Competitiveness of renewable heat supply in well-functioning district heating markets in the Netherlands

(1)

Competitiveness of renewable heat supply in well-functioning

district heating markets in the Netherlands

A quantitative analysis of market failures and costs of district heating technologies Keywords: District heating, Market functioning, Market failures, Renewable heat, The Netherlands

Gerrit Jan Adriaan van de Poll, s2966840 Master Thesis Economics & Finance

Supervisor: Prof. dr. M. Mulder June 4, 2020

Abstract

(2)

1 Introduction

Increasing global concerns regarding climate change have pushed many national governments, such as that of the Netherlands, to introduce new legislation to help combat this climate emergency. In 2019 the Netherlands introduced the Dutch Climate Act, which states that the Dutch government aims to reduce greenhouse gas emissions in the Netherlands by 49% by 2030 and 95% by 2050, compared to the levels of 1990 (Ministry of Economic Affairs and Climate Policy, 2019b). As a part of the effort to reduce greenhouse gas emissions in order to fulfill the commitments of the Dutch Climate Act, the reduction of energy consumption for heating purposes is imperative. Half of the energy consumed in the country is used for heating buildings and other related production processes, primarily in the form of natural gas (Segers et al., 2019). Thus, certain measures must be undertaken to guarantee a sustainable provision of energy for heating.

As an alternative to natural gas-fired heating, district heating (DH) systems hold significant potential in the effort to decrease greenhouse gas emissions (Oei, 2016). The potential of DH systems mainly lies in its ability to employ a variety of energy sources for heat production. These systems utilise a network of insulated pipes to distribute the produced heat (generally in the form of hot water) to end users to satisfy their demands for heating (Frederiksen and Werner, 2013). DH systems allow for the inclusion of energy from multiple sources, such as (1) Combined Heat and Power (CHP) production, (2) residual energy such as industrial waste heat and, (3) Renewable Energy Sources (RES) such as large-scale geothermal and biomass for heat supply (Liu et al., 2019). Although DH systems can make use of RES, in practice, less than a fifth of the total energy supply of DH in the Netherlands is obtained from these sources (Segers et al., 2019). The remainder of the energy supply in DH is primarily obtained from coal and gas-fired heat sources. This relatively small share of renewable heat supply in DH leads us to question the competitiveness of renewable heat supply versus fossil fuel fired heat supply (e.g. coal or gas-fired) in fulfilling the energy requirements of DH systems. As it is desirable to utilise renewable heat supply in DH in order to comply with the Dutch Climate Act, this study aims to investigate the competitive position of renewable heat supply versus fossil-fired heat supply in DH systems in the Netherlands.

(4)

different types of solutions can be implemented by the regulator, and are selected on the basis of the type of market failure.

Based on several of the characteristics of DH systems and previous literature on market failures in DH, it seems likely that market failure are indeed present in DH markets, thus hindering satisfaction of the conditions that ensure the existence of a well-functioning unregulated DH market. These market failures take multiple forms. First, similar to the case of electricity and gas markets where the establishment of physical infrastructure is necessary to ensure supply, it is argued that the distribution of hot water in DH constitutes a natural monopoly due to significant economies of scale (W˚arell and Sundqvist, 2009; Söderholm and W˚arell, 2011). Second, the production of heat may also be subject to economies of scale due to higher efficiencies in larger scale heat production facilities (Danish Energy Agency, 2016). Third, DH systems require strong coordination between production and distribution in order to keep the supply and return temperature of DH systems constant. Therefore, the existence of economies of scope in the vertical integration of production and distribution of heat seem likely (Liu et al., 2019). Fourth, DH operation can lead to the emission of pollutants through the burning of fuels (e.g. coal and gas) in order to produce heat. Unless these effects are internalised in the decision-making process of the polluter, either through regulatory or economic instruments, the choice of the market between different heat production technologies and fuels may become distorted (Söderholm and W˚arell, 2011; Fahlén and Ahlgren, 2010).

Whilst it may seem likely that market failures hinder the functioning of unregulated DH markets, literature on market failures and the necessity of regulation in DH markets in the Netherlands, or in general, remains scarce. Not only this, but very little empirical evidence exists with regards to market failures in DH. Thus far, only a few studies have empirically investigated the presence of economies of scale in production (Wibe, 2001), distribution (Park et al., 2016) and environmental externalities (Fahl´en and Ahlgren, 2010) in DH, and these papers focus only on countries such as Sweden and South-Korea.

Based on the discussion above, it is clear that empirical evidence on market failures is lacking, and thus an analysis of the existence of market failures in DH markets is necessary in order to answer the first research question of this study:

(1) To what extent are the conditions for well-functioning district heating markets satisfied in the Netherlands?

In order to investigate the existence of market failures in DH markets in the Netherlands, we will first explore DH systems themselves, and the relevant literature, to hypothesise the existence of possible market failures in DH. Based on our initial hypotheses on the presence of economies of scale and scope and environmental externalities in the production and distribution of heat, we will then empirically investigate these potential market failures. We will do so through the use of two different methods. First, we will employ the econometric estimation of a cost function to estimate economies of scale and scope in DH firms. For this study, we rely on cost data of Swedish DH firms from the period of 2013 to 2018 in our estimation of economies of scale and scope, since Swedish DH markets, in contrast to DH markets in the Netherlands, are unregulated, and thus regulation does not hinder our ability to capture market failures in unregulated DH markets. In addition, Sweden is the only country where data on DH firms is publicly available. Second, in our investigation of the presence of environmental externalities in DH, we will use environmental cost accounting to estimate the environmental cost of a variety of heat production technologies used in the Netherlands.

(5)

competitive than the latter. Therefore, the second part of this study will involve an investigation into the competitive position of renewable heat supply in comparison to that of fossil-fired heat supply. In doing so, we will distinguish between the short run and long run competitiveness of heat production. In the short run, the competitiveness of heat production is determined by the short run marginal costs (SRMC) of production. To elaborate, with the aim of cost minimisation by the market operator, heat production technologies with the lowest SRMC are the first to be utilised to meet demand (Dominkovi´c et al., 2018). Similarly, in the long run, for a given market price, the heat production technology with the lowest long run average cost would be the optimal choice to build and operate (Hansen, 2019). In the literature, there are two related ways to define the long run average cost. First, the Levelised Cost of Energy (LCOE) depicts the present value of the average cost per unit of energy that the plant is expected to produce over its lifetime (Hansen, 2019). Second, the Annual Revenue Requirement (ARR) depicts the average lifetime cost per unit of capacity of the heat production plant (De La Torre et al., 2008).

Similar to the lack of literature on market failures in DH markets, the existence of literature on the competitiveness of renewable heat supply versus fossil-fired heat supply is also scarce. While the SRMC and LCOE of DH technologies have been estimated in a few studies (Hansen, 2019; Liu et al., 2019; Sun et al., 2016), these studies often only consider a small set of heat production technologies and do not account for the environmental cost of production. Therefore, our second research question is:

(2) Assuming the existence of well-functioning district heating markets, what is the competitive position of renewable heat supply in the Netherlands?

We will address this question through the use of the previously discussed cost estimation methods to evaluate both the short- and long-term competitiveness of both renewable and fossil-fired heat production in the Netherlands. We will use the most recent technical, financial, and emission data on representative heat production technologies.

