CONCEITOS DE MODELO DE NEGÓCIOS

2.6.1 Results with the Cobb-Douglas function

Table 2.5 shows the results of the Cobb-Douglas production function. In estimating the Cobb-Douglas function, we run several models and use the number of patents as the dependent variable. The results suggest that human capital has no effect on innovation but R&D has a positive effect on innovation with an elasticity comprised between 0.220 and 0.238 in the different models. In the model 1, we include only FP5 and it seems to have no effect on innovation. In column 2, we include the FP6 and find a positive effect on innovation with an elasticity of 0.028. When we include the two variables at the same time in the model 3, we see in column 3 that the results remain unchanged. Finally, when we include the sum of the FP5 and FP6, we see that the total effect is almost similar to the effect of the only FP6 (0.027). As explained above, the Cobb-Douglas production function is nested in the translog production function, so we apply a Wald test to check whether the Cobb-Douglas production function is rejected in favor of the translog production function. In our case, at the 5% significance level, the Cobb-Douglas production function is rejected by the Wald test in favor of the translog production function for all the models (see figure 2.4).

Table 2.5: Results with Cobb-Douglas function

Adjusted R² 0.005 0.037 0.038 0.036

Note: The dependent variable is the number of patents and all variables are log-transformed and first-differenced.

We used the within estimation to run our model. The columns give the estimates and the Standard errors (in parentheses below coefficients). Significance levels: ***1%; **5%; *10%

2.6.2 Results with the translog function

We first introduce variables on FPs in a translog function, in a random growth model for the entire EU-27 regions. Table2.14, therefore, shows these estimates, for log-log specification and first differentiated model with individual fixed-effects. First of all, the coefficients of HK and R&D are very stable whatever the way to introduce FPs’ variables. The R&D coefficient always ranges between 0.55 (column 5) and 0.60 (column 1), its square is also always significant and equal to about−0.065. The human capital coefficient ranges between 0.508 (column 8) and 0.90 (column 1) and its square is about −0.11. The interaction term between human capital and R&D spending is never significant. This leads to an average elasticity of the number of patents to R&D spending between 0.184 and 0.236 (table2.22), at the mean point³². The elasticity of human capital is also negative at the mean point between

−0.472 and 0.388. These results are a little stronger when we calculate the elasticities at the median point.

When introducing the FPs’ variable, the amount of FP5 spent in one region seems to have no effect alone on the number of patents in that region. However, the regional FP6 expenditure seems to have a positive impact; its elasticity is equal to 0.028 when it is in-troduced only in level (columns 3 and 4). When its square is also inin-troduced as well as an interaction term with the FP5 (column 5), the elasticity of the FP6 expenditure on the number of regional patents follows a convex form and is equal to 0,029 at the mean point

32Table2.21presents the mean and median of all variables in logarithm according to the considered sample

(table 2.22). Estimates (columns 5 and 6) also highlight a substitution effect between the FP5 and FP6, as the coefficient of the interacting term between these variables is equal to

−0.002 (table 2.13). If the sum of the FP5 and FP6 regional amounts is included as policy variable (columns 7 and 8 of table 2.22), results are quite similar to those obtained for the FP6 variable alone: 0.027 when introduced alone, in column 7, and 0,026 at the mean point (column 8).

In table 2.13, we include in the model the FPs amounts and all their interacting terms with human capital and R&D spending. In column 2, the FP5 regional spending is still never significant. Coefficients associated to the regional FP6 amounts are similar to those obtained previously and the R&D spending and the FP6 spending seems to be substitutes as the interacting coefficient is significant and equal to −0.006. These complete translog estimates show that, for entire EU-27 regions, only the FP6 spending has a significant effect on regional innovation, on average (0.026). This result remains positive but stronger when FP6 amount is replaced by the sum of FP5 and FP6 spending. In these case, the elasticity of the regional number of patents to FPs is equal to 0.043 (table 2.23). We find also that the elasticity of substitution between FP5 and FP6 is negative. This relationship is related to the negative correlation between the variables due to the fact that the FPs spending spilled over the periods. The FP5 program start later than the official lunched period and when the FP6 program started, the FP5 was running but with decreasing expenditures. That’s why during the first years of the lunch of the FP6, its amounts were increasing while the FP5 amounts were decreasing. This explanation is confirmed by the fact that the effect of the sum of FP5 and FP6 spending (SumFP5-6) is positive and significant.

Moreover, when we focus on the complementarity and substitution of factors for the EU-27 countries, results in table 2.13 (column 4) show that there is complementarity between the human capital and the sum of FP5 and FP6 spending (0.007). This means that the decisions to invest in R&D must be joined to the investment in education and human capital to increase the innovation of countries. However, we find substitution (column 3) between the regional R&D spending and the FP6 spending (−0.006), meaning that there is a possible lack of coordination between national and European instruments. This may be due to a substitution effect between regional R&D spending and the europeans subsidies for innovation policies. No complementarity or substitution effect is observed between innovation the factors and the FP5 spending. Therefore, when the EU decides to fund research activities and

innovation policies, it should think at national public R&D subsidies structure but also think at improving training and teaching quality in higher education institutions to strengthen skills and human capital.

