Advanced search
Publication Cover

Climate Policy

Volume 16, 2016 - Issue 1
Submit an article Journal homepage
461
Views
16
CrossRef citations to date
0
Altmetric
SYNTHESIS

The (ir)relevance of transaction costs in climate policy instrument choice: an analysis of the EU and the US

Fabian Joas Potsdam Institute for Climate Impact Research (PIK), Telegrafenberg 31, 14473 Potsdam, Germany; Mercator Research Institute on Global Commons and Climate Change (MCC), Torgauer Straße 12–15, Berlin 10829, GermanyCorrespondencejoas@pik-potsdam.de
&
Christian Flachsland Mercator Research Institute on Global Commons and Climate Change (MCC), Torgauer Straße 12–15, Berlin 10829, Germany
Pages 26-49
Published online: 01 Dec 2014

SYNTHESIS

The (ir)relevance of transaction costs in climate policy instrument choice: an analysis of the EU and the US

Abstract

This article assesses the relevance of ex post transaction costs in the choice of climate policy instruments in the EU (focusing mainly on the example of Germany) and the US. It reviews all publicly available empirical ex post transaction cost studies of climate policy instruments broken down by the main private and public sector cost factors and offers hypotheses on how these factors may scale depending on instrument design and other contextual factors. The key finding from the evaluated schemes is that it is possible to reject the hypothesis that asymmetries in ex post transaction costs across instruments are large and, thus, play a pivotal role in climate policy instrument choice. Both total and relative ex post transaction costs can be considered low. This conjecture differs from the experience in other areas of environmental policy instruments where high total transaction costs are considered to be important factors in the overall assessment of optimal environmental policy choice. Against this background, the main claim of this article is that in climate policy instrument choice, ex post transaction cost considerations play a minor role in large countries that feature similar institutional characteristics as the EU and the US. Rather, the focus should be on the efficiency properties of instruments for incentivizing abatement, as well as equity and political economy considerations (and other societally relevant objectives). In order to inform transaction cost considerations in climate policy instrument choice in countries that adopt new climate policies, more data would be desirable in order to enable more robust estimates of design- and context-specific transaction-cost scaling factors.

Policy relevance

The findings of this study can help inform policy makers who plan to set up novel climate policy instruments. The results indicate that ex post transaction costs play a minor role for large countries that feature similar institutional characteristics as the EU and the US. For instrument design the focus should rather be on efficiency properties of instruments in incentivizing abatement, as well as equity and political economy considerations (and other societally relevant objectives).

1. Introduction

A significant body of literature has evolved in recent years that analyses the optimal design of climate policy instruments for correcting market failures, such as the external costs of emitting harmful GHGs or research and development (R&D) underinvestment due to technology spill-overs (e.g. Fischer & Newell, 2008; Goulder & Parry, 2008; IPCC, 2007). As demonstrated by Stavins (1995) and Coggan, Whitten, and Bennett (2010), the design, implementation, and enforcement of such policy instruments involve transaction costs (TCs) that might affect optimal policy choice. In other words, the ranking of policy instruments regarding the economic cost-effectiveness criterion might change when TCs are included in the analysis (Ofei-Mensah & Bennett, 2013). Transaction costs are defined as ‘the ex ante costs of establishing environmental policy in all of its aspects, and the ex post costs of administering, monitoring and enforcing the policy once established' (Krutilla & Krause, 2010). This article provides an assessment of the relevance of ex post TCs in the choice of climate policy instruments, such as permit trading, taxes, and standards, by reviewing and comparing the available data on their transaction-cost performances. This is done by reviewing all publicly available empirical ex post TC studies of climate policy instruments broken down by the main private- and public-sector cost factors. In order to inform TC considerations in climate policy instrument choice in countries that adopt or change their instruments, hypotheses are offered on how they can be expected to scale depending on various instrument designs or regulatory context factors, such as the number of regulated entities or the level of abatement. Different design options (e.g. the point of regulation and the design of trading schemes) and their influence on TCs are discussed. Where quantitative data are not available, qualitative considerations and hypothetical plausibility deductions supplement the analysis. Ex ante TCs are not considered because very little empirical data are available, even though these costs may be important factors influencing the implementation of climate change policies.

The key finding from the evaluated schemes is that it is possible to reject the hypothesis that asymmetries in ex post TCs across instruments are large and thus play a pivotal role in climate policy instrument choice. Both total and relative ex post TCs incurred by the public and private sectors can be considered low. In Germany, for instance, they range from €28 to €81 million1 annually for different instruments on which relatively robust empirical evidence is available. This represents between 6 and 18 Euro cent (€ct) per regulated ton of CO2 in 2011, or less than 2% of the certificate price in the European Union Emission Trading System (EU ETS) in 2011 (€10), or less than 1% of German net renewable electricity subsidies in 2011 (€12 billion) (Frontier Economics, 2012).

This article is structured as follows. Section 2 reviews the literature on TCs in climate policy instrument choice. Section 3 introduces the TC definition and the methodical approach of this study. Section 4 offers the results and a discussion on TCs for (1) trading schemes, (2) taxing schemes, (3) a comparison of different points of regulation for tax and trading schemes, (4) technology standards, and (5) performance standards. Section 5 concludes.

2. Literature review

Stavins (1995) was the first to demonstrate analytically that in the context of environmental policy instruments, the TCs of trading permits can significantly affect the efficiency of an ETS. He finds that if TCs drive a significant cost wedge between a firm's option to abate internally or to trade permits on the market, this can significantly reduce the efficiency gains of harmonizing marginal abatement costs (MACs) across regulated entities. This hypothesis has been confirmed empirically for several environmental ETSs in the US. Kerr and Maré (1998) found that the TCs of permit trading reduced cost-effectiveness by around 10–20% in the US lead phase-down scheme. Gangadharan (2000) showed that the presence of trading TCs reduced the probability of trading by about 32% in California's Regional Clean Air Incentives Market (RECLAIM) scheme. Hahn and Hester (1989) demonstrated that high TCs decreased trading activity in the Fox River scheme in Wisconsin. In this context, it must be noted that TCs have two distinctly different effects. First, TCs cause direct costs (such as the costs of monitoring, reporting and verification (MRV) of emissions, brokerage, and trading) that can be readily measured and are the subject of this study. Second, TCs can also distort optimal abatement and lead to welfare losses (Stavins, 1995). The latter effect is discussed in the following, but its quantification is not the subject of this study.

