Printer Friendly

Framework Conditions, Innovation and Productivity in European Regions.

Introduction

Investment in innovation is recognised as a key driver of productivity, long-term prosperity and economic growth. Innovation is expected to help address pressing economic and social challenges--including an ageing population, climate change, and numerous health and environmental issues. New products, services and processes will have to be developed, creating growth opportunities for firms as well as new skill needs and job opportunities for workers (OECD 2016). Economic output can increase with investment in physical and human capital. However, in the long run, income per capita rises with innovations that make physical and human capital more productive.

Because of the non-rival character of new ideas and increasing returns to the accumulation of knowledge (Romer 1990), firms are a natural target of the public policy mix to support innovation. In particular, corporate investment in R&D and intangible assets (including software and databases, training of employees or organisation and business processes improvements) is encouraged as it is associated with firm innovation performance and productivity (Haskel and Westlake 2017).

However, innovation is not the outcome of a linear input-output process and it is not homogeneous across regions within a country. The emergence of areas or regions that are particularly successful in innovation-led growth suggests that local characteristics, beyond the innovative capacity of firms, can be important (Kenney 2000; Saxenian 1990). R&D efforts are translated into economic and social development through a complex mechanism, where pro-innovation culture, efficient regulatory environment and institutions play a crucial role (albeit mostly implicit) to support local innovation output and, consequently, the competitive advantage of territories or innovation clusters. Various authors have stressed the importance of different concepts for regional innovation policy, such as innovation systems (Cooke et al. 1997), innovative milieux (Camagni and Capello 2002), learning regions (Morgan 2007) and social filters (Crescenzi and Rodriguez-Pose 2009).

Economic theory suggests that investments in innovation should target less developed territories, as returns to innovation should be higher. However, firms typically invest less in R&D in countries or regions that are far from the knowledge frontier. This has been referred to in the literature as the innovation paradox (Cirera and Maloney 2017). While innovation performance can be explained by many different factors, one reason why returns can be lower in less developed regions is the lack of local complementary factors (such as regulatory and framework conditions, quality of institutions and infrastructure), highlighting the inter-connected role played by the different components of innovation systems (Nelson and Rosenberg 1993).

An environment that facilitates the development and use of new ideas is thus expected to enhance the returns from investment in innovation. Good governance and favourable business conditions at the national and regional level are considered to be basic ingredients for economic development and prosperity (Holmberg et al. 2009; Mo 2001; Charron et al. 2014). Acemoglu et al. (2005) argue that institutions influence firm performance and innovation attitudes and shape the development path of countries.

In Europe, firms located in regions with lower levels of development tend to invest less in innovation and focus less on new products (and products have typically less innovation content), while firms in more prosperous regions, instead, focus more on product development and intangible investment (Bubbico and Wruuck 2018). This can exacerbate existing regional disparities, in a context where convergence in Europe has been slowing down in recent years (European Commission 2017a) and technological evolution can lead to win-takes-all-dynamics well beyond the ICT sector.

Understanding how firms create and adopt innovations is important for the design and implementation of effective public support. The low quality of framework and regulatory conditions in some areas of the EU can be major factors in reducing economic and social cohesion. Structural reforms impact regions in different ways: product market reforms have been shown to affect negatively some lagging regions of the EU (European Commission 2017a, b). In many countries, less developed regions have weaker capacity to absorb innovation and innovation-related investments--on the back of poor or outdated industrial capacity, industrial structure and firm size distribution (e.g. due to the prevalence of low value-added activities), as well as low quality of public institutions, with limited capability to create and disseminate knowledge (Oughton et al. 2002).

Rodriguez-Pose and Di Cataldo (2014) argue that lagging European regions lack the institutional infrastructure needed to attract R&D investment and make investments in technology pay off. Using data on four EU countries, Farole et al. (2017) find that the business environment affects firm performance and penalise firms located in less developed regions. Institutional factors are driven by national policies that are often spatially blind but play out in different ways across regions. However, the lack of comparable EU-wide data on regional institutional performance (on the one hand) and firm innovation performance and investment activities (on the other) has limited research on this issue.

The recent development of the European regional quality of government index complements national measures of the business environment (such as World Bank's Ease of Doing Business ranking) with sub-national measures (Charron et al. 2014, 2016). (1) This has triggered novel regional analysis of the nexus between quality of institutions and returns to investment in innovation (Rodriguez-Pose and Di Cataldo 2014), returns on infrastructure investment (Crescenzi et al. 2016), regional competitiveness (Annoni et al. 2017) and regional development (Ketterer and RodriguezPose 2018).

Regional framework conditions refer to both hard and soft inputs required to increase the innovation generation and absorption capacity of regions, conducive to productivity improvement. The hard regional framework is made of measurable components such as education and infrastructure. However, the systemic view of innovation also emphasises the role of soft elements, including the quality of local institutions and governance, exchange of knowledge between industry, services and research institutions and the adaptation of the regulatory environment. In this context, soft elements are important as innovation policies are typically more complex than traditional development policies. This is due to epistemic conditions (uncertainty and information asymmetry in innovation processes), governance (multi-actor governance) and organisational features (complexity of innovation systems).

This paper explores the links between regional framework conditions, innovation inputs, firm innovation performance and regional productivity growth (Fig. 1). We use a combination of macroeconomic and microeconomic data, ranging from firm-level data of the EIB Investment Survey (EIBIS), regional data of the European Commission on regional competitiveness of institutions (combining the Ease of Doing Business Index and the European Quality of Government Index) and Eurostat regional statistics. The design and implementation of EIBIS is consistent across countries and sectors, which is critical for understanding variation in firm innovation activities across regions in Europe.

We show that framework conditions and firm innovation inputs affect innovation performance and regional productivity growth. We argue that, in addition to R&D investments and the education level of the labour force employed in science and the tech sector, improvements in the regional business environment--such as the quality of institutions and infrastructure quality--can help firms close the gap with the technological frontier and support the economic development of regions in the EU.

The remainder of this paper is organised as follows. The next section presents a picture of innovation efforts and performance in EU regions. The "Linking Innovation Efforts to Innovation Output: Analysis of Patenting and Innovating Firms in Europe" section explores the relationship between firm innovation inputs and innovation output and underlines the importance of framework conditions for innovation performance. "The Contribution of Enabling Factors to Regional Performance" section discusses the links between regional innovation inputs, business environment conditions and productivity growth of EU regions. The last section is a conclusion with implications for regional innovation policy in Europe.

The Innovation Landscape in the EU

The EU is investing less in R&D as a share of GDP than other major global players, such as the US or China, with potential negative implications for innovation and long-term growth (Fig. 2). While R&D expenditures have increased in most EU countries over the last decade, the convergence in R&D investment intensity across countries was mostly driven by the fall in R&D intensity in the top-performing countries (such as Finland and Sweden) after 2008.

