Volatility in the Price of Oil since Hubbert's Peak and Investment Risk

IS VOLATILITY IN THE PRICE OF OIL A CAUSE OF THE BAD BEHAVIOR OF THE FINANCIAL INDUSTRY ?
The former chairman of the US Federal Reserve Alan Greenspan provided the following response during questioning about his “ideology” before a committee of the US congress on the 23rd Oct 2008:

Greenspan said: "I have found a flaw. I don't know how significant or permanent it is. But I have been very distressed by that fact...."

A reflexive media and commentators jumped on a bowed Greenspan for the usual depressing reasons. However, is it possible that Mr Greenspan had let his mask slip for a moment. The part of his sentence that caught my attention was “…how significant or permanent…”. Re-reading his testimony it occurred to me that Mr Greenspan was not saying that he’d been wrong. Instead, I believe he may have been implying that conditions in the economy had changed unexpectedly and fundamentally in a manner that now made his old way of thinking flawed.

This idea attempts to explore the basis of Mr Greenspan’s distress. In short, it proposes the skeleton of a mechanism for how volatility in the price of oil might explain the recent unfortunate actions of the financial industry.

I disclose that I am not a professional economist or mathematician, though I am a working science professional and do have familiarity with basic statistical methods.

Consider the following hypothetical sequence of events:

1. Global oil production peaks around the yr 2000 as predicted by Hubbert.

2. A new pattern of volatility in oil prices emerges shortly thereafter and this pattern continues to build through the present (and into the future).

3. This type of volatility is in the nature of being on the production down slope of a finite resource in great demand (i.e., oil). Variance and occasional very large movements were certainly evident in the upslope of the production curve. However, the frequency and severity of oil price movements that characterize the new pattern on the downhill side of the production bell curve are envisaged to be of a different order of magnitude.

4. Owing to the singular role of oil in the economy, volatility in its price begins to propagate in variable degrees into the volatility of the price of nearly everything else.

5. A general increasing volatility in prices translates into increases in risk of investment - indeed, due to unknowns in future prices and most especially of oil itself, real financial risks across the board are probably rising exponentially relative to price volatilities.

6. A smart, knowledgeable and initially small number of financial insiders begin to intuit the implications (as outlined broadly in 1 through 5) of being on the down slope of the global oil production 4-5 yrs earlier than the rest of us. These people don't need to know each other, but do need to have specialized knowledge and be capable of uncommon insight. Certain extraordinary minds may even foresee the unwelcome consequences of price volatility for risk prior to peak oil in yr 2000.

7. With financial risk increasing, a still small, but growing number of insiders coming to understand what is going on, the incentives within markets shift rather quickly from protecting shareholders interests, to figuring out how to quickly “cash out”. This shift in incentive is the fundamental change (of uncertain permanence) that may be the real cause of Mr Greenspan's distress.

8. The behavior of the primary few spreads within the financial industry. Most of these secondarily affected individuals are probably oblivious to the ultimate cause (i.e., oil price volatility) of their actions. A fin de seicle ethos is pervasive. There is an unspoken or perhaps even quietly discussed urgency that you should make your “nut” and get out. Greed, self-interest and/or stupidity do the rest.

How the "cashing out" occurred is where my idea goes somewhat fuzzy. But the story does seem to conclude with tens possibly and hundreds of trillions dollars in worthless assets in the "shadow banking system". We know the rest - credit crisis, stock market crash, bailouts and a looming severe recession.

The hypothetical mechanism posed above imagines an ultimate cause for our financial disaster beyond ordinary human failings such as greed and selfishness. Human frailty is a necessary, but not sufficient factor. It also attempts to provide a tangible nuts and bolts mechanism and this is a weakness. On the one hand, the workings of complex systems are ineffable. However, the butterfly in South East Asia still has to beat its wings to provoke that storm 2 years later in North Carolina. There are always ultimate causes for events in the natural world. This we have learned. The idea does not require an organized conspiracy to work. Simply put, it poses that increasing volatility in the price of oil following its global peak in production as a necessary and sufficient factor. All else, bad behavior of the financial sector included, self organizes and flows downstream from this.

In the subsequent discussion of data I will refer to the sequence 1 through 8 above as the “hypothetical sequence”.

Data and Analysis

I posted the hypothesis on blog sites on the internet late last year and earlier in January 2009. However, it quickly became clear to me that a fundamental problem was lack of data. A good first step seemed to be to determine the pattern of volatility in oil prices over the last 20 or so years surrounding Hubbert’s predicted peak in global oil production of year 2000.

I found getting hold of a free source of historical oil pricing data on the internet a surprisingly difficult. There appeared to be packages that could be purchased, but eventually I came across what seemed to be a reliable and free data base at the Illinois Oil & Gas Association website.

http://www.ioga.com/Special/crudeoil_Hist.htm

This site provide monthly data from the mid-1980s up until the present on the “HISTORY OF ILLINOIS BASIN POSTED CRUDE OIL PRICES”. Not perfect, but a start.

