Agricultural Subsidies and Global Greenhouse Gas Emissions

1. Agricultural producer support by main types of support

Figure 1 shows the total amount of annual government support to agriculture by type of support for 54 reporting countries. Support is presented for developed (high-income) and developing (low- and middle-income) countries by type: market price support (“MPS”), mainly consisting of border measures; subsidies coupled to input use or level of agricultural output (“coupled subsidies”); direct transfers to producers (“uncoupled subsidies”); and general service support expenditures (“GSSE”), which include other forms of support such as agricultural innovation systems, sanitary and phytosanitary standards, and rural infrastructure. 

Figure 1.

Source: Laborde, D., Mamun, A., Martin, W. et al. Agricultural subsidies and global greenhouse gas emissions. Nat Commun 12, 2601 (2021). https://doi.org/10.1038/s41467-021-22703-1 data source: OECD¹

The two components of support that influence output decisions most directly are subsidies coupled to output and market price support provided through trade measures. Coupled subsidies tend to increase output without lowering demand in the subsidizing countries and hence to increase global emissions. Market price support tends to increase supply in protecting regions but, at the same time, reduces demand for agricultural products in those countries by raising consumer prices, making its impact on global emissions an empirical question to be addressed in this paper. Decoupled support is designed to have no impact on output and, hence, unless accompanied by effective environmental conditions, also no effect on emissions. GSS support, however, includes investments in research and development (R&D) that may be reasonably assumed to reduce both the cost of production and emissions per unit of output.

2. Coupled subsidies vs market price support, 2017–19

A rough indicator of the relative magnitude of coupled price support and market price support is provided by dividing the value of producer support by the value of output at world prices, as shown in Table 1. A key feature of Table 1 is the extraordinary rate of border support in a few high-income countries, such as Japan (57%) and Norway (63%). Farm support rates in China are not as high, but nonetheless substantial as market price support and coupled subsidies add almost 15% to farm output value. Also noteworthy is the negative market price support in India (−12% of farm output) combined with sizeable coupled subsidies (7%). Globally, the rate of support from coupled subsidies averaged 5.5% in 2017–2019, while market price support rates averaged 5.7%.

Table 1.

Source: Laborde, D., Mamun, A., Martin, W. et al. Agricultural subsidies and global greenhouse gas emissions. Nat Commun 12, 2601 (2021). https://doi.org/10.1038/s41467-021-22703-1 

Meanwhile, greenhouse gas (GHG) emissions from agriculture are strongly concentrated in a few commodities with beef, dairy, and rice accounting for over 80% of agricultural GHG emissions (Table 2). The production of these emission-intensive goods is often heavily supported using market price-support measures. This suggests a clear link between agricultural support and GHG emissions. However, the strength of this link requires close examination. At least four factors need to be considered before making any strong inferences: (i) the average rate of support to agriculture, (ii) differences between types of support, (iii) differences in rates of support across commodities and countries, and (iv) impacts of support on production methods and processes.

Table 2.

Source: Laborde, D., Mamun, A., Martin, W. et al. Agricultural subsidies and global greenhouse gas emissions. Nat Commun 12, 2601 (2021). https://doi.org/10.1038/s41467-021-22703-1 data source: OECD¹

The average rate of support to agriculture matters because high rates of support are likely to attract resources into agriculture, increase output and, at constant technology, increase emissions from production. The type of support matters because of its influence on overall incentives to both producers and consumers. Differences in rates of support across commodities may have important impacts on overall emissions given large differences in the emission intensity of commodities and across countries as measured by the CO₂ equivalent of greenhouse gases emitted per unit of output. As noted in a related study², output of individual agricultural commodities is likely to be more responsive to differentials in agricultural support rates than is overall agricultural output to the average rate of agricultural support. The same study also points out that, while there are many cases where high rates of support are paid on emission-intensive commodities, the average rate of support to high-emission commodities was below that for relatively low-emission-intensive commodities for almost all of the 1993–2015 period.

