(i) ETS

An ETS would require operators to acquire allowances for the CO₂ emissions associated with their fuel use—total allowances, and hence emissions, would be capped with allowance trading establishing an allowance or emissions price.¹⁸ In principle (for equivalently scaled instruments), an ETS promotes similar mitigation responses as the pure carbon tax; auctioning of allowances generates the same revenue; and allowance requirements could be modified (i.e., set in reference to a baseline emissions) to mimic a revenue-limiting tax scheme discussed below.

The ETS is potentially less cost effective than a carbon tax in a dynamic sense, to the extent that short term volatility in emissions prices causes significant differences in (discounted) incremental abatement costs at different points in time, though empirical studies for more general carbon pricing schemes suggest this effect is of only moderate importance.¹⁹ Besides being volatile, prices in ETSs to date have also been depressed²⁰, partly due to their incompatibility with other mitigation instruments—for example, if a carbon intensity standard for ships is combined with an ETS this lowers the allowance price without affecting emissions which are fixed by the cap (under a carbon tax the emissions price is fixed so the standard would reduce emissions).²¹

(ii) Carbon intensity standards—design

Carbon intensity standards—as currently implemented by IMO for the technical design efficiency of new ships—are less effective than carbon taxes (for the same implicit CO₂ price). Unless accompanied with a mechanism to review the design specification of existing ships in a comparable manner, they are limited in application to newbuild ships, and therefore do not promote responses in (2), (3) and (4) above. Carbon intensity standards, moreover, do not provide an automatic mechanism for equating incremental costs of CO₂ reductions across different operators which undermines cost effectiveness. Furthermore, the environmental economics literature suggests that non-pricing mitigation instruments are generally less effective at promoting clean technology investment than pricing instruments.²²

Carbon intensity standards may have some reinforcing role if carbon pricing is constrained (e.g., by political acceptability issues) and, arguably, to overcome obstacles to technology deployment (though there is potential for overlap if industry-retained funds are used for similar purposes). Ideally standards should include flexibility provisions, like out-of-compliance fees allowing operators to fall short of the standard (if meeting it is relatively high-cost for them) and rebates for operators exceeding the standard (if meeting it is relatively low-cost for them).

(iii) Carbon intensity standards—operation and design

Generalizing the concept of a carbon intensity standard on design specification to include operation by (i) applying it to existing (rather than just new) ships and (ii) accounting for both technical and operational efficiency in the standard, would substantially increase its environmental effectiveness. Moreover, allowing operators who exceed the standard to sell credits to those falling short of the standard, would promote cost effectiveness (i.e., trading provides an alternative flexibility mechanism to the fee/rebate provision just mentioned).²³ The approach, as previously proposed by the United States²⁴, and as represented below, may do little to promote response (3) above however as, for practical purposes, it would involve significant disaggregation of vessel types (e.g., different size container ships) with different standards applied to those types, which can limit incentives for shifting to larger ship sizes (if this implies meeting a tighter standard).

A technical challenge for carbon intensity standards for operation and design is whether a carbon intensity metric can be obtained which is environmentally effective, compatible with available or collectable data, and does not unintentionally penalize ships with special operational requirements (which would distort shipping markets)—these issues can increase the administrative burden and political acceptability of such standards. More generally, as with ETSs, the standards provide less certainty over (implicit) emissions prices than a carbon tax.

(iv) Offsets

Finally, under an offset scheme (like that for international aviation) operators would be required to purchase credits for emission reduction projects outside of the maritime sector for any excess of their CO₂ emissions above a benchmark level. In theory, by establishing a uniform reward for each tonne of CO₂ reduced (i.e., the need to purchase fewer offset credits), this approach can cost-effectively promote similar, within-industry behavioral responses as under carbon taxes (albeit the revenue-neutral version—see below), as well as promoting outside-industry emissions reductions. In practice however, the supply, and therefore price, of future offsets is highly uncertain (Box 1).

