An outlook to certification of hydrogen
A discrete choice experiment about the willingness to pay for hydrogen in
heating contracts
Master Thesis Economics University of Groningen
Faculty of Economics
Name student and student number: Mark Maljaars, S2746565
Student email address: M.Maljaars@student.rug.nl
Name supervisor University of Groningen: M. Mulder
Name supervisor GasTerra: G. Martinus
Date Thesis: 9 January 2020
Abstract
Hydrogen is increasingly considered as an indispensable link in the global energy transition. A general market for hydrogen is in its early phase, giving rise to multiple issues shaping this market. This study discusses the options of certification of the hydrogen market and investigates consumers preferences for hydrogen. I recommend to create a public European certification authority, which issues all hydrogen certificates in Europe. Furthermore this study finds that the certification market of hydrogen should adopt a book and claim system. A discrete choice experiment is applied to find the willingness to pay for hydrogen. Two types of hydrogen are investigated for heating houses: hydrogen based on electricity and hydrogen based on natural gas. The results show a positive willingness to pay of €8,30 (9,2%) on their monthly bill for hydrogen based on electricity and an increase of €5,29 (5,9%) for hydrogen based on natural gas. The findings further suggest that consumers are willing to pay 1,2% more per month for hydrogen based on electricity imported from outside Europe. In terms of CO2 reduction, a
willingness to pay of €130 per ton CO2 is identified.
1
1. Introduction
At the historic Paris Agreement on climate change, the world recognized the urgency of climate change. All parties involved agreed to limit average global warming to a maximum of 2° Celsius, with the aim not to exceed 1,5° Celsius. (UNFCCC, 2017). Greenhouse gas emissions need to be reduced in order to achieve this goal (Hanley et al., 2018). One of the main components of achieving the goal of the Paris Agreement is to increase the share of renewable energy sources at the expense of conventional energy sources. Hulshof et al. (2019) state that so far, governments used their traditional tools (subsidies and taxes) to promote the usage of renewable energy. In addition to this, certification schemes have been implemented to stimulate renewable energy. The aim of these certification schemes is to overcome the information asymmetry that is present in the energy market and eventually may lead to adverse selection of consumers as they are not able to identify whether the energy consumed has its origin in the renewable or the non-renewable sector.
2
supply; decarbonizing sectors that previously have been hard to decarbonize (i.e. (long distance) transport and industry) (Hinicio and Waterstofnet, 2018; IEA, 2019).
In combination with renewable energy sources like wind and solar energy, hydrogen allows for the gradual replacement of fossil fuels (Balat and Kirtay, 2010; Hydrogen Council, 2017; Van Wijk and Hellinga, 2018). There are two different ways to generate hydrogen: utilizing electrolysis or through Steam Methane Reforming (SMR). Electrolysis requires pure water and electricity, resulting in hydrogen and oxygen, whilst the steam-methane method uses natural gas to produce hydrogen. The latter produces hydrogen and carbon dioxide. If this carbon dioxide (CO2) is emitted, one speaks about grey hydrogen. However, if the CO2 is captured and
stored, known as Carbon Capture and Storage (CCS), the hydrogen produced becomes CO2
-neutral. This CO2-neutral hydrogen is called blue hydrogen. The captured CO2 can be stored in
depleted gas fields or salt caverns (Cappellen et al, 2018). In the process of electrolysis, one can use renewable energy as the electricity input. Doing so prevents any form of CO2 to be
emitted in the process. Therefore the use of renewable energy for electrolysis enables the hydrogen to be labeled green hydrogen (Mulder et al., 2019).
3
this information asymmetry. This leads to the question of how to implement certification on the hydrogen market. To answer this question, multiple certification schemes are evaluated. The elaboration on certification schemes for the hydrogen market will only be valuable if the market of hydrogen will evolve. Therefore it is crucial that there is a demand for hydrogen in general. Due to the versatile properties of hydrogen, it can help to tackle the challenges in the energy transition. The versatility of hydrogen is characterized by its ability to be produced, transported and stored in several ways (IEA, 2019). This on its own, is however not a guarantee that consumers will embrace hydrogen. Therefore, an equally or even more important aspect is to investigate the willingness of consumers to adapt to the usage of hydrogen. Hence, this paper will elaborate on the preferences of consumers for hydrogen. Consumers reveal their preferences by their willingness to pay (WTP) for hydrogen. This results in the following research question: what is the WTP for the different types of hydrogen? To answer this, a discrete choice experiment is conducted in which consumers were asked which heating contract they would pick for heating their houses. In addition to this, the WTP of CO2 emissions for
heating contracts will be investigated.
The problems addressed in this paper are twofold: on the one hand I will qualitatively investigate the properties and characteristics of the hydrogen market, with the goal to elaborate on the possibilities for certification on the hydrogen market. However, the certification market of hydrogen will only arise if there is a positive WTP for hydrogen. Therefore, the second part will investigate the WTP for hydrogen.. Where other studies have investigated the WTP for green electricity (Sundt and Rehdanz, 2015; Roe et al., 2001; Nomura and Akai, 2004) and green gas (Borchers et al., 2007) , this paper innovates as it adds a WTP study for hydrogen to the existing literature.
4
2. Literature review
2.1 Design of green energy certificates 2.1.1 Certification schemes
Mol and Oosterveer (2015) discuss several methods of tracing sustainability, including the two approaches experienced in the electricity and gas market. Some of these methods are applied in different industries like the agro-food value chain. The different sustainability tracing models analyzed are identity preserved or tracking and tracing, segregation, mass balance and lastly, the book and claim model. An overview of this is provided in table 1.
Tabel 1 - Certification Schemes
Certification scheme Description
Identity preserved / tracking and tracing Certified product is from the producer onwards kept apart throughout the supply chain at all stages
Segregation Certified product is mixed with other producer’s
certified products; the product is not traceable back to the producer
Mass Balance Certified product is mixed with non-certified
products during transportation, but monitored administratively
Book and Claim Certified product is mixed with non-certified
products during transportation. A separate market trades the certificates; not traceable.
5
preserved/tracking and tracing method is the end of the value chain as consumers are not able to identify single producers of the product, since the product is mixed with the same product of other certified producers. An advantage of this method is a lower cost base resulting from higher economies of scale and more competition. The third system is the system of mass balancing, in which the product is traceable throughout the entire value chain, ensuring that the certified amount produced by the producer equals the certified amount at the end of the value chain. This system allows the certified product to mix with non-certified products throughout the value chain. Consequently, the end product is likely to be a combination of certified and non-certified products. Lastly, the book and claim model allows producers to register their sustainable product in a central registry at a trading platform. The producer receives a tradeable certificate for its product and is able to sell this certificate on the global market. Final manufacturers should buy the certificates if they want to sell certified products to consumers. The price of a certificate in a book and claim system is determined by supply and demand of the certificates.
