POLICY INFLUENCES ON THE INNOVATIVE PERFORMANCE IN THE RENEWABLE ENERGY SECTOR

(1)

PERFORMANCE IN THE RENEWABLE ENERGY SECTOR

AN EMPIRICAL ANALYSIS OF COUNTRY-LEVEL PATENT DATA

University of Groningen

Master Thesis

Ma Business Administration: Strategic Innovation Management

B. Leussink (1688219)

bartleussink@gmail.com

20

th

_{of January, 2014}

First supervisor: Dr. F. Noseleit

Second supervisor: Dr. P. de Faria

(2)

1

Preface

During my master program, strategic innovation management, I wrote this thesis about the policy effects on innovative performance in renewables. My interest in the energy sector has been triggered for some time. Because of the transition from conventional to renewable energy, the relation between energy and innovation is very relevant. The deepened insights gained from this research have increased my interest and knowledge in this field, which was one of my personal goals with this topic.

(3)

2

Abstract

This study examines the effect of government policy on technological innovation in the renewable electricity sector. Patent application data of 15 European countries is used to evaluate several widely applied support mechanisms to promote renewable electricity generation. It is found that feed-in tariffs and quota obligations combined with TGC are most successful in stimulating innovation. An explanation lies in the fact that these mechanisms have a differential influence on risk associated with investing in renewables. Successful mechanisms are able to lower the associated risk for investing in renewables, whereas less successful mechanisms are not able to provide this security. By implementing effective mechanisms, innovation can be stimulated which prevents the lock-out of renewables, thereby reaching the goal of a sustainable future.

Keywords: renewable energy, governmental policy, support mechanisms, innovative

(4)

3

1. INTRODUCTION

This research focuses on the influence that governmental policy has on the innovative performance in renewable energy, specifically the electricity generating sector. This sector can be characterized as one with many changes over the last decades. Changes that are caused by society demanding a sustainable future for generations to come. Realizing the transition from conventional to renewable energy is considered an important aspect in reaching this sustainability. Relating to this, Kwant (2002) states, that the well-recognized need for sustainable society drives renewable energy policies. This means that the government has an important role to play when it comes to this transition. Many authors (Lauber, 2011; Reiche & Bechberger, 2004; Unruh, 2000) emphasize the importance of governmental policy and the implemented support schemes on the success of the transition.

Since the United Nations Framework Convention on Climate Change, signed in 1992, the transition from conventional to renewable energy became a more important aspect in governmental policy. Together with 191 other countries, the Netherlands signed the treaty to change the future of energy consumption. Their current target for the use of renewable energy is 14% in 2020 and full use of renewable energy in the year 2050 (CBS, 2012). Some advantages the government recognizes are;

- Less greenhouse gas or other emission harmful for the environment or health. - Less dependence on a small number of oil and gas producing countries. - Less dependence on energy sources that become empty.

These advantages are well known, so one should expect the government taking an active role in this. However, in the Netherlands, when looking at the percentage of conventional versus renewable energy, the share of renewables from 2009 to 2010 has decreased with 0,4% (Blokhuis, Advokaat & Schaefer, 2012). This development seems contrary to the targets that are set to achieve by the Dutch government. The hold off in the transition from conventional to renewable energy is controversial, especially because the government has typified the energy sector as one of the so called top sectors (CBS, 2012). According to a recent report from a Dutch government institution concerned with policy and environment, it seems that the government nowadays realizes that the Netherlands has fallen behind;

(6)

5

According to the report, countries like Germany and Denmark have progressed much more in the transition of becoming green. Besides the fact that this should be done for the environment, it now threatens the economic situation of the Netherlands. For Dutch firms to stay competitive, a more efficient use of energy and raw materials is needed and besides that, when ignoring the degradation of the environment now, the future will bring higher costs. (Planbureau voor de Leefomgeving, 2013).

To open the way for this massive transition, innovation has become critical in the energy sector. Lowering prices of renewable energy through innovation will drive conventional energy use to a minimum. The problem arises that firms have no incentive to adopt more costly technologies that reduce emissions but provide no additional cost savings to the firm (Popp, Hascic & Medhi, 2011). Governmental regulations can create the incentives for companies to invest and innovate. According to the CBS (2011), the Dutch government is stimulating innovation in renewable energy technologies as they do with a so called “top sector”. By looking at innovation as a stimulus for the transition from conventional to renewable energy, it appears this is not very successful in the Netherlands. The question stands why, for some countries like Germany, it is possible to effectively stimulate innovation. The existing literature often focuses on a specific part of renewables and is outdated because of the dynamic context the renewables are in. The following research focuses on the country wide innovative performance in renewable energy generation. By taking on a European perspective, it becomes possible to compare different policies and their outcomes in terms of innovative performance. With the Dutch government realizing that they are behind in their policies, it is worth investigating what the cause may be, especially regarding the costs these support schemes bring to national economies.

First, literature research will be done to clarify the role of innovation in renewables. It will become clear which support mechanisms are used by governments and how these relate to firms and their decisions. When fully developed, measurement will take place on the performance of different European countries by looking at the number of patent applications that can be ascribed to several technological areas. Patents are a means of protecting inventions developed by firms, institutions or individuals, and as such they may be interpreted as indicators of invention (OECD, 2009). Within the renewable energy sector, this research considers innovation in seven different technological areas; wind, solar (thermal, photovoltaic (PV) and hybrids), geothermal, marine (excluding tidal) and small hydro energy (tidal, stream, dam less1_).

(7)

6

This given, there will be measurement among seven different patent groups representing the different technological areas.

By comparing on a European scale, insight will be gained in two aspects;

- The policy goal, learning from other countries. The assumption here is that the regulations from other countries can be transferred. This is not always the case because regulations are path dependent, being unable to imitate.

- Scientific goal: explanation of differences between countries/systems, their effect and their development.

Ultimately, the comparative assessment among selected countries may lead to insights concerning the effects of policy on innovative performance in general. The specific case of the Netherlands will be evaluated in the conclusion. With this research it becomes possible to evaluate the evolving policy measures, for instance the Dutch “new energy agreement for sustainable growth” signed in September 2013.

1.1 Research questions

In the above section, it becomes clear that much has to be learnt about the relation between innovation and policy in the renewable energy sector. Therefore, the goal of this study is:

To understand the effects of government policy on the countries innovative performance in the renewable energy sector.

This goal leads to the following questions that need to be addressed;

1. How do different types of policy support mechanisms across Europe relate to the innovative performance in the renewable energy sector?

(8)

7

2. CURRENT LITERATURE

The following section will take a close look at the available literature. Different types of policy support mechanisms are described along with the interaction these policies have on a firm level.

