The effect of environmental policies on
the production of renewable energy
Fien Verheij
student number: 1128811611-08-2017 supervisor: Erik Plug
Master Thesis University of Amsterdam Department of Economics
Abstract
This paper analyses the effect of four different environmental policies on the production of renewable energy, using panel data from 1990 until 2012 from 14 western-European countries. The use of a new policy stringency index, composed by the OECD, is the main contribution of this paper. Furthermore, this research allows for possible simultaneous determination of environmental policy and renewable energy production.
The results show that public R&D expenditures and renewable energy certificates are significantly influencing renewable energy production, contrary to environmental taxes and feed-in-tariffs. Especially renewable energy certificates have a large impact, suggesting regulatory instruments are more effective than price-based policies. Furthermore, there is evidence of simultaneous causality; it is therefore possible that environmental policies respond to the production of renewable energy. More research is needed to confirm this and until then, the results above should be interpreted cautiously.
Keywords: environmental policy, renewable energy, renewable energy certificates, public R&D expenditures
2
1. Introduction
This paper aims to answer the following question: “What is the effect of environmental policies on the production of renewable energy?” There is little discussion on the importance of reducing the emission of greenhouse gases by diminishing the use of fossil fuel energy. In 2015, emissions caused by fuel combustion and fugitive emissions from fuels produced 55 percent of total emissions in the European Union, not even taking into account emissions caused by the transport sector1 (Eurostat, 2017). Limiting greenhouse gas emissions at a global level will therefore require a major shift away from fossil fuels and an extreme increase in the production of renewable energy. In 2010, the European Commission stated: “The energy challenge is one of the greatest tests Europe has to face” (p.3).
Thus, one of the strategic objectives of the European Union is the development of a resilient energy union with a forward-looking climate policy. The EU’s Renewable Energy Directive has set up a binding target of 20 percent of total energy consumption from renewable energy sources by 2020 (European Commission, 2017). In setting this goal, the EU is not only driven by climate change mitigation through the reduction of greenhouse gasses; renewable energy also fosters sustainable production and is emerging as an inclusive driver of economic growth. Furthermore, it can enhance energy security across Europe (European Commission, 2017). All European countries have adopted a national action plan and implemented a different mix of public policies to reach this target.
As the move towards renewable energy is thus becoming more and more pressing, countries have to figure out the quickest and most efficient way to do so. Especially the countries who are lagging behind on their renewable energy goals, such as France and the Netherlands, have to identify the best policies to accelerate their transfer to renewable energy. It is therefore crucial to study the effects of environmental policies.
Various papers analysed the effectiveness of environmental policies, although a limited amount specifically focused on their impact on the production of renewable energy. As the different sectors of renewable energy are relatively infant industries, especially in the 1980s and 1990s, and innovation is the core element of environmental improvements, the majority of the papers examined the effect of environmental policies on technological innovation within energy sectors (Song, 2011).
The first to empirically test this were Lanjouw and Mody (1995), who concluded that an increased stimulation of environmental protection by governments in the 1970s and 1980s
3 led to the development of new pollution control technologies. However, they did not control for other factors that might influence environmental innovation. Jaffe and Palmer (1996) extended this study and measured innovation not only by R&D expenditures, they also used patents in the renewable energy sectors. They concluded that pollution abatement expenditures, the policy measure they examined, had no significant impact on innovation.
Contrary, Brunnermeier and Cohen (2003) only looked at environmentally related patents as a proxy for innovation and found that pollution abatement expenditures did have a positive impact on innovation. As a measure of policy stringency, they included the number of pollution related visits by the government. Although they found a significant result, the magnitude of these results were relatively small: mean patenting increased by 0.04 percent when industry abatement expenditures rose with one million dollar (Brunnermeier & Cohen, 2003).
De Vries and Withagen (2005) tested the effect of environmental policy stringency on innovation in the field of SO2 abatement, measured by patents in that field. They assumed
strictness of environmental policy is captured by international agreements, which became more stringent over time. Compliance towards these agreements will therefore enhance domestic stringency. They however could not find a clear effect of environmental policies on innovation. This was unsurprising as their measure of environmental stringency did not fully reflect policy variability (Johnstone et al., 2009).
Popp (2006), compared data from Germany, Japan and the US and found that a tightening of domestic regulation had a positive impact on technological innovation regarding air pollution control.2 Haščič et al. (2009) examined patenting activity in automotive emission-control technologies for a cross-section of OECD countries. They concluded that both foreign and environmental domestic policies had a positive effect on patenting activity, although its magnitude depended on the type of technology induced.
Johnstone et al. (2009) also used patent counts as a proxy for innovation and found that public policies had a positive and significant effect. This paper showed how different sets of policy measures affect different areas of renewable energy technologies and vary in their effectiveness. Especially renewable energy certificates were found to have a large impact on innovation. Johnstone et al. (2012) used data from the World Economic Forum’s ‘Executive Opinion Survey’ that included responses about policy stringency and therefore measured perceived policy stringency. They confirmed a positive effect of policy stringency on
2 Popp (2006) used patents for SO
4 environment-related technology.
By estimating a simultaneous panel data model of environmental innovation and toxic air pollution, Carrion-Flores and Innes (2010) found a significant but very small effect of environmental policies on innovation in the energy sector.
All papers above focused on innovation. The majority found that environmental policies had a significant and positive effect on innovation, although this effect was often small. However, it is not certain that the stimulation of innovation in energy sectors has been translated into an actual increase in renewable energy production. Only a couple of papers recognized this problem and estimated the effect of environmental policies on the production of renewable energy.
Carley (2009) investigated the effect of different renewable portfolio standards (RPS), one of the most prevalent policy measures in the US, on renewable energy production. She found ambiguous results and concluded RPS failed to increase the share of renewable energy. However, Menz and Vachon (2006) found that RPS did have a significant positive effect on the capacity of wind power in the US. Schmid (2011) examined the production of renewable energy in nine Indian states from 2001-2009 and found that national policies had a more profound impact than state-level environmental policies. Also, quotas had a significant impact on renewable energy production, while feed-in-tariffs did not.
Whereas these papers only focused on a particular country, Song (2011) was the only one that investigated the effectiveness of environmental policies in 26 countries. She looked at the effect of multiple policies on the overall share of renewable energy of total energy production. This study found that quantity restrictions have been more effective than price-based policies. Energy suppliers are more stimulated to produce renewable energy if they are going to be fined when not meeting the quote, as opposed to receiving subsidies for producing renewable energy.
This paper builds upon existing literature by extending the paper by Song (2011). I will estimate the effect of different environmental policies on the production on renewable energy from 1990 until 2012, while Song (2011) used panel data from 1990 until 2004. The environmental policies that are included in this research are public R&D expenditures, environmentally related taxes, feed-in-tariffs and renewable energy certificates.
Furthermore, a new variable will be used to measure the stringency of these environmental policies. Empirical research on the economic effects of environmental policies relies heavily on the complex task of evaluating the stringency of these policies (Albrizio et al., 2014; Koźluk & Zipperer, 2014). Multiple papers have tried to estimate this by using
5 various methods, like some papers described above. They used surveys on the perceptions of stringency, the number of government inspections or the number of people working in the R&D department in an attempt to capture policy stringency (in the latter case the stringency of R&D expenditures) (Botta & Koźluk, 2014). These methods are very limited as they provide little information about actual policy stringency.3 The lack of plausible measures posed a major problem for many papers and has led to biased results.
