The Impact of Renewable Energy on the EU Electricity Prices and CO2 Emissions


Bachelor Thesis, 2015
77 Pages, Grade: 2.0

Free online reading

Table Of Contents

1 Introduction

2 Sources of Energy
2.1 Classification
2.1.1 Fossil Energy
2.1.2 Nuclear Energy
2.1.3 Renewable Sources of Energy
2.2 Renewables in the EU/World Electricity Production
2.3 CO2 Emissions by Source

3 Electricity Pricing in the EU
3.1 Electricity Supply and Demand
3.2 Electricity Price Components
3.2.1 EU Electricity Prices by Component
3.3 Conclusions: Future Price and Cost Trends

4 Renewable Energy in the EU
4.1 Renewable Energy Targets
4.2 Greenhouse Gas Emission Targets

5 Impacts of Renewable Energy Promotion: Panel Data Analysis
5.1 Literature Review
5.2 Data and Methodology
5.2.1 Data Set Summary
5.2.2 Data Sources
5.2.3 Variables
5.3 Theoretical Framework
5.3.1 First Differences Estimation
5.3.2 Fixed Effects Estimation
5.4 Practical Applications of the Theory
5.4.1 Electricity Price and Renewable Energy
5.4.2 CO2 Emissions and Renewable Energy
5.5 Justification of the Model Results
5.5.1 Electricity Price and Renewable Energy
5.5.2 CO2 Emissions and Renewable Energy

6 Conclusion

Appendix A: Data Set of the Model

Appendix B: Theoretical Framework of the Panel Data Model
B.1 First Differences Estimation
B.2 Fixed Effects Estimation
B.3 Fixed Effects versus First Differences

Appendix C: Practical Applications of the Theoretical Model
C.1 Electricity Price and Renewable Energy
C.2 CO2 Emissions and Renewable Energy

Acronyms

Bibliography

Abstract

This work is focused on the topic of electricity pricing in the EU connected with the increasing use of renewable energy sources in electricity production and consumption. It provides background information related to the types of energy sources along with a summary of their benefits and drawbacks regarding both the environmental impact and financial costs. Furthermore, it involves both world and European electricity production statistics followed by an overview of the EU approach concerning the support of environment-friendly energy production methods. The core of the thesis is then an econometric panel data analysis based on the data collected from 13 member states of the EU over the period between 2010 and 2013 analysing two relationships. First, the impact of the share of renewable energy sources in the final electricity production on the European consumer electricity prices. Second, whether the replacement of fossil fuels by renewable energy sources causes a significant decrease in the CO2 emissions. In conclusion, this paper provides suggestions for further research based on the undertaken analyses. The results of the panel data analysis presented in this paper suggest that household electricity prices in the studied EU member states increase with the deployment of electricity production from renewable sources, as expected. On the contrary, a negative effect of the renewables used in the EU energy consumption on the CO2 emissions produced was found by the model regression.

JEL Codes H20, Q20, Q40, Q47, Q48, Q54

Keywords electricity price, renewable energy, CO2 emissions, energy policy

1 Introduction

As energy has recently become the crucial fuel for social and economic development and renewable energy-related activities have had significant impacts on the global economy, the topic of renewable energy production and its effects on consumers and the environment is undoubtedly worth studying. Moreover, since investments in deployment of renewable energy sources have been the highest in Europe compared to the rest of the world over the last decade, it is understandable to choose the EU member states as the appropriate regions for the purpose of this study. By promoting and using more renewables to meet its energy demand, the EU not only lowers its dependence on imported fossil fuels connected with uncertainty and political concerns, it also aims to make its energy production more sustainable and environment-friendly.

Renewable energy replaces conventional fuels (mostly coal, oil and natural gas) in four distinct areas (namely electricity generation, heating, motor fuels and rural energy services). Within the scope of this thesis, we focus on the electricity sector as it plays a decisive role in reaching the EU renewable energy targets. The main objective of this thesis is to analyse the effects caused by the shift of the EU energy consumption and production to the alternative sources of energy. By using the literature reviewed in this study, our observed data and the econometric panel data analysis, we aim to find the relationship between the increasing share of renewables in the electricity production and the changes in electricity prices in the examined EU member states. In addition, we study the impact of renewable energy participation in the energy consumption on the total amount of carbon dioxide produced by the EU countries since the reduction of greenhouse gas emissions is one of the key goals of the EU energy and climate policy.

The thesis is structured as follows. First, a brief overview regarding the types of energy sources and their characterisation along with a summary of both the global and EU electricity production by source is given in Chapter 2 as the essential background for the topic of electricity from renewable energy sources. Next, Chapter 3 includes a description of the EU electricity pricing along with the components comprising the electricity prices for final consumers. In Chapter 4, there are fundamental information concerning the EU renewable energy and climate policy containing the targets to achieve a sustainable energy sector in the long run with considerably lower greenhouse gas emissions produced by the energy production. To reach its energy sector goals, the EU makes use of specific support schemes for the promotion of renewable energy production which are described in this chapter as well. The last chapter covers the econometric model analysing the impacts of the use of renewable energy on the consumer electricity prices and the level of carbon dioxide emissions produced, and is followed by the conclusion.

2 Sources of Energy

2.1 Classification

There is a controversial debate about the effects of the electricity sector reforms (concerning the promotion of renewable energy) on electricity prices (Moreno 2011). The deployment of renewable energy technologies provides several positive effects, mainly with reference to an expected increase in energy self-sufficiency and cleaner environment, but it also leads to some additional costs related to the adjustments in production, prices and transportation systems. Hence, we aim to provide an overview regarding the costs and benefits connected to each energy source.

According to the World Energy Resources (WER) Survey 2013, the value of the global primary energy supply is forecasted to rise to 17,208 Mtoe by 2020, an increase by more than 22% compared to the 2010 level (see Table 2.1). Renewable energy has become a widely discussed topic since its share in the world primary energy supply is expected to increase as well, from 13% to approximately 18% over the 10-year period. It implies that the amount of renewable energy generated on a global basis is estimated to rise by almost 69%, from 1832 Mtoe in 2010 to 3097 Mtoe in 2020. On the contrary, the level of energy generated by using natural fossil sources is predicted to decrease by at least 6 percentage points over the time period.

The following three parts of this section are focused on a basic characterization of fossil, nuclear and renewable energy sources to offer a reader the fundamental background for the further analysis provided in the other chapters.

Table 2.1: Total Primary Energy Supply by Resource (1993, 2010, 2020)

illustration not visible in this excerpt

Source: World Energy Resources 2013 and WEC World Energy Scenarios to 2050

2.1.1 Fossil Energy

Fossil energy is generated by using the remains of decomposition of plants and animals in the nature. The main three types of fossil fuels for energy generation consist of coal, petroleum, and natural gas. These fuels are burnt in fossil-fuel power stations and the heat produced during the burning process is used either directly for heating or converted to mechanical energy or electrical power. As shown in Table 2.1, fossil fuels account for more than 80% of the world primary energy supply but the number is expected to decrease (WER 2013).

The fact that the technology and infrastructure needed for the extraction of fossil fuels already exists and has improved over the last tens of decades, makes such sources less costly than the renewable ones associated with relatively newly developed modern technologies. On the contrary, the intensive extraction and consumption of fossil fuels results in an environmental degradation and high amount of greenhouse gases emission contributing to concerns about the global warming. Regarding the data in Table 2.1, the world energy supply figures indicate the global efforts to decrease the level of countries’ dependence on fossil fuels. The reasons for this decrease are the facts that the fossil-based resources are non-renewable; their production is un-sustainable; and they create high level of environmental pollution and energy security risks for dependent countries (WER 2013).

2.1.2 Nuclear Energy

The main source of fuel for nuclear reactors is uranium. According to WER (2013), “the present survey shows that the total identified uranium reserves are abundant based on the current energy requirements”. A growing trend has been seen in the total nuclear electricity generation during the last two decades albeit the proportion of nuclear-based electricity supply in the total global electrical power production decreased.

Public arguments against the use of uranium in energy generation process are comprised mostly of concerns about the reactors’ operation and final waste disposal, since the radioactive waste as a by-product of nuclear power production is dangerous to most forms of life and hence the whole environment. Moreover, the safety, emergency, containment and storage systems connected to handling of radioactive waste bring about high costs.

By contrast, the defenders of nuclear energy base their arguments on the facts that this type of energy production is environment-friendly regarding the CO2 and other greenhouse gases emissions; the nuclear energy transformation into electricity is almost ten times more efficient than in the case of coal or oil; and the cost of the generated electricity is moderate and relatively predictable over the nuclear reactors’ service life (WER 2013).

2.1.3 Renewable Sources of Energy

Renewable energy (RE) can be produced from a wide variety of sources including mainly sun, wind, water, and biomass. The advantage of using such resources for energy generation is the fact that they are abundant and are not found only in a limited number of territories (compared to fossil fuels and uranium). Moreover, the modern deployment of renewable energy is assumed to lead to a significant energy security, climate change moderation, environmental pollution reduction and economic benefits.

Biomass and biofuels are energy sources derived from living or recently living organisms (referring mainly to plants and plant-based materials). Generally, biomass is used either to produce heat by combustion or to conversion to some type of gaseous or liquid biofuels. Probably the most important attribute of the modern biofuels is the fact that they can be used in diesel engines and are considered to be an alternative to fossil-based fuels used in transport. The other advantages of this energy source are its worldwide abundance and relatively simple combustion technologies connected with the energy production. Although using biofuels causes less CO2 emissions than fossil fuels, it produces some air pollutants such as nitrogen oxide or sulphur dioxide and emits some gas or liquid waste (WER 2013).

Hydro power is power obtained by using the energy of flowing and falling water which is harnessed for further purposes. Currently, the main use of water power is the modern development of hydroelectric power stations. The fundamental advantages of using water power in the energy production contain zero waste and CO2 emissions generated during the process, low operation costs, reliability in conjunction with generating large amounts of power and capability to meet a specific energy demand by possible regulation of the output. By contrast, the opponents of hydro power argue that the construction of hydroelectric dams is very expensive and has negative environmental impacts on the dam areas being absolutely adapted to functioning of the dam. Moreover, the energy generation using water power can be affected by drought or other climate and weather changes (WER 2013).

Wind energy can be also used to generate mechanical power or electricity having a relatively high energy output. The supporters of wind energy see the advantages from both the environmental point of view (the reduction of greenhouse gases emissions, and little disruption of ecosystem caused by wind turbines installation) and the economic efficiency (no fuel or waste costs during the turbine life cycle, simple technology, and relatively quick installation) (WER 2013). Nevertheless, the wind turbine installation is not feasible for all geographic locations and territories and even at a suitable place, the output is proportional to unpredictable wind speed. In addition, the modern wind power generation requires a high initial investment and subsequent ongoing maintenance costs, usually resulting in reliance on government subsidies.