The annual R&D investment gap in the EU is estimated at EUR 140 billion, based on the target of 3% of GDP spending on R&D. Gross domestic expenditures on R&D in the EU 28 were on average EUR 295 billion per year during 2014-2016, corresponding to 2% of GDP. While R&D expenditures have been resilient and have continued to grow throughout the crisis period in Europe (unlike other components of gross fixed capital formation), the R&D investment gap remains significant.

With regard to regional concentration of innovation, Germany represented 30% of total EU R&D expenditures in 2016, followed by France (17%) and the UK (13%). Focusing at the sub-national scale, innovation is even more concentrated geographically, with 50% of total patents and R&D expenditure, and around 30% of employed in science and technology concentrated in only 10% of regions (Fig. 3). Most of these regions are top-performing regions within their countries, corresponding in most cases to capitals. The intra-country dispersion in macroeconomic performance has been persistent over time in Europe (Iammarino et al. 2018; Borsi and Metiu 2015), in line with R&D expenditure dispersion across regions within countries (Table 1).

The link between innovation efforts and innovation output can be explored using firm-level data of the EIB Investment Survey (EIBIS). EIBIS is an EU-wide firmlevel survey on investment dynamics and obstacles and is administered to a stratified random sample of firms in each of the 28 countries of the EU. (2) In EIBIS, firms report whether they invest to develop or introduce new products, processes or services; the new products can be new to the company, new to the country or new to the global market. Firms also report whether they invested in R&D (including the acquisition of intellectual property).

Firms can be classified into five different innovation profiles based on R&D investment and innovation activities (EIB 2017). The five innovation profiles consist of basic firms, adopting firms, incremental innovators, leading innovators and developers. Basic firms do not invest in R&D and do not introduce new products. Adopting firms do not invest in R&D but invest in developing or introducing new products, processes or services. Incremental innovators have substantial R&D investment and introduce products, processes or services that are new to the company (but not new to the country or the global market). Leading innovators also invest in R&D and introduce products that are new to the country or the global market. Finally, developers have substantial R&D investment but do not (yet) introduce new products.

As the database of EIBIS includes location information, it is possible to match firms' innovation profile to regions. Following the approach underlying EU cohesion policy, we classify NUTS 2 regions into three groups, based on the level of regional development: (1) less developed: regions with GDP per capita below 75% of the EU average; (2) transition: regions with GDP per capita 75%-90% of the EU average; (3) more developed: regions with GDP per capita above 90% of the EU average. (3)

Using EIBIS data, we find that innovative firms (i.e. leading innovators) are more likely than non-innovative firms to invest to replace capacity, in particular in less developed and transition regions (Fig. 4, left panel). Innovative firms are also much more likely to invest in capacity expansion (Fig. 4, right panel). Because they are more likely to invest, this indicates that innovative firms can be a source of jobs, growth and investment opportunities, notably in less developed regions.

Leading innovators tend to rely more on internal sources to finance investment; they are also more likely to report that they are finance-constrained. In the EU, internal funds represent 65% of investment finance for innovative firms, compared with 60% for non-innovative firms (Fig. 5, left panel). Poor availability of external finance could be an explanation for the higher share of firms that report being finance-constrained in less developed regions (Fig. 5, right panel). Financial services tend to be geographically concentrated, and finance availability depends also on proximity. At the same time, highly innovative companies in more developed regions may be reluctant to share information in fear of leakages of private information needed for an external financier, which may explain why they have a higher share of internal finance.

In addition to R&D, other types of intangible assets--including software and databases, training of employees and organisational capital--can be important sources of firm performance. In the USA and several EU countries (including Sweden, the UK and Finland), investment in intangibles represents a large part of gross fixed capital formation, which exceeds that in tangible assets (EIB 2016). The rising importance of intangible capital has also been associated with structural features of advanced economies, in particular slow productivity growth and rising inequality over the past 2 decades (Haskel and Westlake 2017).

Firms that allocate a greater share of investment to intangibles tend to innovate more (Fig. 6, left panel). They are more likely to develop or introduce new products, processes or services. R&D investment is the main driver of this positive correlation between intangible assets and the introduction of new products, processes or services. However, investment in software and databases, and in organisation and business process improvements matter as well. This emphasises the importance of complementarity across intangible assets for firm innovation, suggesting that public policies aiming to support innovation in the EU should not only promote R&D investment (EIB 2018b).

Intangible firms tend to be more productive, in particular in other EU countries (Fig. 6, right panel). The association between total factor productivity and intangible investment is also positive in the periphery and cohesion countries, but it is slightly weaker than in other EU countries. (4) In line with the schematic diagram in Fig. 1, Fig. 6 represents the links of innovation inputs with innovation output and performance using firm-level analysis.

A substantial part of the differences in income per capita across countries in the EU is due to differences in firm-level productivity (Fig. 7, left panel). When considering a large set of countries, including both developing and advanced economies, total factor productivity accounts for more than 60% of the variation in GDP per capita across countries (Jones 2016). More innovative firms are based in more productive economies. Firm productivity is associated with innovation activities, but the correlation at the country level is weaker than that with GDP per capita (Fig. 7, right panel).

A similar pattern also holds at regional level, with a wide dispersion in innovative output at all levels of productivity. Firms in cohesion and periphery countries tend to have lower productivity than in other EU regions total factor productivity (Fig. 8, left panel). At the same time, investment in innovation displays high variation within each country group (Fig. 8, right panel).

In line with the schematic diagram in Fig. 1, the right panel in Figs. 7 and 8 represents the links between innovation output and economic performance using firm-level and regional analysis. This suggests that there are critical elements in the institutional environment that can affect the success and performance of innovative firms. These include factors such as product and labour market regulations, the protection of intellectual property, the development of capital markets to finance innovation and the nature of the complementarities between investments in physical capital (such as machinery and equipment) and intangible assets (such as R&D, software and databases, organisation and business process improvements).

Linking Innovation Efforts to Innovation Output: Analysis of Patenting and Innovating Firms in Europe

To analyse the link between innovation efforts and firms' innovation output, we use a standard knowledge production function, considering the firm as the unit of analysis. The firm-level knowledge production function consists of two sets of variables: regional factors and firm-level explanatory variables:

[Z.sub.it] = g(H[K.sub.rt]; [F.sub.rt]; I[N.sub.it]; [U.sub.it]) (1)

where H[K.sub.rt] refers to the availability of human capital in regions r in year t, [F.sub.rt] denotes regional framework conditions, I[N.sub.it] inputs to innovation of firm i in year t (such as investment in intangible assets) and other firm-level variables, and [U.sub.it] captures firm-level unobservable characteristics. The dependent variable [Z.sub.it] is a binary response, and we use two measures of innovation outcomes: innovation activity and patenting.

