The research was conducted in two basic stages. An overall assessment of the methodological correctness of the selected group of indicators for monitoring the success of decarbonisation was made in the first stage of the research, while the differences between EU countries and non-EU countries were assessed in the second stage.
First research stage – basic methodological assessment
The correlations between individual indicators were determined in the first phase of the paper, by applying Pearson's correlation coefficient, and the results are shown in Table 1.
Table 1 Correlations of decarbonisation indicators - Pearson's correlation coefficient
|
CO2
per
capita
|
CO2
Emi-
ssion
|
Year to year CO2
|
Cumu-
lative
CO2
|
Consu-
mption CO2
|
Share
of
CO2
|
Cement
CO2
|
Coal
CO2
|
Flaring
CO2
|
Gas
CO2
|
Oil
CO2
|
Other CO2
|
CO2 per capita
|
|
1
|
.530**
|
-.129*
|
.491**
|
.265**
|
.528**
|
.521**
|
.546**
|
.401**
|
.505**
|
.521**
|
.428**
|
CO2 emission
|
|
.530**
|
1
|
-.211**
|
.958**
|
.993**
|
.981**
|
.979**
|
.987**
|
.926**
|
.989**
|
.975**
|
.972**
|
Year to year CO2
|
|
-.129*
|
-.211**
|
1
|
-.137*
|
-.521**
|
-.280**
|
-.176**
|
-.223**
|
-.046
|
-.185**
|
-.238**
|
-.259**
|
Cumulative CO2
|
|
.491**
|
.958**
|
-.137*
|
1
|
.917**
|
.896**
|
.942**
|
.940**
|
.953**
|
.979**
|
.881**
|
.871**
|
Consumption CO2
|
|
.265**
|
.993**
|
-.521**
|
.917**
|
1
|
.966**
|
.954**
|
.987**
|
.712**
|
.978**
|
.893**
|
.984**
|
Share of CO2
|
|
.528**
|
.981**
|
-.280**
|
.896**
|
.966**
|
1
|
.942**
|
.981**
|
.843**
|
.952**
|
.980**
|
.978**
|
Cement CO2
|
|
.521**
|
.979**
|
-.176**
|
.942**
|
.954**
|
.942**
|
1
|
.959**
|
.948**
|
.969**
|
.956**
|
.959**
|
Coal CO2
|
|
.546**
|
.987**
|
-.223**
|
.940**
|
.987**
|
.981**
|
.959**
|
1
|
.877**
|
.963**
|
.959**
|
.967**
|
Flaring CO2
|
|
.401**
|
.926**
|
-.046
|
.953**
|
.712**
|
.843**
|
.948**
|
.877**
|
1
|
.956**
|
.861**
|
.856**
|
Gas CO2
|
|
.505**
|
.989**
|
-.185**
|
.979**
|
.978**
|
.952**
|
.969**
|
.963**
|
.956**
|
1
|
.938**
|
.929**
|
Oil CO2
|
|
.521**
|
.975**
|
-.238**
|
.881**
|
.893**
|
.980**
|
.956**
|
.959**
|
.861**
|
.938**
|
1
|
.991**
|
Other CO2
|
|
.428**
|
.972**
|
-.259**
|
.871**
|
.984**
|
.978**
|
.959**
|
.967**
|
.856**
|
.929**
|
.991**
|
1
|
** Correlation is significant at the 0.01 level (2-tailed).
* Correlation is significant at the 0.05 level (2-tailed).
The results in Table 1 show that all indicators have high correlations (at the level of p<0.05 or p<0.01). It is already evident at this point that the monitoring model that would include all the above indicators is not methodologically correct, because high correlations clearly indicate a poorly established model, so further testing is required.
CO2 per unit of GDP has been identified as one of the basic indicators of the success of the energy transition, therefore additional attention is paid to observing this indicator in relation to others. This indicator is positively correlated with all other indicators, except the indicator Year to year CO2 change, where the correlation is negative, while there is no statistically significant correlation with Flaring CO2.
