Applied Economics and Policy Working Paper Series AEP WP NO. 2022-05 Ship-owner Response to Carbon Taxes: Industry and Environmental Implications Pierre Cariou Ronald A. Halim Bradley J. Rickard Abstract We consider the effects of a maritime bunker levy on ship-owner profits, trade, and emissions. Standard and augmented gravity models are employed using data from 2016 to estimate the impact of a change in transit time and transit cost on grain and soybean trade flows and on vessel speed. Results for a bunker levy of 50 USD/tonne of fuel, or less, stress that it will not trigger a change in the optimal speed of the vessel which is contrary to most theoretical models that predict an increase in fuel costs will always lead to a reduction in speed and carbon emissions. For markets where the shipowners pass the tax on to final consumers, it is also optimal to keep the same speed (and transit time) as long as the tax is equal to, or less than, 100 USD/tonne. Bunker levies exceeding 100 USD/tonne may be needed to reduce carbon when trade flows are sensitive to trade costs and transport time, as may be the case for many agricultural commodities. Charles H. Dyson School of School of Hotel Administration Samuel Curtis Johnson Applied Economics and Graduate School of Management Management Ship-owner response to carbon taxes: Industry and environmental implications Pierre Cariou Professor, Centre of Excellence in Supply Chain (CESIT) KEDGE Business School Talence, France 33000 E-mail: pierre.cariou@kedgebs.com Ronald A. Halim Principal Transport Economist Equitable Maritime Consulting The Hague, Netherlands E-mail: ronald@equitablemaritimeconsulting.nl Bradley J. Rickard Associate Professor of Food and Agricultural Economics Dyson School of Applied Economics and Management Cornell University, Ithaca, NY 14853 Tel: 607.255.7417 E-mail: bjr83@cornell.edu Abstract: We consider the effects of a maritime bunker levy on ship-owner profits, trade, and emissions. Standard and augmented gravity models are employed using data from 2016 to estimate the impact of a change in transit time and transit cost on grain and soybean trade flows and on vessel speed. Results for a bunker levy of 50 USD/tonne of fuel, or less, stress that it will not trigger a change in the optimal speed of the vessel which is contrary to most theoretical models that predict an increase in fuel costs will always lead to a reduction in speed and carbon emissions. For markets where the shipowners pass the tax on to final consumers, it is also optimal to keep the same speed (and transit time) as long as the tax is equal to, or less than, 100 USD/tonne. Bunker levies exceeding 100 USD/tonne may be needed to reduce carbon when trade flows are sensitive to trade costs and transport time, as may be the case for many agricultural commodities. Keywords: agricultural trade, bunker levy, carbon tax, environmental policy, maritime economics, gravity model JEL Classifications: D22, F18, H23, Q17, R41 Acknowledgments: This research has been supported by the Social Sciences and Humanities Research Council of Canada (SSHRC) project (N 895- 2017-1003): “Green Shipping: Governance and Innovation for a Sustainable Maritime Supply Chain”. May 2022 1. Introduction The ITF/OECD (2019) acknowledges that emissions linked to international transport represent one of the most challenging environmental problems. This has been a significant on-going concern by policy makers, economists, and industry stakeholders (van Veen-Groot and Nijkamp 1999; Cadarso et al. 2010; Zhu et al. 2018) and has been driven, in part, by expanding global supply chains (Nabernegg et al. 2019). Recent reports indicate that international movements of products lead to 2,600 million tons (Mt) of carbon emissions (CO2), and that approximately 53% are due to ground transportation, 7% is attributed to air, 2.5% attributed to inland waterways, and 37.5% is linked to sea transportation (ITF/OECD 2019). International maritime transport emissions are expected to increase by up to 50% by 2050; this is an increase to approximately 1,500 Mt in 2050 compared to the 2018 level of around 1,000 Mt (ITF/OECD 2018; ITF/OECD 2021). The share of shipping in global anthropogenic emissions is also projected to rise from 3% in 2018 to 5% by 2050 (IMO 2020). Due to the challenge posed by shipping emissions, mitigation measures are therefore needed to achieve the goals agreed upon by the IMO in 2018, which aims to cut greenhouse gas (GHG) emissions from shipping by 50% by 2050 compared to