Empirical studies of TCs in the environmental policy literature remain patchy, because most authors have limited their TC definitions and analyses to the narrow neoclassical definition – the cost of using the market mechanism – that has usually been operationalized as the brokerage costs of trading permits. This narrow definition has two problems. First, it covers only a small portion of total TCs. Second, it does not allow for a comparison of TCs with other policy instruments where no trading takes place, such as taxes or standards. Against this background, McCann, Colby, Easter, Kasterine, and Kuperan (2005) and Krutilla and Krause (2010) developed a more inclusive taxonomy (see Section 3) that provides a useful framework to analyse the environmental policy instruments comparatively. This broader perspective has been adopted by several empirical studies in recent years, which revealed that in trading schemes, brokerage costs are small compared to other TC components, such as the costs of MRV (Brockmann, Heindl, Löschel, Lutz, & Schumacher, 2012; Jaraitė, Convery, & Di Maria, 2010; Löschel, Brockmann, Heindl, Lutz, & Schumacher, 2011; Löschel, Kiehl, Lo, Koschel, & Koesler, 2010; VBW, 2011).

Three other recent studies have investigated the relative TC performances of different climate policy instruments. Betz, Sanderson, and Ancev (2010) compared the total costs (abatement costs plus TCs) of achieving a certain reduction target by means of uniform firm coverage under an ETS to a scheme where small emitters opted out of the ETS and, instead, were covered under a performance standard. They found that with modest emissions reduction targets, shifting from uniform to partial coverage led to overall cost savings. Yet, if the level of ambition for CO2 abatement increased, overall savings decreased due to the relatively lower abatement efficiency of the performance standard.

Ofei-Mensah and Bennett (2013) compared three climate policy instruments in the Australian transport energy sector: two standards providing enhanced consumer information (the Fuel Label Program and the voluntary Fuel Efficiency Program) and a hypothetical Tradable Permit and Fee System. They identified strong asymmetries in TCs per tCO2e abated between the standards (about $2.5/tCO2e abated) and the Tradable Permit and Fee System (about $7.2/tCO2e). However, this conclusion critically hinged on their chosen cost metric of cost-per-ton abated and the assumption that the trading system would yield only 6.7MtCO2e total cumulated emissions reductions over a 15-year period. If more cumulative abatement occurred in this time span – e.g. 67 MtCO2e – the TCs of trading according to this metric would drop (in the example, to $0.72/tCO2e abated). This shows that the chosen metric is critical for a comparison with other policy instruments, in this case because TCs depend on the assumed abatement level.

A recent synthesis article by Mundaca et al. (2013a) describes TCs as ‘an important factor in public policy' in the context of climate policy instruments. The article reviews TCs in the context of energy efficiency technologies, renewable energy technologies, offset carbon markets, and the EU ETS. However, the TCs are not presented in a consistent metric across the instruments, which makes it difficult to draw broader conclusions. Clearly, TCs have been high or even exorbitant for some projects of the Kyoto mechanisms (Antinori & Sathaye, 2007; Michaelowa & Jotzo, 2005).

This article complements the existing literature by reviewing and comparing all publicly available empirical studies on climate instrument TCs. The aim is to discern whether TCs critically affect the efficiency rankings of different climate policy instruments.

3. Definitions and approach

This study adopts the following TC definition: ‘Transaction costs are the ex ante costs of establishing environmental policy in all of its aspects, and the ex post costs of administering, monitoring and enforcing the policy once established' (Krutilla & Krause, 2010; based on McCann et al., 2005). In other words, ‘[transaction costs are all costs of the policy] excluding abatement costs.' This definition covers all phases and aspects of costs that are needed to carry out a meaningful comparison across different policy instruments (see Table 1 for a general list of public sector TCs and Tables 2 and 3 in the following sections for a list of private- and public-sector TCs in the case of emissions trading, which further differentiates the general ex post TC components, as indicated in Table 1).

Table 1 TCs associated with public policies

The focus of this article is on ex post TCs, because only very limited data exist on the ex ante stages for climate policy instruments (categories (i)–(iii) in Table 1). In the short term, if these ex ante costs are significant, they may influence a firm's decision and could lead to an investment hiatus, implying adverse impacts on cost-effectiveness. However, in the long run, ex post costs will usually be the dominant factor (Betz, 2010).

Other studies that have investigated the TCs of policy instruments apply concepts such as information costs, search costs, administrative costs, compliance costs, legal expenses, monitoring costs, and enforcement costs (Destatis, 2011; Kossoy & Guigon, 2012; LECG, 2003; Margaree Consultants Inc., 1998; McMahon, Chan, & Chaitkin, 2000; DR Roever & Partner KG 2006; Sandford, Godwin, & Hardwick, 1989; Smulders & Vollebergh, 2001; VBW, 2011). The definition adopted here allows for the integration of all these cost components.

As pointed out by Mundaca et al. (2013b) and Macher and Richman (2008), the existing empirical literature on TCs for policy instruments lacks a well-developed and comprehensive theoretical foundation. Although this article does not attempt to systematically fill this gap, it aims to contribute to the development of TC theory by introducing conceptual distinctions of how ex post TC dimensions for climate policy instruments can scale with respect to differing regulatory contexts and policy designs. The need for such a specification of the scaling of TCs arises when attempting to compare estimates of TCs across empirical cases. This is done in order to draw conclusions regarding the anticipated TCs when introduced in other regulatory contexts, that is, to inform the choices and set-ups of novel climate policy instruments.

Building on the empirical studies reviewed in this article, as well as conceptual considerations, we suggest distinguishing the following dimensions along which TC components of climate policy instruments can be scaled:

  1. The number of regulated entities

    • Firms' MRV costs make up a large share of total TCs. The more entities being regulated, the higher the total private-sector TCs will tend to be.

  2. The size of regulated entities

    • Small installations tend to have smaller total TCs than large installations. However, the TCs per tCO2 regulated tend to be smaller for large installations due to economies of scale.

  3. The number of regulated appliances (in the case of an appliance standard)

    • For homogenous mass-produced goods, such as household appliances or vehicles, certification must be carried out for each model of appliance (e.g. fridge or dryer) that is offered in a certain market. Therefore, TCs scale based on the number of different models of appliances being regulated, but not on the total number of appliances sold.

  4. The level of abatement

    • Identifying abatement options at the firm level and establishing an abatement cost curve is costly. Therefore, increasing levels of abatement should raise TCs.

