Ownership and district heating prices:
The case of an unregulated natural monopoly
Alejandro Egüez alejandro.eguez@umu.se Umeå Economic Studies No. 980
Centre for Environmental and Resource Economics (CERE) Department of Economics
Umeå School of Business, Economics and Statistics (USBE) Umeå University
Abstract
The price of district heating in Sweden is unregulated and differs substantially among different networks. This paper investigates if the price variation can partly be explained by ownership status, i.e., whether the network companies are privately- or municipally-owned. The empirical analysis is based on data on district heating prices, ownership status, and network characteristics for the period 2012-2017. The results show that prices are higher in privately-owned district heating networks than in municipally-owned networks, especially in the fixed component of the price. It is argued that municipal and private companies’
divergent objectives may be part of the explanation for these differences. Finally, district heating prices are positively correlated with the market prices for heat pumps, regardless of ownership, which suggests a general price-setting strategy based on the price of substitutes.
JEL Codes: Q41, L11, L33, L43, L95, D42
Keywords: district heating prices, ownership, public versus private, natural monopoly, two-part tariff
Highlights:
• Data confirms that district heating prices are slightly higher in private networks than in municipally-owned networks.
• The price differential is mainly in the fixed price component.
• The different objectives of municipal and private companies may partially explain these differences.
• District heating prices are correlated with the market prices of heat pumps.
1. Introduction
The district heating sector in Sweden was liberalized in 1996, which made it possible for different actors to enter the market. Many networks remained under municipal ownership, but some were sold to private companies (Magnusson, 2016).1 District heating companies are free to set their prices, as prices are not subject to price regulation (Werner, 2017). The main objective of this study is to ascertain if ownership status affects district heating prices in Sweden.
District heating prices typically have a two-part tariff structure consisting of a fixed fee and a variable component. Prices differ significantly among networks.
For example, the overall price in Stockholm is 18% higher than in Gothenburg and up to 70% higher than in Luleå, which raises the question of the source of these differences.2 The exercise of market power may be possible in some networks, but other factors can also explain these differences. Previous studies have found lower prices in municipal-owned networks than in privately-owned networks (Åberg et al., 2016; Andersson & Werner, 2003, 2005, 2001; Colnerud Granström, 2011; Muren, 2011). However, those studies did not empirically test the underlying reasons for the differences. This paper takes a step further by splitting district heating prices into their fixed and variable components and testing the effect of ownership.
Previous studies have asserted that district heating networks can be considered local natural monopolies, implying potential market power (Joskow, 2007;
Lundgren et al., 2013; Söderholm & Wårell, 2011; Wissner, 2014). According to standard economic theory, a private company is assumed to maximize its profit and, therefore, engage in monopoly pricing. A monopolist may use the fixed fee of a two-part tariff to extract some or all of the consumer surplus. This implies
1 Each network is considered as a different price zone. Networks do not necessarily follow municipal borders.
2 These figures correspond to a typical multi-dwelling building with an annual heating
requirement of 193 MWh. The corresponding monthly bill for district heating in an apartment in Stockholm is 983 SEK, while for a similar apartment in Gothenburg it is 834 SEK and in Luleå 579 SEK (1 EUR = 10 SEK).
See http://nilsholgersson.nu/rapporter/rapport-2019/fjarrvarme-2019/
that output increases towards the efficient output level, and that the variable price moves towards the marginal cost. If consumer demand is homogenous (identical over consumers), a two-part tariff of this type will lead to a variable price equal to marginal cost and consequently to an efficient output level (Cabral, 2017). In practice, the variable price will be somewhere between the monopoly price and the marginal cost because the demand differs among consumers (Waldman & Jensen, 2016). A monopolist without profit-maximizing objectives but with the aim to recover their costs may use the fixed price component to at least cover their fixed costs and avoid losses.
A price differential in the fixed component of the tariff can reveal differences in how much private and municipal companies aim to capture consumer surplus based on their objectives. Therefore, analyzing the effect of ownership on the fixed and variable components of district heating prices is a useful approach to investigate whether the companies’ objectives, characterized by their ownership type, explain the price differentials between Swedish district heating networks.
The present study tested the following hypotheses: H1) district heating prices are higher in privately-owned networks than in municipally-owned ones; and H2) the price differential in H1 corresponds to the fixed component of the price. A reduced form approach was used based on a panel data set from the period 2012 to 2017.
Separate analyses were conducted for multi-dwelling buildings (MDB) and single- family houses (SFH).
Municipal and private companies can differ in their pricing strategies due to different objective functions.3 It is worth noting that municipal energy companies in Sweden are not exempt from profit generation because, in compliance with the Swedish District Heating Act, they are to be run in a businesslike manner, which suggests that profit generation is one objective. However, these companies may have broader social objectives that drive them to sacrifice a certain level of profit to achieve the objectives (Åberg et al., 2016; Broberg & Egüez, 2018; Kaufman,
3 Municipal and state ownership are two types of public ownership, but they can pursue different objectives. Hereinafter, “public” and “municipal” ownership will be used interchangeably. When necessary, state ownership will be mentioned explicitly.
1990; Nygårds, 2011). Therefore, profit generation may not necessarily be pursued as in a profit-maximizing private company, especially if the municipality’s guidelines discourage high prices that can harm inhabitants’ interests. Economic theory assumes that the nature of private companies is to be profit-maximizing.
However, the companies may find that it is not always optimal to set full monopoly prices. First, district heating companies compete with other companies on the broader heating market and may lose some of their market share to other heating forms, at least in the long run. Second, companies face the threat of future regulation (Bonev et al., 2018, 2019).
