Comparison of R&D expenditure to GDP ratio
At the world level, several countries have recognized the benets of supporting R&D investment and started to implement various direct and indirect public policies in order to stimulate private R&D and innovation. Most of the developed and developing countries have spent public resources to boost their innovation and economic system and increase the level of welfare. According to the OECD statistics presented in gure 1, over the 2002-2012 period, R&D intensity grew in the OECD countries (from 2.19% to 2.40%), in the EU-28 (from 1.76% to 1.97%), in the United States (from 2.55 to 2.79%) and in Japan (from 3.12% to 3.34%). Countries like Estonia (more than tripled, from 0.72% to 2.18%), Portugal (doubled), Slovenia (almost doubled), South Korea, Czech Republic and Turkey were the fastest growing OECD countries while in the same period, R&D intensity in China almost doubled, increasing from 1.07% 1.98% and surpassed the EU-28 for the rst time in 2012.
European innovation strategy
At the European level, over the last three decades, the EU has developed its systems of innovation with respect to the strategic orientation of its technology and innovation policies.
The main instruments of the EU are the Framework Programmes (FPs) on Research and Technological Development (RTD) which have funded thousands of collaborative R&D projects to support transnational cooperation and mobility for training purposes. Based on the Maastricht treaty, the FPs were implemented in 1984 to strengthen the scientic and technological bases of industry and to promote research activities (CORDIS, 2002)1. Since their implementation, the FPs have known strong rise in their budgets from e3.75 billion for the rst phase (FP1) to e80 billion for the Horizon 2020 program (FP8).
Moreover, through the Lisbon strategy at the European Council in 2000, the European Commission set a target of investing 3% of Gross domestic product (GDP) on R&D by 2010. However, the Eurostat statistics (gure 1) indicate that although the ratio of R&D expenditures to the GDP has increased in certain European countries (Estonia, Portugal, Slovenia for example), in several other member states, it has decreased (Sweden) or stagnated (France, United Kingdom). As consequence, the average R&D expenditure has increased slightly but has never reached 2% of the GDP in the EU-28 as a whole in this period.
Figure 2 shows the changes in Gross domestic expenditure on R&D to GDP ratio for EU28 compared to USA, Japan, China and OECD countries over the 1995-2012 period. According to the ocial data (OECD, 2018), in 2010, R&D amounted to 2.2% of GDP for the OECD as a whole and amounted only to 1.8% of GDP in average for the EU-28. Therefore, as shown by the gure 2, when making a comparison, we see that only Denmark, Finland, Israel, Japan, South Korea and Sweden were the countries whose R&D-to-GDP ratio exceeded 3%. On average, although some EU countries such as Finland, Sweden and Denmark were among the most performing in the world, the EU-28 as a whole performed lower than Japan, USA, and average OECD countries. Therefore, it may be understandable that the EU needs to increase its R&D eort in order to become more competitive and to catch up the most performing countries like Japan and USA for example.
French innovation policy strategy
In France, one of the largest economies in the EU, the government have spent a lot of public resources to support innovation. Indeed, France is one of the three European leading countries in terms of R&D volume, but have spent between 2% and 2.23% of its GDP over the 1995-2012 period to support innovation. Figure 1 shows that between 2002 and 2012, the French R&D-to-GDP ratio has stagnated (from 2.24 to 2.26%). Although France did not meet the 3% target set by the EU, French public R&D expenditures as a percentage of GDP is always above the EU average but remains below the levels of countries like Finland, Sweden, Denmark, and Germany. According to French Ministry of Research, in 2012, French funding of R&D was about e48.4 billion of which 41% of public contribution. In France, as in several other developed countries, there is a mix of several intervention instruments including the clusters policy implemented in 2004 to develop networking between companies and research and training organizations.
The economic and social context issues as well as the challenges of growth and sustainable development push public policymakers to support R&D and innovation. In struggling to improve the performance of innovation systems policymakers who spend a lot of public resources need to know about the eectiveness of the policies they implement.
