Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries

57 TRENDY EKONOMIKY A MANAGEMENTU TRENDS ECONOMICS AND MANAGEMENT ISSN 1802-8527 (Print) / ISSN 2336-6508 (Online) 2022 39(1): 57–90 DOI: http://dx.doi.org/10.13164/trends.2022.39.57 Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries Nicole Škuláňová, Veronika Šudová Abstract Purpose of the article: The issue of capital or financial structure is still a very current topic, the beginning of which dates back to the middle of the last century. Despite such a long time, there is still no universal theory that would help us understand the behaviour of companies in this area. This is due to the fact that each industry, sector, economy, and even the company itself has a different strategy, and therefore a different financial structure. Unfortunately, it is not possible for researchers to analyse all the companies in the world individually, but it is at least possible to examine individual industries and economies. The purpose of this research is to expand knowledge about the financial structure in the industry – Accommodation and Food Service activities in 8 selected countries of Central and Eastern Europe during the period 2010–2018. A total of 23,991 companies are analysed, which are divided into medium and large. Due to the fact that research in this industry in selected economies was not found, this research could significantly expand knowledge about the financial structure in selected economies and the sizes of companies individually. Methodology/methods: Two methods were chosen to meet the aim – the least squares method and the Generalized Method of Moments. It is a comparison of two regression analyses, a simple one, in which several assumptions must be met, and a modified one, in which only one test follows to verify the credibility of the resulting model. Scientific aim: The aim of this research is to determine whether profitability, liquidity, asset structure, non-debt tax shield, the GDP growth rate, inflation rate, and reference interest rate affect the level of total, long-term and short-term debt. Findings: The main finding of the research is the limitation in the use of the least squares method in terms of fulfilling the basic assumptions and the fact that both internal and external determinants have an influence on the formation of financial structure, however, in terms of significance, the influence of external determinants clearly prevails. Conclusions: The main conclusion is that non-corporate determinants have the most significant impact on the level of indebtedness, with the influence of the reference interest rate clearly dominating in terms of the value of coefficients; while in terms of the frequency of coefficients the GDP growth rate is significant. Keywords: Financial structure, profitability, liquidity, non-debt tax shield, asset structure, GDP, inflation, interest rate. JEL classification: G30, G32 Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 58 Introduction The length of the lifecycle of individual companies always depends on the amount of funds available to the company and can thus finance its business and investment activities. Financial sources can be divided according to various aspects, while the two basic ones are in terms of time (short-term and long- -term) and ownership (own and debt). This breakdown can be found in the basic financial statement – balance sheet, on the liabilities side, where these funds form the capital or financial structure. The key concept for this research is the financial structure, which, unlike the capital structure, which includes only long-term sources of financing, includes all sources of financing. The issue of shaping and optimising the financial structure began to be explored at the beginning of the last century. Since then, economists have been trying to answer the question: “What is the right ratio of own and debt sources of funding?” To date, no universal theory has been found that tells us how companies choose a specific financial structure. According to Myers (2001), this may be due to the fact that countless factors influence the formation of the financial structure. Therefore, new and new studies dealing with this issue are constantly emerging, attempting to find that universal theory. The only thing so far from previous research is that the formation of the financial structure depends on the size of the companies, industries, countries, and the size of the analysed sample. All the mentioned facts are the motivation for this research, which deals with the Accommodation and Food Service activities industry in eight selected countries of Central and Eastern Europe. The main benefit should be the dissemination of knowledge in the field of financial structure for specific economies, which are not examined so frequently. The dissemination of knowledge lies in the fact that selected countries are analysed individually from the perspective of the industry and their size. The companies analysed are divided into medium and large, with each country and size forming one panel (two panels per country), which may not be a matter of course. This will give us slightly more detailed results than if we combined all the countries into a single panel. At the same time, the authors did not find any study that would deal with this industry in the given economies. The authors found several studies involving this industry, but the researches were either focused on a different economy as in the case of Bhaird, Lucey (2010) for Ireland, Mangafić, Martinović (2015) for Bosnia and Herzegovina, Li, Singal (2019) for United States, Sikveland et al. (2022) for Norway. Or the researches included selected industry in one panel together with a number of other industries, and therefore we cannot see the effects of determinants on the financial structure directly in this industry – for example in studies of Šarlija, Harc (2012), Mateev et al. (2013), Strýčková (2015), Lourenço, Oliveira (2017), Yildirim et al. (2018), Matemilola et al. (2018), Matemilola et al. (2019), or Moradi, Paulet (2019). Thus, there is a considerable scope for researchers to examine the industry separately and in all selected economies. The analysis of industry Accommodation and Food Service activities is part of broader research, which focuses on individual industries, primarily from the primary, secondary and tertiary economic sectors. The benefit is also the size of the analysed sample consisting of almost 24,000 companies, all of which are found in the Orbis database for the selected industry. Last but not least, the research attempts to compare two selected methods, namely simple panel regression (least squares method) and modified panel regression using the Generalised Method of Moments. This is an example of the fact that basic statistical methods with all plausibility tests may not always be the right choice and another method can be found and used (especially in the field of finance). Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 59 This paper is organised as follows. Section 1 defines earlier research on the financial structure and selected determinants put forward by this study. Section 2 presents the research methodology, data, and variables and provides the characterisation of the industry, examined economies, and endogenous variable. Section 3 describes the results of the analysis of variable dependencies using the least squares method and Generalized Method of Moments. Section 4 presents the conclusions. 