Our results show that DH markets in the Netherlands, to a certain extent, do not inherently satisfy all the conditions necessary for well-functioning markets to exist due to the presence of market failures. We base this finding on our estimation results of economies of scope in the production and distribution of heat in Swedish DH firms, which indicates that the combination of vertical operational activities will lead to cost-efficiencies. Consequently, in a situation where only a single distribution network is present, an integrated DH firm that is both producing and distributing heat has a cost advantage over a specialised firm that only produces heat, therefore giving the integrated firm market power. In addition, we find that the environmental costs of heat production are currently insufficiently internalised through economic instruments in the price of heat. As a result, the optimal choice of the market between different sources of heat supply can become distorted. Consequently, the introduction of regulation is required in order to ensure that there are well-functioning DH markets in which renewable heat supply is able to compete with fossil-fired heat supply.

Based on our results of the presence of market failures in DH markets, and with the assumption that the necessary regulation has been implemented to satisfy the precondition for well-functioning DH markets, we find that renewable heat supply (after phasing out of coal-fired heat supply) holds a strong competitive position relative to gas-fired heat production. In addition, we find that a decrease in electricity prices in general coincides with an increase in the competitive position of renewable heat supply in DH relative to fossil-fired heat supply, and vice versa.

(6)

Given the capability of DH to incorporate RES and thus contribute to a reduction in greenhouse emissions, there may be temptation to infer that investing in DH will automatically lead to a consequent reduction in greenhouse emissions. However, it is important to remember that investment in DH does not necessarily equate to investment in the use of RES in DH. Our results show that government intervention in the form of regulation is necessary to allow for competition between renewable heat supply and fossil-fired heat supply. Additionally, various policy instruments may be required to increase the relative attractiveness of renewable heat supply, however, the scope of these policies is dependent on the price of electricity.

To the best of our knowledge, this study is the first to address the issue of why the share of renewable heat supply in DH in the Netherlands is relatively small. In doing so, our findings on market failures in DH markets and the relative competitiveness of renewable heat supply versus fossil-fired heat supply serve to contribute to reducing the scarcity of academic literature on each topic.

This study has been structured into eight sections: (1) the introduction; (2) background infor-mation, which serves to discuss the operation and characteristics of DH systems and to provide a comparison between DH in the Netherlands and Sweden; (3) theory on economic efficiency and market failures, and the competitiveness of heat production; (4) research methods, including market failure analysis and cost estimation methods; (5) cost data on DH firms and techno-economic data on heat production; (6) results and their interpretation; (7) discussion on the results, implications and limitations, and; (8) conclusion.

2 Background

In order to identify potential market failures in Dutch DH markets and to understand the drivers of competitiveness in heat production, it is first important to understand how DH systems operate, the characteristics of DH, and the overall context of this study. Each of these factors will be discussed in turn. First, we will discuss how DH systems work, and elaborate on the characteristics of DH. Second, we will discuss the context within which DH is part of the total supply and demand for heat in the Netherlands, the market organisation and regulation of the DH sector, and the future opportunities and challenges of DH in the Netherlands. The final component of this section will serve to address an issue raised in the introduction, whereby regulation of the DH sector in the Netherlands limits our ability to capture market failures. To combat this limitation, this study makes use of cost data on Swedish DH firms, this data is publicly available and DH markets in Sweden are unregulated. However, if our results, which are based on Swedish DH firms, are to be generalised in order to contribute to answering our research question regarding the extent to which the conditions for well-functioning DH markets in the Netherlands are satisfied, then we deem a comparison between the DH sectors in the Netherlands and Sweden necessary.

2.1 Operation and characteristics of district heating

Many of the characteristics of DH that are relevant to understanding the background of our research topic relate to how DH operates. Therefore, before discussing the relevant characteristics of DH, we first must elaborate on the functioning of DH.

(7)

Figure 1: Basic schematic overview of the technical and economic layer of district heating systems.

Thick solid lines indicate the flow of hot water towards the heat user installation and the return flow of cooled water back to the heat production installation. Dashed lines indicate financial flows, and thin solid lines represent either economic interaction between economic agents or the utility received from consuming the heat.

which feed heat, commonly in the form of hot water, into a transport network. This transport network serves to move heat to and from a distribution network. A heat transfer station exists between these networks, and is necessary as a point of separation between the two. Finally, the distribution network manages the supply of hot water to households and other buildings, as well as the return of the cooled water. For an extensive schematic overview of the technical layer see Appendix A.

Individual heat networks may deviate from this basic scheme in some regards. For example, in smaller heat networks, there is not always a distinction between the transport network and the distribution network. In this case, the heat production source feeds directly into the distribution network. In larger networks, it is also possible that a heat production source does not directly feed into the transport network, but rather is connected via an additional pipe. In many cases, the primary heat source is also often supported by one or more auxiliary boilers that provide assistance in order to meet peak demand. In this case, the primary heat source provides only the baseload.

The economic layer of DH systems concerns the economic agents involved in the operation of such systems, as well as the costs that they incur as a result. Based on the technical system design and market design, the economic agents involved and the costs they incur may differ. In general, we can distinguish between the following economic agents in DH systems:

(8)

• Transport network operator: the transporter owns and operates the transport network from a heat source to, but not including, the main heat transfer station. The costs arise from investment expenditure in the network and annual operating costs. As discussed earlier, the need for, and therefore the presence of, a transporter is dependent on the size of the network; • Distribution network operator: the distributor owns and operates the distribution network and all the related components necessary for the distribution of heat up to and including the end-user installation. The costs arise from investment expenditure in the network and auxiliary installations and annual operating costs. In addition to maintenance and operation, the operational costs consist of system imbalance costs resulting in the need for the co-firing of auxiliary heat boilers;

• Heat retailer: the retail heat supplier collects the revenues from the heat supply to consumers and reimburses the costs of all agents involved in the heat supply. The heat supplier does not manage the physical components itself, and;

• Consumer: the consumer incurs costs based on his/her demand for heat.

Based on the discussion of the technical and economic operation of DH, two characteristics of DH systems are especially relevant. First, as the physical distance between the producer and user of heat increases, the costs of DH increases exponentially due to the cost of infrastructure and energy losses that occur during the transport and distribution of heat. Consequently, DH systems have a limited geographic scope (Liu et al., 2019). Additionally, similarly to a power system, DH systems require balance in the form of a constant supply and return temperature (S¨oderholm and W˚arell, 2011). As a result, the need for coordination between heat production and distribution is strong.

The second vital characteristic of DH is its flexibility and capability to incorporate a variety of energy sources for heat production. These hybrid systems manage to achieve both higher energy efficiency and greater fuel savings than heat production at the household level, whilst still minimis-ing environmental impact (Papa et al., 2019). Over the years, DH systems have been developed to operate at lower temperatures. As a result, many different types of heat production technologies are able to be fed into DH networks without compromising the reliability of the overall network operation. The heat production technologies used in DH can be categorised into two groups: Com-bined Heat and Power (CHP) production, and heat only production. CHP plants can either produce both heat and electricity together and in different ratios, or may produce only one of the two. CHP systems use the rejected or waste heat from the power production units in order to produce hot water (Sayegh et al., 2017). CHP production facilities come in a wide range of sizes and can be used with a variety of fuels, including coal, gas, waste, and biomass. In contrast to CHP, heat only pro-duction technologies, such as geothermal heat, industrial waste heat, heat only boilers (HOB), and heat pumps (HP), are used to produce heat alone. These types of technology can run on fossil fuels such as coal and gas, renewable fuels such as waste and biomass, or fuel in the form of electricity. Depending on the source of power production, electricity may be considered a renewable fuel source for heat production.

