همکاری با موسسات تحقیقاتی دولتی و موفقیت در نوآوری : شواهدی از فرانسه و آلمان
|کد مقاله||سال انتشار||تعداد صفحات مقاله انگلیسی||ترجمه فارسی|
|2355||2013||18 صفحه PDF||سفارش دهید|
نسخه انگلیسی مقاله همین الان قابل دانلود است.
هزینه ترجمه مقاله بر اساس تعداد کلمات مقاله انگلیسی محاسبه می شود.
این مقاله تقریباً شامل 16829 کلمه می باشد.
هزینه ترجمه مقاله توسط مترجمان با تجربه، طبق جدول زیر محاسبه می شود:
|شرح||تعرفه ترجمه||زمان تحویل||جمع هزینه|
|ترجمه تخصصی - سرعت عادی||هر کلمه 90 تومان||23 روز بعد از پرداخت||1,514,610 تومان|
|ترجمه تخصصی - سرعت فوری||هر کلمه 180 تومان||12 روز بعد از پرداخت||3,029,220 تومان|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Research Policy, Volume 42, Issue 1, February 2013, Pages 149–166
We evaluate the impact of cooperation with public research on firms’ product and process innovations in France and Germany using Community Innovation Survey data from 2004 and 2008. We find that cooperating with public research increases product innovation, but has no effect on process innovation, which depends more on firms’ openness. Our benchmark estimates, which are very similar in 2004 and 2008, suggest that the increase in product innovation is much higher in Germany than in France. Endogeneity tests show that the French benchmark estimate may be somewhat biased in 2004 but not in 2008, which hints at a persistent gap in the effect of cooperation between France and Germany. We derive two important policy implications from our results. First, public–private collaborations in research should not be encouraged at all costs, since they may not sustain all forms of innovation. Second, the changes in the institutional context of public-private partnerships in research which have been implemented between 2004 and 2008 have yet to prove effective in sustaining the innovation impact of cooperation.
Modern societies supposedly base their wealth on a steadily increasing and widely accessible knowledge base. This implies that new knowledge needs not only to be discovered, but also to be diffused, i.e. it ought to be made readily available to the society, which will then be able to generate value from it. Most lines of research agree on the fact that interactions between industry and science are among the most prominent institutional interfaces for knowledge diffusion. This paper focuses on formal collaborations between firms and public research institutions, and examines their impact on the innovativeness of firms using French and German data from recent Community Innovation Surveys (CIS). Our paper offers several contributions to the literature. First, we propose a detailed comparison of the institutional contexts of cooperation with science in France and in Germany, taking recent changes into account. Second, using recent data, we develop an empirical analysis that considers both product innovation and process innovation, whereas most previous studies only focus on the former. Moreover, our econometric methodology can address both selection and endogeneity issues, which has not been done in related previous studies. In addition, we will extend our main analysis to examine the specifics of the manufacturing and services industries. We will also, as far as our data sources allow, try to grasp the dynamics of the phenomenon we study and try to assess its impact on productivity, in order to derive more policy implications. The remainder of the paper is organised as follows. In Section 2, we state the objective of our research, we discuss the interest and feasibility of an institutional comparison between France and Germany and we sketch our conceptual framework. In Section 3, we present our estimation strategy and choice of variables. Section 4 is dedicated to the presentation and discussion of the results. Building on these results, Section 5 proposes further explorations, from which we can derive more policy implications. We conclude in a final section.