1. Literature overview Forming and optimising the financial structure of a company is a demanding activity due to the number of determinants that can have a significant impact on this process. We can divide these determinants into in-house and out-of-company. The first group of determinants is based on the internal environment of the company and is in a position to influence these determinants. In this research, we will rank profitability, liquidity, asset structure, and non-debt tax shield. The second group of determinants consists of factors that come from the external environment of the company and are often macroeconomic indicators whose impact cannot be influenced by the company. From this group of determinants, this research includes the GDP growth rate, the inflation rate and the reference interest rate of selected economies. In the following paragraphs, the assumptions and previous studies for the individual determinants will be mentioned in turn. Before searching the literature, it is worth mentioning that all determinants can have a positive and negative impact on the level of debt. Literature overview may only be a list of a few studies without further information, but the studies have been selected to include at least one of the selected economies, as most of the studies do. Unfortunately, for all determinants, there are no studies that have analysed selected countries, and therefore studies dealing with different economies are presented also (especially for external determinants). It was important to find out what researches have revealed so far and what results can be expected. Therefore, it is irrelevant to provide more detailed information for the studies, as the impact of a spoecific determinant on the level of indebtedness is important. The rationale for the impacts must be determined by each researcher; it does not matter how the identified impacts were justified by other authors in in the different samples analysed. It was stated in the Introduction that there is room for research in a selected industry in selected economies. The literature overview confirms this statement, as we can see that the studies end in 2017. Thus, there is room for new research in these economies as well, as most of these studies relate to selected economies and we see that the studies are rather outdated. At the beginning of this part, it is appropriate to briefly mention the fundamental studies, as the assumptions of some determinants are derived from them. As mentioned in the introduction to the article, economists have been working on the issue of financial structure since the beginning of the last century, with “The Cost of Capital, Corporation Finance and the Theory of Investment” by Modigliani, Miller (1958) being considered a key study. This study has become a basic source followed by all other economists. At the same time, two basic theories of capital structure emerged from this study, i.e. the trade-off theory and pecking order theory. In the trade-off theory, Brealey et al. (2020) seek the optimum of capital structure through a balance between the tax advantage of debt and the cost of financial distress. In the pecking order theory, Myers (1984) creates a hierarchy of sources of funding, concluding that equity should be preferred to external ones. These two basic theories are then followed by other authors and extended. As the number of studies grows, so do the Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 60 number of known determinants, the number of countries studied, and the industry. The positive impact of profitability on the level of debt is linked to the above-mentioned trade-off theory (Brealey et al., 2020), which states that if companies are more profitable, their financial distress costs decrease, which reduces the likelihood of bankruptcy and companies are more attractive for lending. This relationship has been confirmed by Klapper et al. (2002), Pinková (2012), Aulová, Hlavsa (2013) and Mokhova, Zinecker (2013) for Slovenian companies, and Růčková (2015a, 2015b, 2017) in the Czech Republic and Hungary. The negative impact of profitability, on the other hand, is associated with the pecking order theory (Myers, 1984), which argues that as profits grow, so do its various parts, such as retained earnings, which are a cheap means of financing. This link is far more common in previous research. The negative link also prevails when divided into total, long-term and shortterm debt. This association was stated by Nivorozhkin (2002, 2005), Bauer (2004), Weill (2004), Črnigoj, Mramor (2009), Hernádi, Ormos (2010, 2012), Hanousek, Shamshur (2011), Jõeveer (2013), Mateev et al. (2013), Mokhova, Zinecker (2013), Prędkiewicz, Prędkiewicz (2015), and Růčková (2015b, 2017) for Poland and Slovakia. The positive impact of liquidity on the level of debt is associated with the amount of liquid assets (e.g. marketable securities, bills of exchange), which can be sold relatively easily in the event of an unfavourable period. However, in order for a company to sell such assets, it must own a certain amount of those assets. Unfortunately, illiquid assets (e.g. fixed assets) are difficult to sell and their sale usually carries a higher loss. At the same time, it should apply in the case of (non)liquid assets, liquid assets should be financed by debt, illiquid assets with equity. This impact is supported by the results of, for example, Williamson (1988), Shleifer, Vishny (1992), Mateev et al. (2013) for long-term debt, and Růčková (2015b) for the Czech Republic. There are several explanations for the negative impact of liquidity. The basic explanation is the potential conflict between managers and owners in the event that if managers could freely dispose of corporate assets, they could expropriate the owners through a gradual sale. The second and simpler explanation is that the more liquid the company’s assets, the more its debt would decrease, as higher liquidity can lead to low investment activities and therefore no debt financing is needed. This influence is supported by the results of Myers, Rajan (1998), Morellec (2001), Frieder, Martell (2006), De Jong et al. (2008), Lipson, Mortal (2009), Mateev et al. (2013) for short-term debt, Pinková (2012), Aulová, Hlavsa (2013), and Růčková (2015b) for Poland and Slovakia. The impact of the asset structure depends on the selected variables, the form of indebtedness and certain special cases. As for the variable, the most common indicator is the share of tangible fixed and total assets. Tangible fixed assets are used because they are assets that are used as collateral to obtain debt financing. Therefore, the more such assets a company has, the more room it has for debt financing. According to a study by Titman, Wessels (1988), intangible assets are not used as collateral and, moreover, as has been said for liquidity, these assets are very difficult to sell in the event of existential problems. In terms of the form of indebtedness, long-term debt is expected to have a positive impact, given that fixed assets (e.g. real estate, machinery) are usually used as collateral. Conversely, short-term debt is expected to have a negative impact, as inventories and similar assets are not theoretically used as collateral, although of course there is collateral in practice in the form of inventories or unmined minerals, etc. A positive impact can be found in Michaelas et al. (1999), Klapper et al. (2002), Nivorozhkin (2002), Delcoure (2007), De Jong et al. (2008), Hernádi, Ormos (2010, 2012), Kayo, Kimura (2011), Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 61 Mokhova, Zinecker (2013), and Vo (2017). On the other hand, there are studies by Klapper et al. (2002), Nivorozhkin (2002), Bokpin (2009), Mokhova, Zinecker (2013), and Vo (2017). The latter are special cases in which the previous assumption settings may differ slightly. The first case is the financial system of the analysed economy, because in the case of a financial system oriented to financial markets, the collateral cannot be expected, as it is used only in bank-oriented financial systems, as stated by Antoniou et al. (2008) and Acedo-Ramírez, Ruiz-Cabestre (2014). The size of the company also has a significant effect on the assumptions, as large amounts of tangible assets are usually held by medium-sized and especially large companies, as stated by Michaelas et al. (1999), Klapper et al. (2002), Onofrei et al. (2015) and Lourenço, Oliveira (2017). Last but not least, it depends on the choice of the sector examined, as sectors with a large amount of inventories, such as agriculture or construction, should not, in theory, use stocks as collateral; as reported by Aulová, Hlavsa (2013) and Růčková (2015a). The non-debt tax shield is considered a substitute for the tax shield and should reduce the level of debt compared to it. This expectation stems from a variable consisting of depreciation, which acts as an own source of financing that can be used for corporate financing. A negative relationship has been confirmed, for example, by Michaelas et al. (1999), Wald (1999), Klapper et al. (2002), Song (2005), Hernádi, Ormos (2012), and Acedo-Ramírez, Ruiz-Cabestre (2014). The positive impact on the level of debt is explained either by the equality of depreciation with the value of tangible fixed assets (if there are almost same values, companies would rather use fixed assets as collateral than depreciation) or by the existence of differences in tax regulations in the countries analysed. Delcoure (2007), Hernádi, Ormos (2010), Aulová, Hlavsa (2013), and Mokhova, Zinecker (2013) found a positive effect. The development of the economic cycle is linked to the impact of the development of the GDP growth rate on the level of debt. The positive impact of GDP on debt can be explained, for example, by the period of economic growth, when, in addition to the economy, profits to individual companies usually grow and everyone has optimistic expectations, so creditors are willing to lend to almost anyone and debt can rise. In a recession, the