Given the above, two opposing conclusions can be drawn. First, the indicator CO2 per unit of GDP can be used as the only indicator to assess the success of energy transition or this indicator does not contribute to the tested model. In order to test the hypothesis on these two ways of making a conclusion, an exploratory method was used to estimate the correlation of CO2 per unit of GDP with other indicators, and the results are shown in Table 2.
Table 2 Correlations between CO2 per unit of GDP and selected decarbonisation indicators - Pearson's correlation coefficient
|
CO2
per
capita
|
CO2
Emi-
ssion
|
Year to year CO2
|
Cumu-
lative
CO2
|
Consu-
mption CO2
|
Share
of
CO2
|
Cement
CO2
|
Coal
CO2
|
Flaring
CO2
|
Gas
CO2
|
Oil
CO2
|
Other CO2
|
CO2
per unit of GDP
|
|
.387**
|
.187**
|
-.178**
|
.163**
|
.502**
|
.223**
|
.139*
|
.244**
|
.076
|
.160**
|
.168**
|
.233**
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
** Correlation is significant at the 0.01 level (2-tailed).
* Correlation is significant at the 0.05 level (2-tailed).
Table 3 shows a regression model, which aims to assess the share of individual indicators in explaining the presented model of monitoring the success of decarbonisation.
Table 3 Summarised results of the regression model of selected decarbonisation indicators
Model
|
R
|
R Square
|
Adjusted R Square
|
R Square Change
|
1
|
.993a
|
.987
|
.987
|
.987
|
2
|
.999b
|
.997
|
.997
|
.010
|
3
|
.999c
|
.999
|
.999
|
.001
|
4
|
1.000d
|
.999
|
.999
|
.001
|
5
|
1.000e
|
.999
|
.999
|
.000
|
6
|
1.000f
|
1.000
|
1.000
|
.000
|
7
|
1.000g
|
1.000
|
1.000
|
.000
|
8
|
1.000h
|
1.000
|
1.000
|
.000
|
9
|
1.000i
|
1.000
|
1.000
|
.000
|
10
|
1.000j
|
1.000
|
1.000
|
.000
|
11
|
1.000k
|
1.000
|
1.000
|
.000
|
12
|
1.000l
|
1.000
|
1.000
|
.000
|
a. Predictors: (Constant), Consumption_ CO2
|
b. Predictors: (Constant), Consumption_ CO2, Share_of_ CO2
|
c. Predictors: (Constant), Consumption_ CO2, Share_of_ CO2, Flaring_ CO2
|
d. Predictors: (Constant), Consumption_ CO2, Share_of_ CO2, Flaring_ CO2, Other_industry_ CO2
|
e. Predictors: (Constant), Consumption_ CO2, Share_of_ CO2, Flaring_ CO2, Other_industry_ CO2, Coal_ CO2
|
f. Predictors: (Constant), Consumption_ CO2, Share_of_ CO2, Flaring_ CO2, Other_industry_ CO2, Coal_ CO2, Gas_ CO2
|
g. Predictors: (Constant), Consumption_ CO2, Share_of_ CO2, Flaring_ CO2, Other_industry_ CO2, Coal_ CO2, Gas_ CO2, Oil_ CO2
|
h. Predictors: (Constant), Consumption_ CO2, Share_of_ CO2, Other_industry_ CO2, Coal_ CO2, Gas_ CO2, Oil_ CO2
|
i. Predictors: (Constant), Consumption_ CO2, Share_of_ CO2, Other_industry_ CO2, Coal_ CO2, Gas_ CO2, Oil_ CO2, Cement_ CO2
|
j. Predictors: (Constant), Share_of_ CO2, Other_industry_ CO2, Coal_ CO2, Gas_ CO2, Oil_ CO2, Cement_ CO2
|
k. Predictors: (Constant), Share_of_ CO2, Other_industry_ CO2, Coal_ CO2, Gas_ CO2, Oil_ CO2, Cement_ CO2, Flaring_ CO2
|
l. Predictors: (Constant), Other_industry_ CO2, Coal_ CO2, Gas_ CO2, Oil_ CO2, Cement_ CO2, Flaring_ CO2
|
The results in Table 3 show that 99.99% of the decarbonisation success can be explained only by the indicator Consumption based CO2. By adding the indicator Share of CO2, the model is completely explained (99.99%). All additional indicators, CO2 per unit of GDP as well, that are added to the model do not contribute to further clarification of the decarbonisation success.
