Estimated attribution of the RFS Program on soybean biodiesel in the U.S. using the Bioenergy Scenario Model

Abstract

Biofuels are supported by various governmental policies in the U.S. and globally as an alternative transportation fuel for environmental, geopolitical, and economic reasons. Much debate surrounds the effectiveness of these policies as well as the overall net environmental effect of increased biofuels use. In the U.S., recent studies have shown that the Renewable Fuels Standard (RFS) Program, overall, may not have been the leading driver of the ethanol industry from 2005 to 2020, contrary to common perception. Similar scrutiny has not been applied to biodiesel. This study uses the Bioenergy Scenario Model, a well-vetted system dynamics model, to retrospectively apportion historical biodiesel production between the RFS Program and other potentially influential drivers, such as the economics of biodiesel vs. diesel, the Biodiesel Tax Credit (BTC), California’s Low Carbon Fuel Standard, and other factors. From 2002–2020 about 36% of biodiesel production can be attributed to the RFS Program, 35% to the BTC, and the rest to other factors. Thus, the overall effect of the RFS Program appears much larger on biodiesel than on corn ethanol. The finding that the same policy may have very disparate effects on different biofuels helps inform the design of future policies on biofuels.

1. Introduction

Biofuels have been developed in the United States (U.S.) and globally as one way to reduce the greenhouse gas footprint of transportation, and to wean countries off oil from geopolitically unstable regions. The Renewable Fuel Standard (RFS) Program (U.S. EPA, 2022) has been viewed as a critical policy that drove biofuel production in the U.S. However, the strength of the effect has not been rigorously assessed, and recent research has questioned this supposition. There were non-RFS federal programs, tax credits and subsidies, trade policies, and state mandates that also spurred biofuel production, but it was not until the first version of the RFS Program (RFS1 under the Energy Policy Act of 2005), and especially the second version of the RFS Program (RFS2 under the Energy Independence and Security Act of 2007), that domestic production accelerated (U.S. EPA, 2010). With concerns for environmental impacts (e.g., land use change), largely due to expanded corn and soybean planting, quantifying what proportion of past biofuel growth can be attributed to the RFS Program versus other factors is important for informing future national policies. In this paper we focus on biodiesel, because there is a lack of published literature on biodiesel as compared with corn ethanol; and, because of the statutory cap in EISA on corn ethanol and continued improvements in vehicle efficiency, much of the future growth in the industry is anticipated to be from biodiesel.

Biodiesel production in the U.S. grew from 10 million gallons in 2002 to 1.8 billion gallons in 2020 (Figure 1). How much of this biodiesel production can be attributed to the RFS Program specifically, compared to market forces and other policies, has not been rigorously assessed. This apportionment is critical to evaluate from a policy perspective. If we do not understand what the ultimate effect of a policy is, and how policies interact to influence an industry, we cannot effectively design future policy towards broader goals. Studies that combine multiple factors into a single aggregate driver make it difficult for policymakers to evaluate the effects of the policies that they enact. Much of the early research available on the effect of the RFS Program focuses on corn ethanol, which found that the RFS Program was the major factor that drove the increase in corn ethanol production and consumption in the U.S. (Carter, Rausser, & Smith, 2017). However, more recent research suggests that this assertion may not be robust to scrutiny (E. Newes et al., 2022; Taheripour, Baumes, & Tyner, 2022; U.S. EPA, In Press). There are many factors that affected corn ethanol production; primary among these are the RFS Program, the need for a substitute for methyl tertiary-butyl ether (MTBE) as an oxygenate in gasoline, basic economics of oil vs. corn as a fuel feedstock, as well as state programs such as California’s Low Carbon Fuel Standard (LCFS). Many earlier studies combined some or many of these factors with the RFS Program, thereby conflating the RFS effect with that of other drivers. When these are isolated from one another, the U.S. Environmental Protection Agency (U.S. EPA, In Press) found that the RFS Program alone was only responsible for 0–1.0 billion gallons of ethanol (up to 11% of production) annually from 2005–2012, and 0–2.1 billion gallons (up to 14% of production) annually from 2013–2018 (U.S. EPA, In Press). The central question examined in this effort is similar – given the many factors that influence biodiesel and renewable diesel, how influential was the RFS Program compared with the other drivers?

Figure 1.

Annual U.S. biodiesel production (solid orange bars) and annual change in biodiesel production (black plus signs) for 2002–2020 (EIA December 2022 Monthly Energy Review, https://afdc.energy.gov/data/10325) along with key events affecting production such as when the BTC was available retroactively (black bars; i.e., 2010, 2012, 2014–2015, 2017–2019) and when it was available prospectively (blue bars; i.e., 2005–2009, 2011, 2013, 2016, 2020). Black horizontal line at zero annual change added for reference.

There is no reason a priori to expect the same effect from the RFS Program on biodiesel and renewable diesel as compared to its effects on the ethanol industry, since the feedstocks, markets, and fuel supply chain operations are quite different. For example, biodiesel and renewable diesel in the U.S. are mostly produced from soybean oil and waste fats, oils, and greases (FOG) from animal processing and restaurant activities. Largely because of feedstock costs, biodiesel and renewable diesel are much more expensive than corn ethanol to produce per gallon (ethanol = $3.02/gallon, biodiesel = $6.83/gallon, renewable diesel = $7.61/gallon)¹. Thus, the credit created by the Renewable Identification Number (RIN) of a biofuel under the RFS Program may have more value for biodiesel and renewable diesel producers than ethanol producers. Furthermore, there is no “blend wall²“ for most biodiesel and renewable diesel compared with corn ethanol. Biodiesel production increases were not necessitated by the need to replace something else (i.e., as ethanol had in replacing MTBE). Thus, there are many differences that suggest lessons learned from studies on ethanol may not be entirely transferrable to biodiesel and renewable diesel. Here we seek to better understand the individual and combined effects from various drivers to elucidate this complex interacting system of economics and policies.