opposite is true. The positive impact was confirmed in these studies, e.g. Gajurel (2006), Hanousek, Shamshur (2011), Salehi, Manesh (2012), Çekrezi (2013), Mursalim, Kusuma (2017), and Yinusa et al. (2017). The economic cycle can also explain the negative impact of GDP growth on the level of debt. At the beginning of the paragraph, we mentioned that in a period of economic growth, companies usually grow profits that can be used as their own source of financing for investment activities, and therefore there is no need for debt financing. The negative impact has been confirmed, for example, by Cheng, Shiu (2007), Gajurel (2006), Bastos et al. (2009), Bokpin (2009), Hanousek, Shamshur (2011), Jõeveer (2013), or Mursalim, Kusuma (2017). The impact of the inflation rate on the level of debt is also expected to differ with regard to the form of debt. Long-term debt is expected to have a negative impact, as the inflation rate should reduce existing debt together with a decline in the real interest rate. This relationship can be found, for example, in Gajurel (2006), Cheng, Shiu (2007), Jõeveer (2013), Mokhova, Zinecker (2014), or Öztekin (2015). Short-term debt is expected to have a positive impact, given that creditors can hedge against lower real interest rates, but the hedging is short-term. This relationship can be found in Hanousek, Shamshur (2011), Mokhova, Zinecker (2014), and Yinusa et al. (2017). The influence of the price of external sources of financing can be based on a logical assumption – the higher the interest rate Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 62 (debt financing costs), the more expensive and less preferred debt financing is, and vice versa. However, Yinusa et al. (2017) found other possible explanations, namely the impact of the quality of the institutional, legal and regulatory environment. In their study, they divide economies into developed and developing. The result of the study is that developed economies have a high-quality institutional environment, good creditor protection and legal enforcement of liabilities, etc., while developing economies may lack quality in these areas. It follows that developed economies should have a positive impact and emerging economies a negative impact. 2. Data and methodology The subject of this research includes companies classified in Section I – Accommodation and Food Service activities according to the NACE classification. The input time series come from the Orbis and World Bank databases. A total of 23,991 companies were analysed, of which 22,973 are medium-sized and 1,018 are large and very large companies. The selected industry includes facilities that provide customers with short-term accommodation or the preparation of food, snacks and beverages for immediate consumption. The industry includes both accommodation and catering facilities, as the two activities are often combined in the same facility. The analysis covers the period 2010–2018. Eight economies from Central and Eastern Europe were selected for the analysis of the selected industry, namely the Czech Republic (CZ), Slovakia (SK), Poland (PL), Hungary (HU), Austria (AT), Slovenia (SI), Bulgaria (BG), and Romania (RO). These countries belong to the so-called extended Visegrád Group. It might seem that Austria, Slovenia, Bulgaria, and Romania are not members of the original V4, but these countries are very often associated with this group, as representatives of these countries attend various meetings of this group and cooperate with it in different areas (e.g. ministry of agriculture, energy, climate policy, or territorial development coordination). It is clear from the values of economic indicators that, for example, the Austrian economy is somewhere other than the Romanian or Bulgarian economy in terms of the level of indicators, but it is a relatively well-established merger of these economies into an extended Visegrád Group, whose companies have therefore become the subject of this research. At the same time, the countries concerned were selected due to the lack of studies within these countries for the sector. The aim of this research is to determine whether profitability, liquidity, asset structure, non-debt tax shield, the GDP growth rate, inflation rate and reference interest rate affect the level of total, long-term and short-term debt. With regard to the literature search, the following two research questions were formulated: Table 1. Expected impacts of individual determinants on the amount of individual forms of debt. Total debt Long-term debt Short-term debt Profitability – – – Liquidity – – – Asset structure –/+ + – Non-debt tax shield – – – GDP growth rate –/+ + – Inflation rate –/+ – + Reference interest rate – – – Source: authors’ calculations. Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 63 ● Are there differences in impacts in terms of the different maturities of the used funding sources? ● What impact does the price of financial external sources have on the used sources of financing? Due to the literature search of previous studies, we can create assumptions of the resulting impacts for individual determinants within the research questions. The examined companies belong to the group of medium and large companies, so there is no need to divide the assumptions according to the size of the companies, as there should be no difference between these companies. Table 1 shows the expected impacts for each determinant. A sign “+” indicates a positive impact of a given determinant on the level of the selected form of indebtedness, while a sign “–” indicates a negative impact and a sign “–/+” indicates the possibility of both positive and negative impact. 3.1 Variables In the Results section, several variables are used within the least squares method and the Generalized Method of Moments, which will be described in this subchapter. The endogenous variable is represented by debt, which comes in three forms. Total debt (DER) is the ratio of total liabilities to equity. Long-term (DER_L) and short-term (DER_S) debt differ from total debt in that, instead of total liabilities, there are long-term or short-term liabilities. The right side of the regression equations is made up of seven exogenous variables that take the form of individual determinants. Profitability takes the form of a return on equity (ROE), which in this case is the ratio of earnings before interest and taxes and equity. Profit before tax was selected to abstract from divergent taxation, as each of the eight selected economies has a different tax policy. The L2 indicator (quick ratio) was selected from the liquidity indicators, i.e. the ratio of current assets adjusted for inventories and short-term liabilities. The structure of assets can also be expressed by a number of indicators. In this research, this is the most common share of tangible fixed and total assets. The non-debt tax shield is represented by the ratio of depreciation and total assets. The remaining three variables represent the external environment of companies and are the GDP growth rate at market prices, the inflation rate and the reference interest rate of the given economy. 3.2 Methodology Two methods were chosen to analyse the influence of individual determinants on the level of debt, both of which are regression analyses. In the first case, it is the least squares method and in the second case it is the Generalized Method of Moments (GMM), which is a modified basic regression analysis. The article tries to compare these two methods with regard to the difficulty of verifying their results from the point of view of plausibility. The resulting models of the least squares method must meet several basic assumptions and tests, while the GMM method only needs to perform a single test after analysis. The default equation for both methods looks like this: Yit = α0 + β1·ROEit + β2·L2it + β3·SAit + β4·NDTSit + β5·GDPit + β6·INFit + β7·IRit + εit (1); where: Yit characterizes endogenous variables, i.e. debt in three forms (DERit, DER_Lit, DER_Sit), where DER denotes the debt-to-equity ratio for the i-th number of companies in a given economy in a particular sector during period t (2010–2018); α constant; ROE return on equity; L2 liquidity – quick ratio; SA asset structure; NDTS non-debt tax shield; Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 64 GDP GDP growth rate; INF inflation rate; IR reference interest rate; ε random component, which includes all other factors that affect the amount of debt. The following two subchapters will mention the basic characteristics of the two selected methods. 