The first two models require the certified product to be physically isolated from non-renewable products. Mol and Oosterveer (2015) argue that the these models are preferred when the certified product is recognizable for individual consumers and it impacts the consumer identity. On the other hand, if consumers are not able to identify the sustainable properties, they state that the mass balancing or book and claim models are preferred. In addition, if consumers cannot distinguish the difference in quality of the certified or non-certified product, mass balancing or book and claim models are favored. Moreover, Staaij et al. (2012) argue that the costs of physical segregation are higher than mass balance and book and claim method. Therefore I will solely investigate the mass balancing and the book and claim method in the renewable energy sector.
6
Moraga Gonzáles et al. (2019) notices that under the book and claim system, there is no connection between certificates and the physical trade of the green electricity. Therefore the volume of the certificates is allowed to exceed the capacity of the power interconnection. To illustrate this, they highlight the example of Iceland: Iceland exports green certificates to the continent, whilst there is no power interconnection between them. Contrary to the book and claim system, the certificates under the mass balance approach are connected to the physical transfer of the gas. Returning to the example of Iceland, there is no possibility of transferring green gas from Island to the continent, as exporting certificates are only allowed if a physical connection exists.
Ecofys (2013) evaluates the mass balance approach in the green gas market against its alternatives (book and claim approach and physical segregation) and finds that a central registry decreases the risk of fraud and error, especially the risk of double selling or claiming. Investment costs for a mass balance approach in the green gas market are already present, making switching costs to the book and claim approach high. Mass balance requires the whole supply chain to be audited, whilst the book and claim approach only requires the generation and selling point to be audited. This results in higher supply chain costs for the mass balance approach compared to the book and claim approach. The study of Ecofys concluded that the mass balance approach should remain in use for the green gas market. The main reasons for this are: preventing confusion on the market; no fundamental complaints about the mass balance; the high investments made for the mass balance approach are already made and the balance between the effectiveness and the administrative burden is a fair compromise.
Based on the study of Ecofys (2013), one can argue that it is crucial to implement the most desirable certification system from the beginning, as switching costs are in general too high. The main advantage of the mass balance approach is the higher degree of transparency compared to the book and claim approach, as there is evidence that at point A of the gas network the gas is inserted, whilst at point B the gas is extracted. The book and claim system does not offer such detailed information regarding the inserted gas.
2.1.2 Willingness to pay renewable energy
7
consumers for different types of sustainability in the energy market. Comparable markets like the electricity and the gas market show that consumers are willing to pay a premium for more environmental friendly options (Roe et al., 2001; Borchers et al., 2007) . Oerlemans et al. (2016) state that for green electricity, consumers pay a premium to cover the additional production expenses of the renewable energy source. In addition, Sundt and Rehdanz (2015) find that people are willing to pay a premium for green energy. An interesting finding in the differences in WTP is that the price of GO for European wind is substantially lower than Dutch wind (Afman and Wielders, 2016), hinting that within countries preferences exist favoring ‘home made’ green electricity. As a result of a positive WTP and the ability to trade certificates, Moraga Gonzáles et al. (2019) argue that subsidies do not need to be as large compared to a situation where the former is not present. This is because the positive WTP generates an additional stream of revenues for the producers of renewable energy sources. The WTP for green certificates depends on the wholesale price of electricity (or gas) as well. Both the price for the certificate and the general price for the commodity determine the price for green certificates. As a result, a price increase in the commodity has a negative effect on the demand for green certificates (Hulshof et al., 2019)
2.1.3 Certification on the electricity market
8
reliable, accurate and fraud-resistant. The requirements of the 2009/28/EC have to be implemented by all member states. The directive further states that the GO should specify:
• The energy source from which the energy was produced and its start and end dates of production;
• Whether it relates to electricity or cooling/heating;
• Specifics of the installation where the energy was produced (identity, location, type and capacity of the installation);
• Investment support for the installation and to what extent the energy unit has benefited from a national support scheme;
• The date on which the installation became operational
• A unique identification number, the country of origin and the date.
In their study about certificate systems, Certifhy (2015b) finds that of all countries investigated (Germany, Netherlands, UK, France, Switzerland, Poland and Denmark), Germany is the only one that uses the mass balancing method on a national scale for green gas, whilst the other countries use the book and claim method. For liquid and gaseous biofuels, the fuel quality directive (2009/30/EC) and the renewable energy directive (2009/28/EC) only accept the mass balance method for cross border trade. As a result, renewable methane GOs are also traded internationally using the mass balancing method, whilst on a national level they use the book and claim system. The differences in the applied GO systems between countries are a result of the implementation phase of GOs, as consistency between countries was not required. In the electricity market, one recognized that after the implementation of the 2009/28/EC directive, cross border transfers lag behind. To solve this, the AIB founded the European Energy Certificate System (EECS). The EECS facilitates the transfer of certificates between member states. They implemented an electronic hub to issue, transfer and cancel the several types of GOs.
9
is that GOs should cover all production routes, including export and import with third countries. Next to that, the certificate should be open for different applications for hydrogen. For now, the industry is the main user of hydrogen, but this might change in the future.
An important feature is to separate an informational part and a qualifications part for the hydrogen market, as the definitions for green and/or blue hydrogen might evolve over time. Therefore Certifhy (2015c) argues that the informational part should be factual, whilst the qualifications part of the GO possibly changes over time as a result of policy developments. Such a change applies for instance for the definition or threshold of green and low-carbon, as this threshold could increase over time. Therefore it is necessary to transparently review the existing requirements and certification system as a whole on a regular basis. Lastly, Certifhy (2015c) recommends that, to increase trust of consumers in GOs, a harmonized GO system for all member states is preferred. This enables the implementation of standard rules for conversion, proper bookkeeping in all countries and the prevention of double counting. This will enlarge consumers trust in the GOs.
The 2009/28/EC directive does not require to mention the amount of greenhouse gasses emitted in the production process. This may however be important information to consumers, as this gives an additional opportunity for consumers to evaluate the certificate. There are two common approaches to mention greenhouse gasses on GOs. On the one hand, one could mention the amount of CO2 emitted in the production process. For wind and solar energy, these emissions
would be negligible. On the other hand, one could also ask for a life cycle analysis in which the complete process (i.e. for wind energy, building wind turbines is also included) is evaluated. Renewable electricity producers are in a disadvantageous position if the greenhouse gasses are evaluated with a life cycle analysis, as this provides consumers information that renewables like sun and wind also emitted greenhouse gasses. Certifhy (2019) decided to not use the life cycle analysis, but solely focus on the emissions of the production process. Consequently, the emissions from wind, hydro and photovoltaic are considered zero.
10
implementation of certificates. Another finding is that a common international standard positively affects market volumes, indicating the importance of an international standard. Since there is some degree of autonomy for design and implementation of the GOs, there are still differences between member states in terms of GOs. Hulshof et al. (2019) notice that differences exist in the issuing body of certificates. Some countries privatized its issuing bodies, whilst other countries have public issuing bodies. They further find that private certifiers experience lower market volumes. The authors assume that this is a result of higher fees charged by private certifiers. Next to that, Moraga Gonzáles et al. (2019) argue that consumers may have less trust in private certifiers compared to public certifiers. More trust leads to a more prominent role of certificates, therefore Moraga Gonzáles et al. (2019) advise to increase the role of public authorities in the process of certification.
