignificant (see Table 6).

The last hypothesis (H3) referred to the study of a negative relationship between a company’s financing structure and its asset structure and profitability.

It should be stated that the research covering the electricity supply section confirmed this hypothesis to a fairly large extent. Out of 24 cases analysed, a negative relationship was determined in 16 cases. It is noteworthy that the ROA variable had a negative impact on the dependent variable in all countries under analysis. In the case of the FA/TA variable, a negative impact already occurred in only four countries (Germany, Poland, Romania and Hungary), while a positive impact was observed in five countries (Bulgaria, Czech Republic, France, Spain and the UK). In the remaining two countries, there was no significant impact of the FA/TA variable on the TL/TA dependent variable (see Table 5).

In the second section, mining and quarrying, the research confirmed hypothesis three (H3) to a much lesser extent. Out of the 22 cases analysed, only 10 revealed a negative relationship. The impact of the FA/TA variable on the dependent variable was found to be negative in six countries (France, Spain, Sweden, Hungary, the UK and Italy). In Bulgaria, the survey revealed a positive impact, while in the remaining four countries’ impact was insignificant. For the ROA variable, on the other hand, a negative impact was determined in four countries (France, Poland, Romania and Italy), while the remaining seven countries analysed revealed a nonsignificant impact (see Table 6).

Considering studies by other authors, a positive relationship between the financing structure and company size in the energy sector was shown, inter alia, studies by: Saeed [71], Ghani and Bukhari [60] and Jaworski and Czerwonka [78].

A significant number of studies reveal a positive relationship between the financing structure and asset structure [51,60,67,78].

In turn, a negative relationship between liquidity and the financing structure was confirmed by the studies conducted by the following authors: Liu and Ning [72], Berkman, Iskenderoglu, Karadeniz and Ayyildiz [51], Grabińska B, Kędzior M., Kędzior M. and Grabiński K. [70], Jaworski and Czerwonka [78].

In the case of profitability, the dominant relationship is a negative one, as confirmed by, among others: Saeed [71], Liu and Ning [72], Ghani and Bukhari [60], Chakrabarti and Chakrabarti [67], Grabińska B., Kędzior M., Kędzior D. and Grabiński K. [70], Jaworski and Czerwonka [78].

Among the studies in the mining and quarrying section, the results obtained by Endri et al. [84] should be noted. They studied the impact of oil prices, interest rates, profitability, liquidity and company size on the leverage of mining companies in Indonesia. They showed that the first four factors mentioned have a negative impact on leverage levels, while company size has no impact on debt levels.

It should be noted that, in the research conducted, the most ambiguous determinants of the financing structure include the interest and noninterest tax shields.

The analysis of the research results allows us to conclude that in both the electricity supply and mining and quarrying sections, the studied determinants similarly affected the financing structures of enterprises, and the formulated research hypotheses were partly verified. However, the electricity supply section revealed a greater number of cases of significance. In addition,

Table 7. The relationship between the financing structure and selected determinants in the light of certain theories of capital structure and the obtained research results.

Explanatory Variables	Pecking Order Theory	Agency Theory	Substitution Theory	Signalling Theory	Energy Section	Mining and Quarrying Section
Asset’s structure	−	+	+	n/a	+/−	+/−
Size	+/−	+	+	n/a	+	+/−
Profitability	−	+	+	+	−	−
Liquidity	−	x	+	x	− *	− *
Noninterest tax shield	−	+	−	n/a	+/−	+/−
Interest tax shield	x	x	x	x	+/−	+/−

Notes: + positive relationship, − negative relationship, +/− ambiguous relationship; n/a—no grounds to indicate a relationship; x—not applicable; *—the CL/CA variable used in the research is the reciprocal of the general financial liquidity ratio. Source: author’s own elaboration.

shows the results of the study in the analysed sections against selected capital structure theories. This comparison shows that profitability in both analysed sections strongly and at the same time negatively influenced the share of total debt in the financing structure, which coincides with the assumptions of the pecking order theory. This implies that companies primarily use equity to finance their activities. The second variable that clearly affected the financing structure in both analysed sections is the CL/CA variable, which expresses the inverse of liquidity. The research results reveal that in both the electricity supply and mining and quarrying sections, the CL/CA variable positively influenced the share of total debt in total assets, meaning that as liquidity increased, the share of debt in the financing structure decreased.

Furthermore, research into the electricity supply section reveals a clear impact of company size on the financing structure. This relationship turned out to be a positive one, so large enterprises in this section showed a higher share of total liabilities in their financing structure. It should be added that such a relationship follows from the agency and substitution theories. In the case of the other variables, the research shows that the relationships studied were ambiguous.