3.2 Least squares method The results of the least squares method need to be verified by some tests and at the same time the time series should meet certain specifics. The input time series should be stationary, which means that their probability distribution is constant over time. A number of tests can be used to verify this assumption; in this research, the Augmented Dickey-Fuller test is used, whose null hypothesis states that there is a unit root that indicates non-stationarity. The resulting p-value can be tested at the most common values of significance levels 0.01 and 0.05. If the resulting p-value is smaller, the null hypothesis is rejected and the time series is stationary (Enders, Lee, 2012). A normal distribution of residues in time series is also assumed. According to Asteriou, Hall (2021), residues should be independent and identically distributed with a mean of zero. Determination of skewness and sharpness values is often used to assess normal distribution, but complex tests such as the Jarque-Bera test can also be used. The null hypothesis of this test states a normal distribution of residues. The resulting p-value can be tested again on the most common values of significance levels 0.01 and 0.05. If the resulting p-value is higher, the null hypothesis cannot be rejected and the residues have a normal distribution. As for the tests after the compilation and acquisition of the resulting models, the basic test to verify the overall significance of the model is the F-statistics. This test tests whether an endogenous variable is a linear combination of selected functions of an exogenous variable. The null hypothesis states that the regression parameters are zero and there is no statistically significant relationship between the endogenous and exogenous variables, i.e. the regression model is inappropriately selected. The resulting p-value can be tested at the most common values of significance levels 0.01 and 0.05. If the resulting p-value is smaller, the null hypothesis is rejected and the regression model is chosen correctly (Jamshidian et al., 2007). Another assumption concerns multicollinearity, which means a correlation between exogenous variables. It is necessary to perform a correlation analysis if we have more than one exogenous variable in the model. If the correlation coefficient is high, the quality of the regression model is reduced. A value of –0.9 or 0.9 is considered high in many studies. A correlation coefficient that reaches such a value indicates that one of the exogenous variables is redundant in the model. The presence of multicollinearity can cause an artificial increase in the coefficient of determination, which would state that the model explains more of the behaviour of the endogenous variable, when in fact it would not. There are several correlation coefficients, in this research Pearson’s correlation coefficient will be used, which characterises only the linear relationship, in other words it reflects only the variability around the linear trend. The coefficient can take values in interval <–1; 1> (Asteriou, Hall, 2021). Serial independence states that residues are independently distributed and not correlated. A frequently used test to verify autocorrelation is Durbin-Watson statistic, whose null hypothesis states that there is no autocorrelation between residues (Yin, 2020). The resulting p-value can be tested at the most common values of significance levels 0.01 and 0.05. If the resulting p-value is higher, the null hypothesis cannot be rejected and Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 65 the residues are not autocorrelated. In addition to using p-value results, the value of the coefficients of this statistic can be used. The coefficients are in the interval <0; 4> and ideally the value of the coefficient should be as close as possible to 2. If the value is less than 2, it is called a positive autocorrelation; conversely, if the value is greater than 2, it is a negative autocorrelation. However, a value of 2 is very rarely obtained, so researchers use certain intervals within which it is claimed that residues are not autocorrelated. There is no given interval, we can find e.g. these intervals: <1.8; 2.2> or <1.6; 2.4> (Asteriou, Hall, 2021). The last basic assumption is homoskedasticity, whose null hypothesis states that the variance of residues is constant. To verify this assumption, tests are used to detect the presence of heteroskedasticity, which tells us that the dispersion of residues is not constant. There are a number of tests, such as the Breusch-Pagan test, whose null hypothesis states homoskedasticity. The resulting p-value can be tested at the most common values of significance levels 0.01 and 0.05. If the resulting p-value is higher, the null hypothesis cannot be rejected and the residues have normal variance (Greene, 2020). Last but not least, another coefficient can be used to determine the quality of the regression model, namely the determination coefficient (R2), which, as stated by KurzKim, Loretan (2014), expresses what proportion of the variability of the endogenous variable the given model explains. The value of the coefficient should be in the range from 0 to 1. The higher the value we get, the better the model we hypothetically constructed. However, this indicator is not very reliable because it can be skewed, e.g. if we add more exogenous variables that are not necessarily related to the problem, it may seem that the model contains a lot of exogenous variables that explain the behaviour of the endogenous variable and the coefficient. The determination can be almost 1. A similar bias can be caused by the presence of multicollinearity. 3.2 Generalized method of moments Simple panel regression is a suitable method with respect to the amount of input data. However, the often-used least squares method is not entirely appropriate, as the basic premise of this method is stationary time series, which macroeconomic series in particular may not meet, and thus we could eliminate some variables from our models (Průcha, 2014). Therefore, a modified panel regression was selected in the form of a two-stage Generalized Method of Moments (GMM) system, the development of which had a major impact on research in finance. This method overcomes a number of limitations of other methods – for example, there is no need for the already mentioned stationary data, nor is there a need to create distribution assumptions, which means that variables can show serial correlation and conditional heteroskedasticity (Jagannathan et al., 2002). The GMM method can be found for the first time in the studies: Arellano, Bond (1991), Arellano, Bover (1995), and Blundell, Bond (1998). These studies contain general assumptions of this method: short time series and many observations, linear functional relationship, one endogenous variable on the left that is dynamic depending on its past values, exogenous variables that may not be strictly exogenous (correlation with past or current errors), fixed individual effects and the mentioned autocorrelation and heteroskedasticity within individual observations, but not across them. The GMM model thus solves the endogeneity problem, which means the correlation between the explanatory variable and the error term (Roodman, 2009). The GMM method includes certain internal tools (lagged value of the dependent variable, internal transformation processes) in solving unobserved heterogeneity, simultaneity and Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 72 Medium-sized companies CZ SK PL HU AT SI BG RO BP p-value 0.4785 3.80E–05 4.00E–05 7.77E–203 0.8706 0.3053 5.69E–07 2.33E–26 Total debt α 0.3654 34.9275 –6.8005 –3.5569 –15.7629 –5.3069 –0.7199 –16.2974 ROE 1.29E–05 –4.4975a1.2119a5.2172a5.0092 X0.8096a1.9353a L2 0.0055 –0.0168 0.0051 0.0042 –1.26E–06 –3.25E–05 0.0090 0.0320 SA 7.0792 –5.1697 15.8302 2.9887 –124.2888 16.9918 –15.0254 –2.6817 NDTS X32.9592 11.1225b2.1936 16.5147 52.1469 3.3492c0.2772 GDP –42.4609 –639.0139 –21.6384 143.0846 –129.5699 –641.2237 49.9638 169.5565 IR –162.1402 –3,066.5257 91.3720 31.6203 361.8550 –7,270.3045c767.6774 374.0239 INF 1,076.3191 394.9006 –27.7448 –38.0875 1,078.5790 283.1800 44.3590 –14.6033 DW 2.0290 2.0010 2.0530 1.9840 2.0020 2.0040 2.0000 1.9940 F-stat. p-value 0.6540 0.0000 1.18E–12 0.0000 0.8750 0.3400 0.0001 1.26E–186 R20.20 % 30.40 % 1.00 % 11.60 % 0.10 % 0.30 % 0.20 % 5.10 % BP p-value 0.2310 7.09E–07 0.5558 8.44E–200 0.8819 0.3045 0.0143 5.73E–07 Source: authors’ calculations. Symbols a, b and c indicate significance at 1%, 5% and 10%. Table 5. Results of the least squares method for medium-sized companies of individual economies. (Continuation) Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 73 Table 6. Results of the least squares method for large companies of individual economies. Large companies CZ SK PL HU AT SI BG RO Short-term debt α 0.6930 119.4207 16.3185 11.7986 48.7142 9.206a3.8468a0.1061 ROE 3.8891a–36.9403 