To further confirm that decarbonisation success cannot be explained by the indicator CO2 per unit of GDP, these two indicators were added to the regression linear model as a criterion and predictor variable (Table 4.).
Table 4 CO2 unit of GDP (a predictor variable) and Cumulative CO2 emission (a criterion variable) – regression model
Model
|
R
|
R Square
|
Adjusted R Square
|
|
|
|
|
1
|
.187a
|
.035
|
.031
|
|
|
a. Predictors: (Constant), CO2_per_GDP
The regression analysis results, in Table 4, show that CO2 per unit of GDP manages to explain only 3% of the variance of Cumulative CO2 emission, which is absolutely insufficient.
In the next step (Table 5), CO2 emission per capita was used as a criterion variable, in order to estimate how much of variance of this dependent variable can be explained by CO2 per unit of GD.
Table 5 CO2 unit of GDP (a predictor variable) and Cumulative CO2 emission per capita (a criterion variable) – regression model
Model
|
R
|
R Square
|
Adjusted R Square
|
|
|
|
|
1
|
.387a
|
.150
|
.147
|
|
|
a. Predictors: (Constant), CO2_per_GDP
As it can be seen in Table 5, the analysis shows that only 15% of the CO2 per capita emission variance can be explained by CO2 unit of GDP, which is another confirmation that CO2 unit of GDP should not be taken as a sufficient or even necessary indicator for drawing conclusions about the success of decarbonisation.
The analysis so far has not revealed whether the exploratory method can determine how the indicators should be grouped, in order to find a way to better define CO2 unit of GDP and use it in further analysis. Therefore, the next phase of the research included the implementation of the Principal Component Analysis, which requires the following assumptions:
- All variables in the model must have an interval level of measurement;
- The linearity of the variables used in the analysis was previously confirmed through the Pearson correlation coefficient of the variables (Table 1).
- Using the Kaiser-Meyer-Olkin test (for checking the adequacy of the sample), it was determined that the required minimum value of 0.6 was met (Table 6).
- Using Bartlett's Test of Sphericity, it was confirmed that the data in the correlation matrix differ from zero.
Table 6 Kaiser-Meyer-Olkin test for checking the sample adequacy and Bartlett's Test of Sphericity of data
Kaiser-Meyer-Olkin Measure of Sampling Adequacy.
|
.705
|
Bartlett's Test of Sphericity
|
Approx. Chi-Square
|
5402.513
|
df
|
136
|
Sig.
|
.000
|
The first insight into the values of the examined indicators is presented in the table with communalities (Table 7) which shows how much of variance of each individual parameter can be explained by the selected factors. All parameters are well represented in the selected factors and there is no need here to exclude any of these indicators (all values obtained are > 0.3).
Table 7 Communalities
|
Initial
|
Extraction
|
CO2 _per_capita
|
1.000
|
.639
|
CO2 _emission
|
1.000
|
.994
|
Year_to_year_ CO2
|
1.000
|
.793
|
Cumulative_ CO2 _emission
|
1.000
|
.877
|
Consumption_ CO2
|
1.000
|
.991
|
Share_of_ CO2
|
1.000
|
.960
|
Cement_ CO2
|
1.000
|
.900
|
Coal_ CO2
|
1.000
|
.986
|
Flaring_ CO2
|
1.000
|
.963
|
Gas_ CO2
|
1.000
|
.978
|
Oil_ CO2
|
1.000
|
.845
|
Other_industry_ CO2
|
1.000
|
.969
|
GDP_Change
|
1.000
|
.792
|
GDP_per_capita
|
1.000
|
.881
|
REC
|
1.000
|
.917
|
REO
|
1.000
|
.884
|
CO2 _per_GDP
|
1.000
|
.774
|
Extraction Method: Principal Component Analysis.
|
Based on the Eigenvalues, the proposed model has 3 separate factors (the Eigenvalue of which is above 1). Auxiliary (GDP-based) indicators are also included in this phase of the research, and the results are shown in Table 8.
