1. Introduction
The 21st Conference of the Parties of the United Nations Framework Convention on Climate Change (UNFCCC) in December 2015 resulted in the Paris Agreement, where parties agreed to take action to limit global temperature rise this century to well below 2 °C above pre-industrial levels [
1], often referred as the “2 °C target”. Limiting climate change below 2 °C will require substantial reductions of greenhouse gas (GHG) emissions and the transition to a climate-friendly, low carbon economy. The European Commission’s report "Global Energy and Climate Outlook, Road from Paris" [
2], provides an initial estimate of potential emission reductions by sectors in the global economy that are required to reach the 2 °C target. This estimation is done by comparing a business as usual reference scenario to a 2 °C target scenario for the world (
Figure 1). The estimation indicates that the GHG emissions reduction required by 2030 could be achieved by the power sector (contributing 39% to the total mitigation effort), followed by “other energy” sectors (19%), industry (18%), agriculture (10%), buildings (6%), transport (4%) and waste (4%). These results exclude emissions and sinks for the “Land Use, Land Use Change and Forestry” sector (LULUCF). More precisely, global GHG emissions from the agriculture sector (i.e., only accounting for the agricultural non-CO
2 emissions methane and nitrous oxide) are estimated to rise to 6.283 gigatonnes of carbon dioxide equivalents (GtCO
2eq) by 2030 in the reference scenario, whereas they decline to 4.996 Gt CO
2eq in the 2 °C scenario. This represents a 20% reduction in global agricultural sector emissions by 2030 [
3].
Other model simulations identify similar reduction targets for agricultural non-CO
2 emissions necessary to meet the objectives of the Paris Agreement. For example, the Integrated Model to Assess the Global Environment (IMAGE), the Global Change Assessment Model (GCAM) and the Model for Energy Supply Strategy Alternatives and their General Environmental Impact (MESSAGE) [
4] calculated the need of global agricultural non-CO
2 emissions mitigation in the range of 11–18% by 2030 compared to the reference emissions (a reduction of 0.92–1.37 GtCO
2eq per year). The estimations in Kitous et al. [
2] and Wollenberg et al. [
4] are only two examples showing that the agricultural sector will be impacted both directly and indirectly by a low carbon economy. On the one hand, several studies point out that the agricultural sector has to directly contribute to emission reductions if the global climate change goals are to be met. This contribution has to come through direct emission reductions but also from increased land-use-based carbon dioxide removal [
2,
5,
6,
7,
8,
9], which will have a direct impact on agricultural production [
10,
11,
12,
13]. On the other hand, the agricultural sector will also be indirectly affected, as agricultural intermediate prices respond to the new economic environment. Given these foreseeable challenges, there is a need to adjust existing modelling tools, and eventually develop new ones, capable of analysing the economic impacts of a low carbon economy on agricultural markets in detail.
A variety of agricultural economic models are already equipped and utilized for the analysis of climate change mitigation on the agricultural sector [
11,
12,
13]. However, the Aglink-Cosimo model [
14,
15], as one of the main partial equilibrium agro-economic models used to prepare medium-term agricultural market outlooks [
16,
17], is not yet prepared with all necessary features to account for agricultural emissions and respective mitigation efforts. Given that the agricultural projections produced annually by the OECD and FAO with the Aglink-Cosimo model establish the benchmark for many other agricultural economic models, it is specifically important that Aglink-Cosimo is able to transmit and measure the impact of a less carbon intensive economy on agricultural markets. Moreover, Aglink-Cosimo has important features that make it particularly suitable for analysing impacts on the agricultural sector of policies related to a movement towards a low carbon economy. For example, the model has a global coverage of the main agricultural commodities produced, consumed and traded, a detailed representation of domestic and trade-related agricultural policies, and accounts for substitution effects between agricultural commodities through explicit domestic price transmission equations [
10,
11]. Accordingly, enabling Aglink-Cosimo to transmit and measure the impact of a less carbon-intensive economy on agricultural markets is a major contribution to the future analysis of agricultural emission pathways and related impacts on agricultural market developments.