Support intended to influence production practices and processes, such as subsidies on fertilizers, pesticides, or improved seeds, also matter. In practice, these mostly aim to stimulate agricultural production which may induce more emissions unless improved practices are more resource efficient. Higher use of inputs such as fertilizer may be an additional source of GHG emissions though improved, climate-resilient seeds may help improve environmental outcomes. Some support programs, like the reformed Common Agricultural Policy of the EU, condition support on compliance with environment-friendly production processes and land conservation practices.

The OECD’s GSS estimate includes support for activities such as agricultural research, development, and training intended to raise agricultural productivity. Productivity gains tend to reduce the emission intensity of agricultural production, for instance, through reduced use of intermediate inputs. Some new technologies appear able to both reduce emission intensities and lower costs (see, for example, Mernit³).

This article focuses on the implications of current agricultural support policies for GHG emissions. It applies a rigorous model-based analysis of the impacts of incentives on agricultural outputs and emissions. This analysis provides an opportunity to consider all the influences outlined above—impacts on overall output, differences in incentives across countries and commodities, as well as differences in farm technologies and practices used for production. It also allows us to examine the extent and potential implications of environmental conditionalities incorporated in producer support measures. We consider not just the total emissions per unit of output, but also the source of those emissions—whether they are, for instance, from enteric fermentation by ruminants or methane emissions from rice cultivation.

1. The emission intensity of agricultural production

A key parameter for understanding the impacts of agricultural support on climate change is the emission intensity of production by region, measured by the amount (in kg) of CO₂-equivalent greenhouse gases produced per kg of output. If production of a good in an area with higher emission intensities is replaced by production from an area with lower-emission intensities, global emissions from that production of that commodity will fall for the same level of output. For a proper assessment of the quantitative impact on emissions of agricultural support measures, a general equilibrium approach of the type used in this paper is needed to account for possible shifts between products. If, for instance, a reduction in support for rice leads to an increase in beef production as resources are reallocated, global emissions may increase even if emissions from rice are reduced.

Table 3

Source: Laborde, D., Mamun, A., Martin, W. et al. Agricultural subsidies and global greenhouse gas emissions. Nat Commun 12, 2601 (2021). https://doi.org/10.1038/s41467-021-22703-1 

Emission intensities vary greatly across countries/regions and commodities (see Table 3 and the Methods section for further detail). The emission intensity for bovine meat is by far the largest for any food product, and it ranges from 12.1 kg CO₂ eq per kg of production in the United States to 108.3 kg CO₂ eq in India. There is a clear association between income levels and emission intensity, with the intensity for beef more than twice as high in the group of developing and emerging economies than in high-income, developed countries. Underlying this link is a strong relationship between productivity levels and emissions as productivity increases typically save on inputs and reduce emissions per unit of output. Tubiello⁴ points out that total emissions from agriculture have fallen steadily since the 1980s in the countries subject to emission reduction commitments under the Kyoto protocol—despite substantial increases in incomes and population. Some progress on this front has been made in both developed and developing countries. While higher in most cases in developing than in developed countries, emission intensities have fallen much more rapidly in developing countries since the early 1990s².

2. The impact of agricultural support on GHG emissions

We estimate the impact of current agricultural support measures on GHG emissions through simulations using IFPRI’s global computable general equilibrium (CGE) model, MIRAGRODEP (see Methods section), augmented with models that capture the impacts of changes in outputs and inputs on emissions. We run simulations with the MIRAGRODEP model that compare observed levels of output and emissions by country and commodity with those that would come about in the absence of the government support. We look specifically at the impacts of coupled subsidies and border restrictions (trade measures) and simulate both the impacts of each type of support and their combined impacts. Supplementary Table 1 summarizes impacts on output, while Table 4 and Fig. 2 summarize the main results for emissions. More detailed findings are available in a related working paper⁵. Throughout the analysis, we make the standard economic assumption that changes in output prices result in movements along a supply curve with the underlying production technology remaining constant. We introduce innovations to the underlying technology separately, that is, as the outcome of investments in research and development—perhaps partly price-induced—that are directed towards solving problems such as increasing farm incomes, reducing consumer prices of food and/or reducing GHG emissions.