When the dependent variable is innovation activity, the analysis is based on the first two waves of EIBIS. When we consider patenting activity as the dependent variable instead, we use a panel dataset of firms that participated in EIBIS by relying on Orbis variables matched to the EIBIS database. In other words, we consider two different measures of firm innovation performance, assessing both a shorter and a longer time period, and use both regional and firm-level explanatory variables.

The reason why we use two different types of innovation output is that they capture distinct aspects of innovation. While R&D expenditures are typically used as a proxy for innovation input, patenting activity is widely used as a measure of innovation outcomes. Patents are preceded by the basic and applied research phases within a firm, and they can be followed by a development process that aims to introduce new products, processes or services on the market. Focusing on the end of the two stages of the development process, the considered measures of innovation output are (1) the patent applications emerging at the end of the research phase and (2) whether a firm self-assesses as being an innovator, i.e. having introduced or developed a new product, process or service.

Innovation output (new products, processes or services) is widespread across all sectors. Patenting activity, however, is less broad-based. Manufacturing firms are much more likely to engage in patenting activities than firms operating in other sectors (Fig. 9). This is also consistent with the evidence that manufacturing firms tend to conduct more R&D (EIB 2018b). The share of firms that patent is lower in the periphery and cohesion countries than in other EU countries. The gap between these three groups of countries has narrowed slowly over the past 10 years.

Patenting firms tend to be more competitive and productive than non-patenting ones. According to EIBIS data, they tend to be larger firms and show higher labour productivity and a higher level of capacity utilisation. They are also more likely to export their goods or services and to invest abroad. Clearly, they are also much more likely to introduce new products, processes or services. In addition, patenting firms tend to report a greater share of machinery and equipment that is state of the art, and they outperform non-patenting firms in terms of productivity performance (measured by using either labour productivity or total factor productivity).

Patenting firms show better financial health indicators. According to Orbis data, patenting firms have higher levels of liquidity (proxied by the current ratio) than firms that do not patent. (5) They are also slightly less leveraged (using the ratio of loans plus long-term debts to total assets) and as good as non-patenting firms in terms of their profitability (captured by return on assets).

There is considerable regional variation in patenting activities within EU countries, notably in France, Italy and Spain. Patenting remains relatively weak in periphery and cohesion countries (Fig. 10). However, there are notable exceptions (e.g. several regions in Poland have good patenting performance, while some regions in Germany show low patenting activity).

We complement firm-level data with regional structural indicators to explore how innovation efforts of firms are translated into innovation output depending on regional enabling factors. There is considerable heterogeneity in factors characterising the environment in which firms operate across regions within EU countries. Table 2 reports the estimates of a panel probit regression analysis using the two first waves of EIBIS. The indicator of innovation output (whether firms introduced new products, processes or services) serves as the dependent variable. Since EIBIS includes information on investment in different types of intangible assets (as a share of total investment), this provides an opportunity to analyse their impact on innovative output.

The results in Table 2 highlight the importance of innovation inputs (investment in R&D, software and data as well as organisation and business process improvements). Investment in training of employees (as a share of total investment) is negatively associated with innovation activities, suggesting that there could be a crowding-out effect of training expenses of firms (as firms have limited resources to invest in the various forms of intangibles). From the firm's point of view, regional variables such as the educational attainment of the working age population, the quality of institutions and the quality of infrastructure are "external" factors, being outside of their control. Our results thus provide additional evidence to the recent findings in the literature on the positive role of regional institutional conditions for innovation (Rodriguez-Pose and Di Cataldo 2014).

In the analysis where patenting activity is the dependent variable, firms in EIBIS are matched to Orbis data to construct a panel dataset between 2000 and 2016. (6) We combine this panel dataset with regional indicators of R&D expenditures, human capital, quality of institutions and quality of infrastructure in order. Because there is little information on R&D spending at the firm level in the financial statements reported in Orbis, we use regional data on public and private R&D expenditures. As a proxy for human capital, we use the share of employed people with tertiary education or in Science and Tech (instead of the educational attainment of the working age population used in Table 2). The estimates in Table 3 show that a good mix of skilled labour force, high-quality institutions and infrastructure, and regional R&D expenditures is crucial for promoting firms' patenting activities, as they all increase the probability of patenting activity.

The Contribution of Enabling Factors to Regional Performance

The level of development of a region can be approximated with GDP per capita (Fig. 11, left panel). (7) It is associated with regional structural attributes such as infrastructure quality, human capital endowments and institutional quality. However, this relationship is not linear for all EU regions. We map regions according to the level of structural attributes, based on three components of the European Commission's Regional Competitiveness Index (RCI): higher education and lifelong learning, institutions, infrastructure (Fig. 11, right panel). Although there is an overlap between regions with high income levels and good framework conditions, a number of regions score lower regarding enabling factors than what their income levels would predict--for example in Northern Italy and Spain, and a number of capital regions including Bucharest, Budapest, Prague and Warsaw (in cohesion countries), Dublin and Madrid (in periphery countries), Ile-de France and Vienna (in other EU countries).

Between 2011 and 2015, R&D expenditures per capita increased rapidly in some less developed regions of Europe (Fig. 12, left panel). Several regions where innovation performance was previously lagging were catching up, particularly in cohesion countries (notably some regions in Poland, Hungary, Czech Republic, Slovakia, Romania and Bulgaria) but also less developed regions belonging to periphery countries (several regions in Southern Italy, Portugal and Greece). R&D expenditures per capita also grew rapidly between 2011 and 2015 in regions in other EU countries (Belgium, Germany and the UK in particular).

In recent years, the level of human capital has increased faster in more developed regions (Fig. 12, right panel). Between 2014 and 2017, employment in knowledge-intensive services and employees with tertiary education (both as a share of total employment) increased rapidly in several regions in other EU countries (notably Sweden, Finland, France and the UK), in the periphery (Portugal, Greece and Italy) and in a few regions of cohesion countries.

In the previous section, firm-level empirical analysis shows that investments in human capital and R&D expenditures support innovation output but that they are not the only factors at play. Enabling regional framework conditions is associated with innovation activities at the firm level. This section investigates the nexus between innovation efforts and an indicator of regional performance, namely the growth in regional TFP, to assess the role of framework conditions at the aggregate regional level. TFP can be estimated at a regional level as the residual of a standard growth accounting Solow model: (8)

[DELTA] ln [A.sub.rt] = ln ([Y.sub.rt]/[Y.sub.r,t-1]) - (1 - [gamma])ln ([L.sub.rt]/[K.sub.r,t-1]) - [gamma] ln ([L.sub.rt]/[L.sub.r,t-1]) (2)

where [A.sub.rt] is the level of TFP in region r in year t, [Y.sub.rt] is regional GDP, [K.sub.rt] is a measure of the regional capital stock (see "Appendix" for more details), and [L.sub.rt] denotes total employment in the region. TFP is estimated using the production function (in logarithm) with regional dummies to obtain factor shares.