2.1 Renewable energy

2.1.1. Renewables in general

Renewable energy is defined as energy that is derived from natural processes (e.g. sunlight and wind) that are replenished at a higher rate than they are consumed (IEA, 2013). According to the International Energy Agency (2007), one can define three generations of renewables, reaching back more than 100 years. First-generation technologies emerged from the industrial revolution and include geothermal power and heat, conventional hydropower and biomass combustion. Second generation technologies include solar heating and cooling, wind power, modern forms of bioenergy and solar photovoltaics (PV). These technologies are now entering markets as a result of investments since the 1980s. The initial investment was prompted by energy security concerns linked to the oil price crises of that period but the enduring appeal of renewables is due, at least in part, to environmental benefits (IEA, 2013). Third-generation technologies are still under development and include concentrating solar power, marine energy, enhanced geothermal systems and advanced bioenergy systems. These are not yet widely demonstrated or commercialised and still depend on attracting sufficient attention and funding.

(9)

8

The main issue for renewable energy is developing the capacity to compete with conventional or fossil energy. Projects in renewables often involve higher upfront costs than their conventional energy counterparts (Boomsma, Meade & Fleten. 2012). This creates problems for investors and for the final price of renewable energy, which makes this competition a rough challenge. Schmidt & Marschinski explore the possibility of a technological breakthrough in the renewable energy sector (2009). They see the transition from low-output to high-output in renewables as a discontinuous rise in the supply of renewable energy, and a drop in the supply of fossil energy. In sense, this would mean that renewables are becoming better able to compete with fossil fuels. There are however difficulties for the renewable energy sector to develop. The author Unruh (2000) calls it the carbon lock-in, where the problem arises that inferior technology may become locked-in due to path-dependency. When this is the situation, innovation in renewables would become unnecessary because the market for renewables would not develop. Schmidt & Marschinski (2009) state that the social optimum requires an early transition and that the market failure that occurs is seen as an inefficient equilibrium selection. Preventing lock-in and an early transition seem to be important success factors. Accelerating technological innovation in renewable technologies can contribute to lower the costs of renewables so that they can compete on a level playing field with conventional energy sources (Noailly & Smeets, 2012). The next section elaborates on this innovation aspect.

2.1.2 Innovation in renewables

When fossil fuel prices go up, we can expect innovations to be directed at renewables at the expense of fossil fuel technologies (Noailly & Smeets, 2012). However, to the extent that fossil fuel innovations improve the efficiency of the related technologies, an increase in fossil-fuel prices might also induce more fossil fuel innovation. However, according to Schmidt & Marschinski (2009), this effect is not strong as the scope for cost-reducing innovations in fossil fuel is presumably smaller in this mature sector, and effects of technological progress may be offset by increasing fossil fuel scarcity.

The article of Acemoglu, Aghion, Bursztyn & Hemous (2012) highlights the central roles played by three main factors on the direction of technological change. These factors are incentives to innovate in renewable versus conventional technologies;

(10)

9

- Price effect, directs innovation towards the sector with higher price. So, innovation takes place where it is possible to save on the use of higher priced inputs.

- Direct productivity effect, innovation takes place in areas with an existing stock of knowledge, thereby building on earlier innovations, creating path-dependencies.

These factors influence the transition from conventional to renewable energy sources. The transition can be triggered by increasing world energy demand, by a reduction in the supply of fossil fuels, or by policy intervention (Schmidt & Marschinski, 2009). Policy intervention has to focus on the areas where it is possible to trigger changes. Countries can focus on these factors when developing their support schemes. Especially the market size effect seems to have potential to be influenced by policy. By creating demand for renewables, the market size effect will be positive towards this type of energy. The next section will take a closer look at the role of governments and their options for promoting renewables.

2.2 Types of policy support mechanisms

This section takes a closer look at the different support mechanisms seen across European countries, and possible explanations for these differences. An answer is provided why governmental support is justified in the renewable energy sector.

2.2.1 Different support mechanisms

Policy mechanisms are defined as any concrete activity initiated by the government in order to enlarge the market implementation of renewables (Harmelink, Voogt & Cremer, 2006). Besides the framework conditions mentioned earlier, Reiche & Bechberger (2004) also emphasize the crucial role of the deployed mechanisms for the promotion of renewables. The main mechanisms are feed-in tariffs (FITs), quota obligations with TGC, tenders, and tax regulations. Several authors (Rathmann et al., 2011; Held, Ragwitz & Wietschel, 2005) define the different mechanisms and classify these between price-based and quantity-based mechanisms;

- Feed-in tariffs (FIT); generated renewable electricity can be fed into the grid at a guaranteed tariff for a determined period of time. FITs may also consist of premium tariffs paid in addition to the market price (feed-in premiums). The determined period of time differs between European countries.

(11)

10

revenue from selling TGCs. The certificate price depends on a predefined quota target and is determined on the market. There is a penalty when the quota is not reached. - Tender procedures; A predefined target of additional renewable capacity is set. Through

a bidding system, projects with the lowest generation costs for a certain amount of added capacity can obtain financial support in the form of long-term feed-in tariffs. - Tax regulations; A reduction of tax payment on renewable sources, or a tax to increase

the costs of conventional energy (carbon tax).

Price-based mechanisms Quantity-based mechanisms

Feed-in tariffs Quota obligation in combination with TGC (tradable green certificates)

Tax regulations Tender procedures

Table 2.1 – price- and quantity-based mechanisms.

(12)

11

Quantity-based mechanisms are focusing on the amount of renewables, thereby eliminating the question of how much renewable energy will be generated. The aim of the quota/TCG is to introduce conditions of market competition into the production of green electricity for technologies that are not fully competitive with traditional supply systems (Meyer, 2003). As Ringel (2006) explains, the obliged entities are free to choose whether they fulfil the quota themselves or rather pay another entity for covering their obligation. So, producers, consumers or distributers of energy have a certain quota they have to fulfil, but can fill this quota by trading in green certificates. They prove the fulfilment of their obligation by showing that they have bought the respective amount of electricity generated by renewables (Ringel, 2006). According to Held et al. (2005), this system is rather low in effectiveness, especially related to wind energy, but this may be because of transient effects. The quota system with TGC is a rather new mechanism in all countries using it. As Menanteau et al. (2003) identify, renewable electricity generators benefit in two different ways; by selling it on the network at the market price, and by selling certificates on the green certificates market. The extra costs for producing renewables are compensated by selling green certificates at a price that covers the loss. A disadvantage identified by Ringel (2006) says that at market conditions, only the producer of wind power, hydro and biomass energy will be able to sell their certificates and consequently continue their production, which leaves out far-market solutions. This is because these technologies can reach low energy prices. This does not fall in line with the goal to diversify in a broad way on the electricity generating technologies.