Private and municipal companies could also differ in how effectively the owners can control the business (Boardman & Vining, 1989; Nikogosian & Veith, 2012;
Roland & Stiglitz, 2008; Yarrow et al., 1986). The managers of private companies respond to shareholders, while the managers of public companies respond to the public, often represented by elected politicians. The control mechanisms in public companies may not be as effective as in private companies, where the stock price reflects the expected stream of dividends and, as such, reflects the expectations on the company’s profits. Under the assumption that a private company seeks to maximize profits, managerial performance will be reflected in the stock price. This context gives the principals (shareholders) strong incentives to monitor agents, e.g., by an incentive scheme based on the company’s stock price. The managers of municipal companies, by contrast, are subject to public and political scrutiny (Peltzman, 1971). Their performance depends on how they manage the tradeoff between keeping prices low (affecting profitability) and making some profit.
Municipal companies also have an incentive to be cost-efficient so as to generate profits without increasing prices. If the managers of public and private companies, the agents, strive to satisfy their principals’ interests, they may also be concerned about cost-efficiency. Nevertheless, municipal companies may be more likely to set lower prices if their primary purpose is not one of maximizing profit, as in the case of private companies.4
4 Municipal companies may also have energy conservation objectives and therefore set higher prices.
In regards to price regulation, it is argued that district heating competes with other heating forms, leaving little room for district heating companies to exert market power in the long run, and as a consequence, prices do not need to be regulated. It is also argued that district heating companies use a price-setting strategy based on the cost of district heating substitutes and have moved away from cost-recovery pricing, which was the pricing method before liberalization (Birgersson, 2004). However, high switching costs, lock-in effects, and technical limitations may challenge the real potential for substitution (Söderholm & Wårell, 2011; Hellmer, 2010).5 For example, ground source heat pumps require large areas for energy wells that are not always accessible (Åberg et al., 2020). A self- regulation platform known as the “Price Dialogue” (PD) has been formed as an alternative to price regulation. District heating and real estate companies meet voluntarily on this platform to discuss future prices (Abrahamsson & Schrammel, 2016). This paper contributes to the existing literature by incorporating the PD and variables that capture the competition of heat pumps with district heating into the analysis.
The paper will proceed as follows. Section 2 presents an overview of the Swedish district heating sector. Section 3 is a review of the empirical literature. Section 4 covers the empirical model. Section 5 describes the data and its sources. Section 6 contains the results. Section 7 covers the main conclusions and discussion.
2. The district heating market in Sweden
District heating has the largest market share for heating in the residential sector in Sweden, with half of the heat supply to MDB and SFH, followed by nearly 25% from heat pumps (Swedish Energy Agency, 2020; Werner, 2017). District heating in Sweden represents 13% of the total final energy consumption, which
5 From a consumer perspective, one difference between district heating and heat pumps is that the latter require larger investments in the beginning but have lower running costs. Conversely, the energy company invests in the district heating network, so substantial investments are not required from the consumer if the network is already in place. However, running costs are generally higher than with heat pumps.
translates into about 50 TWh in 2018, of which MDB consumed 24 TWh and SFH 5.5 TWh.6 The production of district heating used 61 TWh as input energy, of which 62% was biomass (including biomass from household and industrial waste), 23% fossil fuels,7 8% industrial processes (waste heat), and 7% heat pumps8 (Swedish Energy Agency, 2020).9
Since the liberalization of the market in 1996, several mergers and acquisitions have occurred, and ownership has been diversified (Magnusson, 2016). Different laws guide the operations of district heating companies and, implicitly, the price levels. The distribution of district heating is considered a natural monopoly, and the Swedish Competition Act (Ch. 2, 7§) explicitly prohibits the use of market power. The Swedish Companies Act (Ch. 3, 3§) establishes that if the purpose of the company is, totally or partially, different from the generation of profits to its shareholders, that should be stated explicitly in the company’s articles of association.10 This is how many municipally-owned energy companies operating in line with the municipality’s guidelines divert from profit maximization.11 At the same time, the Swedish District Heating Act (38§) has established that municipal district heating companies should be managed in a businesslike manner.
District heating prices typically consist of a fixed fee and a variable component.
The fixed part is typically determined by the customer’s load demand (in MW), and the variable part is determined by their energy consumption (in MWh).
6 The final energy consumption was 373 TWh in 2018, of which industry consumed 141 TWh, transport 84 TWh, and the residential and service sectors consumed 147 TWh. These figures exclude losses from nuclear energy for 125 TWh (Swedish Energy Agency, 2020).
7 9.2 TWh mainly accounts for the fossil part of household waste, industrial fossil waste, and peat; 2.2 TWh coal; 1.3 TWh natural gas; and 1.1 TWh petroleum products.
8 These are large heat pumps installed in district heating systems and should not be confused with domestic heat pumps.
9 Waste is an important fuel not only for district heating but also for the co-production of electricity in combined heat and power (CHP) plants. The total energy production from waste is 17.5 TWh, from which 2.2 TWh is used to produce electricity and 15.3 TWh district heating.
By 2018, half of the treated household waste in Sweden was utilized for energy recovery (Avfall Sverige AB, 2019).
10 In Swedish: bolagsordningen. See Aktiebolagslagen 3 kap. 3 §.
11 The Swedish Local Government Act (Ch. 2, 7§) establishes that municipal companies should not be profit-driven and should provide utility to the members of the municipality, i.e., its inhabitants.
Seasonal variations generally apply, whereby the price is usually higher in the winter than in the summer, when the peak demand is lower. There may be additional price components, such as price adjustments based on the return water temperature to incentivize users to install efficient heat exchangers (Song et al., 2016). Companies are free to discriminate between types of consumers and establish price menus with different alternatives for the fixed and variable components of the price. District heating companies primarily use a fixed fee, a variable component, and seasonal variations to increase their ability to compete with heat pumps (Swedish Energy Agency, 2015). In the absence of district heating’s fixed charges and seasonal adjustments, consumers may find it attractive to install heat pumps as a complement to district heating. They would then use heat pumps as the main heating source in the winter and district heating in the summer.
