INNOVATION FINANCING STRUCTURE AS A FACTOR OF ECONOMIC GROWTH: CROSS COUNTRY ANALYSIS Abstract. The article focuses on the level and dynamics of innovation financing in Azerbaijan and Ukraine compared to the world level and the places of Azerbaijan and Ukraine in the Global Innovation Index and trends in their positioning in the dynamics. The analysis reveals negative dynamics in both countries in this sphere. The innovation financing structure's role as a factor of economic growth and international reproductive relations

The article focuses on the level and dynamics of innovation financing in Azerbaijan and Ukraine compared to the world level and the places of Azerbaijan and Ukraine in the Global Innovation Index and trends in their positioning in the dynamics. The analysis reveals negative dynamics in both countries in this sphere. The innovation financing structure's role as a factor of economic growth and international reproductive relations development is substantiated. The dependence of the country's economic growth level (GDP growth per capita) on the value of expenditures on innovation financed by various sectors of the economy (government, the private nonprofit sector, foreign investors and the higher education sector) is studied. The study consists of data for 12 European countries for 2007-2017 (limited calculations in 2017 due to the availability of information on open portals of the World Bank, the EU Statistical Office). At the first stage, the distribution of the relevant indicators was evaluated using the Shapiro-Wilk test. Based on these results the method of calculating the correlation coefficient is chosen: Pearson – for indices that are subject to the ordinary distribution law or Spearman – for indices that are not subject to the ordinary distribution law. A correlation analysis regarding the strength and nature of the relationship between relevant indices and the dynamics of GDP per capita in these countries is performed to identify the duration of time lags, after which this relationship is the most statistically significant. In the second stage, there are three types of regression models for estimating panel data to identify the impact on the economic growth dynamics of innovations financed by different economic sectors: 1) with fixed effects (based on the least-squares method); 2) with random effects (based on the general least squares method (GLS); 3) dynamic model for estimation of Arellano-Bond panel data, which considers time lags (based on the general method of moments (GMM)). In the third stage, using Wald's tests, Breusch-Pagan and Hausman, the adequate model specification is chosen. When choosing a dynamic model of Arellano-Bond, the Sargan test is performed to validate the parameters. The control variables in all three types of models consider net inflows and outflows of foreign investment, inflation (GDP deflator) and labour force participation rate (% of total population ages 15-64). The second and third stages of the study obtained the results as follows. It is empirically confirmed that a 1% increase in the share of government sector-funded R&D expenditures leads to a decrease in annual GDP growth per capita by an average of 0.15% (excluding time), business sector – to the increase by 0.13% with a time lag of 2 years, thanks to foreign sources – to the increase by 0.1% (without time lag); higher education sector – to the decrease by 0.78% (without time lag). It is substantiated that the state should reduce the share of direct investment in innovation. At the same time, it should focus on effective legislation, motivating the business sector and foreign investors to increase investment in research and development to stimulate economic growth in Azerbaijan and Ukraine and the development of international reproductive relations.


Marketing and Management of Innovations, 2020, Issue 3
135 http://mmi.fem.sumdu.edu.ua/en role that government regulations can play in the development of innovation, growth and continuous improvement of the citizens' quality of life. Nelson (2005) argues that the standard neoclassical theory of economic growth is inadequate to explain the economic growth phenomenon. He presents an alternative theory, which emphasizes that economic growth caused by the technological progress, considers this process as involving the coevolution of technology, institutions and industry structure.
As a result of the formation of a multiple regression model, Sokolov-Mladenovic et al. (2016) concluded that an increase in research and development spending as a percentage of GDP by 1% would increase the growth rate of real GDP by 2.2%. The researchers paid special attention to the negative impact of the birth rate in the EU-28 on economic growth. Pessoa (2007) traced the link between R&D spending and economic growth, emphasizing that increasing R&D spending is not a guaranteed way to improve economic growth, especially in countries below the technological frontier. Still, there are other ways in which technology affects growth, other than those based on formal research and development indicators.