Table 4 

Source: Laborde, D., Mamun, A., Martin, W. et al. Agricultural subsidies and global greenhouse gas emissions. Nat Commun 12, 2601 (2021). https://doi.org/10.1038/s41467-021-22703-1 

Figure 2

Source: Laborde, D., Mamun, A., Martin, W. et al. Agricultural subsidies and global greenhouse gas emissions. Nat Commun 12, 2601 (2021). https://doi.org/10.1038/s41467-021-22703-1 

Our findings show that coupled subsidies stimulate agricultural output and emissions, while agricultural trade interventions reduce emissions (as compared with a situation without these interventions). Specifically, coupled subsidies increase global farm output volume by 0.9% (see Supplementary Table 1). Primarily because of this stimulus to production, GHG emissions from agriculture are 34,420 kt of CO2 eq higher (an increase of 0.6%) than they would be without the coupled subsidies. The impact on emissions is smaller than the output effect, because the stimulus provided is less for the most emission-intensive products, such as beef and dairy products, and because the expansion of output resulting from subsidies is larger in richer countries with lower-emission intensities. Table 4 further shows that the impact of coupled subsidies on emissions is similar in magnitude for developed and developing countries. Close to a third of the increase results from stimulus to synthetic fertilizer use. The total impact of subsidies on emissions will be greater once emissions from land use change are added to these estimates.

3. The importance of efficiency improvements for reducing emissions

In sum, while many have criticized current subsidy programs as contributing to global warming, our results suggest that simply abolishing current programs could, in fact, lead to slightly higher emissions. There are, however, many reforms that could be undertaken to improve the performance of agricultural support against goals such as improving economic efficiency, reducing poverty, and lowering emissions. One approach that would serve all three of these goals simultaneously might be to increase the support to agricultural R&D, and particularly R&D that focusses on reducing emission intensities. This might be done by repurposing some of the resources currently provided as distorting subsidies to R&D that is currently counted under the GSSE element of the OECD’s measures of support. Because many studies indicate that the economic returns from R&D focused on increasing agricultural productivity are extraordinarily high⁶, and agricultural productivity growth appears to have a much bigger impact on poverty reduction than productivity growth in other sectors⁷, the required reallocation of resources might be relatively small.

Furthermore, there is every reason to believe that research focused on reducing agricultural emissions—or combinations of cost and emission reduction— would substantially reduce emission intensities. While research with a strong focus on emission reductions as well as productivity increases is relatively new, there are already promising new technologies and practices that could reduce in methane emissions from rice and from cattle by up to 50% (see, for example, refs. ^3, 8 on dietary supplements for cattle and alternate wetting and drying in rice). Hurdles to adoption of some of these new technologies can be formidable⁹, but many types of improved farm management practices could provide substantial environmental benefits at low cost. Because there has been relatively little emphasis in research programs on reducing GHG emissions, it seems likely that the portfolio of lower-emission innovations could be expanded quite rapidly if given greater priority. Innovations that reduce emissions from the largest single source of GHG emissions—enteric fermentation by ruminants—would seem particularly likely to result in both emission reduction and increases in productivity since these emissions involve an obvious waste of a potentially valuable hydrocarbons.

As illustrative examples, we considered two polar cases: (i) R&D that reduces emission intensities by 30% in the countries for which we have data on agricultural incentives while reducing costs only by enough to permit its adoption. and (ii) innovations that reduce the need for all inputs, thus reducing emission intensities and increasing productivity. The first is perhaps like the case of dietary supplements for cattle, at least some of which reduce emissions without greatly stimulating output per unit of input. The second is more consistent with innovations like alternative wetting and drying in rice, which raises productivity of all inputs, while reducing emissions per unit. The 30% reduction is chosen because it seems to be within the range that is suggested as feasible by the currently available—but not yet widely adopted—innovations discussed above.

Figure 3

Source: Laborde, D., Mamun, A., Martin, W. et al. Agricultural subsidies and global greenhouse gas emissions. Nat Commun 12, 2601 (2021). https://doi.org/10.1038/s41467-021-22703-1 

The results of these simulations are presented in Fig. 3. The reduction of almost 20% in global emissions from technical change that reduces emissions without raising productivity simply reflects our partial coverage of support measures. Because we account for production that accounts for roughly two-thirds of emissions, reductions in emission intensities of 30% on that production without any impact on productivity, result in a fall in global emissions of 20%.