Solar energy can be harnessed using a variety of technologies. The most frequently used such technology is called solar photovoltaics (solar PV) denoting a non-polluting method of electrical power generation by converting solar radiation into direct current electricity using PV solar panels. The use of electricity generated from solar energy has increased over the last few years (according to Eurostat data). This trend can be caused by the facts that the installation and dismantling of solar panels is relatively uncomplicated and quick, the energy generation process is reliable and the installed solar panel systems last from 15 to 30 years without almost any maintenance costs. However, the need for high initial investment, limited availability of materials for solar PV panels, and the dependence on sunny weather cause the unsuitability of this electricity production method in some areas (WER 2013).

2.2 Renewables in the EU/World Electricity Production

Since the core of this thesis is to analyse the relationship between an increase in the use of RE sources and the electricity prices in the EU, we provide a short summary regarding the current importance of RE in the electricity production sector. Figure 2.1 concerns the total electricity production by source in 2013, both globally and in the EU. While comparing the two graphs, it is noticeable that unlike the world average figure (67%), the EU share of fossil fuels in the total electricity production was lower than 50%—albeit oil, coal and natural gas have been the mostly used sources of energy throughout the world, accounting for more than 80% of the global energy production (WER Survey 2013). The share (25%) of RE in the EU electrical power generation was above the global average proportion (22%) in the same year.

The most significant RE resource used for the production of electricity was hydro power (both globally and in the EU). On the global scale, the energy for almost 73% of the electricity from renewable energy sources (RES-E) was drawn from hydro power stations. In the EU, the hydroelectricity participation in the RES-E production was around 40%. Concerning the other increasingly used RE sources, wind accounted for almost 14% of the global electrical power generation from renewables, and biomass served as a source for about 8% of the production. The figures for the EU were 28% and 20%, respectively. In addition, solar power comprised 3.6% share of the RES-E generated worldwide. In the EU, the proportion was more than twice higher, approximately 8%. According to the above mentioned figures, about 76% of the RES-E production in the EU came from hydro, wind and solar power—energy sources connected with considerably high initial costs of electricity generation and strongly supported by the EU energy, climate and environmental programmes regarding the following decade (see Chapter 4). In Chapter 5, we will use an econometric model to analyse the effect of using these energy sources on the electricity prices.

illustration not visible in this excerpt

Figure 2.1: Total World and EU Electricity Production by Source in 2013

Source: REN21 Report 2014, WER Survey 2013

2.3 CO2 Emissions by Source

Since distinct electricity generation methods (drawing energy from different sources) produce carbon dioxide and other greenhouse gases in varying quantities through construction, operation (including fuel supply activities) and decommissioning, the lifecycle approach accounts for the emissions from all phases of each electricity production project attempting to calculate the global warming potential of electrical energy sources. Observing the lifecycle emissions from electricity generation allows us to fairly compare the different generation methods on a per kilowatt-hour basis (see Figure 2.2).

illustration not visible in this excerpt

Figure 2.2: Lifecycle CO2 Emissions by Source (in t/GWh)

Source: WNA Report 2011

The above mentioned lifecycle phases of electricity production including construction, operation and decommissioning of the generators are associated with CO2 emissions produced through:

- manufacturing of the electricity generators (e.g. solar panels and wind turbines in case of renewable sources) along with acquiring the necessary components needed for their construction (which are usually imported from developing countries and hence leave a noticeable carbon footprint),
- transportation of the generators to the convenient location (usually by trucks powered by fossil fuels) and later to the recycling place,
- and the recycling process of the generators after they can be no longer used (including the emissions produced during the decommissioning process as well as subsequent waste disposal).

The data in Figure 2.2 were obtained from the World Nuclear Association (WNA) Report 2011 reviewing over twenty studies assessing the greenhouse gas emissions produced by different forms of electricity generation. It is noticeable that all renewable sources included in the statistic (wind, solar PV, biomass and hydro power) perform substantially better than each of the fossil-based fuels with respect to the level of CO2 emitted. Hence, according to the WNA Report and the lifecycle approach, it is rational to expect that the replacement of fossil fuels in the EU electricity generation by RES-E production results in a decrease in the amount of carbon dioxide produced. This hypothesis is tested empirically in Section 5 using the panel data analysis.

3 Electricity Pricing in the EU

3.1 Electricity Supply and Demand

This chapter provides an insight into how electricity prices and costs are evolving and which factors are driving their changes. Since the energy markets were deregulated in 1998, market prices of electricity have been the result of supply and demand. Due to the fact that electrical power cannot be stored, it is produced at the moment of demand. Hence all the factors influencing the demand immediately affect the price on the spot market (commodities or securities market in which goods are sold for cash and delivered immediately). The same applies for the electricity supply. The summary of these factors is given by Figure 3.1.

Figure 3.1: Factors Affecting the Electricity Supply and Demand

illustration not visible in this excerpt

Source: RWE AG

According to RWE website, “on the supply side, the electricity price is heavily influenced by fuel prices (for fossil fuels) and the prices for CO2 allowances”. To determine how much electricity is generated by renewable power stations, the weather and climate, capacities of power plants, and their current technical conditions are crucial too.

On the demand side, electricity price is also influenced by the weather (temperature and cloud cover influencing consumer behaviour directly) along with the state of the general economy but also e.g. holidays (public, school or bank) and fluctuations in the global economy.

3.2 Electricity Price Components

To understand how the price of electricity is finally determined, we have to consider all the elements affecting it, influenced by both market forces and government policies. In Figure 3.2, you can see a summary of such elements.

Figure 3.2: Elements of Consumer Prices

illustration not visible in this excerpt

Source: European Commission (2014)

European Commission (Energy Prices and Costs in Europe, 2014) defines the composition of consumer prices of electricity as follows. The energy component consists of two parts. The wholesale element of the price is connected with the costs of delivering energy to the grid, including fuel purchase (or production), shipping and processing as well as the costs of construction, operation and decommissioning of power stations. The retail element then reflects the costs arising during the sale of energy to final consumers on the retail markets. Next, the network element is composed of transmission and distribution infrastructure costs connected with the maintenance and expansion of grids, system services and network losses. The last element of the electricity prices in the EU are taxes and levies which have a form of either general taxation (VAT, excise duties) or specific levies promoting and supporting targeted energy, environment and climate policies.

3.2.1 EU Electricity Prices by Component

Over the last five years, the European consumer prices of electricity have increased noticeably. Albeit the differences between distinct national prices have been large, almost all EU member states have seen a consistent rise in their electricity prices (see the data set in Appendix A). The energy policies and accompanying environmental targets, both on the national and the European level, have been considered to play an increasing role in determining the final electricity price for consumers. Hence, in order to better understand the relationship between energy policies and electricity prices, it is useful to disaggregate the price into its elements (as in Figure 3.2) and compare them.

The relative share of the energy component in the retail price of electricity has diminished over the last five years. If we compare the data from 2010 and 2013 (the 2014 data has not been available yet), we can notice that the energy component has seen the smallest increase while the tax/levy component has increased the most over the time. Since 2010, according to Eurostat, household electricity network costs went up by 14.2%, taxes and levies rose by more than 20.7% and energy supply costs by approximately 6.5% (see Figure 3.3) for the EU weighted average electricity price.

illustration not visible in this excerpt

Figure 3.3: Electricity Prices by Component in 2010 and 2013 in the EU

Source: Eurostat

Albeit the relative share of energy cost element in the European electricity prices is diminishing, it still composes the largest part of the price. According to European Commission (Energy Prices and Costs in Europe, 2014), the EU household electricity retail prices have risen each year from 2010 to 2013 as opposed to the average wholesale electricity prices which decreased over the period. This fact can be linked to the EU energy policies, mainly with the growth of power generation capacity with low operating costs, such as wind and solar power. However, due to a weak price competition in a number of retail markets (allowing suppliers to avoid passing on the wholesale price reduction to retail prices), the fall in wholesale prices has not resulted in a reduction in the retail prices (EU Commission, 2014).

Regarding the last element of electricity prices, European Commission claims that “in most member states, taxes and levies have financed energy, environment and climate policy measures, including promotion of energy efficiency and renewable energy production.” As can be seen in Figure 3.3, in 2013, the mentioned taxes and levies were the second largest component of the EU average prices of electrical power for households.

The change in percentage proportion of the electricity prices formed by taxes and levies over the last few years is showed in table 3.1. In nine of the thirteen EU member states, for which the data have been collected, the share has risen over the last 3 years. The most noticeable increase was seen in Germany, where taxes and levies stood for 52% of the electricity price in 2014 while in 2012, this figure was only around 16%. On the contrary, Belgium, the Netherlands, Poland and Portugal saw a decrease in the participation of taxes and levies. The EU average figure has increased from 29% in 2012 to more than 32% in 2014.

Table 3.1: Percentage Shares of Taxes/Levies in Electricity Prices by Country

illustration not visible in this excerpt

Source: Author’s computations, Eurostat

In addition, according to European Commission (2014), renewable energy costs added to retail prices composed 6% of the average EU household electricity prices in 2012. The report also asserts that there is a wide range of the costs in form of renewable energy taxes and levies, with Spanish and German shares reaching 15.5% and 16% of household electricity prices respectively, in contrast to Poland and Sweden with less than 1% shares. However, the share is increasing in the majority of the EU member states due to the EU policies supporting the use of renewable sources of energy. The net effect of renewable energy on retail electricity prices has not been the same throughout the EU. While in Spain the effect has appeared to be reducing the prices, in Germany it has been the opposite case. In chapter 5, we will study further the effect of renewable energy on the electricity prices by applying an econometric panel data model.

The last important element of the breakdown of electricity prices consists of the already mentioned network costs. Albeit the relative shares of transmission and distribution costs (as well as the absolute levels) vary greatly across the EU, in all member states the distribution costs exceeded the transmission costs each year over the last five years. Since 2010, the electricity network costs went up by 14.2% for households what has been expected in the context of energy sector transformation but it could be mitigated through better network governance on the national level. The absolute values of electricity network costs, ranging from 2.2 cents/kWh to 9.7 cents/kWh (Eurostat) between EU member states, imply that such costs can have a significant impact on the total electricity prices (European Commission, 2014).

3.3 Conclusions: Future Price and Cost Trends

Over the last five years, the rise in electricity prices has been driven mainly by increases in taxes/ levies and network costs. Hence, the goal of the EU is to ensure that the policies financed by taxes and levies (energy, environment and climate policies) are applied as cost effectively as possible. Member states therefore review their different national practices including the European Commission’s guidance regarding energy sector government interventions (mainly renewable energy and energy efficiency policies) to minimise negative consequences for energy prices (European Commission, 2014). In addition, according to European Commission, “the EU needs to benchmark network costs and practices to ensure that European convergence in network practices improves the efficiency of the distribution and retail markets and so reduces the network cost element of the prices.”

According to the European Commission’s 2030 energy and climate policy framework, the energy costs are expected to be driven by rising fossil fuel prices as well as by the high initial investments needed for the construction of renewable energy power stations (mostly wind, hydropower and solar PV power stations) and the infrastructure connected with it. Specifically for electricity, European Commission claims that the costs are estimated to increase up to 2020 when they are expected to stabilise and subsequently slightly decrease as fossil fuels are going to be already replaced by renewable energy sources with low operation and maintenance costs of the power plants. In Chapter 5, we will analyse whether the increase in the share of renewable energy sources in the electricity production has significantly caused an increase in the European electricity prices over the last five years.