the 2008 level. An economically efficient mitigation option is to set carbon prices that reflect the social cost of transportation (Dominioni et al. 2018; Heine et al. 2017; Heine and Gäde 2018). A number of global or regional Market Based Mechanisms (MBMs) have been proposed (Psaraftis et al 2019; Faber et al. 2009; Kosmas and Acciaro 2017; Cariou et al. 2021; Mundaca et al. 2021) and are recommending the use of taxes (e.g., bunker levies) or emission trading schemes to incentivize the use of low and zero-carbon fuel alternatives. However, there continues to be much discussion on the appropriate level for such bunker levies (Psaraftis et al. 2019). MBMs are controversial as they may lead to inequities across shipowners and across countries, and therefore it has proven difficult to reach a 1 multilateral agreement for this measure. Empirical research has established that, in general, perceived fairness is an important factor for public support of carbon pricing (Sommer et al. 2022). While the vast majority of economists support the adoption of a carbon tax as an efficient mechanism to reduce greenhouse gas emissions, such a policy may also excessively burden the poor (Fremstad and Paul 2019). As stressed by Bureau et al. (2017), “International transport plays a major role in trade globalization ... and emerging countries fear an increase in the cost of international transport that may impede their development and advanced countries do not want to make progress alone in the fight against global warming. In this context, incorporating climate challenges into the regulation of international transport need to be accompanied by transparent economic impact assessment”. These impacts of MBMs on trade and economic outcomes for industry stakeholders are a consequence of two main effects (DNV 2021; UNCTAD 2021). First, there is a direct impact due to higher compliance costs. Second, there is an indirect impact as new regulations trigger a reaction from the industry, by shipowners, on the optimal speed of vessels and therefore, on transit times. The report by the UNCTAD (2021) sums up these two effects into a total logistical cost to estimate the impacts of IMO short-term measures for three scenarios (compliance to new technical requirements, combined with a High or Low level of ambition) for 184 countries and 11 product categories. One of the main findings from the DNV (2021) and UNCTAD (2021) studies concerns the sensitivity of trade to maritime transport cost and transit time, and how such sensitivities will affect trade patterns. For instance, countries that are the most affected could be those which have distant trade partners (with longer transit times) and for commodities with a high share of transport costs relative to the total product value (i.e., high CIF/FOB ratios). Overall, we recognize that maritime regulations will have consequences for both compliance costs and travel times, and our framework will model these considerations 2 explicitly in discussing the importance of demand elasticity (cost and time) on shipowner behavior and on emissions (Section 2). We then provide an empirical application for two agricultural export products and offer a way to measure these elasticities using an augmented gravity model (Section 3). Here we take advantage of the new maritime transport cost database developed by the World Bank and UNCTAD1. Section 4 incorporates our estimates to assess ship-owner profits and their potential reaction to the new market conditions, and finally, the impacts on trade and emissions. Section 5 offers our conclusions and avenues for future research. 2. Impact of a maritime carbon tax on profits, speed, trade and emissions In a simple setting, carbon emissions are positively related to the speed of a vessel and this relationship is shown in the left panel in Figure 1. Any increase in fuel price due to an environmental tax (e.g., bunker levy) will lead to a reduction in the speed and therefore, to a decrease in emissions. Insert Figure 1 around here The extent of the reduction depends on the shape of the speed-profit function (illustrated in the right panel in Figure 1). When assuming that demand is inelastic to transit time (shown by the solid and the dashed dark curves in the right panel of Figure 1), a bunker levy encourages the ship-owner to reduce speed (and associated emissions) from A to B. If we assume that demand is sensitive to transit time then another speed-profit relationship exists (shown by the solid and the dashed red curves in the