In 2014, the OECD launched a new composite index to measure environmental policy stringency, which covers most OECD countries from 1990 until 2012. To the best of my knowledge, this new index was never used in prior research to examine its effect on renewable energy production. Therefore, I will try to use this index for this purpose and aim to contribute to a better insight in the relation between environmental policies and the production of renewable energy. In section three, the composition of this index will be explained extensively.
Another problem that has to be mentioned is that most papers described above did not acknowledge the possibility of simultaneous causality; environmental policy could respond to the production of renewable energy. A couple of papers (Carley, 2009; Johnstone et al., 2011) mention the fact that environmental policy can be endogenously determined, but do not include this in their methodology. This paper not only acknowledges this but will also empirically check the presence of reverse causality.
The main findings of this paper are the following. Public R&D expenditures and renewable energy certificates are found to have a positive and significant impact on renewable energy production, while feed-in-tariffs and taxes do not. Renewable energy certificates are the most efficient policy measure; when the stringency index of this policy goes up with one index point, renewable energy increases with 21 percent. However, the results also suggest simultaneous causality is present in the data, which indicates that the results above are biased. Therefore, they should be interpreted with great caution until further research is done.
The rest of the paper is organised as follows. Section two describes the context of renewable energy production and environmental policies in Europe. Here, the different policies will also be explained. Section three describes the data and how these were collected. The empirical strategy is explained in section four and section five describes the results.
3
The number of government inspections for example give little information about the quality of those inspections and other aspects of the implementation of the policy.
6 Lastly, there will be a conclusion with a brief discussion and the limitations of this paper.
2. Context
The production of renewable energy in the EU has been steadily increasing over the last couple of decades, as shown in figure 1. Between 2004 and 2014, the share of renewable energy of total energy consumption almost doubled, reaching 16 percent of gross final consumption in 2014 (Eurostat 2020 indicators, 2016). Renewable energy is defined as follows: “energy derived from natural processes that are replenished constantly. In its various forms, it derives directly or indirectly from the sun, or from heat generated within the earth” (International Energy Agency, 2017, p. 42). Included in the definition is energy generated from solar, wind, biofuels, geothermal, hydropower and ocean resources, and biofuels and hydrogen derived from renewable resources. As can be seen in figure 1, hydropower4 is still accounting for the largest share of total renewable energy, although solar- and wind power are catching up.
Figure 1. Gross electricity generation from renewable energy sources in the EU
source: Eurostat
The EU was the largest investor in renewable energy until 20135 and, together with the US, the only world region that increased its share of renewable energy from 2005 until 2013 (European Environment Agency, 2016). However, to reach the goal of 20 percent of renewable energy consumption in 2020, more far-reaching measures are necessary than the
4 Potential and kinetic energy of water converted into electricity in hydroelectric plants (International Energy
Agency, 2015).
5 In this year, the EU was surpassed by China, who accounted for 27 percent of global new investments in
7 policies that are currently in effect (Klessmann et al., 2011). The share of renewables also varies greatly among European countries (figure 2). While some member states, like Sweden6, have already reached their target, other countries need to increase this share dramatically. Especially France, the United Kingdom and the Netherlands are lagging behind on renewable energy goals. If no action is taken in the Netherlands, the share of renewables will be stuck at 12.5 percent in 2020, while the target is 14 percent (Energieverkenning, 2016).
Figure 2. The share of renewables in 2014 of gross final energy consumption (%) (green) and the legally binding targets in 2020 (pink)
Source: Eurostat
The debate about climate change and the energy transition has received an enormous amount of attention in the last couple of decades and has been more present on the political agenda than ever. The pressure on governments to take action also increased dramatically. Therefore, the number of environmental policies has risen, just as the stringency of these policies. Generally, the International Energy Agency distinguishes six types of environmental policy measures: research and development, investment incentives, taxes, tariffs, voluntary programs and quantity obligations, such as permits (International Energy Agency, 2006).
First, in the 1970s, numerous countries introduced support for R&D since all renewable energy sectors were still infant industries. Investment incentives, taxes and
8 support polies soon followed (Johnstone et al., 2009). Energy permits are the most recently introduced measures, especially certificates for which the obligations are tradable across generators. This was encouraged by the implementation of the EU Emission Trading System in 2005, while comparable methods were already introduced on a national level in many member states (European Commission, 2016). There are several types of tradable permits, such as energy efficiency certificates, 𝐶𝑂2 trading schemes and renewable energy certificates.
Relying on available data, this paper focuses on four policy measures: R&D expenditures, environmentally related taxes, feed-in-tariffs and renewable energy certificates. R&D expenditures entail all national public R&D expenses on renewable energy.
Environmentally related taxes are defined by the OECD as: “any compulsory, unrequited payment to government levied on tax bases deemed to be of environmental relevance, i.e. taxes that have a tax base with a proven, specific negative impact on the environment” (OECD, 2016, p.2). There needs to be a redistributive element in order for it to be considered a tax. The tax bases include: energy products, transport equipment and transport services, pollution and natural resources, including mining (OECD, 2016).
A feed-in-tariff is a long-term contract that offers a fixed price for the produced renewable energy and thus offers a premium over the energy market for the duration of the contract (OECD, 2016). It is thus a subsidy and aims to accelerate investments in the renewable energy sectors.
Renewable energy certificates are another kind of market-based policy. It provides an obligation to produce a percentage of renewable energy and is thus a quota limit of fossil fuel energy, relative to total energy production. The certificates are issued under a trading system (OECD, 2016). Economic actors who produce less than the maximum permitted amount of fossil fuel energy can sell the certificates to others who need them in order to comply to regulations.
In figure 3, the timing of the introduction of the four different policies in the 14 countries that are included in this research, is shown. Generally, R&D was the first measure that was introduced. The specific tradable permit that this paper is focussing on, renewable energy certificates, is only introduced in five countries so far. A feed-in-tariff is still not introduced in Norway and Belgium, in contrast to the rest of the countries.
9
Figure 3. Introduction of renewable energy policies by type
A Austria, B Belgium, DK Denmark, B Belgium, FR France, DE Germany, IR Ireland, IT Italy, NL Netherlands, NO Norway, P Portugal, E Spain, SW Sweden, CH Switzerland, UK United Kingdom. Source: OECD/International Energy Agency
3. Data
The empirical analysis of this paper is based on panel data retrieved from multiple sources and covers the period 1990-2012. The data are quantified on an annual basis and include fourteen western-European countries7.
In this analysis, I use total primary renewable energy supply as the outcome variable, which is measured in millions of toe8. This is retrieved from the OECD Green Growth Indicators Database and is calculated as follows:
𝑇𝑜𝑡𝑎𝑙 𝑝𝑟𝑖𝑚𝑎𝑟𝑦 𝑟𝑒𝑛𝑒𝑤𝑎𝑏𝑙𝑒 𝑒𝑛𝑒𝑟𝑔𝑦 𝑠𝑢𝑝𝑝𝑙𝑦 =𝑟𝑒𝑛𝑒𝑤𝑎𝑏𝑙𝑒 𝑒𝑛𝑒𝑟𝑔𝑦 𝑠𝑢𝑝𝑝𝑙𝑦(% 𝑜𝑓 𝑇𝑃𝐸𝑆)100 ∗ 𝑇𝑃𝐸𝑆9
The logarithm of this variable is taken to reduce its skewedness and normalize the data (see appendix I). When taking the logarithm, the results can be interpreted as a percentage
7
Austria, Belgium, Denmark, France, Germany, Ireland, Italy, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland and the United Kingdom. The only country that is not a member state of the EU is Norway.
8 A tonne of oil equivalent (toe) is a unit of energy defined as the amount of energy generated by burning one
tonne of crude oil, which is more or less equal to 11.63 MWh (International Energy Agency, 2006).
