Comments on: Fat tail of the day, oil edition

By: Beezer

Beezer — Sat, 07 May 2011 20:02:16 +0000

The margin requirements were raised, and speculators push leverage to the limit. They had to cover and cover fast.

This is the definition of speculation and in our current overly financialized world it happens far often than Gaussian math can accomodate. The earlier comment about Mandelbrotian math being more accurate is correct.

By: TFF

TFF — Fri, 06 May 2011 23:58:44 +0000

I was commenting on this move myself. No rhyme or reason, no news, nothing changed. Except suddenly the trading price of oil dropped by ten dollars a barrel.

By: xyz70

xyz70 — Fri, 06 May 2011 04:15:03 +0000

Eliminate the mortgage deduction, raise the retirement age for Medicaid and social security and institute an effective transactions tax.

By: djiddish98

djiddish98 — Fri, 06 May 2011 00:06:12 +0000

Chart link fail

http://www.gasbuddy.com/gb_retail_price_ chart.aspx?time=48

Just tick the crude box

By: djiddish98

djiddish98 — Fri, 06 May 2011 00:03:09 +0000

Does anyone know why retail gas prices are higher than 2008 levels when crude was ~115?

http://www.GasBuddy.com/gb_retail_price_ chart.aspx?city1=USA Average&city2=&city3=&crude=y&tme=48&uni ts=us

The graph is tough to pin down, but looking at the data point on 4/19, it seems as if gas was around $3.60 national average with crude trading around ~115.

Why the 8% increase in retail?

By: nedofbaker

nedofbaker — Thu, 05 May 2011 23:42:49 +0000

Felix,

I apologize that I only comment when I have a complaint because, honest, I enjoy reading your stuff!

But I second 3oosion above… Mandelbrot observed that financials don’t follow a normal/gaussian distribution. That was a few decades ago. I really don’t understand why otherwise sophisticated financial commentators forget this all the time.

And, really, going after speculators?

By: jmh530

jmh530 — Thu, 05 May 2011 22:05:20 +0000

On a one day horizon, the mean of most financial securities is close enough to zero that it won’t matter much.

By: david3

david3 — Thu, 05 May 2011 20:21:46 +0000

You are missing the mean in your std dev calculation. It was more than a 5.4 std dev movement.

This t-shirt should clear up your confusion.

http://rlv.zcache.com/standard_deviation _root_mean_square_of_values_tshirt-p2358 11735824455090trlf_400.jpg

By: GRRR

GRRR — Thu, 05 May 2011 20:14:50 +0000

Blame the quants and their equations.

By: CDN_Rebel

CDN_Rebel — Thu, 05 May 2011 20:09:57 +0000

Might seem crass to even present it, but I guarantee that conspiracy nuts out there will think this is Obama’s doing through his army of Goldman Sachs cronies all to make George Soros a few extra bucks or buy votes.

I’m pretty impressed by silver dropping 30% in a week because of a minor tightening of capital requirements myself…

By: jmh530

jmh530 — Thu, 05 May 2011 19:51:58 +0000

I have daily data going back to January 1986 for oil prices that show the standard deviation of log changes as 2.58%.

Another interesting quirk of statistical time series is time varying volatility (GARCH). Since I don’t have Matlab handy to do the calculation, I just took the rolling 20 day standard deviation. Most recently, I got like 1.66%, which is consistent with what you have. So over the whole period, volatility was a lot more than it has been just recently.

By: 3oosion

3oosion — Thu, 05 May 2011 19:37:15 +0000

1) Market movements are not normally distributed. This has been known for a long time. We get 100 year floods every few years.

2) No one has described a plausible theory as to how speculation can affect oil or other commodity prices unless there is physical withholding of oil from the market, in the form of high inventories. Supply and demand. For example, http://krugman.blogs.nytimes.com/2009/07 /08/oil-speculation/

By: OneEyedMan

OneEyedMan — Thu, 05 May 2011 19:34:27 +0000

That should be trading days in a century, not day in a century

By: OneEyedMan

OneEyedMan — Thu, 05 May 2011 19:33:46 +0000

The chance of a 5.4 standard deviation move is 3.326629e-8.
There are approximately 25200 days in a century. That’s (1-3.326629e-8)^25200 chance of not seeing at least one move this large. That works out to 99.92% chance of it not happening once in a century. Not never happen, but close.