**4. Results**

The following variables were used to construct a taxonomic measure of development to describe the level of RES use in selected countries:

*X*1—share of electricity generation from renewable sources in total electricity generation (in %);

*<sup>X</sup>*2—electricity generation from water energy (in GWh per capita);

*<sup>X</sup>*3—electricity generation from wind energy (in GWh per capita);

*<sup>X</sup>*4—electricity generation from solar energy (in GWh per capita);

*<sup>X</sup>*5—electricity generation from biomass (in GWh per capita).

These variables were selected based on substantive criteria and constitute a set of potential diagnostic variables.

In the next step, formal (statistical) criteria were checked, i.e., potential diagnostic variables were analysed in terms of the degree of differentiation and correlation between particular variables. In order to assess the degree of variables' diversity, the coefficient of variation ( *Vj*) was calculated [13]. In this article, a limit value of coefficient of variation was assumed at the level of 10%. In the case when the condition is met:

$$0 \le V\_{\bar{j}} \le 10\% \ (j = 1, 2, \dots, p)\_{\prime}$$

where:

*p*—number of potential diagnostic variables.

The variable *Xj* is eliminated from the set of diagnostic variables. In this article, all proposed diagnostic variables were found to be sufficiently diverse.

Next, Pearson's linear correlation coefficients were calculated between each pair of potential diagnostic variables.

Finally, all potential diagnostic variables were used to construct a taxonomic measure of development. The nature of these variables (i.e., the impact on the level of RES use for electricity generation) was then determined. All variables adopted for the study turned out to be stimulants, which means that the higher the values of these variables, the higher the level of development of RES for electricity generation in each country.

In the next step, the values of diagnostic variables were normalised using the standardisation method, and the so-called development pattern was determined, i.e., the object (here: country) with the best values of individual diagnostic variables. Then, the Euclidean distances of individual countries from the development pattern were calculated. The smaller the distance from the development pattern, the higher the level of renewable energy use in a given country.

The last stage consisted of calculating a synthetic measure (the so-called development measure), on the basis of which it is possible to rank the 28 countries surveyed in terms of the level of a composite phenomenon (i.e., the level of RES use for electricity generation). The values of the development indicator are standardised in the interval [0; 1]. The closer the value of this measure is to unity, the higher the level of the complex phenomenon (here: the level of development of renewable energy). Figures 1–4 show the values of the development ratio for the studied countries in selected years.

**Figure 1.** Development taxonomic ratio values (TMR) for the examined countries in 2004. Source: author's own elaboration.

**Figure 2.** Development taxonomic ratio values (RMT) for the examined countries in 2009. Source: author's own elaboration.

**Figure 3.** Development taxonomic ratio values (RMT) for the examined countries in 2014. Source: author's own elaboration.

**Figure 4.** Development taxonomic ratio values (RMT) for the examined countries in 2019. Source: author's own elaboration.

Table 1 presents the ranking of the surveyed countries by level of RES use for electricity production for four selected years (in five-year intervals). Table 1 and Figures 1–4 intentionally present the same data in two graphical forms, as they have varying usability for the reader.

**Table 1.** The ranking of countries by level of RES use for electricity production for four selected years (alphabetical order).



**Table 1.** *Cont.*

Source: author's own elaboration.

In each of the years surveyed, Sweden topped the ranking in terms of the level of RES use for electricity generation. Malta was ranked last in 2004, 2009, and 2014, but Poland was ranked last in the last year examined.

A synthetic measure that takes into account both sources of electricity production directly dependent on climate (hydro, solar, and wind) and biocomponents (Table 2 column A) has identified an interesting group of seven countries at the top of the ranking for 2019. Table 2 should be viewed in conjunction with Table 3. When such a measure is reduced by energy production from biocomponents, the same countries appear in the top seven of the ranking, albeit in a different order (Table 2 column B). This shows a good diversification of renewable sources for electricity generation in those countries that (apart from Finland) are in the middle of the ranking in terms of the share of biocomponents in electricity generation from renewable sources. However, in the case of Estonia, the inclusion of biocomponents as a renewable source of electricity generation (the share of 60.6%) has resulted in the country moving up in the 2019 ranking from 25th to 12th place.

The group of countries that both measures present as the best performers in the development of renewable electricity sources includes Sweden, Finland, Austria, Denmark, Germany, Luxembourg, and Portugal. It should be noted that, apart from Portugal, the other six countries in this group are rich countries in northern Europe, and all of them belong to the group of the so-called old EU member states (adopted before 2004). This group includes both net electricity exporting countries (Sweden, Germany) and significant net importers (Finland and especially Luxembourg—Table 2 column C).


**Table 2.** Taxonomic measures of development of renewable electricity generation in EU countries and net electricity exports—country rankings for 2019.