(13)

12

Tax regulations are recognized by Kwant (2002) as drivers of innovation: the energy tax encourages energy conservation and the use of renewable energy by making fossil fuel energy more expensive. However, Menanteau et al. (2003) have a very different view about this. They agree to certain extent by stating that the simplest, most efficient solution for fair competition between conventional and renewable sources would be to correct the market imperfections by implementing an environmental tax, which would be an incentive to innovate. But, in practice, taxes are faced with the problem of political acceptability and, furthermore, an environmental tax may not be sufficient in itself to stimulate the dynamic learning process required to bring down costs (Menanteau et al., 2003). This failing dynamic learning process to bring down costs contradicts the statement of Kwant (2002) that tax regulations are drivers of innovation. Also, according to Reiche & Bechberger (2004), tax regulations can change from year to year, which creates uncertainty.

2.2.2. Differences among countries

Reiche & Bechberger (2004) try to explain policy differences in EU member states. They come up with several framework conditions which influence the success of the applied mechanisms for the promotion of renewable energies:

- Definition of renewables, for instance concerning hydropower. Most member states exclude hydropower above 10 MW as renewable.

- Geographical conditions and starting points, for instance, the amount of rainfall, sun, wind and geothermal heat and besides that, the availability of fossil resources.

- International obligations like the EU directive and the Kyoto protocol.

- Planning cultures, for instance in the Netherlands, the permit procedure is more time consuming because of environmental and building permits.

- Public awareness, countries population differs in accepting higher prices for green electricity and other disadvantages like noise pollution. In Denmark for instance, wind turbines owned by the public increases acceptance.

- Technical differences, for instance if grid capacity is suited for decentralized energy production.

(14)

13

projects which by then had dominated the energy sector for some time (e.g., different types of nuclear power plants including fast breeders and reprocessing coal liquefaction, nuclear fusion, new turbines).

These framework conditions clearly play a role in the fact that governments have constantly tried to find a suitable support scheme. Governments occasionally alter the choice of support mechanisms. An example is the replacement of competitive tendering in the United Kingdom with the tradable green certificate. Boomsma et al. (2012) acknowledge this by stating that changes can be caused by a shift in the political environment, the tightening of national or international targets, a change in technology, or as governments may no longer be able to finance support schemes under the financial crisis. By looking at the existing knowledge, it seems contrary that governments continually switch to other support mechanisms. In the Netherlands for instance, some authors criticize the way that this is dealt with. Blokhuis et al. (2012) state that a cause of the stagnation in renewable energy generation in the Netherlands is the absence of a nation-wide, clear and consistent long-term policy on the introduction of renewable energy. Other authors confirm this; complex mechanisms and frequent phase- in and –out of the adopted policies are not desirable, as seen in the Netherlands (Costa, La Rovere & Assmann, 2008; van Rooijen & van Wees, 2006). The policy should be clear and consistent to ensure continuity.

2.2.3 Justification of support

(15)

14

According to Owen (2006), it is justified because of the existing market barriers, which is anything that slows the rate at which the market for a technology expands. Renewable energy technologies will be increasingly exploited as fossil fuels become increasingly scarce. This scarcity will be visible in the form of price signals. But according to Owen (2006), these price signals are only right when externalities are incorporated in the electricity tariff of both fossil and renewable energy. In the case of fossil fuels an example would be pollution. In the eye of governmental policy, this would be a way to eliminate this market barrier that is risking the lock-out of renewables. Other researchers (Lauber, 2011) agree with this and besides that identify learning curve costs as an additional justification of governmental support.

There is a side note to this according to Schmidt & Marschinski, because they only see justifications for a certain level of support until a high-output state is reached. After that, softer policy mechanisms are sufficient to eliminate the remaining sources of market failure (2009).

2.3 Policy interaction with firms

As mentioned in the previous part, some governments are more successful in their policy decisions than others. This also affects firms and investors that are looking for specific market conditions generated by policy decisions. The relevance lies in the fact that many innovations come from firms, so attracting these and affecting their decisions could stimulate innovation.

2.3.1 Affecting firm decisions

Reasons why firms keep on investing in dirty rather than clean technologies are apparent, but the government has possibilities to affect firms in their technological decisions. When governments invest in and buy clean energy they generate extra demand for renewables. In sense they increase the market size, mentioned earlier by Acemoglu et al. (2012) as one of the factors that works as an incentive to invest and innovate in clean technologies. This demand driven policy is referred to as public technology procurement (Aschhoff & Sofka, 2009). Public procurement has been revitalized as an innovation policy mechanism on both European and national levels (Aschhoff & Sofka, 2009; Edler & Georghiou, 2007). Researchers (Aschhoff & Sofka, 2009) found that public procurement is effective in particular for smaller firms in regional areas under economic stress and in distributive and technological services. They suspect that public procurement may be especially promising for firms with limited resources.

(16)

15

is similar to this procurement. As Aschhoff & Sofka (2009) state about the process of public procurement “After the government has placed a tender for a specific need and firms have applied, the decision is made by the government. Only one firm or a consortium of firms gets the order to generate and deliver the product or service”. However according to Lauber (2011), only a few countries took the actual step of market creation for a host of new renewable generators.

According to Wüstenhagen & Menichetti (2012), green public procurement strategies may have positive effects on private sector investment in renewables by reducing perceived risk and adding a stamp of credibility to renewable energy technologies. This contradicts the earlier statements of Menanteau et al. (2003), where competition from the bidding process drives down the margins and thereby increasing the risk. Wüstenhagen & Menichetti (2012) say that investors and firms make their decisions based on the level of perceived risk and expected return. The true challenge policy makers are facing is not primarily about “paying a green premium”, but one of influencing strategic choices of those investors who will deploy capital anyway, and are selecting between opportunities in conventional and renewable energy projects (Wüstenhagen & Menichetti, 2012). Investors compare opportunities by looking at their risk-adjusted return. With policy it is possible to change the risk-risk-adjusted return and in that way influence firm investment decisions.

Edler & Georghiou (2007), look at the justifications for the use of public procurement to spur innovation which relates to three levels. First, it is a major part of “local” demand, which is a major factor in the location decision of firms to generate innovation in a given location. Second, there is a range of market and system failures affecting the translation of need into functioning market for innovative products, and public procurement can prove effective in redressing this. Third, the purchase of innovative solutions offers a strong potential for contributing for public missions. The public mission would be lowering the level of emission. Contrary to this, Menanteau et al. (2003) say that the reduced margins inherent in the bidding system have limited the R&D investment capability of manufacturers and their suppliers. These reduced margins are the result of the competitive character of public procurement.