In the absence of price regulation, the need for increased transparency in district heating costs and prices has been a matter of concern for authorities such as the Swedish Energy Agency and the Swedish Competition Authority (SOU 2003:115).12 As a response to this need, the PD was established in 2013 by the energy and real estate industries. This self-regulation platform facilitates discussions between district heating companies and consumer representatives about possible price changes.13 Energy companies voluntarily apply to become members of the PD. Membership requires that companies undertake a consultation process with customer representatives before enacting prices changes.14 The PD aims to increase the degree of transparency and trust between district heating companies and their customers, mainly the owners of multi- dwelling buildings. The PD is a unique feature of the Swedish district heating
12 See also the report (in Swedish): ”Fjärrvärme på värmemarknaden –Rapport över uppdrag att följa utvecklingen på fjärrvärmemarknaden till regeringen (dnr N1999/11368/ESB).”
13 Four organizations have led the PD: 1) Swedenergy, an association of energy companies; 2) Riksbyggen, one of the largest private real estate cooperatives in Sweden; 3) Sveriges
Allmännytta (formerly SABO), the Swedish Association of Public Housing Companies; and 4) Fastighetsägarna, a federation representing real estate owners.
14 The consultation process begins with an information meeting where the energy company presents its price policy and explains the elements that form the district heating price. In subsequent meetings, the customer representatives express their opinions on the proposed price change. Finally, the actual price change is presented in a closing meeting.
sector. As a voluntary and self-regulatory instrument, the PD could be viewed, at least partially, as an alternative to price regulation. Nevertheless, prices between district heating networks in Sweden differ significantly, and the effect of the PD remains unclear (Abrahamsson & Schrammel, 2016).
3. Review of the empirical literature
Evidence from other sectors, such as electricity, water, refuse collection, and airlines, mostly supports the hypothesis that private companies set higher prices than public companies (Branston, 2000; Fiorio & Florio, 2013; García-Valiñas et al., 2013; Martínez-Espiñeira et al., 2009; Meyer, 1975; Peltzman, 1971; Percebois
& Wright, 2001; Ruester & Zschille, 2010; Porcher, 2017). The literature also focuses on the effect of ownership on efficiency and performance by examining costs and profitability, respectively. A reason for not analyzing the effect of ownership on prices may be that prices tend to be regulated in certain natural monopolies, so the focus shifts to efficiency.15 However, if prices are unregulated, it is also important from a societal perspective to consider the differences in prices as monopoly power, if present, also affects consumers through changes in the consumer surplus.
District heating prices are regulated in countries such as Denmark and Lithuania, but remain unregulated in Sweden and Finland (Aronsson & Hellmer, 2009;
Gatautis, 2004; Grohnheit & Mortensen, 2003). Deregulation in Finland occurred in 1999, and Linden & Peltola-Ojala (2010) have examined the effects of deregulation on Finland’s district heating prices using data from 1996 to 2002.
The results of a fixed effects GMM estimation show that the price difference between publicly- and privately-owned heating companies on average was 5.34 EUR/MWh and 2.74 EUR/MWh depending on whether the customers bought energy for small private houses or MDBs. These differences represented 10% and
15 For further references to efficiency comparisons between public and private companies in sectors such as water, electricity, refuse collection, and other sectors, see: Boardman and Vining (1989); Megginson and Netter (2001); Renzetti and Dupont (2003); Shirley and Walsh (2000);
Suárez-Varela et al., (2017); Vining and Boardman (1992); and Yarrow et al., (1986).
7% of the average prices in the sample, which was 53.72 EUR/MWh for small private houses and 41 EUR/MWh for MDBs.
A number of studies have investigated the role of ownership in Swedish district heating pricing (Åberg et al., 2016; Andersson & Werner, 2003, 2005; Colnerud Granström, 2011; Muren 2011). Muren (2011) has investigated the differences in district heating prices between large companies (defined as those with a presence in more than five municipalities) and small companies during the period 2000- 2009.16 Mergers and acquisition processes have led to the expansion of large companies. Prior to liberalization, energy companies functioned within their municipal jurisdiction. Muren has found evidence that prices in large companies are 3-4% higher than in small, local, companies. These results stand in contrast to the expected price effect of economies of scale and suggest that other factors are at play to explain price differences.
Colnerud Granström (2011) has analyzed cross-sectional data from 2009 and found that the price on average was 0.13 SEK/kWh higher for SFHs in the networks of privately-owned companies (1 EUR = 10 SEK).17 In 2009, the average price of district heating for SFH was 0.63 SEK/kWh. There was no statistically significant evidence of differences in prices between privately- and municipally- owned networks for MDB. The reason the ownership effect on prices only holds statistical significance for SFH is not clear. The present study also analyzed SFH and MDB but used a broader and more updated panel data set than Colnerud Granström (2011).
Åberg et al. (2016) have observed a difference in the average prices of district heating from municipally-owned, privately-owned, and state-owned networks.
The study is based on descriptive statistics of data from the Energy Markets Inspectorate and the Swedish District Heating Association. The study also
16 Large companies include E.On (private), Fortum Värme AB (mixed: private and municipal), Vattenfall AB (state), Rindi Energi AB (private), Neova AB (mixed: private and state), and Skellefteå Kraft AB (municipal).
17 Average exchange rate 2016-2019 (SCB Statistics Sweden, 2020).
conducted a survey to understand these differences. The authors conclude that differences in organizational strategies and goals can explain lower prices in municipal companies. They assert that, although district heating prices are unregulated, municipal companies still largely use cost-recovery pricing and prioritize political goals over financial goals. This argument alligns with the finding in Ganslandt (2010) that private district heating companies have better profitability indicators than their public peers.