Balashova (2015) examined the impact of such instruments as funding for research conducted in the government sector (in government research organizations and higher education institutions), funding for research conducted in the business sector (through government procurement, grants, etc.), tax subsidies and benefits provided to businesses for research and development -on the amount of funding by the business sector of internal research and development in OECD countries in the period from 1981 to 2012.
Methodology and research methods. The methods of correlation and regression analysis using the STATA software package for the sample from 12 European countries for 2007-2017 (limited calculations in 2017 due to the availability of information on open information portals of the World Bank, EU Statistical Office) were applied to confirm the hypothesis on the impact of the innovative financing structure on economic growth.
The nature of the distribution of the studied indicator was assessed using the Shapiro-Wilk test (Shapiro and Wilk, 1965). Based on those results the calculating method of the correlation coefficient was chosen: Pearson -for indicators subject to the law of normal distribution (Pearson, 1896), or Spearmanfor indicators do not obey the law of normal distribution (Spearman, 1904). A correlation analysis allowed establishing The strength and nature of the relationship between the indices of innovation funding structure and the dynamics of GDP per capita. It revealed the duration of time lags, after which this relationship is the most statistically significant.
There are three types of regression models to estimate the panel data to identify the impact on the economic growth dynamics of innovations funded by different economic sectors (Baltagi, 2013): 1) with fixed effects (based on the least-squares method) (Allison, 2009;Arellano, 1987); 2) with random effects (based on the general least squares method (GLS) (Schunck, 2013); 3) a dynamic model for estimating Arellano-Bond panel data, which considers time lags (based on the general method of moments (GMM)) (Arellano & Bond, 1991). Wald, Breusch-Pagan, and Hausman tests selected the adequate model specification (Gourieroux and Monfort, 1995;Hausman, 1978). In the case of a dynamic Arellano-Bond model, the Sargan test for parameter validity was performed (Arellano and Bond, 1991).
Results. The study (Samoilikova, 2020) observed the empirical substantiation and formalization of the impact made by the financial regulations indices for innovation development on the overall level of innovation development in different European countries. One of the essential areas of financial regulations is to study the impact made by the structure of financing innovation as a factor in economic growth. One should note that innovative development involves a significant transfer of resources between economic sectors, organizations and countries.
Gross expenditure on R&D (GERD) ( Table 1) includes the capital and current expenditures in four main sectors: the government, business, private non-profit and higher education sectors, covering fundamental and applied research and experimental development.     The trend of changing the share of GERD in GDP differs significantly. It can be explained by individual economic development features as a whole and innovative development of different countries, in particular. Many EU countries have a gradual increase in this index since 2009 after the financial crisis (Estonia, Slovenia, Poland, the Czech Republic). Figure 2 presents the dynamics of GERD in Ukraine and Azerbaijan in comparison with the EU. It clearly shows the declining trend and low level of funding for innovation in these countries, in contrast to the EU countries. The GERD has been analyzed by the main financing sources: the share of R&D expenditures financed by the government sector (Table 2), the business sector (Table 3), the higher education sector (Table 4), the private non-profit sector (Table 5), at the expense of foreign sources (Table 6).  