The second innovation assumes a reduction is achieved in both emission-intensity and production costs by 30%. This would yield a reduction in global GHG emissions from agriculture of just under 10%. The fall by half in the emission reductions reflects the rebound effect associated with innovations that reduce the costs of producing goods. Higher productivity reduces consumer prices and increases consumer demand. The increase in demand offsets some of the reduction in input use and in emissions induced directly by the technological innovation in this scenario. This estimate of the rebound effect from agricultural productivity is consistent with that obtained in another study using a quite different model to analyze the impacts of productivity growth in crops and livestock on greenhouse gas emissions¹⁰. Because price elasticities of demand for food are low, the rebound effect is small enough that a sizeable net reduction in emissions remains following the introduction of the innovation. This is in contrast with technical change in many other goods and services, such as the original case considered by Jevons¹¹, where improved steam engines lowered the cost of power from coal and increased demand enough to outweigh the reduction in the amount of coal needed for any given task.

The net reduction in emissions identified in this analysis should be considered a lower bound because it omits the gains from reductions in emissions from land-use change. With price inelastic demand for food, agricultural productivity improvements reduce the land footprint needed to meet food demand, adding a further reduction in emissions that we plan to evaluate in future work. Even with this important degree of underestimation of the net gains, both these experiments point to the potential for much larger reductions in emissions from productivity growth and, hence, from redirecting support measures to R&D and incentives to adopt climate-smart practices.

The analysis presented in this paper examines the implications of current levels of agricultural support on global GHG emissions from agricultural production. To assess these impacts, we compared the current level of emissions with a counterfactual without these support measures. In this assessment, we focused on emissions from agricultural production only. This allowed us to concentrate on the complexities associated with changing these subsidies and to provide a basis for understanding more comprehensive and far-reaching reforms. For the assessment we created a new database mapping GHG emissions by source, location, commodity, production stage and technology and incorporated this information into IFPRI’s global model to relate agricultural production structures and market behavior to emission intensities by location, production sector, technology, and source of emissions.

Our findings show that current subsidies paid by governments that stimulate production induce both higher global agricultural output (0.9%) and emissions (0.6%). The existing market price support to farmers provided by trade barriers has almost no effect on global agricultural production and reduces GHG emissions by ~2% compared to a situation without such agricultural market protection. Combined, the coupled subsidies and border measures slightly increase global farm output (by 1.1%), while reducing global GHG emissions from agriculture by around 1.7%. These small net impacts arise because border measures in rich countries lower global demand more than they increase supply and induce shifts from relatively high emission-intensity producers to lower-emission-intensity producers in the rich countries. Coupled subsidies, by contrast, provide incentives to expand emission-intensive agricultural activities without any offsetting impact on demand. The upshot is that, on balance, current agricultural subsidies and trade protection appear to have a very modest impact on global emissions. This suggests that substantially reducing the vast current GHG emissions from agriculture will require an overhaul of current incentive structures, shifting support to interventions that more directly target emission reduction, such as GHG taxes on output or consumer demand, or more funding for R&D in productivity-increasing and emission-savings technologies and subsidizing the cost of their adoption.

These findings are preliminary and further research is needed to understand the true impacts, especially since the present scenario analysis did not consider in the impacts on land use change or on the carbon sequestration capacity of forests and soils. Furthermore, the findings should not be taken as conclusions about the effectiveness (or lack thereof) of current agricultural support policies. Present agricultural support policies in most countries are largely based on political-economy considerations and rarely for their impacts on GHG emissions. Proper assessment of policy effectiveness requires assigning policies to the goal that they are to pursue most directly. Future work will also involve accounting for impacts of changes in support on land use change and the carbon sequestration capacity of forests and soils, as well as additional scenarios for repurposing subsidies in ways that are more sensitive to climate mitigation and adaptation of agricultural sectors. Our tentative conclusion is that simply abolishing current agricultural subsidies and market price support would at best have a very limited impact on emissions—and could even increase them slightly. This points to a need to investigate the use of multiple instruments focused on the multiple goals of policy makers—such as economic efficiency, emission reduction, food security, climate-resilience of production, and poverty reduction—if we are to successfully achieve these multiple goals.