نتیجه گیری انگلیسی
4.1. Product innovation We start the discussion of our results with those pertaining to product innovation. In accordance with our estimation strategy, we first comment the estimates of our benchmark Generalized Tobit model. These estimates are presented in Table 2. We can first remark that the estimated coefficient associated with the inverse Mills ratio (see Appendix A for a formal definition) is significantly different from zero in both countries. This means that correcting for sample selection bias was necessary, and that the corrected estimates are likely to be free from this specific bias. For the sake of concision, we now focus our comments on the main result, i.e. the effect of cooperation with public research institutions.As can be seen in Table 2, column (II), the estimated effect of cooperation on the intensity of product innovation is significantly positive both in France and in Germany. The coefficient is larger in Germany than in France. This is also the case for the marginal effect, which is equal to 2% (both at the sample mean and at the sample median) in France and to 5% in Germany (again, both at the sample mean and at the sample median). These marginal effects are all (like the estimated coefficients) significant at the 5% level. Economically, these values mean that cooperating with a public research institution entails an increase of 2 percentage points (pp) in the share of sales due to new products in France, and an increase of 5 pp in Germany. This model corroborates the suspicion (raised in Section 2.2) that the German KTT system might be more effective than the French one. The next step in our analysis consists in determining whether our benchmark estimates are likely to be biased by endogeneity, and therefore whether it is necessary to estimate the Heckit model with endogeneity correction. We thus conducted standard endogeneity tests in the ordinary Heckman-type model (see Wooldridge, 2002). The p-values of these tests, reported under Table 2, suggest that our estimate of the effect of cooperation with public research is likely to be biased by endogeneity in France, but not in Germany. Thus, estimating the endogeneity corrected models is relevant, at least on the French data. Table 3 present the results of the Heckman model with endogeneity correction. At first glance, these results qualitatively confirm those of our benchmark model. After correcting for endogeneity, the estimated effect of collaborations with public research institutions remains significantly positive in France and in Germany. With respect to a quantitative interpretation we prefer to remain cautious, because intuition tells us that endogeneity should bias our benchmark estimates downwards. The reason for this is simply that more innovative firms are also more likely to cooperate without implying any causal relationship. Thus, an endogeneity-corrected estimate should be lower, but we observe the opposite in our results. This parameter inflation is often encountered with IV-methods in finite samples. The problem is known to occur even for instruments that pass the standard over-identification and weak instrument tests (such as those we used when applying our model to the French data).11As a consequence, we do not use the models with endogeneity correction to derive precise quantitative estimates. We simply use these models to check whether, after correcting for endogeneity, the previously observed positive effect of cooperation with public research remains significant. Table 3 indicates that it is actually the case. Furthermore, we will see in Section 5 that, when using CIS 2008 data, the endogeneity issue vanishes even in the French sample, while a substantial difference in the estimated effects for Germany and France remains. This suggests that the marginal effects of 2 pp for France and 5 pp for Germany derived from the benchmark Heckman-type models may be the most reasonable estimates we can get with CIS4 data, even though they may be biased by some degree of endogeneity in France. Exploring the endogeneity issue, however, should be left to other studies that may have more powerful instruments at their disposal, such as quasi-natural policy experiments. Besides our main result, two secondary results also deserve to be mentioned. First of all, according to the benchmark model, cooperating with other partners has no significant impact on the share of innovative sales in France or in Germany. In the French sample, the model with endogeneity correction suggests that collaborations with other partners might even have a negative effect on the intensity of product innovation. However, looking at the selection equations in Table 2 and Table 3 reveals that these collaborations are positively associated with the decision to engage in product innovation or not. More precisely, collaborating with other partners (e.g., customers, competitors, firms within the group or in the supply chain) significantly increases the probability of being a product innovator. In the light of our main result, we interpret this as a sign that cooperation with other partners may primarily provide an “entry ticket” to product innovation, whereas cooperation with public research institutions may be a catalyst to make it more successful. The second interesting result concerns the effect of our indicator of openness. In the benchmark model, the openness of the company is positively associated with the probability to innovate in product, both in France and in Germany, but has no significant effect on the intensity of product innovation. In France, the model with endogeneity correction suggests that openness is indeed positively associated with the probability of a product innovation, and could be negatively associated with the intensity of product innovation. Interestingly, however, openness is, in this model, positively associated with the probability to cooperate with public research institutions. This is consistent with Laursen and Salter (2004), who find that more “open” firms are also more likely to rely on academic knowledge to innovate. 