1. Introduction
Food versus fuel is America’s newest battle. Lying on the front is the question of whether America can increase annual corn based ethanol production from 10.75 billion gallons today to 15 billion gallons in 2015 (Renewable Fuels Association, 2010). The Renewable Fuel Standard (RFS) program was originally created by the Energy Policy Act of 2005 and expanded under the Energy Independence and Security Act (EISA) of 2007. It recently underwent statutory revisions in February 2010 mandating that 15 billion gallons of corn based ethanol must be blended into gasoline annually by 2015 (U.S. Environmental Protection Agency, 2010).
This represents a long history of U.S. state and federal ethanol policies. Although ethanol is used as a transportation fuel today, it was originally used as an illuminating oil prior to the Civil War. A federal tax of $2 per gallon imposed in 1862 to fund war efforts made ethanol’s use cost prohibitive. Even though the tax was removed in 1906, it wasn’t until the Energy Tax Act of 1978 that America’s modern ethanol industry was born (U.S. Department of Energy, 2008).
Today, domestic ethanol production is encouraged through a combination of state and federal policies. Primary federal ethanol policies consist of: RFS ethanol use mandate, Winter Oxygenated Fuels Program, Small Ethanol Producer Tax Credit of $.1 per gallon for plants producing less than or equal to 60 million gallons of ethanol per year, $.45 per gallon ethanol blender’s credit (RFA, 2010), and a $.54 per gallon tariff applying to imported ethanol with the exception of the duty free importation of 240 million gallons of ethanol annually under the Caribbean Basin Initiative (Zhang, 2007). A time line of major policy events in U.S. ethanol production is presented in the graph titled “U.S. Ethanol Supply and Demand” on page 1. The graph links policy changes to levels of U.S ethanol production, consumption and imports.
In addition to federal ethanol policies, there are 106 different state laws affecting ethanol production and marketing. These state policies fall in one of five broad categories: producer incentive programs (i.e. preferential tax treatment)/grant funds, retailer/infrastructure incentives for ethanol blends and E-85, state use mandates, retail pump label requirements, and state fleet fuel purchase/use requirements.
Simulations run by the Food and Ag Policy Research Institute (FAPRI, 2008) suggest state policies have a minor impact on ethanol demand today, shifting the U.S. ethanol demand curve outward by no more than 10% . However, it was state level bans on methyl tertiary butyl ether (MTBE)[1] between 2002 and 2007 (Low and Isserman, 2009) which enabled the U.S. ethanol industry to double its annual capacity from 2.3 billion gallons in 2002 (61 plants) to 5.5 billion gallons (110 plants) in 2007 according to the Renewable Fuels Association (2010). This growth lessened in 2008 as a result of a recession and poor ethanol plant profitability caused by commodity markets collapsing. These conditions sparked a round of bankruptcies and plant shut downs starting in 2008 that found 36 plants (Johnson, 2009) or 21% of total U.S. ethanol capacity idle by spring 2009 (RFA, 2010). Although the majority of idle capacity is presently back in production and 11 new ethanol plants are currently being built, the wide spread effects of ethanol warrants an examination of U.S. ethanol policy and production. The objective of this paper is to answer the question, what impact does government policies have on U.S. ethanol production?
This answer to this question is of interest to policy makers and industry organizations alike. The puzzle in the literature this paper specifically tries to address is what is the effect of states banning the fuel additive MTBE if the Reformulated Gasoline (RFG)[2] program is in place? To fill this whole in the literature, 2SLS is used where the quantity of ethanol and price of ethanol are treated as endogenous. It is found that increasing the percentage of the total population living in states with MTBE bans by 10% will always lead to a .255% increase in the quantity of ethanol demanded regardless of whether or not the RFG program is in place. This estimate should be interpreted with caution. Although the system of equations using annual national level data did not suffer from autocorrelation and had exogenous instruments to resolve the endogeneity of ethanol price in the supply and demand equation, the instruments used to identify the demand curve were not relevant and therefore MTBE was the only government policy variable in the demand equation to be statistically significant. Only one government policy variable, the federal blender’s tax credit was included in the supply equation and it was statistically significant. A 1% increase in the federal blender’s credit results in a 2.4% increase in the quantity of ethanol produced.
