« Is intertemporal trade like international trade? | Main | DIY Economics: Canadian Dollar Edition »

Comments

Feed You can follow this conversation by subscribing to the comment feed for this post.

RE: Start date of 1995. That's as far back as the series for the Overnight Rate goes. The series for the target rate goes back much further, but I wanted to use the actual/effective rate, because it will drift from the target from time to time.

Would that be because Canada is gradually becoming a petro-state?

Mike,

Thanks for the analysis. I like your year-by-year analysis of WTI and CAD/USD and especially how it shows a loss of correlation in 2000 and 2003. I assmue that this part of the analysis was completed by considering the correlation over each calendar year. If you think about it, the use of Jan 1 of each year as the cut-off date is arbitrary and different results could arise if one chose July 1 as the cut-off date. More generally, it could add extra information if one considered a moving buffer of 250 days over the entire data set and then calculated a correlation for the buffer for each point in the data set (omitting the first 249 points of course). Essentially a moving correlation, similar to a moving average. It's just a technical suggestion, but it might prove interesting.

Hi Kosta,

Great idea - I'll try it and post a follow-up!

Cheers,

Mike

Would that be because Canada is gradually becoming a petro-state?

Two things are happening - oil prices have shown a rapid increase over this period. And oil volumes as well (shift from conventional to oil sands and then some). I have mentioned the latter point before.

As the relative weight (price x volume) of the energy sector increases (as part of the Canadian economy) shouldn't one expect the correlation to oil prices to strengthen over time?

Mike..what do your results suggest will happen if the Bank of Canada raises interest rates ahead of the Fed over the next few quarters? Does a low correlation between interest rate differentials and the Canadian dollar imply we should not necessarily anticipate a pull on the currency, as some fear

Hi Mike,

I had some comments about your observations.

In point 3 you observed that over the long run, interest rate differentials did not correlate with WTIOIL. I don't think this should be particularly surprising unless you think that Canadian interest rates somehow influence the price of oil (which I doubt). On the other hand, the high price of oil might lead to Canadian interest rates being raised as the Canadian economy overheats, but high prices are something we've seen only lately. Perhaps if you reduced your analysis to the last 5 years you would see a different relationship.

In point 4 you observed that oil prices and interest rates have somewhat of an inverse relationship. You note that this makes some sense as high interest rates should be disinfltaionary. I agree, but I think that its hard determine is interest rates are high based solely on their absolute level. What are high interest rates? For instance, if the Fed raised the Fed fund rate to 2% today, it would be high, but those rates would be low for 2006. High interest rates imply a contractory monetary policy, but the level of interest rates rates which result in a contractory monetary policy varies with the strength of the economy.

It might be informative to add additional variables to better delineate whether monetary policy is expansionary and possibly the strength of the economy. Inflation rates might be useful, as might GDP growth rates (or maybe just go straight to NGDP growth rate). Or, if you want to stick to just interest rates, looking at the rate of change might useful, where periods of rising interest rates implying a contraction in monetary policy (although again one is still challenged by whether the rate of rise perhaps being appropriate for the economy).

I do realize that there's a limit to the number of variables you want to look at, but just some thoughts.

Thanks for doing this.

The correlation coefficients are Pearson correlation coefficients correct?

I have to wonder about the frequency of the data set. It presumes that agents are responding to commodity price movements within a 24-hour period. I'm sure some traders do but others in the market, especially those moving investment capital around, are taking longer. Oil company valuations, for one, tend to lag commodity price movements. The Chinese are overpaying for assets as if US$90/b or higher oil prices were around the corner. But they are the exception.

Instead of using overnight rates that are determined by domestic players, how about some market-determined rate of return that moves foreign capital flows? Currently, it seems that market-determined rates, such as mortgage rates, are disconnecting from overnight markets. The information on agent expectations contained in these market rates are largely absent from the overnight market. Presumably overnight rates are exogenous instruments of the Crown.

So what's next? Time series modelling with numerous lagged variables?

I'd be curious to see a simple OLS regression of first-differenced logged exchange rates over oil prices. If the Pearson correlation coefficient is above 0.9, I would expect the R^2 of this first-differenced regression to be close to 0.6 with normally distributed residuals and a friendly p-value.

However, if Just visiting from MacLeans' hypothesis is right about the growing importance of oil exports, then the regression should exhibit some noticeable heteroscedasticity.

I just took another look at the graph. The annual Pearson correlation coefficients turn negative in the two recent periods, 2000 and then 2003, that immediately follow years where petroleum prices sharply declined. The sharp spike in natural gas prices in late 2000 followed by much weaker prices in late 2001 and 2002 also played a role I would guess.

Note carefully: Oil trades in a global market; natural gas prices are largely North American determined. Imported liquid natural (LNG) is only viable at high natural gas prices. (I say note carefully because I don't believe the Chinese get it nor do the good folks who put together US foreign policy and have helped transformed the USA into a declining power and the premier terrorist nation state of the early 21st century.)

This observation supports the notion that it is not only commodity price movements that are driving demand for Canadian dollars but slower moving investment flows.

Superimposing year-over-year first-differenced oil prices on the graph may shed more light.

Westslope: Great ideas! I've made available the data I use (see my new post), if you'd like to play around with it. I'd be curious to see what you find.

Theo: "What do your results suggest will happen if the Bank of Canada raises interest rates ahead of the Fed over the next few quarters?" According to the data - not a whole lot. But I believe the data is misleading in that regard. Will save that for a future post.

Kosta:

RE: Point 3. I worded it badly. What I meant to say is that I was surprised that there is no stastical relationship between the Canadian dollar and interest rate differentials.

