Better Late Than Never: The Explanation for a Puzzling Pattern in Gas Prices

More than two years ago, I blogged on this site about a puzzling pattern that I had observed in gas prices. Here is what I wrote:

According to U.S.A. Today‘s “Weekend Gas Gauge,” the current average price of regular gasoline is $4.026 per gallon. A year ago the price was $2.945.

What I hadn’t realized was that the gap between regular gas and premium gas has also increased. Right now, premium gas is 38.5 cents per gallon more expensive than regular gas. A year ago, that gap was 29.5 cents. (In percentage terms the gap hasn’t grown, but it seems to me absolute differences – rather than percentage terms – are the right way to think about this problem.)

Why should premium gas be relatively more expensive today?

I went on to offer a few possible economic explanations for that phenomenon, none of which were particularly convincing. I then opened it up to blog readers to see if they had any better suggestions.

Remarkably, I got an email last week – 30 months after the original post – from an economist at Ohio State named Matthew Lewis, who had finally uncovered the true answer to the puzzle.

Here is what Matthew had to say:

While working on a research project I uncovered the answer to your question. Turns out it is not some strange economic phenomenon, but rather a data error on the part of AAA!
The USA Gas Gauge which you get your prices from actually uses the prices quoted by the AAA website. I have the historical data from the AAA site, and it turns out that they only accurately quote regular grade prices. AAA apparently calculates their midgrade and premium prices using a fixed and arbitrary percentage markup over regular. The markup they have used over the last 5 years is 10% (there is slight rounding error). Of course this artificially generates an increase in the absolute cent/gallon markup when prices rise, which is exactly what you notice in your column. The AAA site is simply providing bad data on midgrade and premium prices. If you look at other sources you find that the absolute cent/gallon markup in fact has not changed much even as the price level has fluctuated. There is still a general rule in most places where midgrade is 10 cents higher and premium 20 cents higher than regular.

Matthew’s answer to this puzzle led me to a couple reflections. First, I was pleased that my basic intuition that the price gap should be constant as prices changed was confirmed. I like it when I see some data that make me balk, and it turns out there was a good reason to balk. Second, it points out the dangers of a discipline like economics in which there are “many degrees of freedom,” i.e. where there are lots of different possible explanations floating around. Usually that means that an economist can come up with some rationale to explain just about any outcome imaginable. Often it won’t be the right explanation, but absent some way to disprove it, many people can be convinced it is the truth.

Lady Gi-Go.

Garbage in, garbage out.


At first I wondered why the rate of premium gas shouldn't be a percentage (say 110%) of regular grade instead of the fixed margin, but then I realized - it costs just as much more to create the premium gasoline regardless of how much the barrel costs!

Zach St

The closing reminded me of a story that gets passed around: Schmidt & Hunter, eminent personnel psychologists, had a room full of psychologists at a conference searching in vain for moderators in randomly generated data. The attendees, all with considerable statistical training, found a lot of patterns. Even with training, people have the "wonderful" ability to see patterns in all manners of randomness.

David Cawlfield

As a chemical engineer, I'm initially a bit surprised that the premium grades of gasoline sell at a constant adder over regular. I would intuit a percentage increase because the premium grade would intuitively require more crude oil to produce. However, the increased octane rating of fuel is mostly accomplished by a process called alkylation. Alkylation consumes some sulfuric acid but very little additional crude oil. In fact, alkylation in combination with catalytic cracking is a means to synthesize more gasoline from a barrel of crude. Sulfuric acid prices have remained very low for a long time, and the labor and fixed costs associated with alkylation are fairly constant. In my opinion, the fixed gap you found reflects the fact that the added refining costs associated with high-octane are relatively independent of the price of crude oil.

Peter Parker

I work as a business consultant, and learned as a rookie that if you have a set of numbers, and a "desired result" (desired by the client, or your boss) then there is ALWAYS a way to get the desired result from whatever the set of numbers. As a corollary, there is ALWAYS some way to explain the desired results from any given set of numbers.
This has of course turned me into quite a cynic with regard to most studies - the first thing I do when I see the results of a study is try and find out who paid for it.


Re #4: "...the premium grade would intuitively require more crude oil to produce."

But intuition is wrong. Generally the octane rating (which is what distinguishes regular from premium) is increased by the use of additives, which often aren't derived from petroleum at all. For years the commonest additive was tetra-ethyl lead (thus high-octane gas was commonly called "ethyl"); nowadays it's often ethanol.

