A thought-experiment. A rather real thought-experiment.
Assume you are Governor of the Bank of Canada. Your job is set monetary policy to bring inflation to the 2% target in the "medium term". You are just about to conclude a meeting where you will decide what target for the overnight rate you will set at the next Fixed Announcement Date. You have tentatively agreed on R% as the new target.
Then someone bursts into the meeting room and shows you Stephen's graph. The US GDP data for the last 2 years have been revised down. The US economy has been much weaker than you thought it was. How do you respond to the news? Do you raise R, lower R, or leave R the same?
(Note: I am not asking whether you think the Bank of Canada should have a higher or lower overnight rate target compared to what it actually is today. I am asking whether you think it should be higher or lower compared to what you thought it should be just before you saw Stephen's graph.)
I think most people's immediate reaction would be to lower R. They would reason that the US economy is weaker than we thought it was, and a weak US economy, other things equal, means weaker demand for Canadian exports to the US, and this weaker demand for Canadian-produced goods means you need to loosen Canadian monetary policy to offset that weaker US demand. So Stephen's news calls for a lower R.
But that reasoning is wrong.
You have not learned that there will be a weaker US economy than you thought it would be in future compared to the present. What you have learned is that the US economy was weaker than you thought it was in the past 2 years. And you already have Canadian data over the last 2 years.
Put it another way. Yes, you have learned that the US economy has been weaker than you thought it was. That's bad news, for Canadian demand. But you have also learned that the Canadian economy has offset that US weakness better than you thought it did. That's good news, for Canadian demand. The bad news about the past, and the good news about the past, cancel out. Canadian GDP, unemployment, and inflation, have been what they have been. You haven't learned anything new about them.
And it's not the past, but the future that matters.
This is how you should approach the question.
You already had Canadian data on: GDP for May, unemployment for June, and inflation for June. And you had made tentative forecasts (or "backcasts") for June (in the case of GDP), July, August, and so on into the future, where those forecasts were conditional on your monetary policy. Then you adjusted your proposed future monetary policy until your conditional forecasts came out to where you wanted them to be, given the circumstances. In particular, you wanted a forecast for inflation at 2%, about 18 months to 2 years ahead, conditional on your proposed future path of monetary policy.
Now you get the news in Stephen's graph.
How would the news about US data revisions affect your forecast for Canadian inflation conditional on your tentatively proposed future path for monetary policy? If that news lowers your conditional forecast for inflation, you need to loosen your proposed path for monetary policy to bring your inflation forecast back to target. If it raises your conditional forecast for inflation, you need to tighten your path for monetary policy.
Not quite so obvious, is it?
For example, you might argue that Stephen's graph means Canada needs to raise R, and tighten monetary policy. You would reason that the US economy has been weaker than it looked, which gives it much more room to grow over the next year or two, aided perhaps by the US fiscal and monetary authorities seeing the need for greater stimulus. And this higher future US growth will strengthen demand for Canadian goods compared to what you had forecast. So Canadian inflation will be higher than you had forecast, so you will need to tighten Canadian monetary policy to offset that higher conditional forecast for inflation, and bring it back down to 2%. Which means raising your target R for the overnight rate.
You might reason like that. But I don't think I would.
My tentative answer, for what it's worth, is this.
There is more to monetary policy than just a time-path for the overnight rate of interest. OK, that's a controversial statement. But even those who see it as controversial will allow that the exchange rate can also handle shocks, especially foreign shocks, to demand for Canadian goods. I'm really not sure what the news implies about future US growth. But I think it does imply that US interest rates will be lower in future than we had thought they would be. In the long run, Canadian interest rates will, on average, follow US and world interest rates. The job of the exchange rate is to handle permanent changes to demand for Canadian goods relative to foreign goods. And the job of the interest rate, relative to world interest rates, is to handle transitory shocks to demand for Canadian goods relative to foreign goods.