  5. The volume of market transactions (in the case of trading mechanisms)

    • The trade of TCs includes brokerage costs, which must be paid for each traded unit (e.g. CO2 certificate). Accordingly, the TCs vary with the number of traded certificates.

  6. Other instrument or TC-specific factors

    • The magnitude of TCs from legal disputes, for example, partly depends on how firms estimate the costs of lawsuits versus the chances of winning the case. The more favorable this ratio, the more likely legal action will be. Other factors, such as the costs of lawsuits, corporate law, and legal norms, can also play a role in this specific cost factor.

  7. Time

    • Time, as a TC-scaling factor, cuts across the basic TC categories and the previous scaling dimensions. Cost factors, and the degree to which they are scalable, are likely to change over time. For example, due to learning-by-doing and technological change, absolute TCs may generally fall over time (e.g. trading costs in the US SO2 allowance market fell 98% over time) (Joskow & Schmalensee, 1998; LECG, 2003).

These dimensions can interact. Consider a limiting case. When, over time, the level of abatement becomes so ambitious that emissions approach zero, fewer entities will require regulation, and absolute trading volumes will fall. TCs will probably rise with respect to abatement cost information, but decline with respect to the other factors.

In addition, there are fixed costs during the inception of the policy instrument that occur only once (such as the administrative, informational, and capital costs involved in setting up novel compliance structures), as opposed to the annual running costs of a scheme over the long term (Jaraitė et al., 2010). Finally, for clarity, it is useful to distinguish TC components and scaling factors at the firm- and aggregate-system level (indicating total TCs). This article focuses on the latter.

The methodical approach of this article is to assemble all publicly available ex post TC data on existing climate policy instruments and to use evidence from other policy instruments that were plausible analogies. The latter approach was taken only where TC data on climate policy instruments were scarce or unavailable (e.g. CO2 tax). Data were obtained from case studies, consultant reports, government reports, interviews, and our own calculations. Where possible, the data were broken down into the ex post cost components defined above. Overall, data were limited but available for the EU ETS, the US SO2 ETS, US RECLAIM, the UK excise duties on hydrocarbon fuels, and the US Residential Appliance Standard. When different studies reported different values for a scheme, we focused on the higher estimates to ensure that our aggregation was on the conservative (i.e. high-cost) side. However, by indicating ranges, we also display the low-cost estimates. For policy instruments where neither data nor useful analogies existed (e.g. technology standards for large point sources) the costs were assessed qualitatively.

The use of the metric for reporting TCs depends on the quality of data and the purpose that it is intended to serve. If the aim is to inform the comparison and choice of climate policy instruments in countries that consider adopting novel policies, there is no single metric that can convey all of the desired information in a satisfying manner. This is because, as argued above, cost components are scaled based on context- and policy-specific dimensions. Quantitative scaling factors cannot (yet) be reliably estimated based on the very few existing data points. This can be illustrated with regard to two prominent TC metrics used in the literature.

First, reporting TCs per regulated GHG units (such as €ct/tCO2e) is an elegant way of expressing TCs in a given ETS, relative to the visible price of an emissions allowance. It is most often used for comparing MRV costs across firms in order to identify MRV TC-scaling effects with respect to firm size (e.g. Jaraitė et al., 2010; Löschel et al., 2011). We make use of this metric when discussing these dimensions, but using this metric more generally to compare TCs across policy instruments and regions could mask potentially significant context-specific factors, such as the mix of large and small regulated entities in a scheme.

Second, reporting TCs per unit of GHG abated only makes sense when considering cases with comparable abatement levels. Consider the following example. In the first trading period of the EU ETS (2005–2007), little abatement occurred (Ellerman, Convery, & Perthuis, 2010, p. 191), while the EU ETS incurred full TCs. Thus, TCs per abated emissions were relatively high. If the cap had been much tighter, TCs per abated tCO2 would have been much lower, without changing the absolute TCs, because these are likely to scale only to a limited extent with regard to abatement levels (see Section 4.1). This cost metric would be very high in one case and very low in another, while absolute TCs would have remained roughly constant. Without further qualification, such information is not very helpful and is potentially misleading for comparing costs across instruments and regions.

To sum up, if multiple data points on TC components in different contexts were available – and they might become available in the future – it could inform numeric estimates of the different scaling factors. This is currently not the case. This study copes with this challenge by reporting total TCs per country, using Germany and the US as examples. For specific (climate) policy instruments, the magnitudes of different cost components are reported in detail. Potential scaling factors, if these instruments were transferred to other contexts, are discussed. Furthermore, some sensitivity analyses of TCs with respect to programme design were conducted (i.e. upstream versus downstream points of regulation for the EU ETS). The main limitations of this approach are the same as for the other metrics, i.e. the results and policy instruments are not directly comparable to other countries due to differences in the coverage and levels of abatement that are induced. Bearing this caveat in mind, the existing data indicated that the total TCs of different instruments in Germany and the US are low compared to other macroeconomic figures. Finally, the relative costs of the different instruments fall within a range of similar orders of magnitude.

4. Instrument comparisons

Policy instruments for reducing GHG emissions can be divided into two groups: market- and non-market-based instruments. Market-based instruments address the market failure of the GHG externality by incorporating external costs. While the regulator sets the price (or quantity), the market determines the quantity (or price). Examples are permit-trading schemes (quantity instruments) and GHG taxes (price instruments) (IPCC, 2014). In the case of non-market-based instruments (i.e. standards), the regulator sets either a limit for maximum emissions (in absolute or relative terms) or prescribes a certain technology. Examples are vehicle performance standards (CO2 emissions per driven km) and technology standards (e.g. banning coal-fired power plants). This section analyses and discusses the available empirical TC data components for market- and non-market-based instruments for GHG mitigation. Section 4.3 compares different points of regulation (upstream or downstream) for GHG trading and taxation schemes. The discussion of each TC component for each instrument is complemented by a plausibility analysis of its potential scaling dynamics.

4.1. Emissions trading

Emissions trading is a quantity instrument, because the maximum amount of permissible emissions for the regulated entities is set by the regulator (Tietenberg, 2006). Emissions permits are distributed either by free allocation or auctioning and can be traded among polluting entities. Firms engage in abatement efforts and permit trading until the equilibrium permit price emerges. Under ideal conditions, the instrument is cost-effective because MACs are indicated by the market permit price and equalized across all participating firms.

