Whilst much has been written about the rise in correlation recently — what’s been less frequently observed is the strange disconnection that’s occurring between correlation and volatility.
The two traditionally move together. That is, correlation tends to rise and fall with volatility.
Yet as FT Alphaville discovered — whilst working on a special report on the subject of how increasing correlation is impacting banks’ structured products desks — what’s really puzzling at the moment is why correlation is refusing to budge lower as volatility has fallen.
It’s a discrepancy that may be causing some pain for those incorrectly hedged in the market.
UBS’ Larry Hatheway, interestingly, picks up on the theme on Friday. As he observes:
One of the outstanding features of the crisis was a jump in market volatility. The other was higher correlation—across and within asset classes. Since then, however, volatility has generally declined, but not correlation.
He succinctly also expresses why it is that low correlation is an important factor for many portfolio managers generally:
Correlation between asset classes is central to the asset allocation decision. It determines the extent to which different instruments in a portfolio diversify each other—in other words how the addition of particular assets affects the riskreturn properties of the portfolio. Assets that exhibit low correlation help to moderate portfolio volatility, while assets with negative correlation can act as hedges.
Hence the pain.
Anyway. While many have tended, including FT Alphaville, to focus on more structural issues behind this rising correlation quandary, Hatheway’s team has so far been partial to a more macroeconomic explanation. As he continues:
We’ve argued previously that ‘riskon, risk-off’ trading behaviour, itself a function of elevated macroeconomic uncertainty, might explain extreme correlations.
But because they did remain puzzled, they continued to investigate — and this time they orientated towards the cross-asset themes that were emerging. The process went as follows, with the production of the below two charts:
In order to tease apart the correlation story, we begin by examining relationships between four asset classes (US equities, US Treasuries, global REITS, and commodities). Importantly, using simple linear correlation comes with a caveat. If there are a small number of observations in a sample, such as a five-year rolling window of monthly returns, then the correlation can be misleadingly high or low if there are outliers. Hence, in what follows, we prefer to use daily return data and rolling one-year long windows where available. We think the median absolute correlation is a good indicator as it is commonly quoted and works with larger datasets.
Now, in the opinion of UBS, what these charts most prominently express is that “something exceptional is happening”.
They explain:
Volatility has declined markedly over the last year, but correlation has remained at historically extreme levels. Volatility for all markets dropped as systemic fears ebbed, but absolute correlation, which started to rise rapidly at the same time as volatility, has not followed suit.
Meanwhile, in terms of cross-asset correlations, some historically significant relationship changes may also be under way:
The current high median absolute correlation is actually the result of three pairs of variables becoming more strongly positively correlated (equity-REIT, REITscommodities and equity-commodities in Chart 3) and the three pairs associated with government bonds becoming more strongly negatively correlated (Chart 4).
The distinction is important.
Risk-asset correlations moved to extreme positive territories, while the correlations of risk-assets to US Treasuries all moved to extreme negatives. This is a symptom of asset allocation shifts driven almost assuredly by economic surprises and entirely consistent with the observation that markets have been flipping between ‘risk-on’ and ‘risk-off’ mode. Otherwise, some of the results would be hard to explain, for example high correlation between REITS and commodities.
The correlation between equities and US Treasuries is particularly interesting. Most recently the two are strongly negatively correlated, but this is not always the case. From 1993 to the end of October 1997, the equity-bond correlation was large and positive. The correlation flips sign three times—going negative in July 1998, positive in September 1999, and then negative again in October 2000. From June 2004 until June 2007 the S&P and government bonds stay fairly uncorrelated then there is another sharp swing into negative territory in July 2007, where it has remained to the present day.
Determining the drivers, though, is not quite as easy as noting the trends.
In the end, UBS continues to fall back on a mixture of cyclical and macro effects (as well as changing risk perceptions) as providing the chief underlying causes for the phenomenon. And they conclude, in part:
The underlying macro environment is clearly important for returns, but these charts reveal its key role for cross-asset correlation as well.
—–
Unusual correlation is linked to a single risk factor—shifting expectations (indeed, fears) about the sustainability of recovery.
Full note (which includes the implications for asset allocations) available in the usual place.
Related links:
Risk on, risk off, risk on, risk off - FT Alphaville
Trading the correlation bubble – FT Alphaville
Stock picks tough as asset paths correlate - FT
Correlated returns, high-yield edition - FT Alphaville