RE: Point 4. Agreed. I should have phrased it as 'rising interest rates' rather than 'high interest rates'.

Thanks for your suggestions! I would like to add additional variables to this data set. Ideally I would like to add variables that I can obtain daily values for.

Mike,

thanks for the responses. I have to agree that it is surprising that the interest rate differential does not have a significant correlation to CAD/USD.

Also, I'd like to second westslope's suggestion of using a market based measure of interest rates. I'd vote for US 10 year treasury yields. You could also look at the spread between treasury yields and the central bank rate as a rough measure of whether monetary policy is contractory or expansionary.

Hi Kosta,

10-year bond yields would be a useful addition to the dataset. Let me see if I can track down that data.

http://en.wikipedia.org/wiki/List_of_countries_by_proven_oil_reserves

Canada has the second largest oil reserves in the world. The value of a currency to foreigners is the value of the exports that the currency can buy. When oil price goes up, the value of Canada's exports, both potential and actual, goes up. I would guess that the reason the correlation is growing is because as time goes on, it becomes more and more likely that the oil sands will be exportable.

I think the reason that the correlation is as high as it is may be due to random chance. -insert standard critique of data mining here-. Also, I don't think that interest rates would be highly correlated, as long as the Canada central bank is credible about holding down inflation. For example, during the financial crisis in the U.S., the interest rate plummeted, but the currency soared. When there is shifts in inflation or chaos in financial markets, interest rates can do wacky things.

My hypothesis:

Decompose the price of oil into "permanent" and "transitory" components.

Where "permanent" price of oil = 33% x Expected Present Value of Price of Oil discounted at 33%p.a.

(33% is the MCI coefficient, roughly).

And "transitory" price of oil = actual price - permanent price.

1. The "permanent" price of oil determines the exchange rate

2. the "transitory" price of oil determines the interest rate differential.

Since the price of oil is highly persistent (serially correlated), most changes are "permanent", and that explains the high correlation with the exchange rate.

This may be a dumb or an obvious question to those that play with these things far more often than I do.

If the objective is to determine today's exchange rate, how good of a predictor is yesterday's exchange rate? (the reason I ask this is that it, in my mind at least, corrects for longer term changes in oil production volumes over time - then add in the daily variables such as WTI etc).


did you do any cointegration testing?

time series here, right? not just cross sectional data ...

Back space.... In a post above, I implied that the Chinese were buying resource company in order to have access to the commodity, and this makes no sense given oil is globally-traded commodity.

Earlier this week, I heard a market observer explain that the Chinese were not interested in the commodity but the profits and access to technology. A profitable hedge. Presumably the fortunes of China's manufacturing sector and the global oil industry are to some extent negatively correlated. That explanation actually makes some sense though I don't quite understand why the Chinese firms are willing to pay such a premium.

Mike: I look forward to seeing the data set. :-)

In a post above, I implied that the Chinese were buying resource company in order to have access to the commodity, and this makes no sense given oil is globally-traded commodity.

Different for bitumen and heavier synthetic crudes coming out of Alberta for refining elsewhere. This may require billions of dollars of upgrades at refineries in home countries. Hence why you'll see more vertical integration or joint venture arrangements (producer/refiner) as expansion proceeds.

Importance of security of supply? Have a look at Venezuela/NB Power deal that fell through in these two posts:

Venezuela's state oil firm PDVSA will ship fuel oil to several foreign clients instead of its patented boiler fuel known as orimulsion, which is gradually being phased out of production, newspaper El Universal reported.

Substitution of fuel oil for orimulsion is foreseen in several supply contracts, the paper reported.
...
Interrupting shipments of orimulsion has put PDVSA in legal hot water. Canadian power company New Brunswick Power has sued PDVSA for US$2bn for breach of a supply contract, while Italy's Enel submitted an arbitration request to the International Chamber of Commerce for US$200mn. No decision has yet been reached in either case.

http://www.bnamericas.com/news/oilandgas/PDVSA_to_substitute_fuel_oil_for_orimulsion_exports

FREDERICTON - Senior officials at NB Power knew that an ill-fated, $2-billion deal to buy cheap fuel from Venezuela was at risk, but kept that information from the provincial government, the Crown-owned utility's new president testified recently.

David Hay was speaking to a legislative committee that is investigating why NB Power began a costly refit of its Coleson Cove generating station when it didn't have a signed contract to buy Orimulsion, a less costly form of fuel made from coal and water.
...
"(But) Stuart MacPerson and David Read appear to have believed or known that the fuel-supply agreement was at risk on Nov. 8."

Still, the utility went ahead with a $750-million upgrade at its generating station in southwestern New Brunswick.
...
If NB Power is unable to get a court to order Venezuela to deliver the fuel - as promised in an informal agreement - the public utility is demanding $100 million a year in damages to cover the added costs of buying a more expensive fuel.

In the worst-case scenario, electricity rates in New Brunswick could increase by as much as 10 per cent if the fuel, which is produced only in Venezuela, fails to arrive.
http://www.electricityforum.com/news/mar04/nbpower.html


Hi Mike;

Have you looked at some of the work that the Bank of Canada has done on instability in this relationship? Issa, Lafrance and Murray had a working paper in 2006 (Bank of Canada Working Paper No. 2006-29) and Philipp Maier and Brian DePratto had a paper in 2008 (Bank of Canada Discussion Paper No. 2008-15)

Cheers,

Simon

The comments to this entry are closed.

Search this site

  • Google

    WWW
    worthwhile.typepad.com
Blog powered by Typepad