Among these countries, the largest share of R&D expenditures financed by the government sector occurs in the Russian Federation, Latvia, Ukraine and Estonia. Instead, the lowest expenditures are in Slovenia and Bulgaria. It is worth noting that Slovenia has the highest share of R&D expenditures in GDP with minimum R&D expenditures financed by the government sector. In Ukraine, the Russian Federation and Latvia, with significant government sector funding, the share of R&D expenditures in GDP is relatively low. However, for example, in Estonia during the selected period, the share of R&D expenditures in GDP is relatively high with significant amounts of funding from the government sector. The situation is similar in other countries. So, it does not allow to make unambiguous conclusions about the relationship of the studied indices based on the data from Tables and graphs. Table 3 demonstrates the indices of GERD share financed by the business sector in the studied countries. The highest rates of GERD financed by the business sector in 2017 are typical for Slovenia, Hungary, Romania, while the lowest -In Latvia, the Russian Federation, and Ukraine. However, these positions are not stable in dynamics. Similarly, we consider the indices of GERD financed by the higher education sector, which is insignificant in size compared to the first two surveyed sectors (Table 4).  Table 4 Ukraine 0,2 0,3 0,3 0,2 0,2 0,2 0,2 0,1 0,1 0,2 0,2 Czech Republic 0,8 1,3 1,2 0,9 0,9 0,9 0,5 0,6 0,7 0 Similarly, to the higher education sector as a source of funding, the non-profit sector finances a small share of GERD in the studied countries (Table 5). In particular, in Ukraine since 2014, the private non-profit sector does not practically finance GERD. There was a similar situation in other countries (Slovenia, Czech Republic) during some periods. Lithuania, Hungary, Slovakia, and Romania have a relatively high level of R&D expenditures financing from the private non-profit sector. Instead, low -in Ukraine, the Czech Republic, Slovenia, etc.
The last of GERD financing sources are funds from foreign sources (Table 6). In contrast to the share of financial support for innovation development from the higher education sector or the non-profit sector of the economy, GERD financing from foreign sources in many countries plays a significant role alongside funding from the government and business sectors. For example, in Latvia in 2011-2013, the share of GERD financing from foreign sources exceeded 50%. Thus, the highest value of this index from 2015 to 2017 is observed in Bulgaria and Latvia, Ukraine and the Czech Republic, the lowest -in the Russian Federation, Poland, Romania. Table 7 demonstrates the results of the investigation on the indices of the considered sources of GERD financing affect the country's economic growth, represented by the index of change in GDP per capita. Confirmation or refutation of the hypothesis regarding the impact of GERD financed by various economic sectors on the change of GDP per capita is justified primarily by the calculation of the relevant correlation coefficients. Before that, it is necessary to check whether the indices of GERD share financed by the government sector (GS), the business sector (BS), the higher education sector (ES), the private non-profit sector (NS) and foreign sources (FS). The mentioned above subjects to testing the normal distribution using the Shapiro-Wilk test (Table 8) based on the data from Table 2-6. Calculations are performed in the STATA software package.
Accordingly, the Pearson correlation coefficient calculation defines the strength and nature of the relationship between the indices that obey the normal distribution law (Shapiro-Wilk test result> 0.05). Instead, the Spearman correlation coefficient calculation allowed identifying the relationship between indices that do not obey the normal distribution law (Shapiro-Wilk test result <0.05). Besides, the approximation of the results to the actual realities of the country's economic and innovative development Marketing and Management of Innovations, 2020, Issue 3 141 http://mmi.fem.sumdu.edu.ua/en determines the feasibility of identifying the correlation coefficients taking into account the time lags between the studied indices to increase their adequacy.  Table 9 demonstrates the generalized results of the assessment of the relationship between the financial support of innovative development (according to GERD financing sources) and the change in GDP per capita.  Table 10 summarizes the criteria to estimate correlation coefficients to define the strength and nature of the relationship between the studied indices. The character of the relationship High 0,9 -1 Very high Source: developed by the authors.
It is supposed that the impact is not statistically significant with a very weak and weak relationship between the indices. Significant influence occurs at a correlation coefficient of 0.7, average -from 0.5 to 0.7.