There are few studies that have analyzed the attribution question for biodiesel. Taheripour et al. (2022) used both the biofuels-focused Global Trade Analysis Project (GTAP-BIO) model, which is a Computable General Equilibrium (CGE) model, and a version of Agricultural Energy Partial Equilibrium (AEPE), which is a Partial Equilibrium (PE) model to simulate the effect of the RFS Program. They found that removing the RFS mandate for biodiesel caused a 1.2 billion gallon decrease in biodiesel consumption for 2011–2016 in modeled results. This is equal to the average production for the same period (1.24 billion gallons) and indicates a large importance of the RFS Program on biodiesel production.

EPA initiated a study with contracting firm ICF to simulate biodiesel use in 2020 in a hypothetical scenario without the RFS Program after roughly 2016³. Simulations based at the Petroleum Administration for Defense Districts (PADD)-scale indicated that gas prices, federal and state mandates, and other economic factors played a large role in determining biofuel production trends but that the RFS Program had a larger impact on biodiesel compared to ethanol production. Their estimated biodiesel production volumes spanned 690 to 1460 million gallons/year for crude oil prices ranging from $62-$67/bbl, respectively. They found that state mandates, especially California’s LCFS, largely set the low end of the production volume estimates and the high end of the estimates resulted if spot price margins between diesel and crude oil stayed at 2016 levels. The ICF study found a threshold crude oil price of $72/bbl ($1.71/gallon), where the RFS Program determined whether refineries could blend biodiesel profitably or not. Without the RFS, and especially if tax credits were not available, refineries were estimated to be unlikely to profitably blend biodiesel below these crude oil prices. Crude oil prices have been lower than this for most of the period of growth in biodiesel production. Counted monthly, crude oil price was higher than $72/bbl for 35% of the 19-year study period.

In this study, we use the Bioenergy Scenario Model (BSM), a system dynamics model modified to simulate the biodiesel industry, to retrospectively estimate the impact that the RFS Program had on the biodiesel industry. This study describes mechanisms that shaped biodiesel production and consumption patterns from 2002–2020 through a nested set of scenarios designed to incrementally test individual drivers. The main drivers tested were (1) price competition between diesel and biodiesel, (2) biodiesel tax credit (BTC), (3) RFS Program, (4) California’s Low Carbon Fuel Standard, and (5) trade policy changes. The BSM is primarily domestic-focused model, so some more recent prominent factors such as trade disputes with China are not captured. However, these factors don’t affect much of the buildout of the industry which had already occurred, and are indirectly captured in how they may affect biodiesel price, which is exogenous to the model. By systematically examining each driver in the model simulation we can weigh the relative importance of each and/or their combinations.

2. Methodology

2.1. Modeling Overview

For this analysis we developed a modified version of the Bioenergy Scenario Model (BSM) (Peterson et al., 2015) to support a biodiesel retrospective analysis to help explain the evolution of the biodiesel industry in the U.S. over the period 2002–2020. The BSM is a system dynamics model in which to explore test cases to estimate the effects of policy and economic factors on ethanol production growth. A system dynamics model is different from either Computable General Equilibrium (CGE) economic models (e.g., GTAP-BIO, Taheripour et al. (2020)), or partial equilibrium (PE) models (e.g., BEPAM, Wang & Khanna (2023)). CGE and PE models solve for equilibrium conditions, with either more of the global economy but less industry detail (CGE), or less of the global economy but more industry detail (PE). In contrast, system dynamics models focus on disequilibrium behavior. In an energy context, these models often are used to represent industry detail associated with supply chains, incorporating time delays, feedback structures, and non-linear relationships. All three approaches are useful for studying any industry, including biofuels.

The BSM is a highly vetted system dynamics model that has been used in analyzing many facets of potential biofuels deployment scenarios for research and governmental purposes (Dunn et al., 2020; Lewis et al., 2019; E. K. Newes, Bush, Peck, & Peterson, 2015; Peterson et al., 2019; Vimmerstedt, Bush, Hsu, Inman, & Peterson, 2015). The BSM has been used to examine the rapid expansion phase of the biofuel industry and to understand potential limiting factors, different investment strategies, and to optimize various aspects of the biofuel industry (Vimmerstedt et al., 2015). The BSM has been utilized in diverse exploratory studies including cellulosic biofuels (E. Newes, Inman, & Bush, 2011), heavy duty vehicle technology (Oke et al., 2023), and other biofuel uses. Recently the BSM was used to better understand and overcome technological and economical challenges associated with developing an efficient algae-to-HEFA biofuel system that can support sustainable aviation fuel production (Atnoorkar, Wiatrowski, Newes, Davis, & Peterson).

While typical analyses with the BSM have been prospective in focus, the revised model used in this study (BSM-retro) is structured such that it is possible to support retrospective analyses that focus on scenarios relating to the evolution of the biofuels industry, including starch ethanol, biodiesel, and renewable diesel. The BSM is made up of several interconnected modules (Figure 2.) that represent major aspects of the supply chain for bioenergy. These include feedstock production, logistics, and markets; development and operation of conversion facilities; downstream inventory, pricing, distribution, exports, and domestic use of fuel ethanol; vehicles; and the oil industry. Details of the model are available in previous publications (E. Newes et al., 2011; E. K. Newes et al., 2015; Peterson et al., 2019; Vimmerstedt, Bush, & Peterson, 2012; Vimmerstedt et al., 2015). Feedback processes within and across modules capture dynamics related to land use; inventory and pricing of agricultural products; industrial learning, investment, and utilization of conversion facilities; and fuel use. Major data sources used to parameterize the model are show in Table 1.

Figure 2.

Simplified diagram of relationships in the model. Rectangles represent accumulations; directed double-lined arrows represent flows which change magnitude of accumulations over time. Policy drivers are indicated in italics.

Table 1.