Table 8 Eigenvalues
Component
|
Initial Eigenvalues
|
Extraction Sums of Squared Loadings
|
Total
|
% of
Variance
|
Cumulative
%
|
Total
|
% of Variance
|
Cumulative %
|
1
|
11.667
|
68.627
|
68.627
|
11.667
|
68.627
|
68.627
|
2
|
2.237
|
13.161
|
81.788
|
2.237
|
13.161
|
81.788
|
3
|
1.241
|
7.302
|
89.090
|
1.241
|
7.302
|
89.090
|
4
|
.732
|
4.306
|
93.396
|
|
|
|
5
|
.376
|
2.210
|
95.606
|
|
|
|
6
|
.323
|
1.902
|
97.507
|
|
|
|
7
|
.200
|
1.178
|
98.685
|
|
|
|
8
|
.127
|
.749
|
99.434
|
|
|
|
9
|
.039
|
.229
|
99.663
|
|
|
|
10
|
.021
|
.123
|
99.785
|
|
|
|
11
|
.014
|
.080
|
99.865
|
|
|
|
12
|
.011
|
.062
|
99.928
|
|
|
|
13
|
.008
|
.048
|
99.975
|
|
|
|
14
|
.002
|
.014
|
99.989
|
|
|
|
15
|
.001
|
.007
|
99.996
|
|
|
|
16
|
.001
|
.004
|
100.000
|
|
|
|
17
|
|
|
100.000
|
|
|
|
The main goal of this analysis is to define the number of components that manage to explain most of the model variance, which would allow the monitoring model to be simplified and made more efficient and accurate. In this case, it is found that three components explain 89% of the variance of the entire model. However, it is necessary to determine which indicators overlap with them - which would indicate a methodological problem. For this purpose, factor rotation was used, by applying the Direct Oblimin method (oblique rotation, due to the determined correlation between the obtained factors) from which the Pattern Matrix is derived (Table 9).
Table 9 Pattern matrix with factor overlap (selected SE European countries)
|
Component
|
1
|
2
|
3
|
Flaring_ CO2
|
1.024
|
|
|
Cumulative_ CO2 _emission
|
.991
|
|
|
Gas_ CO2
|
.988
|
|
|
Consumption_ CO2
|
.985
|
|
|
Other_industry_ CO2
|
.973
|
|
|
Coal_ CO2
|
.963
|
|
|
CO2 _emission
|
.962
|
|
|
Cement_ CO2
|
.927
|
|
|
Share_of_ CO2
|
.896
|
|
|
Oil_ CO2
|
.791
|
|
|
REC
|
-.779
|
|
|
CO2 _per_GDP
|
.657
|
-.343
|
|
GDP_per_capita
|
|
.924
|
|
REO
|
|
.884
|
|
CO2 _per_capita
|
|
-.790
|
|
GDP_Change
|
|
|
.849
|
Year_to_year_ CO2
|
|
|
.784
|
Extraction Method: Principal Component Analysis.
Rotation Method: Oblimin with Kaiser Normalization.a
|
a. Rotation converged in 4 iterations.
|
The results in Table 9 clearly show that the principle of simple structure is almost completely satisfied, since three groups of components are defined. The first component consists of indicators related to the type of energy product or industry that is a source of CO2, but it should be emphasized that high overlaps exist within the first component. The second component comprises two GDP-related indicators, Renewable electricity output and CO2 per capita. Component no.3 is the smallest (it contains only GDP change and Year to year CO2) but is the most methodologically correct.
Furthermore, the indicators not contained in any other component can be obviously found in the defined three components of the model. The only, and at the same time the most problematic indicator in this model, is CO2 per unit of GDP, which is in a high positive correlation with component 1 (.657), but at the same time is in a high negative correlation with component 2 (-.343).
Given the performed analyses, the indicator CO2 per unit of GDP is clearly insufficient to explain the overall decarbonisation, and its use, in general, can be considered questionable in the process of monitoring the decarbonisation success – it is insufficiently clearly defined, it is contained in two components of the model, and there is also a high degree of overlap with other indicators.
Considering all CO2-related parameters, CO2 emissions proved to be the most useful by type of energy product, as well as Cumulative CO2 emissions (which was used as a criterion variable in previous models).
