Working Paper
Biofuels, Deforestation, and the GTAP Model
Published: December 2024
As the global population and incomes grow, increasing demand for food creates economic pressure to expand agricultural land, which replaces forests and woody savannas with cropland or pasture and releases large quantities of carbon that significantly contribute to climate change. Around the world, governments are also promoting the large-scale use of biofuels made from crops that also require land. A credible empirical literature describes a clear path from domestic biofuel production to tropical deforestation that likely causes overall increases in emissions. First, it is well accepted that biofuel policies raise domestic food prices; indeed, they are intended to do so. Second, domestic price increases for basic agricultural commodities propagate worldwide via the import/export activities of global commodity traders. Third, tropical deforestation responds to economic incentives such as increased crop prices. Fourth, the average carbon emissions from clearing land to produce the crops that generate biofuels exceed the resulting savings from reduced gasoline or diesel use by 300-400% for at least thirty years.
These empirically established causal paths from domestic biofuel policy to deforestation in the tropics strongly caution against turning food crops into fuel as a climate policy. However, when setting biofuel policy, governments frequently ignore this evidence and instead use a particular class of complex economic land use models. These models purport to predict in a highly disaggregated way how biofuel demand alters the demand and supply of different crops and land uses in many regions throughout the world. Variants of the popular GTAP model are used for this purpose by the State of California, for international aviation agreements, and by the U.S. Treasury for tax credits at least in 2024. Due to a range of market-mediated responses, including reduced food consumption, GTAP claims that diverting crops to biofuels results in little expansion of cropland and even less loss of forests or other carbon-rich land. As a result, governments assume that switching to biofuels reduces greenhouse gas emissions.
Getting biofuel policy wrong is not a minor issue because current global biofuel policies are already a major driver of growth of demand for agricultural products, and emerging policies have the capacity to transform the earth. Between 2004 and 2024, global oilseeds used for vegetable oil expanded by roughly 250 million acres (an area equivalent to roughly three quarters of U.S. harvested cropland), and during this period, biomass-based diesel fuel contributed roughly 40% of the increased vegetable oil demand. Announced projects would more than triple capacity for renewable diesel (Malins and Sandford 2022). Spurred in part by government policies and international agreements, most of the aviation industry has pledged to decarbonize by 2050, and under some policies and international agreements, airlines can use biofuels to claim reduced emissions (Graver et al. 2022). Biofuels from vegetable oils are among the cheapest claimed “sustainable aviation fuels (Graver et al. 2022). If vegetable oils supply even one-quarter of likely aviation fuel in 2050, their production will require 40% of global cropland at today’s oilseed mix and yields (Authors’ calculations).
Should governments rely on GTAP-style models to conclude that biofuels will not lead to vast deforestation and benefit the climate? We have been able to analyze and run GTAP code directly and can provide new, direct, and independent evidence of how the model works to produce low ILUC numbers. The GTAP model may be an interesting theoretical research tool for some purposes, but we conclude that it lacks scientific credibility as a policy tool, particularly when applied to biofuels and land use. As summarized on the following page, the model produces physically impossible results by a wide margin. The parameters and equations that dictate its economic effects -- the justification for the model’s use -- lack empirical basis. The model also contains multiple pure assumptions that contradict empirical evidence and that independently and even more collectively guarantee a low ILUC.
We argue that governments should replace reliance on GTAP-style models with an alternative, more transparent focus on the opportunity cost of land used for biofuels. This is consistent both with how governments typically model the carbon costs of other kinds of economic activities, from using gasoline to making cars, and with how economists evaluate costs in general: What potential carbon mitigation is lost by using land for biofuels rather than in alternative ways? If not used for biofuels, land can be used to store carbon (in the form of forests). It can also be used to produce food, which not only contributes directly to human welfare but also permits forests and other carbon-rich habitats to remain in their uses and continue to store carbon. Any alternative approach should be consistent with the lack of positive empirical evidence that biofuels will reduce emissions and with the strong empirical evidence linking biofuel use to tropical deforestation.
In this paper, we discuss each of these issues in greater depth. We provide some background information on biofuels, land use, and the existing empirical literature before turning to a more detailed discussion of GTAP modeling.
These empirically established causal paths from domestic biofuel policy to deforestation in the tropics strongly caution against turning food crops into fuel as a climate policy. However, when setting biofuel policy, governments frequently ignore this evidence and instead use a particular class of complex economic land use models. These models purport to predict in a highly disaggregated way how biofuel demand alters the demand and supply of different crops and land uses in many regions throughout the world. Variants of the popular GTAP model are used for this purpose by the State of California, for international aviation agreements, and by the U.S. Treasury for tax credits at least in 2024. Due to a range of market-mediated responses, including reduced food consumption, GTAP claims that diverting crops to biofuels results in little expansion of cropland and even less loss of forests or other carbon-rich land. As a result, governments assume that switching to biofuels reduces greenhouse gas emissions.
Getting biofuel policy wrong is not a minor issue because current global biofuel policies are already a major driver of growth of demand for agricultural products, and emerging policies have the capacity to transform the earth. Between 2004 and 2024, global oilseeds used for vegetable oil expanded by roughly 250 million acres (an area equivalent to roughly three quarters of U.S. harvested cropland), and during this period, biomass-based diesel fuel contributed roughly 40% of the increased vegetable oil demand. Announced projects would more than triple capacity for renewable diesel (Malins and Sandford 2022). Spurred in part by government policies and international agreements, most of the aviation industry has pledged to decarbonize by 2050, and under some policies and international agreements, airlines can use biofuels to claim reduced emissions (Graver et al. 2022). Biofuels from vegetable oils are among the cheapest claimed “sustainable aviation fuels (Graver et al. 2022). If vegetable oils supply even one-quarter of likely aviation fuel in 2050, their production will require 40% of global cropland at today’s oilseed mix and yields (Authors’ calculations).
Should governments rely on GTAP-style models to conclude that biofuels will not lead to vast deforestation and benefit the climate? We have been able to analyze and run GTAP code directly and can provide new, direct, and independent evidence of how the model works to produce low ILUC numbers. The GTAP model may be an interesting theoretical research tool for some purposes, but we conclude that it lacks scientific credibility as a policy tool, particularly when applied to biofuels and land use. As summarized on the following page, the model produces physically impossible results by a wide margin. The parameters and equations that dictate its economic effects -- the justification for the model’s use -- lack empirical basis. The model also contains multiple pure assumptions that contradict empirical evidence and that independently and even more collectively guarantee a low ILUC.
We argue that governments should replace reliance on GTAP-style models with an alternative, more transparent focus on the opportunity cost of land used for biofuels. This is consistent both with how governments typically model the carbon costs of other kinds of economic activities, from using gasoline to making cars, and with how economists evaluate costs in general: What potential carbon mitigation is lost by using land for biofuels rather than in alternative ways? If not used for biofuels, land can be used to store carbon (in the form of forests). It can also be used to produce food, which not only contributes directly to human welfare but also permits forests and other carbon-rich habitats to remain in their uses and continue to store carbon. Any alternative approach should be consistent with the lack of positive empirical evidence that biofuels will reduce emissions and with the strong empirical evidence linking biofuel use to tropical deforestation.
In this paper, we discuss each of these issues in greater depth. We provide some background information on biofuels, land use, and the existing empirical literature before turning to a more detailed discussion of GTAP modeling.