In this paper, the model adjustments necessary to enable Aglink-Cosimo to account for non-CO
2 emissions and to reflect the impacts of a low carbon economy are briefly outlined. This updated model is then used to simulate the economic impacts on agricultural markets of a global 2 °C target that is compatible with the Paris Agreement. Since Aglink-Cosimo is a partial equilibrium model, this scenario analysis requires first capturing the macro-economic impacts in a general equilibrium model framework and transmitting these changes to the Aglink-Cosimo model. In a second step, the agricultural sector’s possible contribution to reductions in GHG emissions is analysed by implementing scenarios with a global GHG emission tax compatible with a 2 °C target. In addition, marginal abatement cost curves (MACC) for the main agricultural methane and nitrous oxide emission sources are introduced to capture the potential effects of technology development for mitigation (see methodological approach in Frank et al. [
18]). This highlights the importance of technological progress for achieving a certain agro-environmental target, which is often neglected in the literature.
2. A Partial Equilibrium Modelling Framework
The Aglink-Cosimo model is a recursive-dynamic, partial equilibrium, multi-commodity market model of world agriculture. The model was developed by the OECD and FAO secretariats, with the double purpose of preparing medium-term (usually about 10 years) agricultural market outlooks [
16,
17], and to provide an economic simulation model for the assessment of policies [
19,
20,
21] and economic changes related to the agricultural sector [
22,
23]. The model endogenously calculates the development of annual supply, demand and prices for the main agricultural commodities produced, consumed and traded worldwide. The present version of the model covers 82 individual countries and regions, 93 commodities and 40 world market clearing prices. Country and regional modules are developed and maintained by the OECD and FAO Secretariats, with important input in terms of data and analysis from country experts and national administrations. In a joint publication, the OECD and FAO provide annually a global outlook for the development of agricultural markets and prices. A large amount of expert knowledge is applied at various stages of the outlook process and Aglink-Cosimo is used to facilitate the consistent integration of this information from a markets intelligence perspective. Moreover, the outlook is built on the basis of specific assumptions on the short- and medium-term development of key macro-economic indicators (such as GDP, exchange rates, population, inflation and energy prices), which seem plausible at the moment of preparing the projections, given the current global environment [
14,
15]. For this paper, the model version released by the OECD-FAO with their 2017 agricultural market outlook was used. It includes market projections up to the year 2030 and a complete land allocation system introduced for 14 countries (Australia, Canada, Switzerland, Japan, South Korea, Mexico, Norway, New Zealand, United States of America, European Union, Argentina, Brazil, China and Russian Federation), taking into account double cropping systems in China, Brazil [
24] and the United States [
16]. Taking this initial model version and its underlying database, the following elements were added to analyse the impact of a less carbon intensive economy on the agricultural sector: (1) an enhanced land allocation system, (2) diminishing food demand elasticities with growing income, (3) increasing factor productivity and long run yield elasticities, (4) a module on GHG emission accounting (i.e., estimation of emission intensities per agricultural production activity), and (5) incorporation of technological progress for emission abatement (i.e., decomposition of technological, production and structural emission reduction effects). These model improvements are briefly outlined below.
First, a complete land allocation system was imposed for all single developing countries (Algeria, Angola, Bangladesh, Burkina Faso, Cambodia, Cameroon, Chad, Chile, Colombia, Congo, Côte d’Ivoire, Democratic Republic of the Congo, Egypt, Ethiopia, Gabon, Ghana, Haiti, India, Indonesia, Iran, Iraq, Israel, Kazakhstan, Kenya, Lao People’s Democratic Republic, Lebanon, Libyan Arab Jamahiriya, Madagascar, Malawi, Malaysia, Mali, Mauritania, Morocco, Mozambique, Myanmar, Nigeria, Pakistan, Paraguay, Peru, Philippines, Rwanda, Saudi Arabia, Senegal, Somalia, South Africa, Sudan, Thailand, Tunisia, Turkey, Uganda, Ukraine, United Republic of Tanzania, Uruguay, Viet Nam, Yemen, Zambia and Zimbabwe) and developing country regions (Other Sub-Saharan Africa, Least Developed Countries (LDC), Subsaharan Africa, Other Asia Developed, Other Asia, LDC Asia, Other Oceania, LDC Oceania, Other South America and Caribbean, Other Middle East, Other Western Europe, and Other Eastern Europe) specified in the model, where initially “pasture land” and “other crop land” (i.e. aggregate of the land used by all other crops not specifically included in Aglink-Cosimo) were exogenous. Incorporating a full land allocation system in the Aglink-Cosimo model is especially important in the context of emissions related to land use and land use change (LULUC). For this purpose, a full matrix of supply elasticities for crop land was estimated, specifically including pasture and other crop land, which, for example, allowed accounting for ruminant production returns on land allocation. Even though CO2 emissions related to land use changes are not yet considered in the emission accounting of the model, capturing changes in total land use gives an indication of the effects that policy changes can have on GHG emission developments.