2.1.4 Book and claim system vs mass balance approach for the hydrogen market
In terms of market liquidity, one favors the book and claim system for the hydrogen market. This enables significantly larger amounts of certificates to be traded. There is no requirement regarding the requirement of physical trade of the consumed energy and certificate. The mass balance in contrast, does face this restriction as discussed earlier. Furthermore, the mass balance system ensures that all stages in the value chain are audited. Consequently, one argues that there is no possibility of double counting in the mass balance system. In a conversation with Vertogas (the issuing body of Dutch green gas certificates) it has been indicated that the main advantage of the mass balance approach is to provide the customer with information.
11
The mass balance system allows certificates only to be transferred to places where there is a physical connection to the production location. This imposes a restriction on the mass balance system, resulting in lower market liquidity with the mass balance system. For green gas, this restriction is largely overcome by the extensive network that is present. Both the gas and the electricity network are widespread in Europe. The current gas network covers nearly all countries in Europe. Therefore one concludes that imposing a mass balance system does not necessarily cause restrictions on the certification market, since the physical requirements are met in the natural gas market. Therefore one would argue that green gas is tradeable throughout Europe. There is however another obstacle that prevents this from happening. Consultations with employees of GasTerra B.V. indicated that there are different quality requirements among Europe for green gas. This differs per country, leading to minimal cross border trade of green gas and consequently green gas certificates. This once more indicates the importance of a harmonized European system.
Contrary to the electricity and the green gas market, the hydrogen market is still in its early phase. This means that most aspects of this market are still in development. This also applies to the certification of hydrogen. Therefore the possibility arises to shape it as desired. In contrast to the electricity and gas markets, the hydrogen market does not yet have access to a widespread network of pipelines or grids. This imposes a large restriction on the tradability of hydrogen in general and consequently also its certificates under the mass balance approach. Certificates would be traded on a very local scale, as the degree of hydrogen infrastructure is currently limited. To overcome this, inserting hydrogen in the current gas network is extensively researched (Hydrogen Europe, 2019; Gasunie, 2018; Detz et al., 2019). This research finds that it is possible to inject hydrogen in the current natural gas grid. However, as a consequence of legal and technical reasons, this is limited to a certain percentage.
12
On the other hand, if one would use the book and claim system, the tradability of certificates is not limited to borders of the grid. This gives rise to the question whether hydrogen markets should evolve locally and expand on its own pace. Opponents of this argue that now is the time to scale up hydrogen and it should be done on an international level, leading to spillover effects (IEA, 2019; TKI nieuw gas, 2018).
However, before certification on the hydrogen market will take place, there must be a demand for it. Demand for hydrogen certificates will only be present if there is a positive WTP for hydrogen. As a result of this requirement, this study investigates the WTP for hydrogen.
2.2 Hydrogen market
The goal of the hydrogen sector is to establish a green hydrogen economy, where hydrogen is mainly produced through electrolysis with renewable energy. Currently, many believe that we are just at the beginning of a possible hydrogen economy (Van Wijk and Hellinga, 2018). In line with this, several countries (i.e. Belgium, Korea, the Netherlands) have published a hydrogen roadmap to gain more insight in the role to play for hydrogen (IEA, 2019). Forecasts for this role of hydrogen in 2050 differs from a 6% hydrogen share of total energy consumption (IRENA, 2019) towards estimations of even 18% (Hydrogen Council, 2017). In order to reach this ambitious goal, it is necessary to fasten the implementation and to make substantial investments in the hydrogen sector. This will result in economies of scale, leading to more efficiency and cost reductions (TKI nieuw gas, 2018; IEA, 2019).
As hydrogen has only recently obtained the momentum, there does not yet exist a general hydrogen market. So far, hydrogen is mainly used for industrial usage; Certifhy (2015) finds that this industrial usage represents more than 90 percent of the hydrogen market and is definable in four categories: chemical (63%), refining (30%), metal working (6%) and other industrial use (1%). A major part of the produced hydrogen is used directly by the producer itself. The other part is mainly sold through bilateral contracts. Certifhy (2015) notes that the price of hydrogen is likely determined by local markets and bilateral contracts. As a result, there does not exist a uniform price of hydrogen.
13
continuous flow of renewable energy for the production of green hydrogen. To place this in perspective, if all the hydrogen consumed today would have been green hydrogen, meaning that it is produced through water electrolysis, it would require an energy demand of 3600 TW, which is more than the annual electricity generation in the European Union (IEA, 2019).
From an environmental perspective, one favors green over blue and grey hydrogen. However, green hydrogen is currently not able to compete on cost and in scale to the other types of hydrogen (IEA, 2019; Mulder et al., 2019; Hinicio, 2018). In addition to this, not all reports indicate the future or green hydrogen that bright. Mulder et al. (2019) state that the fierce competition for green electricity as a result of worldwide electrification is a possible pitfall for green hydrogen. In their study, Mulder et al. (2019) argue that green hydrogen will only be competitive once the price of electricity drops to around 20 euro per MWh. However, over the last years, the average price of electricity was around 45 euros per MWh (Mulder et al., 2019). As a result, it is not expected that green hydrogen will be cost competitive in the short term. In addition, Mulder et al. (2019) argue that as a result of the electrification process, electricity demand is expected to rise in the future. Total electricity demand in the Netherlands might even double. Therefore it is not likely that the price of electricity will decrease towards the required levels mentioned above. Blue hydrogen however is a suitable option to fill this gap in the short term. Several authors argue that blue hydrogen may offer the solution as blue hydrogen is the ‘stepping stone’ for green hydrogen (H-vision, 2019; CE Delft, 2018; TKI nieuw gas, 2018; Cappellen et al., 2018). In order to operate effectively, blue hydrogen requires a solid infrastructure and smooth supply chain, which in later stages will also benefit green hydrogen. By embracing blue hydrogen, investments are expected to take place and the infrastructure should be prepared to transport hydrogen. All these actions are beneficial in later stages for green hydrogen as well. Therefore H-Vision (2019) argues that blue hydrogen is an indispensable step towards the ultimate goal of a green hydrogen economy.
cost-14
15
3. Method and study design
3.1 Method
For investigating WTP of consumers, Breidert et al. (2006) argue that there are two methods to analyze the WTP: the revealed preference or the stated preference method. Revealed preference methods are based on actual or simulated price response data, whilst the stated preference method uses surveying techniques. As there is no actual or simulated data available of the hydrogen market (Certifhy, 2015), the revealed preference method is not applicable. Consequently, the stated preference method will be used and data is obtained by conducting a survey in which consumers reveal their preferences. This method can be distinguished in two types: the contingent valuation method (CVM) and choice modelling (CE) (Oerlemans et al., 2016).