The literature on the determinants of TFP is large, but good candidates, particularly in the endogenous growth literature, include R&D spending, investment in human capital and possibly some variables capturing social capital and framework conditions--as they are likely to play an important role for regions that are far from the frontier. Because regional convergence is an EU policy goal, we include in the estimation a "convergence" term in the form of the gap with respect to the frontier TFP level. Lastly, it is likely that the role of human capital and of R&D spending can be different for different levels of TFP; hence, it is reasonable to include in the final equation to be estimated, some interaction terms. We estimate the following equation:

where TFP [gap.sub.rt] refers to the gap of region r in year t with respect to the region with the highest level of TFP, R&D is R&D spending in Euro per 1000 people, HKIS (lagged by 1 year) stands for the share of people employed in high knowledge-intensive sectors as a proxy for human capital, F[C.sub.rt] refers to proxies for regional framework conditions (including institutional quality, the extent of financial constraints in the region, or perceived obstacles to investment), and [V.sub.rt] captures regional unobservable characteristics that can vary over time.

The results in the first column of Table 4 use fixed effects estimation and show that annual total factor productivity growth over the period 2000-2015 is positively associated with the growth of R&D expenditures and the increase in human capital at the regional level. This is also consistent with the results investigating the nexus between innovation and productivity at the regional level (Asikainen and Mangiarotti 2016; Mannasoo et al. 2018). Other factors matter as well, such as the productivity gap with respect to the region with the highest level of productivity, which indicates convergence in total factor productivity over time across EU regions. In other words, it is easier for regions far from the technological frontier to catch up.

We also find that higher levels of human capital in a region can converge to the frontier faster. In the catching up phase (when the gap with the frontier is wide), strengthening human capital can be more important than increasing R&D spending. However, when the gap with the frontier becomes smaller, higher levels of R&D spending can be a more efficient way to approach the frontier. In the regression analysis, the interaction term of R&D expenditures with the TFP gap is negative and the interaction term of change in the share of employment with the TFP gap is positive. This suggests that, for less innovative regions, the adoption of technologies that already exist elsewhere may be more relevant than developing new technologies to achieve faster TFP growth. (9)

Adding framework conditions to this analysis means partially changing the estimation approach: fixed effect estimation is no more available, given that our proxies for framework conditions and the EIBIS data included in the estimation do not vary over time. (10) The estimates in Table 4 are based on random-effects estimation (except for the first column that uses fixed effects), and they should be interpreted with some caution.

To address potential concerns about the representativeness of EIBIS data at the regional level, we also estimated the same model using a subsample that only includes regions with a relatively high numbers of firms. (11) We find that the results are qualitatively similar when we focus on regions with a large number of firms. This suggests that the inclusion of regions with a small number of observations, which are likely to be less representative, does not bias the estimates in a systematic way.

Overall, the results in Table 4 suggest that framework conditions and the business environment can play a crucial role. Regional framework conditions, proxied by the sub-indices of the European Regional Competitiveness Index (RCI) on quality of the business environment, local government and local infrastructures quality, are also associated with TFP growth. In addition, financial constraints related to firms' activities and long-term obstacles to investment are associated with lower productivity growth. This highlights the importance of the context in which firms operate, notably the role of the financial sector (proxied by whether firms are finance-constrained and whether the availability of finance is an obstacle to investment activities) as firms need a stable and predictable environment.

Conclusion

Europe is facing pressing challenges in the medium to long-term. In a context of rapid population ageing and environmental threats, innovation will remain the key driver to support productivity growth in European regions. This paper uses regional structural indicators and firm-level data of the EIB Investment Survey (EIBIS) and shows that framework conditions and firm innovation inputs affect innovation performance and regional productivity growth. We argue that, in addition to R&D investments and the education level of the labour force employed in science and the tech sector, improvements in the regional business environment--such as the quality of institutions and infrastructure--can help firms close the gap with the technological frontier and support the economic development of regions in the EU.

Understanding how firms create and adopt innovations is important for the design and implementation of effective public support. An environment that facilitates the development and use of new ideas is expected to enhance the returns from investment in innovation. Innovation support should be the result of a complex, tailored policy mix to support firm innovation and competitiveness, based on their level of human capital, the gap with respect to the technological frontier and features of local economic systems, in line with the concept of smart specialisation strategies. In this context, regional policy should support both the creation of a pro-innovation environment--a key enabling factor for firm productivity--and support intangible investments--a key factor driving firms' innovation output--to increase the local absorptive innovation capacity.

This suggests that public innovation policy should go beyond direct support for innovative firms. Policymakers should not only focus on highly innovative firms in the manufacturing sector or on more tax incentives for business R&D investment. They should also aim to create a self-sustaining ecosystem that enables the effective diffusion, circulation, commercialisation, adoption and adaptation of new products, processes and services. This is especially relevant for firms that do not innovate at the technological frontier. Resources devoted to R&D and human capital are not sufficient to support innovation activities; they must be combined and complemented with high-quality institutions and infrastructure and supportive regulatory and framework conditions.

However, even if there is a case to be made for public intervention in terms of market failure, policymakers need to be careful to redress this failure without distorting the business environment. Innovation policy intervention needs to be regularly evaluated to assess whether the policy instrument is the most appropriate to best alleviate market failures at the local level.

We believe that there are several avenues to continue this work further. For example, research that would apply a more detailed and regional view in the US, similar to the analysis presented in this paper for EU regions, could offer more insights into how leading innovators and lagging regions of the EU compare with those of the USA.

https://doi.org/10.1057/s41294-019-00091-2

Acknowledgements The authors thank the anonymous referees for their helpful comments and feedback.

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.Org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/LO/) applies to the data made available in this article, unless otherwise stated.

Appendix

Measure of the Capital Stock for the Estimation of TFP

The capital stock [K.sub.it] included in Eq. (1) is not available at regional level. Following Mannasoo et al. (2018), it is calculated as

[K.sub.rt] = (1 - [delta])[K.sub.r,t-1] + [I.sub.rt]

with

[K.sub.r,2000] = [I.sub.r,2001]/([g.sub.r] + [delta])

where [g.sub.r] is the average rate of growth of gross fixed capital formation in region r in a 5-year period. Instead of using an automatic rule, the depreciation rate [delta] and the years of computation of the average for [g.sub.r] have been chosen for each of the 28 EU countries (using granular grid search) in order to obtain a measure of the evolution of the capital stock that is similar to the national one (available in the AMECO dataset). We then applied the two chosen parameters to each region in that country.

