4.1. Validity and Reliability of Data
Cronbach’s Alpha was estimated for the questions with several answer options or Likert’s scale, which proved that all questions provide relevant information (
Table 4). The reliability results show that the concept has been measured precisely without errors. The higher the value of Cronbach’s Alpha, the better the reliability [
56]. In all estimated questions, Cronbach’s Alpha is above 0.6, and the data are suitable for further research.
Principal component analysis (PCA) was applied for the chosen question blocks that provide information on the factors that determine particular problems. To employ principal component analysis, it is necessary to check whether the data are suitable for structure detection [
34]. Thus, the Kaiser–Meyer–Olkin measure of sampling adequacy test (KMO) and Bartlett’s test of sphericity were estimated for the following blocks of questions: 11, 12, 14, and 15. It was indicated that for the 11th block of questions, KMO = 0.740, Bartlett’s test of Sphericity
= 456.7, df = 55 and
p < 0.001. Thus, the data are suitable for PCA. For the next block of questions dedicated to determining the external that factors influence knowledge transfer, the KMO test (KMO = 0.842) and Bartlett’s test of Sphericity (
= 542.9, df = 55,
p < 0.001) show the suitability for PCA. The KMO test (KMO = 0.775) and Bartlett’s test of sphericity (
= 455.2, df = 78,
p < 0.001) prove the suitability of principle component analysis of the 14th block of questions that indicate the factors determining absorptive capacity in a company. The KMO test (KMO = 0.808) and Bartlett’s test of sphericity (
= 404.8, df = 36,
p < 0.001) for the last block of questions demonstrate appropriateness for PCA.
4.2. Results of Principle Component Analysis
Principle component analysis, as a background for EFA, reduces the number of variables. We assume that the most significant variables are supposed to be explained by the component in which the eigenvalue is more significant than one. The first component is the most general component in which most items load and present the most significant amount of variance. In the case of essential factors for implementing innovation, there is no correlation among the variables; thus, PCA might be applied. In this case, the first component explains 32.19% of the total variance, while the second and third explain 12.8% and 10.6%, respectively. The remaining eight components explain the remaining 45% of variance. The first component includes four variables (“new technologies drive further innovation” (Q11_5), “product or service improvement” (Q11_8), and “new products or patents developed” (Q11_10); and “carrying out research in the company” (Q11_11), with the following vector loadings of variables: 0.617, 0.673, 0.736 and 0.713. Two of the most important variables (“indirectly stimulated the supply of goods” (Q11_3) and “indirectly stimulated the demand of goods” (Q11_4)) are in the second component, with the vector loadings 0.918 and 0.850. The third component is covered by three remaining variables (“increased efficiency and productivity” (Q11_1), “optimized processes” (Q11_2), and “staff development” (Q11_6)), with the vector loadings of the variables as follows: 0.549, 0.761, and 0.729. The results of PCA analysis are the set of unit vectors that make up the transformation matrix (
Table 5).
The most significant correlation in the first component is between the component and the variable “new products or patents developed”. The second component has very high correlation with both variables. However, a stronger correlation exists between the second component and “indirectly stimulated the supply of goods”. The third component has a moderate, strong correlation with all variables. However, the strongest correlation between the component and “increased efficiency and productivity” is estimated.
In the case of the external factor that determines knowledge and technology transfer, the first component explains 39.15% of the total variance, the second and third 11.67%, and 9.38%, respectively, which, in total, explains more than 60% of all loading. The first component includes six variables out of eleven (“business investment in research” (Q12_1), “lack of young researchers” (Q12_4), “high technology export and/or import” (Q12_5), “state support for cluster development” (Q12_7), “state financial support for the transfer or assimilation of new technologies or knowledge” (Q12_8), and “cooperation between research institutions and business” (Q12_9), which produced the following vector loadings of variables: 0.804, 0.696, 0.658, 0.693, 0.638, and 0.714. The second component contains two variables (“cultural and historical similarities between the host and the transferring countries” (Q12_10) and “geographical distance between technology and knowledge transfer and host company” (Q12_11)), with the vector loadings of 0.796 and 0.758, respectively. The two remaining variables (“geographical distance between technology and knowledge transfer and host company” (Q12_3) and “competencies of a foreign partner in the development of innovations” (Q12_6)) are involved, the third component representing importance by the vector loadings of 0.771 and 0.822 (
Table 6).
Furthermore,
Figure 3 demonstrates that the hypotenuse of the right triangle is the projection of the original coordinates and becomes the new X coordinate. There is no y’ component, so its value is zero.
In summary, it might be concluded that the correlation between the first component and individual variable varies from moderately positive to a high positive correlation. The greatest positive correlation is between the first component and the “lack of young researchers”. The second component has a high positive correlation with both variables. A high positive correlation has been estimated between the third component and both variables.
In the case of the factors that determine absorptive capacity (14th block of questions), four components explain more than 59% of the total variance. All the variables were extracted in four components, as absorptive capacity is based on four processes. The first component explains more than 29% of the total variance; the other explains 13.37%, and the third and the fourth 9.19% and 8%, respectively. The rotated component matrix by varimax with Kaiser normalization revealed that five variables with a vector loading of more than 0.5 lay in the first component. The variables “learning skills” (Q14_9), “collective competence to adapt and adopt innovations” (Q14_10), “diversity of competencies” (Q14_11), “the existence of knowledge and information gathering technologies” (Q14_12) and “company’s investment in employee training” (Q14_13) produced the following vector loadings of variables: 0.596, 0.687, 0.634, 0.676 and 0.725. The second component involved three variables (“higher technological education” (Q14_1), “higher education in economics and management (Q14_2), and “number of employees with a master’s or doctoral degree” (Q14_3)), with the vector loadings of 0.797, 0.736, 0.753. The third component involved “individual ability of employees to find, select and absorb relevant information” (Q14_4), “communication within the business sector” (Q14_5), “understanding consumer needs” (Q14_6), with produced loadings 0.776, 0.701, 0.665. Two variables (“belonging to business clusters” (Q14_7) and “participating into exhibitions, events” (Q14_8)), with the vector loadings 0.636 and 0.857, were revealed in the fourth component. The results of the transformation matrix are provided in
Table 7.