Andersson and Werner (2003, 2005) have used multivariate linear regressions and found that the price of district heating was 3.4% and 4.9% higher in privately- owned networks than in networks with at least 50% municipal ownership for the years 2001 and 2003, respectively. In a governmental report, Birgersson (2004) has observed that the district heating prices in privately- and state-owned companies were 5-20% higher than in networks owned by municipal companies in large cities. The report highlights that, after liberalization, district heating companies became more inclined towards using a value-based pricing strategy and to setting prices according to the costs of district heating’s substitutes rather than setting prices to cover costs, which was the practice prior to liberalization. If prices are based on the cost of substitutes, it is likely that district heating prices will positively correlate with electricity prices and negatively correlate with the efficiency of heat pumps.
Public companies may prioritize social objectives, such as minimizing the environmental impact of their production. Lundgren et al. (2013) have used data from 2004 to 2010 to determine if there were differences in environmental performance and productivity between publicly- and privately-owned district heating companies. The results do not show any statistically significant difference.
However, they do reveal that environmental performance and productivity in private companies were more sensitive to policies such as carbon taxes and the EU emissions trading system (EU ETS). The policy compliance of public companies may not drive them to protect the environment. Instead, they may safeguard the environment in line with their statutory objectives.
Overall, previous studies have found that district heating prices are higher in privately-owned networks than in municipally-owned ones. This paper contributes to the literature by analyzing the effect of ownership on district heating’s fixed and variable components for MDB and SFH. This analysis provides greater insights into how district heating companies set prices depending on ownership type. The study also adds to the empirical evidence by using updated data and additional control variables such as the cost of heating with heat pumps and participation in the PD.
4. The empirical model
A reduced form model was used to test the effect of ownership on district heating prices. The dependent variable, the price of district heating, can be viewed as an equilibrium price that is determined by factors affecting the demand, and hence the marginal revenue, as well as the marginal and average cost functions. These variables, together with the companies’ managerial objectives, determines the equilibrium price of district heating.
Factors that were assumed to affect the demand for heating are income, temperature measured by the number of heating degree days, and the price of substitutes, which was determined by the price of electricity and investment costs.
Factors that affect costs are primary energy prices and network characteristics such as length, age, and urban composition. The effect of ownership can be the result of different objective functions. However, it is also possible that ownership affects costs (as a result of differences in managerial efficiency, for example), and hence the equilibrium price. The former effect was assumed to be captured by dummy variables for ownership, whereas the latter was assumed to be captured by an interaction between the cost variable and the ownership dummies. The reason the PD was included as a control variable is that it is a platform for supply and demand sides to discuss district heating prices. Nevertheless, the effect of PD on prices remains an empirical question. The model includes an interaction variable between the PD and ownership dummies.
Based on the discussion above, the econometric specification in equation (1) forms the empirical analysis basis.
𝑃𝑖𝑡 = 𝛽0+ 𝛽1𝑑_𝑂𝑖𝑡+ 𝛽2𝑑_𝑃𝐷𝑖𝑡+ 𝛽3𝑑_𝑂𝑖𝑡× 𝑃𝐷𝑖𝑡+ 𝛽4𝐸𝐿𝑖𝑡+ 𝛽5𝐻𝑃𝑖+ 𝛽6𝑌𝑖𝑡
+ 𝛽7𝐻𝐷𝐷𝑖𝑡+ 𝛽8𝐶𝑖𝑡+ 𝛽9𝑑_𝑂𝑖𝑡× 𝐶𝑖𝑡+𝛽10𝑑_𝑌𝑒𝑎𝑟𝑖+ 𝜀𝑖𝑡 Eq. (1)
Where:
𝑖 = District heating network 𝑡 = Years (2012-2017)
𝑃 = District heating price (SEK/kWh) 𝑑_𝑂 = Ownership dummy
𝑑_𝑃𝐷 = Price Dialogue dummy
𝑑_𝑂 × 𝑃𝐷 = Interaction between ownership and PD 𝐸𝐿 = Electricity price (SEK/kWh)
𝐻𝑃 = Installation costs of a vertical brine-water heat pump (SEK) 𝑌 = Disposable income (1,000 SEK)
𝐻𝐷𝐷 = Heating degree days
𝐶 = Operation and maintenance costs, including fuels (SEK/kWh) 𝑑_𝑂 × 𝐶 = Interaction between ownership and O&M costs
𝑑_𝑌𝑒𝑎𝑟 = Year dummies
The model specification in equation (1) can be used to test whether ownership has an effect on district heating prices for MDB and SFH, after controlling for covariates. Unlike previous studies, the full price and its fixed and variable components were estimated as separate dependent variables to test if the effect of ownership differs between the two parts of the tariff. The choice between a random or fixed effects model depends on the research question. As the overall objective of this study is to estimate the effect of ownership that is mostly time- invariant (see Table 2), random effects models are more relevant even if the Hausman specification tests favor a fixed effects model (Petersen, 2012).
The specification in equation (1) was used for the main model (Model I).
Additional considerations of the explanatory variables for costs led to two alternative specifications (Models II and III). Unlike network characteristics, such as length, age, and urban composition, data on input prices, such as primary energy prices and wages, was not part of the data set, as explained in section 5.
Instead, aggregate data on operation and maintenance (O&M) costs were available for production and distribution. Production O&M costs were assumed to mainly contain the costs of primary energy inputs. With this assumption, production-O&M costs can be included together with the network’s length, age, and urban composition (Model II).18 Alternatively, to test the effect of energy inputs and the co-production of electricity on district heating prices, the network characteristics can be included together with the percentages of selected fuels, such as biomass and waste, over the total energy input used to produce district heating, with and without the co-production of electricity in combined heat and power (CHP) plants (Model III).