2. Literature Review
Controversy surrounding ethanol and the broader renewable fuels movement has enabled economists, political scientists, environmental scientists and engineers to secure extensive grants to study the ethanol industry. This grant funding has incentivized many scholars to focus their research on specific ethanol policy or program effects. In response, little work has been done to estimate U.S. ethanol supply and demand curves. The work that has been done to estimate U.S. ethanol supply and demand curves using 2SLS was pioneered by Kevin Rask. Rask used state level monthly data from 1984 to 1993 to develop an ethanol supply and demand model and to calculate elasticities (1998).
In similar spirit, Luchansky and Monks used monthly national level data from 1997 to 2006 to estimate U.S. ethanol supply and demand curves. Luchansky and Monk also calculated elasticities to measure the response of ethanol production to ethanol, gasoline, MTBE and corn price changes. Problems using endogenous variables were resolved using 2SLS. It was found that when the corn price was treated as endogenous it had a positive coefficient implying that an increase in corn prices leads to an increase in the equilibrium quantity of ethanol. In contrast, when a corn price instrumental variable was used instead of the corn price itself the coefficient on the corn price instrumental variable had the correct sign (Luchansky and Monks 2009).
This paper expands the work of Luchansky, Monks and Rask by using annual national level data from 1982 to 2008. In addition, Luchansky and Monks’ supply curve specification does not account for the value of ethanol byproducts used for livestock feed. Luchansky and Monks include corn oil as a co-product price, but most ethanol plants are dry-grind plants and don’t have the technology to capture corn oil from the corn kernel. In response, this paper includes the soybean meal price in the supply equation as a proxy for the value of ethanol plant byproducts used as livestock feed.
In addition to variable selection, it is also important to consider this paper’s econometric methods. 2SLS is commonly used in commodity markets to estimate supply and demand curves, although it is difficult to correctly specify the structural equation for some markets. For example, C.Y. Cynthia Lin in her paper Estimating Annual and Monthly Supply and Demand for World Oil: A Dry Whole? used 2SLS to estimate aggregate supply and demand curves for world oil (2004). Lin’s use of instrumental variables did not yield coefficients of the expected sign for OPEC demand and many of the parameters in the supply equations. This indicates that Lin’s econometric specifications or economic theory didn’t accurately reflect the complexities of the world oil market (2004). These results suggest that correctly specifying aggregate supply and demand models is difficult and initial attempts to estimate aggregate supply and demand models with 2SLS commonly yields coefficients whose signs contradict economic theory.
Unlike previous works, this paper uses annual data to avoid autocorrelation associated with monthly data. In addition this paper uses national data, but creates an instrument to account for state level changes in MTBE bans over time.
3. Economic Theory
U.S. ethanol price and quantity is determined by the intersection of supply and demand curves. The primary determinant of the ethanol supply curve is corn price. According to the USDA’s January Feed Outlook Report, 32.1% of the U.S. corn crop for the 2009/2010 marketing year, is used for ethanol while the remainder is used for non-ethanol food, seed and industrial use (9.7%), livestock feed and residual use (42.5%), and exports (15.7%), thus, any changes in ethanol policies or production influences corn prices (2010). This means that corn price can not be directly used in the supply equation. The corn price must either be proxied for using a lagged corn price or be instrumented for using 2SLS.
Ethanol production’s impact on input markets is likely heightened by government policies aimed to directly or indirectly increase ethanol demand. Since MTBE was banned in over twenty states from 2000 to 2007 while the RFG program mandating use of either MTBE or ethanol to make gasoline burn cleaner was in place from 1995 to 2005, economic theory suggests that an MTBE ban will increase ethanol demand by a larger amount when the RFG program is in place than when the RFG program is not in place.
4. Data
Ideally state level panel data would be available for the price and quantity of ethanol produced and consumed in the U.S. since 1982. Unfortunately, the U.S. Department of Energy stopped maintaining its state level ethanol series in May 1993. Currently, the U.S. Department of Energy has annual state level data and monthly national level data on the quantity of ethanol consumed and produced. Ideally, state level historical ethanol prices would be available. In order to estimate a system of equations using state level panel data, the quantity of ethanol shipped in and out of each state has to be known. The U.S. Department of Energy (2009) currently tracks ethanol shipments, but only tracks the quantity of ethanol shipped and imported by region, not state and the data series was just started in 2009. It would also be ideal to have a database on all federal and state level ethanol subsidies so a variable of total subsidy per gallon of ethanol could be computed.