Robustness Checks

The analysis in Table 2 uses a panel with 2 years of data (T=2), the two first waves of EIBIS. The number of available observations is 9480 (each wave of the EIBIS survey includes around 12,300 firms) with a small share of firms observed in both periods. The unit of observation of the dependent variable (introduction of new products, processes or services) and many explanatory variables is the firm; only the explanatory variables related to the quality of institutions and of infrastructures and to the available skills in the labour market vary at the regional level.

Table 2 shows the results of a random-effects panel probit regression analysis. In order to test for endogeneity in the model with innovation activity as dependent variable, we run four separate two-stage least squares estimations on a balanced panel of firms that are observed in both waves of EIBIS. As a result, the number of observations using this balanced drops to 774 from 9480. The endogeneity tests are then carried out using the first lag of the four different types of intangible investments as instrumental variables. The null hypothesis for exogeneity cannot be rejected for R&D including IPRs, for software, data, IT networks; and for organisation and business process improvements. However, exogeneity is rejected for training of employees.

Diagnostics tests for multicollinearity (pairwise correlations and variance inflation factor) suggest that mulitcollinearity is not a serious issue for the explanatory variables that we consider (Tables 5 and 6).

The analysis in Table 3 uses a panel with 16 years of data (T = 16), based on the matched EIBIS-Orbis database. The unit of observation of the dependent variable (patenting activity) is the firm, while many explanatory variables vary at the regional level, with the exception of categorical variables used as controls for size, sector and country group of each firm.

Diagnostics tests for multicollinearity (pairwise correlations and variance inflation factor) suggest that mulitcollinearity is not a serious issue for the explanatory variables that we consider (Tables 7 and 8).

The estimation in the first column of Table 4 is based on a panel regression with fixed effects. The unit of observation is a region, and all explanatory variables vary at the regional level. The Hausman test confirms that the choice of fixed effects is adequate in this setting. Inference is based on robust standard error to address heteroscedasticity.

Following Mannasoo et al. (2018), we also use dynamic panel estimation (based on the Arellano-Bond estimator) and find that the results are qualitatively similar to those in the first column of Table 4 (Table 9). Despite the inclusion of lagged TFP growth in the list of explanatory variables, the estimates have the same signs and similar magnitude.

In the other columns of Table 4, we use random effect estimation because we include the time-invariant variables proxying for framework conditions. We include these variables one at the time in order to avoid potential multicollinearity issues.

https://doi.org/10.1057/s41294-019-00091-2

References

Acemoglu, D., S. Johnson, and J.A. Robinson. 2005. Institutions as a Fundamental Cause of Long-Run Growth. In Handbook of Economic Growth, vol. 1, ed. R Aghion and S.N. Durlauf, 385-472. Amsterdam: North-Holland.

Annoni P., L. Dijkstra, and N. Gargano. 2017. The EU Regional Competitiveness Index 2016. European Commission Directorate-General for Regional and Urban Policy Working Paper 02/2017.

Asikainen, A.-L., and G. Mangiarotti. 2016. Contribution of Knowledge Intensive Activities to Regional Competitiveness: Production Function Approach. In Knowledge Intensive Business Services and Regional Competitiveness, ed. J.J.M. Ferreira, M.L. Raposo, C.I. Fernandes, and M. Dejardin, 192-210. Abingdon: Routledge.

Borsi, M.T., and N. Metiu. 2015. The Evolution of Economic Convergence in the European Union. Empirical Economics 48(2): 657-681.

Bubbico, R., and P. Wruuck. 2018. EIB Investment Survey 2017: Regional Results. Luxembourg: European Investment Bank.

Camagni, R., and R. Capello. 2002. Milieux Innovateurs and Collective Learning: From Concepts to Measurement. In The Emergence of the Knowledge Economy, ed. Z.J. Acs, H.L.F. de Groot, and P. Nijkamp, 15-45. Berlin: Springer.

Charron, N., S. Dahlberg, S. Holmberg, B. Rothstein, A. Khomenko and R. Svensson. 2016. The Quality of Government EU Regional Dataset, Version Sep 16. University of Gothenburg, The Quality of Government Institute.

Charron, N., V. Lapuente, and L. Dijkstra. 2014. Regional Governance Matters: A Study on Regional Variation in Quality of Government Within the EU. Regional Studies 48(1): 68-90.

Cirera, X., and W.F. Maloney. 2017. The Innovation Paradox: Developing-Country Capabilities and the Unrealized Promise of Technological Catch-Up. Washington, DC: World Bank.

Cooke, P., M.G. Uranga, and G. Etxebarria. 1997. Regional Innovation Systems: Institutional and Organisational Dimensions. Research Policy 26: 475-491.

Crescenzi, R., M. Di Cataldo, and A. Rodriguez-Pose. 2016. Government Quality and the Economic Returns of Transport Infrastructure Investment in European Regions. Journal of Regional Science 56(4): 555-582.

Crescenzi, R., and A. Rodriguez-Pose. 2009. Systems of Innovation and Regional Growth in the EU: Endogenous Versus External Innovative Activities and Socio-economic Conditions. In Growth and Innovation of Competitive Regions, ed. U. Fratesi and L. Senn, 167-191. Berlin: Springer.

EIB. 2016. Investment and Investment Finance in Europe 2016: Financing Productivity Growth. Luxembourg: European Investment Bank.

EIB. 2017. Investment Report 2017/2018: From Recovery to Sustainable Growth. Luxembourg: European Investment Bank.

EIB. 2018a. EIB Investment Survey, EU Overview. Luxembourg: European Investment Bank.

EIB. 2018b. Investment Report 2018/2019: Retooling Europe's Economy. Luxembourg: European Investment Bank.

European Commission. 2017a. My Region, My Europe, Our Future: Seventh Report on Economic, Social and Territorial Cohesion. Directorate-General for Regional and Urban Policy. Luxembourg: Publications Office of the European Union.

European Commission. 2017b. Competitiveness in Low-Income and Low-Growth Regions. The Lagging Regions Report. SWD (2017) 132 Final, Brussels 10.4.2017.

Farole, T., I. Hallak, P. Harasztosi, and S. Tan. 2017. Business Environment and Firm Performance in European Lagging Regions. World Bank Policy Research Working Paper No. 8281.

Haskel, J., and S. Westlake. 2017. Capitalism Without Capital: The Rise of the Intangible Economy. Princeton, NJ: Princeton University Press.

Holmberg, S., B. Rothstein, and N. Nasiritousi. 2009. Quality of Government: What You Get. Annual Review of Political Science 12: 135-161.

Iammarino, S., A. Rodriguez-Pose, and M. Storper. 2017. Why Regional Development Matters for Europe's Economic Future. European Commission Directorate-General for Regional and Urban Policy Working Paper 07/2017.

Iammarino, S., A. Rodriguez-Pose, and M. Storper. 2018. Regional Inequality in Europe: Evidence, Theory and Policy Implications. CEPR Discussion Paper No. 12841.