5. Data
A panel data set for the period 2012-2017 was used in this study, including observations from 440 district heating networks in Sweden. District heating prices, including VAT, were obtained from the Swedish Energy Markets Inspectorate. The price data contains a yearly aggregate (in SEK) divided into a fixed part and a variable part for typical annual heat requirements of 193 MWh for MDB and 20 MWh for SFH. The aggregate, fixed, and variable components were averaged for these typical annual heat requirements to obtain prices per kWh.19
18 Assuming that the production function of district heating is 𝑔(𝐹, 𝑁) = 𝑎(𝑁) × 𝐹, where 𝑁 are the network’s characteristics, and 𝐹 are the fuel inputs (the only variable inputs). The variable cost is 𝐶 = 𝑟𝐹, where r is fuel prices. The variable cost function becomes 𝐶(𝑟, 𝑞; 𝑁) = 𝑟(𝑎(𝑁)1 )𝑞, where 𝑞 represents output. Since fuel inputs are the only (variable) input,
MC=VAC=𝑟(𝑎(𝑁)1 ). Given this production function, network characteristics such as age, length, and urban composition can be included as regressors together with the costs.
19 The Nils Holgersson reports consider an annual heat requirement of 193 MWh for MDB, which corresponds to a heated area of approximately 1,000 m2, e.g., a MDB of 15 apartments with a heated area of 67 MWh per apartment, including common areas. For reference, the most common apartment in the three main cities in Sweden (Stockholm, Gothenburg, and Malmö) is a two-room apartment with an average size of 57 m2, excluding common areas. See
Information about the ownership status of district heating companies is available in Magnusson (2015).20 Table 1 in the present study shows that, by 2017, 193 energy companies operated 440 district heating networks in Sweden, with a total supply of 51 TWh. Nearly half of the heat was supplied to MDB, 10% to SFH, and the remaining percentage to commercial premises (Swedish Energy Markets Inspectorate, 2019b). Of these, 139 municipally-owned companies supplied two- thirds of the heat, 34 privately-owned companies supplied 14%, and the remaining quarter was supplied by 20 companies that were state- or mixed-owned.
Ownership was grouped into three categories: municipal, private, and state- and mixed-owned companies.21
Table 1:
Ownership mix in the production of district heating in Sweden (2017)
Ownership # of
companies # of
networks
Total supply (GWh)
Supply to MDB
(GWh)
Supply to SFH (GWh)
Municipal 139 289 31,709 13,968 3,938
Private 34 105 6,887 3,468 621
State22 2 14 2,711 1,484 213
Mixed (private and municipal) 12 14 932 396 89
Mixed (private and state)23 1 10 - - -
Mixed (state and municipal)24 5 8 8,692 4,967 173
TOTAL 193 440 50,931 24,283 5,033
Table 2 shows the average prices of district heating by ownership type for MDB and SFH during 2012-2017. On average, a 193 MWh MDB had an annual district heating bill of 164,050 SEK, whereof 38,600 SEK was the fixed fee. The average
https://www.scb.se/hitta-statistik/artiklar/2016/Vanligast-med-2-rum-och-kok-pa-57- kvadratmeter/
20 The ownership status for the period 2012-2017 was verified with the companies via e-mail communication.
21 Fifteen networks owned by cooperatives and foundations were excluded from the original data set.
22 Swedish Vattenfall and Norwegian Statkraft.
23 Neova AB, which is part of Vapo Oy, is owned 50.1% by the Finnish state and 40.9% by Suomen Energiavarat, a conglomeration of Finnish energy companies.
24 Mainly Stockholm Exergi AB (8,217 GWh), which is owned by Stockholm’s municipality, and Fortum, which is mainly owned by the Finnish state.
annual district heating bill for a 20 MWh SFH was 18,000 SEK, of which 3,600 SEK was the fixed fee.
Table 2: Mean of district heating prices (SEK/kWh) by ownership
Ownership MDB (193MWh) SFH (20MWh)
Overall Fixed
component Variable
component Overall Fixed
component Variable
component
Municipal 0.83 0.19 0.64 0.88 0.17 0.71
Private 0.89 0.23 0.66 0.92 0.21 0.71
State and mixed 0.86 0.18 0.68 0.93 0.18 0.75
Total 0.85 0.20 0.65 0.90 0.18 0.72
Table 3 provides a summary of frequencies in panel data format, i.e., it shows the percentages for the overall, between, and within group categories of the analyzed district heating networks by type of ownership. Municipal companies owned two- thirds of the analyzed networks. A percentage higher than 100% in the total of the between groups category indicates that some district heating networks had switched ownership due to mergers and acquisitions. The within group percentage illustrates the percentage of time that district heating networks belonged to a particular ownership category. These numbers indicate that changes in ownership type between 2012 and 2017 were minimal.
Table 3:
Panel data (2012-2017) frequency table by ownership type
Ownership Overall Between Within
Freq. % Freq. % %
Municipal 1,742 66 294 67 99
Private 621 24 107 24 97
State and mixed 277 10 48 11 96
Total 2,640 100 449 (n= 440) 102 98
Data on PD membership was obtained from the organization’s web portal to create a dummy variable that indicates whether or not a company is a member.25 Table 4 shows that although only 37 companies were members of the PD by 2017, they supplied more than 70% of the district heating in Sweden.26
25 See http://www.prisdialogen.se/medlemmar/
26 By 2020, the member companies had increased to 43 (76% of the supply).
Table 4:
Ownership mix of the Price Dialogue members (2017)
Ownership #
members # networks Supply
(GWh)
Municipal 31 76 20,560
Private 4 21 4,917
State (Vattenfall) 1 10 2,522
Mixed (Stockholm Exergi) 1 1 8,216
Subtotal Price Dialogue 37 108 36,216
Total members and non-members 193 440 50,931
% (Price Dialogue / Total) 19% 25% 71%
Vertical brine-water heat pumps have become the most common type of ground source heat pumps in Sweden (Johansson, 2017; Karlsson et al., 2003).27 The installation costs for vertical brine-water heat pumps and electricity prices were selected to represent the price of the substitutes for district heating. An overall positive relationship between district heating prices and the price of substitutes was expected. In practice, the price of substitutes is also affected by a heat pump’s lifetime, capital costs, and efficiency. However, these variables were not included in the econometric estimations herein. Lifetime is essentially a constant, and interest rates are unlikely to vary among municipalities. Even though interest rates may vary over time, it was assumed that they are captured by yearly fixed effects. A heat pump’s seasonal coefficient of performance (SCOP) is an indicator of its efficiency. For example, a SCOP of 4 means that 1 kWh of electricity is assumed to deliver 4 kWh of heat (Li et al., 2018). Location plays a role because external temperature affects the thermal efficiency of heat pumps. However, the SCOP was excluded from the econometric specifications because of its high correlation with heating degree days (HDD), which was used as an indicator of heating demand (Pearson correlation coefficients higher than 85%).