Second, adjustments to the income, own food and cross food demand elasticities were made in the model for developing countries. In particular, these elasticities were transformed to become variables (as opposed to constants), allowing them to decrease in value as wealth increases over time. This adjustment, thus, enables the ability for developing countries to close income gaps with developed economies (i.e., allows developing countries to move along the Engel curve).
Third, another important issue to consider in medium- to long-term analysis is factor productivity, which is expected to increase over time. Therefore, a long-term crop yield response to movements in agricultural commodity prices and input costs, as well as to the share of labour. was also introduced into the model. This adjustment included: (i) the estimation of long-term elasticities responding to historical long-term crop prices and cost signals [
25], (ii) changes in the share of labour in the total cost index, following changes in real GDP per capita [
16], and (iii) a new input demand system, reflecting that the move to a low carbon economy will likely affect the prices of fertilisers, chemicals and energy, which in turn could lead to changes in the input mix.
Fourth, modelling the contribution of the agricultural sector to GHG emission reduction targets involves calculating GHG emissions per agricultural production activity (i.e., emission intensities) and allowing the model to react when GHG emission mitigation policies, such as carbon taxes, are imposed. Therefore, the model was improved to account for the agricultural non-CO
2 emissions methane and nitrous oxide, which the UNFCCC attributes to the sector “agriculture”, differently than CO
2 emissions and removals (LULUCF sector) and CO
2 emissions related to energy consumption at the farm and the processing of agricultural inputs (other sectors). These were calculated following the IPCC [
26] guidelines at the tier 1 level and using FAOSTAT data for the emission factors [
27,
28,
29]. GHG emissions were then calculated in the model per country or region by multiplying the activity data (i.e., hectares of land and heads of livestock) by the calculated emission factors. In order to perform different policy scenarios, the non-CO
2 emission inventories in Aglink-Cosimo were aggregated in CO
2 equivalents.
The calculation of emission factors was based on historical emissions and production data from FAOSTAT, but in order to allow for emission efficiency improvements reflecting the dynamics of production systems, trend functions were estimated. These trend functions for emission intensities were estimated within a robust Bayesian estimation framework that combined data from FAOSTAT on production quantities and emission inventories. The approach is further outlined in Jansson et al. [
30], Pérez Domínguez et al. [
31,
32], and Van Doorslaer et al. [
33]. Regarding carbon taxes, the taxes on emissions were introduced in the individual area harvested and livestock production equations, which allowed the analysis of tax effects in terms of emission reductions and production impacts at the individual country level. The carbon tax was introduced on a “per tonne of carbon-equivalent” basis and was applied to each production activity in each region captured by the model, so that emission intensity across activities and regions was taken into account.
Finally, technological (i.e., technical and management-based) mitigation options are incorporated into the analysis in the (reduced) form of regional marginal abatement cost curves (MACC) for different agricultural non-CO
2 emissions. These MACC are estimated ex-post based on information from Lucas et al. [
34] and are depicted as the exponential function of the maximum potential degree of abatement given a certain carbon tax (i.e., the maximum emission reduction level to be reached when the cost of reducing the last tonne of emissions equals the price of the tax). With this it is possible to further disaggregate the changes in emissions and production related to different carbon taxes into: (a) production effects (i.e., reducing agricultural production), (b) structural effects (i.e., structural change in the agricultural sector due to trade or shifts in consumption preferences for agricultural commodities) and (c) technological effects (i.e., technological progress at the agricultural production level) [
18].
3. Scenario Narratives and Design
This paper assesses the impact of a low carbon economy on the agricultural sector and focus on the potential contribution of the agricultural sector to global GHG emission reduction targets by means of global carbon tax scenarios. Currently, a large share of agricultural non-CO
2 GHG emissions stem from bovine meat and dairy production. In the past, GHG intensities from these livestock production activities have been reduced due to the evolution from less to more intensive productions systems, resulting in increases in commodity output per animal that are larger than the corresponding increases in emissions per animal [
28]. Similarly, agricultural yields have evolved towards more intensive and resilient crop production systems. Taking these past trends into account, the option of retiring land from agricultural production, creating potential carbon sinks, is one possible strategy to reduce CO
2 emissions. This could be combined with changes in consumer’s preferences towards diets containing less animal protein [
35]. A way to accomplish this strategy and to enforce the contribution of the agricultural sector to GHG emission mitigation is to introduce a carbon tax per tonne of GHG emissions. This would effectively target commodities with higher GHG intensities, which typically would be ruminant meat and milk from less intensive livestock productions systems. The resulting commodity price increase would give an incentive to consumers to change their consumption habits to less emission intensive products (e.g., eating less beef).