4 Renewable Energy in the EU

According to European Commission, energy prices and costs increases have been a significant political concern in the EU for decades. Recently, the European energy sector is in the middle of a shift away from the dependence on imported fossil fuels and their no-environment friendly characteristics and hence it needs high levels of investment to develop the power generating system replacing the existing one. Particularly, growth in wind and solar power, both having great impact on energy production costs, was caused by the EU aim to decrease CO2 emissions produced by electricity generation.

There are various ways to anticipate the impacts of the mentioned changes of current energy sector. Decarbonisation targets along with some other environmental and climate policy goals are designed to ensure a sustainable energy sector in the long run. However, in the short run, the costs are expected to be higher (mostly comprised of the initial investments needed for construction of the power stations and its infrastructure). According to European Commission (2014), “European governments expect these changes to deliver both short term benefits for consumers (jobs and quality of life) and long term sustainability objectives.” In this chapter, we focus on the European energy policy regarding the efforts to decrease the greenhouse gas emissions as well as the dependence of energy generation on the fossil fuel combustion (i.e. promotion of energy from renewable sources).

4.1 Renewable Energy Targets

The EU Renewable Energy Directive 2009 (which is still binding in its original version) has set a target of 20% final energy consumption from renewable sources by 2020. To achieve this goal, each EU member state received its national renewables target (reflecting the initial position) ranging from 10% (in Malta) to 49% (in Sweden) and committed to achieve this target until 2020.

In Figure 4.1, there are the target levels for the 13 EU countries analysed by the model in Chapter 5. For a proper designing and reforming of the renewable energy support schemes in each member state, European Commission provides guidance programmes and requires progress reports published by the countries every two years to show how they actually move towards the EU 2020 target (European Commission Guidance, 2013). Moreover, a new framework for climate and energy policies agreed by the European Commission in October 2014 sets a target of at least 27% share of RE in energy consumption in the EU by 2030.

Since individual EU member states have different available resources and unique energy markets, they have adopted distinctive national renewable energy action plans which determine the types of actions each of them should take to meet the renewable energy targets. The plans are composed of sectorial targets for electricity, heating and transport; planned energy policy measures and joint projects with other countries; national policies to develop biomass resources; and the different mix of renewables technologies the counties expect to employ (European Commission, 2011).

illustration not visible in this excerpt

Figure 4.1: Share of RE Sources in the EU Gross Final Energy Consumption

Source: Eurostat

According to the European Commission’s report, to make certain renewable energy technologies employed by each country competitive, public interventions such as support schemes are necessary. Since energy markets alone cannot deliver the desired level of renewables in the EU, the national support schemes are needed to overcome such market failure and encourage increased investment in renewable energy.

To limit a distortion of energy prices and markets caused by insufficiently designed schemes, European Commission has adopted guidance for EU countries designing and reforming the initial renewable energy support schemes suggesting that:

- financial support for renewables should be limited to what is necessary and should aim to make renewables competitive in the market,
- support schemes should be flexible and respond to falling production costs (as technologies mature, schemes should be gradually removed),
- unannounced changes to support schemes should be avoided as they undermine investor confidence and prevent future investment,
- EU countries should take advantage of the renewable energy potential in other countries via cooperation mechanisms set up under the Renewable Energy Directive (2009)

(European Commission: Guidance on the Use of Renewable Energy Cooperation Mechanism, 2013)

The cooperation mechanisms can have a form of statistical transfers, joint projects or joint support schemes. First, in a statistical transfer (an accounting procedure), an amount of renewable energy is deducted from one country’s progress towards its target and added to another’s. Due to such transfers, EU countries have an extra incentive to exceed their targets since they can receive a payment for energy transferred to others (European Commission Guidance, 2013). Second, through the joint projects, renewable energy projects regarding electricity generation are allowed to be co-funded by more EU member states which share the costs but then also the resulting renewable energy for the purpose of meeting their targets without having to transfer the energy between the countries. Third, a joint support scheme is designed to be co-funded by two or more EU countries and it should stimulate an increase in renewable energy production in one or all of the countries. This form of cooperation involves measures as a common quota, common feed-in tariff, or a common feed-in premium (European Commission Guidance, 2013).

The feed-in systems as economic policy mechanisms promoting active investment in and production of renewable energy sources are generally the most commonly used RES-E (electricity from renewable energy sources) support schemes in Europe. The feed-in tariff consists in long-term contracts connected with the electricity generation produced by an infant technology (wind and solar power generation) and accompanied costs for producers who are guaranteed price per kWh of electricity produced and hence face a lower risk arising from the renewable energy generation. The feed-in premium mechanism offers a payment in a form of premium above the electricity market price to the electricity producers using RE sources. It implies that with a decrease in market prices of electricity, producers face a risk of higher costs of production. The value of the risk then corresponds to the design of the premium mechanism (European Commission Guidance, 2013). In general, three main types of feed-in premiums exist.

First, fixed premium does not interact with average prices in the electricity market. The RES-E producers thus face a high risk arising from changes in the electricity prices. Second, feed-in premium with cap and floor prices introduces the highest and lowest level of price support reducing the potential risks but also surpluses. Third, sliding premium is generally computed as a function of the average price of electricity.

In contrast to feed-in systems, quota systems are quantity-based and technology-unspecific (Götz, 2012). While quantity-based support schemes define a certain percentage of RES-E in the electricity mix which needs to be provided by the producers, price-based support sets a fixed price for an energy amount of RES-E (e.g. one MWh). Hence, quota systems typically reach their targets but have an inherent uncertainty about the price. The second typical attribute is that quota systems are types of technology neutral support. It means that compared to feed-in tariff supporting the specific infant technologies in order to create a broader RES-E mix in the future, quotas usually lead to a more cost efficient deployment of RES-E, since every produced MWh of RES-E has the same value and hence producers can choose the cheapest and most cost efficient technology to produce the specific amount of RES-E (Nicolosi, 2010).

4.2 Greenhouse Gas Emission Targets

Along with the promotion and support of RE production, the EU aims to reduce its greenhouse gas emissions. As CO2 is the greenhouse gas mostly produced by human activities and is considered to be responsible for about 64% of man-made global warming (European Commission, 2015), we focus on this gas also in the econometric analysis in Chapter 5.

For 2020, the EU committed to reduce the overall domestic greenhouse gas emissions compared to 1990 levels by 20% (EU ETS, 2013). The EU initiative to reduce greenhouse gas emissions includes adopting various legislations and setting targets, but the key tool has recently been the EU Emissions Trading System (ETS). According to European Commission’s EU ETS (2013) report, it is a cornerstone of the EU policy regarding the climate change concerns and the biggest international system for trading greenhouse gas emission allowances. It operates in the 28 EU member states along with Iceland, Lichtenstein and Norway. The principle which the EU ETS works on is called ‘cap and trade’. It consists in setting a limit (‘cap’) on the emissions of a certain greenhouse gas and a subsequent reduction of this maximum amount. This system over time leads to the overall decrease in the level of the certain greenhouse gas emissions. The other characteristic of the scheme is the fact that each company receives a fixed amount of emission allowances which it can either use or sell to some other company. At the end of each year, a company’s allowances have to correspond to the level of emissions produced. If the emissions are higher than the number of allowances, fines from the EU are imposed.

Regarding EU ETS (2013), “by putting a price on carbon and thereby giving a financial value to each tonne of emissions saved, the EU ETS has placed climate change on the agenda of company boards and their financial departments.” According to Gerbelová (2014), there is a clear reduction in CO2 with the increase in CO2 prices (with 100 EUR/tonne of CO2, there is a 79% decrease expected in 2050 compared to the 1990 level). A sufficiently high CO2 price also promotes investment in clean, low-carbon technologies. Currently, the EU ETS covers around 45% of the EU greenhouse gas emissions and is considered to be the most cost-effective emission reduction method adopted in the EU (EU ETS, 2013).

5 Impacts of Renewable Energy Promotion: Panel Data Analysis

In this chapter, the influence of using renewable energy sources (instead of fossil and nuclear resources) in the EU electricity production on the EU end-user electricity prices is estimated by employing an econometric panel data analysis. Moreover, the impact of renewables in the EU energy production on the amount of CO2 emissions produced by each region is estimated by the model as well. In the following sections, we provide a review of past researches done on the same or closely related topics, data set and methodology characterisation and theoretical background along with the practical application of the model itself.

5.1 Literature Review

The relationship between the modern energy policies, regarding the significant increase in renewable energy (electricity) production, and the changes in energy (electricity) prices have been analysed by many research papers over the last decade. The empirical and theoretical studies using different methodologies and data sets have shown ambiguous results; in some cases they were even contradictory. Mostly, a positive response of the electricity prices to the increased proportion of renewables in RES-E production was found. However, some studies came to the opposite conclusion using arguments specific for the analytical methodology used.

Paraschiv, Erni & Pietsch (2014) analysed the impact of renewable energy promotion (wind and PV) in Germany on the changes in electricity prices. Their analysis revealed that the deployment of RES-E technologies enhance extreme price changes. While the results of their dynamic fundamental model implied that renewable energy caused a decrease in market spot prices, the prices for final consumers (which we are interested in for our analysis) increased overall due to the feed-in tariff costs added to the spot prices. Fernández, Ortiz & Bernat (2013) used their study to analyse the RES-E deployment in Spain and Germany, the EU members with very similar electricity systems both having significant role in the EU energy production. According to the study, public funding, set by the EU to promote investment in renewable energy generation facilities, means an additional cost to electricity pricing systems and can but does not have to lead to an increase in the electricity price for final consumers (depending on aspects specific for each country).

Moreno & López (2011) proposed to use panel data model with the aim of explaining the household electricity prices as a function of several economic variables related to renewable energy sources and electricity market regulation. Their results, using panel data set provided by Eurostat and covering 27 EU countries from 1998 to 2009, suggested that electricity prices increased with the deployment of RES-E, mainly due to high initial generation, distribution and transmission costs. González, de Miera & Vizcaíno (2008) in their study agreed with the general opinion that the private costs of RES-E generation were in most cases above those of conventional electricity but they stressed the fact that it was important to consider the social benefits provided by RES-E production, including the environmental aspects, which some studies had overlooked. On the case of Spanish RES-E generation, they showed that a reduction in the wholesale price of electricity (caused by lower costs of the energy component of the price, see Section 3.2.1) could be greater than the increased costs for the consumers arising from the RES-E support schemes (usually feed-in systems in the EU). Therefore, the net effect of RES-E on retail prices can be to reduce, not raise. A similar analysis was provided by Würzburg, Labandeira & Linares (2013) regarding the Austrian and German region. Their study also showed that the net effect of RES-E production can be positive to final consumers (i.e. decreasing the retail prices) depending on the region and assessment method chosen.