42.0318a3.5013a5.2917 –0.823 2.3782a1.0324a L2 –0.0510 –20.3979 –0.0004 –2.4746 0.0005 –1.7362b–0.0014 –0.0002 SA –27.1508c–630.0940 –294.3458b–51.0841b56.8629 –5.614 –12.6813 0.4186 NDTS 2.1716 –12.8321 –17.3085 –5.9230b54.7948 –7.019 –1.9565 0.2438 GDP –10.3628 1,046.7345 –365.2321 –86.2626 2,016.5907 –5.616 –28.1489 –1.8632 IR –70.4731 28,300.000 486.0959 –16.3175 –4,522.5757 –26.046 –178.0959 –13.5153 INF 5.1337 –9,553.8708 –179.7909 –45.7174 –3,018.6877 –64.4103 34.2959 41.4908 DW 1.9910 2.0100 2.1470 2.0350 2.0250 2.107 1.9790 1.9320 F-stat. p-value 7.66E–40 0.2752 1.02E–79 9.91E–14 0.2990 0.0056 9.51E–119 2.13E–12 R228.30 % 0.80 % 34.00 % 12.90 % 1.10 % 11.40 % 51.30 % 10.00 % BP p-value 0.0014 0.0915 1.59E–105 7.01E–09 0.358 0.703 1.73E–04 0.0315 Long-term debt α –4.8670 1.3750 6.0076 11.2399 3.5370 0.4907 2.8422 –1.3488 ROE 22.8419a0.5974 0.0494 –58.5515a5.3917a–1.8507a0.4112c0.0862 L2 0.2385 –0.5378 0.0805a6.9901 –1.16E–05 0.2749 –0.0041 –0.0005 SA –99.2385c–12.7710 –10.1369 –325.9859 3.5740 –10.7683 –59.9353b0.6999 NDTS 12.7213 –3.8975 –0.0352 37.6857 12.1390c0.6488 8.6482b–1.3262 GDP –16.8841 117.9274 41.8460 326.2077 –361.8124 17.8653 –6.9127 19.3159 IR –545.8545 –676.0491c–313.3192 608.8441 –147.5838 226.4462 –3,014.4809 51.4282 INF 180.7798 49.8471 101.0668 –398.8043 –149.9852 –35.2555 –46.0758 –0.5561 DW 1.9670 1.9880 2.1340 2.0360 2.0110 2.0230 2.0400 2.0270 F-stat. p-value 1.94E–91 0.3230 6.96E–37 1.67E–17 3.62E–17 2.34E–05 0.0207 0.5720 R251.90 % 3.20 % 18.10 % 15.80 % 12.00 % 18.20 % 2.10 % 0.90 % BP p-value 1.57E–39 0.7358 0.0543 4.80E–05 0.0341 0.1443 0.0045 0.6648 Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 74 Table 6. Results of the least squares method for large companies of individual economies. (Continuation) Large companies CZ SK PL HU AT SI BG RO Total debt α –4.1741 120.7958 22.3261 23.0384 52.2512 9.6963a6.6890 –3.1224 ROE 26.7310a–36.3429 42.5256a–55.0503a10.6834b–2.6733a2.7895a2.1304a L2 0.1875 –20.9357 0.0800b4.5155 0.0005 –1.45E+00 –0.0055 4.65E–05 SA –126.3893b–642.8651 –304.4827b–377.0701 60.4369 –16.3826 –72.6166b0.0831 NDTS 14.8930c–16.7296 –17.3437 31.7627 66.9338 –6.3701b6.6917 0.3200 GDP –27.2469 1,164.6619 –323.3861 239.9451 –2,378.4031 12.2490 –35.0617 26.1797 IR –616.3276 27,620.0000 172.7767 592.5266 –4,670.1594 200.4000 –3,192.5768 53.2375 INF 185.9135 –9,504.0238 –78.4241 –443.8712 –3,168.6729 –99.6658 –11.7799 39.2534 DW 1.9750 2.0100 2.1350 2.0360 2.0240 2.0960 2.0220 1.9590 F-stat. p-value 1.91E–107 0.2738 7.30E–77 1.63E–14 0.1550 0.0018 1.41E–22 2.75E–23 R257.50 % 0.80 % 33.00 % 13.50 % 1.50 % 12.90 % 13.90 % 16.90 % BP p-value 6.84E–68 0.0912 3.59E–105 2.17E–04 0.3565 0.8653 0.0168 0.0048 Source: authors’ calculations. Symbols a, b and c indicate significance at 1%, 5% and 10%. Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 75 behaviour of large companies. If we do not take into account the maxima, then less than 7 and 18% is not a very high value. Although the models explain some of the corporate debt behaviour of selected companies, especially in medium-sized companies, the models do not explain anything. Another test that we can observe in the second part is the Breusch-Pagan test (BP p-value) to verify the presence of heteroskedasticity. In order to satisfy the condition of Breusch-Pagan test, the p-value must be higher than the usual significance level of 0.05. We see this condition that not even half of all the resulting models meet. Bold values indicate that this test has been completed. The presence of heteroskedasticity reduces the credibility of the models. The last test is needed to verify multicollinearity. The correlation analysis is used to find out if there are unnecessarily many exogenous variables in the model. Some exogenous variables may develop similarly (e.g. macroeconomic time series) and therefore it is not necessary to include both time series in the model, as this could distort the resulting model and, for example, artificially increase the coefficient of determination. The correlation coefficients should be in the range (–0.9; 0.9) in order for this condition to be met and the model to be considered plausible. Since a total of 16 time series were analysed, 16 correlation matrices were obtained from the correlation analysis, which can be seen in the Annex. We see that there are several conflicts (bold coefficients). This is usually a correlation between endogenous variables, which does not matter, as each form of debt examined separately and whether the forms of debt are correlated with each other is irrelevant. However, we can observe three cases where exogenous variables correlate with each other or indebtedness with an exogenous variable. These are Czech medium-sized companies where there is a correlation between profitability and the non-debt tax shield. One of these variables should be removed for the model accuracy. As part of the regression analysis, the non-debt tax shield was removed, which indicates the letter X in Table 5. In the same table, we can see a second conflict, namely the correlation between profitability and total debt/ long-term debt for Slovenian medium-sized companies. Therefore, profitability has been removed from the model for long-term and total debt, which again indicates the letter X. An important conclusion of this section consists in the fact that although regression models may seem credible through F-statistics, which for most models have confirmed that they are compiled correctly, the opposite is true. After performing the basic tests, the models would have to be modified a lot, while some determinants would have to be removed, which would greatly change the original model and expectations. However, it may seem that we did not meet the assumptions that ensured that we did not reach any conclusions. The opposite is true, as the need to meet a number of requirements has shown well that this method is really very unsuitable for the analysis of corporate panel data. 4.2 Generalized method of moments In Tables 7 and 8, we can see the results of panel regression using the GMM method for companies of both sizes. At first glance, it is clear from the tables that the results are not available for all countries, as the number of rows does not correspond to the numbers of selected economies. The reason is the fact that was discussed in the Methodology section, in which the Sargan test was mentioned, which serves to verify the plausibility of the resulting model with respect to the presence of autocorrelation or heteroskedasticity. Economies within the various forms of debt that we do not see in the tables did not pass this test – they did not exceed the limit of 0.05. The values for economies that have passed the Sargan test (J-stat.) can be seen in the last columns. In the following paragraphs, the individual influences of determinants for individual Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 76 companies will be analysed according to their size, but some results can be summarised for all companies together, because the values of coefficients reach very low values and in fact there is no significant effect on debt. This is lagged value of the debt. In terms of the relationship between current and past debt, a positive impact prevails in both size categories of companies. Given the size of the coefficients, we can only talk about an indication of the effect of this variable. The positive effect implies that if companies used debt financing in the previous period, they are likely to use it in the following period as well, and the level of indebtedness will increase. On the contrary, the negative impact indicates the exact opposite, namely that if companies used debt financing in the previous period, they are unlikely to use it in the following period and the amount of debt will thus decrease. According to previous studies, profitability should have a rather negative impact on the level of debt. This effect was found in medium-sized Slovak, Slovenian, Romanian companies and in large Czech, Slovenian and Bulgarian companies. The remaining impacts identified are positive. The negative effects of profitability are in line with the results of e.g. these studies Črnigoj, Mramor (2009), Hernádi, Ormos (2010, 2012), Mokhova, Zinecker (2013), Růčková (2015b, 2017) and GDP growth rates with the studies Bastos et al. (2009), Bokpin (2009) and Jõeveer (2013). The negative effect means that if these companies are growing in profits, the companies should prioritise rising profits as a source of financing, and the level of debt should therefore fall. For all companies except large Slovenian companies, the impact of profitability is also supported by the negative impact of GDP on debt. This impact is linked to the claim that, during a boom, companies usually grow