Major input data sources used in BSM-retro

Item	Description	Reference
Oil prices	Historical refiner cost of crude oil, 2002–2020	(U.S. Energy Information Administration, 2020)
Gasoline and diesel prices	Based on the AEO Retrospective Review using a linear regression between gasoline (or diesel) and oil prices based on historical data.	(U.S. Energy Information Administration, 2020)
Initial land allocations (including Conservation Reserve program)	2002 values from regional land allocations based on data from the National Agricultural Statistics Service (NASS)	(U.S. Department of Agriculture, 2020)
Commodity crops	Yields, production cost per acre for commodity crops (corn, soy, wheat, other grains, and cotton), 2002–2020	Economic Research Service (2023) (USDA ERS, 2023)
FOG supply and prices	In each region, 5 supply-price pairings used to develop regional supply curves.	(Badgett, Newes, & Milbrandt, 2019)
Biodiesel and renewable diesel industry	Reference data for production, 2002–2020	Alternative Fuels Data Center (U.S. Department of Energy, 2022)
RIN prices	Daily average RIN prices, 2008–2020	Oil Price Information Service (2020)

2.2. Adaptations to the BSM and scenarios examined

We sought to understand what structural and parametric changes would enable the BSM to capture the historical patterns of U.S. ethanol, biodiesel, and renewable diesel production over the period 2002-present. Additionally, we sought to understand what might happen in counterfactual cases in which policy support for biodiesel and renewable diesel had been absent. To meet these objectives, we modified the BSM to account for five major drivers of the biodiesel industry: (1) price competition between diesel and biodiesel, (2) the BTC, (3) the RFS Program, (4) California’s Low Carbon Fuel Standard, and (5) select trade policies (Table S1).

In the model, production of biodiesel and renewable diesel is determined by production capacity and by the utilization of that capacity (Figure 2). Production capacity reflects the cumulative investment in conversion facilities, which is based on the economic attractiveness of investment as reflected by a net present value metric. Utilization is determined by expected per-gallon revenue (including incentives such as tax credits), relative to cost of production. Feedstock prices represent a significant cost of production and are determined by commodity crop prices—particularly the price of soy. Because the model captures the evolution of the fuel ethanol industry as well as the biodiesel industry, dynamics around agricultural land allocation and commodity crop markets in the United States are a central feature of the BSM-retro. The scenarios were run with an increasing number of factors, with each factor added in the general chronological order that they actually occurred historically, starting with a price competition only “baseline” scenario. Thus, because of the sequencing in the real world and in the model, we were primarily interested in the estimated actual effects of various factors, as opposed to potential effects that various factors could have had under alternative histories. See the SI for more details on the BSM-retro model.

Multiple policy inputs were incorporated into the model and turned on or off depending on the test scenario. For the price competition baseline scenario, the BSM-retro is run with only the model’s foundational economics. No biofuels-related policies were enabled. See the SI for more information about the economic parameterizations.

For the BTC, a $1.00/gallon production subsidy was applied for the years 2002–2020. This captures the biodiesel tax credit (BTC), which was created in 2004 (Schnepf, 2013), and other state-specific subsidies that were accessible to the industry before 2004. It is important to note that the BTC program lapsed and was applied retroactively at multiple points during the period 2002–2020 (Figure 1). The model captures the effect of this “stop-and-go” policy by basing investment on expected values for BTC. This effect appears significant as the changes from one year to the next were heavily dependent on whether the BTC was available prospectively or not (Figure 1). When the BTC was available prospectively, growth was usually strong, when it was available retroactively, growth was usually lower. The model assumes that with each re-start of the policy, the effect of the subsequent lapses is diminished as expectations are adjusted. Note that there were also small state production subsidies for 2002–2020 that were included in the BTC-term for analytical convenience.

The model uses values for D4 RINs, based on U.S. Environmental Protection Agency data (U.S. EPA, 2020), as the estimated direct effect of the RFS Program. RIN data archived by the EPA⁴ were sampled at 6-month intervals, beginning in June 2010, and are applied to both transesterification and HEFA processes. The RIN system of ensuring compliance with renewable fuel targets consists of four categories: cellulosic biofuel (D3), biomass-based diesel (D4), advanced biofuel (D5), and total renewable fuel (D6). Each RIN type trades at a different price. The standards are also nested so that D4 RINs can be used to satisfy D6 RFS requirements, for example, if the purchaser chooses to do so. There were periods when the economics made the normally more-expensive D4 RINs the “marginal” RIN and they were used to meet higher RIN requirements. While this nested nature of the RFS Program could lead to some overlap between the biodiesel and ethanol industries, the BSM structure and manner which we ran the study effectively minimized confounding factors between the industries.

The “splash and dash” and “Argentinian dumping” factors were modeled for 2006–2008 and 2016–2018, respectively, and were the only international factors included. According to the U.S. Energy Information Administration (U.S. Department of Energy, 2021), a peak in biodiesel exports occurred in 2008 because of an unintended effect of the BTC in the European Union (EU). This technicality, referred to as “splash and dash”, allowed foreign-produced biofuels that ultimately were consumed outside the U.S. to, nevertheless, benefit from U.S. subsides. Refiners imported biodiesel to the U.S. (B100), added a “splash” of diesel (B99), and then “dashed” the blended fuel to Europe to gain a second subsidy. The technicality was closed by Congress passing the Emergency Economic Stabilization Act (P.L. 110–343) in October 2008 (Tomei & Upham, 2009). The model accounted for the “splash-and-dash” impact by applying a production subsidy that was additional to the value of the BTC during 2006–2008. From about 2016–2018 the U.S. imported large volumes of soybean biodiesel from Argentina. Argentina has always produced large volumes of soybean biodiesel at low prices. From 2016–2018 it flooded the global marketplace, and the U.S. went to the World Trade Organization to make an “anti-dumping” complaint which was ultimately supported. The model accounted for the apparent “Argentinian dumping” of biodiesel by applying a negative incentive to domestic production (Dlouhy & Parker, 2017). The effect of this, as it is modeled in BSM, was to dampen the expected unit revenue for domestic producers during the dumping period.

For the LCFS, the model incorporated the impact of California’s Low Carbon Fuel Standard (California Air Resources Board, 2023) program by modeling an additional incentive that applied to the usage of biodiesel in the LCFS region (i.e., California). Incentives available to other regions accounted for the cost of transporting fuel from other regions into the LCFS region.

