An enduring mystery of the Canadian economy is trying to figure out why the CAD-USD exchange rate is doing whatever it happens to be doing at any given moment. Notwithstanding popular opinion – an opinion that is all-too-often reinforced by commentators who should know better – the CAD-USD exchange rate isn’t like a stock price, rising when the Canadian economy does well. In fact, the correlation between GDP growth and the exchange rate is generally negative, as a casual glance at this graph confirms (click for larger image).

In subsequent posts, I’ll be working through this issue, but I’m first going to look at the ‘petro-dollar’ hypothesis. Mike Moffat at About Economics was asked ‘Is the value of the Canadian dollar related to oil prices?‘ To his surprise and mine, he gets an R-squared of 0.8477. As he says,

The 0.8477 figure is exceptionally high, denoting that there is a very, very strong positive relationship between movements in oil prices and movements in the value of the Canadian dollar.

(emphasis in the original). It turns out that the relationship isn’t so strong after all, and has a lot to do with his choice of sample. Here’s a plot of the exchange rate against oil prices using data going back to 1980 (click for a larger version):

If you only look at data from January 2002, you see the strong upward trend in both series, so a R-squared of 0.85 isn’t really surprising. Of course, if you try to explain the exchange rate over this period using just a linear time trend, you’d do even better: a R-squared of 0.91.

But if you look at the data going back to the 1980s, that relationship is much weaker. For example, oil prices can’t explain how the exchange rate went from 0.73 USD to 0.88 USD and back again during 1985-1995: they stayed at around 20 USD for pretty much the whole time.

I’ve run several regressions using daily data since 1986, and I never got a R-squared greater than 0.01.