9
10 change10, which makes them better suited for interpretation.
As explanatory variables, four different environmental policy measures are distinguished. Firstly, the variable ‘national public R&D expenses on total renewable energy’ (R&D) is retrieved from the International Energy Agency’s Energy Technology Research and Development Database and measured in millions11. As this variable was also highly skewed, the logarithm of the variable is taken as well (see appendix I).
The explanatory variable ‘environmentally related tax revenue’ (Tax) is provided by the OECD’s Policy Instruments for the Environment (PINE) Database. It is measured as a percentage of GDP of the country.
The third independent variable is ‘renewable energy certificates’ (REC) and is a binary variable. It takes on the value 1 if this policy is in effect and 0 otherwise. This was retrieved from the OECD Renewable Energy Policy Dataset, which was lastly updated in March 2013. The fourth independent variable ‘feed-in-tariffs’ (FIT) is taken from the same dataset. This is also a binary variable and takes on the value 1 if the policy is implemented.
However, it is most preferable to not only measure the presence of a policy measure (which is done by the binary variables), also the stringency of a policy is relevant. Furthermore, the independent variables above are difficult to compare since two are continuous and two are binary. Therefore, a uniform measure of the policy stringency of all four different policy measures is used in addition to the variables described above. This is the Environmental Policy Stringency Index (EPS), composed by the OECD.
In this context, environmental stringency is defined as “the strength of the environmental policy signal – the explicit or implicit cost of environmentally harmful behaviour, for example pollution” (OECD, 2016, p.3). This is straightforward for policies like taxes; a higher tax rate will imply higher policy stringency. Subsidies like R&D expenditures and feed-in-tariffs have the same interpretation. The higher the subsidy, the higher the opportunity costs of using fossil fuel energy and the more stringent the policy. For renewable energy certificates, this entails that the higher the percentage of electricity that must be generated from renewable energy sources, the more stringent the measure is (Koźluk & Garsous, 2016). To see the structure of the EPS index, see appendix II. The EPS indices used in this research are marked in red. The total composite indicator of policy stringency (EPS Total), also included in the descriptive statistics in table 1, is only used to check the relation between the policy variables and the EPS index (see footnote 16 in section 5).
10 See footnote 17 and 18. 11
11 The data that were used by the OECD for the composition of the four indices used in this research are the following. For the R&D index, they used R&D expenditures on renewable energy technologies as a percentage of GDP. For the tax index, tax rates on the emission of 𝐶𝑂2, 𝑁𝑂𝑥 and 𝑆𝑂𝑥 are used. For the feed-in-tariff index, the rate in euros per generated KWh for solar- and wind energy is used. Lastly, the percentage of renewable energy that has to be procured annually was used for the renewable energy certificates index (Botta & Kuźluk, 2014). These data were categorized, scored and aggregated into EPS indices12. The EPS ranges from 0 to 6, 6 being the most stringent possible.
Lastly, this research controls for energy prices, which was taken from the International Energy Agency’s Energy Prices and Taxes Database. This database provides data on the average annual industry end-user prices (PI) and residential end-user prices (PR), measured per MWh13.
Energy prices may be negatively related with the production of renewable energy; the higher the prices, the less likely the government is willing to invest in relatively more expensive renewable energy sources (Carley, 2009). Also, the demand for renewable energy will fall when the prices rise. On the other hand, the production can increase as a rise in prices has the potential to make renewable energy more economically feasible (Carley, 2009). However, since there are no data available on the renewable energy prices, these weighted average prices include both prices of fossil fuel energy and renewable energy. Because renewable energy represents a relatively small portion of total energy production, especially in the years 1990-2012, a rise in average price can entail a rise as well as a fall in renewable energy price. Therefore, based on theory, it remains unclear whether to expect a positive or negative relationship between the average prices and renewable energy production. Carley (2009) however, used the same control variables and found that average energy prices across all end-users has a negative and statistically significant effect on renewable energy generation. Therefore, this paper expects to find the same.
Table 1 shows the descriptive statistics of all variables. As this research covers 14 countries over 23 years, it is expected that for all variables 322 observations are present. For the policy measures and the controls, there are thus some missing observations. However, the data are still strongly balanced14.
12
For the exact composition of the indices and the raw data that were used: see
http://www.oecd.org/eco/growth/Do-environmental-policies-matter-for-productivity-growth.htm.
13 In euros, in 2015 prices and exchange rates.
14 According to STATA, the software package used for his research. A panel dataset with a time variable is
12 As to be expected, because R&D expenditures were the first environmental measure to be implemented, this EPS index has the highest mean and thus is the most stringent policy in general. REC, implemented very recently, has the lowest policy stringency mean. The descriptive statistics also show that the mean of residential energy prices is considerably higher than of industry prices. For an overview of all variables, see appendix IV.
Furthermore, a pairwise correlation matrix is presented in appendix III to check for multicollinearity. When two or more regressors are highly correlated, the coefficients of at least one individual regressor will be imprecisely estimated (Stock & Watson, 2012). The only two variables with a correlation higher than 0.5 are PI and PR, which is not surprising as these variables both represent energy price levels. Generally, the correlations are low so I conclude that multicollinearity is unlikely to distort my findings. Lastly, the data were checked for outliers. Two observations of R&D expenditures were more than a hundred times larger than the observations of the years afterwards. Consequently, I removed those observations so they do not distort the data.