Electricity prices for MDB were obtained from the Nils Holgersson reports.
Electricity prices herein include the retail price, distribution (grid) fees, and taxes
27 In 2018, 33 thousand MDB had some type of heat pump; 16 thousand were ground source heat pumps, followed by 14 thousand air-to-water heat pumps. In the case of SFH, 1.37 million houses had heat pumps, from which 482 thousand were ground source heat pumps and 560 thousand air-to-air heat pumps (Swedish Energy Agency, 2019).
(VAT and energy tax) for each municipality, which were averaged based on an annual heat requirement of 193 MWh. The Nils Holgersson reports contain information about MDB. Retail and distribution prices for a 20MWh SFH, including VAT and energy taxes, were obtained from the Energy Markets Inspectorate. Unlike the Nils Holgersson reports, which are summarized by municipality, retail prices for SFH are divided into four electricity zones, and distribution prices are available for each network. This data was matched with the main municipality served by each network using the tool “Nätområden”
provided by the Swedish power system operator.28 Installation costs for heat pumps were obtained from the Swedish Refrigeration & Heat Pump Association for each municipality.29 The available data on heat pump installation costs vary between municipalities, but not over time. It was assumed that technical progress is captured by yearly fixed effects.
Income and urban composition were used as socio-economic and demographic control variables. They were obtained from Statistics Sweden, SCB. These variables were for the main municipality served by the district heating network since networks do not necessarily follow municipal borders. Income may affect the user’s willingness to pay. A positive relationship between income and district heating prices would be expected. However, a rich household may not be willing to pay more in the presence of substitutes. The average household’s annual disposable income (thousand SEK) was used as the indicator of income. The urban composition was captured by the ratio of the average number of SFH over the number of individual apartments in MDB. A positive relationship with district heating prices was expected since a lower concentration of individual apartments in MDBs in relative terms might discourage scale effects.
28 https://www.natomraden.se
29 The Swedish Refrigeration & Heat Pump Association – in Swedish: Svenska Kyl och
Värmepump Föreningen (SKVP) – conducts the PULSEN survey (see https://skvp.se/aktuellt- o-opinion/statistik/pulsen/2018-eng). This survey asks retailers and installers about the cost of a turnkey contract of a heat pump installation for a heat demand of 20 MWh per year. For MDB, the heat demand size of 193 MWh per year has been adjusted in SKVP’s online comparison tool JFK: https://skvp.se/varmepumpar/jfk-online.
Data on annual heating degree days (HDD) was used as a measure of heating requirement. The expected relationship between district heating prices and HDD is unclear. A negative relationship would be expected since an increased demand for heat may favor economies of scale. However, this relationship may also be positive because of the increased utility of heat in municipalities with colder temperatures. HDD data was obtained from Eurostat by NUTS2 regions.30 This data was subsequently matched with the main municipality served by each district heating network. Eurostat’s HDD indicator compares the average air temperature and a threshold of 18°C when the outdoor temperature is below 15°C.31
Data on the district heating networks’ size and age was obtained from the Swedish Energy Markets Inspectorate to capture some of the network and operation characteristics. The length (km) of the district heating distribution network is mainly a measure of size. Network length includes both forward and return directions. A negative relationship between length and district heating prices was expected due to economies of scale. A negative relationship between a district heating network’s age and district heating prices was expected because it is more likely that older networks have already recovered their investment costs.
However, they may require more maintenance. The original data format from the Swedish Energy Markets Inspectorate gives the percentage of the network built before the 1950s, and the percentage of the network built in the 1960s, 1970s, 1980s, 1990s, 2000s, or 2010s. To approximate the average age, a similar approach as used in Colnerud Granström (2011) was adopted, where the average age, in years, was estimated using equation (2):
𝐴𝑔𝑒 = %1950𝑠× (𝑦𝑒𝑎𝑟 − 1950) + ⋯ + %2010𝑠× (𝑦𝑒𝑎𝑟 − 2010) Eq. (2)
30 Sweden is divided into 8 NUTS2 (Nomenclature of Territorial Units for Statistics) national areas: Stockholm (SE11); East Middle Sweden (SE12); Småland, and the islands (SE21); South Sweden (SE22); West Sweden (SE23); North Middle Sweden (SE31); Middle Norrland (SE32);
and Upper Norrland (SE33). See
https://ec.europa.eu/eurostat/cache/metadata/en/nrg_chdd_esms.htm
31 If 𝑇𝑀 ≤ 15°𝐶; then [𝐻𝐷𝐷 = ∑ (18°𝐶 − 𝑇𝑖 𝑀𝑖 )]; else [𝐻𝐷𝐷 = 0], where 𝑇𝑀𝑖 is the average air temperature of the day 𝑖.
Data on operation and maintenance (O&M) costs per unit of input heat (SEK/kWh) was obtained from the Swedish Energy Markets Inspectorate. They are disaggregated into the production and distribution processes. Companies are instructed to report O&M costs, excluding fixed labor costs, which aids in capturing the variable component of O&M costs. Unlike distribution O&M costs, production O&M costs mainly include fuel costs, which contain the fuel’s purchasing, storage, treatment, and transportation costs (Swedish Energy Markets Inspectorate, 2019a). According to the descriptive statistics in Table 5, the most significant component of the total O&M costs in the data is related to production costs that include the input fuels. Data on the energy inputs (in TWh) were used to calculate the shares of selected energy inputs, such as biofuels, waste, industrial waste heat, and hot water, purchased from the industry. Biofuels and waste could be inputs for CHP and standard boilers.