2.2 Characteristics of district heating in the Netherlands

(9)

order to gain a more comprehensive understanding of DH in the Netherlands, we will first clarify the context within which DH is part of the total heat supply and demand in the Netherlands. We can then explore current DH systems in the Netherlands, followed by a discussion on the market structure and regulation of the DH sector in the Netherlands. We will also shortly discuss the policy instruments currently in place in the Netherlands to incentivise investment in renewable heat and disincentivise greenhouse gas emissions resulting from the production of heat for DH. We end with a discussion on the future opportunities and challenges for DH in the Netherlands and in general.

As depicted in Figure 2, in 2015 a total of 1,224 petajoule (PJ) was used for heating in the Netherlands (Ecorys, 2016). This energy demand is consumed by three sectors: namely, for space heating and warm tap water in the built environment (∼50%), for process heat in industry (∼40%), and for heating greenhouses in the horticulture sector (∼10%) (Segers et al., 2019). These ratios remain fairly constant over time, with the built environment consuming relatively more heat for space heating in years with colder winters. In terms of heat supply, natural gas, electricity, and heat in the form of hot water or steam are the three main energy carriers used to provide heat in the Netherlands (Segers et al., 2019). Of the total consumption of heat, natural gas and electricity are primarily used to produce heat on-site at the household level through the use of gas-condensing and electric boilers. The remainder of the heat is delivered via heat networks.

Figure 2: Sankey diagram showing the breakdown of the total heat demand and supply in the Netherlands in 2015 (petajoules)

Blue streams represent energy demand for heating per sector. Orange streams break down energy supply through heat networks and for on-site heat production. Green streams distinguish between heat supply in district heating and community heating. Estimates of heat supplied via district heating differ per source. Figure represents estimate by Ecorys (2016).

Heat networks can be distinguished into either industrial or non-industrial heat networks. In-dustrial heat networks primarily supply steam to industry requiring high temperatures for heating processes. Industrial heat networks may consist of one firm supplying heat to another firm, or as clusters of firms exchanging heat with each other (Segers et al., 2019). Non-industrial heat networks mainly supply heat in the form of hot water (without steam). These networks can be further subdi-vided into networks that distribute heat within a building, also known as community heating, and networks that transport heat from decentralised heat sources, also known as DH.

(10)

annual supply of heat. Large DH networks, which are defined as networks that supply more than 150 terajoule (TJ) of heat to end users every year, are located primarily in major Dutch cities. During 2015, large DH networks supplied approximately 20 PJ to around three hundred thousand linkages (Menkveld et al., 2017). Furthermore, almost three quarters of the heat supplied to these large grids comes from natural gas and coal-fired power stations and natural gas-fired auxiliary boilers. Approximately 15 percent of the heat supplied in 2015 was obtained from renewable energy sources. Currently, the contribution from conventional CHP plants is dominant but is slowly declining. As well as this, a second trend to note in large DH networks is an increase in the use of biomass instal-lations. This concerns both the conversion of installations that previously only produced electricity, as well as the establishment of new installations. On the other hand, small DH networks, which are defined as those which supply less than 150 TJ annually to end users, supplied approximately 2PJ to 50 thousand connections in 2015 (Menkveld et al., 2017). Many smaller DH networks use heat and cold storage and heat pumps.

Although there is often mention of “the district heating market”, there is currently no market for DH with supply and demand at the national level in the Netherlands. Suppliers of DH are not in direct competition with one another due to DH systems being predominantly designed as isolated systems without connections to one another (Liu et al., 2019). As a result of this design, users of DH have no freedom of choice in selecting their preferred DH provider. Additionally, due to the inability of heat to be transported and distributed profitably over long distances, no wholesale market exists for DH. Although there are many different ways for the DH sector to organise themselves under these conditions, in general DH systems in the Netherlands are primarily vertically integrated (CE Delft, 2009). This means that a single firm will act as a full-service heat supplier by producing, transporting, and distributing heat to its consumers (CE Delft, 2009). In terms of ownership, both public and private parties own and operate the production, transport and distribution activities involved in DH (Oei, 2016).

Similar to that of the electricity and gas sector, regulations exist in the DH sector as a method of increasing social welfare (Den Hertog, 2010). Given the dependency of DH users on their provider, the Heat Act, and later the Revised Heat Act, were introduced in 2014 and 2019 respectively in the Netherlands to protect households and small companies (i.e. those with a connection capacity below 100kW). In order to prevent monopoly pricing, the heat price is capped using the NMDA principle. The NMDA is a Dutch principle and translates to mean “not-more-than-otherwise”. In practice, it means that a user of DH should not be allowed to pay more for energy than if that user were connected to a gas network and making use of a gas-fired boiler. Besides this price-cap regulation, the Heat Acts also serve to cover aspects such as financial compensation for outages, conditions for disconnecting consumers, content of contracts, and other relevant obligations. The Dutch regulatory body for consumers and markets, Autoriteit Consument en Markt (ACM), monitors the compliance of suppliers with the Revised Heat Act. DH networks with only large consumers (i.e. those with a connection capacity above 100kW) are not covered by either of the Heat Acts.

(11)

Besides the taxes and fees in place to reduce greenhouse gas emissions, the Dutch government is planning to start phasing out coal-fired plants (Ministry of Economic Affairs and Climate Policy, 2019a)

The final component of this section provides an overview of some of the future opportunities and challenges that DH systems in the Netherlands can expect to encounter. In electricity markets, wind energy and solar energy are playing an increasingly large role (Zerrahn et al., 2018). However, these sources of energy are not continuously available on demand. Due to the intermittent character of wind and solar power, there is a need for balancing power to maintain system balance. DH systems can play an important role in integrating intermittent energy sources into energy systems, as they are able to cope with the nature of intermittency by using electricity to produce heat.

Another opportunity for DH systems is the development towards utilisation of lower temperature heat supply. Under lower temperatures, DH systems may attain lower distribution losses, and hence increase their cost-effectiveness (Castro Flores et al., 2017). Besides lowering energy losses, there are also more types of heat sources available that have the capability to supply low-temperature heat to DH systems.

There are, however, also challenges that exist for the DH sector. Heat demand and density act in part to determine the profitability of a DH system. With the threat of global warming calling for improvements in energy savings measures, heat demand and density are expected to decrease, thus having a negative impact on the profitability of DH systems (Castro Flores et al., 2017). In addition, increasing the share of renewable electricity generation in the energy mix, which has lower short run marginal costs, forces down wholesale electricity prices, a result of the so-called “merit order effect” (Sensfuß et al., 2008). As electricity prices decrease, the competitiveness of individual electric heating solutions such as small-scale electric boilers and heat pumps will increase.

2.3 Comparison of district heating in the Netherlands and Sweden

In comparing DH in the Netherlands to that of Sweden, we will first introduce the characteristics of DH in Sweden, and then provide a comparison of DH in the two countries based on the role of DH in the total heat provision system, the market structure of DH, and the regulatory environment.