3.2 Contingent valuation method versus choice experiment
CVM is a direct survey method, where a hypothetical market is described (Breidert et al., 2006). An increase or decrease in the hypothetical scenario is presented to respondents and they are asked to reveal their maximum WTP or minimum willingness to accept this hypothetical scenario. This is done by offering a change in the quality or quantity of the scenario at a given cost (Mogas et al., 2006).
Mogas et al. (2006) further state that the indirect survey method, the CE, respondents are presented a series of choice sets, each containing usually three or more alternative goods. The various alternatives are differentiated by the value of the attributes and levels. Respondents should indicate their preferred alternative of the choice set. The CVM asks the respondents about a single event, whereas the CE provides the respondents with a choice set where the respondent is asked to choose their preferred option.
16
CVM. Diamond and Hausmann (1994) argue that CVM does not deliver accurate estimates of WTP. In line with this, Hausmann (2012) finds that three major problems exist in CVM. First of all, values will be overstated due to the hypothetical response bias. Secondly, the difference between the willingness to accept and the WTP is too large. Lastly, Hausmann (2012) argues that there is a problem with embedding and scope. This leads Hausmann to conclude that the obtained data from CVM is ‘useless’ and therefore he prefers no number rather than a CVM number. Kling et al. (2012) draw similar conclusions as Hausmann, albeit less convincing concerning the problem concerning embedding and scope. As a consequence, the CVM will not be used in this study.
The indirect survey method, described by Breidert et al. (2006) uses a rating process to estimate a preference structure to ultimately derive the WTP. They further state that the indirect method is preferred as it exhibits higher external and internal validity. Holmes et al. (2017) list several advantages of CE over other valuation methods. The characteristics are in general exogenous and therefore not collinear. Furthermore, the choices consumers face in markets are similar to the ones in the CE, making it relatively easy to answer. In line with previous studies of the willingness to pay, I will use the stated preference model. More specifically, I will conduct a questionnaire and apply the indirect CE approach.
3.3 Choice experiment
17
component and an unexplainable random component. The latter comprises all unidentified factors impacting the choices. The systematic component contains attributes that explain the differences in choice alternatives. (Louviere, 2010; McFadden; 1974; Holmes et al., 2017; Huber and Zwerina, 1996). This is summarized in the following formula:
𝑈𝑖𝑎 = 𝑉𝑖𝑎 + 𝜀𝑖𝑎
Where Uia is the unobservable direct utility and i denotes the respondent and a denotes the or
choice alternative. Via is the systematic component, whilst ε is the random component that
captures individual and alternative-specific factors that influence the utility. The systematic component is observable for the researcher, whilst the random component adds utility that is unobservable for the researcher. A general assumption that is often implemented is that the indirect utility function is a linear function of the chosen attributes (Longo et al., 2007; Holmes et al., 2017).
𝑉𝑖𝑎 = 𝛽0𝑎 + 𝑧𝑖𝑎𝛽 + 𝜀𝑖𝑎
In this linear function, Via is still the systematic component observed by the researcher. zia is a
row vector of the attributes. In this case zia is a 1x4 vector as this experimental design consists
of four attributes. Coefficients are shown in β, which is a column vector of weights associated with the attributes. Lastly, εia is the random error term or unobservable part of the utility person
i has for alternative a.
In line with McFadden (1974), it is assumed that a respondent chooses the alternative that maximizes the respondent’s utility. Therefore, in each of the choice sets the respondents face, a respondent will choose the alternative that provides the highest indirect utility (Longo et al., 2007):
𝑃𝑖𝑎 = Pr (𝑉𝑖𝑎 > 𝑉𝑖1, 𝑉𝑖𝑎 > 𝑉𝑖2, … , 𝑉𝑖𝑎 > 𝑉𝑖𝐴 = Pr(𝑉𝑖𝑎 > 𝑉𝑖𝑗) ∀𝑗 ≠ 𝑎
Where Pia indicates the probability that option a is selected by individual i. Including the error
terms:
𝑃𝑖𝑎 = Pr [(𝑉𝑖𝑎+ 𝜀𝑖𝑎) > (𝑉𝑖𝑗+ 𝜀𝑖𝑗)] ∀𝑗 ≠ 𝑎
Rewriting this:
18
Therefore, one has to make assumptions about the error term. Holmes et al. (2017) and Louviere (2010) argue that researchers cannot observe the full utility of consumers, as there is an unobservable utility in the mind of consumers. This is captured in the error term. Therefore one has to explicitly define the error term. Dependent on how the error term is treated, one automatically ends up with the econometric model to use. Following the work of Bergmann et al. (2006), who also conducted a choice experiment with renewable energy, I assume that the error term of the RUT model is independently and identically distributed (IDD) and follows type 1 extreme value Gumbel distribution. This results in use of the multinomial logit model (MNL) (Holmes et al., 2017).
Therefore the multinomial logit model applies here, leading to the probability that respondent i chooses alternative a out of A alternatives is:
𝑃𝑟𝑖𝑎 = exp (µVia) ∑𝐴𝑗=1exp(µ𝑉𝑖𝑚)
Where µ is a scale parameter and it is the inverse of the standard deviation of the disturbance. In standard multinomial logit models, the scale parameter (µ) cannot separately be identified and therefore is set to unity (Lanscar et al., 2017).
To find the WTP of respondents, this paper follows the generally accepted approach in which one uses the coefficients of the attributes to estimate the tradeoffs between the attributes. Since there is a price attribute included, one is able to compute the WTP for an increase or decrease of attribute levels (Amador et al., 2013; Bergmann et al., 2006; Borchers et al., 2007). Therefore one calculates the WTP as follows:
𝑊𝑇𝑃𝑖 = − (𝛽 𝑛𝑜𝑛 − 𝑚𝑎𝑟𝑘𝑒𝑡 𝑎𝑡𝑡𝑟𝑖𝑏𝑢𝑡𝑒
𝛽 𝑚𝑜𝑛𝑒𝑡𝑎𝑟𝑦 𝑎𝑡𝑡𝑟𝑖𝑏𝑢𝑡𝑒 )
19
4. Experimental design
4.1 Constructing attributes
In CE surveys, respondents are asked to pick one of the options provided to them. These options are described and differ by its attributes. Often, at least one of the options entails a hypothetical situation, indicated by a set of attributes (Longo et al., 2009). This CE investigates the choice of respondents in a situation where their existing heating system was no longer functioning and therefore needs replacement. The main question in constructing the attributes of a choice experiment is regarding the underlying question concerning the incentive of people to choose option A over option B. In this case, why would a respondent choose heating contract A over contract B.
It is preferable to keep the list of attributes as easy as possible, as it is unknown how respondents will react to complex survey questions (Swait and Adamowicz, 2001). Holmes et al. (2017) recommend to hold conversations with managers, scientists and people that typify the population of the sample to give the researcher more insight in understanding the sample group and to maximize the understandability of the survey. This gives information about the credibility of the chosen attributes as well as the degree of understandability and clearness. This will help to determine the attributes used for the survey. After defining the attributes, one should specify the levels.
20
reliability of the parameters. However, the levels should make sense to respondents, leading to a trade-off between statistical preference and practical considerations.