The increasing interest in innovation and the importance of resources devoted to the R&D policies led to a great number of researches which addressed the evaluation of the eectiveness of public innovation policies using several econometric methods. The R&D policies impact evaluation remains subject to key methodological and empirical concern for economic researchers because the complexity of innovation systems and innovation process make it dicult to identify the real eects of a policy. The impacts found in the literature and their magnitude vary depending on the geographical scope, the data used and their level of aggregation, the estimation method and model specication (David et al., 2000; Cerulli and Pot, 2012).
The main objective of this thesis is to empirically investigate the eectiveness of public innovation policies. The basic idea that motives this thesis is that lot of public resources are spent at the European level but also in France to support R&D and innovation, and therefore evaluations are needed to know how much money should be invested, in what sectors and under what conditions to better reach the targets when supporting innovative activities. This thesis focuses on two specic policies, one of which was implemented at the European level and the other was implemented in France.
In the literature, many studies have used econometric techniques to analyze the eects of the EU Framework Programmes (FPs) policy on rms’ innovation and performance. Nevertheless, these studies have focused on micro-data (rm level) and the question asking if the FPs policy has positive eects on innovation at the macro-level (regional) has not been addressed. The rst purpose of this thesis is to analyze the impacts of the EU Fifth and Sixth Framework Programmes policy on the regional innovation of the EU-27 countries.
Furthermore, in the literature, there are few studies assessing the impacts of the French competitiveness clusters policy. Globally, these studies focus on the eects of this policy on the performance of small and medium-sized enterprises (SMEs) and on midsized rms (ETIs).
Their ndings suggest a rejection of the crowding-out hypothesis and suggest weak positive impact on innovation input additionality (private R&D and employment in R&D) but no substantial eects in terms of output additionality (innovation and market performance) are found. However, there is no study controlling at the same more than one instrument. Knowing that in France there is a mix of several public innovation instruments, it may make sense to control some instruments (clusters adhesion and FUI projects participation) when evaluating the eects of the clusters policy. Indeed, the lack of conclusive results on the eectiveness of the policy may be partly attributable to the lack of adequate data and methodology but also to the simultaneity of several instruments of innovation policies. Moreover, we see that almost all the dierent studies partly disregard or remove systematically from their evaluations the large rms which undertake a huge share of total R&D spending (48.5% in 2012) and which are the main beneciary of subsidies. Therefore, the second purpose of this thesis is to examine and better understand the eectiveness of the French Competitiveness clusters policy on the innovation and performance of small and medium-sized enterprises, midsized rms but also large rms by controlling two innovation instruments.
Because of the two main purposes of the thesis, it makes sense to separate the research questions into specic research questions. With respect to the two dierent policies analyzed in this thesis, we have two sets of main research questions addressed.
The rst set of main research questions addressed in this thesis and which focus on the EU innovation policy are: Have the EU Fifth and Sixth Framework Programmes impacted positively the European regional innovation? How the eects, if any, of these programs vary between leading and lagging EU countries? What are the improvements that need to be done in order to make the policy more eective?
The second set of main research questions addressed in this thesis and which focus on the French clusters policy are: What are the factors that determine the participation of rms in the clusters policy? What are the impacts of clusters adhesion on rms’ performance? What are the impacts of FUI projects participation on rms’ performance? Do rms perform better when they jointly belong to clusters and participate in projects? Do the eects of the policy dier between SMEs and ETIs and large rms. What are the improvements that need to be done in order to make the policy more eective?
In this thesis, I use a variety of empirical models in order to answer the dierent research questions. In the following work we use the new development of empirical models and particularly natural experiments studies approach, in order to evaluate the impacts of innovation policies at both macro-level (European regions) and micro-level (French SMEs, ETIs, and large rms).
To answer the rst set of main research questions, we use macro-data to evaluate, in one complete empirical chapter, how the two phases (Fifth and Sixth) of the EU Framework Programmes policy have impacted the regional innovation of the EU-27 countries. we also account for the simultaneity of the two instruments.
