1
Influence of state aid on productivity and innovation
performance in the European Union energy sector
Through a sustainability lens
Vincent Maessen (s4431650) Supervisor: dr. H.L. Aalbers Second examiner: dr. K.F. van den Oever
Abstract:
The European Union strives for a more efficient and more innovative energy sector in order to minimalize the environmental impact on the energy sector by granting state aid. This master thesis researches the effect of state aid on innovation performance and productivity by
using a data set of 188 companies over the years 2012-2018. After conducting a mixed model analysis on this sample a negative significant relationship could be found between state aid
and innovation performance. This effect becomes positive and loses its significance when looking at environmental based state aid. No other significant relationships could be found
between state aid and productivity, and innovation performance and productivity. Furthermore, the results indicate that there are no mediation effects of state aid or
sustainability support with innovation performance on productivity.
2
Table of content
Chapter/paragraph page
Introduction 4
1.1 Research question 5
1.2 Scope and Design 6
1.3 Relevance and contributions to literature 6
1.4 Social angle research 7
1.5 Structure 7
Theoretical framework 8
2.1 Theoretical lens 8
2.2 Rationale state aid 10
2.3 Context: Energy sector 12
2.4 Independent variable: State aid 13
2.5 Independent variable: Sustainability support 14
2.6 Mediating variable: Innovation performance 16
2.7 Dependent variable: Productivity 17
2.8 Controlling variables: Organizational performance & Frim size 19
2.9 Hypotheses development 20 2.10 Conceptual model 23 Methodology 24 3.1 Nature of research 24 3.2 Data selection 24 3.3 Variable description 24 3.4 Method of analysis 27
3
Results 29
4.1 Assumptions 29
4.2 Outcomes main effects mixed model analysis 32
4.3 Mediation effects 36
4.4 Summary results 36
Conclusion & discussion 38
5.1 Conclusion 38
5.2 Theoretical implication 39
5.3 Practical implications 40
5.4 Limitations research 41
5.5 Recommendations for further research 42
References 43
4
Introduction
Electricity becomes ever more important and is present everywhere in the modern world. The global demand for electricity is portrait to be growing at 2.4% annually between 2005 and 2030. The earth is facing an energy crisis, due to an escalation in global energy demand. The global population exceeded already 7 billion people while still growing steadily, besides there is still dependency towards fossil fuel. The burning up of fossil fuels goes way faster than the creation of it. Hence burning fossil fuels depletes natural resources. Furthermore, it results in a higher amount of carbon-dioxide emissions in the atmosphere, some experts believe this is the reason why the temperature is rising globally (Coyle & Simmons, 2014, p.16).
The European Commission states that a reliable energy supply at reasonable prices together with a minimum impact on the environment are crucial for the European economy (European Commission, 2018). The European Union is building an energy union in order to achieve the following goals: Safe energy supply, viable energy, minimalize the environmental impact, and accessible energy to all member states. This EU energy union strategy is made up out of five dimensions: Security, solidarity and trust; a fully integrated internal energy market; energy efficiency; climate action – decarbonizing the economy; achieved by stimulating research, innovation and competitiveness.
The aims of state aid are to stimulate an increase in production efficiency and other sectors (e.g. research & development and implementations of innovation) are also emphasized by Ginevičius, Podvezko & Bruzge (2008). Moreover, innovation is a key factor in becoming more sustainable and achieving organizational success (Hansen, Grosse-Dunker & Reichwald, 2009). Besides, innovation positively affects the productivity of a firm (Aghion et al., 2015; Peloza, 2009; Mansury & Love, 2008).
According to the European Commission (2018) in some circumstances it is needed to provide state aid in order to create a well-functioning and equitable economy. Approximately €62 billion were provided as state aid to the energy sector by the European Union, which is a share of 53% out of the total state aid granted by the European Union in 2017. This is an €4.4 billion increase in energy sector state aid in comparison with 2016. However a company which receives state aid gains extra assets over its competitors who do not. Therefore, state aid is in essence prohibited unless the public goal provided by the state aid outweighs any risk of distortions of competition in the market. The European Commission is in charge of ensuring that the prohibition is being implied and that the exemptions are fairly and equally applied across the European Union (European Commission, 2018).
5 The rational of state aid is to support two fundamental theorems that form the basis of modern welfare economics. The first theorem states that all competitive equilibriums are Pareto-optimal. The second theorem states that every Pareto-optimal allocation of resources results in an economy that is perfectly competitive. Pareto-optimality is a state of allocation of resources from which it is impossible to reallocate in such a way that one individual will be better off without making at least one other individual worse off (Blaug, 2007). These two theorems separate the two welfare elements of equity and efficiency. Following this rationale, state aid can help in two ways. First to push the Pareto-welfare frontier outwards, this means that the total welfare is being enhanced, this is also known as making the cake bigger. The second way is to move along the Pareto-welfare frontier. This means to redistribute the available resources as such that it maximizes the preferences of society. This process is also known as dividing the cake better (Friederiszick, Röller & Verouder, 2006, p. 636). State aid is believed to be a measure that can be used by the governmental institutions to reduce the Pareto-inefficiency and increase social and regional cohesion (Nitsche & Heidhues, 2006, p. 38). In addition state aid granted to energy companies can be beneficial for society, because endeavours to become more sustainable as a society are highly linked to the energy sector (Dovì et al., 2009).
1.1 Research question
Based on the introduction the research question in this thesis is: “What is effect of state aid on
a company’s productivity and innovation performance and to which extent is the company’s productivity mediated by the company’s innovation performance?” This research question is
answered also by also looking at the sustainability angle of the state aid specifically. To answer this question it is important that various sub questions are answered first. The sub questions of the research questions are:
- What is state aid?
- What is a company‘s innovation performance? - What is productivity?
- What is sustainability?