(v) Summary

A key theme from the above discussion is that design details matter. A carbon tax should be considered as a preferable mitigation instrument, so long as a robust and predictable price is established and (large) revenues are either used efficiently or the tax is on the difference between emissions and a benchmark amount to limit revenues. An ETS could be a reasonable alternative, but the same issues apply, and price stability mechanisms are then needed. Carbon intensity standards can have a reinforcing role, though they should be designed flexibly given heterogeneity in compliance costs among ship operators, and applied broadly to promote more mitigation opportunities, even though this adds administrative complexity. Offsets provide an alternative form of emissions pricing, but considerable uncertainties surround their price and credibility.²⁵

(i) Administration

International maritime carbon taxes could be collected on shipping fuels at the refinery gate as an extension of fuel tax administration procedures long-established in most countries and this would involve collection from a small number of large, easily identifiable, taxpayers. Collection at the international level from ship operators, through an IMO-administered fund, appears the more relevant option however, given delegation of maritime mitigation strategy to the IMO, and to avoid difficulties in coordinating policy across national governments.²⁶ Capacity for measuring shipping fuel use by trip is being developed and operators could pay the tax on either an annual or individual route basis, with denial of port access, or ship arrest, for non-compliant operators potential enforcement mechanisms.²⁷

More precisely, a ship operator’s tax liability would be given by

where: τ^CO2 is the tax rate on CO₂ emissions; F^SHIP is the ship’s fuel use; and β^CO2 is the emissions factor for the fuel being used, which are well known (e.g., lower for LNG per unit of energy than for conventional heavy fuel oil sold as bunker fuel).²⁸

Revenue-neutral variant. Under a variant of the carbon tax that limits the amount of revenue raised, the tax liability for the operator would instead be given by:

where BENCH^SHIP is an (exogenous) benchmark level of emissions assigned to the operator, so operators pay taxes or receive subsidies depending on whether their emissions are above or below their benchmark. If the benchmark is set at the emissions that would have been generated on the operator’s routes by the average ship within a cargo classification (e.g., container ships) then overall this carbon tax variant will be revenue neutral (tax payments from operators with above-average emissions intensity would offset rebates to operators with below-average emissions intensity). In this case the benchmark could be calculated by:

where TM^SHIP is tonne-miles for the individual operator and (CO2/TM)^AV is emissions per tonne-mile for the average ship within the relevant classification. Scaling back the benchmark (for all ships) would result in a positive amount of revenue on net, and the scheme would converge to a pure revenue-raising carbon tax if the benchmark is reduced to zero.

The revenue-neutral carbon tax provides the same incentives as the pure (revenue-raising) carbon tax (for a given CO₂ price) for responses (1) and (2) above, but essentially fails to promote response (4) as there is no pass through of (large) emissions tax payments into higher shipping costs for heavy or long-distance products. Response (3) is promoted, as operators benefit from reduced tax payments or increased in rebates from, for example, shifting to larger, more efficient ships (at least within a vessel classification).

(ii) CO₂ prices

In principle, the CO₂ tax rate trajectory might be based on the following possibilities:

• The ‘social cost of carbon’ (SCC) (i.e., the discounted value of worldwide damages from the future global climate change associated with an additional tonne of CO₂ emissions), but reaching agreement across IMO member states would be challenging, not least given widely differing estimates of the SCC in the literature;²⁹

• Global emissions price trajectories consistent with the 2°C target in the Paris Agreement, but a recent review³⁰ suggests global CO₂ prices of $40-80 per tonne (in addition to any pre-existing fuel taxes) would be needed in 2020, which is highly ambitious given the current global average price of about $1 per tonne of CO2;³¹

• Estimated emissions prices countries will need to phase in by around 2030 to implement their Paris Agreement mitigation pledges, but estimates are uncertain, vary considerably across countries³², emissions targets may be adjusted in the interim, and regulatory instruments may be used in part to meet targets;

• Modelling assessments of the price trajectory consistent with emissions goals for the maritime sector (given candidate technologies and expected growth in shipping demand) but extremely high prices (perhaps over $300 per tonne) might be needed in the absence of other technology deployment policies;³³ and

• Prices in other carbon pricing schemes, the main point being that tax rates for maritime considerably higher than prices elsewhere might be challenging from a political perspective.