Thereafter, to answer the second set of research questions, we use micro-data to evaluate the eects of the French competitiveness clusters policy on rms’ innovation and performance (input and output additionality). To do so, because of methodological issues, we divide the work into two dierent chapters. The rst chapter as commonly done in the literature evaluates the eects of the clusters policy on small and medium-sized enterprises’ innovation and performance (input and output additionality). The other chapter goes further and proposes an adequate methodology to bring a deep analysis of the eectiveness of this clusters policy on midsized and large rms’ performance in terms of incentives for private R&D investments, innovation, job creation and market competitiveness. As in the rst empirical chapter, in these two empirical chapters, we consider two policy instruments (adhesion to clusters and participation in FUI projects), account for their simultaneity and compare them.
The thesis is organized into four chapters, one of which is reserved for the theoretical and empirical literature review of innovation policy evaluations and the three others are empirical contributions to public innovation policy evaluations.
The rst chapter presents a literature review of public innovation policies and examines the empirical evidence on the eectiveness of public interventions to support R&D and innovation.
We review the economics of innovation literature by taking theoretical and empirical perspectives on the relation of innovation and economic growth, innovation and technical change, and innovation and rms. We discuss some indicators used to measure innovation.
Moreover, the chapter gives an overview of the theoretical rationales of public intervention to support private R&D and innovation and the main public intervention instruments. Finally, we review the major econometric method of evaluation of public innovation policies and present a literature review on the evidence of several public intervention instruments stimulating private R&D spending and innovation. The empirical literature overview highlights the great heterogeneity in results of empirical studies that tried to evaluate the eects of public support for R&D and there is no consensus in the literature. Although some crowding-out effects have been found in former studies and particularly in the USA, recent empirical evidence on input and output additionality at the rm level suggests that R&D subsidies may mostly stimulate private R&D investment and positively impact innovation outcomes and rms’ performance.
However, the studies analyzing the eectiveness of cluster policies on the rms’ R&D and performance are mixed and non-conclusive. The results found in the literature vary depending on the geographical scope, the data used and their level of aggregation, and the estimation method and model specication (David et al., 2000; Cerulli and Pot, 2012).
As explained by Cerulli and Pot (2012), even if the majority of models focus on testing private R&D additionality, much attention should be devoted to the eects of R&D eort on rms’ performance. In general, the literature conrms the existence of a positive relationship between innovation policies and rms’ innovation, but the eect on economic performance is not so evident. In the following chapters, we use the new development of empirical models and particularly natural experiments studies to evaluate the impacts of innovation policies on private R&D and rms’ performance. Our contribution to this on-going R&D and innovation policy evaluation consists of empirical analyses conducted in three studies.
In the chapter 2, we use macro-data (regional panel) covering the 1995-2012 period and 218 regions of the entire EU-27 and adopt a knowledge production function (KPF) with a translog function specication. We use a random trend model specication that controls for all the unobserved heterogeneity of regions that can aect innovation. It is a panel model approaches which deal with unobservables and endogeneity to study the eects of the European Union Framework Programmes (FPs) policy on regional innovation. This study contributes to the literature in terms of methodology and ndings. In terms of methodology, dierently from other studies which generally use rm-level data to evaluate the eects of the FPs policy, it proposes the use of macro-level data and a relevant macroeconomic method controlling for main innovation inputs, i.e. human capital (HK) and R&D and the FP5 and FP6 spending, as well as main unobserved factors that may aect the innovation process. The use a translog function specication for the production function allows taking into account the complementarity and substitution eects between factors but also the threshold eects and the initial endowments of innovative factors. In terms of ndings, the results bring new evidence on the impacts of the FP5 and FP6 programmes on innovation output at the macro-level in EU-27. Moreover, results reveal for complementarity but also for substitution between factors. Further, a comparison of the impacts between leading countries (EU top performer countries) and lagging countries (EU low 16) shows strong heterogeneity of these results.