As described in the previous section, European countries and therefore regions are very heterogeneous, regarding their economic and innovation system and regarding the amount of FPs spending they benefit from. In order to distinguish different effects of the Framework Programmes, the same econometric models were run on three samples: the top 6 EU, the top 11 EU countries, and the low 16 EU countries, in terms of innovation activities, as described in the previous section. The results are shown in tables 2.15 to 2.20 and the calculated elasticities are presented in tables 2.24 to 2.29.

For the top 6 EU countries, results (table2.15) for R&D variables and human capital have a much more significant scope compared to those obtained for the entire EU-27. Moreover, in these countries, the results (table 2.24 column 7) on the FPs effects are slightly similar than to those of the top 11 countries for both FP5 and FP6 although the effect of the sum of FP5 and FP6 is slightly stronger. Further, when focusing on the more eleven innovative countries (EU top 11), results (table 2.17 and 2.26) for R&D variables are very similar to those obtained for the entire EU-27. The regional level of human capital shows a strong concave elasticity for the 11 more innovative European countries. About the FPs impacts, the FP6 spending has the same effect as in all european countries (about 0.026), but the FP5 spending becomes significant alone as for the top 6 countries and the effect of the sum of FP5 and FP6 spending (SumFP5-6) is stronger (0.048) on regional innovation, whatever the way they enter in the regressions. Results for less performer 16 EU countries (EU Low 16) in table 2.19 suggest that while the human capital seems to have no effect, the effects of the R&D spending is negative and those of FP6 are quite similar to the previous ones of the EU-27 (0.027) in the partial translog. However, in the complete translog, the effect of the FP5 is null and those of FP6 and the sum of FP5 and FP6 seem to be negative (respectively

−0.022 and −0.001). FP5 spending alone never affects regional innovation in these lagged EU regions, in terms of patents.

For the results of the complete translog (see tables 2.16, 2.18 and 2.20), we compute the elasticities at the mean point of variables. These results confirm the previous ones and suggest that the impacts of the FPs policy are larger in the EU top 11 performers countries, in terms of R&D and human capital than in the EU low 16 (lagging countries).

For the EU top 6 performers, results show complementarity between human capital and FP6 or SumFP5-6 spending but also substitution between regional R&D spending and FP6 spending. However, for these countries, no complementarity or substitution effect is de-tected between innovation factors and the FP5 spending. Further, when focusing on the Top EU11 performer countries, results in table2.27show complementarity effects between human capital and FP6 (0.007) and between human capital and the sum of FP5 and FP6 spend-ing (SumFP5-6) (0.008) but also substitution between regional R&D spendspend-ing and FP6 or SumFP5-6 spending. However, for these countries, no complementarity or substitution effect is detected between innovation factors and the FP5 spending. Furthermore, when focusing on the less innovative countries (EU Low 16), findings show that the FP5 has no effect but the FP6 and the SumFP5-6 have significant effects on regional innovation. However, no complementarity or substitution effect is detected between factors. These results may be explained by the earlier introduction of the FPs policies in leading European countries, but also reveal that certainly due to threshold effects, regions must have reached a certain level in innovation abilities in terms of R&D spending and human capital to make these policies efficient (absorptive capacity).

The results of the EU Low 16 countries in table 2.19 reinforce that idea: R&D, HK and the squared of HK are never significant, only the squared R&D is significant and equal to about −0.100. The elasticities at the mean point show that the effect of human capital is null and the effects of regional R&D spending are negative on innovation. The effects of FP6 are also similar than previous ones but slightly smaller, as well as for the sum of FP5 and FP6. FP5 spending alone never affects regional innovation in these lagged EU regions, in terms of innovative process.

All these previous results (elasticities) are calculated at the median point and they are a little stronger than those calculated at the mean point. As you can see, the estimates from the translog production function are more significant than those from the Cobb-Douglas.

Finally, these results confirm the role of thresholds effects and that it is necessary to ob-serve a certain level of innovative factors to make the policy efficient in the long run. These findings are in line with the assertion of Fagerberg (2004) who explained that every innova-tion need a new combinainnova-tion of existing capabilities, skills, and resources. Innovainnova-tion requires high levels of resources but the availability of knowledge and qualified human resources able to implement changes in the process of production is essential. According to (Cohen and

Levinthal,1989), innovation requires financial means to undertake expenditures for research and development which improve absorptive capacities and generate new knowledge and in-novations. All these results highlight the importance of introducing a more flexible function to take into account the complementarity and substitution effects between factors but also the threshold effects and the initial endowments of innovative factors, to assess the impact of policies, especially collaborative policies such as FPs.