The other research question to be analysed by the model in this chapter is whether the amount of CO2 emissions produced by the EU countries significantly depends on the share of renewables in the EU energy production. Vast majority of researches based on this topic showed that there is sufficient evidence that the RE participation in the total EU energy production had an important impact on the carbon dioxide emissions produced by the economy. However, the fossil-based energy industry causing the majority of greenhouse gas emissions has not been typical only for the EU. Shafiei & Salim (2014) showed this fact using the data from all OECD countries; Özbugday & Erbas (2015) proved the long-run reduction in CO2 emissions caused by the replacement of fossil fuels by RE sources in the energy production processes in thirty six different countries; Moore, Lewis & Cepela (2010) came to the same conclusion while studying the United States energy production.

For our econometric panel data analysis we have chosen to study the effect of the EU RES-E production on the electricity prices. According to Moreno (2011), Paraschiv (2014), and the observed increasing trend in both the EU electricity prices and RES-E share in electricity production, we expect our model to show a positive impact of RES-E on the prices. On the contrary, regarding the analysis of the impact of RE promotion on the EU CO2 emissions, we expect it to be negative.

5.2 Data and Methodology

5.2.1 Data Set Summary

The data set (see Appendix A) encompasses 4 subsets of data for each of the 14 selected European regions reflecting a 4-year time period (from 2010 to 2013). The areas include thirteen European countries, namely Belgium (BE), the Czech Republic (CZ), Germany (DE), Spain (ES), Finland (FI), France (FR), the United Kingdom (GB), Italy (IT), Netherlands (NL), Poland (PL), Portugal (PT), Romania (RO), and Sweden (SE) along with a compound region called EU27. The data for EU27 are used only for comparisons with the individual member states and are excluded from the econometric analysis. They were computed either as an average or as an aggregation (specified for each data subset) of the data collected from the 27 EU member states which had entered the EU before the enlargement in June 2013.

The countries are selected according to their energy production share in the total EU energy production (regarding the data collected by Eurostat in 2013). The countries with the highest shares are included in the analysis excepting Denmark (2.4%) for which a sufficient amount of data needed for further analysis was not provided by the data sources. In addition, Portugal (with only 0.6% share in the total EU energy generation) is involved in the data set as it is a country with the highest share of renewable energy sources used for the electricity production. Altogether, the collected data describe 89.2% of the EU energy production (see Figure 5.1).

illustration not visible in this excerpt

Figure 5.1: EU Member States’ Shares in the Total EU Energy Production

Source: Eurostat: Energy Production 2013

The 4 mentioned subsets incorporate the information about each region’s:

(i) electricity prices for domestic households (EUR/kWh)
(ii) the percentage share of electricity generated by using renewable energy sources in the total electricity production
(iii) the percentage share of renewable energy in gross final energy consumption
(iv) the amount of CO2 emissions (Mt) produced by the region in total, per capita and per unit of energy production

Data adjusted to per capita or per unit of production values are incorporated in the analysis since they enable us to clearly compare the data from different regions regardless of either the area’s population or the level of production, respectively. The base currency used in the data set is EUR. The unit of measurement of each variable is mentioned in each specific case of the model application and interpretation.

5.2.2 Data Sources

The examined data have been acquired from several resources. The electricity prices for households have been provided by Eurostat using the new methodology of data collection (from 2007 onwards) and excluding all taxes and levies. The proportions of electricity generated by using renewable energy sources in total electricity production for each of the 14 regions were obtained from Global Energy Statistical Yearbook 2014 published by Enerdata. The percentage shares of renewable energy in gross final energy consumption have been found in the Eurostat database as well as the electricity prices mentioned above. The data are submitted on the basis of an Annual Joint Questionnaire (Eurostat/IEA/United Nations Economic Commission for Europe) employing an internationally agreed methodology.

The accuracy of the basic data depends on the quality of the national statistical systems. However, Eurostat verifies to the highest possible extent whether the reported data respect the prescribed methodology. Hence the data are considered to be highly comparable and accurate. The last subset of the econometric model data set is the amount of CO2 emissions (in Mt) produced by fuel combustion by each region in total, per capita and per unit of energy production. The source of these data was again the already mentioned Global Energy Statistical Yearbook from 2014 which can be found on the Enerdata website.

5.2.3 Variables

Country Each of the examined European regions is assigned a natural number from 1 to 14 as follows: 1 = EU27, 2 = BE, 3 = CZ, 4 = DE, 5 = ES, 6 = FR, 7 = IT, 8 = NL, 9 = PO, 10 = PT, 11 = RO, 12 = FI, 13 = SE, 14 = GB. The numbers altogether form an id dimension for the panel data. Each id variable is constant for all time periods and has only data ordering function in the panel data analysis.

Year Our data set consists of 4 time periods (2010 to 2013, yearly) which are the same for each of the researched countries and serve as time variables of the panel data model. The year 2010 was chosen as a starting point since it has been the first year in which the Renewable Energy Directive 2009 (see Section 4.1) was already in force. All sufficient data for the year 2014 were not found at the time of our research. Hence the data set ends with 2013 data.

Electricity Prices (EUR/kWh) For each country in the data set, the variable elprice reflects the average electricity price for households comprised of electricity basic price, transmission, system services, distribution and other services, and excluding taxes and levies. For the variable EU27 as a country aggregation, the values are calculated by weighting the twenty seven EU member states’ national prices with the latest available national consumption for the households.

Electricity from Renewable Energy (%) The values of the variable elfromRE are computed as the ratio between the electricity production from selected renewable energies (hydro, wind, geothermal and solar) and the total electricity supply for end-users for each id and time variable of the panel data set.

Renewable Energy in Energy Consumption (%) The variable REcons serves as an indicator measuring how intensive is the use of renewable energy and, by implication, the degree to which renewable fuels have submitted fossil and/or nuclear fuels.

CO2 Emissions (Mt) The total amount of CO2 emissions produced by each region each year is represented by the variable CO2. The units of measurement are metric tons. The variables CO2percap and CO2perprod correspond to the level of carbon dioxide emissions adjusted to the region’s population and the total energy production, respectively. These variables serve for an initial data set analysis and comparison of the examined countries. However, in the econometric model, only the variable CO2 is included since we study the impact of RE sources on the total amount of carbon dioxide emitted.

Table 5.1: Summary of the Variables

illustration not visible in this excerpt

Source: Author’s data set and Stata computation

In Table 5.1, the summary of the researched data set is presented by using the Stata statistical software. The number of observations reflects the fact that the data from 13 regions over the 4 mentioned time periods are included in the computation. The data for EU27 have been excluded from the summary as they could distort the results. They represent either averages or summations of the values from the countries already included in the statistics.

According to Table 5.1., the electricity price (represented by the variable elprice) paid in the selected European regions by households is estimated to be 12.665 EUR cents per kWh on average. While the lowest average price, 7.95 EUR cents per kWh, was paid by consumers in Romania in 2012, the highest average electricity price in the data set, 17.72 EUR cents per kWh, applied to Spanish households in 2013.

Regarding the variable elfromRE, the minimum proportion of electricity generated by using renewable energy sources in the total electricity production was recorded in Poland in 2010 at the level of 7.4% while the maximum share of 62.5% was monitored in Portugal in 2013. The overall mean percentage value of renewable energy participation in the total European electricity production was 25.625% over the examined 4-year time period for our data set, while the average share for the EU27 countries was about 2% higher, specifically 27.8%. In seven out of the thirteen countries in the data set, the overall average proportion was below the 25.625% level, namely in Poland (9.7%), the Czech Republic (9.8%), the United Kingdom (11.8%), the Netherlands (12.7%), Belgium (13.4%), France (15.2%) and Germany (22.6%). The above average participations of renewable energy in electricity generation were seen in Sweden (54.8%), Portugal (52.1%), Finland (35.1%), Spain (34%), Italy (31.9%) and Romania (30.1%). Concerning the values of the REcons variable, we can see that the percentage share of renewable energy in the gross final energy consumption measured in the countries included in the data set ranges from 3.3% to 51.9% having the mean at 16.8% level. The values substantially vary due to the differences in the aims of energy policies and approaches to production and consumption of renewable energy in the examined European countries albeit there are some targets set by the EU (see Chapter 4 for more detailed information about the energy policies and the EU approach).

The least intensive use of renewable energy was seen in the United Kingdom in each of the examined time periods whereas, by contrast, Sweden each year showed the highest degree to which renewable sources of energy have substituted fossil and/or nuclear energy sources. Apart from Sweden, also three other countries from the data set exceeded on average the mean value, specifically Finland (33.8%), Portugal (24.5%) and Romania (22.7%). However, the below average values were found in most of the studied regions: in the United Kingdom (4%), the Netherlands (4.4%), Belgium (6.3%), Poland (10.5%), the Czech Republic (10.7%), Germany (12%), Italy (12.8%), France (13%) and Spain (14.2%). See Figure 5.2 for a graphical summary of these values along with the average shares of renewable energy in the electricity production.

illustration not visible in this excerpt

Figure 5.2: RE in the EU Electricity Production and Energy Consumption

Source: Author’s computation using the data in the data set.

The last three variables from the summary are associated with the amount of carbon dioxide emissions produced by each country in the data set. According to Table 5.1, the mean level of CO2 emissions produced by the countries from our sample was approximately 234.27 Mt a year. However, the individual values varied considerably, from the minimum at 38.3 Mt per year observed in Sweden in 2013 to the maximum at 756.8 Mt per year seen in Germany in 2013. Since the Swedish surface area is almost 1.2 times larger than the German one, it is clear that size of the region’s surface does not imply larger carbon dioxide emissions produced.

Nevertheless, some other variables can influence the level of pollution generated by a region, such as the region’s population or the level of the energy production. Hence, the data adjusted to per capita and per unit of energy production values are included in this initial data set analysis. As we can see in Table 5.1, the average amount of CO2 emissions per inhabitant was 7.1735 Mt a year. The lowest carbon footprint observed in our sample was left by an average Romanian in 2013, approximately 3.37 Mt a year, whereas the highest amount of carbon dioxide produced per capita was seen in Finland in 2010, 12.71 Mt a year.

Albeit the variance of the mentioned per capita values is relatively high, the values per unit of energy production vary even more across the data set. The mean amount of carbon dioxide emissions produced per 1 Mtoe of energy was 4.8833 Mt. The least has been emitted by the Swedish energy production, 1.0943 Mt/Mtoe in 2012. The most polluting (in terms of carbon dioxide emissions) energy production has been found in Italy, emitting 12.9667 Mt of CO2 per Mtoe of energy generated in 2010. However, a decreasing trend of CO2 emissions in Europe has been seen in majority of the researched countries. The total amount of produced carbon dioxide has been reduced over the 4-year time period in 11 out of the 13 countries. France and Germany represented the only exceptions. In terms of per capita values, the figures decreased in all regions apart from Germany and Portugal. Eventually, regarding the quantity of CO2 emitted per Mtoe of energy production, all the regions excluding Great Britain, Germany and France saw a decline in the emission level. This short summary implies that Germany is the only country which has not been able to cope with cutting down the greenhouse gas emissions by any measure.