Table 1: Descriptive Statistics
Variable N Mean Std. Dev. Min Max
Dependent variable
Total primary renewable
energy supply 322 7.261 6.219 .155 32.249 Log(total primary renewable
energy supply) 322 1.458 1.219 -1.867 3.474
Policies Public R&D expenditures 298 40.422 43.337 0 261.364
Log(public R&D expenditures) Tax revenue 297 266 3.034 2.721 1.399 .749 -1.266 1.376 5.566 5.386 Renewable energy certificates** 308 .130 .337 0 1 Feed-in-tariffs** 308 .552 .498 0 1
Indices EPS* Public R&D
expenditures 322 2.394 1.315 1 6
EPS* Tax revenue 322 1.488 .612 .25 4
EPS* Renewable energy
certificates 322 .807 1.246 0 5.2 EPS* Feed-in-tariffs EPS* Total 322 322 1.643 1.933 1.853 .849 0 .479 6 4.133
Controls Energy price industrial 287 76.712 39.574 18.413 209.727
Energy price residential 306 147.100 55.111 49.434 297.48
Notes: The number of observations, mean, standard deviation and the minimum and maximum values of each variable. **Binary variable *Environmental Policy Stringency Index
13
4. Empirical Method
I will estimate the effect of both types of independent variables (the policy measures and EPS indices) on the production of renewable energy with a fixed effects model. Since panel data were used for this research, a fixed effects model is most appropriate because it acknowledges the heterogeneity between the different countries. The following model is estimated:
log (𝑅𝐸𝑃)𝑖,𝑡 = 𝛽0+ 𝛽1(𝑃𝑜𝑙𝑖𝑐𝑦𝑖,𝑡) + 𝛽2(𝑃𝑟𝑖𝑐𝑒 𝑅𝑒𝑠𝑖𝑑𝑒𝑛𝑡𝑖𝑎𝑙𝑖,𝑡) + 𝛽3(𝑃𝑟𝑖𝑐𝑒 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑖,𝑡) +
𝜆𝑖 + 𝜆𝑡+ 𝜀𝑖,𝑡 (1)
Where i = 1,…,14 indexes stand for the unit (country) and t = 1990,…,2012 indexes time. 𝜆𝑖 represents the fixed effects term that captures unobserved country-specific heterogeneity that is constant over time. Similarly, 𝜆𝑡 controls for the heterogeneity that is
constant over all countries. The inclusion of these country- and time fixed effects mitigates the threat of omitted variable bias arising from unobserved variables. Besides these fixed effects, the model controls for average residential- and industry energy prices (𝑃𝑟𝑖𝑐𝑒 𝑅𝑒𝑠𝑖𝑑𝑒𝑛𝑡𝑖𝑎𝑙, 𝑃𝑟𝑖𝑐𝑒 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦). log (𝑅𝐸𝑃) is the natural logarithm of the outcome variable ‘total primary renewable energy supply’. The explanatory variable 𝑃𝑜𝑙𝑖𝑐𝑦 represents the different policy measures as described in the previous section: R&D expenditures, tax revenue, renewable energy certificates and feed-in-tariffs. The same is done for the EPS indices:
log (𝑅𝐸𝑃)𝑖,𝑡 = 𝛼0+ 𝛼1(𝐸𝑃𝑆𝑖,𝑡) + 𝛼2(𝑃𝑟𝑖𝑐𝑒 𝑅𝑒𝑠𝑖𝑑𝑒𝑛𝑡𝑖𝑎𝑙𝑖,𝑡) + 𝛼3(𝑃𝑟𝑖𝑐𝑒 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑖,𝑡) +
𝜆2𝑖+ 𝜆2𝑡+ 𝜀2𝑖,𝑡 (2)
Where 𝐸𝑃𝑆 stands for all four EPS indices. For each dependent variable, the regression will be estimated separately to be able to compare their individual effects on renewable energy production accurately. After this, the effects of the four policy measures will be estimated together. Then, the model will be as follows:
log (𝑅𝐸𝑃)𝑖,𝑡 = 𝛾0+ 𝛾1(𝑅&𝐷𝑖,𝑡) + 𝛾2(𝑇𝑎𝑥𝑖,𝑡) + 𝛾3(𝑅𝐸𝐶𝑖,𝑡) + 𝛾4(𝐹𝐼𝑇𝑖,𝑡) +
14 Here, 𝑅&𝐷, 𝑇𝑎𝑥, 𝑅𝐸𝐶 and 𝐹𝐼𝑇 both represent the policy variables as well as the EPS indices of the corresponding policy measures. As the two groups of explanatory variables basically measure the same, I do not expect any large differences between the two. However, to take possible measurement error in the explanatory variables into account, I also perform an instrumental variable regression with the two groups of independent variables. Measurement error can result in correlation between the regressor and the error term 𝜀𝑖,𝑡 (endogeneity), which will lead to biased estimates. An IV approach can mitigate this bias (Stock & Watson, 2012). The variables R&D, Tax, REC and FIT will act as instrumental variables to estimate the effect of the EPS indices on the production of renewable energy. This model has the following form:
log (𝑅𝐸𝑃)𝑖,𝑡 = 𝜂0+ 𝜂1(𝐸𝑃𝑆𝑖,𝑡) + 𝜂2(𝑃𝑟𝑖𝑐𝑒 𝑅𝑒𝑠𝑖𝑑𝑒𝑛𝑡𝑖𝑎𝑙𝑖,𝑡) + 𝜂3(𝑃𝑟𝑖𝑐𝑒 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑖,𝑡) +
𝜆4𝑖+ 𝜆4𝑡+ 𝜀4𝑖,𝑡 (4)
𝐸𝑃𝑆𝑖,𝑡 = 𝜋0 + 𝜋1(𝑃𝑜𝑙𝑖𝑐𝑦𝑖,𝑡) + 𝜋2(𝑃𝑟𝑖𝑐𝑒 𝑅𝑒𝑠𝑖𝑑𝑒𝑛𝑡𝑖𝑎𝑙𝑖,𝑡) + 𝜋3(𝑃𝑟𝑖𝑐𝑒 𝐼𝑛𝑑𝑢𝑠𝑡𝑟𝑦𝑖,𝑡) + 𝜆5𝑖+
𝜆5𝑡 + 𝜀5𝑖,𝑡 (5)
Equation (4) is the second stage of the instrumental variable approach. The first stage, equation (5), includes the same control variables as the second stage, as well as the time- and country fixed effects. Again, 𝐸𝑃𝑆 stands for all index variables and 𝑃𝑜𝑙𝑖𝑐𝑦 represents all four policy variables. Thus, the variable R&D is the instrument for the variable EPS R&D, Tax for EPS Tax, REC for EPS REC and FIT is the instrument for EPS FIT. Similarly to the fixed effects model, first the effects of the EPS indices are estimated individually, while afterwards their effects on renewable energy production are estimated together. The relation between the policy variables and the EPS indices will be presented separately in section 5 (table 2). Since both types of variables measure the presence and stringency of an environmental policy, although with different data, I expect a positive and highly significant relationship.
For the instrumental variables to be valid, they have to be correlated with the EPS indices, but should be uncorrelated with the measurement error (Stock & Watson, 2012). The validity of the instruments will be discussed in section 4. Because the instrumental variable approach is used to correct for measurement error, the estimates are expected to be more accurate when the IV regressions are performed.
The problem of reverse causality is addressed in a different way. Simultaneous determination of the environmental policy variables and the production of renewable energy
15 is a plausible threat to the validity of the results of the models described above. As was stated by Levison and Brunel (2013): “We want to know the effect of regulations on economic outcomes, but cannot easily separate that from the effect of those outcomes on environmental regulations” (p.3). Multiple papers (Carley, 2009; Song, 2011) acknowledged that environmental policy variables and renewable energy production are potentially jointly determined but did not empirically test this. Therefore, this paper not only recognizes that the causal effect could go both ways and therefore environmental policy could respond to the production of renewable energy, it also includes this in its empirical analysis.
To check for simultaneous causality, firstly a two-year lag is taken for every explanatory variable and then the same models as described above are applied. Second, a two-year lead is taken. This way, it is possible to get some insight as to whether environmental policies and renewable energy production are jointly determined. As it might take a couple of years for a policy to be fully implemented and have an effect, it is not surprising to find significant results with the lagged explanatory variables. Albrizio et al. (2014) included lagged values of the environmental policies for the same reason and found significant results. However, when including the explanatory variables with a two-year lead, significant results might be an indicator of simultaneous causality.
5. Results
First, the relation between the policy variables and the EPS indices is estimated. Columns 1-4 are the first-stage regressions of the instrumental variable model (equation 5). The results are shown in table 2. As was expected, each estimate is positive and significant at a 1% significance level (with the exception of the estimate of Tax). Therefore, these are strong instruments for the IV model. The instrument Tax is not significant and is thus invalid. This can be explained by the fact that the EPS Tax index only includes taxes on the emission of 𝐶𝑂2, 𝑁𝑂𝑥 and 𝑆𝑂𝑥, while the variable Tax is much broader15. Therefore, the coefficient of determination is also small. Hence, the estimate obtained with the instrument Tax in the second stage of the instrumental variable model (equation 4) is not reliable. When all policy
16 variables are regressed together on the total EPS index16, only the estimates of Log(R&D) and FIT stay significant.
