Bryan Caplan  

I Win My Long-Run Gas Price Bet

Caligula's Wish... Searching for the Individual...
In July of 2008, the average U.S. price of regular gasoline was $4.062.  As usual, global hysteria followed.  And as usual, I was unperturbed.  So unperturbed, in fact, that I made the following bet with Tyler Cowen and David Balan:
I will bet $100, even odds, that the U.S. price of gas (including taxes) in the first week of January, 2018 will be $3.00 or less in 2008 dollars.
A subsequent clarification specified that the bet was on the price of regular gasoline.

Today, the January CPI arrived, allowing us to finally resolve this ten-year bet.  In 2008, the US CPI stood at 215.3.  In the third quarter of 2017, it hit 244.7.  Since then, there has been further inflation of 0.3%, bringing us to 245.3, for a grand total of 13.9% inflation during this period.  For me to win, then, the average price of regular gasoline in January 2017 must be less than $3.417.

So where are we now?  In January of 2018, the average price was a mere $2.555.  I have therefore won this bet by a margin of over 25%.  (Indeed, even if we count all gasoline, the average price is only $2.671).  I would have prevailed if there'd been 0% inflation - or as much as 14% cumulative deflation.

Now what was this bet really about?  Back in 2008, Tyler argued that there was a deep intellectual error in the work of Julian Simon:
It's amazing how much, on this issue, some people resort to what can only be called technical analysis -- inferring future price movements from past trends -- when they would scoff at that approach in almost any other context.
Why was Tyler so dismissive?  As far as I can understand, he thinks that basic arbitrage theory implies that there can't be predictable price patterns.  My view, in contrast, is that there are many conditions that undermine the power of arbitrage.  So if we see a 150-year price pattern, it's perfectly reasonable to expect it to continue. 

And what is that pattern for gasoline?  For as long as we've had data, gas prices have shown frequent spikes, followed by gradual declines back to long-run trend.  So when prices spiked to over $4.00, I expected the past to repeat itself.  And repeat itself it did.

I expect that Tyler will insist that I just got lucky.  And if I lost roughly half my bets, that would be a wise reaction.  However, this latest victory brings my betting record to 17 wins and 0 losses.  Yes, pride goeth before the fall.  There's at least one outstanding bet that I now expect to lose.  Still, the only reasonable explanation for my 17-and-0 record is that my judgment is exceptionally good.  As I've said before:
[T]wo key practices account for most of my success.

1. I take the "outside view."  When predicting, I start with long-run averages, and presume the "latest news" is distracting trivia.  For example, when I made my unemployment bet with Tyler, I looked at all the unemployment data for 1948 to the present, and assumed the future would resemble the past.  As usual, it did.

2. I spurn hyperbole.  Human beings adore superlatives, but superlatives rarely apply to the real world.  So when I notice someone treating hyperbolic poetry as literal truth, I rush to wager.  For example, when John Podhoretz asserted that the Iran nuclear deal "effectively ensures that it will be a nuclear state with ballistic missiles in 10 years," I smelled hyperbole, and tried in vain to entice him put some money on it.

In slogan form: I owe my track record to numeracy and normalcy.  Step back, calm down, look at the numbers, and target thinkers who say, "This time it's different."
Last point: As always, my opponents have my respect - and deserve yours.  They stuck out their necks and made clear claims.  If every pundit would do the same, this would be a far better - and far quieter - world.

Update: David Balan asked me to post this comment:
I have paid Bryan the $100 that I owed him from losing this bet. And not only did I lose, I kind of "extra" lost, by which I mean that when I made the bet I thought I was very likely to win, and that it probably wouldn't even be close (secular increase in demand from economic growth in China and other developing countries, etc.). So a very clear "W" in the record books for Bryan. (This is in contrast to the other bet that I lost to him, where I think the W should have an asterisk next to it.)

COMMENTS (11 to date)
Blaise writes:

What about your bet on climate change with Yoram Bauman? I agree there's still a long time to go but I will be happy to bet you will lose that bet.

You have a very good betting record but as a forecaster you can't be taken seriously if you win relatively low stake bets such as the price of gasoline but loses one that could be catastrophic for the world. You need some sort of weighting systems for the social relevance of your bets. As a neutral observer, I prefer to follow 17 exaggerated claims and overreact in order to avoid one catastrophic possibility. That's the limit of relying on historical averages: you will not be able to forecast radical changes.

Still I like very much your approach as I also feel people love exaggerated claims and media amplify them. Maybe bets should be based on some sort of credibility points instead of money as amounts are often too low to matter. A pundit who has lost all his credibility points shouldn't be given a voice anymore.

Living in the UK and following Brexit quite closely, I think you still have a decent chance to win your bet. Maybe 25%.