This analysis used data from many sources. We used historical sources for oil prices, conventional diesel prices, biodiesel and renewable diesel production, biodiesel and renewable diesel consumption, and agricultural commodities. Initial values for land allocation and number of biodiesel facilities were also based on historical data. Table 1 shows major data sources included in the model, and the SI has more information.

2.3. Model Validation:

Validation of the BSM-retro is critical to understand the value of comparisons of historical to various test case simulations. For the study period (2005–2020), the model can simulate historical conditions and generates results that are broadly consistent with empirically observed data for a broad set of metrics (see below). Important metrics to examine include commodity crop production, commodity prices, as well as biodiesel production. The observed historical conditions include the combined effects of individual all processes that mirror actual markets, incentives, and industry practices during the study period. Thus, the scenario used in the model validation step is the scenario with all five major drivers enabled. One or more of these processes can be disabled in the model to generate counterfactual simulations for comparison to historical and simulated historical conditions, potentially providing insight into the role each driver played in the evolution of the system.

We found that the BSM simulated biodiesel production that was very close to observed (Figure S5a). The BSM slightly underestimated renewable diesel production for the earlier years (2011–2019), and overestimated production in later years (2019–2021). As published earlier, the BSM simulated ethanol production very well (Figure S5b, Newes et al. 2022 (E. Newes et al., 2022)). For biofuel prices, the modeled historical prices corresponded well with observed prices once the incentives (e.g., BTC) were included (Figure S6). BSM For crops, the BSM simulated corn, soy, and wheat production (Figure S7a–c) and acreage (Figure S7d–f) very well. Prices were less accurately simulated (Figure S7g–i). The BSM underestimated the corn price spikes in 2008 and 2011, but performed better more recently (Figure S7g). The BSM underestimated soy prices from 2004–2013 and was relatively accurate after 2013 (Figure S7h), and underestimated wheat prices for most years (Figure S7i). Since most of the increases in soybean biodiesel was after 2010 (Figure 1, and especially the contributions from the RFS Program), these earlier mismatches for soybean price may be less concerning for our purposes in this study (Discussed further below).

3. Results

Table 2, Figure 3, and Figure 4 (a, b) present results from a series of BSM-retro scenarios designed to estimate the effect these six key factors had on U.S. domestic biodiesel production. The scenario that had all six factors enabled (price competition, BTC, “splash and dash”, RFS, LCFS, and Argentine “dumping”) had production levels relatively commensurate with historical production (Figure 3a and S5a). Using this as our reference scenario, a total of 17.82 billion gallons of biodiesel were simulated to be produced from 2002 to 2020 (Figure 3b, red bar), very close to the observed amount (16.84 billion gallons). We report the impact that each factor had relative to the set of factors it is building upon for each year. The order in which the factors were built may influence the magnitude of each effect as discussed in the Limitations section. Unless otherwise stated, the scenarios were run with actual oil prices.

Table 2.

Estimated Impacts from Modeled Scenarios

Driver	Modeled effect on 2002–2020 cumulative biodiesel production as percentage of cumulative reference production (billion gallons)	Modeled effect on 2010–2020 cumulative biodiesel production as percentage of cumulative reference production (billion gallons)
Price competition	20.0% (3.6 billion gallons)	19.4% (3.1 billion gallons)
BTC	34.8% (6.2 billion gallons)	31.3% (5.0 billion gallons)
“Splash & Dash”	5.4% (1.0 billion gallons)	4.4% (0.7 billion gallons)
RFS Program/RINs	36.0% (6.4 billion gallons)	40.6% (6.4 billion gallons)
LCFS	7.0% (1.3 billion gallons)	7.9% (1.3 billion gallons)
Argentine Dumping	-3.1% (−0.6 billion gallons)	-3.5% (−0.6 billion gallons)

Figure 3.

Observed and modeled annual (a, billion gallons/year) and cumulative biodiesel production (b, billion gallons) from 2002 –> 2020 for different scenarios. Note that small state effects are included in the BTC (explaining nonzero results for the BTC pre-2004). Bar colors on the right also correspond to the legend on the left.

Figure 4.

Proportion of annual biodiesel production contributed by each model factor relative to reference production for 2002 –> 2020 (a, %). Modeled annual biodiesel production (b, billion gallons/year) contributed by each model factor relative to reference production for 2002 –> 2020. Price competition is embedded in the reference.

Over the study period, price competition alone (“baseline,” scenario PC, Figure 3, 4) accounted for 0–0.4 billion gallons/yr annually, or 20% (3.6 billion gallons)⁵ of the cumulative biodiesel production from 2002–2020 (Figure 3, 4). The influence of price competition decreased in importance from about 88% (0.03 billion gallons/yr) in 2002 and 2003 when few other factors were present to about 24% (0.05 billion gallons/yr) in 2006 then stayed at around 20% (0.24 billion gallons/yr) from 2007 through 2020.

Including the BTC as a factor increased annual production to 0.1–1 billion gallons/yr, contributing to 35% (6.2 billion gallons) to the total simulated biodiesel production for 2002–2020 (Table 2). The proportion of production potentially attributable to the BTC was high in the early years before the RFS2 (60–70% from 2004–2009), and decreased after the RFS2 went into full effect (generally 20–40% from 2010–2020) of about 59% (0.47 billion gallons/yr) from 2006 to 2011 then at a generally lower 29% (0.74 billion gallons/yr) for 2012 to 2020.

Around the time the BTC was enacted, the “splash and dash” technicality was capitalized upon. Our BSM-retro simulations indicated that “splash and dash” accounted for an average of about 17% (0.07 billion gallons/yr) annually from 2006 to 2008 but only 5% (0.95 billion gallons) of cumulative biodiesel production from 2002 to 2020.