Thus, the correlation analysis of GERD impact financed by the government, business, private nonprofit sectors, foreign sources and the higher education sector on GDP per capita dynamics in these countries revealed the duration of time lags, due to which this impact becomes statistically significant: http://mmi.fem.sumdu.edu.ua/en -the impact of the GERD share financed by government sector on GDP per capita: high -in Romania (with a lag of 2 years), in Bulgaria, Latvia, Slovenia, Hungary and the Czech Republic (with a lag of 3 years), in Ukraine (with time lag in 4 years); average -in Estonia (without time lag), in Lithuania and Slovakia (with a lag of 2 years), Poland, the Russian Federation (with a lag of 3 years). The character of the relationship for 10 countries in the study sample is inverse, for 2 countries -direct; -the impact of GERD share financed by the business sector on GDP per capita: high -in Romania, Hungary and the Czech Republic (with a lag of 2 years), in Slovenia and Poland (with a lag of 3 years), in Ukraine 4 years); average -in Bulgaria, Estonia, Latvia and Slovakia (without time lag), in the Russian Federation (with a lag of 2 years), in Lithuania (with a lag of 3 years). The character of the relationship for 8 countries in the study sample is direct, for 4 countries -inverse; -the impact of GERD share financed by the higher education sector on GDP per capita: high -in Romania and the Czech Republic (with a lag of 1 year), in Bulgaria (with a lag of 2 years), in Ukraine (with a lag of 5 years) ); average -in Latvia and Lithuania (without time lag), in Poland (with a lag of 1 year), in Estonia, Slovakia and Slovenia (with a lag of 2 years), in the Russian Federation (with a lag of 3 years). The character of the relationship for 7 countries in the study sample is inverse, for 4 countries -direct; -the impact of GERD share financed by the private non-profit sector on GDP per capita: high -in Bulgaria (without time lag), in Latvia and the Czech Republic (with a lag of 1 year), in Estonia and Slovakia (with a lag of 2 years), in Lithuania (with a lag of 3 years), in Poland (with a lag of 5 years); average -in Slovenia (with a lag of 1 year), in Romania (with a lag of 2 years), in Hungary (with a time lag of 3 years). The character of the relationship for 5 countries in the study sample is inverse, for 5 countries -direct; -the impact of GERD share financed by foreign sources on GDP per capita: high -in Romania, Hungary and the Czech Republic (with a lag of 2 years), in Bulgaria, Poland, the Russian Federation, Slovakia and Slovenia (with a lag of in 3 years); average -in Latvia and Ukraine (with a lag of 1 year), in Estonia (with a lag of 2 years), in Lithuania (with a lag of 3 years). The character of the relationship for 11 countries in the study sample is direct, for 1 country -inverse.
Therefore, we test the hypothesis regarding the negative impact of GERD share financed by the government sector, on the country's economic growth. We build a regression model for panel data for the studied countries for the period from 2007 to 2017 to confirm this hypothesis.
Since the economic growth (change of GDP per capita in % to the previous year) ( Table 7) cannot be estimated only by GERD share financed by the government sector (Table 2), we will introduce benchmarks into the model that are important macroeconomic determinants, namely: 1) net outflows of the direct foreign investments (% of GDP) (Table 11): 2) net inflows of foreign direct investment (% of GDP) (Table 12): 3) share of labour resources (% of the total population aged 15-64) (Table 13): The regression model for estimating the impact of GERD share financed by the government sector, net inflows and outflows of foreign direct investment, labour resources and inflation on GDP dynamics per capita can be represented as: GDPGS = α + β1 (GS) + β2 (II) + + β3 (IO) + β4 (L) + β5 (I) +u + ε (1) where α -constant; β -coefficients obtained by the least-squares method (MLS); u -standard error for individual effects; ε -standard error; GDP -annual change of GDP per capita (% to the previous year); GS -the share of GERD financed by the government sector (% of GERD); IO -net outflows of foreign direct investment (% of GDP); II -net inflows of foreign direct investment (% of GDP); L -the share of labour resources (% of the total population 15-64 years); I -inflation rate (GDP deflator, %). Table 15 demonstrates the descriptive statistical features of the model variables. We evaluate a regression model with fixed effects for the studied variables. Regression «within» is a method for estimating the regression model coefficients with deterministic individual effects (FE) and is an initial regression model, rewritten in terms of deviations from the average time variables, eliminating individual effects that are not observed. Each object of the sample is introduced with its own constant. Thus, the model considers the existing heterogeneity, which is not observed. The evaluation of the model is performed by the least-squares method (MLS).  Table 16 demonstrates the assessment results of the impact of GERD share financed by the government sector on the annual change of GDP per capita. The level of significance of the t-criterion for the coefficient L (labour resources) exceeds 0.05, so it cannot be statistically significant (the probability of error acceptance of the hypothesis is 15.5%). Other indices are statistically significant. Moreover, the model requires only the noncorrelation of ε and X for the ability of MLS estimates with deterministic individual effects. The correlation between X and u is assumed, which is a manifestation of the FE-model flexibility. In our case corr(u_i, Xb) = -0.6232.