(12)

Based on the characteristics of DH in Sweden and those of the Netherlands, an overview of the characteristics of DH in each of these countries is depicted in Table 1. Based on Table 1, we see several differences and similarities between the DH sector of both countries.

Beginning with the differences, we note that, in contrast to Sweden, the DH sector in the Nether-lands holds a relatively small share in the total supply of heat. In addition, we see large differences in the ownership structure in terms of public versus private ownership, with the Netherlands leaning towards private ownership and Sweden tending towards public ownership. Concerning the regulatory environment, the DH sector in Sweden is deregulated, whereas the Netherlands has a price-cap reg-ulation in place. Other regulatory differences include the legal obligations of DH firms and network access conditions. In terms of the similarities, both Sweden and the Netherlands feature primarily vertically integrated DH activities. We believe this likeness is indicative of similar underlying effects between DH systems in the Netherlands and Sweden, which enables us to justify our approach in the use of Swedish data for our market failure analysis of DH in the Netherlands. With this in mind, we are able to continue on to discuss the theory relevant to answering our research questions. Table 1: Overview of the district heating sector in the Netherlands and Sweden (Adapted from Donnellan et al. (2018); Oei (2016))

Variable The Netherlands Sweden

Market context

Market share of district heating 5% 55%

Number of DH systems 13 large-scale and ∼350 small scale

networks

500-600 systems operated by over 200 companies

Length of network (km) 4,000 23,400

Ownership structure

Mostly private developers/operators, some municipality owned. Vertical integration among DH activities

Initially owned by municipalities, ownership is now a mix of private and publicly-owned.Vertical integration among DH activities

Regulatory environment

Type of regulation Regulated Liberalised

Pricing structure Price cap based on the

No-More-Than-Otherwise principle

Unregulated. Voluntary pricing and complaints processes in place to aid transparency

Legal obligations

Licences for suppliers, municipalities required to consider DH for new development areas

None, other than submitting annual accounts

Network access conditions Negotiated access for heat producers

(not suppliers)

DH operators are obliged to negotiate with any third parties requesting access

Regulatory body The Netherlands Authority for

Consumers and Markets (ACM)

Energy Markets Inspectorate and Competition Authority

3 Theory

(13)

3.1 Benchmark for market efficiency

Before discussing the conditions for well-functioning markets, it is important to first clarify the concept of a market itself. In economics, a market is considered to be any structure that enables buyers and sellers to meet to exchange goods, services and information. In those markets, all of the buyers and sellers of any type of goods, services, or information, and consequently those who influence its price, are referred to as either market participants or economic agents. Markets themselves are of importance as they facilitate trade and also enable the distribution and allocation of resources in society.

In order to evaluate how well a market functions, it is useful to have a benchmark. The concept of a perfectly competitive market provides a useful benchmark to allow for evaluation of actual markets (Moselle et al., 2010). Under perfect competition, productive efficiency and allocative efficiency are maximised. Productive efficiency and allocative efficiency refer respectively to goods which are being produced and sold at the lowest possible cost to the consumers with the highest willingness to pay for these goods (Taylor, 2012). To achieve perfect competition, certain conditions must be present and fulfilled:

• Identical products: many firms produce identical products;

• Many buyers and sellers: many buyers are available to buy the product, and many sellers are available to sell the product, so that market participants are not able to strategically influence the market outcome;

• Informed buyers and sellers: buyers and sellers have all relevant information to make rational decisions, and;

• Free market entry and exit: firms can freely enter and exit the market.

The above conditions imply that market participants in the economy are price takers (Taylor, 2012). In the short run, a perfectly competitive firm bases its output decision on the quantity that will maximise its profits, or, in a case where profit is unattainable, minimises its losses. In the long run, a perfectly competitive firm will react to profits through an increase in output, or respond to losses through a reduction in output or by exiting the market. Ultimately, a long run equilibrium will occur when economic profits are zero and, as a result, there is no incentive for any firm to either exit or enter the market. In such a situation, both allocative and productive efficiency are achieved.

3.2 Market failures

In reality, markets differ from the benchmark of perfect competition. Fundamental shortcomings in the market design, also known as market failures, inevitably act to prevent the achievement of an optimal outcome within markets (Winston, 2007). In theory, the following types of market failure can be distinguished (Mulder et al., 2019):

• Economies of scale and scope: economic agents can influence market outcomes (market power), which may result in production that is too low or of poor quality whilst still pricing them above the social optimal level. This may occur in the case of activities with large fixed costs, such as investments in networks, which results in the ability of one firm to conduct these activities more efficiently than several firms (natural monopoly);

(14)

• Negative externalities: economic agents do not consider all costs of their activities, which may result in too high a level of activity from a societal point of view. Like positive externalities, negative externalities may occur in the form of production or consumption externalities; • Network externalities: the utility that an economic agent derives from a good increases with

the number of other users of that good, which may result in a limited number of suppliers capturing the full market and, consequently, other firms are unable to enter the market; • Information asymmetry: economic agents do not have the same information on the product

or on an activity of the counterpart, causing adverse selection or moral hazard, respectively. In the case of adverse selection, buyers are not prepared to buy high-quality products as they cannot distinguish them from low-quality products. In the case of moral hazard, economic agents are not operating optimally, such as by taking on too much risk or buying too much of a product, and;

• Hold up problem: economic agents are uncertain about the ex post revenues, which may result in too low a level of investment activity.

Whilst the previously discussed market failures may be present in any market, we are ultimately interested in potential market failures in DH. A characteristic that sets DH and energy markets apart from the more traditional markets is the requirement for a network which allows for the exchange of a product between buyers and sellers. Firms that are active in these type of markets are part of what is known as “network industries” (Gottinger, 2003). Various market failures may apply in network industries due to the characteristics of networks. In network industries, markets usually involve large, risky capital investments, economies of scale and scope, interdependent technologies owned by different market players, and network externalities (Gottinger, 2003). In light of this, market failures in those industries seem likely.

While the presence of market failures in DH markets seems likely, the existing literature on market failures in DH currently only investigates economies of scale in the distribution and production of heat (Wissner, 2014; Wibe, 2001; Park et al., 2016) and environmental externalities in DH (Fahl´en and Ahlgren, 2010). Based on the literature and considering the characteristics of DH, some of the most important market failures that may apply to DH markets are:

• Economies of scale and scope: based on large fixed costs, such as investments in networks and production facilities, significant economies of scale and scope in the production and distribution of heat may result in a natural monopoly position. In such a case, the DH firm may exert market power, and;

• Environmental externalities: burning fossil fuels for heat production leads to high levels of air pollution, and as a result causes damage to others. If the environmental impact of using these fuels is not internalised in the price of heat, an inefficient allocation of resources may result. In the following two sections, we will separately discuss these market failures in more detail.

3.3 Economies of scale and scope

(15)

Panzar and Willig (1981) and Baumol et al. (1982) introduced the notion of economies of scale and scope to characterise the effect of size and output diversification in multi-product firms, respectively. In principle, economies of scale occur when the average costs of a firm decline as its output of a product expands. Economies of scale in a single-product firm may lead to market failure in the form of natural monopoly. However, this depends on the level of output, as illustrated in Figure 3.