One of the primary foci of this study, the influence of the RFS by way of RINs, was the most influential modeled factor, driving 36% (6.4 billion gallons) of cumulative biodiesel production over the whole study period (Figure 3b, Table 2). This was slightly larger than the effect from the BTC (35%). D4 RINs became a strong determinant of biodiesel production starting in 2010, the year that RFS2 rules went into full effect, when they accounted for 25% (0.16 billion gallons/yr) of the annual driving force behind biodiesel production. The RFS dropped in importance to 4% (0.04 billion gallons/yr) in 2011 but then increased to an average of about 44% (0.69 billion gallons/yr) per year for 2012 through 2020. During these later years the annual influence of the RFS Program varied from a low of 27% (0.38 billion gallons/yr) in 2013 and a high of 59% (0.88 billion gallons/yr) in 2016. During the dramatic reduction in the relative importance of the RFS from 2010 to 2011 the difference was made up by a concomitant increase in BTC combined with “splash and dash”. When combined, the factors BTC, “splash and dash”, and the RFS accounted for 76% (1.1 billion gallons/yr) of cumulative simulated biodiesel production from 2010–2020.

The fourth factor added, the California LCFS, was simulated to play a relatively minor role, accounting for 7% (1.25 billion gallons) of the cumulative 2002–2020 total biodiesel production. However, more significant portions of biodiesel production could be apportioned annually to the LCFS in 2015 (13%, 0.19 billion gallons/yr), 2016 (13%, 0.20 billion gallons/yr), and 2019 (28%, 0.51 billion gallons/yr).

The addition of Argentine “dumping” was the only factor added to the BSM that could potentially reduce U.S. biodiesel production. If the U.S. received large volumes of biodiesel from Argentina, producers in the U.S. would have more to compete with, their expected financial return would diminish, and domestic production would decrease. Adding the influence of Argentine “dumping” to our simulations accounted for a reduction in biodiesel production by a total of −3% (−0.56 billion gallons) over the study period. Argentine “dumping” played the largest role in 2016 when it accounted for a −21% (−0.31 billion gallons/yr) production change.

4. Discussion

4.1. Historical Context and Model Agreement

To begin to disentangle the relative effects that different drivers have on the biodiesel industry, we first examine our simulation of biodiesel production with only price competition. We then discuss additional factors that played a large role in the industry, presented in general chronological order.

4.2. Price Competition/Macroeconomics

In contrast to blending ethanol in gasoline, adding biodiesel to conventional diesel has not been economical without substantial policy and financial incentives (Drabik, De Gorter, & Timilsina, 2014; Naylor & Higgins, 2017) (Figure 5). Without these interventions, biodiesel production levels would be determined solely by market forces. Among the most important economic factors related to the biodiesel industry, crude oil and diesel prices have a strong role in determining, and are strongly correlated to, biodiesel price (Naylor & Higgins, 2017). During periods of high oil and high diesel prices, producers will have a greater incentive to produce biodiesel. Considering only the historical prices of diesel, the only periods when high biodiesel blends would be expected to be financially competitive with no external incentives would be December 2013 – March 2015, April 2018 – August 2019, and a few other transient periods (Figure 5, blue line and boxes). These upswings in diesel price were unrelated to biofuel policies and corresponded to global-scale macroeconomic changes.

Figure 5.

The price of diesel in relation to biodiesel ($diesel/$biodiesel) with (dashed red line) and without (solid blue line) the $1/gallon BTC subtracted from the price of biodiesel. Biodiesel was cost competitive with diesel for much longer periods after accounting for the BTC (red boxes) compared to the price of biodiesel without accounting for the BTC (blue boxes). Price ratios above 1.0 suggest biodiesel is cost competitive with diesel, all else being equal. ⁶

In addition to the price of oil, the incentive to invest in biodiesel conversion facilities, and the incentive to use facilities in existence, is also driven by the cost of feedstocks (Eidman, 2007). Therefore, we consider the historical trends in the prices of oil and biodiesel simultaneously with the prices of soybeans, soybean meal, and soybean oil. From late-2005 through 2006 crude oil prices were increasing (partly due to hurricanes Katrina and Rita) but the prices of agricultural commodities were low (Schnepf, 2013).

Previous work (i.e., ICF (2018), etc.) has shown that without the RFS, no biodiesel would be produced in 2020 relative to 2016 with oil prices below $72/bbl. Our baseline BSM-retro simulation shows general agreement. In the first few years of the simulation, our price competition-only scenario follows observations closely, but the results diverge around 2005, when oil spot price (Cushing, OK WTI FOB) is above $1/gallon ($42/bbl) for the entire year and diesel price (U.S. No 2 wholesale) spikes to over $2/gallon ($84/bbl). The role of price competition continues to increase but very slowly throughout the rest of the simulation, showing muted responses to interannual changes every year except 2010. Over the whole simulation period, our BSM-retro scenarios that include price competition singly as the driving factor account for only about 20% of the cumulative production volume (Figure 4a, “Sum”). In addition, our simulations show that from 2004 to 2008 price competition decreases in fractional importance as other factors entered the simulation (Figure 4a). From our simulations, in agreement with other studies price competition alone was not enough to drive the biodiesel production levels over the study period. In sensitivity tests, we found that setting oil price to a constant $100/bbl, instead of following historic oil prices, greatly impacted biodiesel production trends (Figure S1), generating about triple the cumulative biodiesel production compared to using historical oil prices for 2002–2020. However, this is still about half (0.6) of what our scenario with all factors included produces.

4.3. BTC

Over the study period and the sequence used for this analysis, the BTC was the second most influential driver, accounting for a total 35% of simulated biodiesel production for 2002–2020. During our simulation period, the first substantial rise in simulated biodiesel production occurred in 2005. Production continued to accelerate through 2007, increased by a slightly lower rate in 2008 then decreased for a few years. 2005 corresponds to the first year of the BTC, which added a financial incentive that was a constant $1 per gallon for agri-biodiesel. This was a critical factor in shifting average biodiesel production upwards.