6 1.2 Research scope and design
This research will focus solely on the European energy sector, which was comprised of +/-89.345 companies in 2016 (European Commission, 2018). This research uses quantitative data coming from reputable data sources, e.g. World Bank, Eurostat, and the Radboud University Nijmegen databases (Orbis & former theses). The experiment is used as the method of data generation. The experiment is suitable for this research, because it has the characteristic that it helps identifying causal relationships among variables (Vennix, 2011). This research will come up with a hypothesis that will be tested via SPSS statistical analysis. 1.3 Relevance and contributions to literature
The general difficulties which are linked to state aid regulation are notorious (Garcia & Neven, 2005). The problem of state aid to business was analysed by many scientists, both in the EU and all over the world (Black & Hoyt, 1989; Bond & Samuelson, 1986; Doyle & Wijnbergen, 1984). All the studies emphasized only some particular aspects associated with the provision of aid (Ginevičius, Podvezko, Bruzgė, 2008, p. 168). Scientific literature provides mixed outcomes regarding state aid and organizational performance indicators. Van Cayseele, Konings & Sergant (2014), for instance, found that state aid may positively affect the productivity of a firm. However, Nicolaides, Kekelekis & Buyskes (2005) expound that state aid may adversely affect competition and could distort the competition in the market. Aghion et al. (2015), state that productivity is positively influenced when state aid is provided. Farla (2015), however, found no clear relation between aid and economic development. Considering this literature, it can clearly be said that the relationship of state aid on organizational indicators really differs per indicator and per context. By conducting a research that tests the effect of state aid and also state aid specifically meant for a better environment on the innovation performance and productivity of companies might contribute in adding more insights into the extent that the EU achieves their goal regarding increasing innovation, sustainability and productivity by providing state aid to the energy sector. The research also checks whether and to which extent the relationship between state aid on one side and productivity on the other are mediated by innovation performance.
7 1.4 Social angle research
As a master student at the Radboud University Nijmegen a lot of emphasis during my education has been on the impact of business on the society. All tax payers are contributing to the state aid issued by the state aid granted by the European Union and its member states. Therefore the whole society is influenced by the policies of state aid by the EU to some extent. Besides the effect of business, and for sure the energy sector, on environment also influences the whole society. Therefore a research about the effectiveness of state aid and environment supporting state aid aiming to improve the innovation performance of companies within a sector which close to everyone in Europe uses will be highly relevant to the society. 1.5 Structure
The introduction of this thesis is followed up by the theoretical framework, which forms the basis of the research. After the theoretical framework has been provided the hypotheses are formulated. The third section focuses on the methodological framework, including data selection, experimental design and data collection, and data analysis. The results of the data analysis are described in section four. The last section is the conclusion and discussion section. This contains the conclusion, scientific implications, the limitations of the research and the recommendations for further research.
8
Theoretical framework
The theoretical framework will firstly provide the theoretical lens used in this thesis. Then the rationale of state aid is given. Thirdly, the context is elaborated. Fourthly, an overview of the current scientific literature on the topics state aid, sustainability, innovation, and productivity is provided. Fifthly, the hypotheses based on the elaboration of these topics the hypotheses are formulated. Lastly, these hypotheses result in a conceptual model.
2.1 Theoretical lens
Several theoretical lenses have been used to describe the rational of state aid and its implications. Some of them are listed in the table below:
Article Theoretical lens used
Ginevičius , Podvezko & Bruzge (2008) Resource based view
Blauberger (2009) Agency theory
Hölscher, Nulsch & Stephan (2017) Growth theory of the firm
Granting state aid is an exceptional measure that the EU takes to attempt to achieve its goal regarding to make the European energy sector more innovative, more productive, and more sustainable. Rules and laws are developed to make sure that the companies which receive state aid actually use it for the purpose intended by the European Union (European Commission, 2013). Therefore, a principal-agent relationship exists. The emphasis of the European Union to realize their goals considering a more sustainable energy sector thru subsidizing innovation makes the testing of the relationship between state aid and innovation performance relevant for practice. Therefore the theoretical lens chosen is the agency theory, because without the principal-agent relationship between the European Union and the recipients of state aid this research could not have been conducted.
Agency theory
Since the 1970‘s agency theory has been a major part of the economy theory of the firm. Pioneers in developing the agency theory are Spence and Zeckhauser (1971), Alchian & Demsetz (1972), Ross (1973), and Jensen & Meckling (1976). Agency theory focusses on the costs of potential differences in interest between the principal and the agent. Agency theory has covered multiple different disciplines throughout the years. Examples are accounting,
9 economics, finance, marketing, political science, organizational behavioural theory, and sociology (Eisenhardt, 1989). It has been argued that an organization is a special case within the agency theory. An organization provides a link between the existence of a complex set of contracts, which can be both written and unwritten, and between several actors. Agency costs exist due to different interests and contractual arrangements (Pepper & Goore, 2015).
There are two different types of agency theory. These are called the ―Positive theory of agency‖ and the ―Principal-agent theory‖ also called the ―Normative agency theory‖ (Pepper & Goore, 2015; Jensen, 1983).
Normative agency theory/Principal-agent theory
Normative agency theory is concerned with helping to solve two problems that might exist in principal-agent relationships. The first problem arises when the principal and the agent have different desires or goals and it is difficult or expensive for the principal to monitor the agent‘s actions. Therefore, the principal cannot verify what the agent is actually doing and whether the agent behaves appropriately. The second problem is risk sharing, this problem can occur when the principal and the agent have a different attitude towards risk. These differences in risk perception can result in different preferences of action for the principal and the agent (Pepper & Goore, 2015).
Agency costs are generated as a result of the different interests and contractual arrangements between owners and top managers. The focus of principal-agent literature is about determining the optimal contract of behaviour versus outcome between the principal and the agent, where the core of the theory is the trade-off between the cost of monitoring the behaviour of the agent and the cost of measuring outcome and the transferring risk to the agent (Eisenhardt, 1989). The normative theory provides a universal theory of the principal-agent relationship. In different guises like employer–employee, lawyer–client, buyer–supplier, and so on (Pepper & Goore, 2015).
Positive agency theory
Research on the positive agency theory focuses on identifying situations where the principal and agent are likely to have opposing goals. If this is clear the positive agency theory tries to describe the underlying mechanisms that limit the agent‘s self-serving behaviour. Positive agency theory is the most important theoretical lens used in the relationship between managers and owners of a company. Furthermore, the positive agency almost exclusively focuses on this relationship between manager and owner. It has also become the leading
10 theory in the extensive academic research about executive compensation (Eisenhardt, 1989, Bratton, 2005; Pepper & Goore, 2015).
The positive agency theory and agency in general is also criticized. For example, Jensen, the leading agency theorist, together with Murphy (1990) was unable to find a conclusive link between CEO pay and stock price performance. More recently, it has been argued that agency theory performed poorly during the 2008 financial crisis and that some situations which implied strong incentives were far from optimal. This contradicted the agency theory (Roberts, 2010; Pepper & Goore, 2015). It was criticized by organizational theorists as minimalistic and by macroeconomists by lacking rigor. However positive agency theory and agency theory in general can be seen as enrichment to economics by giving a more complex view of organizations (Pepper & Gore, 2015).