On pragmatic grounds, the last option might be the most practical and is used to infer an illustrative price path in the modelling below.

(iii) Addressing differentiated responsibilities

One possible, indirect solution to the CBDRRC issue might be to remit carbon tax revenues to the GCF, which would in turn be allocated for climate adaptation and mitigation projects in targeted developing countries³⁴—funding allocations might also be skewed towards countries most vulnerable to higher shipping costs, to provide more finely-tuned compensation.

Another approach might be to allocate some carbon tax revenue to compensation mechanisms for specific countries (e.g., small island developing states and LICs). There are at least a couple of alternative approaches, though neither by itself may be entirely satisfactory. For example, reimbursing target countries for taxes attributed to their maritime fuel sales would overcompensate some countries (hubs where ships frequently refuel prior to offloading cargo in other countries) while undercompensating others (like small island developing states where ships frequently offload cargo without re-tanking).³⁵ Another possibility is to base compensation on countries’ shares of global import values,³⁶ but import value is not necessarily a reliable predictor of CO₂ (e.g., light electronic equipment has a low ratio of CO₂ to import value) and some of the incidence also stems from higher costs for exporters.

Workable compensation schemes should be practical however, given the generally modest to tiny incidence (see below) of carbon taxation. As measured here, incidence is the loss of consumer surplus—the first order revenue payment plus the second order economic welfare cost, as indicated by the (combined) gray rectangles and red triangle in the figure in Box 2 (there are no losses in producer surplus given the assumption of full pass through of fuel taxes in higher prices).³⁷

(i) Model Description

The model distinguishes (to allow a comparison of segment-specific policies) the two main (and quite distinct) types of shipping, namely wet/dry bulk (e.g., oil products, steel, iron ore, coal, grain). The main behavioral responses for reducing emissions, as classified above, are also distinguished. A discrete time-period model is used, going out to 2040, though distant projections are especially speculative (e.g., due to uncertainty over the future availability and cost of low-emission technologies).

The model begins with 2016 maritime fuel use, 334 million tonnes, equivalent to 1,051 million tonnes of CO₂ emissions, with 55 percent of it allocated to bulk shipping and the rest to container shipping. Fuel use is then projected forward in a BAU scenario (with no mitigation measures beyond those implicit in recently observed fuel use) using global GDP (assumed, based on IMF forecasts, to expand 20 percent between 2017 and 2023, and grow at 2.9 percent a year thereafter) along with income elasticities³⁸ of 0.5 and 0.8 for bulk and container products respectively. Future fuel use also depends on bunker fuel prices, which are based on the crude oil price assumed to remain constant in real terms at $70 per barrel ($513 per tonne) and a (permanent) one-off price increase of $13.4 per barrel ($100 per tonne) from 2020 onwards reflecting low sulfur requirements. Higher fuel prices reduce carbon intensity through technical design, operational, and other improvements (as defined above), with each corresponding elasticity³⁹ taken to be -0.15 (the combined elasticity, -0.45, is approximately consistent with other, technology-based modelling for the maritime sector), and through changes in the demand for tonne-miles, though the latter response is modest given the small share of fuel costs in the price of landed imports. Other factors aside, carbon intensity is assumed to decline autonomously at a rate of 0.5 percent a year (e.g., due to gradual turnover of older, less fuel-efficient ships).

To the extent other fuels are used, these are implicitly expressed in terms of the bunker fuel that would yield the equivalent amount of CO₂ emissions, therefore the CO₂ emissions factor in the model is fixed (the emissions effect of shifting to cleaner fuels is implicitly included in the technical design and operational efficiency elasticities).

