Carry Trade Fundamentals and The Financial Crisis

In one of his recent columns the Economist financial markets pundit Buttonwood declared the end of carry trade pointing towards the fact that the low volatility environment which characterised the glory days (2002-2007) of the strategy has now ended. In a way, I agree and in then again not quite. 

I agree that currency markets not as simple as they were in a pre-crisis framework where the JPY and CHF were rock solid low yielders that could be sold against just about everything for a tasty profit. In a way, the JPY and CHF as traditional funding currencies are still rock solid, as it were, in their role as low yielding currencies as both the BOJ and SNB have engaged in quantitative easing operations. However, they are not the only ones and in a world where the US, the UK, and perhaps now also the ECB are all flirting with the zero bound the picture gets a lot murkier. Add to this that the level as well as the variation in market volatility has increased markedly, it means that the waters navigated by carry traders have become decidely choppier.

One good example here seems to be the USD which seems, thanks to Bernanke's credible commitment, to be playing both sides of the fence acting as a traditional funding currency against the Euro while maintaining the role as "high yielder" against the Yen. Buttonwood also suggests that high yielding currencies from e.g. emerging markets ultimately may be driven by risk aversion which seems more than plausible and thus it would mean that in times of heightened risk aversion the big G7 currencies benefit from safe haven flows. This however may also be too simple since there are bound to be significant flows on the back of risk aversion in an intra-G7 context too which indicates that the safe haven flow argument cannot be applied uniformly. I think that monetary policy and those credible commitments matter here and as I have mused before, Bernanke may be paving the way for the mother of all carry trades if and when the green shoots really find root. But old habits also die hard it seems and as I show below, the CHF and JPY are still behaving very much as if they funding carry trade currencies in a developed economy context.

This is perhaps then where I don't agree with Buttonwood in the sense that one thing is if carry trades are making profits or not; quite another is when they make money and when they don't. Briére and Drut (2009) [1] show that in times of crisis carry trade strategies are thwarted by fundamental strategies and more importantly, Christiansen et al. (2009) suggests that carry trade risk premia are time varying. These two points are quite important I think since they indicate that the profitability and proliferation of carry trades are time varying. One could even stipulate that the extent to which the credit cycle, if such a thing exists, coincides with the business cycle, carry trades will tend to be pro-cyclical exacerbating the deviation from fundamentals one the way up and prolonging the downturn through the devastating effects of develeraging and unwinding of risky positions on the way down. The extent to which carry trades are driven by the inability of some economies, even in a global upturn, to close the yield gap with other economies makes this line of study quite important I think for international finance scholars. That however, will be for another day.  

More to the point, I asked recently whether carry trades were once again back on the table with reference to the ongoing discourse on green shoots and second derivatives in the market. Moreover I provided evidence to suggest that this was the case as market volatility has gone visibly down at the same time as risky assets have appreciated significantly, all this coinciding appreciation of funding carry currencies against key high yielders.

Now, in that entry I also said that I was doing a thorough rewrite of my previous paper on this topic and that paper is now finished as a first draft. Here is the abstract;

This paper takes the form of an event study surrounding the current financial crisis. It proposes a theoretical relationship which can be used to model traditional carry trade crosses on a daily return basis as a negative function of equity returns and a positive function of market volatility. In order to test this theory, an Arbitrage Pricing Theory framework is adopted which is used to estimate the factor betas of carry trade crosses with respect to equity returns and market volatility. It is shown how the variation in the currency crosses explained by the functional relationship as well as the estimated factor betas have increased significantly in relation to the financial crisis. The results indicate that low yielding currencies (the JPY and CHF) can be successfully modeled as a negative function of equity returns and a positive function of volatility in the market. The results furthermore underpin studies that have shown how carry trading activity is highly sensitive towards sudden sparks of volatility and risk aversion, and thus how carry trade fundamentals are time varying.

It is important to point out that it is effectively a new paper all together, but since it is heavily inspired by the old I have chosen to present it as a rewrite. The main differences with the old paper is that I am now trying to model currency crosses in stead of equity indices and that I am including market volatility as an explanatory variable. Moreover, the theoretical framework is tidied up significantly and now it allows for an incorporation of a time-varying aspect to carry trade fundamentals. In fact, this is a central pre-requisite of the paper and also why I choose to call it an "event study" since this, ultimately, is what it is. Finally and for the wonks out there, I have expanded the econometric framework when testing for structural breaks and the results are quite interesting although they need to be interpreted with some care on the quantitative level. Basically, these tests should not be used when you have heteroscedasticity, which I almost certainly have in the full sample regressions. However, most of the econometric sources I have consulted use these tests under heteroscedasticity anyway so I am a bit unsure what to conclude here.  

In any case, if you, by any fluke chance, read the other paper you would want to have a look at this one too. At least I think so.


[1] Please check the bibliography in the paper for the full reference for all the papers mentioned above.