Choice
The split off chosen as the lens for this research is the principal-agent/ normative theory. Due to its universal characteristics it is more suitable to use for a principal-agent relationship that consists here, notably between the government and a company (Blauberger, 2009). The positive agent theory is too much focused on the relationship between owners as principals and managers as agents.
2.2 Rationale state aid
The free-market economy is an economic system where individuals can make their own decisions regarding their production and consumption behaviour. It is characterised as a system with a low amount of governmental intervention and price equilibrium where supply and demand meet (Krugman & Wells, 2006). Smith (1776) describes, in one of the most famous books of economics ―The wealth of Nations‖, the role of the ―invisible hand‖ in the market. The invisible hand is the mechanism that helps reaching supply and demand levels to meet, in other words the invisible hand is the price mechanism. Where for example prices go up if the demand goes up and/or the supply goes down. The theory of the invisible hand is based on a free market mechanism, but in practise this is hardly ever the case. Because the free market is disturbed with government regulations, alliances, trade deals, taxes, import tariffs, et cetera. The market itself is being described as a place where trade and competition occur. Trade is the circulation of productive factors. This is mainly in the form of goods and services. Competition is the struggle of firms to strengthen their position of selling those goods and services within the relevant markets (Rubini, 2009, p. 25).
11 Market failure means that the invisible hand mechanism is disturbed and that the resources are allocated imperfectly. The rationale of state aid is mostly based on fixing the market failures that can lead to market failures (Dahlman, 1979) and that market failures can be reduced by government intervention (Nitsche & Heidhues, 2006). There are four sources that can lead to market failures (Dodgson, Hughes, Foster & Metcalfe 2011; Friederiszick, Röller &Verouden 2006): namely externalities, imperfect and asymmetric information, coordination problems, and market power.
Externalities are also known as spill-over effects. These effects occur when an actor in the economic market is not able to internalize the whole benefit or costs of their actions. In some cases these spill-over effects can lead to market failures (Revelli, 2002). This is the case when the spill-over effects negatively influence the society. An example could be the emission of carbon dioxide. Governments sometimes tent to intervene in markets by subsidies when the externalities are positive, despite that it has been argued that the market mechanism can solve the negative externalities itself (Dahlman, 1979). However, a lot of goods or services have, to some degree, negative externalities. For example every good needs to be transported and this increases air pollution (Chang, 1996). Therefore governments are limited to address the negative externalities by state aid.
Another major source of market failure is imperfect and/or asymmetric information. In the market system the sellers of the products and services normally have more knowledge about the product or service than the buyers. This lack of information on one side results in inefficient decisions and decision making process (Friederiszick, Röller & Verouden, 2006). Start-ups, such as green energy companies, have a lower likelihood of getting a good price from the market due to information asymmetry regarding the relatively new innovations they apply (Reboredo, Quintela & Otero, 2017). Government institutions could grant state aid to overcome this deficit for start-ups (Friederiszick, et al., 2006). Nonetheless these interventions often work counterproductive due to government failure. Government failure is the lack of information and short-sightedness by the government about a certain case (Collie, 2000).
The third source of market failures are coordination problems. This is the case when the costs of uncertainty and contracting may prevent an effective design or conclusion of agreements within a contract to be effective (European Commission, 2009, p. 8).
The final source of market failure is distorted market power. Market power distortion means that the market power is concentrated by one or a few big players within a market, in this case they have very high economies of scale that provide cost advantages (Meiklejohn, 1999). In such a situation state aid may be provided to other firms so that they can establish
12 themselves in that market, which effects in more competition and more innovation (Meiklejohn, 1999; Collie, 2000).
Because market failures can be quite specific, Buelens et al. (2007) argue that there are different forms of state aid that have different effects on market inefficiencies. The optimal form of state aid depends on the nature and policy under which the market failure exists. Besides this, the impact a government intervention can differ per context. Examples are: countries, regions, industries, and even companies (Poynter, 1985).
2.3 Context: Energy sector
During the 1990‘s most of the European electricity and gas supply was still under public control. The European Union (EU) and governments of the member states decided to open up these markets to competition to end the monopolies and stimulate cross-border trade to stimulate the creation of a single European energy market, which should lead to lower prices and better service to the costumers (Scholz & Purps, 2009, p. 62). The EU and the member states decided several things (European Commission, 2019):
distinguish clearly between competitive parts of the industry (e.g. supply to customers) and non-competitive parts (e.g. operation of the networks);
oblige the operators of the non-competitive parts of the industry (e.g. the networks and other infrastructure) to allow third parties to have access to the infrastructure;
free up the supply side of the market (e.g. remove barriers preventing alternative suppliers from importing or producing energy);
remove gradually any restrictions on customers from changing their supplier; Introduce independent regulators to monitor the sector.
Due to the liberalization of the market the sector is estimated to consist of 89.345 companies in 2016 (European Commission, 2018). A code (NACE code) is given to each company to assign them within the segments of the energy sector where they belong to.
The energy sector consists of the following NACE codes/segments: A01: crop and animal production, hunting and related service activities B05: mining of coal and lignite
B06: extraction of crude petroleum and natural gas B07.21: mining of uranium and thorium ores
B08.92: extraction of peat
B09.1: support activities for petroleum and natural gas extraction C15: the manufacture of leather and related
13 C19: manufacture of coke and refined petroleum products
C24: manufacture of basic metals
D35: electricity, gas, steam, and air conditioning supply
E: water supply; sewerage; waste management and remediation activities E38: waste collection, treatment and disposal activities; materials recovery
F: construction
G47.30: retail sale of automotive fuels in specialized stores H49.5: transport via pipeline
M: professional, scientific and technical activities
2.4 Independent variable: state aid
State aid is defined as ‗an advantage in any form whatsoever conferred on a selective basis to undertakings by national public authorities‘ consisting of the following points (European Commission, 2019; Ginevičius et al., 2008; State Aid Guide, 2007):
• It is based on the use of state financial resources (State or municipality budgets or funds, etc.) or is
State supported;
• provides enterprises with exceptional economic benefits which they could not get under market conditions;
• It is intended for manufacturing some particular products or providing some particular services or is granted to some particular enterprises;
• It distorts or can distort competition and affects trade between the EU member-states.
• State aid is considered to be aid meeting the abovementioned conditions (criteria)
Four distinct forms of instruments are identified as state aid (European Union, 2013): Grants and tax exemptions, equity participation, soft loans and tax deferrals, and guarantees.