Descriptive statistics for the non-categorical variables are presented in Table 5.32 Descriptive statistics on costs in the table show minimum values with 0. These refer to networks that were not continually operating in 2012-2017.
32 The variations of all variables were higher between networks than within time, except for the distribution O&M costs.
Table 5:
Descriptive statistics for non-categorical variables (2012-2017)
Variables N Mean Std. Dev. Min Max
Dependent
District heating price (SEK/kWh) for 193 MWh MDB 2,199 0.85 0.11 0.46 1.61
- Fixed component 2,187 0.2 0.13 0 0.88
- Variable component 2,187 0.65 0.16 0 1.26
District heating price (SEK/kWh) for 20 MWh SFH 2,203 0.90 0.10 0.33 1.26
- Fixed component 2,193 0.18 0.12 0 1.04
- Variable component 2,193 0.72 0.13 0 1.26
Costs of substitutes
Electricity price in a 193 MWh MDB (SEK/kWh) 2,598 1.64 0.17 1.17 2.13
Electricity price in a 20 MWh SFH (SEK/kWh) 2,598 1.30 0.10 1.02 1.58
Installation costs of a vertical brine-water heat pump
for 193 MWh (1,000 SEK) 2,598 660 21.3 555 705
Installation costs of a vertical brine-water heat pump
for 20 MWh (1,000 SEK) 2,598 124 1.84 124 133
Socio-economic and demographic
Household annual disposable income (1,000 SEK) 2,598 402 62 298 877
Heating degree days 2,598 4,305 878 3,036 6,671
Urban composition ratio (SFH / apartments in MDB) 2,598 2.02 1.28 0.11 11.37
Network and operation characteristics
Network length (km) 2,296 79 176 1 1461
Network age (years) 2,334 20 7 2 55
O&M costs
Production (including fuels) (SEK/kWh) 2,150 0.36 0.23 0 3.70
Distribution (SEK/kWh) 2,150 0.02 0.15 0 6.40
Total O&M costs (SEK/kWh) 2,150 0.39 0.27 0 6.80
Selected energy inputs (share over total inputs)
Percentage of biofuels used in CHP 2,255 0.08 0.22 0 1
Percentage of biofuels used in non-CHP boilers 2,255 0.65 0.41 0 1
Percentage of waste used in CHP 2,255 0.03 0.12 0 0.88
Percentage of waste used in non-CHP boilers 2,255 0.02 0.1 0 0.95
Percentage of industrial waste heat 2,255 0.07 0.22 0 1
Percentage of purchased hot water 2,255 0.08 0.25 0 1
6. Results
The results for the random effects models are presented in Table 6 for MDB and SFH. They show that district heating prices in private networks are 3% higher than the average price in municipally-owned networks. The coefficient 𝛽1_𝑝𝑟𝑖𝑣 was 0.0273 and 0.0248 for MDB and SFH, respectively, in the models where the overall price was the dependent variable (columns 2 and 5 in Table 6). At least one interaction effect was found to be statistically significant for SFH; the adjusted
coefficient evaluated at the mean value of the interaction variable was 0.0251.33 The average price (𝑝̅) in municipally-owned networks was 0.83 and 0.88 SEK/kWh for MDB and SFH, respectively. Therefore, 𝛽1_𝑝𝑟𝑖𝑣𝑝̅̅̅̅ =0.02730.831 = 3.3% for MDB and 𝛽1_𝑝𝑟𝑖𝑣_𝑎𝑑𝑗𝑝̅̅̅̅ =0.02510.882 = 2.8% for SFH. These ownership effects are robust to the alternative specifications in Models II and III. See Appendix A.
These results are similar to the study by Muren (2011), who has found that prices in local companies, often municipally-owned, were 3-4% lower than companies serving five or more municipalities. Previous studies, such as Colnerud Granström (2011) or Andersson and Werner (2003, 2005), have used multivariate analysis with cross-sectional data for specific years. The present study, in contrast, used panel data and other covariates in the econometric specification. Unlike Colnerud Granström (2011), who has only found a statistically significant ownership effect in SFH, the present study confirmed that prices in privately-owned networks were higher than in public networks for MDB, too. The study by Åberg et al. (2016) does not control for the effect of other variables, but their results are in line with the differences in the mean price of district heating between ownership categories.
Unlike previous research, this study examined whether the price differential was in the fixed or variable component. Results in Table 6 show that the price differential occurred in the fixed component, which on average was 17% higher in privately-owned networks for MDB and 24% for SFH, compared to the average fixed component price in municipal networks. The coefficient 𝛽1_𝑝𝑟𝑖𝑣 was 0.0502 and 0.0414 for MDB and SFH, respectively, in the model where the fixed component of the price is the dependent variable (columns 3 and 6 in Table 6).
In MDB, where at least one interaction effect was found to be statistically significant, the adjusted coefficient evaluated at the mean value of the interaction
33 Networks owned by a private company and that are members of the PD correspond to 6.76%.
The average costs in networks owned by private companies is 0.4 SEK/kWh. Therefore, the adjusted coefficient is 𝛽1_𝑝𝑟𝑖𝑣_𝑎𝑑𝑗= 𝛽1_𝑝𝑟𝑖𝑣+ 𝛽3_𝑝𝑟𝑖𝑣(0.0676) + 𝛽9_𝑝𝑟𝑖𝑣(0.4) = 0.0248 +
0.0229(0.0676) − 0.00316(0.4) = 0.0251.
variable was 0.0324.34 The average fixed price component (𝑝̅) in municipally- owned networks was 0.187 SEK/kWh for MDB and 0.17 SEK/kWh for SFH.