Besides the aspect of public technology procurement, several authors have investigated the factors that affect technological choices by firms. Acemoglu et al. (2012) found main implications in the importance of directed technical change:

(17)

16

- Optimal policy involves both carbon taxed and research subsidies, so that excessive use of carbon tax can be avoided.

- Delay in intervention is costly, the sooner and the stronger the policy response, the shorter will the slow growth transition phase be.

- The use of an exhaustible resource in dirty input production helps the switch to clean innovation under laissez-faire (when intervention is stopped).

The research of Noailly & Smeets (2012) differentiates between specialized and mixed firms. Specialized firms conduct innovation in only one type of technology and do not switch between these different technologies. Mixed firms may switch between technologies over time and substitute fossil fuels for renewable technologies. They find that for specialized firms main drivers of innovation are energy prices, market size and knowledge stocks. Prices and market size drive the entry of new renewable firms into the industry. They also find that innovation by mixed firms is characterized by strong path-dependencies (due to past knowledge stocks).

2.3.2 Policy and firm investments

(18)

17

(19)

18

3. RESEARCH DESIGN

In this section, the research design will be examined. It will start with the research method, in which it becomes clear how the research will be conducted. This includes the approach used to deliver the desired results and the statistical analysis needed for this. After that, the data collection method will explain the methods used in selecting the relevant data in order to minimize bias, especially concerning the use of patents. Finally, this chapter deals with reliability and validity issues.

3.1 Research method

This research starts with an extensive elaboration of the current literature. By looking at the renewable energy sector, the different policy mechanisms, and the interaction of these policies on a firm level, a framework is developed which serves as basis for this research.

In order to do a comparative assessment of innovative performance, a measure of innovation is developed. When looking at innovative activity across European countries, it is preferable to use a source containing data that is directly comparable. Patent data originating from the European Patent Office (EPO) collected by the OECD is designed for statistical analysis. According to Noailly & Batrakova (2010), patents have a close (if not perfect) link to invention. Patents have been used before to study policy in renewables, for instance by Noailly & Smeets (2012), but on a firm-level. This research will use the country level to investigate innovation. The table below sums up the advantages and disadvantages for using patents to reflect innovative activity:

Advantages Disadvantages

Patents provide information about a broad range of technologies, with sometimes few alternative sources of data available.

The propensity to file patent applications differs significantly across technical fields. (think of “patent flooding” strategy)

Patents have a close link to invention, as most of the significant inventions are patented.

Not all inventions are patented. Strategic considerations may lead the inventor to prefer alternative protection (e.g. secrecy) Each patent contains information on the

invention process, for instance the inventors name/address.

SMEs – organizations that lack large-scale production – have more difficulty covering the costs of a patent.

More than one million patents are applied worldwide each year providing insights into the progress of invention.

(20)

19

Patent data is public, free and easily accessible.

Differences in patent law around the world limit the comparability across countries. The spatial and temporal coverage of patent

data is unique. Nearly all of the world’s countries share patent data.

Changes in patent laws over the years call for caution when analysing trends over time.

Patent data is quite readily available from regional (and national) patent offices. Data is already collected.

-

Table 3.1 – advantages and disadvantages of patent use. Source: OECD, 2009.

The data collection method will deal to certain extent with the disadvantages by using the appropriate methodologies to minimize the amount of bias. Besides patent data, policy data is gathered. This will consist of the support schemes that were active in each country for each available year. With this, it is possible to do statistical analysis.

3.1.1. Statistical analysis

Two main sources, the book of Cameron & Trivedi (1998) and the website of the UCLA are consulted for the statistical analysis. When the relationship between a dependent and several independent variables is estimated, regression analysis is appropriate. It helps understand how the dependent variable changes when any one of the independent variables is varied, while other independent variables are held fixed (Cameron & Trivedi, 1998). The dependent variable is tested to see if it is the effect. So the effect would be an increase or decrease in the number of patent applications. It is possible to see if these effects are caused by the independent variables. These will consist of the used support schemes for renewables. All variables are specific for the selected European countries and when possible, for the period from 1990-2009.

The regression model most suited for the specific dataset heavily relies on the dependent variable, in this case the patents. It can be regarded as count data which simply means that the data refers to the number of times an event occurs. As Cameron and Trivedi (1998) say, an event count is the realization of a nonnegative integer-valued random variable. Several techniques deal with count data including the ordinary least squares (OLS), Poisson and negative binomial regression models.

(21)

20

techniques can be used. When analyzing the patent data it becomes clear that patent output is not normally distributed. The data is positively skewed as seen in graph 3.1, a histogram showing the distribution of patent applications in wind energy. The same goes for the other patent groups in the used dataset. This is the reason why OLS regression is not suitable. It is also visible that there are many zeros in the dataset. In this case these zeros represent the true values and are not regarded as excess zeros. Another option called a zero-inflated regression model is therefore not needed (UCLA (a), 2013).

Two methods remain, the Poisson regression and the negative binomial regression. According to Cameron and Trivedi (1998), they both make use of conditional probabilities, based on the fact that the probability of an event or outcome is based on the occurrence of a previous event or outcome. The negative binomial regression is regarded as a standard generalization of the Poisson regression (Cameron and Trivedi, 1998). A Poisson regression is more suited when the mean and variance are close to each other. When this is not the case, the data can be regarded as over-dispersed (UCLA (a), 2013).

(22)

21

parameter to model the over-dispersion in order to deal with this problem. The downside of using the negative binomial method is the existing criticism about using it with fixed effects. Besides that, there is not a magical cut-off determining when one model is better than the other (UCLA (b), 2013). In order to increase the robustness of the study, both methods will be reported. Poisson regression is complemented by the negative binomial regression in order to lower the limitations both methods engender. The Poisson regression will be the main focus and when differences occur concerning the significance levels, it will be evaluated.

A choice has to be made concerning random and fixed effects. When the situation occurs that one country has a high patent output relative to other countries, a bias may occur. Less developed countries have the tendency to act on developments in other countries with high patent output. The two countries are highly correlated over time due to country specific characteristics and therefore it would not be the policy that is measured. An example would be that Belgium has a higher probability to introducing a new policy than Germany, because it also strives for success in the transition to renewables, reacting on the success of Germany. The performance in the past of certain countries in the transition to renewables will be measured instead of the policy mechanisms themselves. Using fixed effects minimizes this bias. Comparison of the Poisson regression has taken place, both with fixed effects and random effects. The random effects were chosen because it is better able to model the over-dispersion in the Poisson regression. In the case of the negative binomial regression, the choice is also made to use random effects, especially because fixed effects have been criticized. The negative binomial model and its associated conditional likelihood estimator does not accomplish what is usually desired in a fixed-effects method (Allison & Waterman, 2002).