More here:

Donald Strong writes:

Eliezer Yudkowsky writes:

I continue to superstitiously pray for your victory in all your bets, since one of them was important. You will of course lose your global warming bet, but protection lasting until then would still be nice.

John M Hall writes:

Julian Simon wasn't making an argument on technical analysis. Rather, his argument was that when relative commodity prices are high, then humans will have some ingenious solution to increase supply, which will push prices back down. It's classical supply/demand analysis, not technical analysis.

You were making effectively the same argument, except there were two ways you were right: the market had forecasted stronger demand growth than occurred and it forecasted less supply growth than had occurred. Oil prices had been bid up based on these assumptions. In this sense, the price of oil in 2008 was efficient, conditional on the information the market had. However, when those information turned out to be wrong, prices quickly adjusted.

Rob writes:

What do you think the price of Bitcoin would be in five years? (At what price would you be equally happy to have the "under" or "over" of an even-odds bet?)

Mark Bahner writes:


Congratulations on this win. In my opinion it was spectacular. Not simply that you won on an *even odds* bet that the price would be less than $3--that's spectacular enough--but you won easily even if the price was not adjusted for inflation. That is really remarkable.

The long-term likelihood of gasoline substantially increasing in price seems very small. (To me, now. I certainly wouldn't have seen it in 2008.)

Fully autonomous electric vehicles operating as transportation-as-a-service seem to be the future of transportation. So that will come close to ending demand for gasoline. (But again, there's no way I would have predicted that in 2008...just like there's no way I would have predicted that the U.S. would be producing more oil in 2016 than in 1966.)

Again, congratulations on this spectacular win.

Steven Ehrbar writes:

Outside view has its merits; so does looking at the actions of the people with the best inside view. I yelled at a lot of self-assured "peak oil" people ten years ago to count how many major oil companies were building production coal-to-liquids plants like Sasol's Secunda.

After all, at the price of crude in 2008, Secunda was reaping huge operating profits turning coal into liquid fuel. If the reason oil prices were so high was we were running out of petroleum that could be extracted at a reasonable price, then with the world's huge proven coal reserves, coal-to-liquids would be profitable for a long time to come. On the other hand, if the reason oil prices were high was transient, building coal-to-liquids facilities would be a bad long-term investment, because they have huge capital costs that couldn't be recovered after petroleum prices came down.

The major oil companies have the money, institutional links, and incentive to hire the best petroleum geologists in the world and gather the best worldwide oil field data. They also have the incentive and money and structure to acquire the best data on other parts of the market conditions (demand, etc.) for petroleum products and their feedstocks. They had the best inside view of the conditions involved.

And you know what? None of them were doing more than small pilot programs in coal-to-liquids technology. There were also some small startups doing their own pilot programs, backed by the occasional small military contract on guaranteeing strategic access to liquid fuel and the like. But nobody in the industry, anywhere in the world, was building a new Secunda.

Shane L writes:

It's been a few years now since I've read it, but it strikes me that Bryan's approach is more or less opposite to that of Nassim Nicholas Taleb's The Black Swan (2007). There, if I remember correctly, he said that projection of past trends into the future was unwise and vulnerable to essentially unpredictable shocks. I wonder if Bryan has read the book and has any thoughts on this.

(Also, I'm really enjoying these bets and salute all participants!)

mariorossi writes:

If I look at my bloomberg terminal, the DEC16 oil contract was trading pretty much flat to spot in july 08 (and slightly lower than nov '08 delivery). If I look at 10y breakeven inflation, it was around 3.0 (it spiked at the same time). So real oil price for july '18 was priced at about 33% lower than jul '08 delivery. I don't think I have a similar number for the RB contracts (the gasoline future), but I doubt it would be very different...

That seems pretty close to the terms of your bet. Doesn't this mean you bet at 50/50 odds that the price would be lower than market forward price?

Congratulations for winning your bet, but I don't quite think this is an example of an outside view. The oil market pretty much agreed with you...

Thomas writes:

@Shane L

It's been a few years now since I've read it, but it strikes me that Bryan's approach is more or less opposite to that of Nassim Nicholas Taleb's The Black Swan (2007). There, if I remember correctly, he said that projection of past trends into the future was unwise and vulnerable to essentially unpredictable shocks.

For the most part this principle doesn't really apply to these types of bets. You generally don't want to be all in on a set of investments that completely collapses in the event of a once in a century type event, but for an even (or near even) money type bet for low stakes such as these, the worry of an outlier event is pretty minimal.

Eric rasmusen writes:

There are pretty much three kinds of Time series: mean reverting, random walks, and linear trend. You can do a lot with those three ideas.

Comments for this entry have been closed
Return to top