Why are forecasters predicting a mild recession for Canada?

Shock Minus Control notes that Canadian forecasters in both the public and private sector are predicting a short, mild recession and he provides a good graph that puts these forecasts in historical perspective. In addition, IMF and OECD forecasts for Canada are more optimistic than for most of the other developed countries.

But how and why can it be that anyone can think that Canada can escape with only a minor slowdown when pretty much everyone is predicting impending doom for the US?

To answer this, I'm going to do a quick, back-of-the envelope calculation, and then I'm going to do some moderately sophisticated econometrics. It turns out that both approaches yield almost identical responses.

The quick version first. Here's the story people tell when they explain why Canada will be affected by a US recession:

• Canada exports a lot of what it produces.
• The US is by far the biggest market for Canadian exports.

Both assertions are true: about 36% of Canadian output is exported, and about 77% of exports go to the US. So about 28% of what is produced here is shipped to the US. That's quite a lot, but it's hard to see how this would work as a mechanism for transmitting a severe recession. If (for the sake of argument) demand for Canadian exports is proportional to US income, then a Canadian recession would be about 30% as deep as the US recession that generated it.

Now the long version. The simplest model I can think of for analysing the transmission of US shocks to the Canadian economy is a structural vector autoregressive model (SVAR) for GDP growth in the two countries. As an identification rule, I'm going to assert that a US shock in period t may have an immediate effect on Canada, but a Canadian shock will not affect the US before period t+1.

A useful feature of SVAR models is that you can trace out 'impulse-response' functions to see how a shock is propagated through time and across countries. Here's what happens if both the US and Canada are hit with negative shocks (solid lines are the estimated posterior medians, the dotted lines are the interquartile ranges):

This seems plausible: simultaneous shocks of equal size will have similar effects in the two countries.

But what if the shocks aren't simultaneous? What are the relative weights for country-specific shocks? Here are the impulse-response functions for US GDP growth for the two sorts of shocks:

The immediate effect of a Canadian shock is zero by construction, but the subsequent effects aren't particularly strong. As far as the US is concerned, it's domestic shocks that matter.

Here are the impulse-response functions for Canadian GDP growth:

As one would expect, A US shock has a significant effect on Canadian GDP growth. But look at the scales on the vertical axes of the last two graphs: the immediate effect of a negative US shock on Canada is about 30% that of the effect it has on the US.

In order to generate a prediction of a severe Canadian recession, your story has to include an important negative domestic shock - simply looking at what's going on in the US is not going to be enough.

Let's put this in the context of that graph from the PBO that Shock Minus Control posted. In both 1981 and 1990, the Canadian domestic shock took the form of a contractionary monetary policy. Between June of 1980 and August of 1981, charted bank 5-year mortgage interest rates went from 13% to 21.75%. And in the five months between November 1989 and April 1990, they went from 11.75% to 14.25%, even as inflation was declining to the Bank's target of 2%. It's likely that Canada would have fallen into recession on those two occasions even in the absence of a US shock. Compare those episodes with 2001 and 2008. In neither case was there a home-grown negative shock, so the Canadian economy didn't suffer very much, even though the US was recession.

This analysis doesn't mean that there won't be a severe recession in 2009, of course: a bad shock could happen, and/or the US economy may deteriorate beyond what is currently being predicted. But until we have a story of how and why these events will occur, it's not clear how a forecasting model is going to generate a prediction of a severe recession in Canada.

Comments

[Declan]

Surely the collapsing housing market and falling commodity prices both constitute domestic shocks, no?

[brendon]

The concern I think, is that despite the overnight rate falling to a 50 year low, credit conditions are still quite tight - and so the impact of very accomodative monetary policy seems to have ben mitigated somewhat.

Moreover, commodity prices have tanked and as Declan points out, housing and construction (massive job creators over the past 5 years) are slowing.

Therfore, the 1/3 x US shock + slowing commodity based investment spending + housing slowdown = a worse recession than that being forecasted.

The rule of thumb I have in my had for transmission of US GDP shocks is about 1/3 as well, nice to see some validation of that here. I read an IMF paper awhile back that estimated trade + finanical shocks that were as high as 0.6, might be more realistic in the current situation.

[Brendon]

Curious about something - what variables are in your SVAR?

[Stephen Gordon]

They were real quarterly GDP annualised growth rates: y = 400*(log(x(t)) - log(x(t-1))).

To the extent that falls in commodity prices are generally associated with US recessions, it's not clear how much of that can be thought of as an independent, domestic shock. But it's certainly worth considering: the Bank of Canada's new model explicitly takes term of trade into account.

It's not too hard to add to the model; I'll see what it generates. But maybe not today.

[Patrick]

What about the effects of contracting credit? In the West, especially Alberta, lots of big cap-ex has been canceled. As current projects wind down, unemployment is going to increase, and many Albertans are stuck with high debt levels and big mortgages from their spending spree during the boom years. I think it could get ugly.

[Patrick]

Just some data relevant to my previous post. From here:

http://www.albertacanada.com/statpub/1112.html

In the October numbers the total project value was $286527.2 x 10^6. For the November update it was $279300.0 x 10^6. That's about a 2% decline in 2 months (start of October to end of November). Granted, two months does not a trend make, nonetheless ...

[Roman]

There has been a housing bubble in Canada just like in the US. Prices in Vancouver are 300x rent and 10x income. Toronto is similar. Calgary and Edmonton are already crashing. Banks are have issued 40 year mortgages with no down payment that require two full time incomes to pay. What if one spouse loses their job? Did anyone account for the massive additional risk involved? Not likely. I'd say we're screwed. Not as screwed as the US mind you, and certainly not as screwed as Britain (tits up?), but everybody here is going to get a bad hit. Most people seem totally unprepared and/or oblivious to the situation. Thanks media!

[learning james]

"it's not clear how a forecasting model is going to generate a prediction of a severe recession in Canada."

Haven't you heard ? The problem is that we have relied too heavily on models and they assume closed systems.

Sentiment has changed and my guess is your models do not capture that either.

Trade is collapsing. Great Depression ?

Financing is collapsing. Banks are unable and unwilling to lend.

Household networth is collapsing (House, RSP, savings, investments)

Employment and income are declining.

Put these 'shocks' into your model.

I like your optimism. But please take off the rose colored glasses.