Source: own elaboration based on Eurostat data. O = old EU countries (pre-2004). N = new EU countries (2004 and later). \* Taxonomic measures of development of electricity generation from renewable sources: A.—solar, wind, hydro, and bio; B.—solar, wind, hydro only. \*\* C.—net electricity exports—as the difference between the percentage ratio of electricity exports to total generation (GWh) and the percentage ratio of electricity imports to total generation (GWh) regardless of source.

> At the same time, in the case of both versions of the measure, the penultimate and last place goes to Hungary and Poland, respectively.

> When interpreting the results, it has to be remembered that the sizes of electricity production (in GWh) from different renewable sources included in the synthetic measures were calculated per capita, and nominal production volumes are not used here. As a result, the construction of the measure does not directly reflect the size of the population and the size of the economic potential of the countries studied.

> Interestingly, five countries out of the seven identified by the synthetic measure are also among the top seven EU countries for 2019 in terms of the percentage of renewable electricity production in total electricity production (Luxembourg, Denmark, Austria, Sweden, and Portugal—Table 3 column A) and in terms of the percentage ratio of renewable electricity production to final consumption (Austria, Sweden, Denmark, Portugal, and Germany—Table 3 column B).

> It is characteristic that the last places in the ranking according to synthetic measures (Table 2 column A and column B) are occupied by a compact group of the so-called new Member States (admitted to the EU in 2004 and later). The countries, whose capacity utilisation of electricity generation from renewable energy sources is considered the weakest by the synthetic measures, besides the already mentioned Hungary and Poland, are also the Czech Republic, Cyprus, Malta, Romania, Bulgaria, and Lithuania.


**Table 3.** Share of renewable sources in electricity production and consumption and total energy consumption (in %) in European Union countries—country rankings for 2019.

Source: own elaboration based on Eurostat data. O = old EU countries (pre-2004). N = new EU countries (2004 and onwards). \* A.— Percentage share of renewable sources in electricity generation (GWh). \*\* B.—Percentage share of renewable sources in electricity consumption (GWh). \*\*\* C.—Percentage share of renewable sources in total energy consumption (GWh).

> This dichotomy of 'old versus new EU Member States' is more strongly accentuated by the synthetic measures than is apparent from the percentage share of renewable electricity generation in total electricity generation (Table 3 column A) or in final consumption (Table 3 column B) in the countries under study.

> The thesis on the influence of the wealth of the surveyed countries on their current development of the use of renewable energy sources for electricity production will be verified later in this article by means of a panel model.

> Of note is the low ranking of France (the 16th and 20th position for 2019) and the Netherlands (the 17th and 17th position for 2019) according to both synthetic measures (Table 2 column A and column B).

> An important issue is the growing number of EU countries dependent on electricity imports. In 2004, it was 13 countries, in 2009, already 16 countries, in 2014, the number increased to 17 countries, and in 2019, it was as high as 20 countries. This means that seven countries have lost their energy independence in this way. The calculation uses the difference between the percentage ratio of electricity exports to total electricity production and the percentage ratio of electricity imports to total electricity generation in individual countries (Table 2 column C).

> Interestingly, the Czech Republic and Bulgaria, which are ranked low in synthetic measures, reflecting the possibility of using renewable energy sources to produce electricity, are also countries that have a positive balance of electricity exports and imports, i.e., produce more electricity than they need domestically. This indicates the generation of surplus electricity for export through intensive production of electricity from non-renewable sources.

In the following section of this article, a panel model defined by formula (3) is used to describe the relationship between the share of RES power generation in total power generation and per capita power generation from water, wind, solar, and biomass energy in all EU countries. Data were collected for all 28 EU countries and cover the 2004–2019 period. The source of data was the Eurostat database.

The theoretical model can be written as follows:

$$Y\_{jt} = \alpha\_0 + \alpha\_1 X\_{1jt} + \alpha\_2 X\_{2jt} + \upsilon\_{jt} \tag{3}$$

$$
\omega\_{jt} = \varepsilon\_l + \mathfrak{u}\_j + \varepsilon\_{jt} \tag{4}
$$

Table 4 presents a description of the individual variables.


**Table 4.** The description of variables.

Source: elaborated by the authors.

The level of GDP per capita was used as a potential factor influencing the level of electricity production from renewable sources, assuming that richer countries with a higher level of development care more about the environment than poorer ones and, thus, invest in renewable energy sources. There is no measure of a country's wealth that is not questioned. The Stiglitz commission's report to then-French President Nicolas Sarkozy is an example of an extensive discussion on the subject. Alternative measures such as ISEW or HDI use GDP or its derivatives (GNI) in their construction. However, perhaps had there been more awareness of the limitations of standard metrics, such as GDP, there would have been less euphoria over economic performance in the years prior to the crisis; metrics that incorporated assessments of sustainability (e.g., increasing indebtedness) would have provided a more cautious view of economic performance. However, many countries lack a timely and complete set of wealth accounts—the 'balance sheets' of the economy—that could give a comprehensive picture of assets, debts, and liabilities of the main actors in the economy [51].

