This is going to cover a lot of ground, hopefully well. But it’s simply too juicy to pass up discussing. One of my friends on Twitter turned me on to this post at FT Alphaville. The first thing you notice is the chart:
It’s kind of busy, but what it is tracking through time are EURUSD, the 2yr EURUSD swap spread and a sovereign CDS spread of unhealthy European countries vs. Germany. For once the CDS data is not the data series to focus on. For this post, we’re going to focus on the swap spread and exchange rate.
The FT Alphaville post explains the basic premise underlying the trade:
Funds trading the Euro-US dollar (EURUSD) typically use a pair trading technique pinched from equities trading models. The currencies model buys EURUSD on a statistical dip below the level of EURUSD implied by the 2-year swap spread. In other words, it assumes a mean reversion between the swap spread and the currency pair.
This is a slightly nuanced take on an effect defined by Irving Fisher many many moons ago called the International Fisher effect. The essence of the effect can be found in this equation:
What it says is that you can estimate the percent change in a currency’s exchange rate based on the interest rate differential between the two countries. I’m including a document that gives you some more details about how exactly this works and its complete with spreadsheets:
So what does this have to do with the EURUSD trade? Simple: first, you calculate an implied exchange rate by plugging in the respective 2yr swap spreads as interest rates. Then, you buy the EURUSD as the implied exchange rate goes below the historic mean. Buying the dips in this manner is a mean reversion play. And in the current environment, this is a huge issue.
Why? The problem with this trade is simple: the Fisher equivalency assumes that credit risk is constant between countries. Now let’s look at the CDS implied data in the chart. Looking at a variety of documents out there, you can see that assumption does not hold. Before there was a recognition on the part of traders everywhere that these credit risk issues are in fact somewhat differentiated, this trade made sense. There’s a rate differential between the ECB and the Fed, and based on return expectations, you would have expected the Euro to outperform the dollar. And as the Alphaville post points out:
The weakness of quantitative models (in general) aside, this is usually fine except for a couple things. For a start it’s been forcing the EURUSD to trade within a very specific range — exactly what the eurozone doesn’t need right now. Plus, the swap spread hasn’t decreased by very much since Germany continues to do reasonably well — or at least better than Greece and its porcine brethren.
In short, the quant model is at once increasing the cracks in the eurozone, and cracking.
Hmm… and as the chart shows, the ski slope down is a lot less steep now than it was 4 months ago:
Part of it has been reaction to the plethora bailout announcements, but part of it could also be model driven because a lot of people probably believe that if a bailout package is used, the credit problems go away.
They don’t. They just get delayed. The real issues are left to another day.
But as I pointed out earlier, the Alphaville post mentioned that these algorithms are buying on dips. Conversely, if folks start readjusting their algos to sell on the rips (i.e. selling at peaks after a run-up), there’s another issue:
It points to renewed waves of selling which will force the Euro lower. The feedback loop can work in both ways: exaggerate buying action on the upside but also exaggerating the selling action to the downside.
One other thing that I think this trading system may have touched – Treasury auctions:
At the end of the day these auctions had decent outcomes, but the question is whether or not they should have been stronger. In my mind, I think that if the Euro was weaker (i.e. we had an algorithm breakdown and selling had taken the Euro lower), the bids for Treasuries would’ve had stronger reactions due to the perceived credit worthiness of the US on a relative basis. In other words, the flight to quality bid would’ve played a bigger role. I haven’t done the research into this yet, but it’s an interesting avenue to explore.
But that will have to wait. I have another, more ambitious post to put together…