The chart in Figure 1 is a simple plot of monthly crude “oil price” over a period from 1986 to 2009.



The sharp rise and fall in oil in 2008 is a part of this story that is all to familiar to motorists.

My next problem was how to calculate an index in the volatility of oil price based on these data. I reasoned volatility should reflect the spread or variability in price over successive months. I am not a math professional, so I had to fall back on easy parametric statistics. I used Microsoft Excel to calculate an index as follows. The monthly oil prices for Jan, Feb and Mar of 1986 were $22.50, $16.00 and $14.00 respectively. So I first calculated the standard deviation (SD) of the first 2 numbers (i.e., $22.50 for Jan 1986 and $16.00 for Feb 1986) as an index of their spread. This SD was 4.60. Next, I calculated the standard deviation of the 2nd and 3rd numbers for  (i.e., $16.00 for Feb 1986 and $14.00 for Mar 1986) to give an SD of 1.41. I went on in this way for successive pairs of months for all 276 months down to Dec 2008.

The chart in Figure 2 plots the index of “Oil Price Volatility” in red in the left hand Y-axis along with the monthly oil price plotted in blue (now in the left hand Y-axis). Eyeballing this chart suggests that volatility is rising, particularly since 2002.



A few notes on Figure 2 and its underlying calculation. First, one has to worry about deviance from normality etc. The approach was not perfect. However, the aim was not to clear the forest, just to cut a path through it. For this purpose the approach seemed sufficient for now. Second, I understand variance (i.e., SD2) is the preferred method of estimating spread within a normal distribution. I did calculate variance and these calculations tended to accentuate the trend seen in red in the figure above. However, I decided that the simpler SD calculation did the job, so stuck with it. Third, I also tried calculating SD and variance for 3 or 5 successive months. The same overall pattern was evident from these calculations as in the figure above from the 2-month calculations. I stuck with monthly pairs as I surmised that what would be lost in noise by using the bimonthly figures would be made up with increased signal from a calculation with higher temporal resolution. Fourth, it would have been great to get my hands on daily rather than monthly data to do the calculations – but as mentioned this is all I could find for the moment on the internet. Finally, I thought about using coefficient of variance (i.e., average for the 2 monthly numbers over their SD x 100). However, I stuck with SD as I wanted the pattern to reflect absolute changes in the price of oil. This is the volatility that I surmised would propagate into the prices of other goods per the “hypothetical sequence” 1 through 8. And in turn the SD of oil price volatility (not COV of oil price volatility) should be eventually reflected in increasing investment risk per the initial “hypothetical sequence”.

Qualitatively, the underlying trend in oil price volatility over time in Figure 2 appeared non-linear. So I used Microsoft Excel to calculate a 3-factor polynomial fit to the scatter plot. This trend as represented in the pink non-linear regression line overlaid the “red” price volatility data in Figure 3. The R2 of the regression line is statistically significant and so on as provided in the chart. Interesting features of the pink regression line include that it starts to move notably from around year 2000 (i.e., Hubbert’s predicted production peak as in the hypothetical sequence). It also climbs in what appears to be an exponential manner from this time.



The trend in oil price volatility is isolated in Figure 4. From this it can be seen that volatility based on the calculation approximately doubled between yrs 2000 and 2004 and then approximately doubled again between 2004 and 2006.



I played around with the trend calculation, attempting to figure out when the line would indicate a significant uptick in oil price volatility. This was done by removing yr 2008 and then yrs 2008 and 2007 and then yrs 2008, 2007 and 2006 and so on from the regression line calculation. From what I could ascertain, the signal for the trend was faintly evident from the end of 2003 and continued to build over successive years from that time. Prior to 2003-2004 it would have been very tough to pick up the rising trend signal from the approach taken here. A time line of the trend becoming statistically manifest from around 4 yrs ago would fit with the suggestion of point 6 of the “hypothetical sequence”.

In conclusion, there does seem to be evidence for a quantitative rise in the volatility of the price of oil starting from year 2000 coinciding with the predicted global peak in oil production. How this volatility would propagate into the volatility of the prices of most other goods and services would take sophistication not at my disposal. Calculating the anticipated impact of this price volatility in other goods on investment risk is in turn a few orders of magnitude harder. However, intuition informs that given the singular and central importance of oil as the main source of energy to our economy, the causal relationships speculated upon should be in the offing.

Assuming that the mathematical mechanics of oil price volatility do eventually impact investment risk, troubling questions emerge. 1. Did certain individuals anticipate, intuit or use mathematical approaches to foresee this problem. 2 If so, did some of these people use this information to enrich themselves and/or influence the ethos of the financial sector wittingly or unwittingly…