Examples of grants and tax exemptions are grants, subsidies, and tax allowances. Equity participation is investing in a company. This could either be to help companies gain capital to do their business or a commercial activity for the focal government. Soft loans and tax deferrals are transfers of aid where the recipient gets a discount on the interest of loans or provided by governmental institutions or a tax deferral. Therefore the aid is much lower than
14 the capital value of the transfer. Guarantees are typically expressed by the nominal amount guaranteed by the government. The aid element is much lower than the nominal amount because guarantees correspond to the benefits that the recipient gets from the government by receiving a premium price for their services or products. In this case the government is willing to pay the businesses more than the market-equilibrium.
It is important that the EU and the member states evaluate the objectives and the impact of the state aid in order for the state aid to be effective and to cause fewer disturbances in the market (European Commission, 2018). The evaluation of the outcome and the correct assessment of the state aid by the member states is a task of the European Commission (Nicolini, Scarpa & Valbonesi, 2017).
A research conducted by Ginevičius et al. (2008) which assessed the effect of state aid, based on multiple types of businesses which were granted EU aid subsidies, states that the results are most successful when state aid is granted to development of production, research & development, and education. The aid intended to increase productivity and R&D is associated with infrastructure development projects, where large financial resources are allocated in terms of absolute value. The aid given to stimulate education is of much lower quantity, resulting in a smaller effect. Despite the smaller effect of state aid on education, the effectiveness of state aid in this sector is higher (Ginevičius et al., 2008).
2.5 Independent variable: sustainability support
In addition, sustainability has been reported as being highly important for the energy sector and for the society as a whole (e.g. Dovì et al., 2009; Adams, Jeanrenaud & Bessant, 2016, Yong et al., 2016). Besides, sustainability is one of the main reasons why the EU grants state aid to the energy sector. Therefore it also is highly relevant to also look if state aid granted due to reasons based on environment purposes specifically has an effect on the innovation performance and the productivity of companies. As mentioned in the introduction one of the major reasons why the European Union grants state aid to the energy sector is because they want to create an energy union in order to achieve the following goals: safe energy supplies, viable energy, minimalize the environmental impact, and accessible energy to all member states by stimulating innovation. Therefore, this research does not only focus on state aid in general, but also zooms in on the state aid specifically granted to foster sustainability in the energy sector, labeled sustainability support. The EU believes that this kind of state aid can help companies to be more innovative in order to become more sustainable (European
15 Commission, 2018). Therefore, sustainability support is defined as ‗an advantage in any form whatsoever conferred on a selective basis to undertakings by national public authorities with as goal to support sustainability‘ (European Commission, 2018; Ginevičius et al., 2008; State Aid Guide, 2007).
Ways to be more sustainable are highly linked to the impact of the energy sector on the environment. Therefore the definition of sustainability in this research is the impact on the environment, where the lower the impact on the environment the more sustainable the company is (Dovì et al., 2009). Examples of limiting the impact of energy businesses on the environment and thus becoming more sustainable are for example reducing the waste generated by a company or decreasing the emission of CO2 and/or equivalent greenhouse gasses. In this research sustainability is therefore defined as the minimization of the impact of companies on the environment (French, 2008; Calvo & Domingo, 2015; Dovì et al., 2009).
Economic agents and markets have both been characterized as part of the problem of negatively impacting the environment as well as the solution to limit the impact on the environment by becoming more sustainable. Businesses have been stimulated and encouraged in achieving a sustainable growth of the economy. The role of innovation to achieve sustainable economic growth was profound, resulting in an increasing interest in innovation by policy-makers, managers, and academics (Adams, Jeanrenaud & Bessant, 2016).
For a world economy to function it is pivotal that there is energy supply and that this energy supply is used and generated efficiently. Therefore, to ensure sustainability it is needed that the use and supply of energy is applied by the principle of minimizing the negative effects on the environment. Furthermore, if it is possible the energy companies should strive for improving the environment (e.g. regenerative development). Ensuring the cleaner production of energy is a key factor in making the society more sustainable by decreasing the emission of carbon-dioxide and other greenhouse gasses, and other pollutant substances created by the production of energy (Yong et al., 2016).
Innovation is needed in order to improve the energy efficiency and thus reduce the impact of the energy sector on the environment (Hansen, Grosse-Dunker & Reichwald, 2009). Still the intrinsic motivation that companies have to innovate is a highly important factor, but most of the times it is not enough to innovate. The problem here is that innovation investment has the characteristics of being highly risky and also costs a lot of money. In order to deal with these high costs and uncertainty it is needed that companies receive subsidies in order to stimulate innovation (Wang et al., 2017).
16 Knowing that minimalizing the impact of the energy sector on the environment is one of the goals of the EU, innovation is a key factor in becoming more environmental friendly, and that subsidies might be needed to provide the assets and decrease the uncertainty for companies to engage in innovation initiatives makes it interesting to look between the differences in effect of state aid granted due to environmental reasons and those with other reasons on a company‘s innovation performance and efficiency.
The difference between the independent variables sustainability support and state aid is that sustainability support only looks at the companies of who were granted state aid because of motivations of increasing sustainability. The variable state aid looks at all the companies in the data set which received state aid, no matter the reason. It is relevant to zoom in at the sustainability support, because one of the goals of the European Union is to have a more sustainable energy sector by means of supporting innovation and productivity (European Commission, 2018).
2.6 Mediating variable: innovation performance
Innovation performance is linked to organizational performance by a cumulative self-reinforcing mechanism which implies that innovative firms outperform non-innovative firms and that good performing firms are more likely to innovate than poorly performing firms (Cainelli, Evangelista & Savona, 2006). By the assessment of the effect of state aid to businesses, it can be stated that state aid to businesses is most effective in the following areas: development of production, Research & Development, and education. R&D leads to increased innovation (Ginevičius et al., 2008).
For a long time the endogenous growth theory has prevented state aid to be one of the main policies to address market failure related to R&D investments (Aghion & Howitt, 1998; Segerstrom, 2000). The endogenous growth theory describes that economic growth is mainly the result of internal and not external forces (Aghion & Howitt, 1998). As a result R&D subsidies to companies are one of the largest and fast-growing forms of state aid in developed countries after the endogenous growth theory lost influence (Nevo, 1998).