The EU ETS is by far the largest application of emissions trading worldwide, and a number of TC studies of the EU ETS have been conducted, particularly in Germany and Ireland. They are usually based on questionnaires handed out to companies or expert estimations.

These studies exclusively focus on private-sector TCs. However, the analysis in this article adds public-sector costs. Due to the relatively good availability of data for both private and public TCs in Germany, we have used this country as a case study for EU ETS data and compared it with Irish data to show that private-sector figures seem to be robust. Public agency costs in Germany were considered to be on the higher end (on a TC-per-regulated-entities basis), compared to the rest of the EU (Transaction Costs of DEHSt, W. Seidel, personal communication, April 20, 2011, Berlin, DEHSt). We therefore used the German public-sector TCs as the basis for the calculation of the EU-wide public-sector TCs.

4.1.1. Private-sector costs

The EU ETS TC components for the private sector are displayed in Table 2.

Table 2 Classification of private-sector ex post TCs in the EU ETS

The first component is ‘assembling information on cost-effective abatement at the facility level’. Before firms can engage in abatement, they need to assemble information on their abatement cost schedule. This involves financial and technical analysis, as well as an analysis on how production and product quality will be affected (Hein & Blok, 1995). This requires additional personnel expenditures and advice from experts, including calculations and risk assessments for payback periods. It can be expected that these costs will rise with increasing abatement ambition because it becomes more difficult to find new abatement options once the easy abatement options have been exploited. However, with only €0.9 million in annual costs regulated to the EU ETS in Germany (Löschel et al., 2011), these costs are currently very small and are seen as negligible compared to other cost components. If these costs were significant, they might affect overall cost-effectiveness along the lines identified by Stavins (1995) by driving a wedge between the theoretically cost-effective allocation of abatement and the actually realized one in the presence of TCs.

The MRV of emissions is crucial to all market-based schemes because regulators require reliable emissions data, and firms have an incentive to under-report and over-emit. In the EU ETS, firms must report their emissions annually, then have these reports verified by an external verifier who must be accredited by a government agency (DEHSt, 2012). A firm's MRV costs can differ strongly with the size of regulated installations. This non-linearity in firms' MRV costs (see Figure 1) is due to economies of scale in the MRV process, which feature relatively high facility-level fixed costs that have been well documented by several studies (Betz, 2005; Frasch, 2007; Heindl, 2012; Jaraitė et al., 2010; Löschel et al., 2010, 2011; Schleich & Betz, 2004). It has been argued that this may lead to a competitive disadvantage for small emitters (Jaraitė et al., 2010). Heindl (2012) suggests that this effect may theoretically lead to a larger optimal firm size and might even induce increasing market power in the permit market. However, given the low orders of magnitude involved – a firm emitting 10,000 tCO2 per year at an average TC of €0.76 would face annual costs of €7600 – it seems that the empirical effect will be practically irrelevant. Yet, TCs of €1.51 per tCO2, as reported by Jaraitė et al. (2010) for Ireland, may lead to disproportionate hardships for certain firms.

Figure 1 EU ETS annual MRV running costs for small and large installations in €/tCO2 regulated.

Sources: Jaraitė et al. (2010) and Löschel et al. (2011).Notes: *median; **average costs.
Display full size

From 2009 to 2011, the average annual private-sector MRV costs in Germany were reported to be between €ct2/tCO2 (Löschel et al., 2011) and €ct9/tCO2 emitted (Destatis, 2011). In terms of absolute average-per-entity costs, this translates into roughly €2500 per annum (p.a.) for small installations (annual emissions below 25,000 tCO2) and roughly €8600 for large installations (annual emissions above 25,000 tCO2).2 For Germany, total MRV TC estimates range from €11 to €40 million p.a. As illustrated in Figure 2, MRV constitutes the largest share of private-sector TCs (61–81% of total private costs; Destatis, 2011; Löschel et al., 2011; own calculation).

Figure 2 Year 2011 EU ETS and hypothetical carbon tax TCs for different cost components for Germany based on data from the first and second trading periods in total costs.

Sources: 1, Destatis (2011); 2, Brockmann et al. (2012), Destatis (2011); 3, VBW (2011); 4, Löschel et al. (2011); 5, DEHSt (2011a); 6, DR Roever & Partner KG (2006), DEHSt (2011b), DEHSt (2008), CEC (2012) ; 7, EEX (2012), Point Carbon (2011); 8, Kossoy and Guigon et al. (2012), M. Weber, personal communication, September 10, 2012, Berlin, GreenStream Network Plc, EEX (2012).If the numbers differed across the studies for the same cost components, the highest estimate was used in order to ensure robustness of the results.Notes: Bar chart (left) provides cost components using conservative (higher) cost data. Table (right) provides cost ranges. CDM TCs are excluded to focus on the direct costs of the EU ETS.
Display full size

Aggregate MRV costs are scaled to the number of regulated facilities and the mix of large- and small-sized entities (small/large emitters have lower/higher absolute, but higher/lower per unit GHG costs) (Frasch, 2007; Jaraitė et al., 2010; Löschel et al., 2011). These costs are independent of the abatement level, because MRV processes have to be carried out for all emissions, independent of the level of abatement. It can be expected that MRV costs will fall over time due to the effects of learning and improved technology.

Concerning the allocation of permits, firms covered by the EU ETS received the largest amount of certificates for free in the first two trading periods (2005–2012). In the third trading period (2013–2020), free certificates will be handed out based on firms' export shares and relative efficiency (benchmarking). Brockmann et al. (2012) found that for the third trading period, the median TC for the free allocation of certificates amounted to a one-time expenditure of €25,000 for personnel and external costs per firm at the beginning of the trading period. Extrapolating to all covered firms and distributed over eight years, this amounts to roughly €2.5 million annually for all firms in Germany. The dynamics of this cost component depends on the number of firms that qualify for free allocation and, therefore, on the eligibility criteria for obtaining a free certificate. The higher the number of eligible entities, the larger the aggregate TCs from this component will be.

Auctioning is the alternative to the free allocation of certificates. The costs charged by exchanges to carry out auctions on behalf of governments are about €ct0.3 per certificate (Point Carbon, 2011). A 10% share of auctioned permits in Germany in 2011 amounted to roughly €0.13 million in TC, while a 50% share (as implemented from 2013 onwards) would amount to €0.7 million in TCs per year. These costs are clearly scalable to the number of auctioned certificates. As in the case of trading, these costs may fall over time.