4.2 Selection of attributes and levels
In this CE about heating contracts, respondents are faced with four attributes: (1) the type of gas used for heating purposes, (2) costs per month in euros, (3) CO2 emissions (kg per month)
and lastly (4) origin. This is shown in table 2.
The first attribute captures the degree of exhaustibility of the gas. Natural gas is exhaustible source of energy, whilst hydrogen produced by an electrolyser is not. Hydrogen produced with SMR with CCS on the contrary, is also dependent on natural gas and therefore exhaustible. The second attribute, costs per month, shows the respondent the costs per month for heating its house. Furthermore, this gives the researcher the ability to investigate the WTP for the other attributes in this CE. Its baseline, 90 euros per month, is based on the current average monthly costs for households. The levels of this attribute are based on the baseline plus a premium. As elaborated upon in the literature review, consumers tend to pay a premium for more environmental friendly options (Oerlemans et al., 2016; Sundt and Rehdanz, 2015). This led to our third attribute, the amount of CO2 emissions. This captures the degree of sustainability.
Lastly, the import dependence is captured by adding the attribute origin. Afman and Wielders (2016) find that the price for a GO of Dutch wind is more expensive than that of a European one. This leads to the suspicion of consumers favoring locally produced energy over energy produced in other countries.
Tabel 2 - Household Heating Choice Experiment: Attributes and Levels
Attribute Type of gas Price (euros per month)
CO2 (kg per month)
Origin
Level 1 Natural gas €90 250 Netherlands
Level 2 Hydrogen based
on natural gas
€100 125 Europe
Level 3 Hydrogen based
on electricity
€110 0 Outside Europe
Level 4 €120
21
In each choice question, respondents were provided with two hypothetical heating contracts. In the CE, it was assumed that respondents needed to replace their heating boiler. Therefore respondents were forced to pick a new heating boiler. If a ‘no-choice’ option would have been included, respondents would actually choose to have no heating in their house at all. Therefore a ‘no choice’ option was not included in this research. Following the reasoning of Louviere et al. (2010), a significant increase in the number of observations should take place if one wants to include interaction terms between attributes. As a consequence of practical implications, this is outside the scope of this research. Therefore interaction terms between attributes are excluded.
4.3 pretesting of the survey
The initial set of attributes was submitted to multiple experts in both the academic as the business field in order to gain feedback to further improve the set of attributes. Evaluating the several recommendations, a few changes were made to the levels of the attributes, for instance the range of the attribute price was increased. This is in line with the recommendation of Rose and Bliemer (2009), as they suggest that a wider range is preferred to a narrow range for the level range. This will theoretically lead to better parameter estimates, as this leads to lower standard errors of the parameters. One should however be cautious, as the levels should always make sense to the respondents. This leads to a trade-off between the statistical preference and the practical considerations.
22
attributes. However, the jeopardy in practice is that pictures might hint at information that is not related to the attributes of the model. The degree of uncertainty regarding the interpretation of images, as described by Cherchi and Hensher (2015), led to the decision to ignore pictures.
4.4 Survey structure
The respondents are asked ten times to pick one out of two heating contracts. After the ten choices regarding the experiment, respondents were asked some socio-demographic questions. The variables of interest were gender, age, educational level, having children, residence and whether one owns solar panels. These questions aim to have a better understanding of the characteristics of respondents and mainly what these characteristics imply. This enables the researcher to investigate more specifically the specifications of the model for different characteristics amongst respondents.
As noticed by Keane (1997), respondent have no incentive to make the same decision in an experiment compared to the real market. Certain aspects one encounters in a real market are absent in the choice experiments, such as searching costs. Furthermore, in the experiment the respondent have to make its decision solely based on the specific attributes assigned by the researcher. This imposes the risk that in a real market a consumer observes attributes that the researcher did not observe. Therefore the following information has been added at the beginning of the survey:
Please imagine that you need to replace your heating boiler and renew your contract for heating your house. In the survey, you are asked to choose between heating contracts. This implies a choice for a boiler that is identical in every other way than the energy input (gas/hydrogen). Therefore I ask you to base your decision of heating your house only on the heating contract. Each contract consists of the same aspects, only its values will vary for every option.
23
Figure 1 - Example of choice set used in the survey
Ideally, one would have a sample group that represents a targeted population. In this research, one would preferably have a sample group that represents the Netherlands. However, as a result of monetary and time restrictions, this is outside the scope of this study. Therefore it is decided to target the GasTerra population. This is the company at which I am conducting my research. GasTerra is a wholesaler in natural and green gas.
The properties that characterize an efficient choice design are threefold: level balance, orthogonality and minimal overlap (Huber and Zwerina, 1996).
Orthogonality means that each pair of levels appears equally often across all pairs of attributes within the design. This only holds if each attribute has the same amount of levels. In this CE, different amount of levels are present for the attributes. In such a case, an orthogonal design would proportionally show the levels equally often.
24
factorial design ensures that there is level balance and minimal overlap. This combines every level of each attribute with all possible combinations of the other attributes. The advantage of this is that all effects are statistically independent.
Rose and Bliemer (2009) argue that when using linear models, orthogonality of data is important, as this prevents multicollinearity. Secondly, orthogonality minimizes the variances of the parameter estimates. This results in the smallest standard errors which in turn maximizes the t-ratios. (Rose and Bliemer, 2009). A full factorial design consists of all possible combinations of the levels of the attributes. This enables the researcher to estimate both main effects and interaction effects (Mangham et al., 2008). One restriction was imposed on the full factorial design, as it is not possible to heat your house using natural gas in combination with zero CO2 emissions.
The experimental design was created using the software of Sawtooth Software Lighthouse Studio 9.5.3. This enables the researcher to pretest the experimental design and gather information regarding the level balance and overlap. The distribution between the attributes was tested to be equal among the respondents. As mentioned earlier, the number of choice sets was increased from six to ten, causing the statistical efficiency to increase even further. This statistical efficiency is measured by the D-efficiency. The increase in the D-efficiency of the experimental design secured that the design is efficient.
Kjaer et al. (2006) argue that one should produce multiple versions of the questionnaire in a way that the attributes and its levels differ between the multiple versions. This should be done to minimize any potential bias caused by the order of the choice sets and attributes. In line with this reasoning, this questionnaire is randomized for every respondent, meaning that all respondents face a different questionnaire with different combinations of the attributes and its levels. This ensured the usage of full factorial design, leading to the most efficient experimental design possible. The experimental design was statistically tested by means of D-efficiency and was concluded to be efficient.
25
behind this is the severe anti-spam policy within the company. Therefore without providing the degree of safety, one is trained to not to open a ‘strange link’ in an email.