R&D subsidies are often designed as direct grants. Research and Development are two related but still different innovation activities. Research activities have fundamentally different characteristics from development activities. Research typically has more tacit knowledge, higher intangibility, greater outcome uncertainty, and larger distance to the market. It is more likely that research investments have a lower appropriability compared to development investments. The reason why this is the case is because research usually
17 involves early-stage activities which result in higher spill-overs and possibly higher expected social returns. Moreover, information asymmetries in research projects have a higher negative effect for early-stage investments, which leads to more binding financial constraints compared to development projects. Simultaneously, research and development are interdependent. The development of products and processes often depends on the quality and outcome of conducted research activities. Firms may be able to get knowledge about how to fix problems through research (Hottenrott, Lopes-Bento & Veugelers, 2017).
Apart from the differences in spill-overs and appropriability, research and development activities vary in their degree of risk and uncertainty. Research is perceived as a more discontinuous process, which provides or may not provide solutions to a problem, whereas development has a more continuous way of looking for solutions, using an existing set of ideas. These differences in risk and uncertainty cause that most investors are more willing to invest in development than in research (Karlsson et al., 2004). A research conducted on a sample of Flemish firms found that research investments are more depending on the financial resources of the firm itself than development projects, this leads to more binding constraints (Czarnitzki et al., 2011).
Since research and development have different characteristics, an optimal subsidy policy should be implemented which takes the differences into account regarding social and private returns, and financial constraints between research and development. Also if different subsidy policies are set for both research and development it is important that the interdependency between these two activities is taken into account. However, previous empirical studies on the effect of R&D grants do not distinguish between grants assigned for development purposes or research purposes nor do these studies address the difference in terms of impact that research and development activities have. The reason why this is the case can mainly be explained by the lack of access to information about the subsidized projects and the amount of private money that a firm invested in R&D activities (Hottenrott, et al., 2017). The major objectives of granting R&D subsidies from a public point of view is to compensate the company for their social returns that their R&D activities possibly contribute to the society and to ease financial market frictions that increase the private costs (Wallsten, 2000; David et al., 2000). Therefore, the main rationale of state aid on R&D projects is to give assistance to private sector incentives to invest in R&D in order to simulate innovation, which are deemed to be insufficient from a social welfare point of view. Governments typically complement the investments of the private sector by giving subsidies to public research institutions (e.g. universities) or by trying to make contracts that are more mission-oriented
18 (David & Hall, 2000). In addition, the government provides R&D subsidies to firms in the private sector by direct grants that can decrease the firms‘ costs of R&D projects (Hottenrott, et al., 2017).
2.7 Dependent variable: Productivity
In the past decades a lot of research has been done about the effect of innovative activities on the productivity of companies, both as a challenge for econometric applications and policy concern. Most empirical studies use R&D as an input to productivity. It is still not clear to which extent innovation effects productivity, despite the large number of empirical studies and expending literature about innovation on a firm‘s productivity levels. This variety in the magnitude of the effect can be explained by the variability and uncertainty characters innovation implies (Hall, Lotti & Mairesse, 2009). These firm level studies find that the effect of R&D on productivity is positive (e.g. Parisi et al., 2006; Van Leeuwen & Klomp, 2006). Also a research about the effect of innovation on productivity in multiple Italian sectors showed that productivity is positively influenced by a company‘s innovation. Innovating firms consistently outperform non-innovative firms in terms of productivity. Furthermore, a strong positive relationship exists between innovation and subsequently productivity and that productivity is linked closely to previous investment in innovation activities (Cainelli, Evangelista & Savona, 2005).
Raymond et al. (2015) conducted a research which investigated the relationship of R&D on innovation and innovation on productivity within Dutch and French manufacturing companies and found that R&D activities significantly affect an organization‘s innovation performance. The intensity and the continuous occurrence of investments had a positive relationship on the innovation performance of the companies. Besides, strong evidence was found that labour productivity was influenced by the intensity and occurrence of product innovation. The occurrence and intensity of product innovation both played a significant and important role in increasing the labour productivity of the companies. The innovation performance, however is not significantly influenced by past labour productivity levels of the firm. Furthermore, the research of Raymond et al. (2015) provides evidence that a unidirectional causality exists, which runs from innovation performance to labour productivity, lacking a feedback effect for both the Dutch firms as well as the French firms.
Productivity is of high importance, because to remain profitable in competitive business environments. It is important that firms constantly improve their production performance. Becoming more innovative is a promising way to achieve this goal. Increasing
19 investments in R&D may lead to an increased stock of knowledge of a firm, which consequently mat lead to innovation and therefore may increase productivity. However R&D investments are a risky investment and so bear the risk of firms not being able to realize positive returns on the investments (Bauman & Kritikos, 2016). These findings raise the question to which extent the rise in productivity by state aid is mediated by innovation performance.
As mentioned in the intro increasing innovation and sustainability alongside with productivity are the main reasons why the EU grants subsidies. A research using data from all medium and large enterprises in China gave the results that firms receiving subsidies exhibited higher total factor productivity levels and that subsidies were significantly associated with new product introductions in sectors which were highly competitive (Aghion et al., 2015). So not only did state aid result in more innovation, but also in higher productivity levels.
Moreover, it is argued that sectorial state aid is more likely to increase firms‘ productivity levels by a larger extent in a competitive business environment with a low concentration of firms in the sector. Also state aid has a more positive effect on productivity if subsidies are granted to more firms within the same sector at the firm level. The effect is also amplified when the subsidies are granted to more competitive firms; these firms are mostly young firms (Aghion et al., 2015).
2.8 Controlling variables: Organizational performance & Company size
Organizational performance is one of the key concepts in research regarding management (Richard et al., 2009, p. 718). Organizational performance does not have a universal definition (Kirby, 2005). However one of the most commonly used concept to measure organizational performance is financial performance. Indicators of financial performance that are used in the energy sector are Return on investment (ROI), return on sales (ROS), return on assets (ROA), and the net income (Kishimoto, Goto & Inoue, 2017; Bezdek & Wendling, 2013; Schroeder & Traber, 2012; Wolf, 2009).
The effect of state aid on organizational performance is not clear. A number of studies find a positive relationship between state aid and organizational performance (Bergström, 2000; Møllgaard, 2005; Van Cayseele, Konings & Sergant, 2014). On the other hand several studies could not found a relationship between state aid and organizational performance (Criscuolo et al., 2012; Farla, 2015; Nicolaides, Kekelekis & Buyskes, 2005). It is argued that the effect of state aid on organizational performance takes some time to be visible (Bergström,
20 2000; Butts & Jagers, 2013). Herzer and Morrissey (2009) contradict the lagged effect of state aid on organizational performance.