The final factor that has to be incorporated in the statistical analysis is the fact that policy measures often need time to create changes. It can be assumed that especially in the case of patents, where invention precedes a patent application, the time of effect does matter. Therefore a time correction of at least one year is implemented between de dependent and independent variables. When applying a forward lead in Stata for the dependent variable Y, it creates the situation that all patent applications are placed one year in the future, thereby creating the effect that policy needs time to develop and have influence.

(23)

22

3.2 Data Collection Method

3.2.1 Collection of patent data

Patents are divided among several technological areas, called the IPC classes, which range from section A to section H. A patent may contain several technical objects and therefore be assigned to several IPC classes (OECD, 2009). In general, when looking at certain technical areas, most of the relevant patents can be found by using several searching techniques. In the case of renewable energy patents, this division into several sections is not practical, because these patents are spread over several technological areas. To overcome this problem, the OECD has designed definitions of various technical fields for instance the one used in this research; environment-related technologies. The patents that fall under this field have been selected by technology experts to ensure the reliability of the data. At the OECD it is only possible to do statistical analysis, not to look at patents on the individual level. This research is among the first that uses this well developed database. This gives the advantage of investigating solar technology in three different areas (PV/thermal/hybrid), which has not been done until now. Graph 3.2 shows the patent applications for all the technological areas for the selected countries.

0 100 200 300 400 500 600

(24)

23

By using the OECD Patent Manual, appropriate choices for selecting the relevant patent set can be made. It is still common for analysts to compare patent counts from different patent offices to assess countries’ performance, although these are usually not directly comparable (OECD, 2009). This problem is solved by using the database of the OECD. The relevant patent database is selected by excluding the JPO (Japan Patent Office) and the USPTO (United States Patent and Trademark Office), and including the EPO (European Patent Office). The reason to use the EPO database is twofold. Firstly, because all patent applications are done under the same quality demands. And second, data availability is high and covers European countries in a comparable and timely manner. Triadic patent families and patent applications filed under the PCT are not included. It is also not possible to combine the patent counts because of the different time paths at different patent offices. In the case of a Euro-PCT application, the information on the effective transfer to the EPO is available 36 months after priority (first filing). This strongly influences the timeliness of patent indicators. So, focusing on one patent office in a certain area minimizes the amount of bias.

Information provided on the front page of a patent includes the address of the inventors and applicants. This information makes it possible to link patents to a particular region (OECD, 2009). Attention must be paid here when interpreting geographical data, notably in terms of activities by firms, as their research activity is spread geographically and the address of invention is not necessarily where the research actually took place. The reference country has to be the inventor’s country of residence because this reflects the country where the invention is made, thereby reflection the governmental policies that apply in that country. Also, when inventors of a patent are inventing in different countries, fractional counts will be used. This means that the share of a patent developed will apply to the correct country.

(25)

24

3.2.2 Collection of policy data

It is a challenge to find suitable information about policy which is comparable over time and countries. Support schemes are not always straightforward and suitable for grouping. Nevertheless, by using one main source and some complementary sources, it was possible to come up with the data. The report of Rathmann et al. (2009), published in accordance with the European Commission, provides data from 1997 till 2009. It is valuable data because the range is towards the more recent years which make it possible to investigate the more recent developments. Complementary sources had the function of increasing the reliability of the main source, which will be explained in the validity section. The report of Rathmann et al. (2009) focuses on renewable policies in electricity generation techniques. This is also the reason why biomass en bio fuels are not in this research because they are in respectively the heating and transportation sectors.

3.2.3 Country selection and timeframe

European countries are relevant to compare among each other because of the similarities in economic and political sense. Patent data of the selected technological areas is available for all the countries seen in graph 3.3. It clearly shows that Germany takes a leading role in the number of patent applications.

0 100 200 300 400 500 600 700

graph 3.3 - Patent applications at the EPO in selected renewables, by inventors country of residence.

(26)

25

The chosen time period of twenty years (1990-2009) is chosen because of several reasons. First of all, patent applications in renewables just started to rise in the chosen period, so extending the period backwards is not very useful. Second, because of interpretation reasons. The application of a patent follows strict administrative and legal rules and procedures, set out in international treaties and national statutes (OECD, 2009). These procedures have a direct impact on the value and the meaning of patent data. This is important as these rules are not fully harmonized across countries and have changed over time, and minor variation in the procedure can have drastic effect on the resulting numbers. A considerable amount of harmonization in patent rules across countries took place during the 1990s, notably through the creation of the Trade-related Aspects of Intellectual Property Rights (TRIPS) Agreement (OECD, 2009).

The limit has been set to 2009 because of data reliability. Publication of a patent generally takes place 18 months after the first filing, or priority date. No statistics are available until 18 months after the priority date, since the application is not published until then. This limits the legally possible timeliness of patent data. As mentioned before, this limit would increase when using Euro-PCT or other patent databases. The now used patents have the maximum time span to reliably analyze. The method of “nowcasting” (predicting patent applications) could eliminate this timeliness but in this case factual data is used.

3.2.4 Descriptive statistics

(27)

26

3.3 Validity

3.3.1 Reliability

Reliability is concerned with estimates of the degree to which a measurement is free of random or unstable error (Cooper & Schindler, 2006). Reliability is strong for the patent data, as it is factual information publicly available for everyone to see. Data on the support mechanisms is less reliable because it is only possible to typify a country within a support scheme, thereby losing sight on some nuances and differences between a group of support mechanisms. However, data is derived from a report made for the European Commission, which is a reliable source.

3.3.2 Internal Validity

When the conclusions drawn from a demonstrated experimental relationship truly imply cause, we can speak of internal validity (Cooper & Schindler, 2006). The number of patent applications has a strong relationship with innovative performance. It is however more difficult for the support schemes to be the only factors influencing innovative performance in renewables. There are more factors which are in play, not fully measured in this relationship.

It was possible to enhance the validity of policy data by comparing specific years with available data from publications of other authors than the original data used. The article of Reiche & Bechberger (2004) for instance, with policy data of the year 2002, was similar to the data obtained from the publication with the original policy data.