Fritz Mills

It seems to me that at the gas stations I go to in the Chicago area, the "premium" for premium is drifting towards 25 cents above regular, while the mid-grade is still 10 cents above regular. I've also noted that when gas prices spike (and folks start substituting down), the premium and mid-grade "premiums" get reduced.


This is similar to the Congressional Budget Office's claim that the Economic Stimulus Package was successful based not on facts, numbers or surveys but based on theoretical multipliers.

"CBO used evidence from models and historical relationships to determine estimated "multipliers" for each of several categories of spending and tax provisions in ARRA ... Each multiplier represents the estimated direct and indirect effects on the nation's output of a dollar's worth of a given policy. Therefore, a provision's multiplier can be applied to the budgetary cost of that provision to estimate its overall impact on output."

John Squire

According to my anecdotal observations, many retailers have adhered to the $.10 and $.20 increments for 89 and 91/92 octane (93 seems exclusive to Shell). The highest priced retailers that I pass (frequently $.20 over other locations) have been running % markup since Katrina. Either they set price based on hyper-local demand, or they use inputs like AAA that assume percentage delta rather than absolutes.

Charlie B


Christmas is coming and the American greed systems come into play.


Are economists responsible for verifying that the data they're using is accurate?


Yes, now can you please tell me where i can pay exactly the price listed on the gas station board... why am sold a product at a listed price that cannot exist i.e. 3.29 9/10.

Jim Christian

Premium costs more to produce, as David Cawlfield (#4) pointed out. Exactly how much more is complicated, but is likely to include additional processing costs for alkylation or reforming and additives. However, there is no reason to believe that the price differential at the pump is the same as the cost differential. Cost is only one factor in setting prices.

You ought to be able to tell if the price differential is based on cost or on customer psychology by looking at the variability in the price differential. The extra costs of making premium should be similar from year to year (with seasonable variations) and the demand should also be similar. (Demand is based on the number of cars that require higher octane.) Low variability in the price differential is likely to indicate that pricing is based on costs. High variability in the price differential (after accounting for seasonal affects) would suggest that customer psychology is more important than cost in setting the price. For instance, is the price of premium always 10 cents more than regular? Or does it always end in 9 cents regardless of the price of regular?



Or should Steven Levitt instead be disappointed that he didn't take the time to look at more than two or three numbers in the price series before posting about this gas "puzzle" in the first place? If the author had taken a little time (30 seconds? 1 minute?), to calculate the differences in price at a few points in time himself, he surely would have seen premiums of exactly 10% in each instance and realized immediately that there was some type of issue with the data.

Instead we get a blog post that was lazily constructed and ultimately a waste of time. I hope this doesn't reflect on the care and quality of the author's work more generally when the mistakes aren't as obvious and can't be corrected by readers....

Elisa Miller

That is very interesting, except that here in Dallas, TX, we've seen the gap between regular and mid-grade go to $.15 and another $.15 for the premium. Sometimes the prices are all over the place. Costco and Sams only carry regular and premium and their difference is between 20 - 25 cents.


Here's what I want to know: why are gas stations allowed to charge more for credit card than for cash? Can you imagine if other businesses did this? (Of course the credit card industry would never allow it) Given the price of gas, it is not practical to have cash on hand. And the extra fees they collect are way higher than the credit card processing fees they are charged, and for which businesses are given a tax deduction for. So why are consumers being exploited this way and why is this particular business allowed this practice? This really, really bugs me!

Homer Sung

What caused the spike of crude oil price in 2008?

Chad Freckmann

Does anyone know of a site that identifies make/model of vehicles that need more than 87 octane to run?

Jaguar/Ferrari and some high performance engines. But don't the majority run on regular?


I was hoping that this "better late than never" heading was finally going to relate the fact the the cost of fueling an automobile is far less, than it was 55 to 60 years ago.

A Chevy cost around $2,200 in the 50s. The average annual income was around $3,500. A gallon cost around 30 cents. Nowadays those numbers have all increased 10 fold.

That would suggest the cost of fuel has remained the same. However, that cost is actually almost half of what it was. Because the average Chevrolet back then, got around 10 miles a gallon. Nowadays the average mileage is over 20 mpg.

In short we now drive twice the distance on the same tank of gas. That means, in terms of the 1950s we are now enjoying 15 cent per gallon gasoline.


We waste billions of dollars blowing stuff up and killing people, and we can't even get cheaper gas than this? what's wrong with this picture?