Table 2: Relation between EPS indices and policy variables
(1) (2) (3) (4) (5)
VARIABLES EPS R&D EPS Tax EPS REC EPS FIT EPS Total
Log(R&D) 0.614*** 0.173*** (0.0636) (0.0350) Tax -0.143 -0.0157 (0.113) (0.0817) REC 1.361*** -0.0293 (0.117) (0.0858) FIT 2.731*** 0.432*** (0.201) (0.0741) Energy price residential 0.00271 0.00379** 0.00296* -0.0154*** -0.000209
(0.00231) (0.00191) (0.00175) (0.00399) (0.00129) Energy price industrial -0.00421 -0.00124 -0.0120*** -0.000747 -0.00327**
(0.00283) (0.00232) (0.00209) (0.00466) (0.00148) Constant 0.577* 1.394*** 0.389* 1.768*** 0.962*** (0.349) (0.369) (0.206) (0.466) (0.271)
Country fixed effects yes yes yes yes yes
Time fixed effects yes yes yes yes yes
Observations 269 233 273 273 206
R-squared 0.534 0.144 0.881 0.643 0.898
Number of countrynum 14 14 14 14 14
EPS: Environmental Policy Stringency Index, REC: Renewable energy certificates, FIT: feed-in-tariffs Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 3 shows the results of the fixed effects model using the policy variables (part 1), the EPS indices (part 2) and the instrumental variable regressions (part 3). To be able to compare all three methods, their results are displayed in the same table. The rest of the tables (control variables, the constant, number of observations, coefficients of determination and number of countries) are showed in appendix V. In part 3, the used instruments are also visible (the regressors in table 2).
Using the fixed effects model with the policy variables, only public R&D expenditures and the renewable energy certificates have a positive and significant effect on log(REP). When R&D is increased with 1%, the production of renewable energy will go up with 0.045%17. If REC are implemented, renewable energy production goes up with 28.5%18.
16
The EPS Total index is only used for regression (5) in table 2; it is the composite indicator of total environmental policy stringency and thus also takes other environmental policy measures into account. See appendix II for its composition.
17
Regression (1) is a log-log model, so a 1% change in R&D is associated with a 𝛽1% change in REP (Stock
17
REC by far has the largest effect, even when the effects of all policy measures together are estimated. Then, tax and feed-in-tariffs also have a positive effect on log(REP).
Table 3: The effect of policy measures and EPS indices on renewable energy production
(1) (2) (3) (4) (5)
VARIABLES log(REP) log(REP) log(REP) log(REP) log(REP)
1. log(R&D) 0.0454* 0.00281 (0.0266) (0.0277) Tax 0.0503 0.125* (0.0622) (0.0646) REC 0.285*** 0.290*** (0.0620) (0.0679) FIT 0.0805 0.110* (0.0501) (0.0586) 2. EPS R&D 0.0113 0.0121 (0.0238) (0.0226) EPS Tax -0.178*** -0.186*** (0.0334) (0.0329) EPS REC 0.109*** 0.0952*** (0.0276) (0.0274) EPS FIT -0.0240* -0.0189 3. EPS R&D 0.0739* 0.398 (0.0446) (1.217) Instrument R&D*** -0.351 -1.539 EPS Tax (0.445) (3.966) Instrument Tax 0.210*** -0.234 EPS REC (0.0479) (1.208) Instrument REC*** 0.0295 -0.115 EPS FIT (0.0191) (0.369) Instrument FIT***
Notes: 1. Fixed effects regressions with and country fixed effects. 2. Fixed effects regressions with time-and country fixed effects. 3. IV-regressions with time-time-and country fixed effects. For controls, constant, number of observations, R-squared, number of countries: see appendix V. REP:Total primary renewable energy supply, EPS: Environmental Policy Stringency Index, REC: Renewable energy certificates, FIT: feed-in-tariffs.
Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
However when looking at the EPS indices in the fixed effects model (part 2), the results are quite different. Now, EPS Tax and EPS FIT suddenly have a negative and significant effect on log(REP). These are unexpected results and not in line with theory, since it is presumed that when the policy stringency of both policy measures go up, the production of renewable energy would increases as well. EPS REC is again positive and significant. For
18
Regression (3) is a log-linear model, so a one-unit change in REC is associated with a (100×𝛽1)% change in
18 both fixed effects models, the coefficients of determination are considerably large, due to the inclusion of country- and time fixed effects19. The unobserved country-specific heterogeneity that is constant over time, as well as the heterogeneity that is constant over countries but not over time, explain a lot of variation within the model.
When using the instrumental variable approach (part 3 of table 3) the results change back to their expected values; the negative estimate of EPS Tax is not significant anymore and EPS FIT even turns positive. Therefore, I conclude that the results estimated by the fixed effects models suffer from errors-in-variables bias; the bias in an estimator of a regression coefficient that arises from measurement error in the regressors (Stock & Watson, 2012). The estimates resulting from the IV regressions are thus more accurate. Again, (the stringency of) R&D and REC have a positive and significant effect (although only when estimated individually). When EPS R&D goes up with one unit, the renewable energy production goes up with 7.4% and when EPS REC increases with one unit, it goes up with 21%. EPS Tax20 and EPS FIT do not have an influence on the production of renewable energy.
This is in line with existing literature (Schmidt, 2011; Song, 2011), which also found that renewable energy certificates are the most effective policy. Like Song (2011) described, it seems that renewable energy producers are more stimulated when quota restrictions are imposed, compared to when they receive subsidies. Moreover, previous studies also found that price-based policies, such as feed-in-tariffs and taxes, have no impact (Schmidt, 2011).
Surprisingly, when all explanatory variables are estimated together with the IV model, none of the estimates stay significant. This indicates that a combination of policies to stimulate the production of renewable energy is not effective. Countries are thus advised to focus on renewable energy certificates rather than implement multiple policies simultaneously.
For all three methods, industry prices have a negative effect on renewable energy, an effect which is small but highly significant21. Residential prices however, have a small positive effect on log(REP). The latter is unexpected. A possible explanation for this could be that the number of residential electricity consumers who produce their own renewable energy has increased dramatically from 2005 onwards, mostly thanks to the increased availability of solar panels which can be installed at home (Ren21, 2016). This can be a consequence of the
19 See appendix V. 20
However, this estimate is unreliable since its instrument (Tax) is invalid.
19 rising residential energy prices22, as people are then particularly stimulated to produce renewable energy themselves. This effect is significant, however very small.