4.5 Hypotheses
The attributes and its levels in combination with the socio-demographic characteristics of respondents, enables the researcher to investigate more specific questions. As described in section 4.1, the first attribute displayed different types of gas. I expect that people favor hydrogen over natural gas (hypothesis 1). The findings of Afman and Wielders (2016) suggest that local renewable energy is more expensive than non-local, leading to the second hypothesis. Next to that, I expect that people are willing to reduce the amount of CO2 emissions and
therefore show a positive WTP for this (hypothesis 3). Blomquist and Whitehead (1998) find a positive relation between education and WTP for renewables, this leads to hypothesis 4. Following a similar way of thinking as Longo et al. (2008), one could argue that respondents recognize the importance of policies that stimulate renewable energy and therefore are willing to pay for the benefits of such policies, since these benefits apply for future generations as well. This reasoning leads to hypothesis 5. Lastly, intuitively people owning solar panels are expected to embrace hydrogen faster compared to people without solar panels. This is displayed in hypothesis 6. Table 3 Shows the hypotheses.
Tabel 3 - Hypotheses
Hypothesis Description
1 Consumers prefer hydrogen over natural gas
2 Consumers prefer locally produced energy sources
3 There is a positive WTP to reduce CO2 emissions
4 Higher educated people have a higher WTP for hydrogen
5 People with children have a higher WTP for hydrogen
26
5. Results
5.1 Completed surveys
An email was sent to all employees of GasTerra with the request to complete the survey. They were provided a link which guided them to the survey. The email has been sent three times in total, within a time period of ten working days. In total, the survey was opened 132 times. Of this, 96 respondents completed the survey, whilst 36 respondents did not finish the survey. These were therefore deleted from the dataset. Reasons for not finishing the survey are assumed to be twofold: on the one hand I assume that the employees of GasTerra were curious about the survey and therefore clicked on the link to have a look at it to make an assessment themselves about the time-indication of the survey. Once the link was opened, the software registered this as a respondent. On the other hand, I have received feedback from employees of GasTerra that respondents might have been discouraged by their first choice sets as a result of a dominated choice. This is however an inevitable consequence of the randomization of the survey among respondents.
Table 4 - Descriptive statistics
Variable Sample group
Number of respondents 96
> 45 years old 45,8%
Male 76,0%
Has university degree 65,6%
Has Children 74,0%
Lives in a city 46,9%
Has solar panels 75,0%
5.2 Descriptive statistics
27
in half, this means that 45,8% is older than 45. Another remark one could make about the descriptive statistics is the level of education. The smallest proportion (7,9%) finished MBO, 25% finished HBO and lastly, 65,6% finished university. There were only two respondents answering the other-option. One of them finished high school and the other would not give this information. Furthermore, 46,9% lives in a city and 75% of all respondents own solar panels. The annual report of GasTerra states that in 2018, there were 152 employees, of which 110 (72%) were male and 42 (28%) females. Therefore, the sample group of this study represents the targeted population well.
5.3 Results of Multinomial logit model
The results of the MNL model are displayed in table 5. For the type of gas, the variable natural
gas is used as baseline, whilst for origin the baseline is the Netherlands. In the first column
(MNL), one observed the output of the MNL model. The coefficient of price is negative and statistically significant, meaning that the higher the price, the less likely to choose that heating contract. Both types of hydrogen as type of gas are statistically significant, which implies heating houses with hydrogen if preferred over the baseline, which is natural gas. This is in favor of the first hypothesis. More specifically, both hydrogen based on electricity and hydrogen based on natural gas are statistically significant at the 1% level. The only difference is the coefficient of hydrogen based on electricity being proportionally larger than the coefficient of hydrogen based on natural gas. This findings do not include CO2 emissions, since
this is a variable on its own. Therefore, these results indicate that there is a positive emotion for hydrogen.
28
statistical significance. Therefore, the second hypothesis stating that consumers prefer locally produced energy sources is true for energy sources from outside Europe. However, the this is not the case for energy sources originating in Europe.
Next to that, the results support the idea that consumers prefer less CO2 emissions. All models
show evidence for this at the 1% level. Therefore one can conclude that the higher CO2
emittance, the less likely a heating contract is picked.
The socio-demographic questions are included in the other columns of table 5. The first column presents the general model, whilst the other columns include the socio-demographic effects. One should however be cautious with interpreting these socio-demographic results, as this only indicates the individual effect. Therefore the socio-demographic models should only be analyzed individually. Furthermore, no statistical test could be found to test for multicollineairty between these models, however it is highly likely that there is multicollinearity between some of the socio-demographic models, i.e. between older than 45 and having children. In terms of model testing, a postestimation test for MNL models is the likelihood-ratio test to test which of the models is the strongest. Unfortunately this test is only applicable in models where the socio-demographic are used as dummies, since this results in different degrees of freedom per model. Since this paper uses subsets including the specific socio-demographic characteristic rather than a dummy variable in the general MNL model, the degrees of freedom between the models do not change and therefore the likelihood ratio test is not applicable. Lastly, the model’s goodness of fit is tested by the Akaike Information Criterion (AIC), which indicates that the general MNL model fits the best as this provides the lowest AIC score.
The take-away of the socio-demographic models in table 5 is the finding that both price and CO2 are the main indicators in all models for picking a heating contract. This is shown by the
results, as both the general MNL model as all socio-demographic subsets present statistical significance at the 1% level both for price as CO2 emittance. Another finding is shows that
29
A significant difference arises if one zooms in on the age of the respondents. An interesting results is displayed in the age group younger than 45 years old, as this group only values price and CO2 emissions in their decision of picking a heating contract. There is clear evidence that
these younger people do not value the type of gas or the origin in their decision for a heating contract. On the contrary, people older than 45 years do value the origin of the type of gas. The place of residence does not influence the decision of which heating contract to pick, as both yield the same results. Therefore we must reject the sixth hypothesis.
A slight difference is noticeable in the significance for hydrogen based on natural gas, since residents of a city show a statistical significance at the 5% level, whereas residents of a town at the 10% level. The last column indicates whether respondents own solar panels, or are planning to buy solar panels. One might argue that this indicates the willingness to adjust to climate-friendly options. As stated in the hypothesis, it was expected that people owning or willing to buy solar panels are more likely to choose for hydrogen as a type of gas. The results support this hypothesis, as there is evidence at the 1% level that owners or potential buyers of solar panels are more likely to pick a contract including hydrogen compared to a contract involving solely natural gas. For people without solar panels, a negative sign is even displayed for hydrogen based on electricity, meaning that a contract including hydrogen based on electricity decreases the likelihood of that contract being picked. There is however no statistical significance to support this claim.