Implying frim size as a control variable can be very useful to see if the relationship only holds for a particular size of the companies (Hottenrott et al., 2017; Beck & Deirguc-Kunt, 2006)
2.9 Hypotheses development
Now the important topics of this study have been explained it is time to formulate the hypotheses in order to answer the research question. The sequence of the assessment of the research relationships are: the effect of state aid in innovation performance, the effect of state aid on productivity, the effect of innovation performance on productivity, and lastly the mediation effect of state aid with innovation performance on productivity.
The first hypothesis involves the direct effect of state aid on innovation performance. This relationship has been studied before by scholars. The results of former research gave ambiguous outcomes. A research conducted by Ginevičius et al. (2008), which assessed the effect of state aid based on multiple types of businesses which were granted EU aid subsidies, states that the results are most successful when state aid is granted to development of production, research & development, and education. Howell‘s (2017) findings, however, did not support those state-backed firms will eventually become successful innovators that will generate significantly large social welfare payoffs. Howell (2017) argued that the Chinese government possibly backed the wrong firms.
The role of innovation to achieve sustainable economic growth was profound, resulting in an increasing interest in innovation by policy-makers, managers, and academics (Adams, Jeanrenaud & Bessant, 2016). Moreover, R&D subsidies to companies are one of the largest and fast-growing forms of state aid in developed countries since the endogenous growth theory lost influence (Nevo, 1998).
The problem of achieving innovation is that innovation investment has the characteristics of being highly risky and also costs a lot of money. In order to deal with these high costs and uncertainty it is needed that companies receive subsidies in order to stimulate innovation (Wang et al., 2017).
Therefore, the major objectives of granting R&D subsidies from a public point of view is to compensate the company for their social returns that their R&D activities possibly contribute to the society and to ease financial market frictions that increase the private costs (Wallsten, 2000; David et al., 2000).
21 Therefore, the main rationale of state aid on R&D projects is to give assistance to private sector incentives to invest in R&D in order to simulate innovation, which are deemed to be insufficient from a social welfare point of view. Governments typically complement the investments of the private sector by giving subsidies to public research institutions (e.g. universities) or by trying to make contracts that are more mission-oriented (David & Hall, 2000). In addition, the government provides R&D subsidies to firms in the private sector by direct grants that can decrease the firms‘ costs of R&D projects (Hottenrott, et al., 2017). The logic that assistance in the form of subsidies is needed in order to initiate innovation in the private sector combined with the notion that innovation is a key factor in becoming more sustainable indicates a positive effect of state aid on innovation performance (Hansen, Grosse-Dunker & Reichwald 2009). The literature therefore suggests that state aid positively influences the innovation performance of companies. Leading to the following hypotheses:
Hypothesis 1a: “State aid has a direct positive effect on the innovation performance of the
receiving firm.”
Hypothesis 1b: “Sustainability support has a direct positive effect on the innovation
performance of the receiving firm.”
The second hypothesis is about the effect of state aid on productivity. Ginevičius et al. (2008) argued that state aid granted for the development of production is one of the most successful forms of state aid. A research using data from all medium and large enterprises in China gave the results that firms who receive subsidies exhibited higher total factor productivity levels (Aghion et al., 2015).
Moreover, it is argued that sectorial state aid, in this case state aid provided to the energy sector, is more likely to increase firms‘ productivity levels by a larger extent in a sector with a competitive business environment and a low concentration level of firms. Also state aid has a more positive effect on productivity if subsidies are granted to more firms within the same sector at the firm level. The effect is also amplified when the subsidies are granted to more competitive firms; these firms are mostly young firms (Aghion et al., 2015).Knowing that the European energy sector consists out about 90.000 companies and that €62 billion of state aid is granted to energy companies across the sector shows that the energy sector has a low concentration and that the energy sector has a competitive business environment (European Commission, 2018).
22 Besides a rationale of providing state aid by European Union is to improve efficiency and therefore reduce the impact on the environment of the energy sector (European commission, 2018). The literature therefore leads to the following hypotheses:
Hypothesis 2a: “State aid has a direct positive effect on the productivity of the receiving
firm.”
Hypothesis 2b: “Sustainability support has a direct positive effect on the productivity of the
receiving firm.”
The last direct effect to be discussed is the effect of innovation performance on productivity. Most empirical studies use R&D as an input to productivity. These empirical firm level studies find that the effect of R&D on productivity is positive (e.g. Parisi et al., 2006; Van Leeuwen & Klomp, 2006). A research about the effect of innovation on productivity in multiple Italian sectors showed that productivity is positively influenced by a company‘s innovation. Innovating firms consistently outperform non-innovative firms in terms of productivity. Furthermore, a strong positive relationship exists between innovation and subsequently productivity and that productivity is linked closely to previous investment in innovation activities (Cainelli, Evangelista & Savona, 2005). Besides, Raymond et al. (2015) conducted a research which investigated the relationship of R&D on innovation and innovation on productivity within Dutch and French manufacturing companies. Strong evidence was found that labour productivity was influenced by the intensity and occurrence of product innovation. The occurrence and intensity of product innovation both played a significant and important role in increasing the labour productivity of the companies. The innovation performance, however is not significantly influenced by past labour productivity levels of the firm. The relationship between innovation performance and productivity is therefore unilateral, only going from innovation performance to productivity. This finding is also supported by Colombo & Rabbiosi (2014). The literature therefore leads to the following hypothesis:
Hypothesis 3: “Innovation performance of a company has a direct positive effect on the
company’s productivity.”
The final hypothesis is about the mediation effect going from state aid through innovation performance to productivity. The facts that R&D subsidies are growing more and more and that the European Union states that state aid could help to improve innovation, indicate that that subsidies increase the innovation performance (European Commission, 2018Nevo, 1998;
23 Hottenrott et al., 2017). Furthermore, state aid is often needed to provide financial support in order to enable companies to invest in innovation. (Dovì et al., 2009; Adams, Jeanrenaud & Bessant, 2016; Yong et al., 2016; Wang et al., 2017).
Moreover, the combination of the literature describing the positive effect of state aid on innovation performance at one side (Wallsten, 2000; David et al., 2000; Ginevičius et al., 2008; Adams, Jeanrenaud & Bessant, 2016; Hottenrott et al., 2017) together with the literature describing that innovation leads to an increase in productivity (Parisi et al., 2006; Van Leeuwen & Klomp, 2006; Colombo & Rabbiosi, 2014; Raymond et al., 2015) indicate that the effect of state aid on productivity is likely to be mediated by innovation performance, making it interesting to research. Leading to the following hypotheses:
Hypothesis 4a: “The effect of state aid on productivity is mediated by innovation
performance.”