The R-squared index is not very high (0.2848), but it can be explained by the fact that this model has significant individual differences (as opposed to dynamic), which ultimately indicates the necessity to consider the individual effects and the ability of the selected model. At the same time, the regression model with fixed effects does not allow to estimate the coefficients for time-invariant regressors since they are eliminated from the model after the transformation «within». Therefore, there is a need for a parallel study and regression model with random effects (RE) ( Table 17). The generalized least squares method (GLS) performs estimation. Compared to the previous model, most of the coefficients are not statistically significant (the significance level P> |z| exceeds 0.05). The index corr (u_i, X) = 0 (assumed) reflects an important hypothesis underlying the model -regressors should not correlate with random effects that are not observed. Also, the interpretation of this model should not be based on the R-sq index, since it is not an informative means of checking the model ability in the regression, estimated using the generalized least squares method. The Wald statistical test proves the significance of the regression in this case. However, the index Wald chi2 (5) = 16.39 is a low value, Prob> chi2 = 0.0058. It is explained by the fact that the Wald test checks the hypothesis that all individual effects are equal to zero. Herewith, a regression model with random effects can only take place where the random effect does not correlate with regressors, which is often not performed.
It is advisable to compare the fixed effects regression model with a standard regression model using the Wald test, which checks the hypothesis regarding the zero equality of all individual effects, to select the adequate model. In particular, we obtain the following result: F test that all u_i = 0: F (11, 115) = 3.31. Prob> F = 0.0006. Since the significance level is p-level <0.01, the main hypothesis is not confirmed, and the fixed effects regression model is better suited for data description than the standard regression model. Second, we compare the random effect model with the standard regression model using the Breusch-Pagan test. The Breusch-Pagan test is a test for the random individual effect. In the studied case, the p-level> 0.01, i.e. the main hypothesis is confirmed, and the model with random effects describes the data worse than the standard regression model.
Third, it is reasonable to use the Hausman test to select an adequate model from the two formed ones (between FE and RE models). The test checks the main hypothesis H0: corr (ui, Hit) = 0 or ui can be considered as random effects. And the alternative -HA: corr (αi, Hit) ≠ 0 or ui should be considered deterministic. The results of the Hausman test indicate that the p-level is <0.01, so the main hypothesis is not confirmed. It enables to conclude that in this study, it is advisable to use a regression model with fixed individual effects.
Therefore, the regression equation constructed according to the accepted model with fixed effects is: GDPGS = -0,15GS + 0,55II -0,54IO + 0,33L+ 0,32I -10,03 (2) The evaluated coefficient β for GERDGOV is statistically significant (the probability of erroneous acceptance of the hypothesis is 1%) and negative, indicating an inverse relationship between GDP and GS. It is empirically confirmed that with the growth of GERD financed by the government sector (in total GERD) by 1%, the annual growth (change) of GDP per capita will decrease by 0.15%. Therefore, the urgent problem is to reduce GERD share financed by the government sector in the structure of GERD and the increase in the share of other financing sources for innovative development.