5.4 Willingness to pay
The WTP results are derived from the MNL model and are shown in table 6. The WTP for hydrogen based on electricity is €8,30 per month, whilst the WTP for hydrogen based on natural gas is €5,29 per month. This indicates that people are willing to pay an additional premium for using hydrogen rather than natural gas. Moreover, the data shows that there is a negative WTP for CO2, implying that an increase of 1 kilogram of CO2 decreases the WTP with
€0,13. This is equivalent to €130 per ton CO2. Next to that, there is a negative WTP for gas
30
Table 5 - Multinomial model
Variables MNL Having
children
No children Younger than 45
Older than 45 No university degree
University degree
Lives in town Lives in city Owns solarpanels
Does not own solarpanels Price (EUR) -0.0741*** -0.0791*** -0.0621*** -0.0849*** -0.0673*** -0.0725*** -0.0759*** -0.0707*** -0.0784*** -0.0736*** -0.0797*** (0.00446) (0.00536) (0.00815) (0.00695) (0.00593) (0.00738) (0.00567) (0.00593) (0.00679) (0.00527) (0.00878) Hydrogen Electricity 0.615*** 0.594*** 0.705** 0.235 0.897*** 0.878*** 0.456** 0.524*** 0.729*** 0.849*** -0.039 (0.145) (0.171) (0.275) (0.218) (0.197) (0.240) (0.183) (0.194) (0.220) (0.169) (0.292) Hydrogen Natural gas 0.392*** 0.446*** 0.288 -0.0204 0.715*** 0.203 0.510*** 0.329* 0.465** 0.500*** 0.090 (0.144) (0.169) (0.274) (0.217) (0.194) (0.243) (0.179) (0.194) (0.214) (0.168) (0. 283) EU -0.166 -0.238 0.0156 0.0562 -0.329* -0.492** 0.0582 -0.179 -0.166 -0.107 -0.332 (0.136) (0.160) (0.258) (0.207) (0.182) (0.226) (0.171) (0.183) (0.204) (0.160) (0.264) Outside EU -0.534*** -0.550*** -0.520** -0.266 -0.730*** -0.679*** -0.436** -0.555*** -0.500** -0.488*** -0.650** (0.136) (0.162) (0.256) (0.206) (0.184) (0.229) (0.171) (0.185) (0.203) (0.160) (0.266) CO2 (KG) -0.00990*** -0.0105*** -0.00852*** -0.0117*** -0.00867*** -0.00785*** -0.0113*** -0.00930*** -0.01066*** -0.01101*** -0.00752*** (0.000644) (0.000765) (0.00121) (0.00102) (0.000843) (0.00105) (0.000829) (0.000859) (0.000978) (0.000769) (0.001239) Constant 9.323*** 9.965*** 7.749*** 10.86*** 8.315*** 8.998*** 9.600*** 8.944*** 9.807*** 9.269*** 10.043*** (0.541) (0.651) (0.984) (0.858) (0.709) (0.888) (0.689) (0.719) (0.824) (0.635) (1.090) Observations 1,920 1,420 500 880 1,040 660 1,260 1,020 900 1,440 480 Pseudo R2 0.264 0.283 0.223 0.299 0.247 0.234 0.291 0.241 0.294 0.290 0.224 Log-likelihood -978 -706 -269 -427 -542 -350 -618 -536 -440 -708 -258
The socio-demographic models show that that people older than 45 are willing to pay respectively €13,33 and €10,63 per month for the types of hydrogen, whilst the people younger than 45 have much lower values that are even statistically insignificant. Another interesting finding concerns the level of education, since respondents with a university degree on average have a WTP of €6,01 per month for hydrogen based on electricity, whilst respondents without a university degree are willing to pay €12,11 per month for hydrogen based on electricity. For Hydrogen based on natural gas, consumers with a university degree are willing to pay a statistically significant €6,72 per month. This is in contrast with the non-significant €2,80 per month consumers without a university degree are willing to pay. Lastly, owners of solar panels have on average an WTP of €11,54 per month for hydrogen based on electricity, whereas for hydrogen based on natural gas this is €6,79 per month. This supports the fourth hypothesis. A remark should be made about the impact of having children on the WTP for a reduction of CO2 emittance. The results show that people with children are willing to pay €0,13 per kg less
CO2, whilst people without children are willing to pay €0,01 per kg less CO2. This indicates
that people with children are more concerned about the emissions of CO2 and are willing to pay
to minimalize the emissions. People with children experience a statistically significant positive WTP for both types of hydrogen (€7,51 and €5,64 per month respectively), whereas childless people only experience a statistically significant WTP of €11,35 per month for hydrogen based on electricity. This findings are in favor of hypothesis 5, stating that people with children have a higher WTP. Owners of solar panels experience a higher WTP for a reduction in CO2
Table 6 - Willingness to Pay
VARIABLES MNL Having
children
No Children Younger than 45
Older than 45 No university degree
University degree
Lives in town Lives in city Owns solarpanels
Does not own solarpanels Hydrogen Electricity 8.30*** 7.51*** 11.35** 2.77 13.33*** 12.11*** 6.01** 7.41*** 9.30*** 11.54*** -0.49 (1.96) (2.16) (4.43) (2.57) (2.93) (3.31) (2.41) (2.74) (2.81) (2.30) (3.66) Hydrogen Natural gas 5.29*** 5.64*** 4.64 -0.24 10.63*** 2.80 6.72*** 4.65* 5.93** 6.79*** 1.13 (1.94) (2.14) (4.41) (2.56) (2.88) (3.35) (0.01) (2.74) (2.73) (2.28) (3.55) EU -2.24 -3.01 0.25 0.66 -4.89* -6.79** 0.77 -2.53 -2.12 -1.45 -4.17 (1.84) (2.02) (4.15) (2.44) (2.70) (3.12) (2.25) (2.59) (0.204) (2.17) (3.31) Outside EU -7.21*** -6.95*** -8.37** -3.13 -10.85*** -9.37*** -5.74** -7.85*** -6.38** -6.63*** -8.16** (1.84) (2.05) (4.12) (2.43) (2.73) (3.16) (2.25) (2.62) (2.59) (2.17) (3.34) CO2 (KG) -0.13*** -0.13*** -0.01*** -0.14*** -0.13*** -0.11*** -0.15*** -0.13*** -0.14*** -0.15*** -0.09*** (0.01) (0.01) (0.02) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) (0.02) Constant 125.82*** 125.98*** 124.78*** 127.92*** 123.55*** 124.11*** 126.48*** 126.50*** 125.09*** 125.94*** 126.01*** (7.30) (8.23) (15.85) (10.11) (10.53) (12.25) (9.08) (10.17) (10.46) (8.63) (13.68) Observations 1,920 1,420 500 880 1,040 660 1,260 1,020 900 1,440 480 Pseudo R2 0.264 0.283 0.223 0.299 0.247 0.234 0.291 0.241 0.294 0.290 0.224 Log-likelihood -978 -706 -269 -427 -542 -350 -618 -536 -440 -708 -258
5.4. Result of review of the certificate market
Evaluating the pros and cons for both the mass balance approach and the book and claim system, I recommend implementing a book and claim system for certification in the hydrogen market. At first, because this will benefit the market liquidity of certificates, which is favorable in the early phase of a new certification scheme. The success of the hydrogen market depends on the embracement of it by consumers. Implementing a mass balance system hinders this approach, as operation and consequently certification would only be on very local scale due to its physical connection requirement. The injection of hydrogen in the natural gas grid or conversion to a 100% hydrogen grid overcomes this restriction. However, implementing this takes time and will delay the evolution of the hydrogen market.
