Back in May, I discussed a study conducted for the Renewable Fuel Association (RFA) by Iowa State University’s Center for Rural and Agricultural Development (CARD). The study claims that from January 2000 to December 2011, “the growth in ethanol production reduced wholesale gasoline prices by $0.29 per gallon on average across all regions,” and reduced average gasoline prices by a whopping $0.89 per gallon in 2010 and $1.09 per gallon in 2011. Ethanol boosters like the RFA and USDA Secretary Tom Vilsack tout this study as proof that federal biofuel policies benefit consumers and should be expanded.
The CARD researchers, Xiaodong Du and Dermot Hayes, attempt to determine the consumer benefit of ethanol by inferring what motor fuel prices would have been over the past decade had there been no increase in ethanol production. Ethanol now constitutes roughly 10% of the motor fuel used by U.S. passenger vehicles. Du and Hayes conclude that without ethanol, U.S. motor fuel supply would be significantly smaller and pain at the pump significantly greater.
This procedure, I argued, is ridiculous. First, it assumes that refiners are like deer caught in the headlights and do not respond to incentives. Even if motor fuel prices increase by up to $1.09/gal nationwide over a 10-year period, we’re supposed to believe refiners would not increase output and take advantage of this opportunity to sell more of their product at higher prices. But that’s exactly what refiners would do. In the process, supply would come back into balance with demand, pushing fuel prices down.
Second, the CARD study ignores the opportunity costs of ethanol policy. Capital is a finite resource. Dollars that refiners are mandated or bribed to invest in ethanol production are dollars they cannot invest in gasoline production. The CARD study implausibly assumes that all the refining capacity diverted by federal policy into ethanol production would have been left idle in a free market and not used to produce gasoline instead.
Admittedly, the CARD study is full of math I don’t understand. But two experts in the field — MIT energy economics professor Christopher Knittel and UC Davis agricultural economics professor Aaron Smith — have just produced a technical critique of the CARD study. Titled “Ethanol Production and Gasoline Prices: A Spurious Correlation,” the researchers make several telling points, some of which are funnier than the standard fare found in the ‘dismal science.’
Knittel and Smith begin with a discussion of basic economics to “place loose bounds” on the potential effects of ethanol production on gasoline prices. They note that the largest component of the price of gasoline is the cost of crude oil.
A barrel of crude oil contains 42 gallons, so every dollar per barrel increase in oil prices raises wholesale gasoline prices by about 2.4 cents. Thus, when oil is $100 per barrel, roughly $2.40 of the price of gasoline will be the cost of crude.
Ethanol production can have only a “minimal impact” on crude oil prices. U.S. ethanol constitutes only 1% of world oil use. In addition, ethanol has one-third less energy content by volume than gasoline, so U.S. ethanol production replaces only 0.67% of world oil. Ethanol’s impact on the biggest factor affecting gasoline prices is likely very small.
Ethanol production could however affect gasoline prices by decreasing refiners’ profit margins. The CARD study concludes that the “crack spread” — the weighted average price of refined products minus the price of crude oil — would have been $0.89 higher if ethanol had been removed from the market in 2010 and $1.09 higher had it been removed in 2011. But, argue Knittel and Smith, crack spreads never stay that high for an entire year. Indeed, the “crack spread has not exeeded 60 cents for more than a few brief periods in the past 30 years.” The reason is that when “the crack spread is high, large profits encourage entry into the refining industry, which in turn puts downward pressure on the crack spread.” Or, as I put it above, refiners are not deer in the headlights; they respond to market signals (prices).
The CARD study implies that, but for ethanol production, the crack spread in May 2010 would have been $1.37 — “20 cents higher than the highest crack spread ever observed in the data.” Knittel and Smith comment: “For this to be a long-run effect — which is the implicit assumption in the RFA’s claims — we would have to expect that these historic high crack spreads would not increase capacity utilization,” even though refinery utilization averaged 86.4% in 2010, “lower than every year from 1992-2007.” In other words, we would have to assume that refiners don’t want to get rich.
For reasons of space, I won’t try to summarize Knittel and Smith’s detailed discussion of “issues related to model specification.” More newsworthy and certainly more entertaining is their discussion of “spurious correlation.” Using the CARD study’s models, they estimate the effects of ethanol production on natural gas prices and unemployment rates — “dependent variables” with no particular connection to ethanol. Here’s what they find.
Based on the CARD models, had no ethanol been produced in the U.S. in 2010, “natural gas prices would have increased by 65 percent.” Similarly, “eliminating ethanol in 2010 would have decreased U.S. unemployment by 65 percent.” (So much for RFA’s claim that ethanol creates jobs!)
The authors comment:
These empirical relationships are a classic example of spurious correlation. Ethanol production during this time period is increasing. Therefore, other variables that have a predominant trend, either upward in the case of unemployment or downward in the case of natural gas prices, are likely to correlate well with ethanol production.
To nail down the point, the researchers provide a “whimsical” example. Using CARD’s models, they find that every additional million barrels of ethanol produced increases the age of daughter Caiden Knittel by 26 days and that of daughter Hayley Smith by two months. Policy implication: Eliminating all ethanol in 2010 would “cause Caiden to be a newborn (12 days old) and would cause Hayley’s age to be negative.”
Knittel and Smith conclude:
The results of Du and Hayes are at odds with the historical levels of either the crack spread or the crack ratio [the price of gasoline divided by the price of crude oil] and are inconsistent with an equilibrium [a long-term balance between supply and demand] in the oil refining industry. While an instantaneous surprise elimination of all ethanol sold in the U.S. might raise gasoline prices for a short period, one cannot assume these instantaneous effects would persist for more than a few weeks. This is precisely what Du, Hayes, the RFA, and Secretary Vilsack have done.