Figure 3: Economies and diseconomies of scale and subadditivity of costs for a single-product firm Average Cost (AC) Output (Q) AC a AC b Economies of scale Subadditivity of costs Q 1 Q 2 Disconomies of scale

Economies of scale of f irmaare only present up to Q1 as average costs ACadecline. After Q1average costs

of f irma increase again (diseconomies of scale). However, up to Q2 it remains less costly to supply output

Q2 with f irma, rather than having both f irma and f irmb producing. Firm cost functions that have this

attribute are said to be subadditive at output level Q2(Sharkey et al., 1982).

Previous literature on the determinants of economies of scale has identified a large variety of contributing factors. In general, we can distinguish between two types of economies of scale: internal and external. Declining average cost may be either the result of firm-specific activity (internal economies of scale), or due to external factors (external economies of scale). Sources of internal economies of scale include the specialisation of labor, efficient capital use, negotiation power resulting in lower interest rates, lower prices for raw materials, lower wages, and engineering principles such as the square-cube law (Smith, 1776; Bruni, 1964). Briefly, the square-cube law states that the surface of a container increases by the square of the dimensions, while the volume increases by the cube. Using the example of a pipe, this means that the total cost of a pipe is proportional to surface area, whereas the capacity is proportional to volume. External economies of scale concern factors that are external to the firm but internal to the industry. Major factors causing external economies of scale include advantages from within-industry specialisation, conglomeration of firms within a certain area attracting labour and supporting industry, supportive legislation, indivisibilities, and better transport links resulting in lower average costs (Caballero and Lyons, 1990).

(16)

In the example of economies of scale in Figure 3, we only consider the case of a firm producing a single-type product. In a multi-product firm, economies of scale in the production of one product can be compensated by diseconomies of scale in another product (product-specific), or by diseconomies of scope. A sufficient condition of a multi-product firm being a natural monopoly requires that economies of scale exist for each good, and that economies of scope hold (Ida and Kuwahara, 2004). Economies of scope exist in the case that the combined production of two or more goods or services results in lower costs than if these goods or services were to be produced separately. Common sources of economies of scope include the use of common inputs, joint production facilities, shared costs, and diversified revenue streams. Besides economies of scope existing in the horizontal term of the definition, economies of scope may also occur between vertical stages of production, and can also be referred to as economies of vertical integration. Economies of vertical integration may occur if strong technological interdependencies between vertical stages are present, and if there is a strong need for coordination and adaptation across stages. If these conditions are satisfied, vertical integration enables better coordination across production stages and avoids the duplication of fixed costs (Garcia et al., 2007).

3.3.1 Empirical estimation strategies in the literature

In our discussion of the empirical estimation strategies for estimating economies of scale and scope, we largely follow the seminal work of Baumol et al. (1982). We will first discuss the definitions of both overall and product-specific economies of scale and economies of scope, followed by a discussion on the considerations incorporated in the empirical estimations of economies of scale and scope.

The degree of overall (ray) economies of scale at output vector y in a multi-product firm can be determined by: SR(y) = C(y) Pn i=1yi ∂C(y) ∂yi (1) where C(y) is the (total) cost as a function of output, and yi(i = 1, ..., n) are the types of output

produced by the firm of vector y. The measure for overall economies of scale SR(y) assumes that the

size of the composite output bundle can vary, but that the relative composition of the output bundle remains constant. That is, in increasing the size of the composite output bundle, every output must be expanded with the same proportion. Overall, economies or diseconomies of scale exist if SR(y)

is larger or smaller than one, respectively.

In order to determine product-specific economies of scale, we first need to determine the incre-mental cost of producing the product. The increincre-mental cost of producing product i is estimated in the following way:

ICi(y) = C(y) − C(yN −i) (2)

where y = yi+ yN −i. ICi(y) amounts to the increase in total costs if the amount yi of product i is

added to the output bundle yN −i. The degree of product-specific (dis)economies of scale of product

i at output vector y can then be determined by: Si(y) = ICi Pn i=1yi ∂C(y) ∂yi . (3)

Similar to the measure of overall economies of scale, product-specific economies or diseconomies of scale are present if Si(y) is larger or smaller than one, respectively. In contrast to overall economies

(17)

Regarding the degree of economies of scope between distinct product sets, we suppose that S with 2 ≤ S ≤ n is the number of product sets, and Yp is any combination of the types of output

from vector y (except for the entire vector y itself). For all p 6= k, no component yi of Yp is an

element of Yk, otherwise we will also be accounting for product-specific economies of scale. Based

on the aforementioned assumptions, the degree of economies of scope between two or more product sets can then be determined as follows:

SCS(y) =

PS

p=1C(Yp) − C(Y1, ..., YS)

C(Y1, ..., YS)

(4) where C(Y1, ..., YS) is the cost of the joint production of sets S. Economies or diseconomies of scope

between the product sets are said to exist if SCS(y) is larger or smaller than zero, respectively.

For the estimation of the degrees of economies of scope as in Equation 4, detailed cost data on integrated and specialised firms is required. Baumol et al. (1982) shows that the presence of cost complementarity between two outputs is a sufficient condition for economies of scope to exist. Cost complementarity requires that the marginal cost of producing one product is reduced when the output of another product is increased (Greer, 2011), such that:

∂2C(ˆy)∂yi∂yj< 0 (5)

for i 6= j and for all ˆy(0 ≤ ˆy ≤ y).

Estimation of scale and scope economies using the above discussed definitions requires the econo-metric estimation of either a cost function or a cost frontier. The difference between these estimation methods is that the estimation of a cost function requires the assumption of cost-minimising be-haviour, which does not control for the presence of inefficiency (Saal et al., 2013).

Whilst some studies estimate a cost frontier, in general most studies that estimate economies of scale and scope make use of the econometric estimation of a cost function. In estimating a cost function, choices need to be made regarding (1) the functional form of the cost function; (2) whether the function is a total or variable cost function, and; (3) the estimation technique. In our discussion of these choices, we follow the discussion of Saal et al. (2013).

(18)

Next, we must discuss the implications of choosing either a total cost or a variable cost function. In estimating a total cost function, given cost minimising decisions by the firm, it is assumed that all inputs are endogenous and all outputs are exogenous. However, if certain inputs are invariable in the short run, selection of a total cost function would be inappropriate, as certain inputs would be fixed or quasi-fixed. In such a case, a variable cost function is more appropriate. Then, under the condition of having fixed input factors, a firm minimises its variable costs based on the given levels of input prices, output, and the level of its fixed inputs. Considering these factors in empirical practice, a long run total cost function approach allows for a fuller estimation of economies of scale and scope that considers the impact of both capital and variable inputs (Saal et al., 2013).

Finally, in regards to the estimation techniques, two general methods can be identified in the literature: namely, the traditional cost function method and the stochastic frontier analysis (SFA) method. Regarding the traditional cost function method, previous studies have applied both single-equation econometric estimation methods (e.g. Ordinary Least Squares), as well as multi-single-equation methods (e.g. Seemingly Unrelated Regressions). The traditional cost function method results in an average response function, whereas the SFA method allows for firm level inefficiency by estimating a best practice frontier (Saal et al., 2013).

