Abstract
This article provides an overview of compliance carbon markets that trade carbon emission allowances and analyzes the properties of carbon as an investable asset class. The authors discuss how local supply and demand factors determine allowance prices, focusing on abatement costs and policy adjustments. They then construct a novel total-return time series for four liquid carbon markets and develop an equally-weighted Carbon Composite. They found that individual carbon markets are uncorrelated to each other and commodities and asset classes unrelated to idiosyncratic market fundamentals. The Carbon Composite generated an annualized excess return of 26.63% and a Sharpe Ratio of 1.50 over 2013–2021, suggesting that carbon as an investable asset class could provide tangible diversification benefits to investors. The authors then explore the outlook for carbon as an asset class and identify theoretical and practical justifications for a prospective carbon risk premium.
Key Findings
▪ Carbon prices are driven by local supply and demand factors, and as markets require deeper emissions cuts, carbon prices need to rise to unlock more costly emissions abatement measures, as has been evident in coal-gas fuel-switching in the EU ETS.
▪ This article constructs a total-return time series for carbon markets and builds a Carbon Composite that has generated a positive excess return with a low correlation to traditional asset classes and commodities over 2013–2021.
▪ Using known climate change policy targets, market supply trajectories, projected abatement costs, and current carbon prices, we estimate a prospective annualized carbon risk premium of 2%–12% to 2030.
From 1751–2014, human activity released approximately 1,434 gigatons of cumulative carbon dioxide (CO2) emissions and contributed to an increase in the concentration of CO2 in the atmosphere from 280 parts per million (ppm) in 1958 to 415 ppm in 2018, far above anything experienced over the past 2.5 million years (Hsiang and Kopp 2018). In 2016, the Paris Agreement was ratified by 191 members of the United Nations Framework Convention on Climate Change with the aim to limit the rise in the global average temperature to well below 2°C from pre-industrial levels by 2100. As new policies are implemented to meet this goal, investment risks and opportunities are likely to be generated. For example, Dietz et al. (2016) estimate the annual climate value-at-risk for financial assets to be $2.5 trillion, while Hong et al. (2019) and Engle et al. (2020) provide evidence of the market’s inefficiency in pricing these risks and opportunities.
Emissions Trading Systems (herein ETSs) are increasingly popular policy tools to reduce carbon emissions. An ETS sets an annually declining total emissions cap, and regulated entities must surrender tradable carbon allowances for each ton of carbon they emit, thereby internalizing the negative externality of their carbon pollution. Inclusion into an ETS is mandatory for firms in the regulated jurisdiction within the predetermined sectoral scope and with emissions above a minimum threshold. The total number of allowances available, equal to the overall emissions cap, are either freely allocated or auctioned to emitting entities that are free to trade them. The auditing of emissions is rigorous, and fines are given for non-compliance. For example, the EU ETS imposes a fine of over €100/tCO2 while also appending the outstanding compliance obligation to the entity’s following year obligation.
The trading of allowances allows for comparative advantage. Firms within the “ecosystem” of compliance entities with lower marginal abatement costs will abate and sell their allowances to firms with a higher abatement cost. In comparison, firms with higher abatement costs will buy allowances until the carbon price equals their marginal abatement cost. An ETS, therefore harnesses the market’s power and the carbon price “signal” to achieve emissions reductions at the lowest possible cost. In most ETSs, regulators recycle revenues from auctioning allowances by investing in energy efficiency and low-carbon initiatives.
Carbon markets have grown substantially since 2016 and are becoming increasingly accessible for investors. The World Bank (2021) details that carbon pricing instruments have risen from 40 systems in 2016 to 64 systems in 2021. This reflects global GHG emissions covered by carbon pricing rising from just over 11% to around 22%. Similarly, the value traded in ETS has risen from about €40 billion to €760 billion between 2016 and 2021 (Chestney 2022).
This article assesses carbon allowances as an investable asset class in numerous ways. First, we provide an overview of global carbon markets and analyze the drivers of carbon prices. Second, we construct a total-return time series for each carbon market by accounting for the collateral return and the roll impact between carbon futures contracts.1 We then create an equally-weighted four-market Carbon Composite and assess its performance, volatility, and correlation relative to other asset classes. Finally, we analyze the outlook for carbon prices considering spot-futures linkages and the drivers of a prospective carbon risk premium.
We find that carbon has generated significant positive excess returns combined with a low correlation to other asset classes, making it potentially attractive as a portfolio diversifier. For example, the Carbon Composite has generated an annualized excess return of 26.63% over the nine years of 2013–2021 with a Sharpe Ratio of 1.50 and a low correlation to all other asset classes. Because we use exchange-traded carbon futures for three of the four carbon markets, we assess the spot-futures linkage. The cost-of-carry approach explains the futures premium between spot and futures contracts, making us confident that systematic factors drive the majority of any excess return. We calculate a forward-looking prospective risk premium to 2030 of 2% to 12% annually based on current prices and climate change policy objectives. We conclude that carbon has generated attractive historical returns with a low correlation to other asset classes and has an attractive forward-looking risk premium that aligns with the Paris Agreement objectives and climate goals.
The rest of the article is structured as follows. In the next section, we provide an overview of carbon markets and analyze the drivers of carbon prices. In the third section, we describe the methodology for constructing the Carbon Composite. In the fourth section, we discuss the statistical properties of carbon as an asset class and its forward-looking outlook, and the final section concludes.
CARBON MARKETS
General Characteristics of Carbon Markets
ETSs are increasingly being introduced in jurisdictions looking to reduce their emissions at scale.2 The EU ETS is the oldest carbon market in the world, having launched in 2005, and covers 31 countries, over 12,000 entities, and 45% of total annual EU emissions. Since its implementation, annual emissions from covered entities have declined by 29% (Dechezleprêtre 2018). Similarly, the Regional Greenhouse Gas Initiative (RGGI) ETS covers emissions from electric utilities in 11 US East Coast states. Since its inception in 2009, it has contributed to a decline of 40% in emissions from covered power plants (Stutt et al. 2017). Based partially on the success of the systems above, China has launched the largest ETS in the world in 2021, covering approximately 40% of total domestic emissions (Cui et al. 2014).3
The compliance carbon markets have experienced strong growth in total value traded over the past few years. In 2017, we calculated the value traded in the EU market to be approximately $20 billion, which increased to more than $770 billion in 2021. The Western Climate Initiative (WCI) and Regional Greenhouse Gas Initiative (RGGI), markets also experienced significant increases from 2017 to 2021, with the combined value traded rising from $7.7 billion to $56 billion. Due to both increased prices and volume, the global value traded in carbon markets in 2021 was $851 billion (Chestney 2022). To put this into perspective, in 2021, the total value traded of just the active futures contracts in these carbon markets was equal to approximately 10% of the equivalent Brent Crude Oil futures contracts.
Individual carbon market prices are driven by local factors influencing demand and unique policy design aspects influencing supply. This article focuses on well-functioning and investable (i.e., liquid) markets. After screening for these two criteria based on the volume traded and accessibility for financial investors, we identify four eligible carbon markets: the European Union ETS (EUA), WCI, RGGI, and New Zealand ETS (NZU).
Compliance carbon markets can vary in policy design across five key policy characteristics. Exhibit 1 provides a summary of how the following characteristics vary across the four carbon markets analyzed in this article:
▪ Total allowance supply: the total annual volume of emissions allowances in the market.
▪ Share of allowance supply deriving from auctioning: the regulator distributes emissions allowances either through free allocation or through periodic auctions.
▪ Linear cap reduction factor: the rate at which the total annual allowance supply declines each year.
▪ Treatment of external offsets: the degree to which an ETS allows external (voluntary or project-based) carbon offset units to be used for regulated entities’ compliance purposes.
▪ Allowance banking: often, allowances can be used in later years, providing intertemporal compliance flexibility and giving firms the ability to plan an abatement strategy and hedge their future carbon cost risk.
▪ Market stability mechanisms: mechanisms implemented by the regulator to make an ETS more resilient to unexpected demand shocks by making allowance supply either responsive to price or market supply versus demand imbalances.
Supply Drivers of Carbon Markets
Fundamental Drivers of Carbon Prices
Carbon prices are predominantly driven by supply and demand dynamics and expectations of future changes in dynamics. This suggests that carbon markets operate in a manner typical of many commodity markets (such as oil, gas, and wheat) wherein prices are driven by market-specific supply and demand dynamics as well as global macroeconomic factors (Breitenfellner et al. 2009; Chiou-Wei et al. 2020; Hulshof et al. 2016; Janzen et al. 2014). This section explores how fundamental supply and demand impact carbon prices. First, we discuss the centrality of abatement costs in determining carbon prices, with a deep-dive analysis into how coal-gas fuel-switching has influenced EUA prices. Then we discuss how relevant market events and sentiment also impact carbon prices.
Abatement costs. Carbon markets aim to reduce total emissions at the lowest cost by creating a scarcity of tradable emissions allowances. This ensures that companies internalize the cost of carbon emissions, thereby incentivizing them to reduce carbon emissions. For companies to reduce emissions while maintaining production requires investment into emissions abatement measures. Since emissions allowances are tradable and the carbon price signal is uniform and widely disseminated, firms with the lowest abatement costs reduce their emissions first. A marginal abatement cost curve (MACC) ranks abatement measures from low to high costs and sums up their cumulative abatement. The theoretical equilibrium carbon price is determined by the marginal abatement cost of the last abatement measure needed for the market to match total emissions demand with total allowance supply (Kesicki 2010).4 Exhibit 2 is an example of a hypothetical MACC for an economy, where each bar represents an abatement measure with an associated cost and volume of emissions abated. In this example, if a 200 MtCO2e abatement level is required for the market to clear, the resulting carbon price would be £60/tCO2e (see dashed line).
Illustrative Marginal Abatement Cost Curve (MACC) for an Economy
Power sector fuel switching from coal to gas has been a key abatement measure in ETSs covering the power sector. Switching from coal to gas reduces emissions as coal power generation emits roughly double the GHGs per kilowatt-hour relative to natural gas power generation. However, coal as an energy source is normally less costly than gas, and therefore gas generation is only profitable with the addition of a carbon price. This encourages electric utilities to generate gas before coal in the power plant merit order. The remainder of this subsection analyzes how the MACC drives carbon prices, focusing on how fuel-switching prices have played a key role in driving EU ETS carbon prices.
In the EU ETS, previous studies have shown that the economics of fuel-switching in power producers has been a primary driver of carbon prices (Bai et al. 2019; Chevallier 2009; Chevallier and Goutte 2017; Rickels et al. 2015). Similarly, Zhu et al. (2019) decompose the drivers of EUA carbon prices over 2009–2016 and find that carbon prices are significantly influenced by electricity and coal prices over the short and medium term, with coal having a significant and negative impact on carbon prices. This supports the fuel-switching driver hypothesis because as coal prices rise, the propensity of power generators to switch to less emissions-intensive natural gas increases, thereby reducing demand for EUA allowances. Studies have also shown EUA prices to be influenced by extreme weather events or unexpected changes in weather outcomes (Alberola et al. 2008; Mansanet-Bataller et al. 2007). Weather impacts feed through the power sector as colder-than-expected temperature drives power demand through heating requirements. Increased power demand is then met by less efficient higher-marginal-cost fossil-fuel power producers, generating emissions and raising carbon allowance demand.
Furthermore, Lutz et al. (2013) find that market fundamentals and macroeconomic risk factors are central drivers of EUA prices, but these relationships exhibit time-varying characteristics best described by a regime-switching model. This supports a representation of carbon markets moving through periods where different abatement measures drive prices. A regime-switching approach also reflects the fact that carbon markets such as the EU ETS operate in distinct regulatory phases, across which the rules sometimes change. For example, as the EU ETS moved from Phases 1 and 2 (2005–2012) to Phase 3 (2013–2020), power generators no longer received a free allocation; more sectors were brought into the market; and new rules were gradually introduced, such as the Market Stability Reserve (MSR) mechanism (European Commission 2015). Similarly, moving to Phase 4 (2021–2030) has entailed an increase in the annual cap reduction factor and a further reduction in free allocations. Recently proposed amendments under the ‘fit for 55’ policy package seek to bring maritime shipping into the market, ratchet up the annual cap reduction factor, and adjust the functioning of the MSR (European Commission 2021). Supporting this, Chevallier and Goutte (2017) find that a class of mean-reverting Levy jump process models better fit the modeling of carbon market prices. The authors suggest that these models better accommodate jumps due to policy rule changes by the European Commission and other political uncertainties.
We undertake empirical analysis and find power sector fuel-switching from coal to gas has been a major driver of carbon prices in the EU ETS. We create a fuel-switching price indicator and assess how movements in this indicator correlate to EUA price movements over 2013–2020 (Phase 3 of the EU ETS). Following Chevallier and Goutte (2017), we define a fuel switching cost (Fuel_switch) as
1where EF is emissions factors, FC is fuel costs, and ηcoal and ηgas are efficiency factors for coal and gas, respectively. We use the same efficiencies and emissions factors as the authors, where the efficiency of a coal plant is 40% and a gas plant is 50%, and the emissions factor of a coal plant is 0.364 tCO2e/MWh and that of a gas plant is 0.21 tCO2e/MWh. We use one year forward fuel prices sourced from Bloomberg,5 and for gas we use the TTF contract (TTFGCY1 OECM Index). For coal we use ARA coal (API2YR1 OECM Index) inclusive of cost, insurance, and freight (6,000 kcal).
We run a series of linear regressions to assess the extent to which daily moves in the fuel-switch price explain daily EUA returns.6 Exhibit 3 illustrates the results of these regressions across different periods of analysis: (1) 2013–2020, (2) 2013–2016, and (3) 2017–2020. Previous studies found that oil prices and local equity indexes (as a proxy for economic activity and macroeconomic factors) also play a significant role in driving EUA carbon prices (e.g., Alberola et al. 2008; Chevallier 2009; Lutz et al. 2013; Rickels et al. 2015). We, therefore include two other control variables: front-month Brent Oil returns (Bloomberg ticker: CO1 Comdty) and Eurostox 50 returns (Bloomberg ticker: SX5E Index).7 The regression analysis takes the functional form shown by
2
EUA Price Driver OLS Regressions
NOTE: * indicates significance at the 10% level, ** indicates significance at the 5% level, and *** indicates significance at the 1% level.
Daily changes in the fuel-switching variable exhibit a significant and positive relationship to daily EUA returns in all periods after controlling for other key drivers. As fuel switching levels rise, EUA prices rise too, suggesting that if a certain level of fuel-switch abatement is required to clear the market, then the EUA price will increase with the fuel-switch level to ensure the continuation of the required abatement.
The coefficient on the fuel-switch variable rises in the period over 2017–2020, suggesting that fuel-switching has recently been a more influential driver of EUA prices. This can be explained by the fact that EUA prices only rose near the fuel-switching band toward the end of 2017,8 as illustrated in Exhibit 4. As the EUA price begins to cross the actual fuel-switch levels, changes in the fuel-switch indicator start to result in real EUA demand change, illustrating why the explanatory power of the fuel-switch variable increases over 2017–2020 as shown in the rising R-squared in Exhibit 3. There is substantial evidence that rising EUA prices since 2017 began to drive increasing levels of fuel switching (Agora Energiewende 2019; ICIS 2021; Marcu et al. 2020). Exhibit 5 further supports this impact channel, showing how a rolling 60-day average share of coal power generation has been sharply declining since late 2017, while the share of gas power generation has been rising, with gas power overtaking coal power in 2019.
EU Power Sector Fuel-Switching Band and EUA Price
NOTES: Fuel-switch band based on power plant efficiencies defined as: Lower – gas 43%, coal 27%; Higher – gas 61%, coal 53%.
Average Coal and Gas Generation as a Share of Total Power Generation in the EU
Carbon prices will need to continue to rise to meet 2030 emissions targets, given that carbon markets induce the lowest cost mitigation first and require deeper emissions cuts to achieve longer-term targets. In 2020, the EU announced plans to increase their 2030 climate ambition to reduce emissions by at least 55% below 1990 levels. Delivering this ambition will require increased stringency in the EU ETS such that the cap will need to decline to around 920 MtCO2e by 2030 (European Commission 2021). Average annual emissions over 2018–2020 have been 1,522 MtCO2e, implying that total annual emissions will need to decline by 602 MtCO2e over 2021–2030 (39.5%). This would equate to a required average annual emissions reduction rate of 4.9% per year over 2021–2030. Power sector coal phase-out and increased fuel switching will contribute to this abatement. However, a significant volume of industrial abatement will still be required to meet this target by 2030. Industrial abatement options are more costly than power sector fuel-switching, suggesting carbon prices will need to rise significantly to stimulate investment into these abatement options.9 ,Pietzcker et al. (2021) built an EU ETS supply and demand model using a detailed power sector model and an industrial MACC to analyze the impact of the increased EU ETS stringency to meet updated EU ambition targets. The authors find that increased stringency could lead to EU-wide coal power generation phasing out by 2030 and EUA prices reaching €129/tCO2e by 2030.
This section has focused on analyzing the abatement cost drivers of carbon prices in the EU ETS. However, similar dynamics will occur in all carbon markets. The WCI ETS and the NZ ETS are composed largely of road transport emissions, and abatement costs in the transport sector are widely acknowledged to be substantially higher than those available to the power sector (e.g., Anandarajah et al. 2013; Bothe et al. 2021; Flachsland et al. 2011; Wang and He 2017). The RGGI market covers only power producers. However, there is limited further opportunity for low-cost coal-gas switching in this market (until additional coal-reliant states join the market). In 2019, coal power generation comprised only 2.2% of all RGGI States’ net power generation, while natural gas power generation already made up 45% of power generation (EIA 2021). Meeting climate targets in RGGI will require more than just switching to natural gas in the face of increasing power demand from rising electrification and closures of coal and nuclear power plants (NESCAUM 2018). This will necessitate increasing renewable power investment. However, intermittent renewables are currently not appropriate for baseload power substitution and may require complementary investment into utility scale batteries. Utility scale battery costs in the US averaged $834/KWh in 2017, but costs are declining rapidly as this reflects a 61% reduction since 2015 (EIA 2020). While costs will likely continue to fall rapidly, RGGI carbon prices would still need to rise substantially to support the wider adoption of batteries and carbon capture solutions.
Market events and sentiment. Policy events and sentiment also drive carbon prices based on how they impact expected supply and demand balances. This section discusses the evidence regarding how relevant policy events and wider market sentiment impact carbon market prices.
Exhibit 6 shows how the EUA price responded to the policy announcements on the introduction of the MSR.10 In 2017, an official announcement detailed that the mechanism could remove a hypothetical volume of allowance supply of around 400 MtCO2e by 2019. This announcement was followed by a 12-month price appreciation of around 220%. These price impacts of the MSR announcement make sense, given that the policy implies a significant short-run impact on reducing supply (Bruninx et al. 2020; Mauer et al. 2020; Perino 2018). The MSR has also helped support prices and increase market resiliency in the face of the COVID-19 demand shock (Azarova and Mier 2021; Gerlagh et al. 2020).
The Market Stability Reserve in the European Union
In the WCI, political challenges caused market prices to briefly trade below the quarterly auction price floor, but prices quickly recovered following the resolution of the challenges, as illustrated in Exhibit 7. In 2016, a consortium of California business groups challenged the legality of the ETS, resulting in existential uncertainty for the carbon market and prices falling briefly below the auction price floor (Megerian and Vartabedian 2016). However, in April 2017, the California appellate court found in favor of the legality of the ETS, and prices subsequently began to appreciate (Danish 2017; Fehrenbacher 2017). During the Trump administration (2016–2020), several further legal challenges were launched against the California ETS. However, US courts favored the California regulator in each case, and the Biden administration has since dropped all challenges (IETA 2021).
Political Events Impacting California Carbon Prices
RGGI market prices have also responded to wider market and policy events that had implications for fundamental supply and demand in the market, illustrated in Exhibit 8. In the early 2010s, the shale gas revolution resulted in natural gas prices falling significantly, leading to natural gas power generation rising above coal power generation and hence a rapid decline in emissions (Yan 2021). This led to an oversupply of allowances and caused prices to be determined largely by the RGGI auction price floor over 2010–2013 until the RGGI regulator agreed to reduce annual allowance caps over 2014–2020 (EDF 2015).
RGGI Prices Responding to Key Market and Policy Events
Market Stability Mechanisms in New Zealand
Similarly, the proposal and subsequent repeal of the US Clean Power Plan (CPP) was a key driver of RGGI prices over 2014–2017. In June 2014, the US Environmental Protection Agency first proposed the CPP, a national mechanism to regulate power plant emissions. It did so by introducing state-wide emissions targets for the power sector while providing states flexibility in achieving emissions targets. The flexibility included utilizing carbon markets (US EPA 2015). RGGI was thought to be a potential model for compliance with the US CPP, with observers noting that it could result in more states joining RGGI (Ceres 2015; Walton 2015). RGGI prices began to appreciate from mid-2014 onwards, partially because of new cap adjustments imposed by the RGGI regulator (discussed above) and expectations surrounding the growth of the RGGI market. They required stringency increases under the CPP (EDF and IETA 2018). In August 2015, the CPP was challenged by 27 states in the US Court of Appeals requesting an emergency stay. In February 2016, the US Supreme Court ruled against the CPP (Hurley and Volcovici 2016), causing RGGI prices to fall by around 40% (Congressional Research Service 2019).
In the NZ ETS, policy announcements of changes in market rules have impacted prices prior to the actual implementation of rule changes. In the early years of the NZ ETS, the market was fully linked to the Clean Development Mechanism (CDM) Certified Emissions Reduction (CER) offsets, which meant that CERs were fully fungible with NZU allowances. CER prices collapsed (due to the oversupply of units and the Global Financial Crisis of 2008), causing NZU prices to fall to 2 NZ$/tCO2e in early 2013. In late 2012, the NZ government confirmed that the NZ ETS would fully delink from the CDM by June 2015 (Leining and Kerr 2018). This announcement led to the NZU price gradually appreciating in anticipation of future delinking while CER prices remained between 0.10–0.50 US$/tCO2e over the same period.
Another feature of the NZ ETS was the Fixed Price Option (FPO), which previously allowed entities to purchase an unlimited number of allowances at a fixed price of 25 NZ$/tCO2e (i.e., a price ceiling). In late 2019, the NZ government announced that the FPO level would rise to 35 NZ$/tCO2e in 2020, causing an immediate price appreciation (Energy Market Services 2020), followed by a retracement due to COVID-19 and then a further rise back up to the 35 NZ$/tCO2e level. Subsequently, the price breached the FPO level in anticipation of further anticipated reforms to the NZ ETS, including the introduction of auctions in 2021 and the replacement of the FPO with an auction Cost Containment Reserve (CCR) that would have a trigger price beginning with 50 NZ$/tCO2e (NZ Ministry for the Environment 2019).
There is also evidence that market sentiment influences carbon price movements in the EU ETS. Ye and Xue (2021) construct a carbon tone index through textual sentiment analysis of news articles on the EU ETS. The authors find the tone index to have significant power to predict EUA moves. Similarly, Cummins et al. (2016) analyze market sentiment using Twitter posts relating to European emissions and find that sentiment has a significant relationship to both EUA price and price volatility. There appears to be no evidence that wider climate announcements or events not directly related to market fundamentals have any impact on allowance prices. These wider events (such as the UN Congress of Parties (COP) conferences, presidential elections, or significant IPCC report publications) could indirectly affect carbon markets. This may be an avenue for future research.
CARBON AS AN INVESTABLE ASSET CLASS
We now look at carbon markets from the perspective of an investable asset class. For each market, the EUA, the WCI, the RGGI, and NZU, we create a total-return time series based on the price of one metric ton of carbon dioxide equivalent (tCO2e) in US dollars and then construct an equally-weighted Carbon Composite portfolio.11
Constructing a Continuous Total-Return Time Series
We use daily data on December expiry futures contracts for EUA, WCI, and RGGI and spot physical carbon prices for NZU. Where futures contracts are used, we construct a continuous contract time series that accounts for the collateral return and the roll yield and reflects the total returns generated for an investor who rolls from one contract to the next, avoiding taking physical delivery.12 This investor will earn the short-term cash yield on cash collateral posted as a margin for holding the futures contracts. The roll is calculated in two steps. First, we calculate an adjustment ratio and apply this to all previous prices outside of the roll window to account for the roll yield. Second, we assess daily open-interest and volume for prices in the roll window to determine the optimal roll period and move from one contract to the next equally over a 10-day period. We define the roll period as the last five trading days of November and the first five trading days of December: the time when most market participants choose to roll between contracts and when liquidity is abundant.
Adjustment ratio. We define the adjustment ratio as the average price of the current futures contract during the 10-day roll period divided by the average price of the next futures contract during the same 10-day roll period. This captures the roll yield in the most liquid period of the contracts, where most investors roll from one contract to the next. We apply this adjustment ratio backward to the observed futures prices of the current contract. In practice, if futures prices are in contango, the adjustment ratio reduces returns to an investor.
Prices during the roll window. During the 10-day roll period, we move incrementally into the next contract. We do this by reallocating the investor’s position each day by 10% into the next December contract. For example, on day 1 of the roll period, the investor holds 90% of the current futures contract and 10% of the next contract. On day 2, she holds 80% of the current futures contract and 20% of the next contract. On day 10, she holds 100% of the next futures contract, and the roll has been completed.
The Carbon Composite
We then combine the carbon market continuous time series to create a Carbon Composite portfolio. The Carbon Composite is created by equally weighting the four carbon markets and rebalancing annually.
Two data providers currently provide indexes based on carbon market futures; however, key design differences make the Carbon Composite unique from these indexes. Both the Intercontinental Exchange (ICE) and IHS Markit have launched carbon market futures indexes using futures contracts in the EU ETS, RGGI, and WCI. Exhibit 10 describes each index in detail. The Carbon Composite developed in this article has three novel features that make it distinct from either of the indexes currently available:
1. The Carbon Composite includes the NZ ETS. As there is only a spot market and no futures market currently in the NZ ETS, this also means that the Carbon Composite is the only approach to contain both futures and spot physical carbon contracts.
2. The Carbon Composite is the only index to use an equally-weighted approach, while both the other indexes use a value-weighted approach. An investor with strict liquidity constraints may therefore find their potential investment size to be limited if following the Carbon Composite. However, this weighting methodology lowers the volatility of the index, whereas a value-weighted approach results in the highest weight in the most volatile market. Over time, an equally-weighted approach will also allow investors to benefit from a longer-term convergence in global carbon prices (Dellink et al. 2014; Haug et al. 2015; Tuerk et al. 2011).
3. Each index has a unique roll methodology, with the ICE index rolling over three months, the Carbon Composite rolling over 10 days, and the IHS Markit index rolling on a single day.
Comparator Carbon Futures Indexes
To analyze the statistical properties of carbon in the next section, we use daily data of comparator contract indexes. These contracts include gold, silver, Brent crude, natural gas, coal continuous futures, the MSCI World Total-Return Index, the Barclays Global Aggregate Bond Index, and the Bloomberg Commodity Total-Return Index, from Bloomberg and the ICE between January 1, 2013, and December 31, 2021. For the risk-free interest rate, we use the 1-month USD annualized and continuously compounded LIBOR rate. We divide this by 365 to calculate implied daily yields.
RESULTS
In this section, we first discuss the performance and statistical properties of both individual carbon markets and the Carbon Composite. We then analyze the prospective carbon risk premium by assessing the spot-futures linkages and the potential forward-looking return outlook.
Performance and Properties of Carbon
We use the continuous carbon total-return time series to analyze the statistical properties of the individual carbon markets and the Carbon Composite over a nine-year period from January 2013 to December 2021.13
Individual carbon markets. The four markets generated positive annualized excess returns over the analysis period, ranging from 6.85% to 38.13% (Exhibit 11). All markets except the WCI have been characterized by high volatility as measured by annualized standard deviation, resulting in low risk-adjusted returns as defined by the Sharpe Ratio.
Carbon Market Performance
Carbon market annualized nominal returns have been significantly higher than those of global equities, bonds, and commodities over the same period (Exhibit 12).
Annualized Nominal Return and Sharpe Ratio
There is a low correlation between the four markets and between the carbon markets and traditional asset classes (Exhibits 13 and 14). This is because carbon prices are formed due to unique supply and demand characteristics within each carbon market and allowances are not fungible across markets, as discussed earlier.
Correlation Matrix between Carbon Markets and Other Assets
Carbon Market Indexed Returns
A low correlation does not provide utility to investors if the correlation increases during periods of market stress. Therefore, we analyze each market’s performance during major equity market drawdowns of 10% or greater (Exhibit 15). Each market has exhibited varied performance during these declines. For example, from February 2020 to March 2020, global equities declined 34%, while EUA carbon declined by 36%, RGGI carbon declined 19%, WCI declined 24%, and NZU declined 25%. In contrast, other equity market drawdown periods have demonstrated little correlation with carbon markets. On average, carbon markets have risen during these drawdown periods, and this may make carbon an attractive portfolio diversifier.
Carbon Markets Performance during Drawdown Periods
Carbon Composite. The Carbon Composite produces higher returns and lower volatility than the simple average of the four carbon markets (Exhibits 16 and 17). The low correlation between the four carbon markets lowers portfolio volatility and increases the Sharpe Ratio. Rebalancing more frequently does not affect these results significantly. The Composite also outperforms global equities and commodities (Exhibit 18).
Carbon Composite Performance
Equal-Weighted Carbon Portfolio Performance
Carbon Composite Index Returns
We next assess the performance of the Carbon Composite during equity drawdowns that exceed 10%. We find the Carbon Composite declines in only one of the three periods. On average, the performance of the Carbon Composite during equity drawdowns between 2013–2021 is 14%, versus −24% for equities and −20% for commodities (Exhibit 19).
Carbon Composite Performance during Drawdown Periods
Given the low correlation to traditional asset classes, we investigate the impact of adding a 10% pro rata allocation of the Carbon Composite to a 60/40 equity/bond portfolio (Exhibit 20). The pro-rata addition of carbon into a traditional portfolio results in a significant increase in both the annualized return and Sharpe Ratio compared to a benchmark 60/40 portfolio. The annualized return increases by 2.24%, and portfolio volatility declines from 8.25% to 8.07%, leading to a higher Sharpe Ratio. Consequently, multi-asset investment strategies may benefit from adding the Carbon Composite to diversified portfolios. Exhibit 21 suggests that a 45% allocation to carbon in a traditional 60/40 equity/bond portfolio would be optimal for the investor looking to maximize risk-adjusted returns.
Carbon Composite Addition to a Traditional Portfolio
Impact of Adding Carbon to a 60/40 Portfolio
Another important consideration for portfolio management is whether carbon is a better diversifier than other commodities. The Carbon Composite outperforms other commodities (Exhibit 22) and generates higher risk-adjusted returns between 2013–2021, despite the 2021 European Energy Crisis causing rapid increases in TTF and coal prices (Exhibit 23).14
Indexed Commodity Returns
Commodities Performance
Twelve-month rolling correlations between carbon and other commodities (Exhibit 24) show that the Carbon Composite is largely uncorrelated to commodities that do not have a fundamental link to the carbon markets. Exhibit 24 shows that the Carbon Composite has exhibited a weak positive correlation to Rotterdam coal (19% on average), TTF gas (23% on average), and Brent oil (16% on average) over 2013–2021. Rotterdam coal, TTF gas, and Brent oil show a persistent correlation to the Carbon Composite through this period, which can be explained by the fundamental impact these fuels have on setting power plant merit orders and fuel switching abatement costs in several of the markets. Koenig (2011) analyzes gas, coal, and oil correlations to carbon allowance prices and finds that the correlation is dependent on carbon prices being at a level required to incentivize fuel-switching decisions and that correlation decouples as allowance prices exceed or fall below fuel-switching levels.
12-Month Rolling Correlation between Carbon and Commodities
The remaining commodities (gold, silver, and HH gas) exhibit a much weaker correlation to the carbon composite of 10.6% on average over 2013–2021. Gold and silver have no fundamental link to carbon prices and thus a lack of correlation as expected. HH gas prices would only potentially influence the smallest carbon market (RGGI), but HH gas prices over this period were so low that the RGGI carbon price has largely been outside the coal-gas switching price range (Lee 2014), and in the framework of Koenig (2011) would likely have a decoupled correlation to carbon allowances.
Forward Looking Outlook for Carbon
Since three of the four carbon markets that we analyze use liquid futures markets as the pricing source, we begin this section by discussing the linkage between carbon spot physical-futures. We then explain the prospective carbon risk premium to 2030.
Carbon spot-futures linkage. The establishment of a futures market linked to carbon allowances enhances the efficient functioning of an ETS, allowing a wider ecosystem of traders to bring liquidity and price discovery to the market. Exhibit 25 discusses the benefits of a liquid futures and options market with low transaction costs, providing risk management tools and price discovery in carbon markets.
Liquid Futures Markets Providing Significant Benefits to Carbon Markets
As we are using carbon futures for EUA, CCA, and RGGI, it is key to ensure that any calculated risk premia are not a misidentified function of the futures relationship to spot prices. We, therefore, model the futures premium given the spot-futures relationship using the cost-of-carry approach.
Cost-of-carry approach. For the cost-of-carry approach, a basic theory of storage posits that the futures price of a commodity, Ft,T, reflects its current spot price, St, foregone compound interest between periods t and T, yt,T, storage costs, and a convenience yield. Since holding carbon allowances has virtually no storage costs, and allowances are bankable, storage costs and the convenience yield are zero (Chevallier 2010). The cost-of-carry approach is, therefore
2Taking logs of Equation (2), we have
3The futures price consequently depends on the spot price and foregone interest, with an effective arbitrage boundary on this relationship since carbon is a cash-funded security. The futures premium is calculated as
4We show the price of various expiry futures contracts for European carbon allowances in 2013 (Exhibit 26) and 2018 (Exhibit 27). In 2013, the price of a contract expiring in December 2020 was between €6–€10, while in 2018, the price of a contract expiring in December 2025 was between €7–€30. This larger differential is partly due to the introduction of the MSR raising expectations of future prices, as discussed earlier.
Carbon Futures Prices in 2013
Carbon Futures Prices in 2018
In Exhibit 28, we show that the premium in carbon futures of various maturities is not attributed to the spot-futures relationship using the cost-of-carry approach, since the unexplained premium is small. The maximum unexplained premium between carbon spot-futures is €0.25 or 4% of the price of carbon on January 2, 2013. Results are similar in the other carbon markets.
Unexplained Futures Premium: Cost-of-Carry Approach
We therefore conclude that historical returns from investing into futures linked to carbon are driven by systematic factors relating to supply and demand in each market and not related to the unexplained premium of the futures contract versus the spot price.
Carbon risk premium. Looking forward, we estimate a prospective carbon risk premium by assessing estimates of carbon prices required for alignment with the Paris Agreement 2°C temperature target.
We use a range of 2030 carbon price estimates from the International Energy Association (IEA) Perspectives on the Energy Transition (2018) and the mitigation pathways database hosted by the International Institute for Applied Systems Analysis (IIASA) (Huppmann et al. 2019). The IEA estimates that a carbon price of $120/ton by 2030 is needed within OECD countries to stimulate enough emissions reductions to achieve the Paris Agreement. The IIASA provides a collection of the quantitative mitigation pathway modeling underlying the IPCC’s Special Report on Global Warming of 1.5°C (SR15). This database provides estimates of the carbon price required by 2030 across 15 different models15 and 52 different pathway scenarios. The median 2030 carbon price forecast generated by these models suggests that a price of US$125/ton will be required by 2030 for the world to be on a trajectory aligned with limiting climate change below 2°C by 2100, as illustrated by the box-whisker plot in Exhibit 29.16
Climate Model Estimates: Carbon Prices by 2030 to Be Aligned with a 2°C Scenario
Taking a possible range of 2030 carbon price estimates of $80 to $150 per ton and an average carbon price (on December 31, 2021) of $46.77, we calculate a prospective risk premium of 4% to 12% based on a standard internal rate-of-return model and a 10-year risk-free rate of 2% as shown in Exhibit 30.17 Given this policy preference, low correlation, and the demonstrable excess return generated by carbon since 2013, carbon as an asset class may provide attractive diversification and performance benefits to investors. It is important to note that this prospective risk premium is based on the prices of physical carbon allowances and not the expected return for holding and rolling the futures contracts available in these markets. As of December 31, 2021, the weighted level of contango or negative roll yield in the carbon composite was approximately 2% per annum, which would reduce the risk premium by this amount for investors gaining exposure to carbon purely through the futures market.
The Prospective Carbon Risk Premium to 2030
Aside from model-based forecasts of required carbon prices by 2030, all four carbon markets within the Carbon Composite have attractive supply and demand profiles over the next 10 years.
Essential to the design of carbon markets is the annually declining allowance cap. For 2021–2030, this annual reduction is currently 3.5% in the WCI ETS, 3.8% in the RGGI ETS, and 2.2% in the EU ETS. However, the European Commission’s recent ‘fit for 55’18 policy package would almost double the EU ETS linear cap reduction factor to 4.2%.19 Further to the linear cap reduction factors, both the EU and RGGI have additional supply-reducing mechanisms in place, as illustrated in Exhibit 31.
Carbon Market Supply and Potential Demand Growth (2021–2030)
The EU ETS, RGGI, and the WCI will require faster emissions reductions than recently achieved if they are to meet climate targets. Exhibits 32, 33, and 34 show that if we assume that annual emissions declines continue at the same fixed level as experienced over the past few years, then emissions in each market will exceed their next climate target.20 This implies that additional abatement investments and/or drawdowns on entities’ banked allowances will need to be incentivized for allowance demand to meet allowance supply. This incentivization will require a rising carbon price, as discussed earlier.
EU ETS Emissions Scenario versus 2030 Cap
NOTE: Simple emissions forecast assumes EU emissions continue to decline by 43 MtCO2e experienced over 2013–2019.
RGGI Emissions Scenario versus 2030 Cap
NOTES: Simple emissions forecast assumes RGGI emissions continue to decline by 4.49 million short tons CO2 experienced over 2013–2019. Forecast includes New Jersey and Virginia 2019 emissions as these states join the RGGI market in 2020 and 2021, respectively. Only 2025 emissions target shown as 2030 targets are likely to be updated during the next RGGI 2022 program review (discussed below).
WCI ETS Emissions Scenario versus 2030 Cap
There is also reason to believe that further ETS policy stringency will occur over time. The EU’s announcement of 2030 climate ambition ratcheting up reflects a growing societal pressure for politicians to act on climate change. Parties focusing on environmental issues have gained significant traction in recent elections throughout Europe (Graham-Harrison 2019; Mallet et al. 2019). However, ambition raising is also evident through the enaction of the ‘ratchet mechanism’ within the PA. The Paris Agreement (PA) intends for countries to reassess their Nationally Determined Contributions (NDCs) every five years and ratchet up ambition in the case of ambition gaps. 2021 is the first five-year cycle reassessment (Yeo 2015). Any rise in national ambition will be likely to feed into stricter ETS policy design, as we have seen in the EU. Numerous other countries have already declared their first ambition ratchet, with many governments opting to target a net-zero emissions deadline date in their updated NDC; see Exhibit 35 (UN 2021).
Furthermore, most existing ETSs have periodic review cycles through which policy adjustments are implemented if additional stringency is required. RGGI has already completed two program reviews, each of which resulted in increased stringency, such as implementing bank adjustment mechanisms to reduce supply and introducing a second price floor mechanism in the form of the Emissions Containment Reserve (ECR). In 2021, RGGI began its third program review, which will focus on policy design, ETS supply caps, and environmental justice considerations, among other topics (RGGI Inc. 2021a). Similarly, California has periodic reviews of its Climate Change Scoping Plan to meet its long-term targets. The next update of this plan is scheduled for 2022, and a key part of the analysis will be to identify the role played by various climate policies and whether any revisions are required. California’s Independent Emissions Market Advisory Committee (IEMAC) has already called for several revisions to improve the functioning of the WCI ETS, including raising the emissions reduction contribution of the ETS by reducing supply to account for the large accumulated allowance bank (Burtraw et al. 2020).
CONCLUSION
This article provides an overview of compliance carbon markets and analyzes key fundamental drivers of carbon prices. It builds a new Carbon Composite that reflects the total return to investors in a globally diversified, liquid, and investable carbon portfolio.
We find that carbon markets, similar to other commodities, have prices driven largely by underlying fundamental supply and demand factors. We discuss the centrality of abatement costs in forming carbon prices and find that fuel-switching requirements have become an increasingly significant determinant of EUA prices over 2017–2020. We also provide qualitative evidence that policy adjustment events influencing supply and demand have significantly impacted price formation across all carbon markets. Carbon markets have also demonstrated price sensitivity and volatility related to policy decisions and macroeconomic conditions. Therefore, a carbon investment strategy may benefit from both an informed asset allocation across markets and ongoing risk management governed largely by the fundamental and policy outlook in each carbon market.
We then analyze the statistical properties of carbon both on an individual market basis and as part of the Carbon Composite. We find a low correlation between the individual carbon markets and between the markets and other asset classes. When individual markets are combined into a global Carbon Composite portfolio, carbon as an asset class has generated high risk-adjusted returns producing an annualized excess return of 26.63% and a Sharpe Ratio of 1.5 over 2013–2021. Due to its low correlation to traditional asset classes, the inclusion of carbon into an equity/bond portfolio has increased returns without increasing volatility, thereby improving portfolio efficiency.
Based on an understanding of the remaining carbon budget for the 2°C threshold combined with a marginal abatement cost curve, multiple third-party sources have estimated that a carbon price of at least $80 to $150 per ton will be required by 2030. This translates into an annualized prospective carbon risk premium to 2030 of between 4% and 12%. This risk premium is also supported by current ETS policy supply and demand characteristics, where we show that an acceleration in emissions reductions will be required for the EU ETS, RGGI, and CCA to meet their upcoming emissions targets. Increasing ETS stringency is also likely to drive further price appreciation in the future, given the NDC ‘ratchet mechanism,’ growing societal pressure, and the cycle of periodic ETS reviews.
Carbon markets may therefore have value as both a strategic asset class for a long-term investor and as a powerful climate change policy instrument. This combination of positive expected returns, low correlation, and alignment with climate change policy makes carbon an attractive “green” asset since investing in carbon aligns the investor with climate change policymakers and the Paris Agreement objectives.
ENDNOTES
↵1 We focus on futures since they constitute the majority of the value traded in these markets.
↵2 Eight ETSs are currently operational (EU, US/Canada [WCI], US [RGGI], South Korea, New Zealand, Kazakhstan, and Switzerland, China), four scheduled, and nine being considered (e.g., Mexico, Colombia, Brazil, Chile, Turkey, Vietnam, and Thailand).
↵3 For a review of the merits and workings of carbon markets, see Newell et al. (2013) and Narassimhan et al. (2018).
↵4 While MACC curves are useful constructs, care must be taken to recognize the limitations of specific MACC curves, which often have limited consideration of intertemporal dynamics and uncertainty and may have significantly influential underlying assumptions with respect to fuel prices and technology costs (Kesicki and Ekins 2012).
↵5 One-year forward contracts are used to reflect that fact that most power utilities hedge their forward sales of power to at least one year into the future.
↵6 Using front December EUA futures contracts to analyze EUA returns.
↵7 Some previous studies also show EUAs to also be related to electricity prices (e.g., Alberola et al. 2008); however, we have left this variable out of the regressions due to a high degree of collinearity between the fuel-switch variable and electricity price returns, which could bias the fuel-switch coefficient. This is explained by the fuel-switch variable being composed of natural gas and coal prices, which are key determinants themselves of electricity prices.
↵8 The fuel-switching band reflects the variety of efficiency distributions of power plants in a real-world power sector (Chevallier and Goutte 2017).
↵9 Joas et al. (2019) study industrial abatement costs and find direct reduction abatement costs for steel to be EUR 60/tCO2e with natural gas and EUR 99/tCO2e with hydrogen; similarly in the chemicals sector, green hydrogen from electrolysis is estimated to be EUR 170/tCO2e and methanol-to-olefin to be at EUR 160/tCO2e, while in the cement sector carbon capture from oxyfuel process is around EUR 70/tCO2e.
↵10 The MSR withdraws or injects allowances depending on the level of the Total Number of Allowances in Circulation (TNAC), thereby supporting market tightness in the face of unexpectedly low emissions demand.
↵11 We use the spot exchange rate to convert non-USD carbon prices to USD carbon prices.
↵12 Where spot prices are used, we construct a simple return series.
↵13 We exclude the period before 2013 to avoid analyzing the pilot phases of the carbon markets during which rules were still being refined and market participants getting familiar with market functioning.
↵14 Gas TTF represents Netherlands TTF gas prices, while Gas HH represents US Henry Hub gas prices.
↵15 Models assessed: AIM/CGE 2.0; AIM/CGE 2.1; GCAM 4.2; IMAGE 3.0.1; MESSAGE V.3; MESSAGE-GLOBIOM 1.0; MESSAGEix-GLOBIOM 1.0; POLES ADVANCE; POLES CD-LINKS; POLES EMF33; REMIND 1.7; REMIND-MAgPIE 1.7-3.0; WITCH-GLOBIOM 3.1; WITCH-GLOBIOM 4.2; WITCH-GLOBIOM 4.4.
↵16 One outlier price estimate of US$1,424/ton by 2030 not shown in Exhibit 26.
↵17 If the newly implemented UK-ETS carbon market is included in this calculation, the December 31, 2021, average carbon price rises to US$58, and the prospective risk premium decreases to 2%–10% by 2030.
↵18 This proposal is the policy adjustments that the European Commission has developed to support the EU in reaching its recently strengthened 2030 climate target of lowering emissions by at least 55% below 2005 levels (previous target was 40%). This package proposes a number of policy revisions, one of which is a strengthening of the EU ETS through various means, including updating the 2030 cap, the “linear reduction factor,” expanding into the maritime transport sector, and imposing a carbon border carbon adjustment.
↵19 To be accompanied by a once-off downward adjustment of the cap when the policy is implemented to align the cap with the trajectory it would be on if the 4.2% had been implemented in 2021.
↵20 All emissions forecast scenarios ignore 2020 emissions, given the impact of COVID, and instead forecast emissions from 2019 reported levels.
- © 2022 Pageant Media Ltd
Jump to section
Similar Articles
Cited By...
- No citing articles found.