Because the rules for granting free certificates are not clearly defined, firms can exercise their rights to legally dispute allocations in court.3 A survey by VBW (2011) found that these figures translated into €5400 in litigation per installation in Germany, or €10.7 million overall p.a.4 The scalability of this component is not obvious. However, it can be expected that legal action is scalable to the success rate in relation to the costs of litigation. However, other factors such as legal traditions, corporate law, and political acceptance of regulations by companies may also have an effect.

Permit trading is central to an ETS because it leads to equalization of MACs across all firms. In his seminal study on TCs, Stavins (1995) shows that TCs for permit trading, which he defines as the ‘direct financial costs of brokerage services’, can lead to the inefficient allocation of abatement across firms. From a cost-effectiveness perspective, trading costs drives a wedge between firms' MACs. TCs reduce beneficial permit exchanges and corresponding adjustments in the allocation of abatement, thus reducing the cost-effectiveness of the policy instrument (Stavins, 1995). In the EU ETS in 2011, the trading volume amounted to 9.7 billion certificates (Kossoy & Guigonet, 2012). Trading TCs ranged from €ct0.1/tCO2 to €ct10/tCO2, with costs apparently falling towards the lower end of this range over time (Convery & Redmond, 2007; EEX, 2012; J. Hacker, personal communication, January 24, 2012, Berlin, UMB; Kossoy & Guigonet, 2012; M. Weber, personal communication, September 10, 2012, Berlin, GreenStream Network Plc). In the EU ETS, the overall trading costs amounted to roughly €46 million in 2011 and about €11 million p.a. for Germany. These trading costs are small and there is no empirical indication that they discouraged trading (Jaraitė et al., 2010). Therefore, the impacts of negative cost-effectiveness – Stavins' central concern – are likely to be minor for the case of the EU ETS. Trading TCs is scalable to the volume of trading because brokerage costs arise for each transaction. In the EU ETS, it can be observed that the majority of trades are carried out by market intermediaries who are seeking arbitrage opportunities, and not by regulated firms (M. Weber, personal communication, September 10, 2012, Berlin, GreenStream Network Plc). It is therefore likely that price fluctuations (that give rise to profit opportunities for market intermediaries) determine trading levels rather than the amount of regulated GHGs or the amount of abated emissions, even though in a larger trading system, overall higher aggregate trading levels would be expected compared to smaller schemes. It can also be expected that trading costs will fall over time, as illustrated by the comparison with other environmental permit-trading schemes in the following.

4.1.2. Public-sector costs

Table 3 presents the EU ETS TC components for the public sector.

Table 3 Classification of public-sector ex post TCs in the EU ETS

The compliance agency is the central authority that administers the ETS. It oversees the MRV process by conducting sample checks, allocates free certificates, appoints exchanges to auction certificates, provides information, and engages in litigation. The total annual costs for these tasks amounted to roughly €15 million or roughly €7700 per regulated entity for Germany from 2005 to 2011 (CEC, 2012; DEHSt, 2008; DR Roever & Partner KG, 2006; Seidel, transaction costs of DEHSt, personal communication, April 20, 2011, Berlin, DEHSt).

The compliance agency for TCs is scalable based on several factors, of which the number of regulated entities is probably the most decisive. Furthermore, there are initial start-up costs, such as setting up administrative processes and communicating and coordinating these processes with the regulated entities. It is likely that the cost of the compliance agency is also scalable to the quality of its performance because rigorous and more frequent verification would be more costly than relaxed control. The per-entity costs may decline with the increasing number of regulated entities due to economies of scale.

The public registry is the electronic database for reconciling certificate accounts and the emissions data of all regulated entities and is usually run by the compliance agency. The costs related to the registry are mainly software and servicing costs. In Germany, they averaged about €0.76 million per year for the first and second trading period (DEHSt, 2011a).

This is largely a fixed component because the costs are determined by the licensing and maintenance costs of the software. However, this cost also contains a variable component because it is likely that the costs of the software are scalable to the number of accounts. Due to improved technology, such costs may fall over time. However, they could also rise due to higher regulation requirements and safeguards against fraud (European Voice, 2013).

4.1.3. Offsets for Kyoto mechanisms

In the EU ETS, firms can also use credits from the Kyoto offset mechanisms (the Clean Development Mechanism [CDM] and Joint Implementation [JI]) for compliance. These credits are usually cheaper (due to lower abatement costs) than the EU ETS allowances, but tend to feature higher TCs (due to costly project cycles). In the second trading period, 302 million Kyoto offset credits were used for compliance in Germany (CEC, 2014; DEHSt, CDM und JI in zweiter Handelsperiode, W. Seidel, personal communication via e-mail, 2014; Emissions-euets.com, 2014). Based on data from Krey (2005), Antinori and Sathaye (2007), and Michaelowa and Jotzo (2005), we found a range of TCs, from €15 to €32 million p.a., which resulted from the use of offset credits for Germany. However, it must be noted that these TCs do not affect the companies in the EU ETS directly because the TCs are already incorporated into the offset price, which is usually lower than the costs of the European Union allowances (EUAs). Nevertheless, these costs hamper the efficiency of the EU ETS link to the CDM via the distortive effect identified by Stavins (1995).

Due to the various cost components of the CDM process cycle, a separate analysis of the dynamics of these cost components should be subject to further research; however, this is outside the scope of this study. For the EU ETS, the total TCs incurred by the offset credits is clearly scalable to the number of credits used for compliance.

4.1.4. Summary, discussion, and comparison with other environmental permit trading schemes

Figure 2 aggregates the ex post TC estimates for the EU ETS in Germany. The total annual costs in the one-year trading period from 2010 to 2011 amounted to €81 million in Germany or €ct18/tCO2 regulated. Private-sector costs are 68–80% of total TCs, while public-sector costs are 20–32%.

Several studies have analysed the effects of trading costs on trading schemes targeting other environmental pollutants (Joskow & Schmalensee, 1998; LECG, 2003; Stavins, 1995). These schemes feature much higher trading costs, especially during inception. The RECLAIM scheme featured costs from €3 to €72 per traded unit (tNOx and tSOx), the US NOx budget trading scheme featured trading TCs from €5 to €27 per unit (LECG, 2003), and the US SO2 scheme is currently in the range of €0.15, but started at around €23 (Joskow & Schmalensee, 1998; LECG, 2003) (Figure 3). Due to the effects of learning by market participants and brokers, as well as improved technology, the costs have dropped by more than 98% over time.