5.3 Theoretical Framework

In our model, we use panel data with the 13 selected European countries as the cross-sectional units, and years from 2010 to 2013 as the time dimension. The addition of a time component to the static nature of cross-sectional data brings with it a greater leverage on questions of causality. Due to this fact we can more effectively estimate the impact of one variable on the other with a panel data set. More specifically, in this chapter we are interested in two major research questions, whether higher shares of electricity from RE in total electricity production cause an increase in consumer prices of energy, and whether a higher proportion of RE in gross final energy consumption leads to a considerable decrease in CO2 emissions produced by the European countries. Sources of econometric theory forming the basis for the analysis provided in this chapter are textbooks written by J. M. Wooldridge (2002, 2009).

Before we formulate our model for the estimation of the mentioned effects, there is another rationale for using more complex panel data analysis instead of simple cross-sectional analysis. If we use cross section from only one period (e.g. 2010) and run a simple regression with one independent variable, we probably obtain results suffering from omitted variable problems. One possible solution is to try to control for more factors affecting the dependent variable in a multiple regression analysis. However, many factors can be hard to realize and control for. In this case, we can use panel data to view the unobserved factors affecting the dependent variable as consisting of two types, those that are constant for each cross-sectional unit and those that vary over time, and manipulate with them differently in the analysis.

5.3.1 First Differences Estimation

We can write a panel data model with a single observed explanatory variable, letting i denote the cross-sectional unit and t the time period, as:

illustration not visible in this excerpt

In the notation, i = 2, 3 ... 14 denotes the countries in the data set according to their assigned id numbers (see Section 5.2.3.), t = 2010, 2011, 2012, 2013 stands for the time period. The variables [Abbildung in dieser Leseprobe nicht enthalten] are binary variables equal to one for t = 2011, 2012 or 2013, respectively, otherwise they equal to zero. Due to the inclusion of the yearly dummy variables in the model, we allow the intercept to change over time. The variable a_i captures all unobserved, time-constant factors which influence y_it and is called unobserved effect or fixed effect since it is fixed over time. The error u_it is referred to as the idiosyncratic error. It represents unobserved factors changing over time and affecting y_it.

Since we assume that the unobserved effect a_i is uncorrelated with x_it in our analyses, we can use the first-differences (FD) estimation to obtain the estimate of〖 β〗_1 and eliminate the unobserved effects from the regression equation (5.1). By using the differencing method, we acquire the following equation for t = 2011, 2012, 2013:

[Abbildung in dieser Leseprobe nicht enthalten] (5.3)

If the equation (5.3) satisfies the first four assumptions listed below, the FD estimator (pooled OLS estimator) is unbiased. If all six assumptions are satisfied, usual standard errors and test statistics are valid.

Assumption FD.1. For each i, the model is:

illustration not visible in this excerpt

where the parameters β_j are to be estimated and a_i is the unobserved effect.

Assumption FD.2. Each period we observe the same random sample.

Assumption FD.3. Each explanatory variable changes over time (for at least some i) and no perfect linear relationships exist among the explanatory variables.

Assumption FD.4. For each t, the expected value of the idiosyncratic error given the explanatory variables in all time periods and the effect[Abbildung in dieser Leseprobe nicht enthalten] or by implication, [Abbildung in dieser Leseprobe nicht enthalten].

Assumption FD.5. The variance of the differenced errors, conditional on all explanatory variables, is constant: Var[Abbildung in dieser Leseprobe nicht enthalten]…T. Hence the differenced errors are homoskedastic.

Assumption FD.6. The differenced errors are serially uncorrelated. It means that for all t≠s, the differences in the idiosyncratic errors are uncorrelated (conditional on all explanatory variables): Cov[Abbildung in dieser Leseprobe nicht enthalten].

5.3.2 Fixed Effects Estimation

The other method for estimation of the unobserved effects panel data models, eliminating the fixed effect a_i, is the fixed effects (FE) transformation (or within transformation). Again, we consider an unobserved effects model with a single explanatory variable.

For each i we then have:

illustration not visible in this excerpt

(5.5)

where the equation (5.5) represents the equation (5.4) averaged over time. To eliminate the factors in a_i, we subtract (5.5) from (5.4) and obtain:

illustration not visible in this excerpt

(5.6)

Since we have disposed of the fixed effects included in a_i, we can use the pooled OLS to estimate〖 β〗_1, as well as in the FD case. The obtained fixed effects or within estimator is then unbiased if the first four assumptions, identical to FD.1 through FD.4 listed above, are fulfilled. Under all six assumptions (the fifth and sixth FE assumptions are mentioned below), the FE estimator of β_1 is the best linear unbiased estimator. Hence, the linear unbiased FD estimator should be worse than the FE estimator under such conditions.

Assumption FE.5. The variance of the errors, conditional on all explanatory variables and the unobserved effect, is constant: Var[Abbildung in dieser Leseprobe nicht enthalten] for t=1…T. Hence the errors are homoskedastic.

Assumption FE.6. The idiosyncratic errors are uncorrelated (conditional on all explanatory variables and a_i): Cov[Abbildung in dieser Leseprobe nicht enthalten], for all t≠s.

Further information regarding the FD and FE estimation processes along with a comparison of these two methods are included in theoretical appendix, Appendix B.

5.4 Practical Applications of the Theory

In this section, we estimate our panel data model specifications using the theory explained in Section 5.3 and Appendix B. Each specific model equation with a single observed explanatory variable allows us to control for a predefined factor that is expected to affect the dependent variable.

5.4.1 Electricity Price and Renewable Energy

In our first model specification, we estimate the following equation:

illustration not visible in this excerpt

where i = 2,3 … 14 denotes the 13 European countries according to their assigned id numbers (see Section 5.2.3.) serving as the control group; t = 2010, 2011, 2012, 2013 stands for the time period over which the data have been collected; d2011, d2012, d2013 are year dummy variables; a_i is the unobserved effect; and u_it is the idiosyncratic error.

Using the Stata software, we estimate the model to discover whether there is a significant relationship between the proportion of RES-E in total electricity production in the EU (the variable elfromRE) and the European prices of electricity for households (the variable elprice). According to the reviewed literature (see Section 5.1) and the fact that the electricity generation from RE sources is relatively uncompetitive, uncertain and connected with high initial costs; we expect it to have a positive effect on the electricity prices in the EU. As we decided to use a log-log model, the estimated coefficient β_(1 )on the variable elfromRE signifies the elasticity of electricity price with respect to the share of renewable energy sources in the total EU energy production.

We use FD and FE estimation methods to obtain the estimate of β_1 since the variable elfromRE is expected to be correlated with the unobserved effects in a_i (fixed or roughly constant over the 4 years in each of the countries). Factors assumed to be contained in a_i are e.g. already built infrastructure for power plants using fossil, nuclear or renewable energy sources; the access to fossil and nuclear energy sources; and the natural conditions suitable for development of renewable energy generation in each of the countries (such as the weather, duration of average day and sun light, terrain structure, geographical location etc.).

First Differences

To obtain unbiased and consistent pooled OLS estimator and valid test statistics using the FD estimation method, all six FD assumptions have to be satisfied (see Section 5.3.1). We verify these assumptions using Stata, running the FD regression and obtaining the parameters’ estimates for the following equation:

illustration not visible in this excerpt

According to Stata outputs of several tests (see Section C.1 in Appendix C) we consider the assumptions to be fulfilled. The estimate of β_1 is β ̂_1= .16967 (standard deviation is equal to .04822) with p-value = .001 (see Table 5.2). Hence, the variable lnelfromRE is statistically significant at 5% (or even 1%) significance level as .001 < .01. Since we have already estimated the value of the coefficient β_1, we can now interpret the relationship between the dependent and independent variables. For instance, a 10% increase in the share of renewable energy sources in the total EU electricity production is predicted to cause an increase of the electricity price in the examined European countries by approximately 1.67% on average based on our collected data. The coefficients on all three year dummy variables d2011, d2012 and d2013 are statistically significant at 5% significance level with p-values equal to .004, .02 and .036 respectively. These variables serve as different intercepts for each of the years from 2011 to 2013 and account for secular changes (e.g. market trends) influencing the dependent variable that are not being modelled.

The R-squared of the model specification is R2 = .5515. It implies that approximately 55.15% of the variation in the electricity prices in the EU countries is expected to be explained by the variation in the independent variables included in the model. The value of the R-squared is not very high albeit the model includes the time dummy variables which often cause a noticeable increase in the R-squared since they often account for effects that explain much of the variation in the dependent variable. While separately regressing the variable lnelprice solely on lnelfromRE, we indeed obtain the R-squared with a lower value, specifically R2 = .2154. Hence, the variation in the share of renewable energy sources in the total electricity production is estimated to explain about 21.5% of the variation in the electricity prices in the studied European regions.

Fixed Effects

As well as in the case of FD estimation, the assumptions needed for acquiring an unbiased and consistent pooled OLS estimator have to be verified before we interpret our regression results. In Section C.1, Appendix C, we describe the justification of each assumption’s verification. Once all the six FE assumptions are fulfilled, we can estimate the model equation (5.7) and interpret the outcome of the regression using FE transformation.

The results of the FE regression run in Stata (see Table 5.2) show a positive effect of the explanatory variable lnelfromRE on the dependent variable lnelprice. Specifically, e.g. a 10% increase in the proportion of the RE sources in the total EU electricity production is estimated to cause approximately 1.92% increase in the electricity price for the European households. The variable lnelfromRE is statistically significant at 5% significance level as well as all the time dummy variables included in the model. The exact FE (and FD) regression results can be seen in Table 5.2 on the following page. In addition, an interesting part of the FE regression output is Rho denoting the proportion of the total variation of dependent variable which is explained by the fixed effect a_i. In our case, Rho = .9805, hence only less than 2% of the total variation in lnelprice is caused by the idiosyncratic error.

Fixed Effects versus First Differences

In Table 5.2, we can see the summary of the FD and FE regression results obtained by using Stata. Both estimation methods indicate a positive effect of the participation of the RE sources in the European electricity production on the prices of electricity. Both estimates of the coefficient on the variable lnelfromRE are very statistically significant. However, using the FE transformation, the coefficient (.192486) is estimated to be larger than the FD estimate (.169669) and the expected lnelfromRE standard errors in the FE estimation are lower. It implies that the FE estimate is more significant, both statistically and economically.

While noticing the values of the R-squared, we have to take into consideration the fact that each of them has a different meaning. The R-squared from the FD regression denotes that approximately 55% of the sample variation in the lnelprice is explained by the variation in the independent variables included in the model. On the contrary, the value of the within R-squared from the FE regression means that about 71% of the lnelprice variation within each of the countries in the data set over the 4 years (excluding the fixed effects a_i) is explained by the explanatory variables. Since both the FD and FE assumptions were satisfied before running the regressions, the FE estimator is considered to be the best linear unbiased estimator and thus better than the FD estimator. Moreover, during the FD estimation we lose the first year observations due to which we can miss some important data.