right panel of Figure 1); here the optimal speed is higher than when demand is inelastic to transit time. In this latter case, the maximum profits correspond to the point where transported quantities reach the full capacity of a vessel (shown as point C in Figure 1, at approximatively 15 knots). Although the consequence of a bunker levy leads to a reduction in emissions (from point C to D), the case outlined in Figure 1 3 illustrates that the initial level of emissions (A versus C) and the extent of the reduction (from A to B versus from C to D) can be significantly different. The speed-profit relationship can also differ if transported quantities are sensitive to transport costs, and when shipowners decide to shift the burden of the tax to their customers instead of paying the bunker levy. Under this situation (not reported in Figure 1), this may increase ship-owner profits as they do not pay the bunker levy, but this may reduce the total volume of trade and therefore, their revenue. These examples illustrate how the level of the bunker levy and the sensitivity of trade to transportation cost and time are critical to evaluate the reaction from the industry and the impact on the environment. However, in most settings, the volume of demand transported by a vessel is usually assumed to be constant, or it is inelastic to speed. The idea that trade may be sensitive to time and cost was discussed in Halim et al. (2019). They showed that a bunker levy in the range of 10 to 50 USD/tonne of CO2 will affect trade, and that a carbon tax applied to all transport modes might even stimulate a shift toward maritime transport from all other modes. Mundaca and al. (2021) show that a global tax of 40 USD/tonne of carbon will reduce CO2 emissions by 7.65% for the heaviest traded products (at the 6-digit HS level of aggregation) transported by sea and that the greatest CO2 emission reductions are for products with relatively low value-to-weight ratios. Beghin and Sweizer (2021) highlight that agricultural products will be largely affected by future maritime regulations as moving these goods between markets is costly relative to the farm gate value (Hummels 2007). If the impact of MBMs will be particularly important in the short term for countries that import a significant proportion of their food supply, long-term improvements in fuel efficiency of ships may reduce this effect and as revenues increase we will see a decrease in transportation costs over time (Stocheniol 2011). 4 Korinek and Sourdin (2009) confirm the general importance of the cost of shipping goods in overall agricultural trade costs using an augmented gravity model. Doubling the bilateral transport costs is associated with a 42% decline in the average value of bilateral country-pair agricultural imports. A doubling of the transport costs of cereals between two given countries would lead to a 37% decrease in their trade. However, most studies focus on measuring the macroeconomic impacts of GHG reduction measures on trade volume and GDP, without taking into account the potential responses of the shipowners that, as illustrated in Figure 1, can play an important role on the final outcome. In this research, we aim to address the abovementioned gap and provide three contributions in applied economic research. First, we incorporate the perspective of the carriers or shipowners and assess their likely response to the introduction of a bunker levy. Halim et al. (2019) and Vivid Economics (2010) suggest that the response depends on the ability of shipowners to transfer added costs to consumers, which is related to the elasticity of demand to maritime cost. Rojon et al. (2021) suggest that this response depends on the magnitude of the compliance costs associated with the bunker levies. We focus on how the level of tax and the elasticity of demand to cost and transit time affect ship-owner behavior, profits, trade patterns, and emissions. Second, given the potential importance of trade elasticity to cost and time, we propose a methodology that adopts an augmented gravity model to estimate these elasticities for two specific agricultural exports: grain2 and soybeans. We complement a former study by Korinek and Sourdin (2009) which is, to our knowledge, the only study with a focus on the impact of maritime trade cost on agricultural trade. Third, we answer the call by Beghin et Sweizer (2021) for more applied research on agricultural trade to better understand the impact of new environmental policies applied to maritime transportation. By incorporating estimates from the gravity model into a framework that focuses on ship-owner gross profits for grain and 5 soybeans, we shed new light on the potential reaction of shipowners to environmental regulations and on their impact on agricultural trade and emissions. 