Following this underlying narrative, three global carbon tax scenarios are tested against a business-as-usual medium-term reference situation without a carbon tax (baseline). In the carbon tax scenarios (Tax50, Tax100 and Tax150) the macroeconomic effects inherent in moving to a global low carbon economy are specifically accounted for. Moreover, the potential incorporation over time of new mitigation technologies linked to the carbon price scenarios are taken into account. In practice three separate homogenous carbon price paths for all countries, with the exception of Least Developed Countries (LDCs), are introduced, with carbon prices gradually increasing from 0 in 2020 to, respectively 50, 100 and 150 USD/t CO2eq in 2030. With this scenario setting the impact that emission mitigation policies could have on agricultural production and consumer diets can also be highlighted. Mitigation policies in LDCs are not simulated, such as to avoid potential negative effects on regional production, aggravating food insecurity.
Given that Aglink-Cosimo is a partial equilibrium model, the total impact of a low carbon economy cannot be directly evaluated. The majority of emission reductions will have to be made by other sectors of the economy [
2], and imposing a carbon tax on the global economy will induce macroeconomic effects (e.g., changes in prices for crude oil, fertilisers and pesticides, as well as changes in real GDP) that in turn will impact the agricultural sector. As macroeconomic variables are exogenous in the Aglink-Cosimo model, the macroeconomic impact of a low carbon economy has to be first captured and quantified in a Computable General Equilibrium (CGE) model and then transmitted to the agricultural economic model. For this, a set of carbon tax scenarios is simulated using the Modular Applied GeNeral Equilibrium Tool (MAGNET) model and the GTAP database version 9 with base year 2011 [
36]. MAGNET is a multi-regional, multi-sectoral, applied general equilibrium model based on neo-classical microeconomic theory [
37]. Two versions of this model were used for this paper. The first version was a standard model using the dynamic steering system to compile a GHG emissions model and associated databases. Adjustments were then made to this initial model so that the primary agricultural sector was excluded from carbon taxes in a second model version, i.e., the carbon tax was removed from equations modelling primary agricultural taxes within the model. The same baseline scenario was run on both model versions projecting the GTAP database over four time periods (2011–2017, 2017–2020, 2020–2025, and 2025–2030). Carbon tax scenarios were then imposed as counterfactual simulations in the years 2020, 2025 and 2030 in both models, where the respective nominal Aglink-Cosimo carbon taxes were deflated to real 2011 USD. The resulting percentage changes in the price of energy (i.e., aggregated price change of crude oil, gas, coal), as well as changes in the price of chemicals (i.e., proxy for mineral fertilisers and pesticides) were transmitted to the agricultural economic model. Since a carbon tax on crude oil and pesticides cannot be directly imposed in Aglink-Cosimo, the energy and pesticide price changes are taken from the first version of MAGNET (i.e., including carbon taxes on all economic sectors). Conversely, a carbon tax is directly imposed in the Aglink-Cosimo model for fertilisers by taking the price change from the second MAGNET model version (i.e., excluding carbon taxes for primary agriculture). For a more detailed description of the MAGNET model and its use for the scenario analysis presented in this paper, please see the
Supplementary Materials.
In a similar manner, the implementation of a carbon tax in Aglink-Cosimo does not capture any change in emission intensities per agricultural activity. Such a change will occur when cost-efficient mitigation technologies and management practices get adopted, as long as the carbon price exceeds their implementation costs. To capture this technological effect, the Common Agricultural Policy Regional Impact Analysis (CAPRI) model is employed, which is a partial equilibrium, large-scale economic, global multi-commodity, agricultural sector model [
38]. The CAPRI model does not have the same detailed global agricultural coverage as the Aglink-Cosimo model, but is able to calculate global marginal abatement cost curves (MACC; [
13]). Consequently, the three carbon tax scenarios are implemented in the CAPRI model to identify the mitigation potentials through increased adoption of technology by agricultural producers as carbon taxes change. The simulated emission mitigation in CAPRI was then decomposed into production, structural and mitigation technology effects, and the resulting changes in emission intensities were then transferred into the Aglink-Cosimo carbon tax scenarios to get a complete picture of the effects of a low carbon economy on the global agricultural sector. The methodological approach of the paper is illustrated in
Figure 2.