The BTC had a complex temporal pattern, where it was in effect consistently and prospectively from 2005–2009, but from 2010 until present day it has been re-instated on an annual basis. In some years it was re-instated before or near the beginning of the year of effect; in other years it was reinstated retroactively. In years where it was proactively re-instated, biodiesel producers likely had much more incentive to produce biodiesel because of the certainty of the credit in that year. The BTC appears to have a large effect on the annual change in biodiesel production, especially when the BTC was available prospectively for producers (Figure 1). The intermittent character of the BTC is reflected in the cyclic changes in biodiesel production rates, with production increases corresponding to years when the BTC was in effect (especially proactively) and production rates slowing when the BTC was not in effect (Figure 1). This strong mirroring pattern suggests that the BTC was a dominant factor controlling biodiesel production. However, a growing body of literature (de Gorter, 2010; de Gorter, Just, Korting, & Radich, 2019; de Gorter & Just, 2009) discusses the importance of the BTC relative to the RFS Program and generally concludes that the RFS mandates are the dominant driver for biodiesel production and that the BTC acts as a subsidy for fuel consumption. Our results suggest the effects from these two policies were comparable (discussed in the next section).

4.4. RFS

In 2007 the U.S. government passed the Energy Independence and Security Act (EISA), which built on the existing RFS by increasing mandated biofuel volumes and creating four nested RIN categories for cellulosic biofuel, biomass-based diesel (BBD), advanced biofuel, and total renewable fuel. We do not consider the RFS1 to have had any effect on biodiesel production for three reasons: (1) there was no specific biodiesel requirement under the RFS1, (2) biodiesel was much more expensive than corn ethanol to produce, and thus (3) we expect that most refiners fulfilled their RFS1 requirements through blending with corn ethanol. Internal EPA data support this assumption as most early RINs were associated with corn ethanol and imported sugarcane ethanol from Brazil (U.S. EPA, In Press). EPA implemented the RFS2 in stages, with all four categories established by 2010, the year that requirements specific to soy biodiesel were released. The BSM-retro captures the effects of the RFS through the RIN market, which is the market trading scheme used by the EPA to ensure obligated parties are meeting their biofuel blending requirements. Our simulations show that the RFS Program became an important driver behind biodiesel production in 2010 (Figure 4).

The RFS Program explains about one third (Average 38.4% (0.6 billion gallons/yr), Figure 4) of the driving force behind simulated production during 2010–2020, from a minimum of 3.7% (0.04 billion gallons/yr) in 2011 to a maximum of 59.1% (0.9 billion gallons/yr) in 2016 (Figure 4). The RFS alone accounts for an average of 45% (0.8 billion gallons/yr) for 2017–2020. When added to the modeled factors that came into play before the RFS2 (price competition, BTC, and “splash and dash”), the combined BSM-retro scenario (price competition, BTC, “splash and dash”, and RFS Program) captures 95.6% (15.1 billion gallons) of the reference scenario’s cumulative biodiesel production for 2010–2020.

The Third Triennial Report to Congress on Biofuels (RtC3) contains a preliminary analysis of the potential maximum impact that the RFS Program had on the biodiesel and renewable diesel industries by comparing total domestic production of biodiesel and renewable diesel to volumes required by state mandates and incentives. Using this approach, the volume attributable to the RFS and BTC combined averaged 70% (1.19 billion gallons/yr) for 2010–2019, with a low of 52% (0.14 billion gallons/yr) in 2010 and a high of 74% (1.9 billion gallons/yr) in 2016. Our simulations suggest the contributions from the BTC and RFS are roughly split, attributing an average of about 37.9% (0.6 billion gallons/yr) of biodiesel production per year for 2010–2019 and 40.1% (5.6 billion gallons) of the cumulative reference production (13.9 billion gallons) over the interval solely to the RFS. When combined, the factors BTC, “splash and dash”, and RFS accounted for 76% (1.1 billion gallons/yr) of total simulated biodiesel production from 2010–2020.

4.5. LCFS

Although we have established that the combined effects of price competition, the BTC, and the RFS Program account for most biodiesel production from 2002–2020, the California LCFS has had a steadily increasing effect in recent years and is expected to become a dominant factor in the biofuels industry in the future (Gerveni, Hubbs, & Irwin, 2023). Although the LCFS program was enacted in 2007, it did not go into effect until 2011, and it took a few more years until biofuel volumes used to meet the program’s requirements meaningfully increased (Figure S2). Our simulations first show an impact on national-level production due to the LCFS in 2013, when it contributed 0.7% (0.01 billion gallons/yr) to simulated domestic biodiesel production. This compares favorably to the reported 0.06 billion gallons of biodiesel used in California during 2013⁷. The difference between our estimate and the actual use could be explained by California meeting some of their biodiesel demand via imports, which increased rapidly in 2013 (EIA, 2022).

Notably, renewable diesel consumption has increased much more rapidly than biodiesel consumption, due to it having a lower calculated carbon intensity than biodiesel and not requiring engine modifications for higher blends. Actual biodiesel consumption increased by an average 11.4% annually during 2012–2019 while actual renewable diesel consumption increased by an average 54.5% annually during the same period (Figure 1, 6). In 2013 California consumed 60 million gallons of biodiesel and 117 million gallons of renewable diesel. By 2019, California consumed 212 million gallons of biodiesel and 618 million gallons of renewable diesel.

Figure 6.

Observed (dashed black line) and modeled annual renewable diesel production (billion gallons/year) from 2002 –> 2020 for the reference scenario (solid red line) and a scenario with all factors except for the LCFS (solid blue line).

Even though the LCFS is aimed at reducing California’s transportation carbon footprint through enhanced biodiesel and renewable diesel use, the program is altering national biodiesel and renewable diesel consumption patterns (H. Huang, Khanna, Önal, & Chen, 2013; Lade & Lawell, 2015; Whistance, Thompson, & Meyer, 2017). Several studies (e.g., (H. Huang et al., 2013; Whistance & Thompson, 2019; Whistance et al., 2017) have illustrated how the RFS mandate and the LCFS interact. Literature generally concludes that, between the two, the RFS Program has the greatest influence on biodiesel production and the LCFS plays a strong secondary role (H. Huang et al., 2013; Lade & Lawell, 2015; Whistance & Thompson, 2019; Whistance et al., 2017). The LCFS acts like a consumption subsidy, helping obligated parties meet the RFS mandates. There is also evidence that both programs assist each other in promoting biodiesel production (H. Huang et al., 2013; Whistance et al., 2017). However, it is possible that the coexistence of the RFS and LCFS programs does not increase overall U.S. consumption and production but instead causes a shift, whereby the increase in renewable diesel and biodiesel consumption in California is approximately equal to a concomitant reduction in renewable diesel and biodiesel use in the rest of the U.S. (Whistance & Thompson, 2019; Whistance et al., 2017).