3. Empirical Model 3.1. Ship-owner response to maritime regulations To understand how a bunker levy affects ship-owner decisions regarding vessel speed, we first present the main determinants of their profits. The ship-owner’s daily gross profit or the Time Charter Equivalent (TCE) can be presented, in its simplest form3, as follows (Evans and Marlow 1990). !"# = !.#! − &" − '. )"(+) (1) $"#.%% With )"(+) = -. +& where !"# is the daily Time Charter Equivalent (in USD/day), . is the freight rate (in USD/tonne of cargo), / the transported quantity (in tonnes of cargo), 0 the distance travelled (in nautical miles), + the speed (in knots, i.e. nautical miles/hour), OC are the operating costs, ' the price of fuel (in USD per tonne) and )" the fuel consumption (in tonnes per day). The fuel consumption-speed relationship or FC(s) is critical to understand the reaction of shipowners to a change in fuel price (p). The traditional assumption (IMO 2020; Adland et al., 2020) is to use a speed-to-fuel consumption elasticity equal to 3, which reflects the engine power-speed relationship, so that )"(+) = -. +&, with 1 = 3. When assuming that the transported quantity (W) is not affected by a change in transport cost (Cost) or in transit time (Time), so that !"'!#$%& = 0 and !"'!+,-. = 0, and for a given voyage between two countries equal to d, the speed is the main factor on which the ship-owner can play to maximize the TCE. The optimal speed s* that maximizes TCE can be determined from the first-order condition /!#+0'!- = 01, so that: & +∗ = 3().!.#4'(&&.*.+., (2) 6 The optimal speed therefore changes with the ratio of freight rate to fuel price 23'45 and of the vessel specific characteristics included in parameters - and 1. Corresponding to this optimal speed, the amount of carbon emitted per day or per trip can be estimated and is a function of the fuel consumption-speed relationship and on an emission factor (ef) specific to each type of fuel (according to the fuel’s carbon content): "&( = 5- . )"(+) (3) We now consider that W, the transported quantity for a given route is sensitive to transit , time 3!675 = ()..4 that reflects the quality of service or value of time so that !"'!#$%& ≠ 0 with //012 = 9. !6753&. We also assume that W is sensitive to transit cost (Cost) so that !"' 3!+,-. ≠ 0 with /45.6 = :. ";+< ". For the latter case, it means that if shipowners transfer the cost of the bunker levy to final customers, the demand (W) is also affected. Then, following the introduction of a bunker levy, the ship-owner has two options. First, the ship-owner can pay for the bunker levy (t), whereby the price of fuel increases from p to (p+t) and the optimal & ().!.# speed changes from s* to a new optimal speed with +∗ = 3 )*+, '(&&.(*86).+.,4 . Second, the ship- owner may decide to pass the bunker levy to final customers and therefore shift the burden of the tax through an increase in transportation costs. Under this second configuration, the optimal & ().!.# speed changes from s* to a new optimal speed with +∗ = 3 )*+,,./%04'(&&.*.+., . Despite the fact that the ship-owner may not have to pay for the tax, they may experience a larger decrease in profits if demand is highly sensitive to transit cost. It means that the level of the bunker levy, the sensitivity of demand to transit time, and the sensitivity of demand to transit costs are going to influence the ship-owner’s choice to pay or pass-through the tax. Consequently, this will also affect the level of trade (W) and the effectiveness of the policy on CO2 emissions. In the next section we describe the data and model we use to evaluate the effects of a bunker levy applied to maritime shipments of grain and soybeans. We propose to estimate the 7 transit time (=:) and transit cost (=() elasticities with an augmented gravity model that has been widely used to examine international bilateral trade. We will then employ the estimates from the gravity model in the ship-owner’s profit function and subsequently assess the potential impact of a bunker levy on the vessel speed, trade and emissions. 