GDP & Oil Prices

Tue, Mar 1, 2011 - 1:45am

What a month it has been in the oil markets! Brent crude prices accelerated their price trend, breaking above their trading band, while WTI broke $100 for the first time since 2008. The price increases were due to the market’s indigestion of Middle East riots, civil war and regime changes. At the same time there was an unprecedented divergence between Brent (Scotland) and WTI (Cushing, Oklahoma) crude as a result of a surfeit of crude supplies flowing into Cushing and the mid-American PADD 2.

Worry that high oil prices would derail the recovery was a common theme in the financial media. While this is a valid and important debate, chances are good that prices are not yet to the level necessary to have a marked impact. That being said, the instability in the Middle East is one more straw on the back of a heavily laden camel (otherwise known as the US economy).

I spend a lot of time talking about oil, both at work and in my personal life. Sometimes people are amazed at how much I focus on oil, but that amazement is equaled by my amazement at their disinterest. Since oil is an integral part of 99% of modern Americans' lives (perhaps the Amish and a few survivalists aren’t dependant), from how we get our food to how we move around in the world, an understanding of this issue is of paramount importance! It really is more important than Charlie Sheen’s latest rant, even if it isn’t as fun to think about in the moment.

Part of this importance comes from its effect (or lack thereof) on our overall economy. So we return to this question: are oil prices high enough now to derail a recovery that finally seemed to be gaining some steam. My short answer is a qualified no, but with the stressed caveat say that the price action is disturbing to me, and it is easy to imagine a price move that would create economic recession, if not downright chaos.

The first important thing in deciding this question is to look at oil’s price advance and compare it to previous price advances. Since WTI crude is in its own little world, I’ll go with Brent crude prices. We’re actually looking for the refinery acquisition cost which is an average of a whole bunch of different crude streams, but Brent is available in real time, so we’ll use that number for the most recent price moves. In the past week, Brent moved up $9.60 or just over 9%. It closed last week roughly 31% above where it was one quarter ago. Those are both impressive metrics. However, it is still below the price reached in 2008 by roughly 22.5%. This is important because consumers and businesses have a memory of what they have paid in the recent past, and most can still recall gas prices in 2008 that were higher than they are today.

There are many ways to look at oil shocks. The simplest way is to say an increase in oil prices hurts the economy and a decrease in price helps it. But it is easy to poke holes in this simplistic explanation – for example if prices went down to $1/barrel in April, and then increased to $2/barrel in May would you say the increase that occurred between April and May was going to hurt the economy? Clearly not, since we are used to paying so much more. Additionally, as much as it would be fun to fill up at the gas station for a shiny quarter, businesses would undoubtedly feel uncertain about the chaotic movement and may well shelve investments and plans until they got a better feeling for where things were going. Evidence for falling oil prices helping the economy is mixed at best. To hurt the economy, it seems to be necessary for the price to be “uncomfortable.” And since what is “uncomfortable” is arbitrary depending on who you are talking to, it makes the most sense to define it in terms of what people have paid in the recent past. There is also a good argument to be made that the higher the percentage of world GDP is dedicated to oil expenditures (or of an individual country’s) the greater effect it will have on the economy if prices then go up. Again, to take an extreme example, you might have “gotten used” to paying $1/gallon in the late 1990s, but oil was such a small part of GDP at that point, that the tripling of oil from $10 to $30 (and gas to $1.50) probably didn’t hurt the economy that much.

To summarize, I would argue three things about the recent oil price increase:

  1. Prices are still below their 2008 apogee and are therefore unlikely to have a “shock” effect to American businesses and consumers. [See Figures 1 and 2 for diagrams of my preferred shock variable, and its predicted effect on the US economy. Note its current value is zero.]
  2. However, the US is dedicating a relatively large percentage of GDP expenditures on crude. Therefore we are vulnerable to a large price spike if it were to come. [Figure 3]
  3. Prices have been advancing quickly, particularly in the last week. Given the already high price of oil, this rapid increase is surprising, and reminds me more of the price action prior to the first gulf war, than to geopolitical events in the last decade. If prices double or triple from here, the economy will probably be in trouble.

Preferred Oil Shock Variable

Figure 1: This funny looking graph shows the oil shock variable that I most prefer, using an expenditure-weighted variable that only takes a value when oil prices are at a new three-year high.

Predicted Drag On GDP From Oil Shocks 1953-Present

Figure 2: This graph shows the predicted negative effect on GDP (subtracting from trend growth) from the same model.

Crude expenditures as a % GDP

Figure 3: Expenditures are elevated as a % of GDP, at levels similar to the mid 1970s, and more recently, 2007. We could expect any oil shock (were it to occur) to have a large effect. Any price above $150 would most likely create a reaction in the economy.

Econometric tests have shown a mixed relationship between oil price shocks and recession, but the best models (which have quite robust results in terms of information criterion) specify that an oil shock must have a price that exceeds the previous three year high. James Hamilton’s more recent work is an example of such – he sets as a pre-requisite for an oil shock that prices must establish a quarterly high which is higher than the previous 12 quarters (additionally shocks are treated as asymmetric so a fall in the cost of oil returns a value of ‘zero’ for the oil shock variable). I’ve developed a similar model which also takes into account the percentage of GDP spent on oil. In either case, the quarterly price of oil would have to exceed $117 before the shock variable registered a positive value for Q1 2011!

Certainly it can be argued that the 3 year time horizon for a price shock is a bit arbitrary – the underlying idea is that people form expectations and habits based on the recent price of oil. The specification could be improved by having a logarithmic decay to represent people’s memories and habits. (People wouldn’t regard $100 a shock if it was $125 yesterday, but surely they would if it was $125 20 years ago and then had been $10 in the intervening decades.)

Finally, to argue the oil/GDP point a bit on a theoretical basis (and create the mother of all run-on sentences): whether oil prices are a direct determinant of GDP, whether they act through some intermediary mechanism, whether they are the result of another mechanism which also predicts GDP, or whether the relationship to GDP is purely spurious, econometric results clearly indicate that you cannot discount the possibility that oil price spikes (as defined by Hamilton) do negatively affect GDP. The Granger-Causality numbers on this are nearly unity. And interestingly, oil prices as a determinant have gradually supplanted interest rate term structure as the GDP determinant with the highest information criterion.

Obviously a surfeit of data underlies our inability to make a more conclusive determination, and Murphy’s Law dictates that by the time the determination is made the relationship will have likely changed…a big reason why I was not attracted to econometrics as a field!

All the Best!

About the Author

Director
cello2econ [at] gmail [dot] com ()