Table 4: The lagged effect of policy measures and EPS indices on renewable energy production
(1) (2) (3) (4) (5)
VARIABLES log(REP) log(REP) log(REP) log(REP) log(REP)
1. log(R&D) 0.0399** 0.0607** (0.0175) (0.0279) Tax 0.0237 0.151** (0.0654) (0.0605) REC 0.334*** 0.347*** (0.0643) (0.0661) FIT 0.0259 0.0609 (0.0492) (0.0514) 2. EPS R&D 0.0334 0.0298 (0.0253) (0.0242) EPS Tax -0.147*** -0.171*** (0.0378) (0.0367) EPS REC 0.101*** 0.0754** (0.0297) (0.0299) EPS FIT -0.0401*** -0.0342*** (0.0127) (0.0128) 3. EPS R&D 0.147*** 0.817 (0.0510) (2.978) Instrument R&D*** -0.245 -3.378 EPS Tax (0.675) (11.89) Instrument Tax 0.238*** -0.525 EPS REC (0.0522) (2.900) Instrument REC*** 0.00723 -0.292 EPS FIT (0.0206) (1.099) Instrument FIT***
Notes: 1. Fixed effects regressions with time- and country fixed effects. 2. Fixed effects regressions with time- and country fixed effects. 3. IV regressions with time- and country fixed effects . All explanatory variables are lagged with 2 periods, as well as the instruments. For controls, constant, number of observations, R-squared, number of countries: see appendix VI.REP: Total primary renewable energy supply, EPS: Environmental Policy Stringency Index, REC: Renewable energy certificates, FIT: feed-in-tariffs. Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
Table 4 shows the same three models as in table 3, only with a two-year lag for every explanatory variable. The results are similar. For the fixed effects model with the policy variables (part 1), only 𝑙𝑜𝑔 (𝑅&𝐷)𝑡−2 and 𝑅𝐸𝐶𝑡−2 have a significant effect. When estimated together, 𝑇𝑎𝑥𝑡−2 and 𝐹𝐼𝑇𝑡−2 also have a positive and significant impact on log(REP). In part
20 2, 𝐸𝑃𝑆 𝑇𝑎𝑥𝑡−2 and 𝐸𝑃𝑆 𝐹𝐼𝑇𝑡−2 give unexpected negative results, which changes as the IV method is used. Then, as can be seen in part 3, only the policy stringency indices of R&D expenses and renewable energy certificates have a significant and positive impact. What is striking in comparison to table 3, is that these estimates are considerably larger. When 𝐸𝑃𝑆 𝑅&𝐷𝑡−2 goes up with one unit, the production of renewable energy goes up with 14.7 percent, twice the effect 𝐸𝑃𝑆 𝑅&𝐷𝑡 has. Furthermore, when 𝐸𝑃𝑆 𝑅𝐸𝐶𝑡−2 goes
up with one unit, renewable energy production goes up with 23.8 percent, also an increase compared to the effect of 𝐸𝑃𝑆 𝑅𝐸𝐶𝑡. This can be interpreted as evidence that renewable energy production responds with some lag to environmental policy. It thus takes a couple of years until a policy is in full effect and applied in practice, something which is backed by existing papers in the field (Popp, 2006; Albrizio et al., 2014). Again, all results become insignificant when estimated together. When all policies are in effect at the same time, even renewable energy certificates have no impact anymore. The estimated coefficients of PI and PR in table 4 are very similar to those in table 323.
Lastly, the same three models are estimated with the two-year lead variables of all explanatory variables and shown in table 5. Using the fixed effects model, the estimate of 𝑙𝑜𝑔 (𝑅&𝐷)𝑡+2 is not significant anymore. 𝑅𝐸𝐶𝑡+2 stays significant although its effect is smaller; now the production of renewable energy only increases with 19.4 percent. When the fixed effects model is applied to the EPS indices, 𝐸𝑃𝑆 𝑇𝑎𝑥𝑡+2 and 𝐸𝑃𝑆 𝐹𝐼𝑇𝑡+2 are also negative and significant, while 𝐸𝑃𝑆 𝑅𝐸𝐶𝑡+2 is positive and significant. In part 3, it is visible
that 𝐸𝑃𝑆 𝑅&𝐷𝑡+2 does not have an effect, contrary to 𝐸𝑃𝑆 𝑅𝐸𝐶𝑡+2. If 𝐸𝑃𝑆 𝑅𝐸𝐶𝑡+2 goes up with one index point, renewable energy production increases with 15 percent. Just like in table 3 and 4, the EPS indices have no effect when estimated together with an IV model.
All estimates in table 5 are thus considerably smaller than in the latter two tables. However, some are still significant. Albrizio et al. (2014) also found significant results when using lead variables of environmental policy measures and concluded that this was due to the fact that a change in policy is often known in advance. Firms can therefore anticipate to policies which have not been implemented yet. Thus, the production of renewable energy can respond to measures that will be in effect in the future. However, I was unable to find more evidence for this line of reasoning in existing literature. Furthermore, the significance and the size of the estimates are too large for it to be only accredited by the anticipation of renewable energy producers. Therefore, the results in table 5 suggest that endogeneity is present in the
21 data. This thus indicates that renewable energy production and environmental policies are simultaneously determined, which leads to bias and inconsistency in the estimates. The results presented in the tables above should therefore be interpreted with caution.
Table 5: The leaded effect of policy measures and EPS indices on renewable energy production
(1) (2) (3) (4) (5)
VARIABLES log(REP) log(REP) log(REP) log(REP) log(REP)
1. log(R&D) 0.00988 -0.00177 (0.0196) (0.0264) Tax 0.0887 0.142** (0.0578) (0.0589) REC 0.194*** 0.230*** (0.0579) (0.0651) FIT 0.0729 0.0912* (0.0452) (0.0508) 2. EPS R&D 0.0130 0.0105 (0.0215) (0.0206) EPS Tax -0.170*** -0.177*** (0.0332) (0.0332) EPS REC 0.0689*** 0.0411 (0.0258) (0.0256) EPS FIT -0.0258** -0.0278** (0.0117) (0.0115) 3. EPS R&D 0.0489 0.202 (0.0405) (0.305) Instrument R&D*** -0.566 -0.886 EPS Tax (0.503) (1.044) Instrument Tax 0.150*** -0.0434 EPS REC (0.0460) (0.272) Instrument REC*** 0.0293 -0.0476 EPS FIT (0.0191) (0.0986) Instrument FIT***
Notes: 1. Fixed effects regression with time- and country fixed effects. 2. Fixed effects regression with time- and country fixed effects. 3. IV regression with time- and country fixed effects. All explanatory variables are lead variables with 2 periods, as well as the instruments. For controls, constant, number of observations, R-squared, number of countries: see appendix VII. REP: Total primary renewable energy supply, EPS: Environmental Policy Stringency Index, REC: Renewable energy certificates, FIT: feed-in-tariffs. Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
6. Conclusion
This paper estimates the effect of four different environmental policies on the production of renewable energy, using panel data from 14 western-European countries from 1990 until
22 2012. The use of a new index that measures the stringency of environmental policies, composed by the OECD, provides a better insight into the effects of these policies.
The results show that only public R&D expenditures and renewable energy certificates have a positive and significant effect. When the policy stringency of R&D expenditures goes up with one unit, the production of renewable energy goes up with 7.4%. When the stringency of renewable energy certificates goes up with one index point, renewable energy production even increases with 21%. The price-based policies (environmentally related taxes and feed-in-tariffs) do not have a significant effect. This is supported by previous publications on this topic (Schmidt, 2011; Song, 2011). Furthermore, the policies have a larger effect on renewable energy production after 2 years, which can be interpreted as evidence that the policies need some time to be fully effective. However, when the effects of these policies are estimated together, they don’t influence the production of renewable energy anymore. This indicates that focusing on one policy is crucial.
Countries that want to increase the production of renewable energy are thus advised to focus on renewable energy certificates, as they are most effective. This can be explained by the fact that quantity restrictions, such as the renewable energy certificates, give freedom to the suppliers in choosing the type of renewable energy they want to use, as opposed to feed-in-tariffs and taxes (Song, 2011). They obligate energy suppliers to meet a certain percentage of their supply to come from renewables, but do not distinguish between different kinds of renewable energy sources. This makes compliance with the policy measure easier and therefore spurs the production of renewable energy. After all, the success of policies depends mostly on the political mandate as well as on the interest of the stakeholders involved in the process and the capacity of their participation and cooperation (do Valle Costa et al., 2008). The use of renewable energy certificates is not only the most efficient measure, it is also more cost-effective for governments than funding R&D projects or giving out subsidies. All in all, this paper advises to implement renewable energy certificates to increase the share of renewable energy in a country.
However, there is also evidence that renewable energy production and the environmental policies are simultaneously determined. Environmental policies thus possibly respond to the production of renewable energy. This indicates that the results are biased and thus severely weakens the conclusion described above. Therefore, together with the fact that existing papers in the field have not empirically tested the possibility of simultaneous causality yet, the most pressing issue for future research should be to further examine this. At the same time, it is important to be aware of other limitations of this research.Due
23 to limited data, only the years 1990 until 2012 are included. It is crucial for further research to extend this time span and include more recent years, especially because the production of renewable energy has increased exponentially in the last couple of years.