The main argument in favor of the mass balance approach concerns the reliability and traceability of the certificate. It is argued that there is less trust in the book and claim system as it is more sensitive for fraud. However, evidence for this statement is lacking in the literature and therefore this does not outweigh the benefits of the book and claim system. The arguments in favor of the mass balance approach as mentioned by Ecofys (2013) are based on an existing mass balance approach in an existing market and therefore this evaluation is not one-on-one applicable for the hydrogen market as this market is still in development and certification is absent. The study of Ecofys (2013) indicates that switching costs are generally substantial and switching between systems would cause confusion on the market. Therefore an uniform approach must be chosen from the beginning.
35
6. Discussion and conclusion
6.1 FindingsThis study reviewed existing certification schemes in similar markets with the goal to make a recommendation for certification on the hydrogen market. Special focus was on certification of renewable electricity market and certification on the green gas market, to have a better understanding about certification markets. However, the certificate market will only prosper once consumers have a positive willing to pay for certificates. Therefore this study was twofold: on the one hand I elaborated upon certification for the hydrogen market, whilst the other part focused on the WTP for hydrogen.
In examining the WTP for hydrogen, this paper used a discrete choice experiment in which respondents were asked to choose their preferred contract for heating their house. Both ways of producing hydrogen, using electricity or using natural gas, displayed a positive WTP. Hydrogen based on electricity shows a WTP of €8,30 per month, whereas hydrogen based on natural gas shows a WTP of €5,29 per month. This supports the findings of Longo et al. (2007) that most studies find a positive WTP for renewable energy. Currently households pay on average €90 per month for heating houses with natural gas. This study finds that consumers are willing to increase their monthly bill for heating their house by 9,2% if hydrogen based on electricity is used as energy source, whilst for hydrogen based on natural gas this is 5,9%. In a survey published by the OECD (n.d.), they investigated the maximum % increase in the annual bill consumers are willing to pay to buy renewable energy. The results showed that for the Netherlands, the mean WTP was 4,9%.
Furthermore, respondents are willing to pay €130 for a reduction of one ton CO2. This is almost
identical to the findings of Alberini et al. (2018) who found a WTP for CO2 emissions avoided
of €133 for Italian respondents and €94 for Czech respondents.
36
to the conclusion that consumers are willing to pay €1,07 or 1,2% extra per month for hydrogen based on electricity imported from North Africa compared to natural gas from the Netherlands. Therefore the results of this study supports the idea of Van Wijk et al. (2019).
Socio-demographic variables are included in the survey to check whether these cause a different WTP for hydrogen. This is of particular interest for the energy transition, as it gives more insight in potential users of hydrogen. Longo et al. (2007) argues that respondents with children have a higher WTP to reduce greenhouse gasses. The findings of this research indicate that respondents with children have a WTP of €130 to reduce one ton of CO2 emissions, whereas
respondents without children only display a WTP of €10 per reduced ton CO2. This is in line
with the results of Longo et al. (2007).
The findings of this paper indicates that there is significant support for hydrogen as a substitute of natural gas. Multiple studies, i.e. IEA (2019) and TKI nieuw gas (2018), argue that the time has come to scale up hydrogen and stressed the importance of the role of hydrogen in the energy transition. This study points out that there is a positive WTP for hydrogen among consumers. However, actual usage of hydrogen among consumers is still very limited as a result of a lack of investments. Therefore in terms of policy recommendation, I suggest to increase subsidies and investments that will benefit the growth of the hydrogen market.
37
6.2 Limitations
A remark should be made about the use of stated preference models, as this potentially give rise to a hypothetical bias or strategic behavior (Holmes, 2017). If people are confronted with new situations, the cognitive difficulty may be high, leading to behavioral responses. This may result in estimates that do differ from real-life responses. This study used the multinomial logit model to analyze the data of the choice experiment. In general, this model fits well for CEs, however there is some critique regarding the independence of irrelevant alternatives (IIA) assumption of the MNL model. This assumption states that the ratio of choice probabilities between two alternatives in a choice set is unaffected by other alternatives in the choice set (Holmes et al., 2017). Therefore this assumption is sometimes argued to be too restrictive. As a result, I suggest future research to compare the results to comparable models like the mixed logit model or the nested logit model.
The sample size of this CE was relatively small, as 96 respondents completed the survey. Increasing the sample size generally provides more reliable estimates. GasTerra is a wholesaler in natural gas and green gas. Therefore, a remark must be made about the sample group, as this survey was only distributed amongst employees of GasTerra. Given the industry GasTerra is operating in, it is likely that a cognitive bias is present in conducting this survey. Therefore the results might be underestimated as one expects that employees of GasTerra have an incentive to favor natural gas. Furthermore, it must be noted that the population of GasTerra is not representative for the Netherlands, as the distribution of men and woman are not equally distributed. Therefore a larger and more diverse sample group would benefit the representativity of the survey.
38
Acknowledgements
39
Reference list
Afman, M.R., Wielders, L.M.L. (2016) Factsheet: ontwikkeling prijzen garanties van oorsprong. CE Delft.
Amador, F. J., González, R. M., & Ramos-Real, F. J. (2013). Supplier choice and WTP for electricity attributes in an emerging market: The role of perceived past experience, environmental concern and energy saving behavior. Energy Economics, 40, 953-966.
Balat, H., & Kırtay, E. (2010). Hydrogen from biomass–present scenario and future prospects. International Journal of Hydrogen Energy, 35(14), 7416-7426.
Bergmann, A., Hanley, N., & Wright, R. (2006). Valuing the attributes of renewable energy investments. Energy policy, 34(9), 1004-1014.
Breidert, C., Hahsler, M., & Reutterer, T. (2006). A review of methods for measuring willingness-to-pay. Innovative Marketing, 2(4), 8-32.
Blomquist, G. C., & Whitehead, J. C. (1998). Resource quality information and validity of willingness to pay in contingent valuation. Resource and Energy Economics, 20(2), 179-196. Borchers, A. M., Duke, J. M., & Parsons, G. R. (2007). Does willingness to pay for green energy differ by source?. Energy policy, 35(6), 3327-3334.
Cappellen, L., Croezen, H., & Rooijers, F. (2018). Feasibility study into blue hydrogen-Technical, economic & sustainability analysis. CE Delft.
CE Delft (2018). Waterstofroutes Nederland.
Certifhy (2015). Overview of the market segmentation for hydrogen across potential customer groups, based on key application areas. Deliverable No. 1.2
Certifhy (2015b). A review of past and existing GoO systems. Deliverable No. 3.1
Certifhy (2015c). Recommendations on the establishment of a well-functioning EU hydrogen GoO system. Deliverable No. 3.3
Cherchi, E., & Hensher, D. A. (2015). Workshop synthesis: Stated preference surveys and experimental design, an audit of the journey so far and future research
perspectives. Transportation Research Procedia, 11, 154-164.
Coast, J., & Horrocks, S. (2007). Developing attributes and levels for discrete choice experiments using qualitative methods. Journal of health services research & policy, 12(1), 25-30.