Hypothesis 4b: “The effect of sustainability support on productivity is mediated by innovation
performance.”
2.10 Conceptual model
The hypotheses lead to the following conceptual model:
State aid Innovation performance Productivity Sustainability support
24
Methodology
Several choices will be discussed in the methodology section. The choices are explained in chronological fashion. First the nature of research is discussed. Second the elaboration of the data selection for the variables: state aid, innovation performance, productivity, organizational performance, and firm size. After the elaboration of the data selection the method of analysis is discussed. Finally the methodology section is finished up by giving clarification considering the data analysing process.
3.1 Nature of research
This research applies a quantitative approach. A quantitative nature is the most suitable when the aim of the research is to determine or test whether a relationship exists between multiple variables (Bleijenbergh, 2013, p.35; Vennix, 2011, p. 89). Qualitative research focuses more on the question how a relationship is established and focuses more on the interpretative naturalistic approach (Denzin & Lincoln, 2011, p. 3; Saunders, Lewis, & Thornhill, 2009). Quantitative data is available for all the variables that have to be measured.
3.2 Data selection
In this section, the variables are operationalized in such a way that they are measurable in order to contribute to science. When this is realized a SPPS analysis can be conducted on the data in order to find if the mentioned hypothesis can be verified or falsified. The database used in this research comes from Orbis of the years from 2012-2018 and consists out of 188 companies. Orbis is an online database which contains information about more than 200 million private companies.
3.3 Variable description
State aid
State aid is measured by simply labelling companies that have not received state aid with a 0 and companies that received state aid with a 1. This simplifies the relationship between state aid and innovation performance, because the relative or nominal amount of state aid is not taken into account (Nicolaides, Kekelekis & Buyskes, 2005, p. 38). Besides that this measure makes the dataset more complete it also has been done because there is still a lot of disagreement in the current literature about the effect of state aid on companies, let alone the effect of amounts of state aid. In addition an extra variable is added within the state aid
25 group, which indicates if the reason of granting the state aid was based on environmental concerns or not. A 0 score in this variable means that the reason was not based on environmental concerns and a 1 score means that the reason for granting the state aid was based on environment concerns.
Sustainability support
The state aid as described above will be differentiated between state aid granted due to sustainability reasons and state aid granted due to other reasons. A score of 0 means that the state aid was granted due to other reasons or that there was no state aid given to the company. A score of 1 in the variable sustainability support means that a company received state aid with as goal supporting sustainability. Differentiating for sustainability support helps in giving inside in the output of innovation and productivity for this particular type of state aid (European Commission, 2018; Ginevičius et al., 2008; State Aid Guide, 2007). By doing so an indication is provided of how well sustainability support is able to provide the means of increasing sustainability (Hansen, Grosse-Dunker & Reichwald, 2009; European Commission, 2018).
Innovation performance
Innovation performance is measured by the number of patents (patent stock) by several researchers, e.g. Colombo and Rabbiosi (2014) and Hottenrott et all. (2017). Looking if the patent stock increases for an organization which receives state aid compared to a company which does not is therefore a viable measure to test the relationship of state aid on the innovation performance. In this research the patent stock of each company is measured from 2012 to 2018. The effect of the IV‘s (state aid & sustainability support) on innovation performance can be measured by dividing the sample between groups of who received state aid/sustainability support and one had not received state aid/sustainability support and consequently look how much the patent stock changed over the years 2012-2018 for the different binary values of state aid and sustainability support.
26
Productivity
The productivity of the company is measured by the operator revenue per employee as it is done by Hottenrott et al. (2017). The dataset provides the operating revenue per employee form 2012 to 2018. The mixed model analysis can therefore measure the effect of state aid, sustainability support, and innovation performance on productivity. In this fashion it can be checked whether receiving state aid or sustainability support helps in improving the productivity of the company. Furthermore, the effect of innovation performance on productivity can be measured by checking if an increased patent stock leads to increased productivity. Even the mediating effect of state aid through innovation performance on productivity can be measured by conducting the Sobel-test (Sobel, 1982).
Organizational performance ROA
This study measures organizational performance by giving the Return on Assets of the last available year. Return on assets is a very common way to measure organizational performance; lots of scholars have measured organizational performance in such a way (Augustine, et al., 2016; Bezdek & Wendling, 2013; Kishimoto, Goto & Inoue, 2017; Schroeder & Traber, 2012; Sukpanich & Rugman, 2007; Wolf, 2009). The ROA ratio is calculated by dividing the net income by the total company‘s assets (Adams & Ferreira, 2009; Bhagat & Bolton, 2008; Martínez, Stöhr & Quiroga Sahaya, 2012). Organizational performance is being used as a control variable in order to check whether the relationship between state aid on one hand and innovation performance and productivity on the other is robust for difference in organizational performance. The ROA ratio of the last available year is used, because this gives the most actual value of the organizational performance.
Company size
The size of the company is simply measured by the number of employees. Hottenrott et al. (2017) also controlled for company size effects of state aid on R&D indicators by looking at the number of employees of the company. The population of the groups exists of two groups small/medium sized enterprises and large enterprises. Small/medium size enterprises have up to 250 employees and large firms have over 250 employees (Hottenrott et al., 2017; Beck & Demirguc-Kunt, 2006). A 0 is assed to SME‘s and a 1 is assed to companies with more than 250 employees. To determine the size of the company the number of employees is taken of the last available year.
27 3.4 Method of analysis
In this research it is important that several relationships become clear. Not only the effects of state aid on innovation performance, innovation performance on productivity, and state aid on productivity are important but also the controlling for organizational performance and using firm size as a control variable is important. Therefore it is needed to carry out a multiple regression analysis. A mixed model analysis is useful for this research, because mixed model analysis makes it possible to analyse cross-sectional as well as longitudinal variables in one model at the same time (Baltagi, 2008; Peugh & Enders, 2005). A mixed model analysis is a good fit for a study which uses repeated measures on the same statistical units. In this research the statistical units are the companies and the repeated measures are the values of the variables innovation performance and productivity for the years 2012-2018.
The equation for a linear mixed methods regression can be written as:
Where, Yi is a vector of the observations of a measurable variable (innovation performance
and productivity), bi is a vector of the fixed effects of the observations (state aid, environmental state aid, organizational performance, and firm size) gi is the vector with the
random effects of the observations (innovation performance when productivity is Yi), and ci
represents the residuals of the sample group (Lee & Van der Werf, 2016).
The control variables can be added in SPSS and several assumptions have to be met: homoscedasticity, linearity, multi-collinearity, homoscedasticity, normality, and independent errors (Field, 2013, p. 876).