3.3.2 Empirical evidence on scale and scope economies in district heating

Following the initial introduction of economies of scale and scope by Panzar and Willig (1981) and Baumol et al. (1982), the estimation of economies of scale and scope in firms has received much interest in the economics literature of a variety of industries, including network industries such as telecommunications (Bloch et al., 2001), water (Garcia et al., 2007), and utilities (Filippini and Farsi, 2008).

However, thus far, little research has been conducted on economies of scale and scope in the DH sector. Through the use of parametric econometric cost models, Wibe (2001) shows that economies of scale are not prevalent in Swedish DH production. It is important to note that Wibe (2001) was originally written in Swedish, and that we interpreted the main idea of Wibe (2001) based on the work by S¨oderholm and W˚arell (2011). Park et al. (2016) find economies of scale to be both present and statistically significant in the South Korean DH sector. Using a variable cost function, Park et al. (2016) estimate a translog cost function. Several other studies have examined the efficiency of the production of heat in the DH sector using stochastic frontier analysis (Agrell and Bogetoft, 2005; Lygnerud and Peltola-Ojala, 2010; Munksgaard et al., 2005).

3.4 Environmental externalities

In this section, we will discuss the concept of environmental externalities in a structure similar to that of the previous section. We will first introduce the concept of environmental externalities in general. Following this, we will discuss empirical estimation strategies for environmental externalities in the general literature, and will end with empirical evidence on environmental externalities in DH.

(19)

Figure 4: Illustration of the impact of external environmental costs on demand and supply in the social optimal partial equilibrium

Price (P) Output (Q) Q 0 Q 1 P 0 P 1 Demand Eq 0 Eq 1 Private supply Social supply

Eq0 is the private equilibrium with external environmental costs being unaccounted for in the price where

demand is Q0 and price is P0. From a societal point of view, if the external environmental costs are

internalised in the price, the cost of supply increases and, as a result, the social supply equilibrium lies at Eq1defined by new lower demand Q1 and new higher price P1.

3.4.1 Empirical estimation strategies in the literature

In the example of Figure 4, the value of the environmental impact is expressed in monetary terms per unit of output. In order to do so, one needs to quantify the physical magnitude of the environmental impact per unit of production, and place a value on the environmental impact. We will briefly discuss the quantification and valuation of environmental impacts in turn.

Quantification of the physical magnitude of environmental impacts begins by identifying and defining the relevant environmental externalities. Based on Hohmeyer and Ottinger (2012), we dis-tinguish between three overall stages of environmental externality impacts: initial loadings (e.g. tons of CO2 emissions), intermediate effects (e.g. ambient SOx concentrations), and ultimate impacts

(e.g. increased mortality). Ideally, when quantifying the environmental impacts, one would take into account all life cycle activities, all relevant pollutants, and take site-specific measurements. However, by doing so, one increases the complexity of the methodology, and consequently increases the uncertainty of the results (Fahl´en and Ahlgren, 2012).

(20)

3.4.2 Empirical evidence on environmental externalities in district heating

Similar to the amount of empirical evidence on scale and scope economies in DH, little empirical evidence exists with regards to environmental externalities in DH markets. Thus far, only Fahl´en and Ahlgren (2010) find that economic policy instruments do not fully internalise all external costs in Swedish DH systems.

Whilst it is not focused on environmental externalities itself, a larger body of literature exists on environmental cost accounting in DH. Fahl´en and Ahlgren (2012) find that less comprehensive external environmental cost accounting methodologies may be considered appropriate for finding the lowest social-cost solutions in DH systems. In addition, according to Dones and Heck (2005), while the environmental costs associated with upstream processes do contribute a substantial share to the total life cycle environmental costs of DH, it is actually the environmental costs from the core combustion of fuels in heating plants that account for the majority. Furthermore, Freeman (1996) finds that the omission of the environmental costs associated with upstream and downstream processes of DH does not signifinfluence the economic ranking of the social costs of different sources of heat production.

Whilst CHP plants are able to co-produce power and heat, heat pumps must consume power in order to produce heat. The production decision in those type of heat plants thus has an impact on power production decisions elsewhere, and the contrast between these two processes has different implications in terms of associated environmental costs. In the case of plants which utilise CHP production, it is important to account for the avoided environmental costs that would otherwise arise as a result of alternative marginal power production. It has been shown that accounting for these avoided costs has a significant effect on the economic ranking of DH plants in Swedish DH systems (Carlson, 2002; Fahl´en and Ahlgren, 2010; Holmgren and Amiri, 2007). However, this advantage of CHP production was found to be greatly reduced if the alternative power production would be based on natural gas instead of coal (Holmgren and Amiri, 2007; Carlson, 2002).

As a part of the discussion on the trade-off between complexity and comprehensiveness of envi-ronmental externalities, Zvingilaite (2011) finds that by accounting for the health-related damage costs resulting from the emission of N Ox, SO2 and P M 2.5, in addition to the costs of CO2, the

least-cost ranking of heat and power plants in the Danish energy system changes.

3.5 Competitiveness of renewable heat supply

The final component of this section serves to discuss the theory on the competitiveness of renewable heat production. The competitiveness of renewable heat supply, as with any other type of production, revolves around the cost of production relative to its competitors. In discussing the theory on the competitiveness of heat production, we will distinguish between the short run and long run competitiveness of renewable heat supply versus fossil-fired heat supply.

3.5.1 Short run competitiveness of renewable heat supply

(21)

costs. The short run marginal cost (SRMC) represents the cost to produce an additional unit of thermal energy, given an existing plant or installation. Based on the SRMC, the market operator constructs a supply curve for its entire system in the form of a merit order, as depicted in an arbitrary example in Figure 5.

Figure 5: Price setting in district heating markets based on short run marginal cost pricing

Capacity (MW) Marginal cost e /MWh Demand Waste Coal Gas Peak units P Q

The merit order ranks available sources of production based on ascending order of price, reflecting the short run marginal costs of production together with the capacity of production. Accordingly, in a centralised management, production facilities with the lowest marginal costs are the first to be brought online to meet demand, and the plants with the highest marginal costs are the last to be brought online. In this way, production costs are minimised.

Considering the literature on the SRMC of renewable heat supply versus fossil-fired heat supply, we only find the work of Liu et al. (2019). Liu et al. (2019) find that renewable heat production technologies tend to have low SRMC. In addition, they find that increased amounts of renewable heat from geothermal and industrial waste heat push more expensive gas- and coal-fired heat production technologies out of the merit order, thus reducing the operating hours of fossil-fired heat supply. 3.5.2 Long run competitiveness of renewable heat supply

In the long run, all factors of production are variable. This means that in the long run, the total cost of the plant (lifetime operating plus capital costs) must be taken into account in least-cost production planning. From this point of view, for a given market price, the heat production technology with the lowest long run average cost would be considered as optimal. The long run average cost of heat production can be defined in two relatively similar ways: per unit of energy and per unit of capacity. We will discuss each in turn.

(22)

including other parameters such as energy system efficiency and emissions, the least-cost ranking of different heat production technologies differs significantly.

