Gas Stats

Over the last few months, gas prices having been receiving a lot of attention. According to AAA, gas prices reached an all-time high on July 17 this year at a national average of $4.11 for regular unleaded. It's been talked about as one of the most important issues of the upcoming election. PollingReport.com has a lot of information on various polls conducted by various groups. This particular page gives detailed polling information regarding the priorities of the voters. As you peruse, you can see that "the economy" is generally number one. "Energy" and/or "gas prices" also show up quite a bit. I'd expect when not specifically called out, people think of gas prices and the economy as quite similar. This set of polls is directed specifically at energy issues and includes people's opinions on what is causing the rise in gas prices.

In discussing the relationship of the dollar and oil prices, Ron Paul (R-TX) had this statement before the House Financial Services Committee. In summary, Paul suggested a strong link between inflation and oil prices. I've been thinking about this for a while myself and was curious to see how much of a relationship there actually is between gas prices and some measure of inflation. I also seemed to recall new stories in the past that gas prices increase in the summer which makes since due to increased travel in the summer vacation season. This article from 2005 corroborates my recollections as well as basic supply and demand economics.

So, I decided to build a simple statistical model using regression. Basically, this model takes certain variables as inputs and tries to predict the value of another variable. I decided to build a model that would try to explain (or predict) gas prices based on "inflation" (more in a moment) and the month of the year (to pick up seasonal travel patterns). To capture the impact of inflation, I decided to look at the money supply adjusted for the strength of the U.S. Dollar. There is significant debate amongst economists in discussing the link between inflation and the money supply. There's plenty to read about the thoughts of the monetarists, Keynesians, and the Austrian School economists which I do not intend to get into here. For the money supply, I chose to use M2 (read the money supply article if you care to understand) and sourced the information straight from the FED. I wanted to use data from Shadow Stats, but didn't want to pay the subscription to get the data. The FED is nice enough to make the money supply data available for free. Additionally, I used the FED's own trade-weighted dollar index (I used the "major currencies" index) which tracks the value of the dollar relative to a basket other world currencies. Finally, my gas prices came from the Energy Information Administration - yet another division of the federal government (at least the data is free).

The following charts summarize my process and results:

1. This first chart shows the price of gas (regular grade, nationwide average) from January 1991 until July 21, 2008. As you can see, and already knew, gas prices have seen a sharp increase beginning sometime around early 2002. There is also a little bit of an increase right around 1999, but it reaches a plateau and dips again before the big climb.

2. The next chart shows the M2 money supply in billions of U.S. Dollars. Again, you can see an increasing trend. This one begins around 1995 and rises at a steady pace thereafter. I've just spent more time than I'd prefer at this moment trying to research the cause of such a shift. I had little luck. There are several essays from branches of the FED discussing the "breakdown" of M2 predictability and its impact on macroeconomic indicators around this time. Also, generally speaking, the FED itself can increase the money supply by lowering interest rates and increasing its purchase of treasury securities.

3. The next chart show the relative strength of the U.S. Dollar over the same period of time. This should also show no surprises, at least for the most recent period, as we've heard a lot about the weakness of the Dollar. As mentioned earlier, this data comes from the FED. There may be some argument to say that the weakness of the currency may be due to the increased money supply. However, I argue that we need to look at an adjusted M2 for currency strength to indicate the relative purchasing power of the Dollar through time. Comments/criticism welcome.

4. This chart now plots the adjusted M2 against gas prices over the period of study. The adjusted M2 is calculated by simply take M2 from the second chart and multiplying by the Dollar Index in the third chart and dividing by 100 (since the index is scaled by 100). M2 is on the left Y-axis and gas prices are plotted against the right Y-axis. As you can see, there is a strong relationship between the growth of the money supply and gas prices. This can be explained two ways. First, you could argue that M2 drives or follows inflation, and thus prices. On the other hand, you can argue that the money supply is representative of aggregate demand in the economy (remember, demand is measured by price - the more money that exists, the more people can pay).

5. Now, this chart shows the performance of the simple statistical model which attempts to predict prices based on the adjusted M2 and the month of the year. The model is pretty accurate. It has an R-squared of 0.898 which means that the model explains about 90% of the variability in gas prices. That's pretty good. Additionally, almost all of the variable have strong explanatory power. Generally, every one billion dollars of adjusted M2 yields an increase of three cents per gallon. The model also says that gas prices are highest from April through July as we would expect. One thing that may be apparent is that actual prices are much higher than predicted for the last several observations (blue line is higher than the red line).

6. My final chart displays the error of the statistical model. It is expressed by the percentage variance between the predicted price and the actual price. Positive values indicate that the actual price is higher. The interesting point is that the model consistently predicts prices within a +/- 20% variance. This probably doesn't seem that good, but it's not too bad for a quick and dirty analysis. Also, note that there are periods of persistent over-estimation and under-estimation. This indicates that the model could be improved with other data that reflects these impacts. But, remember, we are already explaining 90% of the volatility.

What's the point of all this?! We should expect a decrease in gas prices. This is for two reasons: first, as we exit July and enter into August, gas prices do tend to decrease (roughly four cents until October where it is a more pronounced decline); second, we are in a period where gas prices are higher than expected based on the money supply. What's causing this? There are a lot of opinions out there: speculation, demand from growing economies, supply constraints, and geopolitics are the main reasons that people discuss. Certainly each of these play a role, some more than others, but simple inflation and seasonality explain much of the issue. Here are few other interesting articles for further reading.