Sunday, January 10, 2010

Oil and the Dollar, 2008 vs 2009

As a follow up to my post last Monday, I went ahead and did a little statistical regression analysis on the correlations between Crude Oil and the US Dollar/Swiss Franc exchange rate (my favored dollar value barometer), and compared the 2008 correlations with those of 2009.
Here's a plot of 2008's Crude Oil prices vs. the USD/CHF exchange rate (Swiss Francs per US Dollar), both end of day prices:
And here's the same plot for 2009:
2009's plot looks a little bit better.
Running a full regression analysis on the 2008 linear model gives me the following output:
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 495.79 21.82 22.72 <2e-16>
dollar08 -365.87 20.13 -18.18 <2e-16>
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 18.72 on 250 degrees of freedom
Multiple R-squared: 0.5692, Adjusted R-squared: 0.5675
F-statistic: 330.4 on 1 and 250 DF, p-value: <>
vs 2009's:
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 311.029 6.946 44.77 <2e-16>
dollar09 -229.369 6.393 -35.88 <2e-16>
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
Residual standard error: 5.399 on 250 degrees of freedom
Multiple R-squared: 0.8374, Adjusted R-squared: 0.8367
F-statistic: 1287 on 1 and 250 DF, p-value: <>
That's right. The 2008 Oil Price/Dollar Strength correlation was about .57, while 2009's was about .84 (the bolded, red R_Squared numbers). That means that in 2008, about 57% of oil's price variation could be attributed to its relationship with the dollar. By contrast, in 2009, about 84% of the variation in the price of oil could be attributed to its relationship with the dollar. (NB: If I really fudge around the data and institute a lag, I can get an R_Squared as high as 75% for the 2008 correlation, but that's with a 12-day lag, compared to 2009's day-of correlation which, incidentally, does not improve whatsoever with a lag. Still, even optimized as much as is possible, 2009's correlation is significantly greater than 2008's.)
So in the short time of one year, we've seen the correlation between the value of the dollar and the price of oil increase dramatically. The thing to keep in mind is that with bivariate data such as this, R_Squared doesn't care which variable is dependent upon which, so while today I'm happily discussing how oil prices are affected by the dollar, I could just as well be saying the reverse. And as I've said before, as the global economy grows and emerging powers such as India and China begin consuming greater amounts of oil, it may not be long before we no longer discuss how a strong dollar drove oil prices higher, but rather how skyrocketing global oil prices turned the dollar into a worthless currency.
Full Disclosure: Short March 2010 Mini Crude Oil (QMH10) as of writing.

2 comments:

Eric said...

woo stats. not to mention all the other really good reasons why the dollar will continue to erode... you could probably get an even better fit with a nonlinear regression. not that it would mean much more.

Charles said...

Hey Eric,

Thanks for the comment. I actually played around with a non-linear correlations (squares, logs, etc.) and couldn't find one that was signifincantly better than pure linear. I mean, check out that graph, after all.

Though if you have some suggestions I'd love to know your thoughts.

Thanks!