5. Conclusions
Limiting climate change to ensure global temperature increases remain 2 °C below pre-industrial levels by the end of the century requires substantial reductions of GHG emissions and the transition to a climate-friendly, low carbon economy. A transition to a lower carbon intensive economy has large implications from both regional and global perspectives. Moreover, it needs to consider not only the environmental dimension but also the economic and societal ones. Policies aiming at a decarbonized economy can have important collateral effects in terms of people’s discontentment, as, for example, recent movements in France have shown [
43,
44]. Furthermore, the increase of prices linked to discretionary climate change mitigation policies can have negative effects on poor economies, and could increase food insecurity [
12] and migration flows [
45]. These elements highlight how the necessary transition to a low carbon economy must be carefully designed. Accordingly, the implementation of GHG mitigation policies in a specific sector needs to be “fair” in the sense of not only taking into account the long-term benefits (i.e., the GHG mitigation goal and limiting climate change) but also short and medium-term costs (transition), and it should be global, such as to minimise emission leakage and effectively reduce GHG emissions [
46].
Using the Paris Agreement as a framework for limiting global temperature rises, in this paper an empirical study is performed on how policies aiming at a global lower carbon intensive economy could be transmitted into agricultural markets. For the analysis, an updated version of the Aglink-Cosimo model is employed to simulate three carbon tax scenarios, specifically accounting for the macroeconomic and technological effects inherent in moving to a global low carbon economy (captured with the MAGNET and CAPRI models). Within this scenario design, homogenous taxes on agricultural non-CO2 emissions (i.e., methane and nitrous oxide) are implemented globally, with the exception of least developed countries, and increased progressively to 50 USD per tonne of CO2eq, 100 USD/t CO2eq, and 150 USD/t CO2eq, respectively, by 2030. Simulation results show that global GHG emissions from the primary agricultural sector are reduced by between 10% and 19% in 2030 compared to the baseline.
The analysis indicates that for the net mitigation of global agricultural GHG emissions, it specifically matters where (i.e., in which country or region) production is affected by climate change mitigation policies. Larger (lower) effects are expected in countries (and commodities) with relatively high (low) emissions per production unit. The results highlight the importance of GHG emission reduction policies on agricultural markets over a medium-term time horizon, as the sector is affected by both direct (i.e., through emission abatement commitments within the agricultural sector) and indirect (i.e., through changes in prices for fossil fuel intensive goods and macroeconomic variables) mitigation policies. For instance, it is shown how emission reductions compatible with the Paris Agreement can have significant effects on agricultural production, especially when looking at the regional impacts. These results also underline the importance of taking climate-change-related policies into account when producing agricultural market outlooks. In this respect, enabling Aglink-Cosimo to account for agricultural emissions and respective mitigation efforts is an essential development, especially considering that the model is used by the OECD and FAO to produce agricultural market projections that establish the benchmark for many other agricultural economic models. However, for future research the Aglink-Cosimo model needs to be further developed, for example to include the adoption of new GHG emission abatement technologies and the contribution of structural change within farming. Moreover, the model should be modified to account for CO2 emissions and removals related to land use and land use changes, to get a broader picture of the possible contribution (and resulting impacts) of the agricultural sector to a global low carbon economy. These aspects are important for the future assessment of both the full potential of the agricultural sector to contribute to achieving the goal of the Paris Agreement, as well as the related impacts to agricultural market developments and potential effects on food security.
Our results show that the technological development induced by the carbon tax can substantially help mitigate GHG emissions, and hence the need to reduce agricultural production levels globally. Technological development is especially important in some developing countries that would be relatively more affected by global carbon taxes, as they are usually characterised by higher emission intensities (kg CO2eq/kg commodity) and are less competitive on the global agricultural commodity markets. This points towards the importance of both (i) technology change and transfer, to reduce emission intensities especially in developing countries (i.e., need to modernize agricultural production systems), and (ii) more sophisticated and differentiated policy approaches for the agricultural sector, to achieve a significant contribution towards the move to a global low carbon economy.