Table 5.2: Regression Results (lnelprice on lnelfromRE)

illustration not visible in this excerpt

Source: Author’s data set and Stata computation

5.4.2 CO2 Emissions and Renewable Energy

For this model specification, we use the same approach as in the previous case. We base our analysis on the estimation of the following equation:

illustration not visible in this excerpt

where i = 2,3 … 14 denotes the 13 European countries; t = 2010, 2011, 2012, 2013 stands for the time component; a_i is the fixed effect; and u_it is the idiosyncratic error. The variables CO2 and Recons are described in Section 5.2.3. The major aim of estimating this model specification is to find the answer to the question whether an increase in the proportion of renewable sources of energy in total energy consumption of the specified EU member states (the variable REcons) has a significant effect on the level of CO2 emitted by these regions (the variable CO2).

Since RE resources are considered to be the “cleaner” alternative to the fossil-based energy production, we expect the growth of its share in total energy consumption to have a negative effect on the CO2 emissions generated. Again, we estimate the model using the Stata software. We assume that the variable REcons is correlated with the fixed unobserved effects a_i (including e.g. the already built infrastructure for power plants or the natural conditions such as the weather, average day duration etc.) hence we use FD and FE estimation methods to obtain the estimates of β_1 as well as we have done it in Section 5.4.1.

First Differences

As in Section 5.4.1, we have to verify the six FD assumptions before we interpret our model results. The first three assumptions (FD.1 through FD.3) are verified directly by considering the format of the model equation (5.10) and the data set. The other three assumptions can be satisfied by using several tests (regarding endogeneity, autocorrelation and heteroskedasticity) and running regression of the following equation:

illustration not visible in this excerpt

Once all the FD assumptions are considered to be fulfilled (see Section C.2., Appendix C) we can focus on the results of the FD regression. The estimate of β_1 is approximately β ̂_1= -3.745 with p-value = .017. Hence, the variable REcons is statistically significant at 5% significance level (.017 < .05). The minus sign of the value of β ̂_1 indicates that our initial expectations about the variables’ relationship were correct. According to the results of the FD regression, the relationship between the variables REcons and CO2 can be interpreted as follows: if the proportion of RE resources in the total energy consumption increases by e.g. 1 percentage points, the amount of CO2 emissions produced by the examined European regions is estimated to decrease by approximately 3.745 megatons per year on average. In addition, the R-squared of the model specification is R2 = .1571. Hence, approximately 15.71% of the variation in the level of CO2 emissions caused by the EU countries is estimated to be explained by the variation in the renewable energy sources’ participation in total energy consumption in the EU countries.

Fixed Effects

To obtain the estimate of β_1 from the equation (5.10) and then to be able to compare the results with the FD estimation, we use the FE transformation as well as in the previous section. Since, the assumptions FE.1 through FE.6 are considered to be satisfied (see Section C.2, Appendix C), we can proceed to FE regression results. The regression output indicates a negative effect of the explanatory variable REcons on the dependent variable CO2. Specifically, an increase in the share of RE sources in the EU energy consumption by e.g. 1 percentage point is estimated to cause a decrease in the yearly amount of CO2 emitted by the EU countries by approximately 5 megatons on average (see Table 5.3). The only explanatory variable of the model, REcons, is statistically significant at 5% significance level. In addition, the Rho of the FE regression, denoting the proportion of the total variation of dependent variable explained by the fixed effect a_i, is equal to .99765. It implies that only approximately .00235% of the total variation in CO2 is caused by the idiosyncratic error.

Fixed Effects versus First Differences

The outputs of both the FD and FE regressions are summarized in Table 5.3. The FD estimation as well as the FE transformation indicates that the proportion of RE sources in the EU countries’ energy consumption has a negative effect on the CO2 emission level, as we expected. For both estimation methods, the estimates of the coefficient on REcons are statistically significant. By using the FE method, we have obtained an estimate with noticeably higher negative effect (-5.0017) than in the case of the FD estimation (-3.74481). The standard errors of the β_1 estimates are lower for the FE estimator (1.097557) than those acquired by the FD regression (1.359718). It implies that the FE estimate is both statistically and economically more significant.

The value of the R-squared for the FD regression denotes that approximately 15.71% of the sample variation in CO2 is explained by the variation in REcons. By contrast, the R-squared obtained from the FE regression is so called within R-squared indicating that about 35.34% of the CO2 variation within each of the countries in the data set over the 4-year period (excluding the unobserved effects a_i) is explained by the variation in REcons. Albeit in both FD and FE estimations we have verified all assumptions necessary to acquire an unbiased consistent estimator, only the FE estimator is considered to be the best linear unbiased estimator under FE.1 through FE.6. Hence we assume that it performs better than the FD estimator.

Table 5.3: Regression Results (CO2 on REcons)

illustration not visible in this excerpt

Source: Author’s data set and Stata computation

5.5 Justification of the Model Results

5.5.1 Electricity Price and Renewable Energy

As expected from the literature review (see Section 5.1), our econometric model showed a positive effect of the RES-E share in the total electricity production on the final price of electricity for the EU households. We used electricity prices excluding taxes and levies in our analysis since these financial charges considerably vary across the countries in the data set and are specific to each member state’s economic and political regime. Hence, we specifically analysed the impact of the rising support for RES-E production (binding for all EU members) on the energy and network element of the EU electricity prices. Since power stations using the RE sources (mainly wind, hydro and solar power) are connected with high initial construction, transmission and distribution costs creating an additional cost burdens for electricity end-users (including households), it makes sense that the mentioned impact on the EU electricity prices has been showed to be positive and significant.

The high initial investments, regarding the energy and network components of the electricity prices, are needed mainly for building infrastructure, construction of power plants, and transmission and distribution of the power. These investments are very similar for each EU member state (deciding to build a new RES-E network) and are expected to increase the cost of providing renewable electricity, especially during early years. They include for instance: prospecting for publicly acceptable and suitable place with good access to RE resources and transmission lines; developing standards and permitting issues for renewables; marketing costs of communicating the benefits of renewables to consumers who are used to buying electricity from traditional sources; and installation, operation and maintenance costs including power plant constructions but also e.g. worker trainings regarding the proper treatment of the new technologies.

5.5.2 CO2 Emissions and Renewable Energy

The results of the second model specification (see Section 5.4.2) indicate that the increase in the use of RE sources in the total EU energy consumption leads to a decline in the amount of CO2 emitted by the EU member states. This regression output corresponds not only to the past researches focusing on the same topic, but also to the lifecycle approach of analysing the level of CO2 emissions associated with each energy source and the process of construction, operation and decommissioning of the source-specific electricity generators (see Section 2.3). Since renewable sources are responsible for much lower amounts of CO2 emissions as compared to fossil fuels according to the lifecycle approach, the negative relationship between the level of CO2 produced and the share of RE sources in the electricity production was expected to be observed.

However, the provided analysis does not include the entire carbon footprint of the electricity production since it examines only data concerning the operation of the generators. This implies that the regression results are biased due to omitting CO2 emissions associated with the other two phases of the generators’ lifecycle, namely construction and decommissioning. Since the two mentioned phases are expected to lead to a positive level of CO2 produced in case of both fossil and RE sources, after including these data into the analysis, the results might change depending on how significant carbon footprint produced during construction and decommissioning of the generators is associated with the RE sources. If the level of CO2 emissions resulting from the mentioned processes was significant, the examined effect would be smaller in absolute terms (hence, closer to zero).

6 Conclusion

The aim of this thesis was to create a sufficient overview of the EU renewable energy and climate policy, its targets towards next few years and the impacts of the increasing share of renewables in the EU energy consumption and production on final consumers and the environment. More precisely, we focus on renewables in the electricity production (RES-E) since it plays a decisive role in achieving the EU renewable energy targets and since the changes in electricity prices affect vast majority of the EU inhabitants. The paper starts with a summary of the characteristics, advantages and drawbacks regarding both renewable and non-renewable sources of energy along with an overview of electricity production by source to show how important renewable energy sources are in the current electricity generation processes. Then, the thesis follows with the description of the EU electricity prices components and an analysis of the factors which generally influence the price changes. In addition, the EU renewable energy and climate policy approach and targets are included in the work to show that the role of renewable sources in the energy sector is expected to be even more important than it already is.

The last part of the paper consists of the econometric model analysing the effects of the renewable energy use on the electricity prices for final consumers and the amount of carbon dioxide emissions produced in the EU a year. We have decided to use panel data analysis as, while using the first differences and fixed effects methods of estimation, it allows for the effects that are unobserved and fixed over time in our model to be correlated with the explanatory variables and eliminated through the regression. Hence we can dispose of the potential omitted variable problem and study the effects of explanatory variables on the dependent variables over a given time period. The results of our model analysis suggest that household electricity prices in the studied EU member states increase with the deployment of RES-E production. Such effect on prices was anticipated, since the majority of renewable energy technologies increase electricity generation, distribution and transmission costs. Moreover, in the EU the largest part of investments for electricity production over the last few years was devoted to new wind power stations and solar photovoltaics which are connected with the highest initial costs when compared to conventional generation methods. On the contrary, a negative effect of the renewables used in the EU energy consumption on the CO2 emissions produced was found by the model regression, as it had been expected while formulating the model since the lifecycle CO2 emissions (covering construction, operation and decommissioning of the power stations) were considerably lower for renewable sources in comparison with fossil-based fuels.

This thesis serves well as an overview in the field of renewable energy and electricity production, consumption and pricing in the EU. It provides the essential background for this topic along with the detailed analysis of two specific impacts of the deployment of renewable energy technologies on the European level. However, within the scope of this thesis, we cannot hope to cover all the possible consequences of the promotion of renewable energy sources in Europe. Nevertheless, this fact makes a space for further research and study. Such work could concern, for instance, the question how the rapid replacement of fossil fuels by renewables in the EU electricity production affects the changes in each particular component comprising the value of the EU electricity prices (energy, network and taxes/levies component separately); or how e.g. the economic development, employment in rural areas and security of energy supply can be affected by this trend. In addition, it would be also interesting to repeat this study in a few years and ascertain whether the high initial costs of renewable energy power stations gradually pay off and allow the EU electricity prices to decrease, taking the advantage of the relatively low operation and maintenance costs of RES-E stations and zero costs of obtaining the energy source (as wind, water and solar energy can be usually used free of charge unlike oil, coal or natural gas). The range of the possible future studies based on this thesis is wide and we think each of them would be exciting to follow.

Appendix A: Data Set of the Model

illustration not visible in this excerpt

Note: The bold figures represent the minimum and maximum values.

Source: See Section 5.2.2.

Appendix B: Theoretical Framework of the Panel Data Model

Since our data set used for the econometric analysis in Chapter 5 consists of both cross-sectional and time series dimensions following the same units over time, we call it panel data set. In other words, by panel data we mean data containing repeated measures of the same variable taken from the same set of cross-sectional units over time. In our applications the units are the 13 selected European countries and time periods are years from 2010 to 2013.