Besides the per unit costs of energy, the Annual Revenue Requirement (ARR) represents the present value of the average lifetime cost per unit of capacity. The ARR is often used in tandem with the Screening Curve Method (SCM). The SCM estimates the least-cost production mix for production planning purposes (Phillips, 1969). It combines the ARR as a function of load hours together with the load duration curve (plot of energy demand in descending order of magnitude versus time in hours). While the LCOE has the advantage of providing a single measure for comparison, the ARR in the SCM gives insight into the share of fixed costs and variable costs per unit of capacity. An arbitrary example of the SCM combining the ARR and load duration curve in the context of least-cost power production planning purposes and applied to a DH market is found in Figure 6.

The ARR or SCM has not previously been employed in the literature to compare the competi-tiveness of DH production technologies.

Figure 6: Screening Curve Method depicting the least-cost production mix in a district heating market ARR ke /MW Load lev el Load hours 0 8760 Load hours 0 ₈₇₆₀ Waste Gas Coal Gas Waste Coal (Baseload)

(23)

4 Methodology

Having identified and discussed the potential market failures in DH and the cost estimation methods of heat production, this section serves to elaborate on the methods that will be employed in answering the two research questions of this paper. With relevance to answering our first research question, we will discuss the research methods for estimating the presence and size of market failures in the form of economies of scale and scope and environmental externalities, respectively. In relation to our secondary research question, regarding the competitiveness of renewable heat supply, we will discuss the SRMC, LCOE, and ARR.

4.1 Estimating market failures

4.1.1 Estimating economies of scale and scope

In estimating economies of scale and scope in DH activities, we distinguish between economies of scale in the production and distribution of heat. To determine whether, and to what extent, economies of scale and scope are present, we employ a parametric econometric cost function. We base our econometric cost function on a multi-product quadratic total cost model used by Mayo (1984), as given by: C = α0+ n X i αiYi+ 1 2 n X i n X j αijYiYj+ n X i m X k βkpk (6)

where C refers to the total cost, Y to outputs i, j = 1, ..., n, and p to input prices k = 1, ..., m. While Equation 6 is not homogeneous of degree one in input prices (see Greer (2011) for proof), this is no concern for the purpose of our study, as will be discussed later in this section.

Based on the cost model in Equation 6, we estimate a multi-product long run quadratic total cost function, as it is consistent with economic theory and with reasonable behavioural assumptions (Baumol et al., 1982; Saal et al., 2013). By implementing a quadratic cost function, we allow for the emergence of possible (dis)economies of scale and scope, as well as cost complementarities (Kwoka, 2002). We use panel data estimation techniques as it assists us in managing our data constraints. We will first introduce the equation before explaining each of the variables in turn.

T Ci,t= β0+ β1∗ H.P rodi,t

+ β2∗ H.Distri,t + β3∗ 1 2 ∗ H.P rod 2 i,t + β4∗ 1 2 ∗ H.Distr 2 i,t

+ β5∗ H.P rodi,t∗ H.Distri,t

+ µt+ νi,t

(7)

(24)

term in output is multiplied by a half for simplicity. To understand why the interaction term is not multiplied by a half, see Equation 6. µtis a dummy variable to account for any yearly patterns. νi,t

is the composite error term containing firm fixed effects αiand the idiosyncratic error term υi,t. β0

is a constant, and the subscripts i and t refer to firms and time in a yearly interval, respectively. In practice, the total cost of production and distribution of DH are affected by the input prices of fuel, labour, and capital. However, these input prices are not included in Equation 7 for several reasons. To accurately include the input price of fuel in the cost function, we need complete infor-mation on the type of heat production technology and fuel mix of each DH firm. However, Swedish DH firms are not required to provide this information. Similar to the issue of the fuel input price, no data is available to accurately estimate the price of capital and labour. As input prices cannot be explicitly incorporated in the cost function, we consider them to be included in the fixed components and the estimated coefficients (Martins et al., 2012).

Given the discussion on cost function theory, we first of all expect that the estimated coefficients of the output variables, β1and β2, are positive, as an increase in output should also always increase total

cost (i.e. monotonicity in output). Second, we expect that the signs of the estimated coefficients of the squared output terms, β3and β4, are negative, which gives us an initial indication of the presence

of economies of scale in the production and/or the distribution of DH. While the quadratic sign of the output variables gives us an initial indication of the presence of economies of scale, through use of estimated coefficients from the regression, Equations 1-3, and the data itself, we are able to estimate and test for overall and product-specific economies of scale. Third, we expect that the sign of the coefficient of the interaction variable, β5, is negative, as this would indicate cost complementarities,

and hence provide evidence of economies of scope in the production and distribution of heat. 4.1.2 Estimating environmental externalities

To estimate environmental externalities in the production of heat, we employ a simplified approach of environmental cost accounting as suggested by Fahl´en and Ahlgren (2012). In our approach, we only account for the environmental costs from the core combustion of fuels in heating plants, as they account for the majority (Dones and Heck, 2005). However, before we are able to estimate the environmental costs of heat production, we must first set a benchmark for which we can compare the environmental costs of heat production to. Since the price of heat is linked to the gas price due to the NMDA principle in the Netherlands, we consider the environmental fees and taxes included in the price per gigajoule (GJ) of gas as the benchmark.

Having set the benchmark, we estimate the short run marginal environmental costs of a variety of heat production technologies based on the initial loadings of pollutants, multiplied by the shadow prices of the social cost of pollution. As DH systems can make use of a variety of heat sources and fuels, the environmental costs of producing heat vary. We consider the environmental costs of heat production for heat only boilers (ECHOB), CHP plants (ECCHP), and heat pumps (ECHP) per

GJ of produced heat using a variety of fuels and emitting a variety of pollutants, as depicted in Equations 8-10.

For CHP plants and heat pumps, we take into consideration the avoided and increased environ-mental costs of marginal power production (ECM P) (Henning and Carlson, 2002). Since a CHP

(25)

of marginal power production are estimated according to Equation 11. It is also important to note that, for industrial waste heat, the environmental costs that arise from industrial processes are not considered to be associated with industrial waste heat itself.

In the environmental cost estimation of the production of heat in the Netherlands, we distinguish between the emission of sulphur dioxide (SO2), nitrogen oxides (N Ox), methane (CH4), nitrous

oxide (N2O), ultra-fine particulate matter (P M2.5), and carbon dioxide (CO2). Given that a variety

of heat production technologies, fuels, and pollutants are considered, we thereby assume that the environmental costs are fully covered in the estimations.

ECHOB= P iCems,i∗ qems,i,HOB ηh,HOB (8) ECCHP = P iCems,i∗ qems,i,CHP ηh,CHP − ηe,CHP ηh,CHP ∗ ECM P (9) ECHP = 1 ηh,HP ∗ ECM P (10) ECM P = P iCems,i∗ qems,i,M P ηe,M P (11) In Equations 8-11, EC refers to the environmental cost per unit of output. HOB, HP , and CHP denote the type of heat production technology in the form heat only boiler, heat pump, and combined heat and power, respectively. Cems,irefers to the cost of emission i per unit of output, qems,irefers

to the amounts of emission i per production technology per unit of output, ηh refers to the heat

efficiency of each heat production technology, ηe,CHP refers to the electrical efficiency in CHP plants,

and ηe,M P refers to the electrical efficiency in a marginal power production plant.