Also, the instrument Tax, when using the instrumental variable model, is invalid. Therefore, the estimates obtained with this instrument are unreliable. The quality of the empirical analysis would thus improve if a well suited, strong instrument is used in future research. This would contribute greatly to a better insight in the effect of environmentally related taxes on renewable energy production.
Thirdly, the policy measure renewable energy certificates, included in this research, is only one of the many types of quantity restrictions that are implemented by various governments across Europe. It is thus important for future research to also investigate the impact of other quantity restrictions and tradable permits, such as energy efficiency certificates and 𝐶𝑂2 trading schemes. Since renewable energy certificates are only
implemented by 5 of the 14 countries24 in this research, it would be particularly interesting to see the effect of comparable policy instruments that are implemented in more nations.
Fourthly, it is preferred to control for private R&D in addition to public R&D expenditures as this is also expected to have a positive effect on renewable energy. However, as firms are often very reluctant in disclosing their specific R&D expenditures, these data were not available unfortunately.
Lastly, since climate change became a more pressing issue on the political agenda, the public opinion on this topic changed greatly as well. More and more, people are becoming environmentally conscious, which can have an significant impact on the demand and thus production of renewable energy. However, ‘environmental awareness’ is a variable that is very difficult to quantify and is not available yet. Issuing a large survey on people’s opinion about the importance of climate change and using these data as a control variable would therefore be an important contribution to future research.
24
24
7. References
Albrizio, S., Koźluk, T., & Zipperer, V. (2014). Empirical Evidence on the Effects of Environmental Policy Stringency on Productivity Growth. OECD Economics Department Working Papers(1179).
Botta, E., & Koźluk, T. (2014). Measuring Environmental Policy Stringency in OECD Countries; a composite index approach. OECD Economics Department Working Papers(1177).
Brunnermeier, S., & Cohen, M. (2003). Determinants of environmental innovation in US manufacturing industries. Journal of Environmental Eonomics and Management, 2(45), pp. 278-293.
Carley, S. (2009). State renewable energy electricity policies: An empirical evaluation of effectiveness. Energy Policy(37), 3071–3081.
Carrión-Flores, C. E., & Innes, R. (2010). Environmental innovation and environmental performance. Journal of Environmental Economics and Management(59), 27–42. Do Valle Costa, C., Rovere, E. L., & Assmann, D. (2008). Technological innovation policies
to promote renewable energies: Lessons from the European experience for the Brazilian case. Renewable and Sustainable Energy Reviews, 65-90.
Energieonderzoek Centrum Nederland (2016). Nationale Energieverkenning.
Energy Technology Research and Development Database. (n.d.). Retrieved April 24, 2017, from International Energy Agency:
http://wds.iea.org/WDS/Common/Login/login.aspx
European Commission. (2010). Energy 2020; A strategy for competitive, sustainable and secure energy. Report from the Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions. European Commission. (2016, June). Climate Action: The EU Emissions Trading System (EU
ETS). Retrieved May 2017, from https://ec.europa.eu/clima/policies/ets_en European Commission. (2017). Renewable Energy Progress Report: Report from the
Commission to the European Parliament, the Council, the European Economic and Social Committee and the Committee of the Regions. Brussels.
European Environment Agency. (2016). Renewable energy in Europe 2016; Recent growth and knock-on effects. Luxembourg: Publications Office of the European Union. Eurostat. (2016, June). Europe 2020 indicators - climate change and energy. Retrieved May
2017, from Eurostat: statistics explained: http://ec.europa.eu/eurostat/statistics-explained/index.php/Europe_2020_indicators_-_climate_change_and_energy
25 Eurostat. (2016, July). Renewable energy statistics. Retrieved May 2017, from Eurostat:
statistics explained: http://ec.europa.eu/eurostat/statistics-explained/index.php/Renewable_energy_statistics
Eurostat. (2017). Greenhouse gas emission statistics. Retrieved June 2017, from Eurostat: statistics explained:
http://ec.europa.eu/eurostat/statistics-explained/index.php/Greenhouse_gas_emission_statistics
Eurostat. (June, 2016). Europe 2020 indicators - climate change and energy. Retrieved May 2017, from Eurostat: statistics explained: http://ec.europa.eu/eurostat/statistics-explained/index.php/Europe_2020_indicators_-_climate_change_and_energy
Hascic, I., De Vries, F., Johnstone, N., & Medhi, N. (2009). Effects of Environmental Policy on the Type of Innovation: The Case of Automotive Emission-control Technologies. OECD Journal: Economic Studies .
International Energy Agency. (2006). Energy Policies of IEA Countries.
International Energy Agency. (2015). Renewables: Annual Questionnaire and Historical Revisions. Eurostat, United Nations Statistics Division, Industrial and Energy Statistics Section.
International Energy Agency. (2016). Energy Technology Research and Development Database. Retrieved April 2017, from http://data.iea.org/
International Energy Agency. (2017). Energy Prices and Taxes Database.
International Energy Agency. (2017). Renewables information; database documentation 2017 preliminary edition.
Jaffe, A. B., & Palmer, K. (1996). Environmental regulation and innovation; a panel data study. NBER Working Paper Studies; working paper 5545.
Johnstone, N., Haščič, I., Poirier, J., Hemar, M., & Michel, C. (2012). Environmental policy stringency and technological innovation: evidence from survey data and patent counts. Applied Economics, 17(44), 2157-2170.
Johnstone, N., Popp, D., & Hašcic, I. (2009). Renewable Energy Policies and Technological Innovation: Evidence Based on Patent Counts. Environmental and Resource
Economics(45), 33–155.
Klessmann, C., Held, A., Rathmann, M., & Ragwitz, M. (2011). Status and perspectives of renewable energy policy and deployment in the European Union—What is needed to reach the 2020 targets? Energy Policy(39), 7637–7657.
Koźluk, T., & Garsous, G. (2016). How stringent are environmental policies? A review of OECD work on indicators of environmental policy stringency (EPS). OECD.
26 Koźluk, T., & Zipperer, V. (2014). Environmental policies and productivity growth – a
critical review of empirical findings. OECD Journal: Economic Studies, 155-185. Lanjouw, J. O., & Mody, A. (1996). Innovation and the international diffusion of
environmentally responsive technology. Research Policy(25), 549-571.
Levinson, A., & Brunel, C. (2013). Measuring Environmental Regulatory Stringency. OECD. Menz, F. C., & Vachon, S. (2006). The effectiveness of different policy regimes for
promoting wind power: Experiences from the states. Energy Policy(34), 1786–1796. OECD. (2013). Renewable Energy Policy Dataset, version March 2013. Compiled by the
Empirical Policy Analysis Unit of the OECD Environment Directorate (Johnstone, N., Haščič, I., Cárdenas Rodríguez M., Duclert, T.) in collaboration with an ad hoc research consortium (Arnaud de la Tour, Gireesh Shrimali, Morgan Hervé-Mignucci, Thilo Grau, Emerson Reiter, Wenjuan Dong, Inês Azevedo, Nathaniel Horner, Joëlle Noailly, Roger Smeets, Kiran Sahdev, Sven Witthöft, Yunyeong Yang, Timon Dubbeling).
OECD. (2016). OECD Policy Measures for the Environment.