B.1 First Differences Estimation

In Section 5.2.1 we use a single observed explanatory variable model, letting i denote the cross-sectional unit and t the time period, as:

illustration not visible in this excerpt

where i = 2,3 … 14 denotes the countries in the data set according to their assigned id numbers (see Section 5.2.3.), t = 2010, 2011, 2012, 2013 stands for the time period and the variables [Abbildung in dieser Leseprobe nicht enthalten] are yearly binary variables. The intercept for t = 2010 is δ1, for t = 2011 it is δ12, for t = 2012 it equals to δ13, and when t = 2013 we have the intercept of δ14. Since 2010 is in our case considered to be the base year, the three dummy variables help us to find the influence of the time when the data were observed (2011, 2012 or 2013) on the value of the dependent variable, holding all factors influencing the dependent variable fixed, and compare this value with the value in 2010. For instance, the coefficient δ2 on the year dummy variable 〖d2011〗_t shows us what the difference between the values of y_it in 2011 and 2010 is, holding all other factors affecting y_it fixed.

The variable a_i captures all unobserved, time-constant factors which influence y_it (such as geographical features of a country; different historical factors with an effect on y_it or even some not exactly constant factors which are, however, roughly constant over the relatively short time period). Generally, it is called unobserved effect or fixed effect since it is fixed over time. Due to the variable a_i, the model in (5.1) is also called fixed effects model. The error u_it is often referred to as the idiosyncratic (specific) or time-varying error. It represents unobserved factors changing over time and affecting y_it. The idiosyncratic error along with the unobserved effect is called the composite error[Abbildung in dieser Leseprobe nicht enthalten]

To estimate the parameter of interest, β_1, we can generally use directly the method of pooled OLS. However, for pooled OLS to produce a consistent estimator of β_1, we have to assume that the unobserved effect a_i is uncorrelated with x_it. Since we will assume the opposite in our analyses, the estimator in this case would be biased and inconsistent. If we want to allow the unobserved factors included in a_i affecting y_it to be correlated with x_it, we can use differencing method to obtain the first-differences (FD) estimator. The key assumption in this case is that the idiosyncratic errors are uncorrelated with the explanatory variable in each time period:

illustration not visible in this excerpt

It implies that the explanatory variables are strictly exogenous after we take out the unobserved effect a_i. If a_i is correlated with x_itj, then under (5.2), x_itj will be correlated with the composite error:[Abbildung in dieser Leseprobe nicht enthalten]. To eliminate a_i by using differencing method, we (or any statistical software we use) just difference adjacent periods and then run pooled OLS regression. In our 4-period case, we subtract time period one from time period two, time period two from time period three and finally time period three from time period four. We obtain the following equation for t = 2011, 2012 and 2013:

illustration not visible in this excerpt

If the equation (5.3) satisfies the first four assumptions of the listed below, a pooled OLS estimator (the FD estimator in this case) is unbiased. To acquire consistent OLS estimator, ∆u_it has to be uncorrelated with ∆x_it. Moreover, we must assume that ∆u_it are uncorrelated and homoskedastic over time for the usual standard errors and test statistics to be valid. Hence we will further test serial correlation and heteroskedasticity in the first-differenced equation in our model specifications. The important assumptions for the first differences estimation are as follows:

Assumption FD.1. For each i, the model is:

illustration not visible in this excerpt

where the parameters β_j are to be estimated and a_i is the unobserved effect.

Assumption FD.2. Each period we observe the same random sample.

Assumption FD.3. Each explanatory variable changes over time (for at least some i) and no perfect linear relationships exist among the explanatory variables.

Assumption FD.4. For each t, the expected value of the idiosyncratic error given the explanatory variables in all time periods and the effect [Abbildung in dieser Leseprobe nicht enthalten], or by implication, [Abbildung in dieser Leseprobe nicht enthalten].

Assumption FD.5. The variance of the differenced errors, conditional on all explanatory variables, is constant: Var([Abbildung in dieser Leseprobe nicht enthalten]2 for t=2…T. Hence the differenced errors are homoskedastic.

Assumption FD.6. The differenced errors are serially uncorrelated. It means that for all t≠s, the differences in the idiosyncratic errors are uncorrelated (conditional on all explanatory variables): [Abbildung in dieser Leseprobe nicht enthalten].

B.2 Fixed Effects Estimation

The other method for estimation of the unobserved effects panel data models is the fixed effects (FE) transformation which is, as well as the FD estimation, one of the ways to eliminate the fixed effect a_i which is expected to be correlated with the explanatory variable(s) in any time period. In our model specifications we will compare the results of the FD and FE estimations and test which of them is more efficient under certain assumptions. For the description of the FE transformation (also called the within transformation), we consider an unobserved effects model with a single explanatory variable, for each i we then have:

illustration not visible in this excerpt

where the equation (5.5) represents the equation (5.4) averaged over time, with [Abbildung in dieser Leseprobe nicht enthalten]it and likewise for x ̅_i and u ̅_i. To eliminate the fixed factors in a_i appearing in both equations we subtract (5.5) from (5.4) and obtain:

illustration not visible in this excerpt

where[Abbildung in dieser Leseprobe nicht enthalten] is the time-demeaned data on y (and similarly for x ̈_it and u ̈_it).

Now we have disposed of the fixed effects included in a_i and as well as in the FD estimation we can use the pooled OLS to estimate〖 β〗_1. The pooled OLS estimator based on time-demeaned variables is called the fixed effects or within estimator since the OLS on (5.6) uses time variation in y and x within each cross-sectional observation. The assumptions for the fixed effects estimation are listed below:

Assumption FE.2. See Assumption FD.1.

Assumption FE.2. See Assumption FD.2.

Assumption FE.3. See Assumption FD.3.

Assumption FE.4. See Assumption FD.4.

As we can see, the first four assumptions are identical to the assumptions for the FD estimator. Under them, the FE estimator is unbiased (as well as in the case of first differences). The key assumption is the strict exogeneity assumption (FE.4.).

Assumption FE.5. The variance of the errors, conditional on all explanatory variables and the unobserved effect, is constant:[Abbildung in dieser Leseprobe nicht enthalten] Hence the errors are homoskedastic.

Assumption FE.6. The idiosyncratic errors are uncorrelated (conditional on all explanatory variables and [Abbildung in dieser Leseprobe nicht enthalten], for all t≠s.

Under the all first six assumptions, the FE estimator of β_1 is the best linear unbiased estimator. Hence, the linear unbiased FD estimator should be worse than the FE estimator under such conditions.

B.3 Fixed Effects versus First Differences

While comparing two different estimators we often use unbiasedness and consistency as the criteria. However, since both FE and FD estimators are unbiased under the Assumptions FE.1 through FE.4 as well as asymptotically consistent (with T fixed as N →∞), the decision on which estimator is better to use then depends on considering some other factors.

Hence we focus on the error structure. If u_it is serially uncorrelated, the FE estimator is more efficient and used rather than the FD estimator. On the contrary, when u_it follows a random walk (i.e. very substantial positive autocorrelation), then the ∆u_it is serially uncorrelated and the FD estimator is more efficient. We can also test directly whether the differenced errors (∆u_it) are serially uncorrelated. If the null hypothesis of no serial correlation is rejected and there is an evidence of substantial negative autocorrelation in the differenced errors, the FE estimator is considered to perform better (Drukker, 2003).

Appendix C: Practical Applications of the Theoretical Model

Based on the theoretical background regarding the econometric panel data analysis offered in Section 5.3 and Appendix B we estimate our model with its several specifications using the first differences and fixed effects estimation methods. Using the Stata software, we test the assumptions that have to be fulfilled for obtaining a reliable slope estimate for the independent variable along with its standard deviation. The slope estimate is necessary for measuring the partial effect of the independent variable on the dependent variable. Moreover, the Stata output includes p-values for test statistics (which are helpful while testing hypotheses, recognizing statistical significance etc.) and the value of R-squared as well. The R-squared, a goodness-of-fit measure, denotes the proportion of the sample variation in the dependent variable explained by the independent variable.

In the fixed effects regression, we obtain three distinct values of R-squared. Nevertheless, we often do not have to focus on all of them. The first is called the overall R2 and is interpreted as the usual R-squared from the regression of the dependent variable on the explanatory variable. The second one is called the between R2 obtained from the regression of time-demeaned data which consists in collapsing the data and removing the time component by taking the means of our variables for each panel unit individually. It implies the between R2 measures the variation between the individual cross-sectional units. However, since we are interested in a good amount of within information (the variation within one individual over time) that can be exploited by the FE estimator, we rather focus on the value of the within R2 offering the goodness-of-fit measure for individual mean de-trended data taking no account of all the between information in the data.

C.1 Electricity Price and Renewable Energy

For the first model specification we estimate the equation:

illustration not visible in this excerpt

with i = 2,3 … 14 denoting the 13 European countries; t = 2010, 2011, 2012, 2013 stands for the time period; d2011, d2012, d2013 are yearly dummy variables; a_i is the unobserved effect; and u_it is the idiosyncratic error. The equation (5.7) is called log-log model specification since the natural logarithm transformed values of y are being regressed on natural logarithm transformed values of x. The output of the log-log model regression is interpreted as the percentage change in the value of the dependent variable caused by 1% change in the value of the explanatory variable.

First Differences

While using the first difference regression in Stata, the assumptions FD.1 through FD.6 have to be verified and fulfilled for us to obtain unbiased and consistent OLS estimator and valid test statistics (see Section 5.3.1). The first assumption is fulfilled since the log transformation ensures the desired linearity in parameters. The second and third assumptions can be verified as well due to the way we have collected the data set (see Section 5.2.1) and since the value of elfromRE changes over time. Moreover, if there is found a perfect collinearity while running the regression, Stata omits the problematic variable and states the fact to inform us. The last three assumptions will be inspected after running the first difference regression and obtaining the parameters’ estimates for the following equation:

illustration not visible in this excerpt

According to Stata output (using commands .predict res, r and .summ res, d), the expected value of the idiosyncratic errors from the regression equation (5.8) is [Abbildung in dieser Leseprobe nicht enthalten]= .00001 which is really close to zero. Hence we consider the fourth FD assumption to be verified. Next, we test for heteroskedasticity using Breusch-Pagan Lagrange multiplier test (obtained by Stata command .bpagan lnelfromRE d2011 d2012 d2013). The Breusch-Pagan Chi-squared statistics yields χ^2= 4.937 with p-value = .1764. Hence there is not enough evidence of heteroskedasticity as we cannot reject the null hypothesis of homoskedasticity at 5% or even 10% significance level (.10 < .1764). Finally, we have to verify the last FD assumption that there is no serial correlation between the differences in the idiosyncratic errors conditional on all explanatory variables in the model. We use the Wooldridge test for autocorrelation in panel data models (Stata command .xtserial lnelprice lnelfromRE d2011 d2012 d2013). The F statistics yields F= 4.389 with p-value = .0581. Thus we do not reject null hypothesis of no autocorrelation at 5% significance level and there is not enough evidence of serial correlation between 〖∆u〗_it .