(ii) Policy scenarios

Six alternative mitigation policies are considered, where policies are compared for a given explicit price, or implicit ‘shadow price’, they place on CO₂ emissions—policies therefore differ in their effectiveness at reducing CO2, depending on the behavioral responses they promote. The policies include:

Pure (revenue-raising) carbon tax—this policy increases future fuel prices according to the CO₂ emissions factor for bunker fuels (3.15 tonnes of CO₂ per ton of bunker fuel). For illustration, a carbon tax starting in 2021 and rising at $7.5 per tonne of CO₂ each year (equivalent to $24 per ton of bunker fuel) to reach $75 per tonne of CO₂ ($240 per ton of fuel) by 2030 and $150 per tonne of CO₂ ($480 per ton of fuel) by 2040. In carbon pricing schemes elsewhere (see Table 2), prices were around $5-$20 per tonne of CO₂ in ETSs and $5-$30 per tonne in carbon tax regimes in 2017, however: prices are likely to rise over time; Scandinavian countries have much higher tax rates; Canada is requiring provinces to phase in a US$40 per tonne carbon price floor by 2022; and France’s carbon tax (for non-ETS emissions) is slated to rise to $100 per tonne by 2022. The carbon tax trajectory illustrated here seems broadly in line with (and perhaps at the high end of) prices that might emerge in other carbon pricing schemes in the next decade or two.

Table 2

Carbon Prices, Selected Countries and Regions, 2017

Source. WBG (2017) and previous editions of this publication, and authors calculations (for Colombia and Canada). Note.

^aSlated price for 2022 (in 2017$).

Table 2

Carbon Prices, Selected Countries and Regions, 2017

Government/region	year introduced	Price US$/ton CO2, 2017 unless noted	Coverage, % of GHGs	Government/region	year introduced	Price US$/ton CO2, 2017 unless noted	Coverage, % of GHGs
CARBON TAXES				Sweden	1991	140	42
Chile	2014	5	55	Sweden	1991	140	42
Colombia	2017	5	40	Switzerland	2008	87	33
Denmark	1992	27	45	UK	2013	24	25
Mexico	2014	1-3	46	TRADING SYSTEMS
Finland	1990	60-65	15	California	2012	15	85
France	2014	100a	35	EU	2005	6	45
Iceland	2010	12	50	Kazakhstan	2013	2	50
Ireland	2010	24	40	Korea	2015	18	68
Japan	2012	3	66	N. Zealand	2008	13	52
Norway	1991	56	50	RGGI	2009	4	21
Portugal	2015	8	25	PRICE FLOORS
South Africa	2016	10	80	Canada	2016	40a	80

Source. WBG (2017) and previous editions of this publication, and authors calculations (for Colombia and Canada). Note.

^aSlated price for 2022 (in 2017$).

View Table

Table 2

Carbon Prices, Selected Countries and Regions, 2017

Government/region	year introduced	Price US$/ton CO2, 2017 unless noted	Coverage, % of GHGs	Government/region	year introduced	Price US$/ton CO2, 2017 unless noted	Coverage, % of GHGs
CARBON TAXES				Sweden	1991	140	42
Chile	2014	5	55	Sweden	1991	140	42
Colombia	2017	5	40	Switzerland	2008	87	33
Denmark	1992	27	45	UK	2013	24	25
Mexico	2014	1-3	46	TRADING SYSTEMS
Finland	1990	60-65	15	California	2012	15	85
France	2014	100a	35	EU	2005	6	45
Iceland	2010	12	50	Kazakhstan	2013	2	50
Ireland	2010	24	40	Korea	2015	18	68
Japan	2012	3	66	N. Zealand	2008	13	52
Norway	1991	56	50	RGGI	2009	4	21
Portugal	2015	8	25	PRICE FLOORS
South Africa	2016	10	80	Canada	2016	40a	80

Source. WBG (2017) and previous editions of this publication, and authors calculations (for Colombia and Canada). Note.