Panel data models (1) were estimated using the GRETL software (GNU Regression Econometrics Time-Series Library). In turn, for the estimation of panel data models, the following were used:


The KMNK estimator is used when all objects covered by the study are homogeneous, and the differences between the empirical and theoretical values of the explained variable Y result only from the random component [52].

The FE and RE estimators are used in the case of sample heterogeneity. Individual effects are the source of sample non-homogeneity. The FE estimator assumes that the individual effects are non-random and can be estimated. In the case of the RE estimator, it is assumed that the individual effects are random and that they are part of the random component. In this case, it is not possible to estimate the value of individual effects; it is only possible to estimate their dispersion [52].

When selecting the type of panel model (simple model, i.e., without individual-dual effects or models with unidirectional individual effects, i.e., FEM—fixed effect model or REM—random effect model) the following tests are used: Wald test, Breusch–Pagan test, and Hausman test. The aforementioned tests allow to assess the correctness of the estimated model. These tests are discussed in many studies in the field of econometrics [53–55].The choice of the estimation method was based on the decision procedure presented in the econometrics literature [52–57]. First, a simple panel model (without individual effects) was estimated using the classical least squares method, and diagnostic tests of the model were conducted.

Table 5 presents the results of Wald, Breusch–Pagan, and Hausman tests, based on which a decision is made on the choice of an appropriate model. These tests allow for the verification of the assumptions about the correctness of panel model estimation.


**Table 5.** The result of the Wald, Breusch–Pagan, and Hausman tests.

\* The adopted level of significance is 0.05 (i.e., *α* = 0.05). Source: author's own calculation.

> Analysing the results of the Wald test, it can be stated that a fixed effects model (FEM) is the correct model in all cases for describing the relation of the share of RES production in total electricity production and of production from water, wind, solar, and biomass and GDP per capita and governmen<sup>t</sup> expenditure on energy per capita.

> The results of the Breusch–Pagan test in each case indicate the random effects model (REM) as the better model. Finally, the results of the Hausman test allow for the conclusion, with the risk of error at the level of 5% (α = 0.05), that in the case of power generation from water energy and biogas energy per capita, the model with random individual effects is appropriate for describing the examined relationship, while in the remaining cases, the model with fixed individual effects (FEM). However, further analysis confirmed the occurrence of heteroskedasticity of the random component in the models *Y1*, *Y*3, and *Y4*. To address this shortcoming, the weighted least squares (WLS) method was used to estimate the parameters of share of electricity production from RES in a total electricity production model, electricity production from wind per capita, and electricity production from solar per capita. In other cases—electricity production from water per capita and electricity production from biogas per capita—the random effects models were estimated.

Table 6 presents the estimation results of the above models.


**Table 6.** Results of model estimation.

Source: author's own calculation. a \*\*\* The statistically significant variable at the level of 1%; \*\* at the level of 5%; \* at the level of 10%.

> In the case of the model describing the share of RES energy generation in total energy generation, the model describing solar power production per capita and the model describing biogas power production per capita, both potential explanatory variables proved to be statistically significant. In the case of the model describing electricity production

from wind, only GDP per capita was statistically significant. In the case of the model describing electricity production from water, only public spending by countries on energy as a percentage of GDP was statistically significant at the 0.1 significance level. This proves large public expenditure on the development of solar electricity production in EU countries, which is significantly greater than for other RES sources.

At the same time, it confirms the significance of economic potential and wealth of countries for better use of RES for electricity generation, which was initially indicated by the analysis of country rankings according to the taxonomic measure of development.

The results obtained allow us to state that in the case of the model describing the share of electricity generation from RES in total electricity generation, two explanatory variables, i.e., GDP per capita and public expenditure on energy as percent of GDP, positively influence the explained variable (i.e., the share of renewable energy generation in total energy generation). The evaluation of the parameter with the independent variable *<sup>X</sup>*1*jt* (0.000486) should be interpreted as follows: if GDP per capita increases by one percentage point, the share of electricity generation from RES in total electricity generation will increase on this account by about 0.000486 on average, assuming constancy of the other variables. The interpretation of the evaluation of the parameter with the explanatory variable *<sup>X</sup>*2*jt* (10.7911) should be as follows: if public expenditure on energy per capita increases by 1 percent point, then the share of renewable energy production in total energy production will increase for this reason by approximately 10.7911 on average, under the assumption of constancy of the other variables.