4.6. Trade Proxies: “Splash and dash”, Argentine Dumping

Two more modeled factors that had a relatively minor effect on biodiesel production were “splash and dash” and Argentine “dumping”. The existence of the BTC contributed to a period of active international biodiesel trade when a technicality allowed exporters to earn a tax credit for every gallon of U.S. biodiesel blended into diesel anywhere in the world (Lamers, Rosillo-Calle, Pelkmans, & Hamelinck, 2014). While not nearly as impactful as the BTC or RFS, the “splash and dash” effect increased simulated biodiesel production by an average of 17% during the peak years of the anomalous trade period 2006–2008. Even though the BSM-retro only includes the “splash and dash” phase until 2008, the temporary financial boon created by higher prices received by U.S. producers (de Gorter, Drabik, & Just, 2011) gave domestic producers slightly more incentive to invest in production capacity, which bolstered production levels after the closure of the technicality (Figure 3, 4). Further, the EU’s trade incentives that were behind the “splash and dash” phase extended beyond 2008, and U.S. producers continued to export larger quantities than normal to Europe⁸.

The final factor to consider is the brief window of time when the U.S imported large volumes of biodiesel from Argentina. Building on a strong oil crop industry, the government of Argentina created a national biodiesel strategy in 2001 that used tax exemptions and eventually laws, along with a National Commission on Biofuels, to bolster the nation’s soy biodiesel industry (Naylor & Higgins, 2017). This resulted in Argentina becoming one of the world’s leaders in biodiesel production and export, with the U.S. as an especially important trading partner (Beckman, 2015). During 2013–2017 for example, an average of 52% of U.S. imports were from Argentina (EIA, 2022). Imports from Argentina fell to nearly zero after 2017 due to the U.S. Department of Commerce creating an anti-subsidy duty in December 2017 and the U.S. International Trade Commission setting an anti-dumping duty in April 2018 (ITA, 2017; USDA, 2018). Argentine “dumping” had a limited impact on simulated domestic biodiesel production. Despite asserting a strong −21% influence on simulated U.S. biodiesel production in 2016, Argentine “dumping” had a minor role in simulated biodiesel production, overall. Simulated results suggest that domestic producers may have responded by investing less in their facilities’ capacity, which influenced future production levels by −1% to −5% from 2018–2020. This is consistent with reports of stalled plant modification plans as recorded by the National Biodiesel Board prior to the anti-dumping determination in 2017 (National Biodiesel Board, 2017).

4.7. Comparison with other studies

There are relatively few modeling studies that have assessed historical biodiesel production, especially compared with corn ethanol studies. However, among those that we are aware of, most show general consilience and report that the majority of biodiesel production is driven by RFS Program mandates (Table 3) (Babcock, 2012; Hayes et al., 2009; J. Huang, Yang, Msangi, Rozelle, & Weersink, 2012; Meyer, Binfield, & Thompson, 2013; Taheripour et al., 2022; U.S. EPA, 2010). These studies found that biodiesel production would have been low without the RFS Program mandates. With the RFS Program mandates for biodiesel in effect, a 1-billion-gallon mandate resulted in a biodiesel production increase of about 0.7–1.9 billion gallons. There are important caveats to each of these studies. For example, Taheripour et al. (2022) and Babcock et al. (2012) simulated ethanol simultaneously with biodiesel so that the biodiesel attribution value is partially dependent on ethanol production. Also, the studies used different years and sets of assumptions for their baseline scenarios and ended their simulations in different years. The results from our BSM-retro simulations compare well within the set of studies. This is different from studies that evaluated ethanol, which find much more divergent estimates (Austin, Jones, & Clark, 2022) – suggesting confidence in the estimated effects of the RFS Program on biodiesel. The systems dynamics approach utilized here, although agreeing with other studies in direction, suggest that the BTC and the RFS Program together are responsible for the bulk of biodiesel production in the U.S., which is commonly ascribed to the RFS Program alone in these earlier studies which do not include the BTC.

Table 3:

Estimates of biodiesel production volumes attributed to RFS from various models/studies.

Study	Production without RFS (billion gallons)	Production with RFS (billion gallons)	Year applicable	RFS mandate (billion gallons)	Increase in production with RFS (per billion-gallon mandate)
U.S. EPA (2010)	0.4	1.670	2022	1.8	0.7
Babcock (2012)	0.04	0.9	2011	0.8	1.1
Huang et al. (2012)	0.2	1.9	2020	1.9	0.9
Meyer et al. (2013)	0.4	1.3	2017–2021 avg	1.0	0.882
Taheripour et al. (2022)		1.2	2011–2016 total
RtC3 (U.S. EPA, In Press)		1.19	2010–2019 avg
BSM (Current paper)		1.09	2011
BSM (Current paper)		1.86	2020
BSM (Current paper)		1.68	2017–2021 avg
BSM (Current paper)		1.24	2011–2016 avg
Observed		1.81	2020

4.8. Limitations and Uncertainties

Our analysis is based on the potential effect of each factor added onto already existing drivers. This approach assumes each factor is independent whereas, in practice, producers and blenders likely decide production levels based on a longer-term view and a more subjective perception of economic risk. These intangible effects of market and policy certainty have been estimated for ethanol but not for the biodiesel industry. Parameterizing market certainty in the BSM as a reduction in expected required rate of return for investors, Newes et al. (2022) saw ethanol growth accelerate by 1 year and production levels increase by up to 2 billion gallon/year. These effects, however, attenuated over time such that the risk-reduction effect of the RFS Program on ethanol was to move the growth of the industry towards earlier years, and did not increase its overall size as these facilities were projected to be built anyway later due to other factors. Biodiesel production facilities also require substantial investment in time and capital. Therefore, a more thorough assessment of the potential impact of the RFS Program via increased market certainty could strengthen our estimate of the influence of the RFS Program as well as non-RFS factors.