3.3.3 External Validity

(28)

27

4. RESULTS

Table 4.1 and table 4.2 show the coefficients and the associated significance levels of different support mechanisms related to the selected renewable technologies. Comparing the Poisson regression and the negative binomial regression, some differences appear. First of all, the Poisson regression shows higher significance levels especially for the feed-in tariff and the tender. However, for both methods the coefficients point in the same direction. The quota/TGC is in both methods significantly positive related to innovative performance in several areas. A difference however is clearly seen with wind technology. The tender shows more differences, also concerning the coefficient levels. In general they are negative with both models, but significance is not seen using the negative binomial regression. This may be due to the fact that the tender is not used very often. The tender related to solar hybrids with high negative levels is related to the fact that tenders were not in place for most countries that applied for patents in this technological area, visible in the significance level. The tax regulation support mechanism is widespread in coefficients along all technologies. Concerning wind energy, there is a small significant negative relation using Poisson but using the negative binomial regression this seems to disappear.

Table 4.1: Coefficients of the different support schemes using Poisson regression.

Wind Solar Thermal Solar PV Solar hybrids Geothermal Marine Small Hydro Feed-in tariff Coef. 0.217 1.164*** 0.559*** 2.035*** 0.598 0.276 0.269 P>|z| 0.125 0.000 0.000 0.000 0.179 0.419 0.441 Quota/TGC Coef. 0.165 1.260*** 0.472*** 2.106*** 0.414 0.201 0.095 P>|z| 0.270 0.000 0.000 0.000 0.382 0.580 0.794 Tender Coef. -1.443*** -0.800*** -0.913*** -15.667 -1.076 -1.503*** -1.491*** P>|z| 0.000 0.000 0.000 0.986 0.073 0.000 0.000 Tax Coef. -0.824* -0.861 -0.221 0.537 0.029 -0.665 -1.007 P>|z| 0.007 0.087 0.521 0.564 0.967 0.227 0.138 Notes: *** refers to a level of significance of P < 0.001 (** P < 0.005 & * P < 0.01)

(29)

28

Table 4.2: Coefficients of the different support schemes using negative binomial regression.

Wind Solar Thermal Solar PV Solar hybrids Geothermal Marine Small Hydro Feed-in tariff Coef. 0.526 0.904** 0.576 1.935** 0.528 0.096 0.224 P>|z| 0.063 0.004 0.038 0.001 0.301 0.808 0.569 Quota/TGC Coef. 0.869** 1.596*** 0.965** 2.327*** 0.923 0.581 0.938 P>|z| 0.003 0.000 0.001 0.000 0.064 0.170 0.026 Tender Coef. -0.631 0.052 -0.103 -20.485 -1.083 -1.101 -0.835 P>|z| 0.055 0.861 0.702 0.999 0.171 0.013 0.048 Tax Coef. -0.035 0.328 0.081 0.949 0.630 -0.214 -0.233 P>|z| 0.911 0.372 0.806 0.260 0.329 0.690 0.684 Notes: *** refers to a level of significance of P < 0.001 (** P < 0.005 & * P < 0.01)

The dependent variable is the patent count in a given technological area.

By looking at the statistical analysis it is possible to draw conclusions by looking at both methods2_{. Looking at individual support mechanisms and their performance, it can be said that}

the feed-in tariff is positively related to all technologies. Especially related to solar technology there are high significance levels. However the Quota/TGC, by looking at both statistical methods, seems to be even more effective in stimulating innovative performance in solar technology. Also this support mechanism is positively related to all technologies and for 50% of the cases this is significant to some extent. The two remaining support mechanisms are both less positively associated with innovative performance. Looking at both methods, the tender is negatively related in 92% of the cases. It seems that it is especially negatively related to wind, marine and small hydro technology in an often significant way. In the case of solar technology, the results are somewhat conflicting. The tax regulations are in line with the tender in the way that they are both negatively related to innovative performance. However, no strong positive or negative relations are seen and only in one case, wind technology, there is statistical significance.

(30)

29

5. DISCUSSION & CONCLUSION

In the following section, the findings from this study will be explained in relation to the existing literature. Differences with current understandings will be evaluated and discussed. The conclusion will sum up the most important results from the study and give some implications for the Netherlands. Finally, the limitations and future research possibilities will be described.

5.1 Discussion

According to the results, policy has an impact on the innovative performance in the form of patent applications, which is in accordance with the existing literature. Positive relations are seen for individual support mechanisms, and not so much along certain technological areas. This points out to the idea that the choice of support mechanisms has a differential influence on the technologies. It is not the case that one specific technological area is positively influenced by all types of support mechanisms.

The FIT is, according to existing literature, effective in stimulating the transition to renewables. According to the results, this price-based mechanism stimulates innovation in especially the solar technologies. The number of patent applications in solar PV would increase with a factor of 1.75 and in solar thermal with 3.20 based on the IRR. Germany has been able to innovate to great extent looking at the number of patent applications, and have been pioneering in the adoption of the FIT. However, this support mechanism does not stimulate innovative performance in wind energy. An explanation lies in the fact that wind technology, relative to solar, is regarded as a mature technology, which weakens the relation between adoption and innovation. When looking at the characteristics of the support schemes, the FIT does not discriminate between certain technologies. It takes away the risk of investing in a technology and thereby guaranteeing that the costs above the market price will be covered. FITs are granted to all new projects and continue for the whole pay off period. This effect will be stronger when the period of guarantee is longer, as in the case of Germany. An explanation of the positive relation of FIT towards innovative performance in solar could be that this is the most expensive technology. FITs cover relatively more of the electricity generating costs with this technology. Another or a combined explanation could be the price effect, directing innovation towards higher priced inputs where more can be saved.

(31)

30

imperfections and stimulating innovation, but in practice it seems to be ineffective. This can be explained by the characteristics of the tax regulations, not being able to lower the uncertainty associated with investing in renewables. This support mechanism does not provide investment security because of the absence of long-term commitments. It does not give incentives to innovate because of the risk it brings when dedicating resources, which has a strong effect because of the high upfront costs of renewables. Because it does not take away risk, this support mechanism is mainly used for existing technologies, where innovation is not that important compared to the second- and third-generation technologies. It could be used when investments are already done, pointing to the idea that it is used for existing technologies with less innovative potential.

The empirical findings say that the tender is negatively related to innovation in many areas. Based on the IRR, the number of patent applications in wind technology and solar PV would be expected to decrease with respectively a factor of 0.24 and 0.40, when the tender is in place. For marine and small hydro these values are even lower, respectively 0.22 and 0.23. Producers compete for projects based on their performance estimates and energy price. The explanation seems to be that the nature of the bidding systems means that profit margins are reduced and profitability rates lowered. Low margins increase the risk associated with investing. Covering the costs of the projects becomes the main goal, not able to invest in technologies. The projects are chosen according to the electricity price, which is only low with certain technologies. This support mechanism points to the cheaper technologies, often the ones that have received an amount of innovation in order to deliver a low price. It could be that the tender is stimulating the adoption of renewables but also points in a direction where innovation potential is low, therefore creating path-dependencies towards certain, more mature technologies.