^aSlated price for 2022 (in 2017$).

Revenue-neutral carbon tax—this policy causes the same fuel price increase as the pure carbon tax, though there is no first-order pass through of tax revenues in higher shipping costs.

Carbon intensity standards—three variants of this policy are considered and implemented in the model through various shadow prices. One applies (denoted CIS—DES) to the technical design efficiency (of new ships)—it imposes a shadow price that promotes technical design efficiency but does not exploit the other three mitigation responses discussed above. Second is a standard (denoted CIS—DES/OP) that also applies to operational efficiency (of new and used ships), and exploits (cost-effectively) the first two of the above behavioral responses. Third is a carbon intensity standard (denoted CIS—DES/BULK) promoting technical design efficiency improvements for bulk shipping only. In each of these policies, the shadow prices are aligned with the CO₂ prices under the pure carbon tax.

Offsets—finally, an offset scheme is modelled that has the equivalent effect on promoting carbon intensity reductions as the pure carbon tax (because for each tonne of CO₂ reduction there is less need to purchase emissions offsets at the same price as assumed under the carbon tax) though, as under the revenue-neutral carbon tax, there is no pass through into shipping costs of charges for infra-marginal emissions. Given the lack of data for parameterizing the future offset supply curve (which, as noted above, is highly speculative) two purely illustrative scenarios are considered—a ‘low cost’ scenario where the marginal cost of offset reductions is approximately the same as that for reducing the CO₂ intensity of shipping, and a ‘high-cost’ scenario where the marginal cost of offsets is three times as high. The offset supply curve determines the additional emissions reductions that occur outside of the maritime sector, though in practice there might be considerable difficulty in establishing that offset projects are additional (i.e., would not have gone ahead in the absence of the offset payment).

(iii) Caveats

One caveat is that the model is static in the sense that fuel use adjusts instantly to fuel price changes (rather than gradually as the shipping fleet turns over), though this simplification seems reasonable given that policies are likely anticipated, phased in gradually, and the model’s focus is on longer-term impacts—the elasticities in the model therefore represent long-run responses (allowing for significant turnover of the vessel fleet).

Most of the price-responsiveness of fuel use in the model reflects reductions in carbon intensity rather than in shipping volumes. One implication is that, even though revenue-neutral carbon taxes and emission offset prices do not charge for infra-marginal emissions the resulting difference in environmental effectiveness compared with a pure carbon tax is not very significant. Another implication is that it should be reasonable to omit the capital and labor costs of efficiency improvements in computing fuel use changes from mitigation policies,⁴⁰ and fiscal or market power distortions in the shipping market in computing their economic welfare effects.⁴¹

Furthermore, the model does not capture the possibility of non-linear responses that might result from sudden switching from current fuels to a clean fuel alternative—however, this possibility seems a distant prospect and, as already noted, with current technical knowledge would likely require carbon prices far above those considered below.

(i) BAU Scenario

Figure 1 shows the BAU scenario with no mitigation measures (beyond those implicit in recently observed fuel use). World GDP is 33 percent and 77 percent higher in 2030 and 2040 respectively, compared with 2020. The expansion in bunker fuel use or CO₂ emissions is far more gradual, however—14 percent by 2030 and 31 percent by 2040 (CO2 emissions are 1,172 and 1,343 million tonnes in 2030 and 2040 respectively), that is, CO₂ to GDP falls by 14 percent and 26 percent by 2030 and 2040 respectively below the 2020 level. This reduction reflects both the assumption of below-unity income elasticities and improving energy efficiency. ⁴²

View Full Size

BAU GDP, Energy Efficiency, and Fuel Trends, 2020=100

Citation: IMF Working Papers 2018, 203; 10.5089/9781484374559.001.A001

Source. See text.

