The better the Bank of Canada is doing its job, the more it will appear that the Bank of Canada should be broken in two. The worse the ECB does its job, the more it will appear that the Eurozone is an Optimal Currency Area. If central banks were even slightly more competent than random number generators, then the existing set of currency areas would probably appear worse than an equal number of currency areas chosen at random.
Suppose that two Canadian provinces, or two US states, or one Canadian province and one US state, were always hit with common shocks. So when A has a 10% positive demand shock, so does B. And when A has a 5% negative supply shock, so does B.
That would be one criterion for saying that A and B are part of an Optimal Currency Area. (There are other criteria too, like whether there is high labour mobility between A and B, but I'm going to ignore those and concentrate on the "correlated shocks" criterion). The reason is that if the shocks are highly correlated, the same monetary policy that works well for A will also work well for B. One size fits both.
So if we wanted to estimate whether A and B are in the same OCA, can we just look and see whether (say) output in A is highly correlated with output in B? Not so fast.
Because when we observe output in A, we are not observing the shock in A. We are instead observing the sum of three things: the shock in A; the response of monetary policy to that shock; shocks in monetary policy.
Take an (extreme) example. Suppose that the Bank of Canada did really really bad monetary policy. Suppose that the US Fed did really really bad monetary policy. Every year, both central banks flip a coin, and if it falls heads they massively loosen, and if it falls tails they massively tighten.
If they both flip the same coin, and so do exactly the same stupid things at exactly the same time, output in Canada and the US would be very highly correlated, so they would appear to belong to the same OCA.
If they flip different coins, and so do different stupid things, output in Canada and the US would have a very low correlation, so they would appear not to belong to the same OCA.
Now lets consider Canadian provinces.
Suppose the Bank of Canada was very bad at its job, and made massive changes to monetary policy on the flip of a coin. The output of all Canadian provinces would be very highly correlated. Canada would appear to be an OCA.
Suppose instead that the Bank of Canada had a crystal ball and was perfect at doing its job. It perfectly and instantly offsets all shocks, so that Canadian output never fluctuates (or the Canadian output gap is always zero, if you prefer). If we divide Canada into two equal-sized parts (call them "East" and "West", or "North" and "South", or "urban" and "rural", or whatever) then by assumption the output of East and West would be perfectly negatively correlated, and Canada would appear to be the worst possible candidate for an OCA.
The better the Bank of Canada is doing its job, the more it will appear that the Bank of Canada should be broken in two.
The same goes for the Fed.
If the Bank of Canada and the Fed were both doing their jobs perfectly, then it would always appear to be the case that {East Canada plus East US} and {West Canada plus West US} would be better currency areas than {East plus West Canada} and (East plus West USA}. Because it's (almost) impossible that East Canada would be perfectly negatively correlated with East US. Or you could appear to do better by joining East Canada with West US, and West Canada with East US. Whatever. Absolutely anything would appear to be better than the status quo.
You don't even need to assume perfection. Provided both the Bank of Canada and the Fed were just slightly better at doing their jobs than a random number generator, it will probably appear that the East-West currency border should be replaced by a North-South currency border. Or a diagonal border. Whatever. Because doing your job even slightly competently creates a negative correlation between areas inside your currency zone.
(Does that point demolish a whole literature, or what?)
There's probably a lesson here for the Eurozone too.
This is just a variant on Milton Friedman's thermostat.
Now I need to figure out the right way to do it. I think you need to look at the correlations between the forecastable components of output fluctuations.
(A couple of days ago an economics blogger linked to a paper saying that some Canadian provinces and US states formed an OCA. It was written by three economists from Greece(!!!). Now I can't find it, because my memory has failed again.) Stephen has found the Matt Yglesias post, and the paper (pdf). Thanks Stephen!
Worthwhile Canadian Initiative 
Nick,
Is “correlated shocks” a legitimate criterion in the literature on this?
Portfolios with high internal correlation are risky. That’s the whole point of modern portfolio theory in finance (if they still call it that). Why wouldn’t that be immediately obvious in a portfolio of regions covered by a single currency?
Nick: The paper is here. (Matt Yglesias’ Slate blog post on the paper is here.)