Table of contents :
1 Public innovation policy evaluation: theoretical and empirical literature
1.2 The economy of innovation
1.2.1 Denition of innovation
1.2.2 Innovation and growth
1.2.3 Innovation and rms
1.2.4 Geography of innovation and knowledge spillovers
1.2.5 Indicators and measures of innovation
1.3 Rationales and instruments for public innovation policies
1.3.1 Rationales for public innovation policies
188.8.131.52 Markets failures related to spillovers
184.108.40.206 Other specic markets failures
220.127.116.11 Failures related to the innovation system
1.3.2 Innovation policies instruments
18.104.22.168 R&D direct subsidies (grants or funds)
22.214.171.124 R&D tax incentives
126.96.36.199 Collaborative R&D and innovation policies
1.3.3 Possible ineciency and risks of public intervention
1.4 Econometric evaluation of innovation policies
1.4.1 Evaluation challenges, counterfactual and selection bias
1.4.2 Econometric models
1.5 Empirical evidence of public innovation policies
1.5.1 Impacts of R&D subsidies on private R&D and innovation
1.5.2 Impacts of R&D tax incentives on rms innovation
1.5.3 Impacts of cluster policies on rms innovation
1.6 Concluding remarks
2 Evidence on the impact of the 5th and 6th Framework Programmes on regional innovation
2.2 The EU Framework Programmes policy
2.2.1 The Fifth Framework Programme
2.2.2 The Sixth Framework Programme
2.3 Related literature
2.4.1 The knowledge production function
2.4.2 The Cobb-Douglas production function
2.4.3 The translog production function
2.4.4 The econometric specication: a random trend model
2.5.1 Sources and variables
2.5.2 Descriptive statistics
2.6.1 Results with the Cobb-Douglas function
2.6.2 Results with the translog function
2.8 Appendices Chapter 2
2.8.1 Descriptive statistics by country
2.8.2 Tables of estimation results
2.8.3 Calculated Elasticities
3 Assessing the impacts of the French competitiveness clusters policy on SMEs’ Performance
3.2 The French competitiveness clusters policy
3.2.1 Denition and implementation of the policy
3.2.2 Funding and budgets of the French policy mix
188.8.131.52 Direct subsidies
184.108.40.206 Indirect subsidies
3.3 Related literature
3.3.1 Eects of the clusters policy on input additionality
3.3.2 Eects of the clusters policy on output additionality
3.4.1 Quasi-experimental design
3.4.2 Econometric strategy
3.5.1 Data sources and variables
3.5.2 Descriptive statistics
3.6.1 Estimated propensities to participate in the policy
3.6.2 Balancing rms’ characteristics before/after matching
3.6.3 Average treatment eects of the policy on rms’ outcomes
220.127.116.11 Eects on innovation input additionality
18.104.22.168 Eects on output additionality
3.6.4 Sensitivity analysis
3.8 Appendices Chapter 3
3.8.1 Literature review
3.8.2 Descriptive statistics
4 Impacts of the competitiveness clusters policy on midsized and large rms’ performance
4.2.1 Data sources and variables
4.2.2 Descriptive statistics
22.214.171.124 Sample structure: ETIs and large rms
126.96.36.199 Dierences between treated and nontreated rms
188.8.131.52 Heterogeneity according to type of cluster
184.108.40.206 Dierences in the number of years a rm participated
4.3 Econometric strategy
4.3.1 The two-way xed eects model
4.3.2 Modeling the heterogeneity of the policy eects
4.3.3 Testing the validity of the model
4.4.1 Eects on innovation and economic performance
220.127.116.11 Mixed eects on innovation input and output additionality
18.104.22.168 Strong positive eects on employment
22.214.171.124 Mixed eects on economic performance
4.4.2 Heterogeneity of eects
126.96.36.199 Heterogeneity of eects according to type of cluster
188.8.131.52 Heterogeneity of eects over years
4.6 Appendices Chapter 4
4.6.1 Descriptive statistics
4.6.2 Typology of competitiveness clusters
4.6.3 Estimation tables without control variables
Conclusions and discussion
List of gures
List of tables