4.2. Process innovation We now discuss the results concerning process innovation, obtained using the same estimation strategy as in Section 4.1. In both models (with or without endogeneity correction), the dependent variable of the selection equation is now an indicator of process innovation. The intensity of process innovation is built out of the four proxies presented at the end of Section 3.1. We conducted a factor analysis on these variables, and found out that they all loaded to a single factor (the first component) with an eigenvalue above one (we report the factor loadings in Table 4a). This implies a single underlying (or “latent”) continuous variable, which we can use as a measure of the intensity of process innovation. In practice, we built this variable using the scores of the first factor, and we used it as the dependent variable in the intensity equation of our econometric models.There are at least two advantages to this intensity variable. First, it provides a single general measure of process innovation, whereas each proxy only captures one of its aspects. Second, the continuous nature of the variable allows us to rely on the same econometric models as those we used for the analysis of product innovation. The only disadvantage is that the effect of the explanatory variables cannot be interpreted quantitatively (using economically relevant units such as percentage points, for instance) as was done with our measure of product innovation (the share of innovative sales). In Table 4b, we present the estimates of the benchmark model without correcting for the potential endogeneity of cooperation with public research. According to these estimates, neither in France nor in Germany does cooperation with public research institutions have any impact on the intensity of process innovation. In France, though, cooperating with other partners is associated with a higher intensity of process innovation, which is not the case in Germany. By contrast, in both countries, the degree of openness of the firm is positively associated with both the probability to introduce a process innovation and the intensity of process innovation.Further tests, reported under Table 4b, indicate that the estimates are not plagued by endogeneity issues. We will therefore neither present nor discuss the results of the Heckit model with endogeneity correction (which are available upon request). Comparing our results on process innovation with those obtained in Section 4.1 on product innovation, we can conclude that cooperating with public research does enhance the latter, but not the former. The degree of openness is, by far, a more important determinant of process innovation than formal cooperation, both in France and in Germany – although cooperation with other partners also tends to increase the intensity of process innovation in France.12 Intuitively, these results do make sense. Indeed, actively seeking feedback from suppliers and customers (and taking it seriously) may be a good strategy to improve the production process, thus generating process innovations. By contrast, the development of a new product may require external resources which are better acquired through cooperation with academic researchers. As a robustness check, we examined the impact of cooperation with public research on each proxy variable rather than on our continuous measure of the intensity of process innovation. We estimated by Maximum Likelihood a variant of our Heckman-type model in which the intensity equation is specified as an Ordered Probit model (to take into account the categorical nature of the proxy variables). The results of these estimations (which are available upon request) confirmed the general conclusion outlined in the previous paragraph. 4.3. Difference across manufacturing and services In Section 4.1, we found that cooperating with public research increases the intensity of product innovation in France and in Germany (where the increase might be higher). By contrast, in Section 4.2, we found that this form of cooperation has no effect on process innovation – which seems to be more influenced by firms’ “openness”. These findings have been derived from estimates obtained on the whole sample in each country. We will now look for differences across sectors by conducting separate estimations in the manufacturing industry on the one hand and in the services on the other.13Table 5 summarizes the main results of this sector-specific analysis (full results are available upon request).We first re-examine the results regarding product innovation. The benchmark model suggests that in France, cooperating with public research increases product innovation in the services only, whereas in Germany it increases product innovation in the manufacturing industry only. However, tests reveal that the benchmark model is likely to suffer from endogeneity biases in France, in both the manufacturing and the services samples, whereas this is never the case in Germany. Estimating the model with endogeneity correction in France, we find that cooperating with public research actually increases the intensity of product innovation in the manufacturing industry as well as in the services. These results might be due to the fact that the French economy has become more and more service-driven over the recent period, whereas the manufacturing industry remains a cornerstone of the German economy. In a nutshell, as far as product innovation is concerned, the results of the sensitivity analysis refine and confirm those obtained in Section 4.1: cooperating with public research institutions increases the intensity of product innovation in France and in Germany. In Germany, this increase is certainly driven by the manufacturing industry only. In France, by contrast, it is likely to be driven by both the manufacturing industry and the services. We now re-examine the results concerning process innovation, using the continuous measure of innovation intensity described in Section 4.2. According to the benchmark model, the effect of cooperation with public research on the intensity of process innovation is never significant. This finding holds for both countries and both industries, except in the services in Germany, where a negative effect can be observed. This effect is only significant at the 10% level, though, and can thus be safely and reasonably neglected. By contrast, the openness of a firm is always positively associated with the intensity of process innovation, in both industries, and in both countries. As was already the case in the main analysis, we did not detect any endogeneity bias in the sensitivity analysis. Overall, these findings simply confirm those of Section 4.2.