It seems like you would need a structural VAR or estimated DSGE with separate but possibly correlated productivity shocks for each region. The correlation here (after taking into account existing monetary policy and monetary policy shocks) would be the measure of how well they fit as an OCA, at least along this dimension.
Nick, Great post. It occurs to me that this also supports McCloskey’s complaint about the misuse of statistical “signficance.” Two regions might have a very strong negative correlation in a statistical significance sense, but the size of the shocks may be rather small, and hence not economically signficant.
Thanks Stephen! Post updated.
Norman: maybe. I have my doubts about structural VARs.
Scott: Thanks! Yep, Dierdre McClosky’s “Ooomph” is going to be part of it. But it’s the magnitude of the output fluctuations relative to what could be prevented with a good monetary policy, that matters.
I need to think more about the right way to do it.
JKH: The OCA literature is more concerned with things like output and unemployment than with asset prices. There may be a connection with optimal portfolio choice. But with portfolio choice you want negative correlations inside a portfolio, while in OCAs you want positive correlations of shocks inside the OCA.
You have reading to do, Nick.
I ran the research group at the Bank of Canada that studied the OCA problem back in the 90s. Recall that at the time, many in Canada actively promoted a currency union with the US because (1) the Europeans had decided on a currency and were actively promoting the massive economic benefits it would bring, (2) we had just entered into NAFTA and it was argued that currency union was necessary to make the trade union work, (3) we were in the Great Canadian Slump and those critical of the BoC tight monetary policy liked the idea of swapping out their policy for Greenspan’s much looser policy. As a result, we had a full research agenda looking at OCA issues and how to estimate just such problems. As you might imagine, there was much empirical work on these issues as well by the IMF and academics interested in the European question.
That empirical work in the 90s sorted out the problem that you identified in your post; what mattered was not the output (or inflation) correlations, but those correlations after we compensate for the impact of independent monetary policy (which will disappear after a monetary union. The result was a 2-step methodology: use a structural VAR (or other structural model) to estimate the monetary policy shocks, then use a factor model to estimate the common and idiosyncratic components.
I recall Bayoumi and Eichengreen wrote some of the key papers on this (one with over 1000 cites is Bayoumi, Tamim, and Barry Eichengreen. Shocking aspects of European monetary unification. No. w3949. National Bureau of Economic Research, 1992.) The team at the BoC, particularly a very talented economist called René Lalonde (Hi René!) wrote multiple papers looking at the OCA properties of Europe, Canada, and North America. (e.g. Dupasquier, Chantal, René Lalonde, and Pierre St-Amant. “Optimum currency areas as applied to Canada and the United States.” Exchange rates and monetary policy (1997): 131-170.) This careful empirical work took into account precisely the problem that you’ve just noticed.
IIRC, some of the most interesting findings from this research included (1) finding that the EU did not look like nearly as good of an OCA as the USA — peripheral countries like the UK particularly stood out as bad candidates, (2) if you wanted to make Canada a better OCA, you’d dump Alberta and Newfoundland, while the regions that made the most sense as a Canadian currency zone were Ontario and Québec — which was an interesting finding in the context of the 1995 referendum on sovereignty, (3) no Canadian region looked like it would be better off in monetary union with the US than in the Canadian dollar zone.
Simon: so I’m reinventing the wheel? OK. In one sense, that’s a relief.
Just had a v. quick look at the OCA paper that started the discussion: seems they weren’t aware of this empirical literature either. (You’d think it would be hard to miss a paper with over 1000 cites…..and you’d be wrong. Stuff like this helps keep academics humble — I mean, what are the odds that anything we write will ever be known by anyone “outside of a small circle of friends”?)
Another point worth noting in that paper is that they seem to be analyzing data for a single year. That is, they are not looking at the variability of anything, nor at any “shocks”. They are instead grouping states and provinces on the basis of the levels of various variables like GDP, unemployment, deficits, exports.
I don’t know what this has to do with OCA. Do you? Does Yglesias?