The impact of financial support for country's innovative development on its economic growth was examined through the business sector according to the dynamic model of panel data evaluation (Arellano-Bond linear dynamic panel-data estimation (Arellano & Bond, 1991)). It is used to study the economic phenomena evolution, avoiding the displacement of aggregation. Linear dynamic models of these panels include lags of the dependent variable as covariates and contain unnoticed fixed or random effects at the panel level. It means that the dynamic model makes it possible to consider how the share of GERD financed by the business sector of the previous period affects the current situation).
The dynamic model, achieved by introducing lagged variables, leads to significant changes in the interpretation of the regression equation. Regressors describe a complete set of information that determines the observed values of the dependent variable without a lagged dependent variable. The prehistory of the regressors is considered with the introduction of the lagged variable. Therefore, it causes any influence on the measurement process. In this case, since both MLS and FE-estimates are incapable of the final values of T, the instrumental variables method or the generalized method of moments (GMM) is used to obtain adequate estimates in this model. Table 18 provide descriptive statistical features of the variables of the model for estimating the impact of GERD share financed by the business sector on GDP dynamics per capita. The results of estimating the impact of GERD share financed by the business sector on the annual change of GDP per capita are shown in Table 19. One should also notice that the dynamic model for estimating the Arellano-Bond panel data considers the fact that some regressors in the model are not completely exogenous. They can be influenced by the past and the present value of the dependent variable (GDP dynamics per capita). In our case, only the labour resources index (economically active population) can be considered a completely exogenous variable. Other variables are considered endogenous.
The estimated coefficient β for BS index is statistically significant (the probability of erroneous acceptance of the hypothesis is 0.6%) and positive. It indicates a direct relationship between GDP and BS.
It is empirically confirmed that with the growth of the share of GERD financed by the business sector (in total GERD) by 1%, the annual growth (change) of GDP per capita will increase by an average of 0.13% over time 2 years. The impact of GERD financed by the higher education sector on the economic growth change was estimated similarly to the dynamic model for Arellano-Bond panel data. It should be emphasized that net inflows and outflows of foreign direct investment and the inflation level are not completely exogenous variables. Instead, the share of GERD financed by the higher education sector and the labour force index considered being exogenous. Table 20 demonstrates the descriptive statistical features of the model variables.  Table 21 shows the estimating results of the impact made by the share of GERD financed by the higher education sector, on the annual change of GDP per capita.
The estimated coefficient β for the ES index is statistically significant (the probability of erroneous acceptance of the hypothesis is 5.4%) and negative. It indicates an inverse relationship between GDP and ES. It is empirically confirmed that with the growth of GERD financed by the higher education sector (in total GERD) by 1%, the annual increase (change) in GDP per capita will decrease by an average of 0.78%.
The impact of R&D financing from foreign sources on the economic growth change was estimated according to the dynamic regression model for panel data.  Source: developed by the authors. Table 23 shows the estimating results of the impact made by GERD share financed by foreign sources on the annual change in GDP per capita. Thus, the regression equation based on the dynamic model of estimation of Arellano-Bond panel data for evaluating the impact made by GERD share financed by the foreign sources on the dynamics of GDP per capita is as follows: GDPFS it = -33,97 -0,23GDPFS i,t-2 + 0,1FSit +0,40IIit -0,42IOit + 0,46Lit + 0,13Ii,t-1 The estimated coefficient β for the FS index is statistically significant (the probability of erroneous acceptance of the hypothesis is 1.5%) and positive. It indicates the direct relationship between GDP and FS.
It is empirically confirmed that with the growth of GERD financed by the foreign sources (in total R&D expenditure) by 1%, annual growth (change) of GDP per capita will increase by an average of 0.1%.