(32)

31

When looking at the factors that distinguish the FIT and the quota/TGC from the tender and tax regulations, it becomes visible that, when looking from the perspective of a potential innovator, one is more risk eliminating than the other. The tax regulations do not give the guarantee that it will be in place for a long period, which increases the amount of risk for investments, unable to predict how the future return will be of investments. The tender creates risk by competition and lowering profit margins and risking low returns. Both these aspects have a negative effect on innovativeness. Both successful mechanisms eliminate risk to a certain extent, especially compared to the less effective mechanisms. With the quota/TGC and the FIT it is clear how the risk-adjusted return will be influenced, because both methods stimulate return in a fairly predictable way.

5.2 Conclusion

In this research it is investigated how different support schemes relate to innovative performance in renewable energy. This is measured by using patent data from several European countries over a period from 1990-2009 and among seven technological areas. This sector is growing continually and is supposed to become a main source of energy in the future. Governmental support can give direction to the development in a way that creates an innovative industry, able to satisfy growing energy needs and compete with conventional sources.

The empirical results from the study imply that governmental policy mechanisms are suitable to direct this technological path. Breaking down path-dependencies that are existing in fossil fuel technology and preventing lock-out of renewables. It is found that feed-in tariffs and quota in combination with tradable green certificates are most effective measures for stimulating innovation, mainly because they are lowering the risk of investing in renewables.

(33)

32

The tender system may be appropriate to increase the amount of renewables, however, historically this would not seem the most logical option, looking at for instance the UK. The government is depending on declining costs, something made possible through innovations. Looking at this research, this is a controversial choice, as innovation is negatively related to the tender mechanism. It could be that the government is depending on other factors. As capacities in other countries will grow, prices will lower and besides that, innovations from other countries could be used in the Netherlands, thereby ultimately reaching the price goals.

It is however wise to choose a quantity-based mechanism, as the government is planning precisely how much added capacity is needed because of international obligations. Besides that, it is positive that attention is paid to the investors. Having a long-term clear and consistent policy, thereby reducing risk, seems central in the agreement.

5.3 Limitations

Besides limitations that can be overcome with patent data, the use of patents has some issues not solvable. As mentioned before, not all inventions are patented. Firms can prefer secrecy, or rely on other mechanisms in order to gain market dominance. Also there is evidence of differing patenting behaviour across industries and countries over time. The value distribution of patents is known to be skewed; some have very high technical and economic value whereas many are ultimately never used. Simple counts, which give the same weight to all patents regardless of their value, can therefore be misleading. However, this limitation is lowered by using the EPO database.

As Reiche & Bechberger (2004) explain, geographical conditions are different among countries. This plays a role in the low innovation levels for some technologies, for instance geothermal and marine and small hydro. Besides that, it was not possible to get policy data from the complete time period of patent applications. The low use of some support mechanism for instance the tender caused some bias to the results, as visible with the solar hybrids. Also, it is not measured if innovation is caused by general innovation policies besides the support policies.

(34)

33

REFERENCES

Articles

Acemoglu, D., Aghion, P., Bursztyn, L., & Hemous, D. (2012). The environment and directed technical change. American Economic Review, 102 (1), 131-166.

Allison, P.D., & Waterman, R. (2002). Fixed-effects negative binomial regression models.

American Sociological Association, 32(1), 247-265.

Aschhoff, B., & Sofka, W. (2009). Innovation on demand – Can public procurement drive market success of innovations? Research Policy, 38 (8), 1235-1247.

Blokhuis, E., & Advokaat, B., & Schaefer, W. (2012). Assessing the performance of Dutch local energy companies. Energy Policy, 45, 680-690.

Boomsma, T.K., Meade, N., & Fleten, S.E. (2012). Renewable energy investments under different support schemes: A real options approach. European Journal of Operational Research, 220 (1), 225-237.

Costa, C.V., La Rovere, E., & Assmann, D. (2008). Technological innovation policies to promote renewable energies: Lessons from the European experience for the Brazilian case.

Renewable and Sustainable Energy Reviews, 12, 65-90.

Edler, J. & Georghiou, L. (2007). Public procurement and innovations – Resurrecting the demand side. Research Policy, 36 (7), 949-963.

Harmelink, M., Voogt, M., & Cremer, C. (2006). Analysing the effectiveness of renewable energy supporting policies in the European Union. Energy Policy, 34 (3), 343-351.

Held, A., Ragwitz, M. & Wietschel, M. (2005). Evaluation of renewable promotion schemes in the European Electricity market. Fraunhofer Institute for Systems and Innovation Research. 41-49. Retrieved from: http://d-nb.info/976574063/34#page=51

Kwant, K.W. (2002). Renewable energy in The Netherlands: policy and instruments. Biomass

(35)

34

Lauber, V. (2011). The European Experience with Renewable Energy Support Schemes and Their Adoption: Potential Lessons for Other Countries. Renewable Energy Law and Policy

Review, 120 (2), 120-132.

Lund, H. (2010). The implementation of renewable energy systems. Lessons learned from the Danish case. Energy, 35 (10), 4003-4009.

Menanteau, P., Finon, D., & Lamy, M.L. (2003). Prices versus quantities: choosing policies for promoting the development of renewable energy. Energy Policy, 31 (8), 799-812.

Meyer, N. (2003). European schemes for promoting renewables in liberalized markets. Energy

Policy, 31 (7), 665-676.

Noailly, J., & Smeets, R. (2012). Directing Technical Change from Fossil-Fuel to Renewable Energy Innovation: An Empirical Application Using Firm-Level Patent Data. CPB

Discussion Paper, 237.

Noailly, J., & Batrakova, S. (2010). Stimulating energy-efficient innovations in the Dutch building sector: Empirical evidence from patent counts and policy lessons. Energy Policy, 38 (12), 7803-7817.

Owen, A.D. (2006). Renewable energy: Externality costs as market barriers. Energy Policy, 34 (5), 632-642.

Popp, D., Hascic, I., & Medhi, N. (2011). Technology and the diffusion of renewable energy.

Energy Economics, 33 (4), 648-662.

Rooijen (van), S.N.M., & Wees (van), M.T. (2006). Green electricity policies in the Netherlands: an analysis of policy decisions. Energy Policy, 34(1), 60-71.

Reiche, D., & Bechberger, M. (2004). Policy differences in the promotion of renewable energies in the EU member states. Energy Policy, 32(7), 843-849.

(36)

35

Schmidt, R.C., & Marschinski, R. (2009). A model of technological breakthrough in the renewable energy sector. Ecological Economics, 69, 435-444.

Unruh, G.C. (2000). Understanding carbon lock-in. Energy Policy, 28(12), 817-830.