In the presence of data constraints, this analysis uses U.S. national annual data from 1982 to 2008. National level data for the quantity of ethanol produced, quantity of ethanol consumed, level of federal blender’s ethanol credit, well head natural gas prices and city average retail prices for all types of gasoline including taxes came from the United State’s Department of Energy’s Energy Information Administration (2009). Ethanol price data were received from Nebraska’s Energy Office (2009). Average annual corn and soybean meal prices were computed using the Chicago Board of Trade’s daily closing prices for the nearby contract. This data was received from Montana State University’s Department of Agricultural Economics and Economics (2009).
All of the price data above were converted to real terms using annual GDP implicit price deflator values with 2008 as base year. The GDP implicit price deflator and the U.S. average per capita income were retrieved from the U.S. Bureau of Economic Analysis (2009). State and national level population data was received from the U.S. Census Bureau’s 2010 Statistical Abstract(2010). Data on the number of licensed driver in the U.S. was received from the U.S. Department of Transportation (2009). In addition, the data on which states implemented MTBE bans into law were provided by the United State Environmental Protection Agency (2007). All additional information concerning state and federal ethanol programs were obtained from the Renewable Fuels Association (2010).
Table 1. Summary Statistics For U.S. Ethanol Market, 1982 to 2008Mean / Std. Dev. / Min / Max
Endogenous Variables
Ethanol production, per U.S. licensed driver(gallons / driver) / 8.55 / 8.23 / 1.25 / 37.84
Ethanol consumption, per U.S. licensed driver(gallons / driver) / 8.78 / 8.74 / 1.25 / 39.36
Real Ethanol Price ($/gallon) / 1.98 / 0.56 / 1.22 / 3.35
Real Corn Price ($/bushel) / 3.64 / 1.01 / 2.29 / 5.98
Exogenous Variables Exclusive to Supply Equation
Real corn price lagged one year ($ / bushel) / 3.65 / 1.03 / 2.29 / 5.85
Real natural gas price ($ / thousand cubic) / 4.07 / 1.00 / 2.06 / 8.07
Real soybean meal price ($ / ton) / 263.20 / 1.78 / 172.90 / 381.60
Real federal blender's credit ($/ gallon) / 0.75 / 0.17 / 0.51 / 1.09
Exogenous Variables Exclusive to Demand Equation
Real Unleaded Gasoline Price ($/gallon) / 1.96 / 0.48 / 1.41 / 3.32
MTBE instrumental variable / 0.10 / 0.19 / 0.00 / 0.51
Winter Oxygenate Program 0/1 Dummy / 0.63 / 0.49 / 0.00 / 1.00
Reformulated Gasoline Program 0/1 Dummy / 0.41 / 0.50 / 0.00 / 1.00
U.S. average per capita income / 32.81 / 5327.73 / 23327.00 / 40222.00
N = 27 annual observations
Table 1 summarizes all key variables used in the study. There is a large difference between the maximum and the minimum quantity of ethanol produced per licensed U.S. driver. The quantity of ethanol produced per licensed U.S. driver was 1.25 gallons in 1982 and steadily increased to 37.84 gallons in 2008. It is also worth noting that the mean for the instrumental variable MTBE is only .1 because the first state level MTBE ban did not occur until 2000.
5. Theoretical Framework
Testing the affects that state and federal ethanol policies have on U.S. ethanol production would ideally be done implementing each policy one at a time for an extended period of time to quantify the short and long-run impacts each policy has when they are the sole ethanol policy used. Then a series of long-run experiments would be conducted by implementing different ethanol policies such as a blenders credit and tariff on imports simultaneously to quantify the interaction effects of different government ethanol policies. Over the period in which these experiments are conducted, all variables such as population and per-capita income have to be held constant so that the study’s results aren’t confounded by changes in variables besides the ethanol policies themselves. Unfortunately, Figure 1 shows that empirically estimating the impact of government policies on U.S. ethanol production is not easy, because there is not a clear systematic relationship between the quantity of ethanol demanded and the price of ethanol. This occurs because of the large number of governmental policies affecting U.S. ethanol industry simultaneously and also structural changes that have occurred in demand, such as consumers preferences for vehicles able to burn higher amounts of ethanol and also structural changes in ethanol supply such as improvements in ethanol processing technology.