I think Canadian interest rates will need to be lower on the news, mainly because US interest rates will need to be lower (i.e. rise less quickly). But, if the news about US data revisions does have any implications for future US demand compared to current US demand, those implications will be mostly permanent. The exchange rate will mostly handle it, if needed. And remember, a lower Canadian interest rate will increase demand for Canadian goods. So it would need to be offset by a higher exchange rate, other things equal, to lower demand for Canadian goods back to where we forecast we wanted it to be.
I have little faith that my answer above is correct. It's a hard question.
The main point of this post is simply to say that the obvious reasoning is wrong. Learning about the US past only matters for Canadian monetary policy if it affects our conditional forecasts of the Canadian future.
Worthwhile Canadian Initiative 
Well, it does reinforce my gut notion that much economic reporting up to an including our own dear Stephen has been far too rosy. It just doesn’t match what I see, especially with regards to hiring.
Right now our exchange rate is in an area that I know many manufacturers from personal experience will find excruciating. A low Canadian dollar with respect to the US dollar was the crack-cocaine of corporate Canada in the 1990’s and early 2000’s. Everybody used it, everyone loved it, everyone got addicted to it. But dear me the withdrawal is dreadful….
I knew enough from newspapers that the exchange rate can be a fickle friend.
Secondly, I have no idea what is going to happen in Washington DC this weekend, but I believe that American debt will be downgraded by the ratings agencies. We have come too close to the brink and seen over the precipice to go back. That by itself is a game-changer. What happens next is anyone’s guess. We just don’t know, we’ve never been here before and every government finance and economics textbook says that the place where we are doesn’t exist; it is a delusion, it will never happen in reality.
I really think we have to start using the word Depression now.
Nick, I think you make a good point. It seems intuitive, at least, to a non-expert.
As an aside, I recently read that both Policy Alternatives and the Canadian Labour Congress have called for a large new jobs program. Both reports they released on this contain no real discussion of macroeconomic theory. In documents that pivot on macro theory this seems odd. It turns out that the CLC report assumes there is no NAIRU. Policy Alternatives’ numbers must mean they think it is at most 6% but there are hints that they don’t believe NAIRU theory either. Neither report even recognizes that this is controversial. I’m confused by this.
Determinant: are you saying you think the Canadian data is wrong too? (I mean that Canadian GDP is lower than StatsCan says?). If those revisions in US data cause you to want to revise down your own estimates of recent past and current Canadian GDP, relative to potential, then that would be an argument for loosening Canadian monetary policy after having seen Stephen’s graph.
Donald: Thanks. I was wondering if I had made it clearly enough.
I haven’t read the PA and CLC reports. There is a bit of unwillingness to think in terms of NAIRU/Natural rate of Unemployment in some quarters. But maybe, just maybe:
1. They could argue that particular features of their program might lower the NAIRU. (Not obvious how it would, but NAIRU’s aren’t written in stone, or fixed independently of all policies, just monetary policy.)
2. They see them as temporary measures to help us get back to the NAIRU.
3. They know the programs won’t reduce unemployment, but they might raise equilibrium wages. Almost everybody argues for almost every policy on the grounds that it will “create jobs”, even if it has nothing to do with jobs.
http://worthwhile.typepad.com/worthwhile_canadian_initi/2010/07/hst-and-jobs-more-studies-in-applied-orthogonality.html
I’m saying the Canadian and American economies are deeply intertwined. I have worked for three different companies primarily focused on exporting their products to the United States. That is a simple, uncontroversial fact of life to anybody who has worked in southern Ontario. It is readily apparent that a weak US economy means the Canadian economy will have one hand tied behind its back. There’s nothing exotic about this, you just have to notice where the sales orders comes from.
If the US is so weak then I can’t see an argument for raising r. Yes, I have long doubted Stephen’s reporting on the Canadian economy. No offence to you Stephen, my eyes and ears simply tell me something different than your statistics do.
The terrible truth is we can’t support industry and export services with low rates without running a large danger of igniting our own property bubble.
Exchange rate intervention might be a better policy that lets the housing market stay even while supporting US-related businesses.
