Sunday, June 5, 2011

The International Forex Network, 1998-2010

. Sunday, June 5, 2011















This weekend I got bored of cleaning some BIS banking data, so decided to play around with their foreign exchange data while watching the Cardinals beat up on the Cubs. There's less of those data, so it was easy to quickly get it cleaned and loaded into R. From there, I made the above graphs. The BIS only collects these data every three years, so the above visualizations represent the last five surveys, covering 1998-2010 (data here)*. These are bilateral ties, e.g. the USD<->EUR ties represent the nominal dollar value of all transactions between those two currencies. The thickness of the tie represents the amount of those transactions, divided by a constant (75) for all periods to make the visualization better. The size of the nodes represents the percentage of total forex transactions involving that currency.

These are quick-and-dirty. I used a simple Fruchterman-Reingold layout to emphasize centrality. Some currencies didn't exist for the whole series -- the euro in 1998; in later periods the franc, mark, ECU (XEU), and "Other EMS" which were rolled into the euro -- but I just gave them zero ties rather than spend the time to remove the actual nodes. (Hey, it's a weekend blog post.) The non-existent currencies are easy to see, as they are disconnected from the rest of the network. Just pretend they aren't there. Also, as I type this I realize that node size for "Other currencies" (Oth) is wrong because I inadvertently left some minor ones out, but the ties are correct and the node size wouldn't change by much since those are all small currencies.

Still, there's some interesting stuff to see. Most immediately obvious, the amount of foreign exchange increased noticeably from 1998-2010, as evidenced by the increasing thickness of the ties across the period. There was a 20% jump just from 2007-2010, which the BIS attributes mostly to technological improvements that lowered transaction costs and high-frequency trading.

Perhaps more surprising is the fact that the shape of the network has changed very little over the past dozen years. The US was the most central node, and the largest in 1998. The increased activity in the intervening years hasn't changed that at all. In 1998, the US was involved in 86.8% of all foreign exchange transactions**. In 2010, the number was 84.9%. The US's centrality (by this measure) peaked in 2001, when 89.9% of forex transactions involved the dollar. Similarly, despite much fanfare the euro has not moved to an especially central position. In 1998 the German mark (30.5%) and French franc (5%) were involved in 35.5% of forex transactions; in 2010 the euro (which absorbed not only the mark and france but other currencies as well) was one of the currencies traded in 39.1% of transactions. The yen decreased very slightly from 1998-2010 (21.7% to 19%), and the pound sterling increased slightly (11% to 12.9%), but in general the network changed very little. The Chinese yuan increased its share by 900% from 2004-2010... but was still on one side of less than 1% of forex transactions in 2010 (from 0.1% to 0.9%). This is shocking: despite all their growth over the past dozen years, in which their GDP has nearly quintupled, the world's second largest economy (third if we consider the eurozone as a collective, as we should for these purposes) is involved in fewer than 1/200 foreign exchange transactions. Nothing else changed much either. There are more thick ties in 2010 than there was in 1998, but all of them include the USD.

USD<->EUR transactions accounted for 28% of all transactions in 2010, close to its 2001 peak of 30%. USD<->JPY was second, with 14%. No other pair had more than 9%, and no pair that excluded USD had more than 3%. The extreme inequality in these relationships is shown by the fact that almost every currency in the network above is tied to the USD in all periods. Very few are tied to any others, short of EUR<->GBP and EUR<->JPY. China, in particular, is conspicuously weakly-tied considering the fact that it is the world's second-largest economy and it engages in so much trade.

There's been a lot of talk in recent years about a "post-American world", and the rise of a multilateral international monetary system to replace the US's "unipolar moment" in the 1990s. Several countries have spoken loudly about trying to displace the dollar as the world's reserve currency, replacing it with the IMF's SDRs or an international basket. These data indicate that such discussion may be premature. While it's possible that such a transition could happen in the future, there has been very little movement in that direction over the past dozen years. Given the fact that complex networks with an unequal topology have a habit of reinforcing themselves over time, we should qualify claims that the US dollar's role in international currency markets is in terminal decline.

*The BIS site says that there have been eight surveys, but the data I downloaded only had the five I present.