profits, which are a suitable source of financing. As for the positive impact of profitability on the level of debt, here too most of the results are supported by the same effects in terms of the impact of GDP. These effects mean that, for example, in the case of economic growth, companies usually grow profits and thrive overall, which reduces the risk of bankruptcy and therefore lenders are willing to lend them additional funds. The positive impact of profitability can be found, for example, in the studies of Klapper et al. (2002), Pinková (2012), Aulová, Hlavsa (2013), and Růčková (2015a, 2015b, 2017) and the GDP growth rates in the studies of Salehi and Manesh (2012), Mursalim, Kusuma (2017), and Yinusa et al. (2017). All economies for which statistically significant coefficients were found performed well for at least half of the period under review. Economies such as Hungary, Romania and Slovenia were hit hard by the global financial crisis, which was still lingering at the beginning of the period under review, and the Hungarian and Romanian governments even had to seek financial assistance, but economic problems were overcome and Romania and Hungary grew during the period, on average over 2.5% per year. In some years, the rate was even over 5% and more. The Slovenian economy grew by an average of 1.7% per year, as in addition to the initial real property and mortgage crisis, the economy also went through a banking crisis. The remaining economies did not suffer the significant effects of the global financial crisis – e.g. the Polish economy, as one of the few economies in the world, did not experience an economic downturn during the whole period and showed a good growth rate throughout the year, averaging 3.6% per year. The Bulgarian and Slovak economies grew on average 2.5% per year, and there were also no significant economic problems in these countries during the period under review. The Czech economy recorded a decline in 2012/2013, when household consumption and investment fell in particular. However, apart from these years, year-on-year growth averaged around 3 %. In Austria, the development of basic economic indicators (debt, Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 77 unemployment) was not favourable, but apart from the introduction of a deposit guarantee, the economy was no longer constrained. Due to the favourable economic development, companies were able to choose whether to use their own sources of financing or debt financing. Economic conditions gave them the opportunity to do so. In terms of liquidity, it depends on the size of the companies, as for medium-sized companies the coefficients are very low, which is not uncommon for this determinant and is only an indication of the effect on the level of indebtedness. The negative impact clearly prevails here, which we can see in all cases in large companies, with the difference that in large companies the coefficients reach much higher values. The negative impact means that companies do not have highly liquid assets, as these assets are usually acquired on debt. To confirm this statement, it would be appropriate to look at the detailed structure of the assets. There may also be conflicts between owners and managers and expropriation of owners, but this is unlikely to occur in all cases. The first explanation is far more likely. Unfortunately, the detailed structure of the assets was not analysed in this research. A negative result is followed by results such as Myers, Rajan (1998), Mateev et al. (2013), Aulová, Hlavsa (2013), and Růčková (2015b). The relationship between asset structure and debt levels should be positive for Table 7. GMM results for medium-sized companies. Medium-sized companies DER(–1) ROE L2 SA NDTS GDP IR INF J-stat. Total debt CZ 0.0527a–0.0030a5.1516a32.9204a173.0985a0.4371 SK –4.9185a0.0030a–8.1956a551.9702b0.2038 PL –0.0079b3.2266a8.5093c52.1978a112.1304a0.3027 HU –0.0030a0.0214b2.2913b35.8212a–68.5682b0.0734 SI –2.2163a6.0380c111.9825a0.2981 BG 0.0068b–0.0084a–6.7458a–1,306.0020b0.5717 RO 0.0200a–1.9330a5.9717a–108.3550a102.2906a0.4834 Short-term debt CZ 0.0686a1.4336a22.5941a90.7396a0.4663 SK –4.5274a–0.0049b–107.6472a1,167.4910a0.1400 PL –0.0431a5.4147b–6.9645a–86.8116a62.2194a0.1955 HU 0.0047a2.6482a–0.0067a 2.1044b7.2949a–38.0782b0.2650 AT –0.0203a–15.4493a–7.3107c129.9113a0.2059 BG 0.0664a–0.0012a12.6994a0.5643 RO 0.0039b–1.5120a3.0467b–2.3016a–173.4417a0.5044 Long-term debt CZ –0.0300a–0.0042b4.6038b20.7818b40.7564c0.5169 PL 0.0385a1.6955a–3.5844a44.4212a32.0995b0.1955 SI –0.0021a–0.1677c2.7280a–1.5276a–4.9497c0.1036 BG –0.0371b21.7240a849.1702a27.0029b0.5071 RO 0.0614a–0.4084b3.3892c–4.7931a–6.0896a0.1811 Source: authors’ calculations. Symbols a, b and c indicate significance at 1%, 5% and 10%. Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 78 long-term debt and negative for short-term debt, as reported by Klapper et al. (2002), Nivorozhkin (2002), Song (2005), Cheng, Shiu (2007), Mateev et al. (2013), and Vo (2017). The positive effect means that the higher the share of tangible fixed and total assets, the higher the value of debt. This relationship is based on the assumption that tangible fixed assets can be used as collateral, usually for long-term debt, but not for shortterm debt. However, corporate reality is different from theory and collateral can often be used for short-term liabilities. Our results show that the level of corporate debt, regardless of the size and form of debt, is clearly positively affected by the structure of assets. If we look at the ratio of tangible fixed to total assets, the result is not surprising. This ratio is on average 71% for medium-sized companies and 68% for large companies. These are high values, from which it is clear that companies have a large amount of assets that they can use to secure when obtaining debt financing. The non-debt tax shield should have a negative effect on the amount of debt. We confirmed this result in all cases except medium-sized and large Hungarian companies. A negative result follows the results of studies such as Wald (1999), Klapper et al. (2002), or Hernádi, Ormos (2012). Companies with negative coefficients benefit from depreciation, which serves as their own source of financing, and should therefore acquire more assets that can be depreciated and, if possible, assets that have higher depreciation rates. These are mostly fixed assets, of which companies have a large number, as mentioned in the influence of the asset structure. On the other hand, positive impacts have been found, for example, by Delcoure (2007), Hernádi, Ormos (2010), and Mokhova, Zinecker (2013). One possible explanation for the positive impact is roughly the same value of tangible fixed assets and depreciation. If these two groups of assets were more or less equal, it would be more advantageous for companies to use collateral than a non-debt tax shield. However, after an additional analysis, it was found that the value of tangible fixed assets and depreciation did not equal or approach them. For medium-sized companies, the value of tangible fixed assets is about 4 times higher than the value of depreciation, for large companies this value is about 2 times higher. Therefore, the differences in tax regulations may explain this, as we do not have detailed internal accounting of all Hungarian companies in which the answer could probably be traced. Another determinant is the reference interest rate, which was expected to have a negative impact on the level of debt. We see that the resulting impacts are diverse, but in terms of company size they agree. We see a negative impact on Polish, Romanian and Hungarian companies, regardless of size, and on medium-sized Bulgarian companies. The result for Bulgarian companies is surprising and difficult to explain, as the average interest rate was very low – 0.05%. However, in the remaining economies, interest rates were high, leading to higher debt financing costs, which meant lower debt. The Romanian reference interest rate averaged 3.6% with a maximum of 6.3% (2010). The Polish interest rate averaged 2.5% with a maximum of 4.5% (2011). The Hungarian interest rate averaged around 3.1% with a maximum of 7% (2011). It is clear that the values are really high compared to the rest of the economies, but it must be added that they have been gradually declining since their peaks in 2010/2011 and the Hungarian interest rate reached 0.9% at the end of the period analysed, which already has the advantage of lower debt financing costs. Unfortunately, interest rates were higher for most of the period under review, and their resulting development outweighed the resulting coefficient. The remaining economies have a positive impact on debt levels. The result is that interest rates were low in