OECD. Policy Instruments for the Environment Database. (n.d.). Retrieved April 24, 2017, from OECD: http://www2.oecd.org/ecoinst/queries/
OECD Statistics. (n.d.). Retrieved April 11, 2017, from Green Growth Indicators: http://stats.oecd.org/
Popp, D. (2006). International innovation and diffusion of air pollution control technologies: the effects of NOX and SO2 regulation in the US, Japan, and Germany. Journal of Environmental Economics and Management(51), 46-71.
Renewable Energy Network (2016). Renewables 2016; Global Status Report. Paris. Schmid, G. (2012). The development of renewable energy power in India: Which policies
have been effective? Energy Policy(45), 317–326.
Song, Y. W. (2011). The Effect of Renewable Energy Policies on Renewable Energy Production. International Atlantic Economic Society(39), 195–196.
Stock, J. H., & Watson, M. W. (2012). Introduction to Econometrics. Essex: Pearson. De Vries, F., & Withagen, C. (2005). Innovation and environmental stringency; The case of
27
Appendix I
Reducing skewedness by taking logarithms
28
Appendix II
Structure of the OECD Environmental Policy Stringency Index
The indices marked in red are used in this research. Source: Botta and Koźluk (2014)
Appendix III
Pairwise correlation matrix
R&D Tax REC FIT PI PR
R&D 1.0000 Tax -0.1756 1.0000 REC 0.2968 -0.1177 1.0000 FIT 0.1876 -0.1342 -0.2249 1.0000 PI 0.2050 0.0213 0.4684 -0.1159 1.0000 PR 0.2740 0.3548 0.3363 -0.1405 0.7824 1.0000
EPS R&D EPS Tax EPS REC EPS FIT PI PR EPS R&D 1.0000 EPS Tax 0.3112 1.0000 EPS REC 0.1509 0.1994 1.0000 EPS FIT 0.0153 0.0578 0.1114 1.0000 PI -0.2119 0.0460 0.3431 0.0325 1.0000 PR -0.0108 0.1704 0.4134 0.1175 0.7972 1.0000
29
Appendix IV
Overview variables
Variable Abbreviation Measurement Source
Dependent variable
Total primary renewable energy supply
REP Tonne of oil
equivalent (toe) in millions OECD (2017), Green Growth Indicator Database Policy measures Public R&D expenditures on total renewable energy
R&D In millions, euro (2015 prices and exchange rates) IEA’s Energy Technology Research and Development Database Environmentally
related tax revenue
Tax % of GDP OECD’s Policy
Instruments for the Environment (PINE) database Renewable energy
certificates
REC Binary OECD Renewable
Energy Policy Dataset, version March 2013
Feed-in-tariffs FIT Binary OECD Renewable
Energy Policy Dataset, version March 2013 Environmental Policy Stringency Indices Public R&D expenditures on total renewable energy
EPS R&D Index OECD (2014)
Tax revenue (on CO2, NOx, and SOx)
EPS Tax Index OECD (2014)
Renewable energy certificates
EPS REC Index OECD (2014)
Feed-in-tariffs EPS FIT Index OECD (2014)
Controls Energy price industrial
PI In euro per MWh
(2015 exchange rate)
IEA’s 2015 Energy Prices and Taxes Database Energy price residential PR In euro per MWh (2015 exchange rate) IEA’s 2015 Energy Prices and Taxes Database
30
Appendix V
rest of Table 3
The effect of policy measures and EPS indices on renewable energy production
1. PI -0.00505*** -0.00284** -0.00356*** -0.00371*** -0.00419*** (0.00118) (0.00128) (0.00111) (0.00116) (0.00117) PR 0.00593*** 0.00394*** 0.00535*** 0.00548*** 0.00493*** (0.000965) (0.00105) (0.000933) (0.000995) (0.00102) Constant 0.661*** 0.625*** 0.616*** 0.599*** 0.501** (0.146) (0.204) (0.110) (0.116) (0.214)
Time f.e. yes yes yes yes yes
Country f.e. yes yes yes yes yes
Observations 269 233 273 273 206 R-squared 0.716 0.716 0.736 0.716 0.736 Number of countries 14 14 14 14 14 2. PI -0.00352*** -0.00392*** -0.00234** -0.00342*** -0.00277** (0.00118) (0.00112) (0.00118) (0.00117) (0.00112) PR 0.00508*** 0.00611*** 0.00510*** 0.00439*** 0.00569*** (0.000965) (0.000934) (0.000935) (0.00102) (0.000962) Constant 0.620*** 0.733*** 0.572*** 0.724*** 0.699*** (0.130) (0.111) (0.114) (0.121) (0.129)
Time f.e. yes yes yes yes yes
Country f.e. yes yes yes yes yes
Observations 287 287 287 287 287 R-squared 0.720 0.748 0.736 0.724 0.768 Number of countries 14 14 14 14 14 3. PI -0.00474*** -0.00327** -0.00104 -0.00369*** -0.00197 (0.00123) (0.00140) (0.00130) (0.00121) (0.00415) PR 0.00573*** 0.00527*** 0.00473*** 0.00594*** 0.00640 (0.000975) (0.00189) (0.000982) (0.00114) (0.00427) Constant 0.618*** 1.114** 0.534*** 0.547*** 1.828 (0.168) (0.479) (0.117) (0.132) (2.602)
Time f.e. yes yes yes yes yes
Country f.e. yes yes yes yes yes
Observations 269 233 273 273 206
Number of
countries 14 14 14 14 14
Notes: f.e.: fixed effects. PI: Energy price industrial, PR: Energy price residential. Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
31
Appendix VI
rest of Table 4
The lagged effect of policy measures and EPS indices on renewable energy production
1. PI -0.00491*** -0.00280** -0.00364*** -0.00363*** -0.00383*** (0.00113) (0.00140) (0.00112) (0.00119) (0.00117) PR 0.00598*** 0.00368*** 0.00547*** 0.00516*** 0.00502*** (0.000926) (0.00114) (0.000919) (0.000983) (0.00103) Constant 0.604*** 0.773*** 0.610*** 0.639*** 0.326 (0.135) (0.221) (0.110) (0.117) (0.219)
Time f.e. yes yes yes yes yes
country f.e. yes yes yes yes yes
Observations 267 207 287 287 194 R-squared 0.726 0.713 0.747 0.720 0.764 Number of countries 14 14 14 14 14 2. PI -0.00348*** -0.00357*** -0.00242** -0.00274** -0.00253** (0.00121) (0.00117) (0.00121) (0.00120) (0.00117) PR 0.00479*** 0.00535*** 0.00452*** 0.00339*** 0.00468*** (0.00101) (0.000986) (0.000972) (0.00104) (0.00103) Constant 0.580*** 0.750*** 0.623*** 0.784*** 0.731*** (0.143) (0.121) (0.121) (0.126) (0.142)
Time f.e. yes yes yes yes yes
country f.e. yes yes yes yes yes
Observations 259 259 259 259 259 R-squared 0.724 0.740 0.736 0.734 0.765 Number of countries 14 14 14 14 14 3. PI -0.00516*** -0.00326* -0.00117 -0.00345*** -0.00211 (0.00124) (0.00175) (0.00132) (0.00126) (0.00928) PR 0.00616*** 0.00451* 0.00449*** 0.00476*** 0.00955 (0.00107) (0.00241) (0.00102) (0.00117) (0.0186) Constant 0.407** 1.099 0.551*** 0.655*** 3.156 (0.199) (0.740) (0.129) (0.137) (8.406)
Time f.e. yes yes yes yes yes
country f.e. yes yes yes yes yes
Observations 241 207 259 259 194
Number of
countries 14 14 14 14 14
Notes: f.e.: fixed effects. PI: Energy price industrial, PR: Energy price residential. Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1