But their maps are pretty…
Nick wrote:
Suppose instead that the Bank of Canada had a crystal ball and was perfect at doing its job. It perfectly and instantly offsets all shocks, so that Canadian output never fluctuates (or the Canadian output gap is always zero, if you prefer). If we divide Canada into two equal-sized parts (call them “East” and “West”, or “North” and “South”, or “urban” and “rural”, or whatever) then by assumption the output of East and West would be perfectly negatively correlated, and Canada would appear to be the worst possible candidate for an OCA.
This makes no sense to me at all, Nick. I mean, I get what you think you are demonstrating here, I just deny that the output between East and West Canada would be perfectly negatively correlated with perfect monetary policy.
Are you using shorthand for “output deviation from the national average” or something? Because actual output between East and West Canada would be highly correlated with perfect monetary policy, wouldn’t it? If a positive supply shock is hitting East Canada, there’s probably a positive supply shock hitting West Canada too, right? At least, way more than a positive supply shock hitting Bangladesh at the same time?
@Bob Murphy: when Nick says that the Bank of Canada is perfect at doing its job he means that it manages to hold Canada’s GDP growth rate constant. Growth in Canada’s regions won’t be constant. If you calculate the correlation between anything and a constant you get zero. Also, since the growth rates of any two-region partition of Canada averages to Canada’s constant growth rate, growth rates of the two regions have to move in opposite directions. They are perfectly negatively correlated.
I don’t get this “optimal currency area” business at all. We never speak of optimal areas for commodities or financial securities. Those things just naturally go from where they are abundant to where they are wanted. And if some area has a tight central bank while another has an easy central bank, money will just flow from the easy zone to the tight zone.
Simon: though I personnally think that ON-QC is not a OCA per se, it is way better than Canada as a whole. That thinking is common amongst QC economists and can justify the promise of keeping the C$. ( An other reason is to calm the fears of the LIV, fears stoked by the federal side.) Politically it would not have flown but a stable currency realignment with the Q$ at a 10-15% discount from the US$, slowly rising as productivity adjusted, would be a nice substitute.
Anecdote : at the time of Meech Lake, I was slated for a post within the ON inter-provincial relations bureaucracy (long story). A good topic of conversation was ” Should the 4 original provinces make a new pact among ourselves and just declare the rest out.” In my view, it would have solved a lot pf problems…
Simon: You are right about that paper. Quite apart from my critique here, it makes no sense at all. It is only looking at data for one year. It does not even try to measure shocks. All it’s doing is using principal component analysis to measure which states and provinces are “alike” in some totally atheoretical sense. Whether they are “alike” in the sense of being candidates for an OCA, or as candidates for finding a good marriage partner, is left as an exercise for the reader. But yes, the pictures are very captivating.
Bob: I see your point about supply shocks. Good monetary policy should let some supply shocks affect GDP. (It depends what we mean by “supply shocks”). It doesn’t affect my general point here though, provided most shocks are demand shocks.
Mike: if you don’t think prices are sticky, or if you think there’s a very high degree of currency substitution across national boundaries, it will be hard to make sense of OCAs. There is indeed a question there for OCA theory. If (say) Alberta would do better with the USD, why don’t individual Albertans use the USD?
Jacques Rene: after the Eurozone debacle, I would have thought that any talk of currency areas that don’t reflect national boundaries would have died out. I hate to think what might have happened had Simon’s group’s empirical analysis showed that the US and Canada were an OCA, and we had created the Amero.
Nick: That’s why everybody knowledgeable in QC knew we wouldn’t keep the C$. But the LIV wouldn’t understand it before the fact. And we have to take them into account ( and be accused of hypocrisy…). Still today, most people in Europe wants to keep the Euro. The Gazette-G&M still talk about how the Euro event show the need for a common currency in Canada( and I don’t thinkn itis because they understand fiscal federalism and balance-of-payments problems…)
A currency is a potent symbol…
And the Euro would have had a fighting chance has it been backed by a central bank instead of a currency board. The whole thing show how you can’t integrate if you don’t have the same vision of how things should work.
Britain, France and Germany builds different tanks and planes,not only as protectionnist-industrial policy measures but also because they have different views on how battles should be run and the materials needed for them.
Over time I would imagine that a rural vs. Urban OCA refining to make more sense. Hmmmm.
Just for the fun of it:
http://www.voxeu.org/article/renminbi-bloc-here-asia-down-rest-world-go