Fixed Effects

The other method of obtaining the estimate of β_1 from the equation (5.7) is the fixed effects (or within) transformation. Before we estimate the model using the Stata software we again have to verify the assumptions needed for acquiring an unbiased and consistent OLS estimator. The first three assumptions FE.1 through FE.3 (see Section 5.3.2) are fulfilled as well as the FD.1 through FD.3 since we estimate the same model specification using the same data set as in the previous case. However, the strict exogeneity assumption (FE.4) has to be tested in a different way than in the first difference estimation. First, we specify the equation (5.7) as:

illustration not visible in this excerpt

where w_(i,t+1) is a subset of the explanatory variables of the model in the time (t+1), in our case it is the variable [Abbildung in dieser Leseprobe nicht enthalten], for t = 2010, 2011, 2012. According to Wooldridge (2002), under strict exogeneity, the parameter π_(1 )= 0. While estimating the equation (5.9) in Stata, we obtained the expected value of π ̂_(1 )= .0016 with the p-value equal to .210, hence the null hypothesis H_0: π ̂_(1 )=0 cannot be rejected at 5% (or even 20%) significance level and we consider the FE.4 assumption to be verified. Finally, in order to be sure that the FE estimator is unbiased and consistent, the last two assumptions of the fixed effects estimation, FE.5 and FE.6, have to be fulfilled as well. We verify them by using the Breusch-Pagan test and Wooldridge test, respectively, as well as in the case of the FD estimation and neither serial correlation of the idiosyncratic errors nor heteroskedasticity is found in the model.

C.2 CO2 Emissions and Renewable Energy

In Section 5.4.2, we use the same approach as in Section 5.4.1. Our second model specification is based on the estimation of the following equation:

illustration not visible in this excerpt

where, as well as in the model equation (5.7), i = 2,3 … 14 denotes the 13 European countries according to their assigned id numbers; t = 2010, 2011, 2012, 2013 is the time dimension of the panel data set; a_i is the fixed effect; and u_it is the idiosyncratic error. For the description of the variables CO2 and REcons, see Section 5.2.3. In comparison to the model equation (5.7), the time dummy variables d2011, d2012, d2013 are excluded from the equation (5.10) since they showed to be very statistically insignificant in this model regression and the results fit better without including them.

First Differences

Before we use the first difference regression in Stata, we have to verify the six FD assumptions needed for acquiring the unbiased and consistent estimator and valid test statistics (see Section C.1). The first three assumptions, i.e. FD.1 through FD.3, are verified directly by considering the format of the model equation, the way the data set has been collected and the fact that we have a model with a single explanatory variable hence there cannot be any linear relationship among the explanatory variables (FD.3).

The assumption of strict exogeneity in the explanatory variables, FD.4, can be tested the same way as in Section 5.4.1. We run the FD regression and obtain the parameters’ estimates for the following equation:

illustration not visible in this excerpt

Then we use the commands .predict resid, r and .summ resid, d in Stata and look at the expected value of the idiosyncratic errors from the equation (5.11) which is approximately equal to zero [Abbildung in dieser Leseprobe nicht enthalten]. Thus, the FD.4 assumption is also considered to be fulfilled. The last two assumptions, FD.5 and FD.6, are tested by the Breusch-Pagan test and Wooldridge test, respectively (see Section C.1 in this appendix for more information). The Breusch-Pagan Chi-squared statistics yields χ^2= 3.637 with p-value = .0565 and the Wooldridge F statistics yields F= 1.784 with p-value = .2064. Hence there is not enough evidence of either heteroskedasticity or serial correlation between the differences in the idiosyncratic errors as we cannot reject the null hypotheses of homoskedasticity and no autocorrelation, respectively, at 5% significance level.

Fixed Effects

As well as in Section C.1, we also use the fixed effects (or within) transformation to obtain the estimate of β_1 from the equation (5.10) and then compare the results with the FD estimation. Similarly to the previous cases, the assumptions needed for us to acquire an unbiased and consistent OLS estimator have to be verified in the first place. The assumptions FE.1 through FE.3 (see Section C.1) are fulfilled as well as the FD.1 through FD.3 as we estimate the same model equation (5.10) with the same data set in both cases. To verify the assumption FE.4, we specify the equation (5.10) as:

illustration not visible in this excerpt

where w_(i,t+1) is a subset of the 〖REcons〗_(i,t+1), for t = 2010, 2011, 2012. According to Wooldridge (2002), under strict exogeneity, the parameter π_(1 )has to be equal to 0. By using Stata, we obtained the expected value π ̂_(1 )= .008 with the p-value equal to .678. Thus, the null hypothesis H_0: π ̂_(1 )=0 cannot be rejected at 5% significance level and we consider the FE.4 assumption to be fulfilled. To verify the last two assumptions, FE.5 and FE.6, we once more use the Breusch-Pagan test and Wooldridge test, respectively, as well as in the case of the FD estimation. Since neither serial correlation of the idiosyncratic errors nor heteroskedasticity is found, we can proceed to the regression results assuming that the FE estimator is the best linear unbiased estimator.

Acronyms

illustration not visible in this excerpt

Bibliography

Drukker D. M. (2003): “Testing for Serial Correlation in Linear Panel-Data Models.” The Stata Journal 3(2): pp. 168 – 177.

Enerdata (2014): “Global Energy Statistical Yearbook 2014.” Web, 9 February 2015.

European Commission (2011): “Technical Assessment of the Renewable Action Plans.” Publications Office of the European Union. ISBN 978-92-79-21049-5.

European Commission (2013): “Guidance on the Use of Renewable Energy Cooperation Mechanism.” Communication from the Commission.

European Commission (2013): “The EU Emissions Trading System (EU ETS).” European Union Publications Office. ISBN 978-92-79-32962-3. (Web: http://ec.europa.eu/clima/policies/ets/index_en.htm, 7 March 2015)

European Commission (2014): “Energy Prices and Costs in Europe.” Communication from the Commission.

European Commission (2015): “Causes of Climate Change.” Web, 15 March 2015.

European Parliament and European Council (2009): “Directive 2009/28/EC.”

Eurostat (2014): “Energy Price Statistics.” Web, 10 February 2015.

Eurostat (2014): “Energy from Renewable Sources.” Web, 10 February 2015.

Eurostat (2014): “Share of Renewable Energy in Gross Final Energy Consumption.” Web, 20 February 2015.

Fernández P. F., E. V. Ortiz & J. X. Bernat (2013): “The Deployment of Electricity Generation from Renewable Energies in Germany and Spain: A comparative Analysis Based on a Simle Model.” Energy Policy 57: pp. 552 – 562.

Gerbelová H., F. Amorim, A. Pina, Ch. Ioakimidis, P. Ferrao & M. Melo (2014): “Potencial of CO2 Taxes as A Policy Measure Towards Low-Carbon Portuguese Electricity Sector by 2050.” Energy Journal 69: pp. 113 – 119.

González P. del R., G. S. de Miera & I. Vizcaíno (2008): “Analysing the Impact of Renewable Electricity Support Schemes on Power Prices: The case of Wind Electricity in Spain.” Energy Policy 36(9): pp. 3345 – 3359.

Götz B., A. Voss, M. Blesl & U. Fahl (2012): “Comparing Different Types of Support Systems for Renewable Electricity.” Institute of Energy Economics and the Rational Use of Energy (IER), University of Stuttgart.

International Energy Agency (2012): “Energy Technology Perspectives 2012: Pathways to a Clean Energy System.” IEA: pp. 2 – 7.

Moore M. R., G. McD. Lewis & D. J. Cepela (2010): “Markets for Renewable Energy and Pollution Emissions: Environmental Claims, Emission-Reduction Accounting, and Product Decoupling.” Energy Policy 38(10): pp. 5956 – 5966.

Moreno B. & A. J. López (2011): “The Impact of Renewable Energies and Electric Market Liberalisation on Electrical Prices in the European Union: An Econometric Panel Data Model.” Department of Applied Economics (University of Oviedo): pp. 27 – 31. ISBN 978-1-61208-138-0.

Nicolosi M. & M. Feursch (2010): “Implications of the European Renewables Directive on RES-E Support Scheme Designs and its Impact on the Conventional Power Markets.” International Association for Energy Economics: pp. 25 – 29.

Özbugday F. C. & B. C. Erbas (2015): “How Effective Are Energy Efficiency and Renewable Energy in Curbing CO2 Emissions in the Long Run? A Heterogenous Panel Data Analysis.” Energy Journal 82: pp. 734 – 745.

Paraschiv F., D. Erni & R. Pietsch (2014): “The Impact of Renewable Energies on EEX Day-Ahead Electricity Prices.” Energy Policy 73: pp. 196 – 210.

Renewable Energy Policy Network (2014): “Renewables 2014: Global Status Report.” REN21: pp. 4 – 12. ISBN 978-3-9815934-1-9.

RWE Corporate Website. “How the Electricity Price Is Determined.” Web, 27 February 2015.

Shafiei S. & R. A. Salim (2014): “Non-Renewable and Renewable Energy Consumption and CO2 Emissions in OECD Countries: A Comparative Analysis.” Energy Policy 66: pp. 547 – 556.

Wooldridge J. M. (2002): “Econometric Analysis of Cross Section and Panel Data.” Massachusetts Institute of Technology: pp. 280 – 310. ISBN 0-262-23219-7.

Wooldridge J. M. (2009): “Introductory Econometrics: A Modern Approach.” Michigan State University: pp. 443 – 505. ISBN 978-0-324-58162-1.

World Energy Council (2013): “World Energy Resources 2013 Survey.” World Energy Council: pp. 5 – 23. ISBN 978-0-946121-29-8.

World Nuclear Association (2011): “Comparison of Lifecycle Greenhouse Gas Emissions of Various Electricity Generation Sources.” WNA: pp. 2 – 8.

Würzburg K., X. Labandeira & P. Linares (2013): “Renewable Generation and Electricity Prices: Taking Stock and New Evidence for Germany and Austria.” Energy Economics 40(1): pp. S159 – S171.

77 of 77 pages

Details

Title
The Impact of Renewable Energy on the EU Electricity Prices and CO2 Emissions
College
Charles University in Prague  (Faculty of Social Sciences)
Course
Institute of Economic Studies
Grade
2.0
Author
Year
2015
Pages
77
Catalog Number
V314216
ISBN (Book)
9783668134942
File size
1210 KB
Language
English
Tags
impact, renewable, energy, electricity, prices, emissions
Quote paper
Marek Čech (Author), 2015, The Impact of Renewable Energy on the EU Electricity Prices and CO2 Emissions, Munich, GRIN Verlag, https://www.grin.com/document/314216

Comments

  • No comments yet.
Read the ebook
Title: The Impact of Renewable Energy on the EU Electricity Prices and CO2 Emissions


Upload papers

Your term paper / thesis:

- Publication as eBook and book
- High royalties for the sales
- Completely free - with ISBN
- It only takes five minutes
- Every paper finds readers

Publish now - it's free