the economies. Slovakia and Slovenia are members Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 79 of the euro area and are therefore subject to the monetary policy of the European Central Bank, which sought to help economies as much as possible during the period under review in the face of crises and economic problems, with an average reference interest rate of 0.3%. The Czech central bank also tried to help the economy and set interest rates very low, averaging 0.4%. Such low interest rates bring very low costs of debt financing, which thus becomes very attractive, and therefore the level of debt to companies has risen. The inflation rate was expected to have a negative impact on long-term debt and a positive impact on short-term debt, as we saw in the studies of Gajurel (2006), Cheng, Shiu (2007), Hanousek, Shamshur (2014), Mokhova, Zinecker (2014), Öztekin (2015), and Yinusa et al. (2017). The resulting coefficients for the impact of the inflation rate on the debt ratio for short-term debt meet our assumptions; however, for long-term debt, the results differ significantly from expectations. In almost all cases, we can observe a positive effect of the inflation rate on the level of debt. The positive impact is strange, as it tells us that debt levels rise as inflation rises. For short-term debt, the positive effect is justified, as short-term inflation can be linked to the interest rate (debt financing costs). However, this cannot be practised in the long run. In the previous paragraph, we discussed the level of reference interest rates in individual economies, which were very low, especially in some economies. The average inflation rate in selected countries ranged from 1.2 to 2. %. These are not high values, but even this inflation rate could reduce interest rates to values where it was advantageous to buy more and more, and therefore there is a positive impact. A brief summary is appropriate at the end of the subchapter. Regardless of the size of Table 8. GMM results for large companies. Large companies Total debt DER(–1) ROE L2 SA NDTS GDP IR INF J-stat. Total debt CZ –28.7690a–205.3771a–22.5093b597.6002a0.4391 SK –0.0466a11.0080a–10.2300a333 5051a–32.1580a0.3821 SI 0.1642a–1.8513a7.5260b279.2540a43.0938a0.6131 RO 0.3749a2.1641a–1.7661b24,6207b0.7172 Short-term debt SK –6.8453a35.7621a186.1524a–316.3017a0.4773 PL –0.0789c15.1581c–291.4571b147.7846a0.9739 HU –44.8175a82.4348a36.9101a0.6449 AT 0.9670a3.9827a177.7547a1,520.1060a118.9317b0.0921 SI –0.0490a–2.0277a7.5292a175.9515a0.2687 BG –0.0340a–2.1036a–25.2762c18.9791b0.1139 Long-term debt CZ –23.8832a9.4005c–174.6068a317.0186a63,2566c0.1185 SK 0.0576a57.7827a144.8003a72.1349a0.2418 SI 0.6306a–1.9076a–3.5470a–23.7863a10,6419a0.3426 RO 0.8458a9.6773c–21.0019a0.4298 Source: authors’ calculations. Symbols a, b and c indicate significance at 1%, 5% and 10%. Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 80 the company, the main finding is that the level of indebtedness of companies in the Accommodation and food service activities sector is most influenced by the determinants of the external environment of companies. From the point of view of the level of coefficients, the influence of the reference interest rate clearly dominates, which has a negative effect on the indebtedness of Polish, Romanian, Bulgarian and Hungarian companies and a positive impact on the indebtedness of the remaining companies. However, the impact of economic development or inflation is not negligible. Compared to the least squares method, we obtained far more results from the original data, which at least partially showed us how the level of debt in selected companies could be affected. If we used only the least squares method, we would not come to any conclusions regarding the failure to meet a number of tests and assumptions. 5. Conclusion This research dealt with the financial structure of companies from the industry Accommodation and Food Service activities. The analysed companies are located in eight selected European economies. Specifically, these were the Czech Republic, Slovakia, Poland, Hungary, Austria, Bulgaria, Romania, and Slovenia. The aim of this research was to determine whether profitability, liquidity, asset structure, non-debt tax shield, the GDP growth rate, inflation rate, and reference interest rate affect the level of total, long-term and short-term debt. Within this goal, two research questions were formulated: ● Are there differences in impacts in terms of the different maturities of the used funding sources? ● What impact does the price of financial external sources have on the used sources of financing? A total of 23,991 companies were analysed, of which 22,973 are medium-sized and 1,018 are large and very large companies. The companies were analysed for the period 2010–2018. The least squares method and Generalized Method of Moments were used to determine the impacts of selected factors. It was a comparison of two regression analyses. The research attempted to compare these two methods with regard to the difficulty of verifying their results from the point of view of plausibility. The resulting least squares models must meet several basic assumptions and tests, while the GMM method only needs to perform a single test after analysis. Regarding the results of the comparison of the two selected methods, the modified regression analysis in the form of the Generalized Method of Moments is a far more suitable method than the least squares method. As stated in the theory, the GMM method has found its application mainly in the field of finance, which is clearly confirmed by this research. Compared to the least squares method, we obtained far more results from the original data, which at least partially showed us how the level of debt in selected companies could be affected. If we used only the least squares method, we would not come to any conclusions regarding the failure to meet a number of tests and assumptions. However, this is also a good finding, as it is so clear that this method is very unsuitable for corporate panel data analysis. The results of the GMM method showed a number of relationships and the effects of individual determinants on the level of debt of selected companies. Given the number of determinants, economies and endogenous variables, it is clear that the results are plentiful and cannot be summarised in a few sentences. However, the main finding of the research is that the level of indebtedness of selected companies is very significantly influenced by the determinants of the external environment of companies. From the point of view of the value of coefficients, the influence of the reference interest rate clearly dominates, while from the point of view of Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 81 the frequency of coefficients, the impact of the GDP growth rate is significant. However, the impact of the inflation rate is not negligible either. With regard to the mentioned results, only the effects of economic development and reference interest rates are summarised here. The effects of the interest rate vary from one economy to another, but it can be stated that the direction of the impact more or less follows the basic assumption – the higher the cost of acquiring debt financing, the less we will acquire it. Therefore, the reference interest rate has a positive impact on the level of indebtedness in economies that have been supported by central banks by keeping interest rates zero or very low for most of the period under review. These are the Czech Republic (average rate during the observed period 0.4%), Slovakia, Slovenia, and Austria (0.3%). On the other hand, in higher rate economies, the impact of the interest rate on the debt level was negative. These are Romania (3.6%), Hungary (3.1%), and Poland (2.5%). One exception is Bulgarian companies, which were found to have a negative impact, with an average reference interest rate of almost – 0.05%. This may be due to the fact that the Bulgarian economy was doing relatively well and at such low rates, companies did not want to go into debt and preferred to use their own resources to finance investment activities. This is a wise decision, because in the event of a crisis, companies will not be over-indebted and will not have to run into existential problems. The results for economic development were mixed, but at the same time the results for profitability were supported. For medium-sized Slovak, Slovenian, Austrian and Romanian companies and for large Czech and Bulgarian companies, a negative impact of GDP growth on the level of debt was found. The negative impact means that if these companies grow in profits (which is usually in times of economic prosperity), companies should prioritize rising profits as a source of funding, and debt levels should therefore decline. A positive impact has been found in the remaining economies, which means that, for example, in the case of economic growth, where companies tend to grow profits and thrive overall, the risk of bankruptcy is reduced and lenders are willing to provide additional funding. Both of these impacts are expected, as the selected economies performed well during the analysed period and there were no major economic problems that would hit the country hard. Acknowledgment This article was supported by SGS/16/2020 Influence of selected internal and macroeconomic determinants on financial structure of companies in selected countries of Central and Eastern Europe. References Acedo-Ramírez, M. A., Ruiz-Cabestre, F. J. (2014). Determinants of capital structure: United Kingdom versus continental European countries. Journal of International Financial Management & Accounting, 25(3), pp. 237–270. DOI:10.1111/jifm.12020. Antoniou, A., Guney, Y., Paudyal, K. (2008). The Determinants of Capital Structure: Capital MarketOriented versus Bank-Oriented Institutions. Journal of Financial and Quantitative Analysis, 43(1), pp. 59–92. DOI:10.1017/S0022109000002751. Arellano, M., Bond, S. (1991). Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations. The Review of Economic Studies, 58(2), pp. 277–297. Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 88 Hungarian large companies ROE L2 SA NDTS GDP IR INF DER DER_L DER_S ROE 1 –0.0243 –0.1275 0.0674 0.1012 –0.0794 –0.1180 –0.3604 –0.3905 0.3257 L2 –0.0243 1 –0.1082 –0.1808 –0.0108 0.0345 0.0058 0.0190 0.0199 –0.0069 SA –0.1275 –0.1082 1 –0.0004 –0.0417 0.0516 0.0179 0.0846 0.0960 –0.1385 NDTS 0.0674 –0.1808 –0.0004 1 –0.0434 0.0317 0.0228 –0.0775 –0.0737 –0.0735 GDP 0.1012 –0.0108 –0.0417 –0.0434 1 –0.7630 –0.6790 –0.0342 –0.0348 –0.0010 IR –0.0794 0.0345 0.0516 0.0317 –0.7630 1 0.7560 0.0406 0.0429 –0.0221 INF –0.1180 0.0058 0.0179 0.0228 –0.6790 0.7560 1 0.0340 0.0379 –0.0463 DER –0.3604 0.0190 0.0846 –0.0775 –0.0342 0.0406 0.0340 1 0.9977 0.2960 DER_L –0.3905 0.0199 0.0960 –0.0737 –0.0348 0.0429 0.0379 0.9977 1 0.2301 DER_S 0.3257 –0.0069 –0.1385 –0.0735 –0.0010 –0.0221 –0.0463 0.2960 0.2301 1 Austrian medium-sized companies ROE L2 SA NDTS GDP IR INF DER DER_L DER_S ROE 1 –0.0007 –0.0029 0.0332 –0.0036 –0.0097 –0.0108 0.0009 –0.0001 0.0269 L2 –0.0007 1 0.0211 –0.0053 –0.0192 0.0041 0.0091 –0.0007 –0.0007 –0.0007 SA –0.0029 0.0211 1 –0.0113 0.0143 –0.0242 –0.0020 0.0055 0.0054 0.0042 NDTS 0.0332 –0.0053 –0.0113 1 0.0032 –0.0034 0.0012 –0.0023 –0.0025 0.0034 GDP –0.0036 –0.0192 0.0143 0.0032 1 0.1403 0.2905 0.0019 0.0015 0.0108 IR –0.0097 0.0041 –0.0242 –0.0034 0.1403 1 0.6509 0.0111 0.0116 –0.0096 INF –0.0108 0.0091 –0.0020 0.0012 0.2905 0.6509 1 0.0134 0.0135 0.0034 DER 0.0009 –0.0007 0.0055 –0.0023 0.0019 0.0111 0.0134 1 0.9994 0.3982 DER_L –0.0001 –0.0007 0.0054 –0.0025 0.0015 0.0116 0.0135 0.9994 1 0.3659 DER_S 0.0269 –0.0007 0.0042 0.0034 0.0108 –0.0096 0.0034 0.3982 0.3659 1 Austrian large companies ROE L2 SA NDTS GDP IR INF DER DER_L DER_S ROE 1 0.0115 –0.1165 0.0112 0.0384 0.0469 0.0144 0.0731 0.3061 0.0165 L2 0.0115 1 –0.0767 0.0167 –0.0021 0.1143 0.0456 0.0036 –0.0056 0.0047 SA –0.1165 –0.0767 1 –0.0263 0.0016 –0.0374 –0.0191 0.0110 –0.0373 0.0203 NDTS 0.0112 0.0167 –0.0263 1 0.0403 –0.0284 –0.0042 0.0075 –0.0098 0.0104 GDP 0.0384 –0.0021 0.0016 0.0403 1 0.1403 0.2905 –0.0372 –0.0277 –0.0357 IR 0.0469 0.1143 –0.0374 –0.0284 0.1403 1 0.6509 –0.0394 –0.0153 –0.0408 INF 0.0144 0.0456 –0.0191 –0.0042 0.2905 0.6509 1 –0.0421 0.0098 –0.0492 DER 0.0731 0.0036 0.0110 0.0075 –0.0372 –0.0394 –0.0421 1 0.6230 0.9860 DER_L 0.3061 –0.0056 –0.0373 –0.0098 –0.0277 –0.0153 0.0098 0.6230 1 0.4838 DER_S 0.0165 0.0047 0.0203 0.0104 –0.0357 –0.0408 –0.0492 0.9860 0.4838 1 Slovenian medium-sized companies ROE L2 SA NDTS GDP IR INF DER DER_L DER_S ROE 1 –0.0009 –0.0451 –0.0099 –0.0004 0.0259 0.0046 –0.9501 –0.9582 –0.1174 L2 –0.0009 1 0.0188 –0.0046 0.0247 –0.0123 0.0054 0.0002 0.0003 –0.0017 SA –0.0451 0.0188 1 0.0151 0.0016 0.0101 0.0220 0.0313 0.0322 –0.0064 NDTS –0.0099 –0.0046 0.0151 1 0.0112 –0.0136 0.0019 0.0066 0.0067 0.0009 GDP –0.0004 0.0247 0.0016 0.0112 1 –0.6065 –0.4957 0.0046 0.0037 0.0161 Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 89 IR 0.0259 –0.0123 0.0101 –0.0136 –0.6065 1 0.6096 –0.0331 –0.0323 –0.0218 INF 0.0046 0.0054 0.0220 0.0019 –0.4957 0.6096 1 –0.0138 –0.0121 –0.0317 DER –0.9501 0.0002 0.0313 0.0066 0.0046 –0.0331 –0.0138 1 0.9982 0.2884 DER_L –0.9582 0.0003 0.0322 0.0067 0.0037 –0.0323 –0.0121 0.9982 1 0.2303 DER_S –0.1174 –0.0017 –0.0064 0.0009 0.0161 –0.0218 –0.0317 0.2884 0.2303 1 Slovenian large companies ROE L2 SA NDTS GDP IR INF DER DER_L DER_S ROE 1 0.1178 –0.1819 –0.6938 0.1063 0.0028 –0.0469 –0.2792 –0.3889 –0.0972 L2 0.1178 1 –0.4641 0.1823 0.0561 –0.1570 –0.0981 –0.0883 –0.0806 –0.0655 SA –0.1819 –0.4641 1 –0.0673 0.0157 –0.0410 –0.0338 –0.0810 0.1160 –0.2156 NDTS –0.6938 0.1823 –0.0673 1 –0.0769 –0.0387 –0.0481 0.1506 0.2001 0.0602 GDP 0.1063 0.0561 0.0157 –0.0769 1 –0.6065 –0.4957 –0.0085 –0.0337 0.0150 IR 0.0028 –0.1570 –0.0410 –0.0387 –0.6065 1 0.6096 0.0354 0.0996 –0.0288 INF –0.0469 –0.0981 –0.0338 –0.0481 –0.4957 0.6096 1 –0.0370 0.0260 –0.0763 DER –0.2792 –0.0883 –0.0810 0.1506 –0.0085 0.0354 –0.0370 1 0.7740 0.8549 DER_L –0.3889 –0.0806 0.1160 0.2001 –0.0337 0.0996 0.0260 0.7740 1 0.3332 DER_S –0.0972 –0.0655 –0.2156 0.0602 0.0150 –0.0288 –0.0763 0.8549 0.3332 1 Bulgarian medium-sized companies ROE L2 SA NDTS GDP IR INF DER DER_L DER_S ROE 1 –0.0046 –0.0222 0.3368 0.0120 –0.0115 –0.0041 0.0321 0.0545 –0.0124 L2 –0.0046 1 –0.0247 –0.0268 0.0016 0.0027 –0.0032 0.0056 0.0100 –0.0027 SA –0.0222 –0.0247 1 0.0569 –0.0542 0.0898 0.0452 0.0128 0.0157 0.0033 NDTS 0.3368 –0.0268 0.0569 1 0.0087 0.0048 0.0193 –0.0026 –0.0014 –0.0031 GDP 0.0120 0.0016 –0.0542 0.0087 1 –0.3285 –0.3007 0.0022 0.0020 0.0015 IR –0.0115 0.0027 0.0898 0.0048 –0.3285 1 0.5802 0.0096 0.0022 0.0157 INF –0.0041 –0.0032 0.0452 0.0193 –0.3007 0.5802 1 0.0114 0.0082 0.0109 DER 0.0321 0.0056 0.0128 –0.0026 0.0022 0.0096 0.0114 1 0.8770 0.7472 DER_L 0.0545 0.0100 0.0157 –0.0014 0.0020 0.0022 0.0082 0.8770 1 0.3360 DER_S –0.0124 –0.0027 0.0033 –0.0031 0.0015 0.0157 0.0109 0.7472 0.3360 1 Bulgarian large companies ROE L2 SA NDTS GDP IR INF DER DER_L DER_S ROE 1 –0.0114 –0.1491 –0.0777 –0.0205 –0.0053 0.0336 0.3591 0.0599 0.7137 L2 –0.0114 1 0.0344 –0.0556 –0.0689 0.0069 0.0135 –0.0165 –0.0116 –0.0164 SA –0.1491 0.0344 1 0.0743 –0.0304 0.0699 0.0370 –0.0176 0.0534 –0.1389 NDTS –0.0777 –0.0556 0.0743 1 0.0130 0.0572 0.0290 –0.0912 –0.0686 –0.0829 GDP –0.0205 –0.0689 –0.0304 0.0130 1 –0.3285 –0.3007 0.0024 0.0254 –0.0414 IR –0.0053 0.0069 0.0699 0.0572 –0.3285 1 0.5802 –0.0568 –0.0768 0.0113 INF 0.0336 0.0135 0.0370 0.0290 –0.3007 0.5802 1 –0.0180 –0.0539 0.0580 DER 0.3591 –0.0165 –0.0176 –0.0912 0.0024 –0.0568 –0.0180 1 0.9059 0.6252 DER_L 0.0599 –0.0116 0.0534 –0.0686 0.0254 –0.0768 –0.0539 0.9059 1 0.2360 DER_S 0.7137 –0.0164 –0.1389 –0.0829 –0.0414 0.0113 0.0580 0.6252 0.2360 1 Nicole Škuláňová, Veronika Šudová: Comparison of Methods of Evaluation of the Financial Structure in Selected Industry and Countries 90 Romanian medium-sized companies ROE L2 SA NDTS GDP IR INF DER DER_L DER_S ROE 1 –0.0003 –0.0182 0.0030 0.0163 –0.0115 –0.0086 0.2248 0.1713 0.1838 L2 –0.0003 1 –0.0044 –0.0048 0.0094 –0.0051 –0.0002 0.0053 0.0125 –0.0001 SA –0.0182 –0.0044 1 0.1383 –0.0619 0.0629 0.0560 –0.0036 0.0158 –0.0126 NDTS 0.0030 –0.0048 0.1383 1 –0.0137 0.0224 0.0205 –0.0003 –0.0034 0.0014 GDP 0.0163 0.0094 –0.0619 –0.0137 1 –0.8043 –0.6293 0.0028 –0.0001 0.0035 IR –0.0115 –0.0051 0.0629 0.0224 –0.8043 1 0.8229 0.0047 0.0063 0.0025 INF –0.0086 –0.0002 0.0560 0.0205 –0.6293 0.8229 1 0.0057 0.0076 0.0029 DER 0.2248 0.0053 –0.0036 –0.0003 0.0028 0.0047 0.0057 1 0.5911 0.9071 DER_L 0.1713 0.0125 0.0158 –0.0034 –0.0001 0.0063 0.0076 0.5911 1 0.1966 DER_S 0.1838 –0.0001 –0.0126 0.0014 0.0035 0.0025 0.0029 0.9071 0.1966 1 Romanian large companies ROE L2 SA NDTS GDP IR INF DER DER_L DER_S ROE 1 –0.0123 –0.0926 0.0054 –0.0167 –0.0137 –0.0172 0.4019 0.0361 0.3041 L2 –0.0123 1 –0.1636 –0.0495 –0.0078 –0.0239 –0.0273 –0.0084 –0.0036 –0.0105 SA –0.0926 –0.1636 1 –0.0588 0.0006 –0.0187 –0.0072 –0.0331 –0.0561 –0.0212 NDTS 0.0054 –0.0495 –0.0588 1 –0.0262 –0.0001 –0.0125 0.0005 0.0132 0.0052 GDP –0.0167 –0.0078 0.0006 –0.0262 1 –0.8043 –0.6293 –0.0487 –0.0231 –0.0585 IR –0.0137 –0.0239 –0.0187 –0.0001 –0.8043 1 0.8229 0.0720 0.0562 0.0699 INF –0.0172 –0.0273 –0.0072 –0.0125 –0.6293 0.8229 1 0.0777 0.0556 0.0793 DER 0.4019 –0.0084 –0.0331 0.0005 –0.0487 0.0720 0.0777 1 0.3084 0.9178 DER_L 0.0361 –0.0036 –0.0561 0.0132 –0.0231 0.0562 0.0556 0.3084 1 0.2166 DER_S 0.3041 –0.0105 –0.0212 0.0052 –0.0585 0.0699 0.0793 0.9178 0.2166 1