• Download Figure
• Download figure as PowerPoint slide

(ii) Carbon Taxes

The (pure) carbon tax, rising at $7.5 per tonne from 2021 onwards, reduces CO₂ emissions by 14 percent below BAU levels in 2030 and 23 below BAU in 2040 (Figure 2a), which would roughly stabilize emissions at just over 1,000 million tonnes in these years—approximately the BAU level in 2020. The policy increases bunker fuel prices by about 40 percent above BAU levels in 2030 and 75 percent above BAU levels in 2040 and 96 percent of the CO₂ reductions reflect reductions in the carbon intensity per tonne-mile (i.e., behavioral responses (1)-(3) from Section 2 combined), while 4 percent reflects reductions in tonne-miles (response (4)). Half of the CO₂ reductions in 2030 comes from bulk shipping (whose emissions share is gradually declining over time the BAU) and half from container/other shipping. CO₂ reductions from the revenue-neutral carbon tax are almost as large as under the pure carbon tax as this policy has the same effect on reducing carbon intensity per tonnemile but (to an approximation) does not affect tonne-miles.

View Full Size

Impacts of Carbon Taxes

Citation: IMF Working Papers 2018, 203; 10.5089/9781484374559.001.A001

Source. See text.Note. Welfare calculations abstract from linkages with the broader fiscal system (see Box 2).

• Download Figure
• Download figure as PowerPoint slide

The two carbon taxes differ dramatically in revenue raised—the pure tax raising 0.07 percent of world GDP, or $76 billion in 2030, and 0.11 percent of GDP, or $155 billion, in 2040 (Figure 2b). By design (and given there are no pre-existing fuel taxes), the revenue-neutral carbon tax has no revenue implications. Intermediate cases, with some positive amount of revenues raised, could be obtained by adjusting the benchmark emissions accordingly.

Figure 2(c) indicates the economic welfare cost of the tax (as defined by the triangle in Box 2), which does not account for the benefits of reducing future global warming. Although the economic welfare costs of the two policies rise over time faster than the tax rate, welfare costs are still modest—roughly 0.006 percent of GDP, or $6.2 billion, in 2030 and 0.016 percent, or $23.4 billion, in 2040 for both carbon tax policies.

The sum of the revenue and welfare cost from Figures 2(b) and(c) indicate the overall burden or incidence of the carbon tax at the global average level, again underscoring the relatively small size of these impacts relative to global GDP—under the pure tax, a modest 0.075 percent of GDP in 2030, and under the revenue-neutral variant, a pretty tiny 0.005 percent of GDP.

(iii) Policy Comparisons

As indicated in Figure 3, the CIS—DES policy reduces CO₂ emissions by about 5 and 8 percent below BAU levels in 2030 and 2040 respectively, that is, it has about a third of the effectiveness of that for the (pure) carbon tax, as it only promotes one of the four behavioral responses. Limiting the policy to bulk ships only (the CIS—DES/BULK policy) further reduces environmental effectiveness, by about half, so this policy only has about 15 percent of the effectiveness of the pure carbon tax. On the other hand, the carbon intensity standard promoting technical and operational improvements across new and existing ships (CIS— DES/OP) is twice as effective as the CIS—DES policy. The offset policies have about the same impact on reducing within-sector maritime emissions as the revenue-neutral carbon tax, given they are taken to establish the same emissions price (and hence reward for reducing within-industry emissions) but (to an approximation) do not pass through a first-order tax payment into tonne-mileage prices. The policies also reduce CO₂ emissions outside of the maritime sector, thereby implying total emissions reductions that, in the high- and low-cost offset cases, are 28 and 85 percent higher than those under carbon taxation. Whether offset schemes could establish the level of prices assumed here is highly questionable however and, most likely, not all offsets would be fully additional.

View Full Size

CO₂ Reductions Under Alternative Mitigation Instruments

Citation: IMF Working Papers 2018, 203; 10.5089/9781484374559.001.A001

Source. See text.

• Download Figure
• Download figure as PowerPoint slide