Wüstenhagen, R., & Menichetti, E. (2012). Strategic choices for renewable energy investment: Conceptual framework and opportunities for further research. Energy Policy, 40 (1), 1-10.

Reports

CBS, (2011). Renewable energy in the Netherlands 2011. Statistics Netherlands. The Hague, The Netherlands.

CBS, (2012). Economic Radar of the Sustainable Energy Sector in the Netherlands. Statistics

Netherlands. The Hague, The Netherlands.

IEA, (2007). Renewables in Global Energy Supply: An IEA Fact Sheet. IEA Publications.

OECD, (2009). Patent Statistics Manual, OECD Publishing.

Planbureau voor de Leefomgeving, (2013). Vergroenen en verdienen, op zoek naar kansen voor de Nederlandse economie. Den Haag.

Rathmann, M., Winkel, T., Stricker, E., Ragwitz, M., Held, A. & Pfluger, B. (2009). Renewable Energy Policy Country Profiles. RE-Shaping. Retrieved from: http://www.reshaping-res-policy.eu/

Rathmann, M., de Jager, D., de Lovinfosse, I., Breitschopf, B., Burgers, J., & Weöres, B. (2011). Towards triple-A policies: More renewable energy at lower cost. RE-Shaping. Retrieved from: http://www.reshaping-res-policy.eu/

Books

(37)

36

Cooper, D. & Schindler, P. (2006). Business Research Methods (Ninth Edition). NY: McGraw-Hill Education.

Websites

UCLA(a); Stata FAQ. How can I analyze count data in Stata? Statistical Consulting Group. From http://www.ats.ucla.edu/stat/stata/faq/count.htm (accessed November 26, 2013).

UCLA(b); Stata FAQ. How can I use countfit in choosing a count model? Statistical Consulting Group. From http://www.ats.ucla.edu/stat/stata/faq/countfit.htm (accessed November 26, 2013).

(38)

37

APPENDIX

Appendix A – Poisson regression (panel/F1./re)

Wind energy

Random-effects Poisson regression Number of obs = 180 Group variable: country Number of groups = 15 Random effects u_i ~ Gamma Obs per group: min = 12 avg = 12.0 max = 12 Wald chi2(4) = 103.14 Log likelihood = -1019.6071 Prob > chi2 = 0.0000 --- Wind | Coef. Std. Err. z P>|z| [95% Conf. Interval] ---+--- FIT | .2168676 .1415117 1.53 0.125 -.0604902 .4942254 Quota/TGC | .1646605 .1492305 1.10 0.270 -.1278258 .4571469 Tender | -1.443478 .1814282 -7.96 0.000 -1.799071 -1.087885 Tax | -.8242724 .305579 -2.70 0.007 -1.423196 -.2253485 _cons | 2.586038 .3533177 7.32 0.000 1.893548 3.278528 ---+--- /lnalpha | .4506347 .3123412 -.1615429 1.062812 ---+--- alpha | 1.569308 .4901595 .85083 2.894499 --- Likelihood-ratio test of alpha=0: chibar2(01) = 3642.07 Prob>=chibar2 = 0.000

Solar thermal energy

(39)

38

Solar photovoltaic (PV) energy

Random-effects Poisson regression Number of obs = 180 Group variable: country Number of groups = 15 Random effects u_i ~ Gamma Obs per group: min = 12 avg = 12.0 max = 12 Wald chi2(4) = 132.54 Log likelihood = -892.37156 Prob > chi2 = 0.0000 --- PV | Coef. Std. Err. z P>|z| [95% Conf. Interval] ---+--- FIT | .5595472 .112223 4.99 0.000 .3395941 .7795003 Quota/TGC | .4722142 .1171327 4.03 0.000 .2426383 .7017902 Tender | -.9134929 .1104153 -8.27 0.000 -1.129903 -.6970828 Tax | -.2211763 .3446724 -0.64 0.521 -.8967218 .4543691 _cons | 2.129833 .3615077 5.89 0.000 1.421291 2.838375 ---+--- /lnalpha | .5586846 .3090496 -.0470415 1.164411 ---+--- alpha | 1.748371 .5403334 .9540478 3.204034 --- Likelihood-ratio test of alpha=0: chibar2(01) = 3888.54 Prob>=chibar2 = 0.000

Solar thermal-PV hybrids

(40)

39

Geothermal energy

Random-effects Poisson regression Number of obs = 180 Group variable: country Number of groups = 15 Random effects u_i ~ Gamma Obs per group: min = 12 avg = 12.0 max = 12 Wald chi2(4) = 6.57 Log likelihood = -176.10876 Prob > chi2 = 0.1602 --- Geothermal | Coef. Std. Err. z P>|z| [95% Conf. Interval] ---+--- FIT | .5980504 .4445656 1.35 0.179 -.2732821 1.469383 Quota/TGC | .4138011 .4732991 0.87 0.382 -.5138481 1.34145 Tender | -1.076185 .6008612 -1.79 0.073 -2.253852 .1014809 Tax | .0292071 .7005247 0.04 0.967 -1.343796 1.40221 _cons | -.7512954 .6173972 -1.22 0.224 -1.961372 .458781 ---+--- /lnalpha | .9842767 .4164503 .1680492 1.800504 ---+--- alpha | 2.675876 1.114369 1.182995 6.052698 --- Likelihood-ratio test of alpha=0: chibar2(01) = 203.10 Prob>=chibar2 = 0.000

Marine energy

(41)

40

Hydro energy (tidal/stream/damless)

Random-effects Poisson regression Number of obs = 180 Group variable: country Number of groups = 15 Random effects u_i ~ Gamma Obs per group: min = 12 avg = 12.0 max = 12 Wald chi2(4) = 49.09 Log likelihood = -263.73937 Prob > chi2 = 0.0000 --- Small Hydro | Coef. Std. Err. z P>|z| [95% Conf. Interval] ---+--- FIT | .2696707 .3497866 0.77 0.441 -.4158984 .9552399 Quota/TGC | .0952687 .3646527 0.26 0.794 -.6194374 .8099748 Tender | -1.491216 .3460728 -4.31 0.000 -2.169506 -.8129259 Tax | -1.007184 .6787829 -1.48 0.138 -2.337574 .3232056 _cons | .5530943 .4778922 1.16 0.247 -.3835571 1.489746 ---+--- /lnalpha | .462672 .3495491 -.2224317 1.147776 ---+--- alpha | 1.588312 .5551932 .8005696 3.151176 --- Likelihood-ratio test of alpha=0: chibar2(01) = 282.74 Prob>=chibar2 = 0.000

Appendix B – Negative binomial regression (panel/F1./re)