Jones, C.I. 2016. The Facts of Economic Growth. In Handbook of Macroeconomics, vol. 2, ed. J.B. Taylor and H. Uhlig, 3-69. Amsterdam: Elsevier.

Kenney, M. 2000. Understanding Silicon Valley: The Anatomy of an Entrepreneurial Region. Stanford, CA: Stanford University Press.

Ketterer, T.D., and A. Rodriguez-Pose. 2018. Institutions Versus 'First-Nature' Geography: What Drives Economic Growth in Europe's Regions? Papers in Regional Science 97: S25-S62.

Mannasoo, K., H. Hein, and R. Ruubel. 2018. The Contributions of Human Capital, R&D Spending and Convergence to Total Factor Productivity Growth. Regional Studies 52: 1598-1611.

Mo, P.H. 2001. Corruption and Economic Growth. Journal of Comparative Economics 29(1): 66-79.

Morgan, K. 2007. The Learning Region: Institutions, Innovation and Regional Renewal. Regional Studies 31(5): 491-503.

Nelson, R., and N. Rosenberg. 1993. Technical Innovation and National Systems. In National Systems of Innovation: A Comparative Study, ed. R. Nelson, 3-28. Oxford: Oxford University Press.

OECD. 2016. OECD Science, Technology and Innovation Outlook 2016. Paris: OECD Publishing.

Oughton, C., M. Landabaso, and K. Morgan. 2002. The Regional Innovation Paradox: Innovation Policy and Industrial Policy. The Journal of Technology Transfer 27(1): 97-110.

Rodriguez-Pose, A., and M. Di Cataldo. 2014. Quality of Government and Innovative Performance in the Regions of Europe. Journal of Economic Geography 15(4): 673-706.

Romer, P. 1990. Endogenous Technological Change. Journal of Political Economy 98(5): 71-102.

Saxenian, A. 1990. Regional Networks and the Resurgence of Silicon Valley. California Management Review 33(1): 89-112.

Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Debora Revoltella [1] * Andrea Brasili [1] (iD) * Rocco L. Bubbico [1] * Annamaria Tiiske [1] * Christoph Weiss [1]

The views expressed in this paper are those of the authors and do not reflect the views of the European Investment Bank.

[mail] Andrea Brasili

a.brasili@eib.org

[1] Economics Department, European Investment Bank, Luxembourg, Luxembourg

(1) The European Quality of Government Index (EQI) has accompanied efforts of the World Bank lo measure the business environment at sub-national level in Europe, where the country coverage was limited (Farole et al. 2017).

(2) See EIB (2018a) for methodological details about EIBIS.

(3) The classification of EU regions intro three groups is based on Article 90 of Regulation (EU) No. 1303/2013 of 17 December 2013.

(4) We classify EU Member States in three groups: Cohesion, Periphery and Other EU countries. Cohesion countries: Bulgaria, Croatia, the Czech Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Romania, Slovakia and Slovenia. Periphery countries: Cyprus, Greece, Ireland, Italy, Portugal and Spain. Other EU countries: Austria, Belgium, Denmark, Finland, France, Germany, Luxembourg, Netherlands, Sweden and the UK.

(5) Bureau Van Dijk's Orbis is a database of private and listed companies with information on financial accounts.

(6) The data on each firm from EIBIS were merged with the Orbis database and the match was done by a third party, which provided anonymised data to the EIB. This means that the EIB does not have the name, the address, the contact details or any additional individual information that could identify the firms in the data.

(7) To proxy for economic development, Iammarino et al. (2017) identifies five distinct groups for EU regions using low, medium-, medium+, high and very high per capita income levels (based on ODP per head in PPS). An alternative measure could be personal income per capita.

(8) "Appendix" provides more details on the estimation and includes robustness checks and diagnostics tests.

(9) The estimates in the first column of Table 4 are qualitatively similar using dynamic panel regression estimation, instead of fixed effects (Table 9 in "Appendix").

(10) This is the reason why we cannot use dynamic panel estimation when we investigate the relevance of framework conditions.

(11) We use the (admittedly arbitrarily) threshold of a minimum of 30 observations per region.

Caption: Fig. 1 Innovation function. Note Adapted from Cirera and Maloney (2017)

Caption: Fig. 2 R&D investment intensity in 2000-2016 (in %). Note China except Hong Kong. No data for China in 2000. Data in 2016 only available for the EU. For each year, "higher bound EU" refers to the value of the EU Member State with the highest R&D intensity; "lower bound EU" refers to the value of the EU Member State with the lowest R&D intensity. Source: Eurostat

Caption: Fig. 3 Regional concentration of patents, R&D expenditure and highly skilled employees. Note Data at NUTS 2 level for EU regions. The data are not available for all the variables and all the regions. Source: Eurostat

Caption: Fig. 4 Share of firms that invest in capacity replacement and share of firms that invest in capacity expansion (in %). Note Investment in replacement: replacing capacity (including existing buildings, machinery, equipment and IT); investment in capacity expansion: expanding capacity for existing products or services. Firms in EIBIS are weighted with value added. Innovative firms are firms that report to invest in R&D and introduce products that are new to the country or the global market. Source: Authors' calculations based on EIB Investment Survey (EIBIS wave 2017)

Caption: Fig. 5 Share of internal funds to finance investment and share of firms that are finance-constrained (in %). Note Internal finance: internal funds and retained earnings (including cash and profits). Finance-constrained: the firm was (1) rejected when seeking any external financing for an investment; (2) quantity constrained (dissatisfied with the terms and the amount received in the last request for external financing); (3) price constrained (the firm did not apply because it thought the conditions of external financing would be too expensive); or (4) discouraged from seeking any external financing (the firm did not apply because it thought the application would be turned down). Firms in EIBIS are weighted with value added. Innovative firms are firms that report to invest in R&D and introduce products that are new to the country or the global market. Source: Authors' calculations based on EIB Investment Survey (EIBIS wave 2017)

Caption: Fig. 6 Share of intangible investment, investment in new products, processes and services, and total factor productivity. Note Local linear regression of share of investment in new products, processes or services on share of intangible investment (left panel). Local linear regression of total factor productivity on share of intangible investment (right panel). Data on all firms are pooled from the three waves of EIBIS (waves 2016, 2017 and 2018). Firms in EIBIS are weighted with value added. Source: Authors' calculations based on EIB Investment Survey (EIBIS)

Caption: Fig. 7 GDP per capita, total factor productivity and investment in new products, processes or services at the country level. Note Data on all firms are pooled from the three waves of EIBIS (waves 2016, 2017 and 2018) to construct an average of total factor productivity for each country. Firms in EIBIS are weighted with value added. Source: Authors' calculations based on Eurostat and EIB Investment Survey (EIBIS)