Important goodness-of-fit measures for mixed model analysis are the Log-likelihood, Akaike‘s information criterion (AIC), and the Schwarz‘s Bayesian criterion (BIC) (Bogdan, Ghosh & Doerge, 2004). The lower the values of these criterions, the better the model fit (Field, 2013, p. 870 & 878). Despite not being part of the mixed model analysis in SPSS, the R-squared has also been calculated manually to show how much the independent variables explain the variance in the dependent variable.
28 3.5 Data analysis process
Firstly the assumptions are tested. The testing of the assumptions is pivotal, because they can ensure that the data is fit for moderator analysis. After this it is important to look at the unusual points in the data set, because they might bias the estimates of variables. The next step is to analyse the data through a mixed model regression and check whether state aid has a significant statistical effect on innovation performance and productivity. Finally, the results of the assumptions, unusual points and the complete mixed model analysis will be summarized.
The main effects are controlled by organizational performance and firm size and will be tested for several points in time to check if state aid realizes growth in the patent stock, patent stock increases productivity, and state aid increases productivity. Also the mediation effect of state aid trough innovation performance on productivity is measured. Having the results gives the possibility to either reject or accept the hypotheses. This is a relative short term analysis. Short-term overreaction to information is quite common (Hur & Singh, 2016; Lobe & Rieks, 2011; Maor, 2012; Wang & Xie, 2010).
29
Results
The results section consists out of four parts. First the assumptions of the mixed model analysis are checked. After this the results of mixed model analysis between the independent variables and the dependent variables are given. Third, the results of mediation effects are given. Lastly, there is a brief summary about the results of the mixed model analysis 4.1 Assumptions
Before the mix model regression analysis can be conducted it is important that certain assumptions are checked. It is important that the assumptions are checked, because it is only possible to make generalizations using mixed method regression analysis if the required assumptions are met (Field, 2013, p. 192-195).
Linearity
The first assumption that will be tested is the linearity of the dependent variable and the mediating variable. A way to test whether the output is linear is to look at the scatterplots. The scatterplots show that the data is quite linear. The absence of a clear pattern means that the data is unbiased, but the increased error on the left side of the plot indicates that the variance is not constant. Therefore, it can be stated that the data is to a large extent linear (Field, 2013, p. 192). Furthermore, the data is considered to be linear if the residuals are more or less symmetrically horizontally distributed across the scatter plot (Norušis, 2006, p. 507). Looking at the scatterplots it can be stated that the upper and the lower half of both variables look quite similar (appendix A+B).
Multi-collinearity
collinearity exists when two or more variables are closely linear related. Multi-collinearity becomes problematic when the Variation Inflation Factor (VIF)-value exceeds 10 and the tolerance value is below 0,2 (Field, 2013, p. 885-886). The VIF provides an index number that shows how much the variance of the variable increases due to collinearity. In other words the VIF shows how much a predictor (IV) has a linear relationship with another predictor. The tolerance value is 1- R2 or 1/VIF,where R2 is the regression of the independent variable on the other independent variables included in the mixed model regression analysis (Cohen et al., 2013). To run this a simple regression can be conducted between these two variables. Running this simple regression and looking at the coefficient table gives VIF-values
30 from 1,045 to 1,182. The tolerance values differ from 0,846 to 0,957 (appendix C). These values indicate that the multi-collinearity is far from problematic.
Homoscedasticity
It is of great importance that the residuals of the data are normally distributed from the mean (Kim, 2015). To determine whether the errors of the residuals are normally distributed it is important to look at the normal Q-Q plot. The residuals are perfectly normally distributed if the residuals on the Q-Q plot are precisely on the regression line. Taking the windsorized logarithm of both the operating revenue per employee and the number of patents it can be seen that the residuals are very close to the regression line (appendix figure D+E). Therefore, the residuals can be considered as normally distributed and thus it can be stated that the assumption of homoscedasticity has been met for both variables.
Normality
Normality can be tested by means of the Kolmogorov-Smirnov test and the Shapiro-Wilk test (Kim, 2015). The closer the Kolmogorov-Smirnov test is to 0 and the closer the Shapiro-Wilk test is to 1 the more likely it is that the data is normally distributed. When the Logarithm was taken of both original variables the tests came close to the desired values (appendix figure F+G). Because these tests are significant it can be concluded that the data is not normally distributed. It is almost impossible to have a dataset which has a perfect normally distributed data; because in the case that only a few values differ the tests will score significant. Therefore the significant values of the Kolmogorov-Smirnov and the Shapiro-Wilk test are not problematic (Kim, 2015). Looking at the histogram of both variables it can be seen that the distribution of the values does not differ much from a perfect normal distribution, the so called Gauss-curve for the productivity. In the innovation performance it can be seen that there are a lot of 0 values. This is the case because a lot of companies did not produce patents (appendix H+I).
31
Independent errors
The final assumption that was tested is the assumption of independent errors. It is important that the estimations of the errors are independent of one and other. The independent errors can be measured by the Watson test (appendix J). The closer the value of the Durbin-Watson test is to two the more independent the errors are. One rule is that values between 1 and 3 are generally accepted (Field, 2013, p.874). In this case the Durbin-Watson test outcome is 0,579. The closer the outcome of the test is to 0 the more the errors are positively correlated (Field, 2013, p.874). However, this research is based on a dataset of 188 companies with a time series from 2012 to 2018. Autocorrelation is a very common in time-series regression and is called chronic dependency (Ludlow & Perez, 2018). This is the reason why the errors are somewhat positively correlated. The Durbin-Watson score of 0,579 is therefore not problematic.
Measures considering the assumptions
As already mentioned, to make the data fit for mixed model analysis it was needed to take the logarithm of the dependent and the mediating variable. With this fashion it was possible to make the distribution of the data more normal. Because the logarithm function of the original variable was taken, it was needed to asses a 1 to all the 0 values for the patent numbers, because it is impossible to take the logarithm of 0.
To solve the extreme measures that influence the average too much, also called outliners, the data has been winsorized. Winsorization is an effective way to deal with extreme outliners. Besides making the data more robust to outliners, winsorization also helps to improve statistical efficiency. The downside of using winsorization is that bias is introduced in the data, because the extreme values are cut off to a certain point (Ghosh & de Vogt, 2006). The reason why winsorization has been used nonetheless is because too many extreme values mutate the mean and the standard deviation too much for the whole data set and it is a better alternative than leaving the extreme values out altogether.