A potentially important drawback of the BSM-retro approach is that the driving effect of the RFS Program is simulated through historical RIN prices which are exogenous to the model. Because the model does not endogenously estimate RINs, the potential effect of the RFS Program under alterative histories cannot be assessed with the current model structure. Therefore, it is possible that by using exogenous RIN values in the BSM-retro, the influence of market forces and other biofuel-related factors on RIN prices is not captured and that some BSM-retro analyses could attribute too much to too little biodiesel production to the RFS Program.

Our attributional analysis of the RFS Program and biodiesel growth is more robust during earlier years compared to recent years. This is because renewable diesel demand has been increasing substantially in the U.S. and this likely had feedbacks on the demand and resulting production of biodiesel that were not thoroughly accounted for in our analysis. Additionally, this rapid change in growth is difficult to parametrize in mathematical models and adds uncertainty to the results. While actual biodiesel production rates decreased by −0.05 billion gallons from 2018–2020 (Figure 1), renewable diesel production has continued to increase dramatically, increasing by 0.23 billion gallon from 2018–2020 (Figure 6, S3). The California LCFS is thought to be driving much of this renewable diesel demand, but other drivers such as the need for more sustainable biofuels and their potential use as drop-in fuels and for aviation will continue to enable this increase, facilitated by improved production technologies.

5. Conclusion and Policy Implications

Various policies in the U.S. encourage the development of transportation fuels that lessen dependence on imported oil while simultaneously reducing greenhouse gas emissions. Meanwhile, organizations that support domestic agricultural production strive to expand markets and income for American farmers. Furthermore, the three dominant biofuel types by volume—ethanol, biodiesel, and renewable diesel—each have different sets of benefits as well as challenges and have developed largely along different trajectories in response to differing economic factors and governmental policy interventions. Therefore, understanding how the industry has developed, including an assessment of how much each policy and non-policy factor contributed, is critical for informed decision making to encourage a more sustainable biofuels industry. Considering only economics, recent research (Meyer et al., 2013; E. Newes et al., 2022; Taheripour et al., 2022; U.S. EPA, 2018, In Press) shows that the ethanol industry at 10% blend levels is likely self-sustaining, without government support. In contrast, biodiesel blending has not been economical without policy interventions. We used the BSM-retro to apportion the relative importance of each driver in a retrospective analysis of biodiesel production volumes.

It is likely that the RFS Program had a very different impact on the biodiesel industry than on the ethanol industry. While both biofuels share a few traits that tie their economic viability, such as vulnerability to petroleum and agricultural commodity prices, sufficient differences exist that have necessitated the RFS and other governmental programs to boost biodiesel production beyond what would have been produced absent these policies. Recent research has indicated that domestic ethanol production has been less dependent on the RFS Program than biodiesel production. Newes et al. (2022) performed a similar analysis as here on ethanol and found that the influence of the RFS on ethanol production was minimal during the early years of growth (2006–2011), then increased during more recent years (2013–2019). These estimates are qualitatively similar to other estimates (Lark et al., 2022), once differences in attribution are accounted for. The extensive RtC3 report (U.S. EPA In Press), using a variety of lines of evidence such as models, RIN analysis, and literature review, determined that, overall, the RFS Program had a relatively minor role in driving ethanol production. This is largely due to key events that affected ethanol demand more than biodiesel demand, such as the need to replace MTBE, the VEETC, a shift from “splash blending” to “match blending”, the blend wall (Figure S4), and other non-RFS related factors.

For biodiesel, the effect of the RFS Program is different because factors like the blend wall and match blending are not present, and the economics are much less favorable compared with ethanol. Since biodiesel is expensive to produce (due largely to feedstock prices and capital investment costs) (Gebremariam & Marchetti, 2018; Haas, McAloon, Yee, & Foglia, 2006), it cannot compete with petroleum-based diesel without additional incentives. Hence, the federal government’s RFS Program was key in encouraging biodiesel development, not only due to RINs but also potentially through providing more market certainty.

Despite its inconsistent nature, the BTC was an important factor, especially during the early growth phase of the industry. The combination of the BTC plus the RFS2 provided sufficient incentive above price competitiveness for almost all the increased biodiesel production and led to a shift in the viability of biodiesel as a transportation fuel. However, our modeling found that once the RFS2 started and D4 RINs took effect, the RFS Program became the larger driver of biodiesel production. Research indicates that in years with both the BTC and the RFS active, the RFS mandates had most of the influence behind biodiesel production and the BTC acted primarily as a fuel consumption subsidy (Drabik et al., 2014). Efforts to disentangle the BTC from the RFS are useful for devising more efficient incentive mechanisms and could inform future efforts by stakeholders that want to encourage biofuel use (de Gorter, 2010; de Gorter, Drabik, & Just, 2013; de Gorter et al., 2019; de Gorter & Just, 2009; Drabik et al., 2014). The effect of one policy (e.g., the RFS2) depend on the presence and magnitude of effect of other policies (e.g., the BTC), suggesting that policymakers would benefit from simulations such as these when developing future policies. Methodologically, researchers and decision makers could also use systems dynamics models to explore potential policy impacts, in addition to the more common partial and general equilibrium economic models, as systems dynamics models allow for detailed examination of complex feedbacks that may be more difficult to capture using other approaches. This modeling approach could also be used to explore various pathways towards the production of sustainable aviation fuel, which likely faces similar complex barriers to growth. Total world energy consumption is expected to increase by about 50% compared to 2020 (mostly in non-OECD countries) (EIA, 2021). Overall biofuels demand will likely increase (Alizadeh, Lund, & Soltanisehat, 2020; Debnath, Khanna, Rajagopal, & Zilberman, 2019; Lusk, 2022). Policies are likely to continue to evolve. Thus, studying past policies can help inform more efficient incentive mechanisms for promoting alternative fuels use in the future.