Caption: Fig. 8 GDP per capita, total factor productivity and investment in new products, processes or services at the regional level. Source: Authors' calculations based on Eurostat and EIB Investment Survey (EIBIS) matched to Bureau van Dijk's Orbis database

Caption: Fig. 9 Share of firms that published a patent over the past 5 years. Note Firms in EIBIS (matched to Orbis) are weighted using value added. Source: EIB calculations based on EIB Investment Survey (EIBIS waves 2016 and 2017) matched to Bureau van Dijk's Orbis database

Caption: Fig. 10 Share of firms that have published at least one patent over the past 5 years, EU regions. Note No data available on UK regions. Regions with no representative data (i.e. less than 30 observations) are omitted. Source: EIB calculations based on EIB Investment Survey (EIBIS waves 2016 and 2017) matched to Bureau van Dijk's Orbis database

Caption: Fig. 11 Income and innovation-enabling factors in regions. Note Income is based on GDP per capita using purchasing power standards. Low (1), medium- (2), medium+ (3), high (4) and very high (5) per capita income levels refer to up to 74%, 75-99%, 100-119%, 120-149% and 150% or above of the EU average. Innovation-enabling framework conditions are based on the following pillars of the Regional Competitiveness Index: (1) Higher education and lifelong learning, (2) Institutions, (3) Infrastructure. Regions are classified as follows: regions in category 1 and 7 score in the bottom (top) quartile regarding all 3 dimensions; regions in category 2 and 6 score in the bottom (top) quartile in 2 out of 3 dimensions. Regions in category 3 and 5 score in the bottom (top) quartile in 1 out of the 3 dimensions. Regions in category 4 regions do not score in either the top or bottom quartile in any of the 3 dimensions. Source: Authors' calculations based on Eurostat and Regional Competitiveness Index (RCI)

Caption: Fig. 12 Growth of R&D expenditures per capita (in %), 2011-2015, and growth of employment share in knowledge-intensive services (in %), 2014-2017, EU regions. Note Intramural R&D expenditures in all sectors. Data for France refer to changes in R&D expenditures between 2009 and 2013. Source: EIB calculations based on Eurostat
Table 1 R&D expenditures dispersion
between and within EU countries (EUR per capita).
Source: Authors' calculations based on Eurostat

                 SD    Min   Max    Number of NUTS
                                    2 regions

Between 28 EU    457   39    1504
  countries
Within countries
All regions      634   6     3737   255
UK               681   98    3688   40
Germany          762   225   3737   38
France (a)       515   77    1561   22
Spain            168   12    586    19
Italy            171   116   623    19
Poland           71    21    313    16
Czech Rep.       288   42    961    8
Hungary          100   31    323    7

Countries are ordered based on the number of NUTS 2 regions

(a) Data on French regions refer to 2013

Table 2 Firm intangible investment (innovation input),
regional framework conditions and firm innovation output.
Source: Authors' calculations based on the EU Regional
Competitiveness Index (RCI) 2016 database and EIB
Investment Survey (EIBIS waves 2016 and 2017)
matched to Bureau van Dijk's Orbis database

Dependent variable: innovation indicator

Educational attainment of the                   0.132 ***
  working age population                        [0.001]
Quality of institutions                         0.096 **
                                                [0.040]
Quality of infrastructure                       0.071 *
                                                [0.052]
Intangible investment: R&D                      0.024 ***
                                                [0.000]
Intangible investment: software, data,          0.002 ***
  IT networks                                   [0.000]
Intangible investment: training of employees    -0.005 ***
                                                [0.000]
Intangible investment: organisation and         0.010 ***
  business process improvements
                                                [0.000]
Firm size                                       Yes
Sector                                          Yes
Country group                                   Yes
Sample size                                     9480
Pseudo-[R.sup.2]                                0.61

Panel probit regression analysis. Firm innovation is an
indicator variable that takes value 1 if the firm reports
in the EIB Investment Survey (EIBIS) that it invested to develop
or introduce products, processes and services, and 0 otherwise.
The explanatory variables include educational attainment of the
working age population; institutional quality and infrastructure
pillar scores of the RCI database; investments of firms into
intangible assets. We also control for sector, firm size and
country group (cohesion, periphery, and other EU countries).
Pseudo-[R.sup.2] calculated rescaling the log likelihood of the
model using the log likelihood of model that only includes a
constant. p values based on robust standard errors in squared
parentheses: *** p<0.01, ** p<0.05, * p<0.1

Table 3 Regional framework conditions and firm patenting
activities. Source: Authors' calculations based on the EU
Regional Competitiveness Index (RCI) 2016 database, Eurostat
and EIB Investment Survey (EIBIS waves 2016 and 2017)
matched to Bureau van Dijk's Orbis database

Dependent variable: patenting indicator

Share of people employed with tertiary     1.686 ***
  education and/or in Science and Tech
                                           [0.000]
Quality of institutions                    0.220 ***
                                           [0.002]
Quality of infrastructure                  0.143 ***
                                           [0.001]
Regional R&D expenditures                  0.348 ***
  (EUR per 1000 people)                    [0.000]
Firm size                                  Yes
Sector                                     Yes
Country group                              Yes
Sample size                                223,350
Pseudo-[R.sup.2]                           0.41

Panel probit regression analysis. Firm patenting
is an indicator variable that takes value 1 if a
firm has published a patent at least once in the
last financial year (according to Orbis), and 0 otherwise.
The explanatory variables include the share of employed
persons with tertiary education or employed in science
and technology as a share of total employment; institutional
quality and infrastructure pillar scores of the RCI database;
regional R&D expenditures (EUR per 1000 people). We also
control for firm size, sector and country group (cohesion,
periphery, and other EU countries). Pseudo-[R.sup.2] calculated
rescaling the log likelihood of the model using the log likelihood
of model that only includes a constant, p values based on robust
standard errors in squared parentheses:
*** p<0.01, ** p<0.05, * p<0.1

Table 4 TFP gap, growth in R&D expenditures, change
in share of employment in knowledge-intensive sectors
(KIS), framework conditions and regional TFP growth.
Source: Authors' calculations based on Eurostat (2000-2015),
European Quality of Government Index (EQI) 2017, EIB
Investment Survey (EIBIS waves 2016 and 2017) matched
to Bureau van Dijk's Orbis database

Dependent variable: annual regional TFP growth

TFP gap (vs. the frontier)   0.040 ***    0.023 ***   0.017 ***
                             [0.000]      [0.000]     [0.001]
Growth in R&D exp. per       2.534 **     1.806 *     1.740
  capita
                             [0.019]      [0.100]     [0.113]
Change in share of emp.      1.197 ***    1.193 ***   1.188 ***
  in KIS (lag)
                             [0.000]      [0.006]     [0.000]
Interaction TFP gap and      -0.012 ***   -0.004 **   -0.003 *
R&D exp.
                             [0.000]      [0.032]     [0.083]
Interaction TFP gap and      0.004 ***    0.004 ***   0.004 ***
  change in KIS
                             [0.000]      [0.000]     [0.000]
Index of institutional                    0.678 ***
  quality
                                          [0.005]
Share of finance-con-                                 -6.238 ***
  strained firms                                      [0.001]
Obstacles to investment
  Availability of finance