The impact of GERD financed by the private non-profit sector on changes in economic growth was evaluated. Descriptive statistical features of the model variables are given in Table 24. The estimating results of the impact made by the R&D expenditures share financed by private non-profit sector on the annual change in GDP per capita are shown in Table 25.  The significance level of the t-criterion for GERD share financed by the private non-profit sector is 0.546. It shows the statistical insignificance of the regression coefficient obtained for this index in the model with fixed effects. Similarly, the regression coefficient for GERD share financed by the private nonprofit sector, calculated according to the model with random effects (significance level of t-criterion is 0.967), is not statistically significant. http://mmi.fem.sumdu.edu.ua/en The estimating results of the impact made by GERD share financed by the private non-profit sector on the annual change in GDP per capita according to the dynamic regression model for estimating panel data (Arellano-Bloom) are shown in Table 26. Exogenous variables include labour force index and GERD financed by the private non-profit sector. In this model, the significance level for the regression coefficient for the index of GERD share financed by the private non-profit sector is 0.143. It exceeds the allowable value of 0.05 and indicates the statistical insignificance of the regression coefficient in the obtained model. Thus, the formed models for estimating the impact of GERD share financed by the private non-profit sector on the annual change in GDP per capita are inadequate. Accordingly, financing GERD from this source is inefficient.
Conclusions. According to the above mentioned, there could be made the conclusions as follows.
1. Models to estimate the influence made by the financial maintenance sources structure of the country's innovative development on the economic growth (growth (change) of GDP per capita) are constructed: -R&D financing by government sector: GDPGS = -0.15GS + 0.55II -0.54IO + 0.33L + 0.32I -10.03 -R&D financing by the business sector: GDPBS it = -41.96 -0.16GDPBS i, t-2 + 0.13BSi, t-2 + 0.5IIit -0.53IOit + 0.57Lit -0.13Iit -R&D financing by the higher education sector: GDPES it = -39 -0,22GDPES i, t-2 -0,78ESit + 0,37IIit -0,41IOit + 0,58Lit -0,16Iit -R&D financing by foreign sources: GDPFS it = -33.97 -0.23GDPFS i, t-2 + 0.1FSit + 0.40IIit -0.42IOit + 0.46Lit -0.1Iit 2. The following hypotheses are empirically confirmed: -with an increase in the share of GERD financed by the government sector (in total GERD) by 1%, the annual increase (change) in GDP per capita will decrease by an average of 0.15% (excluding time lag) (probability of the hypothesis erroneous acceptance is 1%); -with an increase in the share of GERD financed by the business sector (in total GERD) by 1%, the annual increase (change) in GDP per capita will increase by an average of 0.13% with a time lag of 2 years (probability of the hypothesis erroneous acceptance is 0.6%); -with an increase in the share of GERD financed by the higher education sector (in total GERD) by 1%, the annual increase (change) in GDP per capita will decrease by an average of 0.78% (excluding time lag) (the probability of hypothesis erroneous acceptance is 5.4%); -with an increase in the share of GERD financed by foreign sources (in total R&D) by 1%, the annual increase (change) in gross domestic product per capita will increase by an average of 0.1% (excluding time lag) (the probability of erroneous acceptance of the hypothesis is 1.5%).
the inefficiency of R&D financing by the private non-profit sector is empirically confirmed. 3. Taking into account the European countries' experience to stimulate economic growth and development of international reproductive relations in Azerbaijan and Ukraine, the state should reduce the share of direct investment in innovation, focusing on effective legislation. It motivates the business sector and foreign investors to increase investment in research and development. Thus, the hypothesis confirms that business is the best consumer of R&D results since it has more opportunities to monetize the innovation results faster and implement them in production. The length of chain between the emergence of innovation and use results, which are manifested in GDP growth, is reduced. The state's role has to provide effective mechanisms to stimulate the transfer of innovation to the business environment. When the country is the main investor in science and development, the model is less effective in terms of economic growth dynamics.
Author Contributions: conceptualization, methodology, supervision, formal analysis, software, writing-original draft A. S.; investigation, writing-review and editing, writing-second draft A. R.
Funding: This research received no external funding.