Second research stage – comparison between EU and non-EU countries
Further analysis considers differences that may exist in the values of indicators and possible model of decarbonisation in selected countries, so that EU and non-EU countries are analysed separately. To this end, Pattern Matrices have been developed, which show the number of components and the overlap of individual indicators. The results of this part of the analysis for the countries that joined the European Union (Estonia, Latvia, and Lithuania) are shown in Table 10.
Table 10 Pattern matrix with factor overlap (selected SE European countries – EU countries)
|
Component
|
1
|
2
|
3
|
4
|
GDP_per_capita
|
-.977
|
|
|
|
Share_of_ CO2
|
.968
|
|
|
|
REC
|
-.961
|
|
|
|
REO
|
-.874
|
|
|
|
CO2 _emission
|
.796
|
|
|
|
Cumulative_ CO2 _emission
|
.685
|
|
|
|
Consumption_ CO2
|
.660
|
.479
|
|
|
Gas_ CO2
|
|
.961
|
|
|
Oil_ CO2
|
|
.921
|
|
|
Flaring_ CO2
|
|
.833
|
|
|
Coal_ CO2
|
.485
|
-.766
|
|
|
CO2 _per_capita
|
.485
|
-.731
|
|
|
CO2 _per_GDP
|
.622
|
-.704
|
|
|
GDP_Change
|
|
|
.842
|
|
Year_to_year_ CO2
|
|
|
.831
|
|
Cement_ CO2
|
|
|
|
.876
|
Other_industry_ CO2
|
|
|
|
.398
|
Extraction Method: Principal Component Analysis.
Rotation Method: Oblimin with Kaiser Normalization.a,b
|
a. Rotation converged in 7 iterations.
|
b. Only cases for which Country_divided = EU are used in the analysis phase.
|
Table 10 shows that a model with 4 components was proposed for EU countries, which are not clearly differentiated from each other, as is the case when the sample of all European countries was observed (Table 9). The only four factors that are methodologically correct are those related to the trend of change (GDP change and Year to year CO2 change), such as Cement CO2 and Other industries CO2. A significant degree of overlap was detected in the case: Consumption based CO2, Coal CO2, CO2 per capita and CO2 per unit of GDP, where the correlation is positive or negative. All of the above calls into question the methodological correctness of the model for assessing the success of decarbonisation.
In non-EU countries (Russian Federation, Moldova, Armenia, Azerbaijan, Georgia and Ukraine), the model has the least methodological objections than in EU countries. The pattern matrix is shown in Table 11.
Table 11 Pattern matrix with factor overlap (selected SE European countries – non-EU countries)
|
Component
|
1
|
2
|
3
|
Flaring_ CO2
|
.987
|
|
|
Other_industry_CO2
|
.958
|
|
|
Cement_ CO2
|
.949
|
|
|
GDP_per_capita
|
.921
|
|
|
Consumption_ CO2
|
.870
|
|
|
CO2 _per_capita
|
.865
|
|
|
Coal_ CO2
|
.830
|
|
|
Oil_ CO2
|
.817
|
|
|
CO2_emission
|
.806
|
|
|
Share_of_ CO2
|
.661
|
-.510
|
|
REO
|
-.604
|
|
|
REC
|
|
.965
|
|
CO2 _per_GDP
|
|
-.930
|
|
Cumulative_ CO2 _emission
|
|
.815
|
|
Gas_ CO2
|
.605
|
-.672
|
|
GDP_Change
|
|
|
.978
|
Year_to_year_ CO2
|
|
|
.961
|
Extraction Method: Principal Component Analysis.
Rotation Method: Oblimin with Kaiser Normalization.a,b
|
a. Rotation converged in 14 iterations.
|
b. Only cases for which Country_divided = Non_EU are used in the analysis phase.
|
Table 11 shows that in this case, as well, there are factors that are problematic - Share of CO2 and Gas CO2. In these countries, it is more correct to use CO2 per unit of GDP to assess the success of decarbonisation, as it is an indicator that is clearly present in the second component, without overlapping with others, together with Renewable energy consumption and Cumulative CO2 emission.