6. Conclusions

Decisions on the financing structure of companies are among the most difficult and, at the same time, most important management problems of any enterprise. This is because, in the practice of every enterprise, all large investments are made by employing external capital. Management should shape the structure of financing sources in a way that ensures their company’s continued growth on the one hand and minimises the financial risks of the business on the other. The financing structure is also one of the key factors influencing the current economic performance of a company. Therefore, research into the formation of financing structures and their determining factors is still relevant.

The research presented in this paper covers the energy and mining sectors, i.e., those areas of the economy that are currently undergoing a profound transformation requiring very large investments and major changes in asset structure, which, in turn, implies the search for new financial resources.

The main aim of the paper was to examine the dependence of the financing structures of these companies on selected microeconomic determinants. The research verifies the significance of the impact, strength and direction of the influence of these determinants on the size of debt using data on companies from twelve European Union countries spanning the period between 2012 and 2020. In addition, a general analysis and evaluation of asset, liability and equity structures was conducted.

The research revealed that the indebtedness of enterprises in the EU countries covered by the research in the electricity supply section is at a relatively higher level relative to the mining and quarrying section. Referring to other studies, it should also be stated that in the EU countries, the indebtedness of enterprises in the above sections is at a relatively low level compared to sections such as construction, manufacturing, trade and information and communication [83].

In the countries and sections analysed, the share of fixed assets in relation to total assets remains at a fairly even level. However, the share is at a much higher level in the electricity supply section. The average share of fixed assets in relation to total assets in the electricity supply section between 2012 and 2022 was approximately 70%, and in the mining and quarrying section at approximately 50%. It should also be noted that in almost each of the countries, the share of fixed assets in total assets exceeds the share of liabilities in total assets. The only exceptions involve companies from France, Germany and Italy in the mining and quarrying section.

The research also showed that current assets in the electricity supply section are financed using short-term liabilities to a greater extent than in the mining and quarrying section. In turn, the profitability of companies in the mining and quarrying section was higher compared to the electricity supply section.

The study provided evidence of a relationship between the determinants assumed in the model and the financing structure of the companies under analysis (Hypothesis H1). A significantly higher relevance was revealed by the variables for the companies in the electricity supply section. In this section, out of 72 cases, a statistically significant relationship was revealed in 60, while in the mining and quarrying section, out of 66 cases analysed, this significance was established in 36 cases.

The verification of hypothesis two allows us to conclude that the CL/CA and lnTA variables had the greatest positive impact on the overall financing structure in the electricity supply section. In the case of interest and noninterest tax shields, a greater number of cases involved a negative (12) impact rather than a positive (5) one.

In the second section analysed, mining and quarrying, hypothesis two was fully confirmed, i.e., in all countries under analysis, only for the CL/CA variable. For the other variables, a positive impact on the financing structure was found in only seven countries, and a negative impact also occurred in seven countries.

Hypothesis three, on the other hand, assumed a negative dependence of the financing structure on such variables as asset structure and profitability. In the case of profitability in the electricity supply section, the research confirmed the correctness of this hypothesis in all of the countries analysed, while asset structure had a negative impact on the financing structure in four of the countries studied, and in five countries the impact was positive.

In the second section, i.e., mining and quarrying, a negative relationship between the financing structure and the asset structure was determined in six countries, while a positive one was determined in one country alone. In the case of profitability, the research confirmed a negative relationship with financing structure in only four countries.

As already mentioned, the study covered 12 countries of the European Union, including 6 countries from the ‘old’ EU and 6 newly acceded countries. This selection of countries was aimed at conducting additional research to answer whether the two different economic systems in the second half of the last century were factors that played a role in differentiating the impact of the studied determinants on the structure of corporate finance (in these two groups of countries).

The research did not find that the origin of a country in a particular economic system had a significant impact on the factors shaping the structure of corporate finance. For this reason, the article does not present these research results.

The overall picture that emerges from the research suggests that the analysed sections, i.e., electricity supply and mining and quarrying, are in a favourable economic and financial situation, laying a solid foundation for investment and further development. The results of the study showing a negative impact on profitability and liquidity mean that companies in these sections are not operating under pressure to increase their debt levels. In addition, their sheer level of profitability, relatively low debt level, as well as the dominant share of the state in their capital structure account for their high creditworthiness. It should also be noted that, in the case of the energy sector, the above conclusions coincide with the results of research by other authors (Grabiński and Kędzior).

The studies presented in this paper ended in 2020, which was also the beginning of the SARS-CoV-2 pandemic, as well as the ensuing high turbulence in the energy and coal markets and high inflation. Thus, they did not confirm the economic impacts of these phenomena on the operations of the companies studied. It will undoubtedly be worthwhile to conduct further similar studies showing their impact on changes in the financing structures of the above sectors.