Therefore, 𝛽1_𝑝𝑟𝑖𝑣_𝑎𝑑𝑗𝑝̅̅̅̅ =0.03240.187 = 17.3% for MDB and 𝛽1_𝑝𝑟𝑖𝑣𝑝̅̅̅̅ =0.04140.17 = 24.4% for SFH. These results show that privately-owned companies, compared to municipally-owned ones, used the fixed component of the price to extract a larger share of the consumer surplus and thus guarantee a higher markup.
Appendix C presents a sensitivity analysis based on Model I with restricted samples for networks larger than 30, 40, and 50 kilometers. The results of these restricted models show two main differences compared to the unrestricted models in Table 6. First, the ownership effect on price increased for networks owned by privately- and municipally-owned companies. For example, in networks larger than 50 kilometers, district heating prices in private networks were 6% higher than the average price in municipally-owned networks for MDB and 4% higher for SFH.35 The corresponding difference was 3% using the unrestricted sample in Table 6.
A second difference in the restricted sample models is that the effect on prices of state- and mixed-owned networks became even higher than the effect on prices of privately-owned networks compared to municipal. In the unrestricted sample in Table 6, district heating prices in state- and mixed-owned networks were 1% and 2% higher than the average price in municipally-owned networks for MDB and SFH, respectively, while the difference between private and municipal networks
34 The adjusted coefficient is 𝛽1_𝑝𝑟𝑖𝑣_𝑎𝑑𝑗= 𝛽1_𝑝𝑟𝑖𝑣+ 𝛽3_𝑝𝑟𝑖𝑣(0.0676) + 𝛽9_𝑝𝑟𝑖𝑣(0.4) = 0.0502 − 0.0232(0.0676) − 0.0404(0.4) = 0.0324.
35 For networks larger than 50 kilometers (columns 4 and 7 in Appendix C), 𝛽1_𝑝𝑟𝑖𝑣 = 0.048 is for MDB. The average price (𝑝̅) in municipally-owned networks for MDB is 83.1. Therefore,
𝛽1_𝑝𝑟𝑖𝑣
𝑝̅̅̅̅ =0.0480.831= 5.8%. 𝛽1_𝑝𝑟𝑖𝑣= −0.0264 for SFH. However, the interaction effects were found to be statistically significant, so the adjusted coefficient evaluated at the mean value of the interaction variables is 0.0395, which was obtained in the following way. The adjustment considers that networks owned by a private company and that are members of the PD correspond to 6.76%. The average cost in networks owned by private companies is 0.4 SEK/kWh. Therefore, the adjusted coefficient is 𝛽1_𝑝𝑟𝑖𝑣+ 𝛽3_𝑝𝑟𝑖𝑣(0.0676) + 𝛽9_𝑝𝑟𝑖𝑣(0.4) =
−0.0264 + 0.0328(0.0676) − 0.159(0.4) = 0.0395. The average price (𝑝̅) in municipally-owned networks for SFH is 88.2. Therefore, 𝛽1_𝑝𝑟𝑖𝑣_𝑎𝑑𝑗𝑝̅̅̅̅ =0.03950.882 = 4.48%.
was 3%. State- and mixed-owned companies in larger networks set higher prices than privately-owned companies. For example, in networks larger than 50 kilometers, district heating prices in state- and mixed-owned networks were 14%
and 10% higher than the average price in municipally-owned networks for MDB and SFH, respectively.
Table 6: Regression results of Model I with price components as dependent variables (Random effects) Dependent variable: Price of district heating (SEK/kWh)Overall Fixed component Variable component Overall Fixed component Variable component (MDB) (MDB) (MDB) (SFH) (SFH) (SFH) Ownership (Reference: Municipal) Private 0.0273**0.0502***-0.02260.0248**0.0414***-0.0156 (0.0139)(0.0168)(0.0196)(0.0112)(0.0130)(0.0181) State and mixed -0.0009470.0328-0.0360-0.01960.0118-0.0250 (0.0199)(0.0307)(0.0364)(0.0252)(0.0150)(0.0252) Price Dialogue (Reference: Not a PD member) PD member 0.0229**0.003830.01520.006050.004829.56e-05 (0.00990) (0.0138)(0.0166)(0.00557) (0.00800) (0.00879) Interactions PD × Ownership (Reference: Municipal) [d_PD=1] × [𝑑_𝑂=private]-0.0123-0.02320.01180.0229***-0.009010.0305*** (0.0115)(0.0169)(0.0179)(0.00639) (0.00673) (0.00859) [d_PD=1] × [𝑑_𝑂=state and mixed] -0.0238*-0.0566***0.03480.0333***7.60e-050.0339*** (0.0125)(0.0188)(0.0212)(0.0106)(0.00954) (0.0125) Electricity price (SEK/kWh)0.127***0.005320.137***0.145***-0.01840.184*** (0.0300)(0.0433)(0.0449)(0.0399)(0.0421)(0.0473) Installation costs heat pumps (1,000 SEK) 0.000552*0.000404-3.44e-05-0.001101.88e-05-0.000859 (0.000285)(0.000316)(0.000415)(0.00196) (0.00225) (0.00277) Disposable household income (1,000 SEK) 0.000205*1.26e-050.000190-8.11e-05-6.90e-05-1.08e-05 (0.000110)(0.000114)(0.000117)(9.21e-05)(7.34e-05)(0.000102) Heating degree days -2.94e-071.18e-05-1.89e-05**-2.31e-05***9.24e-06*-3.41e-05*** (7.09e-06)(8.52e-06)(9.50e-06)(5.38e-06)(5.55e-06)(7.65e-06) Total O&M costs (SEK/kWh)0.02400.0502**-0.03060.008190.0166-0.00665 (0.0237)(0.0212)(0.0323)(0.0171)(0.0204)(0.0317) Interactions Total O&M costs × Ownership