Uncertainty about the
  future

Sample size                  1394         1390        1394
Number of regions            150          149         150
[R.sup.2]                    0.082        0.159       0.155

Dependent variable: annual regional TFP growth

TFP gap (vs. the frontier)   0.024 ***    0.018 ***    0.017 ***
                             [0.000]      [0.002]      [0.002]
Growth in R&D exp. per       1.798        1.753        1.745
  capita
                             [0.101]      [0.111]      [0.112]
Change in share of emp.      1.194 ***    1.192 ***    1.186 ***
  in KIS (lag)
                             [0.000]      [0.000]      [0.000]
Interaction TFP gap and      -0.004 **    -0.003 **    -0.003 *
R&D exp.
                             [0.029]      [0.041]      [0.055]
Interaction TFP gap and      0.004 ***    0.004 ***    0.004 ***
  change in KIS
                             [0.000]      [0.000]      [0.000]
Index of institutional       0.569 ***
  quality
                             [0.001]
Share of finance-con-        -4.477 ***
  strained firms             [0.007]
Obstacles to investment
  Availability of finance                 -3.188 ***
                                          [0.000]
Uncertainty about the                                  -1.726 **
  future
                                                       [0.019]
Sample size                  1390         1394         1394
Number of regions            149          150          150
[R.sup.2]                    0.163        0.159        0.152

Panel regression with regional total factor productivity
growth as the dependent variables and selected explanatory
variables: TFP regional gap (with respect to region with
highest TFP level in each year); regional growth in R&D
expenditures per capita; regional share of employment in
R&D; change in regional share of employment in knowledge-
intensive sectors (KIS); index of regional institutional
quality (European Quality of Government Index-EQI 2017);
share of finance-constrained firms (EIBIS); share of firms
that report availability of finance to be an obstacle to
investment; share of firms that report uncertainty about
the future to be an obstacle to investment. Annual regional
data from 2000 to 2015. p values based on robust standard
errors in squared parentheses: *** p<0.01, ** p<0.05, * p<0.1

Table 5 Diagnostic tests for multicollinearity:
pairwise correlations

                                               a        b

Educational attainment of working age      a   1
  pop.
Quality of institutions                    b   0.757    1
Quality of infrastructure                  c   0.447    0.426
Intangible investment: R&D                 d   0.076    0.070
Intan. inv: software, data, IT networks    e   0.059    0.059
Intan. inv: training of employees          f   0.045    0.045
Intan. inv: organ, and business process    g   -0.020   0.001
  impr.

                                           c       d        e

Educational attainment of working age
  pop.
Quality of institutions
Quality of infrastructure                  1
Intangible investment: R&D                 0.082   1
Intan. inv: software, data, IT networks    0.061   -0.067   1
Intan. inv: training of employees          0.055   -0.089   -0.016
Intan. inv: organ, and business process    0.015   -0.032   -0.043
  impr.

                                           f        g

Educational attainment of working age
  pop.
Quality of institutions
Quality of infrastructure
Intangible investment: R&D
Intan. inv: software, data, IT networks
Intan. inv: training of employees          1
Intan. inv: organ, and business process    -0.044   1
  impr.

Table 6 Diagnostic tests for multicollinearity:
variance inflation factor (VIF)

Variable                                      VIF    1/VIF

Educational attainment of working age pop.    2.39   0.418
Quality of institutions                       2.02   0.495
Quality of infrastructure                     1.33   0.752
Intangible investment: R&D                    1.03   0.975
Intan. inv: software, data, IT networks       1.02   0.984
Intan. inv: training of employees             1.01   0.986
Intan. inv: software, data, IT networks       1.01   0.995
Mean VIF                                      1.4

Table 7 Diagnostic tests for multicollinearity:
pairwise correlations

                                      a       b       c       d

Share of emp. with tertiary ed.   a   1
  and/or in Sc. and Tech
Quality of institutions           b   0.553   1
Quality of infrastructure         c   0.393   0.643   1
Regional R&D expenditures         d   0.344   0.465   0.423   1

Table 8 Diagnostic tests for multicollinearity:
variance inflation factor (VIF)

Variable                                     VIF    1/VIF

Share of emp. with tertiary ed. and/or in    2.11   0.473
  Sc.and Tech
Quality of institutions                      1.79   0.558
Quality of infrastructure                    1.44   0.695
Regional R&D expenditures                    1.36   0.736
Mean VIF                                     1.68

Table 9 Dynamic panel estimation (based on the
Arellano-Bond estimator). TFP gap, growth in R&D
expenditures, change in share of employment in
knowledge-intensive sectors (KIS) and regional
TFP growth. Source: Authors' calculations based
on Eurostat (2000-2015), European Quality of
Government Index (EQI) 2017, EIB Investment
Survey (EIBIS waves 2016 and 2017) matched
to Bureau van Dijk's Orbis database

Dependent variable: annual regional TFP growth

TFP growth (t - 1)                       -0.311 ***
                                         [0.0001
TFP gap (vs. the frontier)               0.073 **
                                         [0.0001
Growth in R&D exp. per capita            3.808 ***
                                         [0.000]
Change in share of emp. in KIS (lag)     1.640 ***
                                         [0.0001
Interaction TFP gap and R&D exp.         -0.024 ***
                                         [0.000]
Interaction TFP gap and change in KIS    0.006 ***
                                         [0.000]
Sample size                              1118
Number of regions                        137

Dynamic panel regression (based on the Arellano-Bond
estimator) with regional total factor productivity growth
as the dependent variables and selected explanatory variables:
TFP regional gap (with respect to region with highest TFP
level in each year); regional growth in R&D expenditures per
capita; regional share of employment in R&D; change in regional
share of employment in knowledge-intensive sectors (KIS); index
of regional institutional quality (European Quality of Government
Index--EQI 2017). Annual regional data from 2000 to 2015.
p values based on robust standard errors in squared
parentheses: *** p<0.01, ** p<0.05, * p<0.1
COPYRIGHT 2019 Association for Comparative Economic Studies
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2019 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Revoltella, Debora; Brasili, Andrea; Bubbico, Rocco L.; Tuske, Annamaria; Weiss, Christoph
Publication:Comparative Economic Studies
Article Type:Report
Geographic Code:4E
Date:Jun 1, 2019
Words:9879
Previous Article:Transition Dynamics in European Labour Markets During Crisis and Recovery.
Next Article:Introduction: 2017 INFER 19th Annual Conference, Bordeaux.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters