Failings of market failure as a guide to policy

The orthodox economic framework for analysing climate policy in terms of correcting market failures contains important insights, but is too narrow to identify a complete set of policies. 

This post is a little more abstract than usual, with no data or charts.  It looks at the framework that often underlies analysis of policies for reducing greenhouse gas emissions, though only taking a brief look at a subject that warrants much more extensive treatment.

The orthodox policy framework around emissions mitigation, especially among economists and particularly in developed English-speaking economies, consists of identifying, assessing and fixing market failures.  At the risk of slightly caricaturing this approach, essentially it stipulates that the costs of damages from greenhouse gas emissions are not normally factored into economic decisions, a major market failure which needs to be addressed by pricing emissions.  Other market failures may also need to be addressed if emissions reduction is to be achieved efficiently.  For example, there is a divergence of incentives between tenants and landlords (a principal-agent failure), which needs to be addressed by other interventions such as building standards.  Once market failures have been identified and fixed, markets can be left to do their job of making efficient production and consumption decisions.

This framework contains some valuable insights.  Carbon pricing has the potential to help reduce emissions at lowest cost, because there will be many different opportunities to reduce emissions, and a market will help reveal where the low cost abatement can be achieved, and incentivise cost effective action.  And asking what might stop energy markets from operating efficiently can provide important – though not fully comprehensive – insights into where policy interventions might be needed.

However in practice the characteristics of the energy sector imply that the necessary range of policies and activities is much broader than the framework of market failures alone would suggest.  (In other sectors, especially land use, which accounts for a substantial proportion of global greenhouse gas emissions, the limitations of the market failure framework are if anything even greater, with a diverse set of policy approaches needed – but more on this will need to await a future post).  A narrative that focusses exclusively on fixing failures so that markets can do their job is likely to lead to important dimensions of policy being missed.  At best the framework  of market failures highlights where problems may exist, but in many cases says little about how they can best be addressed.

For example, the characteristics of electricity production and delivery mean that electricity markets must be designed, and the form they take can have a major effect on the extent to which they enable a transition to a system with lower emissions.  Electricity markets may, for instance, differ in the way they deal with a preponderance of generation with very low marginal costs, increased interaction between supply and demand (enabled by smart metering and related trends), and likely greatly increased use storage.  An understanding of economic principles will surely help achieve effective market design, but labelling the need to design a market as a market failure does little to help.

Related to this, electricity and gas flow through networks with strong natural monopoly or oligopoly characteristics.  Network development, operation and pricing are inevitably subject to regulation, which may be more or less effective in stimulating low carbon investment.  Again, characterising network monopoly as a market failure, while doubtless valid, does little to tell you how networks need to be designed to enable lower carbon energy systems.

Furthermore, people’s behaviour often does not conform to the simple rational profit maximisation that much mirco-economic theory assumes.  This is more of a failure of (traditional) economic theory than of markets themselves.  This has major effects in areas such as energy efficiency, and a range of policies is typically required to address behavioural characteristics.  Behavioural economics and psychology will offer more insight here.

And, crucially, decarbonisation requires a fundamentally new energy system, including new technologies, notably for electricity storage.  The role of technology policy is too large a subject to go into here, but it is worth noting that the state and various types of institutions other than private corporations, such as universities and research institutes, have played a major role in fundamental innovation in a range of sectors.  It is likely that many such institutions have a role to play in decarbonisation, and looking at their role within a framework of market failure is unduly constraining, and risks misinterpreting their appropriate role and likely behaviour.

More broadly market based instruments such as carbon pricing are just that: instruments for achieving a goal.  The goal cannot be defined by markets, not least because choices affect those still to be born and the type of world they will inherit may differ fundamentally, not just marginally, from the present, depending on the choices made today.

Furthermore, all markets are embedded in wider society, including political and legal institutions.  These institutions and their actions must continue to have legitimacy if the transformation to a lower carbon society is to be effectively achieved.  Enormous efforts are required to retain public support for actions to address the climate change problem, and for the institutions that must implement emissions reductions policies, including carbon pricing.  Maintaining and enhancing this social and political capital and infrastructure is a task that goes far beyond fixing market failures.

It is possible to force-fit some of these issues into a framework of market failures, but doing so risks limiting the range of policy options and associated actions that are considered.  A much broader and more active approach is needed, asking what needs to be done and what tools and approaches are best suited to reaching a goal.  Markets have a major role to play in an effective policy programme.  But talking exclusively within a framework of market failures is likely to miss important dimensions of good policy design and many aspects of the activities that need to be undertaken to achieve the scale of decarbonisation needed to reduce the risks of climate change to more acceptable levels.

Adam Whitmore -  4th April 2014

The case for the EUETS market stability reserve looks in need of further evidence

The market stability reserve proposed for the EUETS would be a significant change to the world’s largest carbon market.  It also raises wider question about attitudes to the efficiency of carbon markets over time.

In response to the continuing oversupply of EU Allowances (EUAs) in the EUETS the European Commission has proposed a “market stability reserve” to come into operation in 2021, at the start of Phase 4 of the EUETS.  The intention of the mechanism is that EUAs will be withdrawn from or returned to the market depending on the size of the cumulative surplus of allowances, which currently stands at around 2 billion tonnes.

Specifically, when the surplus of EUAs exceeds 833 million tonnes allowances are put into the reserve, with 12% of the surplus being put in each year.  When the surplus (excluding the allowances in the reserve) falls below 400 million then 100 million allowances are returned to the market each year.  The need for reporting of the volume allowances means that in effect the calculation of the surplus is based on the surplus 18 months to 2 years earlier.  The quantity of allowances put into or withdrawn from the reserve does not depend directly on the price of allowances.  The parameters and date of introduction of the mechanism all remain subject to change.

It appears likely in practice that a large surplus of allowances will persist to 2020, and that the mechanism will effectively place EUAs into the reserve from the early to mid-2020s, returning them to the market in the late 2020s and 2030s.  The size of the reserve is likely to peak at somewhere around 1 billion allowances in the mid- 2020s (with quite a wide range around this).  The timing of the return is largely due to the chosen rate of return of 100million p.a., and the time lag in the calculation of the size of surplus.   The chart below shows for illustrative purposes only one of many possible scenarios for the flow of allowances to and from the reserve, some of which would include later return of allowances.

Illustration of possible flows into and out of the EUETS market stability reserve …

 Flow into reserve

Apparent lack of environmental benefits

The proposed mechanism will only affect cumulative emissions over the long term if either allowances in the reserve were cancelled, or the presence of the reserve affects future caps, for example allowing a tighter cap to be set in the 2030s.

Cancellation of allowances is excluded from the proposals, and could in any case easily be achieved separately, so does not seem likely to provide a rationale for the mechanism as a means of providing environmental benefits.

There is also no apparent intention to enable tighter caps in future as a result of the number of allowances in the reserve.  And it is not clear that the presence of a formal separate reserve (as distinct from privately held surplus allowances) would lead the EU to tighten the post 2030 cap, although this remains a possibility.

There is thus no clear reason why the mechanism as proposed would provide any additional environmental benefit.  Consequently it seems likely to be beneficial only if it leads lower total cost of meeting the same emissions reduction goals.

Will the mechanism reduce the costs of abatement?

The effect of the mechanism on total costs seems likely to depend on the expected level of EUA prices in the 2030s relative to the 2020s, and the response of the private sector in the 2020s to expected prices in the 2030s, specifically they extent to which they will bank allowances, which is not restricted under the EUETS.

It might be that in the late 2020s the costs of abatement, and thus allowance prices, are expected to be constant or lower, in the 2030s compared with the late 2020s.  This might be, for example, because the costs of low carbon technologies are falling rapidly, or gas prices are falling and fuel switching is still the marginal price setting form of abatement.  In this case the reserve would force banking of allowances (in the broad sense of holding allowances for use at a future date) that the private sector would not undertake.  This would raise prices in the late 2020s and depress them in the 2030s, but with some net increase in (discounted) average prices resulting from inefficiently mandated scarcity in the market in the late 2020s.  The reserve might thus (on expectation) increase total costs.

In contrast, it might be that in the late 2020s prices are expected to rise strongly into the 2030s.  In this case there could be extensive private sector banking from the 2020s into 2030s.  The stabilisation reserve might have little effect, as it would simply (partially) replicate the effect of private sector banking.  The total number of allowances held for future use, and thus the supply demand balance in the market, largely unaffected.  The reserve might thus match the efficiency that private sector banking could accomplish (in the case of rising prices), while sometimes proving less efficient (in the case of flat or falling prices), leading to no net benefit.

Alternatively, the stability reserve could increase efficiency if for some reason private sector banking were not happening in sufficiently large quantities, despite an expectation of rising prices.  This might be due to perceived to perceived political risk or short commercial planning horizons.  However, the presence of severe inefficiencies from lack of private sector banking does not appear very likely.  By the late 2020s the EUETS will have been in place for more than two decades, and as such will be a very well-established policy mechanism.  Credible announcements will likely have been made on the cap for Phase 5 (the 2030s).   Even now allowance prices are well above zero even though the market is expected to remain long for perhaps a decade, implying that the future value of allowances does indeed influence their current price.

A final possibility is that higher prices in the short term would stimulate greater investment in low carbon technologies, preventing lock-in and reducing abatement costs in the long term.  This would be a further effect of perceived political risk or limited commercial time horizons.

The case for the stability reserve seems to rest on either permanent reductions in supply, which are currently unintended and could be introduced in other ways, a reluctance by the private sector to bank allowances in the event of expected price increases, or increased low carbon investment triggered specifically by some increase in scarcity in the late 2020s.  Proponents of the mechanism have not as far, as I am aware, demonstrated that these market features are likely to arise, or that the stability reserve would be effective in addressing them, let alone that it would be the most effective way of doing so.  The case for the stability reserve still needs to be made.

Adam Whitmore – 21st March 2014

Making a low carbon future better as well as cheaper

Framing of decarbonisation pathways needs to take the value in use of low carbon technologies into account.  This can provide a fuller and more positive guide to policy than analysis of marginal abatement costs alone. 

Much analysis of pathways for decarbonising economies takes as its starting point Marginal Abatement Costs (MACs), looking at the cost per tonne of reducing emissions.  This is a useful perspective, for example highlighting the cost effectiveness of improved insulation in buildings.  However, framing decarbonisation as a problem of costs incurred in reducing emissions risks ignoring other characteristics of a low carbon economy.  A broader and more positive framing needs to consider how a more attractive low carbon future can be realised.  This broader framing emphasises some of the potential benefits of low carbon technologies, as well as focussing on non-price barriers to adoption.  Such a framing can offer a more useful guide to the range of policies needed to develop low carbon pathways.  (I should also note that carrying out reliable MAC analysis can itself pose significant challenges.  Some of these are reviewed at the end of this post, but here the main focus here is on those issues difficult to accommodate within the MAC framework).

MAC analysis tends to assume that the reduction in emissions is the main difference between two products which are otherwise very similar (very close substitutes for each other).  This is largely valid for commodities such as electricity, although even here issues such as timing and reliability of generation need to be considered.  However for most consumer goods improving characteristics in use can greatly increase their value.  Making low carbon products cheaper is crucial.  But if they are also better than the higher carbon alternatives this will lead to much more willing and rapid adoption.

Electric vehicles illustrate how non-price attributes can provide additional value to consumers and others, but can also create barriers to adoption.  Electric vehicles have a number of characteristics, which, at least in my experience, make them preferable to their internal combustion engine equivalents.  They are quiet and pleasant to drive, as the Nissan Leaf, the world’s best-selling electric car to date, and the more recent BMW i3 both demonstrate.  Refuelling by simply plugging in overnight is convenient, and there is no need to visit petrol stations, which are not generally pleasant places to be despite the best efforts of oil companies to make them more appealing.  Low centres of gravity lead to good road holding, and electric motors are instantly responsive, making for smooth and often rapid acceleration.  Performance has been one of the main selling points of the Tesla S, and responsiveness is one reason electric motors are finding their way into hybrid drive trains even on high performance cars such as my local car factory’s premiere product, the astonishingly quick and vastly expensive McLaren P1.

While car markets are highly competitive the variation in price for similar cars shows that consumers are often prepared to pay a premium for a car with improved characteristics.  For example, variants of the Volkswagen Golf hatchback range in price from £17,000 to £26,000.  Electric vehicles may similarly be able to realise premiums that reflect their benefits, with the Tesla S already the bestselling car in a third of the richest US postal codes.

Wider benefits may also play a role in adoption of low carbon technologies.  Local air quality is improved by the absence of emissions of particulates and other local pollutants.  This has led some cities to encourage electric vehicles, with, for example, plans for all new London taxis to be zero emission by 2018.  Such non-GHG benefits are produced jointly with greenhouse gas emissions reduction and will in some cases dominate the case for change.

There are also non-price barriers to the use of EVs that can also affect uptake, most notably availability of recharging points to enable longer journeys.  To ease these difficulties governments and the industry are expanding charging networks.  However range limitations remain, along with price, the biggest obstacle to uptake for most electric vehicles.   The Tesla S largely overcomes the range problem with its 300 mile range, but at over £60,000 excluding the government incentive it is not a cheap vehicle.  Plug-in hybrids largely avoid the range problem by retaining an internal combustion engine or on-board generator, but with some compromises of their own.

Similarly, there is much that manufacturers can do to make other low carbon products more appealing.  The chart below shows the spectrum from different types of lighting.  The quality of the light is very different in each case, with the light from compact fluorescents (CFLs) clearly much less continuous than from other sources.  Whatever else, these are clearly not exact substitutes.  It was perhaps premature for the EU to regulate incandescent electric light bulbs out of the market when many people found the light from the substitutes less appealing, and, while many may  prefer them to CFLs, there may be much manufacturers can still do to improve the quality of light from LEDs, alongside continuing reductions in costs.

Source: (1)

To take one more example of non-price characteristics from among many, there is surely room for improvement in the aesthetics of rooftop solar panels, at least in some contexts, and a number of innovators are working on this.

Fortunately, gauging and meeting consumer preferences is something markets do rather well, at least when consumers know what they want and can tell what has been delivered.  So markets have an important role to play in decarbonisation.  But it will be the behaviour of markets for low carbon products as well as markets carbon such as the EUETS that will be crucial to successful decarbonisation.

Decarbonising an economy is difficult and complex.  It can be made easier if new technologies not only have lower carbon dioxide emission than the alternatives, but are also better in other respects.  Policy can help promote this by stimulating innovation, enabling early adoption and removing barriers.  If the future not only has a safer, more stable climate, but is also brighter, cleaner, better looking, and more fun to drive around it will be a lot easier to persuade people that it’s a future in which they wish to invest.

Adam Whitmore – 28th February 2014


Challenges in applying a marginal abatement cost framework – Electric vehicles as an example

MAC analysis is further limited by difficulties of application in practice.  Several factors complicate estimates of the cost of abatement, and some of these are illustrated here by reference to electric vehicles.  These factors can be, and sometimes are, taken into account in careful analysis of abatement costs.  However they are difficult to treat properly, because of the scope of the modelling frameworks and the amount of information they require make them very demanding to assess.

First, the quantity of emissions avoided, and thus cost of abatement, is very dependent on the emissions intensity of the source of electricity.  For example, Norway, currently the world leader in the deployment of EVs, has a mainly hydro based grid, leading to relatively large emissions reductions.  However emissions from electricity generation will be greater in countries with fossil based systems, which will lead to a lower reduction in emissions and higher abatement costs, other things being equal.  However just how much lower may depend on factors such as when EVs are charged, and what the marginal generating plant on the system is at that time. 

Furthermore, lifecycle emissions of the vehicle itself can vary greatly between EV models and even between the same model made with materials from different sources.  For example, the emissions from smelting aluminium for a lightweight body can be very different depending the source of the electricity used in smelting, and emissions will be different again in making a carbon fibre body such as that used for the BMW i3 and i8.  An additional complication is that many lifecycle emissions can fall outside the jurisdiction being assessed, and may be covered by a quite different set of policies.   

Costs can also change greatly over time, sometimes to an unanticipated extent.  Batteries account for a large proportion of the cost of an EV, but costs are falling rapidly.  In the last five years costs have more than halved and energy densities, which set the size of battery pack, have more than doubled.  This trend seems likely to continue as a result of continuing R&D and increasing deployment.  

Technology spill-over benefits from early deployment are difficult to account for in a MAC analysis.  They are among the reasons EVs currently attract financial incentives in many jurisdictions, for example a £5000 grant in the UK, $7500 Federal Tax Credit in the USA, and exemption from VAT and purchase tax in Norway.  Other incentives can also play a role in stimulating early adoption, including exempting EVs from tolls or congestion charges, allowing EVs on High Occupancy or bus lanes, providing free parking, and mandating tight emission standards.


On battery costs:  US DoE report published as part of their EV programme shows costs for batteries declining to around $300-350/kWh.  For a comparison changing the specification of a Tesla S from a 60kWh battery to an 85kWh battery (the two models are otherwise quite similar) increases the price by £6170 excluding VAT, which is $400/kWh (see here).    

Data in the main body of the post on the sales of the Tesla S in prosperous postcodes is from

The EU’s recent proposal for a 2030 EUETS target does not look very ambitious

The EU’s recently announced greenhouse gas emissions target for 2030 looks like just enough to keep the 2050 target credible, but seems unlikely to be perceived as highly ambitious by other jurisdictions.  

The European Commission has recently proposed a target of reducing EU greenhouse gas emissions to 40% below 1990 levels by 2030.  Sectors covered by the EUETS (power generation and large industry) will be required to reduce emissions to 42% below 1990 levels.  This post takes a look, using some rough-and-ready analysis, at how onerous the EUETS target would be if implemented.   The Commission also announced a proposal to establish a “market stability reserve” for the EUETS.  I will return to this proposal in a future post, but for now the analysis excludes its effect.  The analysis also excludes the temporary delay of allowances sales over the next few years (backloading), which does not affect cumulative totals to 2030 in the absence of the stability reserve.

A target of a 40% reduction by 2030 is on a straight line track from the 20% mandated by 2020 towards the least stringent end of the 2050 target, which is an 80-95% reduction from 1990 levels.  This appears to be the minimum reduction likely to retain the credibility of the 2050 target, especially given the current surplus of allowances in the EUETS.  A smaller reduction by 2030, requiring deeper cuts to be achieved more rapidly towards 2050, would likely have increased the perceived probability that the 2050 targets would not be adhered to.

There is currently a surplus of EU allowances of around 2.2 billion tonnes, equivalent to about one full year of emissions covered by the scheme.   This scale of surplus has arisen mainly due to the severity of the recession in Europe.  Emissions currently remain below the cap, and even as the cap tightens it will take more than a decade for the surplus to disappear.

This is illustrated in the chart below.  The cumulative cap on emissions between now and 2030 (green line) starts at level of the current surplus.  It then increases, but less rapidly each year as the annual cap comes down.  This is compared with the illustrative case of annual emissions are constant at 2012 levels (solid blue line), so cumulative emissions grow linearly.  In this case, with no reduction in annual emissions, the surplus disappears in around 2026.  However in practice power sector emissions are expected to fall over the period (see below), reducing cumulative emissions (dashed blue line).  This leads to the surplus disappearing only in 2029, and reduces the cumulative shortfall by 2030 to quite low levels, assuming emissions from industry are constant.  Aviation is excluded from these totals.  Although internal flights remain covered by the EUETS, the associated cap remains unclear.

Cumulative emissions (excluding aviation) and the cumulative cap (including current surplus) show a deficit emerging only in the late 2020s …

 Cumulative surplus

The power sector is the largest source of emissions covered by the EUETS, so is crucial to demand for allowances.  There may be some increase in electricity demand over the period, and hence in demand for allowances.  The increase may be smaller with strong efficiency measures, or larger if there is very rapid uptake of electric vehicles, and will also vary more generally with GDP growth over the period.  There is also likely to be a decrease in nuclear generation to 2030 as older plant comes to the end of its working life and is not replaced by an equal amount of new plant.

However the growth in demand and fall in nuclear output seem likely to be more than offset by continuing growth in generation from renewables.  This implies a net decrease in the need for fossil generation, leading to lower emissions in the absence of changes to the fossil fuel mix.  However there may be some increase in emissions from internal EU  aviation, although any increase is likely to be much smaller in absolute terms than the decrease from the power sector.  Trends in emissions from industry, assumed to stay constant here, will also affect the total.

Together these trends might lead to a cumulative excess of expected emissions over the cap of around a billion tonnes by 2030 (about 3% of the total), including some growth in emissions from domestic aviation.   Projections of emissions over more than a decade and a half are obviously uncertain, and the cumulative total could easily vary by a billion tonnes or more from this total.  Nevertheless, it seems likely that the shortfall in allowances cumulatively over the period will be somewhere in the low- to mid- single figures percent of the total over the period, with the market remaining in surplus until the late 2020s.   The additional abatement required to eliminate the shortfall in this case could be achieved by a moderate amount of fuel switching.  And scenarios where a surplus of allowances persists through to 2030 are not hard to construct.

Any substantial scarcity that does emerge seems likely to be as a result of banking of allowances into the period after 2030, either as a result of private sector banking or the operation of the market stability reserve, which effectively mandates a certain amount of banking of any large surplus.

The EU’s apparent intention to (just about) keep on a track towards its 2050 targets is surely welcome.  However the proposed 2030 target for the EUETS thus does not seem very demanding.  It seems unlikely that such a cap will be to be taken by other countries as a sign of strong EU leadership on emissions reduction.  It also seems unlikely that the EUETS alone will become effective at stimulating large scale investment in low carbon technologies over the next decade and a half.  This risks endangering progress to reduce emissions after 2030.  Additional policy instruments will likely be needed if the EU is to succeed in building the low carbon infrastructure needed to put itself on a path to largely decarbonising its economy by the middle of the century.

Adam Whitmore  -  14th February 2014

Notes on data and assumptions  

The 40% target requires a 20% point reduction by 2030 from the already mandated 20% cut due by 2020.  If this were followed by 40% points (40% down to 80%) over the subsequent two decades an 80% cut would be achieved by 2050.   20% of 1990 levels per decade thus takes the cap towards the top end of the 2050 target range of an 80-95% cut by 2050.

2012 emissions include the industrial emissions additionally covered in Phase 3.  Emissions from large industry are assumed to remain constant over the period.  The linear reduction factor is assumed to increase from 1.74% p.a. to 2.2% p.a. in 2021.

The estimates of power sector trends are based on the IEA 2013 World Energy Outlook New Policies Scenario.  This scenario shows demand growth in EU power generation of 0.4% p.a. over the period, leading to an additional 260TWh of generation by 2030 compared with 2011.  It also shows a decline of 10% in nuclear, but his may include optimistic assumptions about new build.  A decrease in nuclear generation of 20% (180 TWh p.a.) seems plausible, and I’ve used this estimate.  This leads to potential additional demand from fossil generation of 440TWh (260TWh + 180TWh).  The IEA estimates that generation from renewables, including hydro, will approximately double between 2011 and 2030, increasing by 730TWh p.a..  This leads to a net reduction in demand for fossil generation of around 290TWh (730TWh – 440TWh) by 2030.  The estimate of the saving takes account of the profile of these trends, for example the more rapid fall-off in nuclear in the 2020s.  Additional TWh of low carbon power are assumed to reduce emissions by 0.4t/MWh, equivalent to displacing mainly gas. 

The electricity sector projections take their base year as 2011 while the emissions data base year is 2012, but this is taken account of in the calculations. 

Internal aviation emissions are currently around 84 mtpa, but the position of aviation within the EU post-2020 is currently unclear.  The calculations assume that international aviation is dealt with under a separate agreement through ICAO, or not at all. 

The calculations exclude any additional reductions if other jurisdictions take action.  Any reductions in the cap due to international action may in any case be accompanied by increased use of offsets within the EU.

The continuing spread of carbon pricing

Carbon pricing continues to spread around the world, with major schemes in Chinese provinces now in place.

For my first post this year it seems timely to review progress on implementing carbon pricing around the world.  As I’ve previously noted, the spread of carbon pricing during the past decade has been remarkable.  Once confined to a few small economies in northern Europe, it has become a worldwide phenomenon, with more than a dozen major carbon pricing schemes either in place or under development around the world.  The major step forward in last year has been the start of five regional pricing schemes in China, although trading in these markets remains relatively illiquid.  The expansion of carbon pricing in China is set to continue this year as two more trial schemes go live.  A year from now, assuming current programmes run to schedule, carbon pricing will be in place in jurisdictions that together account for between a fifth and a quarter of total global CO2 emissions from energy and industrial processes. 

Not all emissions in these jurisdictions are priced, as governments use other policy instruments to reduce emissions in particular sectors, for example surface transport in the EU.   Nevertheless, by next year over 10% of the world’s energy and industry CO2 emissions are likely to be priced.

On the Chart below the top (blue) line shows how the percentage of emissions in jurisdictions with pricing has grown over the last decade.  The total includes all energy and industry CO2 emissions taking place in each jurisdiction with carbon pricing.   Thus, if all jurisdictions in the world had carbon pricing in place the total coverage would be shown as 100%.  The lower (green) line shows the percentage of energy and CO2 emissions that are actually priced.  For example, the EU accounts for around 11% of emissions, but only a little under half of these are priced by the EUETS.  The gap between the blue and the green lines is the proportion of emissions covered by other policies, or by no policy.  Even if carbon pricing were to be extended to 100% of jurisdictions it is likely that some emissions would remain unpriced. 

CO2 emissions from land use and emissions of other greenhouse gases are excluded from the calculations.  Including these would reduce the proportion of emissions in jurisdictions with pricing, in part because of a large volume of emissions from deforestation in countries without national carbon pricing, notably in Brazil and Indonesia.  Nevertheless the trend is remarkable, and implies that any country considering carbon pricing is very much part of the worldwide policy mainstream.

Coverage of carbon pricing is increasing …

Coverage chart January 2014

By far the most significant new development over the next few years is likely to be the extension of carbon pricing across all of China, which the Government has indicated it wishes to see in the next few years.  This alone would raise global coverage of carbon pricing to over 40%.  Indonesia is also looking at carbon pricing with a prospective voluntary market potentially leading to a compulsory market in due course. 

We may also see somewhat more widespread carbon pricing in the USA.  EPA regulation of existing power plants under the Clean Air Act will oblige states to put in place implementation plans.  This may lead states to establish emissions trading schemes, or (more likely) join the Regional Greenhouse Gas Initiative (RGGI), which covers the power sector only.  Indeed some states are understood to have already expressed an interest in doing so, although it is not yet clear in which states interest is strongest.  Expansion of Western Climate Initiative trading schemes beyond California and Quebec also remains possible.

There also appears to be a trend towards carbon pricing in Latin America.  Mexico may strengthen its currently very limited carbon tax (excluded from the chart) over time.  Provinces in Brazil have looked at emissions trading schemes, and discussions on an ETS are now underway in Chile.

Future trends are, however, far from clear, and the commitment by the Australian government to repeal its carbon pricing legislation is an indication that consistent progress is far from guaranteed.

For those involved with carbon pricing day-to-day it is often easy to forget just how recent it is, and just how much progress has been made in a short time. There is still only a single decade of experience, compared with many decades, and in some cases centuries, for other types of regulation.  As a regulatory “technology” large scale carbon pricing remains more recent than the ipod, and there is still much to learn and a long way to go.  But the achievement to date is both substantial and encouraging.

Adam Whitmore – 16th January 2014

Notes on inclusion and exclusion from the chart:    The small carbon tax introduced in Japan in 2012 by modifying energy taxes is excluded, as is the carbon tax in Mexico, which is small and has limited scope.   The Tokyo emissions trading scheme is excluded as its current status is unclear.  The Swiss scheme is included in the total for the EU.  The status of the Kazakhstan scheme is currently uncertain and I have allowed for a year’s delay to 2015.  For simplicity the California and Quebec schemes are shown with full coverage from their introduction in 2013, although they do not reach this in practice until next year.  Question marks indicate measures which have yet to be enacted. 

Comparison with World Bank study:  The results here are similar to those from a World Bank Study of carbon pricing from 2013, but on a slightly different basis.  The World Bank study quotes just over 10Gt out of 50Gt of emissions taking place in jurisdictions with pricing.  This is for all GHGs.  The total shown here is around 8Gt out of 34Gt (2012 data) for carbon dioxide emissions from energy and industry.  The main difference in the total Gt of emissions in jurisdictions with pricing appears to be due to the inclusion in the World Bank study of a number of jurisdictions where pricing is at an earlier stage than shown here, notably Turkey, Ukraine and Brazil.  However South Africa is excluded from the World Bank total, as is British Columbia.  The total emissions actually priced is quoted by the World Bank as 3.3 Gt, which is 10% of carbon dioxide from energy and industry and 7% of total GHGs, but this excludes some of the China pilots for which no data was available.  I have included estimates in these cases.  If the other pilot Chinese schemes were included  in the World Bank totals their estimates of coverage would likely increase by about a percentage point (to 11% of energy and industry and 8% of total GHGs), roughly in line with the totals quoted here.  The World Bank study can be found at


Data: Emissions data is for 2012, from the EDGAR database, with no adjustment for changes in relative volumes over time.  Shares at subnational level are estimated based on a range of data.  Data sources include  and Zhao et. al., China’s CO2 emissions estimated from the bottom up: Recent trends, spatial distributions, and quantification of uncertainties  Atmospheric Environment, Volume 59.

Early reductions in carbon dioxide emissions remain imperative

New analysis using simple and transparent assumptions shows clearly the importance of early reductions in CO2 emissions.

Many studies have looked at the rates of emissions reduction necessary to meet certain targets, most commonly looking at limiting temperature rise to 2°C.  Some recently published analysis takes a complementary perspective.  It looks at the implications for expected temperature rise of delaying the date at which CO2 emissions peak, assuming that the rate of decline from peak is limited by technical and economic constraints.

But robust simplicity of the model used for the analysis, which can be reproduced on a simple spreadsheet, draws out the implications of climate science very clearly.  (There is nothing new in the science itself.)  The model has the following input variables that together determine cumulative emissions:

  • The current level of emissions.  This is around 10.5GtCp.a. (which is about 39Gt CO2 p.a. – the model works in tonnes of carbon so there is a conversion factor of 3.7 to put the data in the more familiar tonnes CO2).
  • The rate of growth of emissions before the peak is reached.  This is set at 1.8% p.a., based on the growth rate over the last two decades or so.
  • The date at which emissions peak.  Examining the effect of changing this variable is the main objective of the analysis.
  • The rate of decline in emissions achievable after the peak.  Rates from 0.7%p.a. to 2.4%p.a. are considered, assumed to continue indefinitely.

The model assumes an instantaneous switch from growth to decline of emissions at the specified rates.  A smoother peak appears more likely, but the precise shape of the peak makes little difference to the results.

A more material assumption is that emissions growth before the peak continues at the historic rate of 1.8% p.a., which determines the level of peak emissions at any given date.  This rate is based on an extrapolation that includes rapid growth in the 2000s, much of which has been due to growing coal use in China.  Global CO2 emissions growth may well not continue at the same rate in future.  The IEA’s Current Policies case shows CO2 emissions growth from energy use of 1.4% p.a. over the period 2011-2030.  The New Policies Case shows growth of only 0.8% p.a..  BP forecasts a 1.2% p.a. growth over the same period, again with some policy action.  However the lowest of these growth rates does include substantial policy efforts to reduce emissions, and by using historic rates of emissions growth the model appears intended to capture the consequences of continuing low levels of action.

The final variable in the model is the amount that temperature rises in response to cumulative emissions, as measured by the transient climate response to cumulative emissions (TCRE).  Climate models indicate that this response is approximately linear over the relevant range.  Values of 1.5°C per trillion tonnes of carbon (TtC) and 2.0°C per trillion tonnes of carbon are examined, both well within current estimates of the expected range.

The output from the model is peak temperature rise expected as a result of the chosen inputs, although it is recognised that no single parameter can completely capture changes to the climate.

The results for different parameter values are shown in the charts below.   The diamonds show historic emissions, which are extrapolated to a peak year.  The area under the curve is cumulative emissions.  The coloured regions indicate the warming that can be expected for each date at which emissions peak, with the transient climate response (TCRE) and rate of decline in emissions after peak as labelled on each chart.

Charts from Stocker and Allen retouched

 Source:  Allen and Stocker (2013)

The implications of the modelling are striking.  First, as many studies have found, it appears that the internationally agreed target of limiting warming to 2°C is very challenging.  It only appears possible either if global CO2 emissions peak this decade, which seems unlikely, or climate sensitivity turns out to be at the lower of the two values examined and emissions can be reduced at 2.4% p.a. This is an ambitious rate of decline to sustain globally over decades.  It requires (broadly speaking) new and replacement infrastructure to be almost entirely with low carbon and a rapid reduction in emissions intensity of energy, when there has been almost no reduction globally since 1990.

Second, with 1.8% emissions growth rates before the peak, warming a 4°C or more looks likely if emissions can only be reduced by 0.7% p.a. after the peak, even with climate response of 1.5°C.

Third with emissions rising at their historic rate each decade of delay in peaking emissions increases eventual warming by around 0.4°C.

Fourth, slower emissions growth before the peak, and thus a lower peak level of emissions, can significantly reduce maximum temperatures. For example with a 0.8% p.a. growth rather than 1.8% growth to a global emissions peak in the 2030s temperature is reduced by about 0.5°C, assuming in both cases an immediate 1% p.a. post-peak decline. (This is not explicitly shown as sensitivity by the authors but can easily be derived from the model.)

These last two results make it clear that suggestions that reducing CO2 emissions being less urgent than previously thought are unfounded.  A corollary of the analysis is that action on short-lived climate pollutants doesn’t take the pressure off the need to limit CO2 emissions, although it does produce benefits over the next few decades.

The implications of this analysis are too far reaching to describe and justify fully here, but a few stand out.  They are mainly familiar, but take on renewed force in the light of this analysis.  First, the scale and required rate of change in emissions appears unlikely to be driven by a single policy instrument such as a carbon price, valuable though that may be as part of a package of measures.  A carbon price, however desirable, looks unlikely to be enough to drive the sort of systemic changes, including, for example, in grid operation, that will be necessary.  And even if carbon prices could in principle achieve much of the necessary change they look unlikely to be widespread enough and high enough in practice over the next 10 or 20 years, even though the implementation of carbon pricing continues to progress.  Other policy interventions will be needed.  Second, a large proportion of the decarbonisation will need to be accomplished with the technologies we now have either already at scale, or with the potential to grow to very large scale within (at most) a few decades.  Third, no one body or level of institution is likely to be able to drive the necessary change alone.  Action is now taking place at a range of scales – individual, local, province, national and multi-national – and this looks likely to continue to be needed.

Many are already aware of the needs for these sorts of actions.  And much is being achieved.  But much more remains to be done.  It is to be hoped that this new analysis gives renewed impetus to action.

Adam Whitmore –6th December 2013


The modelling described here is reported in: Impact of delay in reducing carbon dioxide emissions, Myles R. Allen and Thomas F. Stocker Nature Climate Change (2013) Stocker co-chaired IPCC Working Group 1 and is based in Bern.  Allen is based at Oxford University.

The model assumes a sharp peak in emissions.  Instead a smoother transition can be modelled by replacing an immediate change from current growth rates to the long term rate of decline with a gradual change in the rate of growth.  Illustratively, a case is examined in which the growth rate changes linearly from initial growth rate to the long term rate of decline over a 20 year period, assuming emissions tracks are the same before and after this period.  This shows that total emissions are reduced by less than 10GtC compared with a sharp peak, thus reducing cumulative emissions by less than 1% of the total.  The effect on results is thus small.

Using lower values for the rate of growth to peak has a larger effect on the results than assumptions about the shape of the peak.  The alternative emissions growth rates quoted in the text of 0.8-1.4%p.a. are from the IEA 2013 World Energy Outlook and the BP 2030 Outlook.  If the peak is in 2040 cumulative emission are reduced by 15%.  In both cases 540GtC of cumulative emissions to date are assumed.  Indicative analysis shows that this can reduce maximum temperature rise by of the order of 0.3-0.7°C, with values outside that range possible depending on parameters chosen, assuming an immediate switch to decline.  

A 2.4% p.a. decline in emissions sustained globally over decades looks likely to be difficult to achieve.  It is roughly equivalent to retiring existing infrastructure after an average lifetime of 40 years, and only building new infrastructure that produces no emissions, both for replacement and to meet new demand – although the decline modelled is proportional rather than linear as implied by this analogy.  Over the last quarter century there has been almost no reduction in the carbon intensity of energy use, and global energy intensity has only decreased by a little over 1% p.a., with continuing growth in energy use, which is projected to continue.  It may be possible to reduce some CO2 emissions, notably from deforestation, rather more quickly than those from energy, which is one good reason for targeting those sources of emissions. 

The approximately linear relationship between cumulative CO2 emissions and resultant peak warming is expressed as the transient climate response to cumulative carbon emissions, or TCRE.  TCRE is formally defined as the warming due to cumulative carbon dioxide emissions per trillion tonnes of carbon (TtC) released into the atmosphere (1 TtC is slightly less than double the emissions so far from fossil-fuel use and land-use change since 1750, which are estimated to total around 540GtC by the end of 2010). TCRE is closely related to the more familiar transient climate response (TCR), which is defined as the warming at the time of doubling of CO2 after it has increased at 1% per year for 70 years. TCR more generally indicates the warming due to any gradual increase in radiative forcing over a 50- to 100- year timescale.  An approximate rule-of thumb is that the TCRE is about 90% of the TCR.  Source: Allen and Stocker.  (See also the IPCC Fifth Assessment Report Working Group 1 for further information on TCR and TCRE).

The UK needs to take a more serious look at importing renewable electricity

Imported solar electricity looks likely to be cheaper than nuclear in the UK by the early 2020s when new nuclear is due to come on line.  Solar and other imported renewables deserve a closer look as one means to decarbonising the UK power sector.

The recently agreed Contract for Difference (CfD) for new nuclear power at Hinkley Point in the UK fixes the price of electricity at £92.5/MWh ($149/MWh) for 35 years, indexed to inflation.  This is expensive compared with the average market electricity price in 2012 of £45/MWh ($73/MWh) (see the end of post for notes on data and calculations).  It is also expensive abatement, at around £136/tCO2 ($217/tCO2) against a gas plant at current market electricity prices.  This compares with the UK carbon price planned for 2020 (including the UK’s own carbon price support levy) of £30/tCO2 ($47/tCO2).

The price for new nuclear compares more favourably with the prices for alternative sources of low carbon power in the UK at the moment.  It is cheaper than the current going rates for offshore wind, solar and onshore wind in the UK (£155/MWh, 125/MWh and 100/MWh respectively).  It is also lower than the cost of Carbon Capture and Storage (CCS), which is estimated at around £160/MWh at the moment (with an abatement cost of about £400-600/tCO2 avoided after taking account of the residual emission from power plants with CCS).

However, solar electricity in Germany and Italy can be produced for around £90-95/MWh, close to the price of electricity from Hinkley, based on the Feed In Tariffs in the first half of this year for ground-mounted systems.  (The Feed In Tariff in Germany has since been reduced to £83/MWh but it is not yet clear how much capacity will be built at this price).

A number of adjustments need to be made to obtain a like-for-like comparison between the prices solar and nuclear.  (The calculations of these adjustments are indicative, but precise numbers are anyway impossible to derive given the long timescales and uncertainties involved.)

First, the cost of back-up capacity and system balancing increases the cost of solar compared with nuclear.  There is some increase in system balancing costs.  Together these are estimated to add £13/MWh to the cost of solar.  Any offsetting adjustment for the system capacity effects of nuclear is excluded from this number (see notes).

Second, nuclear and solar have very different patterns of output with correspondingly different values.   Solar runs only in the daytime, with output higher in summer.  Nuclear output is largely constant, typically running at full output almost all of the time.  With present wholesale market price patterns in the UK the additional value of daytime power for solar outweighs the additional value for nuclear from supplying winter evenings when prices are highest, leading to a premium for the value of the solar production pattern on average over the year of 6/MWh.  (This includes crediting the full value of winter peak hours to nuclear, which may to some extent double count the value of capacity if the above capacity premium is also included in the comparison.)  However, price patterns may change in future for many reasons, for example because large amounts of solar generation may reduce daytime prices.  Summer daytime prices in Germany already show some weakening due to the amount of solar on the system, although they remain generally above overnight prices.

Finally the length of the contracts also differs.  The solar Feed In Tariff in Germany runs for 20 years whereas the Hinkley contract is for 35 years, inflation indexed in both cases, making the Hinkley contract more valuable.  The additional value is estimated as £7/MWh or more (this is the additional price that might be needed if the Hinkley contract were to run for only 20 years).

The adjustments thus offset each other (with, coincidentally, no net effect) leaving the costs of UK nuclear appearing very similar to the present costs of solar in Italy and Germany.  In addition of course transmission costs to the UK would need to be paid for imports of solar from continental Europe.

But it is more appropriate to compare the price of the nuclear contract with the price that solar – which can be installed very quickly – will require when the nuclear plant comes on line a decade from now (assuming it is completed on schedule).  Solar costs look almost certain to fall over this period.  Prices for solar in Germany are roughly a quarter of what they were a decade ago and are expected to fall significantly again over the next decade.  Imports to the UK from sunnier regions in southern Europe, such as Italy or (especially) southern Spain could be cheaper still.  It is impossible to be certain how far costs will fall, but a mid-range estimate (based on long term historical trends) is that prices of perhaps as low as £50-60/MWh or less could be attained for imported solar delivered to the UK by the time Hinkley comes on line, even allowing for the costs of transmission.  Solar would then have a clear cost advantage over nuclear.   Even new solar in the UK appears likely to be available at below the cost of power from Hinkley by then.

Solar alone will not be a complete solution to decarbonising the UK power sector.  But there are also other options for importing renewable power to the UK, including onshore wind from Ireland and Scandinavian hydro, which can provide considerable flexibility.  While they may lack some of the employment and industrial policy attractions of building capacity within the UK, imports can allow access to a scale of renewable electricity supply that the UK, with its poor solar resource and (at least in England) exceptionally high population density, may otherwise struggle to reach.

There are also other advantages to importing renewables.  A diversity of locations, with varying demand and production patterns linked by adequate transmission capacity, can help balance the system, and ensure that power is available at lowest cost.

Policy has an important role to play in making such imports of power viable.  Policy can enable and incentivise the construction of transmission capacity, more flexible system design and operation, and appropriate commercial arrangements.

The Hinkley nuclear deal signals that the UK Government is serious about meeting the need for new low carbon power.  The CfD mechanism provides a new way to finance new nuclear power, and it makes the costs of the project transparent.  And the UK does appear to need new nuclear plant if it is going to meet its ambitious decarbonisation objectives.

But options for importing renewable power do not as yet appear to have received the same degree of policy focus as new nuclear.  Imports of renewables may have a significant role to play in diversifying supply and limiting the total costs of providing low carbon power.  They deserve much more serious and ambitious attention than they have yet received.

Adam Whitmore  - 13th November 2013

Notes on calculations 

Price of the Hinkley C announcement is taken from the DECC announcement The quoted price excludes the reduction of £3/MWh if another similar plant is built.  Other UK prices for low carbon power are draft prices for CfDs.

No adjustment is made for any of the other features of the contract.  In particular no adjustment is made for the value of the loan guarantees for which the project is reported to qualify.

The price for CCS is taken from the report of the UK’s CCS cost reduction task force.  There is an aspiration to reduce costs of CCS to around £100/MWh by the 2020s, but it is not clear whether such a large reduction can be achieved given the slow progress in developing CCS in the power sector globally.  There are similar aspirations to reduce the cost of offshore wind.

Calculations of abatement cost are based on a comparison with emissions from a CCGT of 0.35 tonnes/MWh, using the current market price of electricity, with emissions from plant with CCS at 0.07 – 0.15 tonnes/MWh depending on fuel type and capture rates.

The cost of solar in Germany used is the 2013 feed-in tariff for ground mounted systems less than 10MW of 11.02€c/kWh in April 2013.  The feed-in tariff in Italy was at a similar level. 

The estimated cost of back-up capacity is based on a study from the solar parity project.  The calculation ignores the cost of system capacity associated with nuclear that arises because the proportion of total capacity it accounts for is lower than its proportion of energy it accounts for.  This leads to an additional capacity cost associated with nuclear which to some extent offsets the additional capacity cost of solar, so the capacity cost of solar should in reality be estimated as the difference between the capacity costs of nuclear and solar rather than the full value quoted here, which would decrease the relative cost of solar. 

The price premium for solar is based on 2012 price patterns.  The wholesale electricity spot price in Great Britain is weighted by solar output in Germany over 12 representative days (one per month), allowing for the one hour time difference between Germany and the UK to get the output pattern for notional imports (data was Italy or Spain would yield slightly different results, but the general pattern would be the same).  This is compared with the time-weighted average over the year, which is assumed to be the price received by nuclear power. 

Peak prices during winter evenings contain an element of payment for capacity.  This is added to the value of nuclear relative to solar in this calculation.  However an element of capacity value has already been included as a separate item in the adjustment (see above).  As noted in the main text, there is potentially an element of double counting here, with the value of capacity for nuclear being available at the peak included both in this and the capacity adjustment itself.  This calculation combined with the capacity cost above may thus understate the present premium value of the solar output pattern, and so the assumption is somewhat favourable for nuclear.

Adjustment for contract duration is based on comparison of cash flows from a 20 year and 35 year contract.  The calculation assumes the 2012 baseload market prices prevail in real terms in years 21-35 of the contract (approximately 2044 – 2058), but adjusted to allow for a carbon price of £30/tCO2, based on the government’s stated price target for 2020.  There is a stated goal of increasing the carbon price beyond this level, but the achievability of such a high carbon price remains unclear, and is not regarded as a base case for the purposes of this calculation.  The calculation assumes 8% real terms discount rate (real, pre-tax).  A further adjustment could be made to take account of the lower risk of the revenue stream from the fixed price contract, which would further increase the value of the longer term nuclear contract. 

Another study by the Solar Parity Project estimates a likely reduction in the cost of solar by 2023 of around 40%.  Costs of solar power from southern Europe in 2013 can readily be estimated as anywhere between around £35-65/MWh or depending on rates of deployment, technology learning rates, how much of present low cost of panels is due to cyclically low prices, whether installation costs can match German levels, and what load factor is achievable.  Corresponding costs for UK solar seem unlikely to be greater than around £80/MWh, although the UK government’s own estimate for 2020 is much higher, at around £100/MWh, for reasons which are unclear (see solar roadmap document

If solar gets cheap enough it is possible that the UK’s poor solar resource would be less of a problem considering cost only – the extra output in southern Europe might be offset by the extra transmission costs at very low cost levels.  However this seems a distant prospect, and anyway land availability and the high population density in England would continue to favour imports.

Exchange rates used are $1.61/£ and €1.19/£.