**Percentages in this paragraph, and this paragraph only, are out of 200% rather than 100% because each transaction involves a pair.

5 comments:

Emmanuel said...

As someone who's actually worked as a forex trader, I must say this is basic and unsurprising.

First, of course the US dollar is going to be the "central" currency insofar as all basic currency pairs are referenced to the US dollar. In order to derive the GBP/JPY cross rate, for instance, you need to determine USD/JPY and GBP/USD.

Second, the last time I checked, the yuan was not a freely exchangeable currency, meaning it should have next to no volume in international transactions at the current time. So there's really no point engaging in "we've got bigger FX volume than the Chinese RMB" braggadocio insofar as the other party doesn't even play that particular game.

Any greenhorn trader could've told you both these things without messing with SNA software.

Kindred Winecoff said...

As someone who *hasn't* worked as an FX trader, I must say that your experience as an expert is worth less than my couple of hours screwing around while watching baseball on a weekend.

First, your "In order to derive the GBP/JPY cross rate, for instance, you need to determine USD/JPY and GBP/USD" is irrelevant, b/c the data I've presented isn't about "deriving" the cross-rate, but about *actual trade volumes*. It's about what currencies really matter. And globally-speaking, there's only 1. Maybe 1 and 1/2.

Second, the Chinese haven't made the RMB fully convertible? You don't say. But I'm not the one hyping up the coming Chinese takeover of the global economy, like you do on your worst days and plenty of others do on every day.

The point of this post wasn't braggadocio. It was to look at what's happened in Forex in the past dozen years. I think I did that pretty dispassionately. And I think the conclusion in my last paragraph is pretty obvious.

If you disagree, then write your own post and open up your comments. I'm sick of you taking pot shots at me every other day while refusing to allow comments at your own place. Your "comments aren't necessary anymore" excuse is crap, since you are (by far) the most frequent commenter here. I'd suggest you either stop pot-shot commenting, or start allowing others to comment. Anything else is hypocrisy.

Put up or shut up.

Daniel Bunn said...

This may just be a programming nuance that I'm not familiar with, but why don't the currency labels stay in the same spot from chart to chart? The Yen goes from 12 o'clock to 1, to 9, to 4 then to 7. I think it would be easier to see the volume progression if they stayed in the same positions, but I'm not familiar with the programming you use to generate the charts. Other than that, it's pretty interesting.

Emmanuel said...

Sorry about enabling comments. They went dead at my place a long time ago and will stay that way. I've had a long time to think it over and my mind hasn't changed.

You kind of miss the point about cross rates. There is no way you can trade "third currencies" without involving the USD. In plain English, there is no way you can exchange a non-USD currency for another non-USD currency without involving the USD. Hence the observation that almost 90% of FX trades involve the USD. Of course USD volume is going to be the lion's share. It's a mechanical observation alike saying night follows day--no points there.

There's no need to be cranky here; it's a simple statement of how FX currency trading works globally. See an earlier post regarding Fred Bergsten on cross rates. It is not that hard to understand.

Kindred Winecoff said...

Daniel, It kind of is a programming nuance. Basically, the Fruchterman-Reingold graph model is a force-directed algorithm that emphasizes centrality, but does so through a "random" process. Basically, every time you plot a F-R graph it will look a little different, even if you're using the same data. Using different data, ie different time periods, and it will also change. The only reason why the graphs look similar but rotated is because the data are very similar and there's a small number of nodes.

Emmanuel, I know how cross rates work. Just as there is no way to exchange two non-USD currencies without involving the USD, there is no way to exchange the USD without involving non-USD currencies. What's true of the USD is also true of non-USD. You're right: that's a mechanical observation. But it doesn't tell us anything about what the composition of volume should look like. EUR/JPY is obviously possible; it happens commonly. It just happens a lot less than USD/EUR or USD/JPY. Just as EUR/JPY involves the USD, USD/EUR involves the JPY. So why is one so much more common than the other?

In other words, I'm not talking about rates. I'm talking about exchange.

The International Forex Network, 1998-2010
 

Share it

PageRank

SiteMeter

Technorati

Add to Technorati Favorites