4.2 Estimating the competitiveness of renewable heat supply

4.2.1 Estimating short run competitiveness of renewable heat supply

To investigate the competitiveness of renewable heat suppy, we estimate the SRMC of a variety of heat production technologies and fuels. In estimating the SRMC, we consider the cost of fuel, the variable operation and maintenance costs, the environmental cost, the revenue from selling the co-generated electricity in CHP plants, and the efficiency of the heat production technology. We consider the same set of heat production technologies as was included in the estimation of environmental externalities. For CHP plants, there is the issue of cost allocation between heat production and electricity production. In order to resolve this issue, following the work of Sun et al. (2016), we first set the price of electricity, and then estimate the SRMC of heat accordingly. In this way, all the SRMC are allocated to the production of heat, and the short run marginal benefit from selling electricity at the market price is deducted from the SRMC.

Based on the above, we estimate the SRMCs of heat production technologies as follows: SRM Ci= CF,i+ CV O&M,i ηh,i −ηe,i ηh,i ∗ Pe+ ECi. (12)

In Equation 12, SRM Ci is the short run marginal cost of heat production technology i, CF,i is the

cost of fuel per unit of output, CV O&M,i is the variable operation and maintenance costs per per

unit of output, ηh,i refers to the heat efficiency, ηe,i is the electrical efficiency (0 if not applicable),

(26)

Following the estimation of the SRMC of different heat production technologies, we rank the SRMC of different heat production technologies in ascending order.

4.2.2 Estimating long run competitiveness of renewable heat supply

To investigate the long run competitiveness of renewable heat supply, we estimate the long run average cost per unit of energy output (LCOE) and per unit of capacity (ARR). We consider the same set of heat production technologies as during our estimation of environmental externalities and the SRMC.

In estimating the LCOE, we consider the yearly investment costs, fixed operation and mainte-nance costs, and the yearly heat production in addition to the factors that were considered in the SRMC estimation. We then estimate the present value of the average cost per unit of energy that a heat production technology is expected to produce over its lifetime as follows:

LCOEi,0=

Pn

t=0

T CI,i,t+T CF O&M,i,t+SRM Ci∗Hi,t

(1+r)t Pn t=0 Hi,t (1+r)t (13)

where LCOEiis the Levelised Cost of Energy of heat production technology i at time t = 0, T CI,i,t

is the total investment costs in any year t, T CF O&M,i,t is the total fixed operation and maintenance

costs, SRM Ci is the short run marginal cost as estimated in Equation 12, and Hi,t represents the

total heat produced. The expected economic lifetime of the production plant is denoted by n years, and r refers to the discount factor in the form of the Weighted Average Cost of Capital (WACC).

In estimating the ARR, we consider the same costs as in Equation 13 but in a different form. We estimate the ARR as a function of load hours as follows:

ARR(LH)i= AICi+ T CF O&M,i+ SRM Ci∗ LH (14)

where ARRiis the Annual Revenue Requirement per unit of capacity of heat production technology

i, AICi is the annualised investment costs per unit of capacity based on the expected economic

lifetime of the production plant and the WACC, and T CF O&M,i is the total annual fixed operation

and maintenance costs. The variable component consists of the short run marginal cost SRM Ci,

as estimated in Equation 12, as a function of load hours LH assuming that load hours involve full utilisation of capacity.

Following the estimation of the LCOE and the ARR as a function of load hours, we rank the LCOE and ARR of different heat production technologies in ascending order (assuming equal number of load hours). In addition, we represent the ARR as a function of load hours in a similar fashion as in Figure 6.

5 Data

(27)

5.1 Cost function data

In this study, we use publicly available Swedish DH firm data from the Swedish Energy Market Inspectorate (2020). The unbalanced panel consists of consolidated DH firms in Sweden over a time period of 2013 to 2018. Unfortunately, we were not able to extend the study period before or beyond the currently considered time frame. We carefully eliminated any DH firm with clearly unreasonable observations (e.g. distribution length in meters instead of kilometers) or that no longer exist (e.g. due to mergers or acquisition) by individually checking each of the included DH firms within the data set. After these filtering steps, we obtain a final sample consisting of a total of 211 consolidated DH firms in the considered time period.

The selection of the total cost and output variables used to estimate the multi-product quadratic total cost function was guided by the literature review and the availability of the data. The dependent variable is the total operating cost in euros per year. The two outputs used are (1) the amount of heat supplied to the network as a proxy for the amount produced in gigawatt hour (GWh) per year and (2) the amount of heat sold in GWh per year as a proxy for the amount of heat distributed.

The summary statistics for the employed variables can be found in Table 2. As can be seen in Table 2, there are no DH firms that produce only heat, based on the observation that the minimum of total heat distributed is not equal to zero. Additional tables concerning the summary statistics by year and group can be found in Appendix B. Note that we converted the cost data from Swedish krona to euro using annual average exchange rate data from the central bank of Sweden (see Appendix C).

Table 2: Summary statistics of DH firms in Sweden as included in the sample

Variable Obs Mean Std. Dev. Min Max

Total cost (ke/year) 1,137 17981.64 55252.95 95.765 614336.9

Total heat produced (GWh/year) 1,137 278.490 805.571 0 8694

Total heat distributed (GWh/year) 1,137 261.285 754.676 0.6 8237.1

Year 1,265 2015.502 1.707 2013 2018

Multiple output cost functions, used for the estimation of economies of scale and scope, often rely on output variables that are highly correlated (Saal et al., 2013). We find that this also holds true for our data. The correlation between the employed variables can be found in Table 3. The scatter matrix in Appendix B provides additional information regarding the relationship between the employed variables.

Table 3: Correlation matrix of the dependent and independent variables considered in the estimation of the cost function

Total cost Total heat produced

Total heat distributed

Total cost (ke/year) 1

Total heat produced (GWh/year) 0.9799 1

(28)

5.2 Cost estimation data

In estimating environmental externalities, the SRMC, LCOE, and ARR of heat production tech-nologies, an extensive set of input parameters and assumptions is required. An overview of the type of data, its application, and the source can be found in Table 4. The full data set and assumptions can be found in Appendix C.

Table 4: Overview of the input parameters and assumptions in estimating the environmental ex-ternalities, short run marginal costs, levelised cost of energy, and annual revenue requirement as included in Appendix C

Type of data Contains Year Can be found in Source

Technical Data on the typical yearly load hours, capacity, efficiency, and lifetime per heat production technology

2020 Table 11 Danish Energy

Agency (2020)

Financial Data on the fixed and variable costs per heat production technology

Agency (2020)

Emission Data on the quantity of pollution heat per production technology

Agency (2020)

Fuel price Data on fuel prices per fuel type

Agency (2020) Pollution price Data on pollution prices

per pollution type

2015 Table 15 CE Delft (2017)

Other Data and assumptions on

Competitiveness of renewable heat supply in well-functioning district heating markets in the Netherlands

Competitiveness of renewable heat supply in well-functioning

district heating markets in the Netherlands

Contents

1

Introduction

2

Background

2.1

Operation and characteristics of district heating

2.2

Characteristics of district heating in the Netherlands

2.3

Comparison of district heating in the Netherlands and Sweden

3

Theory

3.1

Benchmark for market efficiency

3.2

Market failures

3.3

Economies of scale and scope

3.4

Environmental externalities

3.5

Competitiveness of renewable heat supply

4

Methodology

4.1

Estimating market failures

4.2

Estimating the competitiveness of renewable heat supply

5

Data

5.1

Cost function data

5.2

Cost estimation data