The results showed the strongest positive correlation between the first component and the variable “company’s investment in employee training”. The variable “higher technological education” among the other variables has very high positive correlation with the second component. Meanwhile, a strong positive correlation exists between the third component and the “individual ability of employees to find, select and absorb relevant information”. Furthermore, a high correlation between the fourth component and “participating in exhibitions, events” is estimated.
Principle component analysis of the last block of questions was devoted to analyzing the factors that determine the transfer from a linear to a circular economy and disclosed that there are two components that explain 55.4% of the total variance. The first component includes six variables, with a vector loading of more than 0.5. These variables are “use of renewable energy sources” (Q15_1), “extending the use of the product by repair, refurbishment, and resale” (Q15_2), “participation in knowledge dissemination networks” (Q15_3), “development of new production processes” (Q15_7), “use of secondary raw materials” (Q15_8), “application of reverse logistics” (Q15_9), with the following estimated vector loadings of the variables: 0.730, 0.778, 0.527, 0.560, 0.772, and 0.687, respectively. The second component has the other three variables “technological innovation through digitization” (Q15_4), “technological innovation through digitization” (Q15_5), and an “exchange of knowledge and good practice” (Q15_6), with estimated loadings of 0.817, 0.697, 0.698. (see
Table 8 and
Figure 4).
The PCA shows that the first component has a moderate or high positive correlation with the individual variables. The strongest correlation exists “between extending the use of the product by repair, refurbishment and resale” and the component. The second component also has a moderate or high positive correlation.
In conclusion, it might be stated that PCA reduced the number of variables into an extracted number of components. Additional loadings were estimated, which shows the correlation between the component and each variable. In addition, a separate row of the component defined a linear composite of the component score that would be the expected value of an associated variable. Although the variables are correlated with the component, the components are uncorrelated, since they are orthogonal to each other in the sample space. Additionally, the higher loading of the variable reveals the greater importance to that component. Based on the PCA, we can state that variables of the construct “benefits of innovations” might be explained by three components named “improvement of products and services”, “demand and supply”, and “efficiency and productivity”. The most significant importance of the first component can be attributed to “new products or patents developed”. To the most considerable extent, the second component is affected by the “stimulated supply of goods”. “Optimized processes” is the most important variable for the third component. After PCA, the construct “external factors” was reduced to three components. Out of the “business investment in research”, “cultural and historical similarities between the host and the transferring countries”, and “competencies of a foreign partner in the development of innovations” are the most significant for the component they were assigned. Similar results show that the study concentrated on less developed countries and firms’ abilities to identify, absorb, transfer and exploit knowledge. In that case, collaboration between partners in networks, especially between business and research institutions, is extremely important [
48]. Variables in the 14th block (absorptive capacity) were reduced to four components with estimated different loadings. The essential variable for the first component is the company’s investment in employee training. “Higher technological education” makes the most significant impact on the second component. For the third component, “individual ability” is the most important. However, a study that analyzed a less developed country’s absorptive capacity and innovativeness proved that networking and knowledge acquired from learning from a foreign capital company is more important than individual ability [
50].
Meanwhile, the 4th component’s most significant impact is the variable “participation in exhibitions”. The 15th block of questions (variables) were reduced to two components. Thus, for the first component, the most significant variable is the “use of secondary raw materials”. On the other hand, “technological innovation through digitization” makes the most significant impact on the second component, which confirms the results of a study on Brazil’s fashion industry [
24]. Thus, the produced loading might also be used as a predictor for future research.
4.3. Correlation and Regression
The variable “innovations” is based on the results of the survey’s 7th block of questions and refers to innovations implemented within the last five years. Several developed or implemented innovations demonstrate the impact of successfully acquired knowledge, transformation, and exploited it. This variable has been used in previous studies [
12,
13,
14,
34] to indicate innovative productivity.
Table 9 presents the results of the correlation between the analyzed variables.
The correlation shows that, in our case, the variable “innovations” has a fragile but significant relationship with “benefits of innovations”, “external factors”, and “transfer from linear to a circular economy”. Meanwhile, our case has no correlation between innovations and absorptive capacity. Social innovations and innovations have moderate significant relationships. Moderate and significant correlations exist between absorptive capacity and the three last variables. At the same time, there is no relationship between absorptive capacity and social innovations. A fragile but significant relationship is estimated between social innovations and “benefits of innovations” and transfers from linear to the circular economy. Although, correlation indicates the association between variables. However, statistically, it only means the degree to which a pair of variables are linearly related. Hence, it does not provide information about the causality and impact of the variables on the other. Thus, for further research, additional econometric modelling might be employed.
Only the second model is significant and it explains 65% of the trend (
Table 10). The correlation and regression analysis unveiled that there are no interlinkages or weak relationships that exist between innovation and other constructs, except social innovation (
Figure 5). On the other hand, some researchers claim [
50] that nonlinear relationships exist between innovation and technology transfer.