3.2. A Description of the Data Our analysis used data on export quantities (in tonnes) for 36 grain exporting countries and 25 soybean exporting countries to 84 importing countries in 2016 (COMTRADE). We use data from 2016 as this is the year for which information on bilateral maritime costs (Costij) are available from the World Bank-UNCTAD maritime costs database. The trade data and the cost data were merged using the HS product level code. In addition, we included traditional covariates in the gravity models such as the bilateral maritime distance (dij) from the CERDI-SeaDistance (Bertoli et al. 2016), which is based on the shortest sea route between the coastal region of the country pairs, with the relevant port being the coastal port of the country with the highest traffic. A proxy of transit time (Timeij in number of days) was calculated by dividing the maritime distance (CERDI) by the average speed of a Panamax bulk carrier, the most common vessel to transport grain and soybeans, with an average sailing speed of 12 knots in 2016 (IMO 2020). We also consider a suite of covariates that are commonly used in gravity models including the existence of a contiguous border between neighboring countries (CNTGij), common language (LANGij) and the existence of former colonial ties (CLNYij) trade patterns (all from the CEPII database). Insert Table 1 around here Table 1 reports some descriptive statistics for the top five grain and for the top five soybean exporting countries in 2016. The United States accounts for 39% of grain exports and the top five countries represent approximatively 80% of world grain exports. Overall, across the two commodities there are large differences in transport costs per kilogram (from 0.01 8 USD/kg to 0.14 USD/kg) which can be explained by the type of product exported, the distance travelled, and the size of vessel used. The share of maritime transport costs in total FOB value is between 7.9% and 10.3% for grains and the average travel time is approximatively 28 days. Exports are more concentrated for soybean exports, with the United States and Brazil accounting for more than 83% of world export volume, amounting for 44% and 39% respectively. Transport routes for soybean trade are longer as reflected by transit time of 49.5 days compared to 28 days for grain trade routes. Table 1 shows that average transport costs for soybeans are lower than for grain (0.01 USD/kg compared to 0.06 USD/kg) and the share of FOB export values is also lower (around 3.17% for soybeans compared to 9.7% for grains). These differences could be due to a number of factors including port congestion and global demand for bulk ocean services (USDA 2014; Steadman et al. 2019). Ocean freight rates from South America to Asia are for instance often less expensive than from the U.S. Gulf (O’Neil 2015, Delmy 2015) because of dry-bulk vessel route patterns, lower cost port charges, higher Panama Canal tolls, and less burdensome navigation restrictions. 3.3. A Gravity Model Considering Maritime Transit Costs and Times In addition to the standard computable general equilibrium models (UNCTAD 2020), many approaches using gravity models have also been considered to assess the impact of MBMs (Mundaca et al, 2021). We begin with a model that only includes distance between partner countries, denoted as dij, in equation (4a), and then in equation (4b) we replace dij with maritime transport costs (Costij) and maritime transit time (Timeij) as explanatory variables. !"#!" = %# + %$%!"'!" + %&()*+!" + %',-)+!" + %((,).!" + %)!".! + %*!"/" + 0!" (4a) !"#!" = %# + %$!"*123!" + %%!"(456!" + %&()*+!" + %',-)+!" + %((,).!" + %)!".! + %*!"/" + 0!" (4b) In equations (4a) and (4b), lnWij corresponds to the logarithm of nominal bilateral international trade flows (in volume) from exporter i to importer j, β0 is a constant term, CNTGij 9 is an indicator variable capturing the presence of contiguous borders between trading partners, LANGij denotes a dummy variable for the existence of a common official language between trade partners, and CLNYij is an indicator for the presence of colonial ties between countries. The covariates lnYi and lnEj are the logarithms of the values of exporter output and importer expenditure, respectively. In line with Anderson and van Wincoop (2004) and Yotov and al. (2016), we re- estimate equations (4a) and (4b) and account for multilateral resistance terms by adding a set of exporter fixed effects and importer fixed effects in equations (5a) and (5b) as follows: 78"12 = 91 + ;2 + %12!"'-.+%3+<#=12 + >3?@<=12 + >4+?3?@<=12 + >4+?578!12 + >678+<#=12 + >378?@<=12 + >4+?578#$%&12 + >778+,-.12 + >678+<#=12 + >378?@<=12 + >4+?