Figure 3 Ranges of ETS per unit (permit) of trading costs compared to the private sector.

Sources: Joskow and Schmalensee (1998), Margaree Consultants (1998), LECG (2003), EEX (2012), Convery and Redmond (2007), J. Hacker, personal communication, January 24, 2012, Berlin, UMB, M. Weber, personal communication, September 10, 2012, Berlin, GreenStream Network Plc.
Display full size

The claim of inefficient abatement and resulting welfare losses due to the TCs of permit trading (Stavins, 1995) seems to be more relevant for these schemes (e.g. Gangadharan, 2000), especially at the time of their inception, than for the EU ETS where it seems sensible to assume that the relevance of this effect is negligible.

The comparison of public-sector TCs reveals that US environmental permit trading schemes are leaner than the EU ETS. While the EU ETS and the NOx budget ETS in the US have running costs of €3,900 and €7,700 per regulated entity, respectively, the US SO2 scheme is far less costly with only €780 per regulated entity (DEHSt, 2008, 2011b; LECG, 2003; McLean, 1997; DR Roever & Partner KG, 2006). For the US SO2 scheme, the total firms’ MRV costs amounted to approximately €100 million and the trading costs were approximately €3.5 million. Including the €1.6 million from public-sector costs, this amounts to TCs of €105 million (Joskow & Schmalensee, 1998; LECG, 2003; US Environmental Protection Agency (EPA), 2001).5

All of these metrics suggest that the TCs of the EU ETS are not prohibitively high in absolute terms from a macro-economic perspective. In any case, when choosing a policy instrument, it is the relative costs of the instruments that matter. This point is addressed in the concluding remarks.

4.2. Tax

Emissions taxation is a price instrument, because the regulator sets the emissions tax, while the quantity of emissions is determined by firms' market behaviours. Much like a permit-trading scheme, a carbon tax can achieve the equalization of MACs across all regulated entities. Under standard assumptions of competitive markets, trading and taxation schemes are symmetric. Asymmetries in instrument performance have been discussed for conditions of uncertainty (Hepburn, 2006; Weitzman, 1974) or incentive structures for subnational jurisdictions in a federal system (Shobe & Burtraw, 2012). This study investigates how taxing and trading compare with TCs performance.

A comparison of TCs between the tax and trading schemes needs to be careful not to compare ‘apples with oranges’, as is sometimes done in public debates where real-world ‘second best' schemes (such as the EU ETS) are compared with theoretical ‘first best' taxation schemes. In particular, specifications regarding the point of regulation and rent distribution need to be treated in a conceptually equivalent manner. A downstream trading scheme with a large portion of grandfathered certificates (such as the EU ETS) should be compared with a downstream taxation scheme with an equivalent level of tax exemptions. Even though the incentives for abatement may be different, the two options (grandfathering and exemptions) should be regarded as analogous responses to identical political economy considerations.6

Firms' MRV costs as the major TC cost component would be the same in both tax and trading schemes for the firms that are subject to MRV, because data requirements for the MRV process are identical. Also, other private TC factors would be identical except for trading and auctioning costs, which do not accrue for taxes.

Drawing on data from the previous section, a downstream taxation scheme for Germany with tax exemptions equivalent to the free allocations in the EU ETS would thus result in TCs of approximately €41 to €71 million (see Figure 2). This is approximately €10 million (the costs of trading and auctioning) less than the TCs of a trading scheme, which seems negligible from a macro-economic perspective.

A different approach to determining the TCs of a CO2 tax is to compare it to other taxes that require similar administrative efforts by the private and public sector. Because a CO2 tax would be designed to act as a levy on fuel according to its CO2 content, excise duties on hydrocarbon fuels are a good proxy. Data from a study on administrative and compliance costs of excises on hydrocarbon oils are available for the UK (Sandford et al., 1989, p. 155). The tax covers light oil, road vehicle fuel, fuel oil, and gas oil. The major findings from this study are that the majority of costs are the MRV costs of installations and the reading of meters (Sandford et al., 1989, p. 156). From 1986 to 1987, public-sector costs amounted to €14.2 million and private-sector costs were €31.7 million for all of the UK.

The TCs for taxation schemes feature the same cost components (except for auctioning and trading) as an ETS. Thus, the same considerations apply concerning the scaling of these cost factors with respect to different context and design variables (see Section 4.1).

4.3. Carbon-pricing TCs: upstream versus downstream points of regulation

The point of regulation is the location in the fossil-fuel processing chain where the regulator carries out MRV and requires the submission of emissions permits. Options can be distinguished according to up-, mid-, and downstream regulations. For upstream schemes, the point of regulation is during the exploitation or importation of fossil resources. In this case, all other downstream consumers would not be subject to MRV and trading, thus eliminating all TCs for downstream firms. The upstream carbon price can be expected to be devolved to the downstream level, as is standard procedure with vehicle fuel excise duties. This approach would eliminate the inefficiency of non-linear MRV costs for small companies, as discussed in Section 4.1 and illustrated in Figure 1. By contrast, midstream regulation is carried out at the processing or storage level, and downstream regulation is exercised at the firm (as in the EU ETS) or even the final consumption level (Flachsland, Brunner, Edenhofer, & Creutzig, 2011).

The TC savings that result from adopting an upstream scheme compared to a downstream scheme depend on two factors: (1) the difference in TCs per emitted tCO2 between small and large entities and (2) the ratio between small and large entities within a scheme.7, 8

The calculations in Table 4 show that switching the point of regulation from downstream to upstream could result in savings ranging from €2 to €9 million p.a. for Germany. Small firms that currently feature high relative TCs would benefit most from these savings. However, the savings are limited in absolute terms because most emissions (over 99%) in Germany come from large sources that already have small TCs. The scope for total TC reductions from a macro-economic perspective is limited. This may be different for other sectors that might be included in the EU ETS, in particular heating and transport. The cost savings roughly calculated here also need to be pitched against the full ex ante (opportunity) TCs of negotiating and implementing such a major regulatory change. However, the calculations suggest that a newly set up ETS should opt for upstream coverage from inception from an ex post TC perspective.

Table 4 Potential savings from switching to EU ETS upstream regulation in Germany

4.4. Technology standards

Technology standards are the most traditional instruments for environmental policy making. One major reason is that they are quite easy to design and monitor (Sterner, 2003, p. 80); i.e. ex ante and ex post TCs are often considered to be low. The drawback of this instrument is that an equalization of MACs across facilities is not possible. Strongly differing MACs among firms leads to low overall cost-effectiveness of the instrument. This has led to a relative decline in the adoption of technology standards compared to market-based instruments in recent years (Aldy & Stavins, 2011).

From a TCs perspective, some forms of technology standards for point sources have the advantage of requiring very little MRV. Because, in many instances, it is not worth the effort to remove technologies once installed, sporadic monitoring is sufficient to enforce compliance, and public-sector efforts and data requirements are low (Hepburn, 2006). Other forms of technology standards, such as the requirement to inspect domestic heating systems for CO2 emissions in German MRV must be carried out regularly. Such forms of technology standards might therefore incur substantial MRV TCs; however, data are unavailable. Data for a standard for the regulation of large point sources – e.g. the planned CO2 standards for power plants in the US (US EPA, 2013) or the regulation of SO2 via scrubbers in Germany – are also unavailable. For the purpose of this study, no general quantitative estimates for MRV costs of technology standards for large point sources could be identified or could be plausibly determined by analogy. It seems safe to assume that they will be lower than those of any other policy instrument considered in this study.

4.5. Performance standards

A performance standard limits the amount of emissions for a certain unit of output (e.g. CO2 per ton of cement, CO2 per driven km, kWh per litre refrigerator capacity). In contrast to a technology standard, it has the advantage of allowing firms to choose from various emissions reduction options during the production process, rather than being obliged to fulfil specific technical requirements. It also has the advantage that the standard can be adapted to technological progress (e.g. Japan's Top Runner Program). Non-tradable performance standards have the disadvantage that harmonization of MACs across firms is impossible.

From a TC perspective, a performance standard for heterogeneous point sources of pollution entities (e.g. cement factories or steel mills) does not entail significant cost reductions. This is the case because, analogous to an ETS, all emissions have to be monitored and verified, and trading costs also arise (in the case of white certificate trading). The costs of monitoring are reduced if a performance standard is set for homogeneous non-point source pollution goods that are subject to mass production (e.g. vehicles or household appliances) and no regular inspection is required. Each model of an appliance (e.g. fridge or dryer) that is offered in a certain market must be certified only once. TCs for three different performance standard programmes are briefly discussed in the following paragraphs.

The US Appliance Standard covers a variety of appliances, such as refrigerators, space heaters, water heaters, and other electrical equipment. For compliance, the products must perform better than a maximum limit of energy consumption per unit of the relevant output. Standards setting, test procedures, certification, and enforcement are carried out by the US Department of Energy.

The calculation of annual TCs from different sources is summarized in Table 5. The total annual public sector TCs for the US Appliance Standard range from €4.1 to €17.4 million in 2011. These TCs are substantially lower than for the other climate policy instruments reviewed in this article, indicating that regulation via appliance standards is relatively cheap. This advantage needs to be considered against the abatement opportunities that are related to this instrument (which are likely to be limited) and the inferior overall cost-effectiveness of the technological standards. The reason for the lack of cost-effectiveness is incomplete MAC harmonization, the salience of which will increase with the rising stringency of the programme. This point is discussed further in Section 5.

Table 5 Annual public sector TCs for the US Appliance Standard in €2011

In general, the TCs of an appliance standard are likely to scale with respect to the number of different appliances covered, as MRV (i.e. certification) must be carried out for each model. There are also ongoing fixed TCs for governments and firms. If regular inspection of a device is required (e.g. home heating systems in Germany), the number of appliances, as well as the inspection procedure, will affect the total TCs for that instrument.

A hypothetical performance standard for large point sources in Germany, which would cover the sectors regulated under the EU ETS, would result in TCs of roughly €41 to €71 million p.a. The rationale behind this calculation is that the costs of MRV, registry, and the public agency will be roughly identical to those of an ETS or tax.

TCs for the EU Emission Vehicle Standard and the US Corporate Average Fuel Economy (CAFE) programme are likely to be low, but no data are available. The scaling properties for these TCs are likely to be analogous to an appliance standard. Each model (but not every single vehicle) has to be certified prior to sales, but regular inspection of each vehicle is not necessary. Furthermore, there are also ongoing fixed TCs for governments and firms.

5. Conclusions

Table 6 and Figure 4 summarize the reviewed TC data for different climate policy instruments. They show that ex post TCs of policy instruments in large industrialized countries are relatively low in macroeconomic terms. For example, the TCs for the EU ETS in Germany are less than 1% of the subsidies for renewable electricity in Germany, which was €12 billion in 2011 (Frontier Economics, 2012). Furthermore, the comparison shows that TCs do not differ strongly across instruments.

Figure 4 Aggregated total TCs for different climate policy instruments.

Notes: Ranges are indicated where data from multiple sources were available. Grey bars indicate that the calculations are based on hypothetical schemes. Black bars indicate that the data originated from empirical studies on schemes that were in operation. For data, see Section 4.
Display full size

Table 6 Aggregated TCs for all instruments considered

Compared to the potential orders of magnitude of macroeconomic inefficiencies from non-optimal policy instruments, especially in the longer run, the magnitude of TC differences (as indicated in Table 6) appear to be very small or negligible. For example, using a numerical dynamic general equilibrium model, Kalkuhl, Edenhofer, and Lessmann (2013) calculate that relying exclusively on a feed-in tariff instead of an optimal carbon tax can increase the costs of climate policy by 0.8% of total consumption in terms of balanced-growth equivalents (see also Fischer & Newell, 2008).

Figure 5 offers a framework to discuss TCs in relation to different abatement levels in order to discern whether TCs are relevant to the total cost-effectiveness ranking of climate policy instruments (for a similar line of argument, see Betz et al., 2010). It depicts two scenarios for the regulation of GHGs in one country9 by comparing the abatement levels (low and high), the sums of the total system-level abatement costs, and total TCs of technology standards and market-based instruments.

Figure 5 Optimal policy instrument choice for different issues (issue A, low level of abatement; issue B, high level of abatement) when both the cost-effectiveness, defined as total abatement costs (ACs), and TC properties of instruments are taken into account

Note: Total ACs are assumed to be lower for market-based instruments due to the equalization of MACs.
Display full size

Standards tend to have lower TCs than market-based instruments, but are inferior in terms of MAC harmonization across regulated entities. The magnitude of this adverse cost-effectiveness property will increase with the overall quantity of abatement, thus increasing the relative total abatement costs of the instrument with rising programme stringency. Hence, standards may be superior instruments at low levels of abatement (issue A), but for ambitious abatement programmes (issue B), market-based instruments can be expected to perform better due to lower total abatement costs. It can be argued that climate policy in large countries is generally a B-level policy issue because ambitious emissions reductions will be required in the long term.

