Model building, Playboy and the impact of ethanol on gasoline prices
I recently read a number of provocative articles (or their summaries) by MIT’s Christopher Knittel and Aaron Smith. They faulted a pair of respected researchers from Iowa State University, Dermot Hayes and Ziaodong Du, in somewhat harsh tones. According to Knittel, the Iowa State pair, in their ethanol-related studies over a three year period (from 2009 through 2012), exaggerated the impact of ethanol on gas prices using relatively low present day ethanol blends.
I thought I was reading the script for a new urban crime show about drugs. Knittel, frequently, used terms like crack ratio and crack spread, ostensibly to note the weak link, found by Hayes at Iowa State, between the prices of ethanol and oil and both to gas costs at the pump. According to the authors, the price of gasoline is not substantially affected by the crack ratio; that is, the relative value of gasoline compared to oil or the price of gasoline divided by the price of oil and the current volume of its ethanol content.
Knittel’s papers angered Hayes, of the Iowa study. He claimed that, over time, the crack ratio and crack spread reflected a pretty strong causal relationship to gas prices. Language in his response to Knittel’s critique reminded me of those wonderful days when I was a dean, listening to different faculty, sometimes personally and sometimes based on methodology, criticize other faculty based on differing research results. The search for academic truth is often a noble road, but paraphrasing Robert Frost, a “road less traveled” — a road often full of human frailty and intellectual potholes.
Despite their critique of each other, both Knittel and Hayes’ studies are important and both, when read in context, should help one better understand the role of ethanol in affecting the cost of gas at the pump. Knittel is more right than wrong when he indicates that the crack ratio and spread does not fully explain the effect of ethanol on gas and oil prices, over time, and he is also correct in challenging the model used by Hayes to identify a reduction of $0.89 to $1.09 on gas prices because of higher ethanol production and higher crude oil prices.
Hypothetically, in isolation from other variables, the higher the crack ratio, the higher the price of gasoline. Further, if the price of ethanol is relatively low or on a downward trend, increased use of ethanol in gasoline blends, in theory, would cause the crack ratio to go down and the spreads to be higher, assuming gas prices remain the same or increase. Good news for consumers! Right? Maybe? Not always? Not at all? Not sure? What if?
I cannot claim real modeling expertise and would not, even for a minute, arbitrate between Knittel and Hayes concerning their use of models and its result — in terms of Hayes, significant impact of ethanol, in terms of Knittel, minor impact of ethanol.
But in terms of the policy argument between them, I suspect Knittel comes out the winner (full disclosure: I did graduate from MIT and while I love Iowa’s rolling hills, I do not like the climate and the fact that the state does not have a great symphony, nor a NFL football or American League baseball team). He points out that the crack ratio’s fluctuations in the ‘80s occurred when oil prices both declined and increased. Ethanol was not a factor and the movements in the crack ratio were not based on ethanol production. He seemingly, correctly, faults the folks in Iowa for not using the crack spread model in their 2011 and 2012 papers to evaluate the impact of eliminating ethanol because the two models —crack ratio which they used and crack spread which they didn’t — produce significantly different results and policy implications.
What does the dispute over models and model use have to do with public policy? A lot! The ethanol supporters touted the Iowa studies to support their claim that increased ethanol use reduces costs to consumers in a major way. Conversely, the ethanol critics suggest that the Knittel analysis debunks the assertion that use of ethanol as a blend will reduce gas prices in a major way.
Knittel suggests the Iowa studies vastly overstate the cost-related benefits of ethanol to the consumer and that Iowa’s model disregards or blurs the effect of price changes and swings in price of both ethanol and oil. Knittel also indicates that that the relationships between oil and gas prices, as well as oil, gas and ethanol prices are much less precise and more complicated than indicated by Hayes’ modeling efforts. Prices of all three fuels are much more subject to behavior and external events than acknowledged by either Knittel or Hayes.
The dialogue between Knittel and Hayes is helpful in sorting cost and price issues regarding ethanol and gasoline. I hope they continue at it, with less emotion, and with analyses better grounded in methodological analyses that generate a better job of linking model building with experience and empiricism. Meanwhile, no matter whether you believe the effect of ethanol on gas prices is high, moderate or low, if the U.S. government acquiesces in the use of higher ethanol blends like E60 and E85, and if the cost spread between ethanol and gasoline continues, an increasingly visible positive impact on fuel prices will likely be witnessed at the pump. Apart from any possible price differential related to use of higher blends, increased use of ethanol as an alternative transitional transportation fuel is in the public interest. According to most reputable studies, such use will respond well to many environmental problems caused by gasoline and it will help reduce America’s need to import oil…a continuing security problem.
Epilogue: I once taught a reasonably popular class on policy development and models. To liven up the class, I told the students that economic and policy models are abstractions of reality and to the extent that the models’ abstractions helps students understand reality, they are “good” models. They asked for examples. It was a late evening and I was tired. I told them to go look at the centerpieces in Playboy and Playgirl. Both presented models of airbrushed men and woman. At our next class, I asked the students if the models increased their understanding of men and women. They were bright and eager students, at least for this assignment, and they indicated, “No.” The models tilted too far toward abstractions and too far away from real world experience. They seemed to learn a lesson about the value of at least some models.