« Statistics Canada's Attention Deficit Disorder | Main | One good thing about Bitcoin »


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

Dunno Livio. Your Figure 1 looks to me like there might be some sort of relationship. Sure, there's a long-term downward trend in manufacturing employment, in Canada like in most other advanced countries. But I was a bit surprised that Figure 1 showed some sort of relationship in the fluctuations.

Re Figure 1: Even when there is large depreciation from 1992 to 2003, there is a bit of recovery in the manufacturing share of employment but it does not return to its prior level. At best, what depreciation of the dollar appears to have done is slow down the decline in manufacturing's employment share. Then once the dollar appreciates, it enhances the decline. Re Figure 2: This two arms picture is curious. Again, overall there seems to be a relationship but as the dollar depreciates - along the upper arm, the manufacturing employment share is quite sensitive to depreciation, along the lower arm, not as much. I have not drilled deep enough into the data but this may represent different time periods in the relationship. This suggested to me that there might be other factors affecting this relationship and I thought there might be elements of a spurious correlation between the two variables (my time series econometrics is not that strong). So I first differenced and put together Figure 3. If you might a straight line to that you get a very slight positive slope - the r-squared was 0.00242.

Livio: my tentative hypothesis would be: manufacturing share = A -B.Time -C.exchange rate.

There's a declining trend, but the exchange rate also matters.

What I can't figure out is why your Figure 3 doesn't work. Maybe annual first differences(?) is just too quick to expect any reaction. We don't expect to see manufacturing employment jump up and down that quickly in response to the exchange rate?

Strictly, it should probably be a trade-weighted real exchange rate, not just nominal USD. But I don't think that would matter much at all, over this time period.

The first differences in Figure 3 were actually month to month. If I annualize the monthly differences (eg. Jan89-Jan88,etc), the relationship is a bit stronger but the r squared was still only 0.0440. I'm going to add this second graph to the post.

Aha! Looking a bit better. But 12 months is still a short time to get a factory to increase production.

Is the coincident difference also the best measure? Manufacturing employment may respond to the prior trend of exchange rates, especially if over the short to medium term firms that deal in cross-border trade use hedging instruments to lock in exchange rates.

Thinking out loud, there may also be a relationship to the purchasing power parity level of exchange. If the true exchange rate is weaker than PPP, then it would be to the benefit of a fully flexible firm to move investment to Canada from the US, as paying identical real standards of living would be cheaper. Likewise, if the true exchange rate is stronger than PPP then the converse holds. The problem is, of course, the PPP level is itself a moving target.

sorry, test


trying to explain the exchange rate by a commodity index like the CRB, should take care of the 2008 spike, and I believe actually also of a lot of the 1992 - 2001 trend (I need to get some old CRB data).

Adding the variable of secular decline of the manufacturing share (happening in Europe as well) ? Maybe by putting this into the analysis via the US manufacturing share?

In the German small town I was born, manufacturing was close to 50% as late as 1989, and is now down to < 20%. Pretty brutal structural change, but China and Eastern Europe also have to earn their money, somehow.

Reducing the noise in Figure 3, by taking only quarterly values, more typically for orders filled ? And still preserving spike like 2008.

And, like Majromax said, some lag variable of the order of one year for adjusting production

There seems to be a 1 to 2 year lag between changes in the exchange rate and the manufacturing. I would vote for taking a difference an examining the correlation function.

Hi there:
Would the results be still the same if you compare the manufacturing share of GDP or export or any other metric with the exchange rate (not employment but the value/volume of manufacturing itself)?
Just wondering!

Figure 1 might have added context if you plotted China's WTO entry in 2001.

The comments to this entry are closed.

Search this site

  • Google

Blog powered by Typepad