Determinant: then I think you have missed my point. OK, we have now learned that the Canadian economy has been fighting with one hand tied behind its back. But we also know, from statsCan, how effectively (or not) the Canadian economy has been fighting. Which means we have also learned that its other hand has been stronger than we thought it was. So we have both bad news and good news about the past. (It’s bad for the US, obviously, but I’m talking just about Canada.) And the bad and good news cancel out, for the past. What do they say about the future?
BTW: I may not be responding to comments for a few days. The MX6 needs some exercise.
Determinant: On second thoughts, or are you saying that you have now revised down your estimate of recent past Canadian GDP, in the light of Stephen’s graph? If so, OK.
Thanks for the reply. The post you linked about Mintz/Ivanova was a good one. My Stalin posters and I appreciate anecdotes about communists.
Nick,
Did you know that during the Great Contraction/Depression in the USA, there was not a single bank failure in Canada! And they had no Central Bank until 1935!!!
Just because US and Canadian economic growth rates are strongly correlated doesn’t mean that the BEA and StatsCan GDP revisions are correlated. My initial reaction was about the problems that the US faces, and how far they are away from a recovery. The only thing the Bank of Canada had to think about is how to react if the Fed embarks on QE3.
There’s no obvious reason why we would want to revise Canadian GDP numbers down. US employment numbers have long been disappointing, and now we have at least a partial explanation for why. Canadian employment growth has been much stronger.
Maybe Simon van Norden will stop by and comment.
There’s more to Canada than Southern Ontario. It’s easy to overestimate the problems with the Canadian economy if you live here instead of, say, Calgary.
But yeah, the high dollar is hurting a lot of exporters. Myself included.
“Just because US and Canadian economic growth rates are strongly correlated doesn’t mean that the BEA and StatsCan GDP revisions are correlated.”
I think it suggests this, if the economies are correlated. As the methodologies and data sources are similar. GDP revisions are not random, they are due to more information arriving at later times, due to some underlying economic process that can only be measured with a lag in comparison to other economic processes that can be measured more quickly.
As an example, suppose that we have only two time series, payroll data (arriving first) and tax data (arriving later). Let’s say tax data is the “true” measure of income that is predicted imperfectly by payroll data, with an upward revision coming from an increase in the relative share of capital income and a downward revision arising from a decrease in capital’s share. If capital’s share is correlated in both economies, then the revisions will also be correlated.
A good inference rule will make the revisions appear randomly distributed (across time) for a given economy, but it need not succeed in making the errors appear random across correlated economies. Take a look at chart 2b of //www.bankofcanada.ca/wp-content/uploads/2010/06/tkacz.pdf
@Nick:
I was at odds with Stephen’s estimates and thought they were too high. It may come to the same thing in that I am a GDP bear.
Hmmm.
Suppose we know that there is a correlation between true Canadian GDP and true US GDP. Suppose that Canadian GDP and US GDP are both observed with noise, but the two noise series are independent. Wouldn’t a Bayesian “backcaster” nevertheless revise down his estimate of Canadian GDP when he learns about US noise and revises down his backcast of US GDP? The optimal backcast of Canadian GDP is presumably some linear function of what StatsCan says it is plus what the US equivalent says US GDP is?
That sounds right. Your description accurately describes my thinking, and I am a sucker for anything Bayesian.
I’ll have a look at the correlations between the revisions of GDP growth rates for various countries later today (in reference to Chart 2b above), but here are some quick thoughts:
1. Eyeball-metrics suggest that there is a small, positive correlation between U.S. and Canadian GDP revisions. The correlations seem highest around business cycle turning points (2001 and 2008), and virtually non-existent in between (e.g. 2005 & 2006).
2. Canadian export and import data are already subject to large revisions due to the difficulty in measuring these, so I won’t be surprised if exports get revised down quite a bit (historical mean revision to export growth is already close to -1%). However, this will likely be offset by upward revisions to consumption, which tends to be under-estimated (historical mean revision is +0.34%). Overall impact on Canadian GDP growth revisions should therefore be somewhat subdued.