Related to these considerations, the incorporation of TC considerations into policy instrument choice might principally also affect the optimal choice of the level of policy stringency in the context of an overall efficiency analysis (i.e. balancing the marginal costs (including TCs) and benefits of abatement using a specific instrument). However, given the small TC differences across climate policy instruments observed in this study, it seems that TCs are practically not relevant in such considerations, at least in regulatory contexts similar to the EU and the US, and instead the relative abatement cost performance as well as other societally relevant evaluation criteria are relevant to the optimal choice of climate policy instruments. However, several limitations to this finding have to be considered.

First, as discussed in Sections 3 and 4, TCs scale with respect to different country- and design-specific factors. For example, TCs may be substantially higher in other jurisdictions with different institutional structures, governance capacities, and market infrastructures. This may be true for other Organisation for Economic Co-operation and Development (OECD) countries, but especially for emerging economies and developing countries. Research on schemes such as the CDM and Reducing Emissions from Deforestation and Forest Degradation (REDD) have shown that TCs can be large and are therefore likely to distort optimal abatement (Antinori & Sathaye, 2007; Boettcher et al., 2009; Fichtner, Graehl, & Rentz, 2003; Michaelowa & Jotzo, 2005). The same may apply to jurisdictions with few abatement opportunities (such as cities). In such cases, the differences in TCs of different instruments may outweigh the advantages of MAC harmonization (indicating that the issue is A-level in the framework of Figure 5). Furthermore, in very small jurisdictions with little abatement, such as cities, differences in TCs may outweigh the advantages of MAC harmonization, which would also indicates A-level situations and related policy choice. To better inform the choice and set-up of novel climate policy instruments with respect to relative TCs, more empirical TC data from applications in different contexts would be required to be able to robustly estimate the magnitude of these scaling effects.

Second, the TCs in this study relate to average ex post costs over a longer time period. Short-term changes in regulation, such as new MRV requirements, may affect investment behaviour and may imply financial hardship for certain firms in the short term, which can create costs at the system level. Furthermore, other ex ante costs such as setting up administrative processes and communicating and coordinating these with regulated entities, but especially the societal costs of political bargaining, may be important factors that can limit the development of climate change policies. A comprehensive analysis of the interdependence of TCs and policy instruments would, therefore, need to put a stronger focus on ex ante TCs. The interplay between ex ante TCs, ex post TCs, and the cost-effectiveness of instruments should be subject to further research.

Third, it has been pointed out that economies of scale, resulting in different MRV costs for firms, may lead to a competitive disadvantage for small emitters, but the effect seems to be small in the case of the EU ETS (see Figure 1 and Section 4.1.1).

Fourth, TCs may alter the shape of a firm's marginal abatement supply curve and subsequently affect cost-effectiveness and efficiency. This may be the case if TCs for assembling information on cost-effective abatement at the facility level vary among different abatement options. The respective TCs for different abatement options at the firm level and their potential to distort programme cost-effectiveness should, therefore, be subject to further research.

Despite these caveats, current evidence indicates that, for large jurisdictions in OECD countries such as Germany and the US, ex post TC considerations play a minor role in climate policy instrument choices. In these contexts, the focus of policy-choice considerations should instead be on the cost-effectiveness properties of the instruments for incentivizing abatement and other societally relevant objectives (e.g. equity or political economy considerations).

This conjecture differs from experiences in other areas of environmental policy, such as pollution trading in the Minnesota River basin and the US- led phase-out trading programme, where very high ex post TCs were observed and thus considered to be important in the overall assessment of optimal environmental policy choice.

Clearly, some caution is warranted given that the underlying data are somewhat meager and patchy outside of the EU ETS. Therefore, the results of this study should be viewed as preliminary until further evidence is provided. Such research should be based on more observations that are founded on clear and uniform reporting rules for TCs. The relevant question here is whether better data would alter the general findings of this study. As this discussion has attempted to demonstrate, it seems unlikely that, for large countries with similar institutional settings to the US and Germany, new data will change the picture dramatically and make ex post TCs a substantial factor for climate policy instrument choice from a macroeconomic perspective.

Notes

1. Throughout this article, all € values have been adjusted for inflation to the year 2011.

2. This calculation is based on firms' MRV cost data, provided by Löschel et al. (2011). The high cost estimates (used in Figure 2) could not be used because no differentiation was made between the small and large firms in the studies by VBW (2011) and Destatis (2011).

3. In the first trading period, 806 appeals were lodged against 1600 allocation decisions and 604 appeals against cost decisions in Germany (UBA, 2009). There were 373 appeals against allocation decisions in the second trading period (UBA, 2009).

4. It has been stated that the enthusiasm for legal disputes is a special German phenomenon (Seidel, personal communication, April 20, 2011 Berlin, DEHSt; VBW, 2011). Litigation costs in Spain and France were substantially lower at only €960 and €60 per installation or €0.9 million and €60,000 total p.a., respectively (VBW, 2011). No data are available for other instruments and other countries.

5. Firms' MRV costs are substantially higher for the US SO2 scheme than for the EU ETS. In 2001 there were 2100 regulated entities with average MRV costs of €47,000 (or $50,000) (EPA, 2001). Initial capital costs were also large. Ellerman (2000) estimated them to be approximately $700,000 per generating unit (at 371 MWe per unit).

6. We assume that, due to political economy considerations (i.e. the rent distribution interests of firms), the taxing scheme would also be designed in a downstream manner.

7. In this context, ‘large entities' refers to the producers and importers of fossil fuels.

8. There are 900 large installations (>25,000 tCO2) and 1100 small installations (<25,000 tCO2) in Germany that make up almost 99% of emissions (Community Independent Transaction Log (CITL), 2013).

9. The numbers of regulated emissions and entities are assumed to be identical in both scenarios.

References

 

Related research

People also read lists articles that other readers of this article have read.

Recommended articles lists articles that we recommend and is powered by our AI driven recommendation engine.

Cited by lists all citing articles based on Crossref citations.
Articles with the Crossref icon will open in a new tab.