By James Williams, Hedgeweek – It is widely acknowledged that many CTAs have failed to deliver any meaningful performance over the last few years. The period 2011-12 was particularly tough. The Newedge Trend Index was down -7.93 per cent and -3.52 per cent respectively and despite market trends beginning to re-emerge in 2013, the Newedge Trend Index still only returned +2.67 per cent.
That said, one needs to look beyond the ‘average’ performance of managed futures strategies. Last year there were many strong performers. Lyxor’s own medium- to long-term strategy – Lyxor Epsilon Managed Futures – returned over 15 per cent in 2013 and has increased more than 35 per cent over the last 12 months to the end of November 2014. Its 2013 market performance of almost 17% was above most competitors as well, with the performance of the Newedge Trend Index standing at -4% Managed futures is certainly recovering its form, therefore, even though global economic growth remains delicate.
Fast gains, slow losses
“Trend-following strategies are one of the best diversifiers among alternative asset strategies and also one of the most robust strategies; if you look at drawdowns and volatility they stand up to scrutiny over the long term. When you look at long-term statistics there are funds with track records of 20, 30 years so you can review a lot of detailed information. The bulk of strategies have a Sharpe Ratio of 60 to 70 per cent and roughly 20 per cent maximum drawdowns.
“That said a trend-following strategy is not a silver bullet. From time to time trends weaken and these strategies naturally suffer and lose money. Very quickly though, when trends return, they are capable of recovering their losses. This is the way managed futures work; they make quick gains over trending periods and slow losses when trends dissipate,” explains Guillaume Jamet (pictured), Principal Portfolio Manager, Lyxor Epsilon.
Trend-following strategies are akin to insurance strategies that pay a premium – i.e. small losses over non-trending periods – so as to be positioned to reap substantial payoffs in favourable market environments characterised by trends and low correlation.
2011 and 2012: dissipation of trends
The primary reason as to why this was such a challenging period for trend followers is largely down to the fact that global markets were both range-bound and strongly correlated. Correlation amongst asset classes surged as investors were unable to make the distinction between risk factors for different assets and the market somewhat started to trade as one single asset. Such increase in correlation implicitly led to a reduction in the investment universe, which runs against the diversification principles on which trend-followers rely to generate returns.
Political headline risk and central bank intervention created rangy markets. ” In rangy markets, if you are looking at one asset there are two possibilities. If you are a long-term trend follower you won’t move your position very quickly or react quickly and performance ends up being range-bound just like the markets. If you are shorter-term you will be frequently allocating and re-allocating but not always in the right direction. You may lose money from churning and trading costs; buying after an increase and selling after a decrease.
“This is what CTAs were facing in 2011 and 2012. High correlation is bad for every type of trend follower. Short-term or long-term, neither likes it when there are fewer markets to trade,” says Jamet.
Navigating market volatility
Oftentimes one hears how large swings in global markets and high volatility regimes are advantageous to trend followers but Jamet is a little more cautious on this relationship.
“Intuitively, a systematic trend follower buys when markets increase and sells when markets decrease so the more volatility there is in the market the more likely the strategy will churn and lose money. A trend follower in effect ends up paying for realised volatility so they shouldn’t favour high volatility,” says Jamet.
For Jamet, the key is how to best incorporate volatility as a buy or sell signal in a trend-following investment models. For instance, to avoid the churning that Jamet refers to, Epsilon’s statistical models may build equity positions in the portfolio when volatility is still low in equity markets by detecting trends that emerge in other markets likes bonds, gold and which might signal a drop in equities (ergo, higher volatility).
“Volatility is not purely an equity play. Equity market volatility leads to continuous price moves in other markets and this is where trend followers make money. For example, in 2008 trend followers didn’t make money by shorting equity markets. They made money because gold and silver prices were surging, bonds were surging and so on.”
“There is a correlation between equity market volatility and other asset classes but not causality. You will lose money churning on equity markets where volatility is high.”
“That’s why I don’t buy the story of CTAs being used as an equity hedge. You can’t use a strategy that is uncorrelated to equities and then claim that you are equity hedged; you can’t be uncorrelated and negatively correlated at the same time. Managed futures is a diversifier. It’s not a hedge,” states Jamet.
Jamet and his team digested some key lessons from the poor performance of managed futures in 2011 and 2012 and without changing the long-term objectives of the statistical model they began to make enhancements to better cope with various market regimes. What resulted was the roll out of an alternative allocation model in September 2012.
Coping with highly correlated markets
As of the end of November 2014, the Lyxor Epsilon Correlation Index, a correlation indicator computed by Lyxor, had reached 19%, compared with a peak of 39% in September 2012, and an average of 27% since Lyxor computes this index in 2004. An index reading of 19% at the end of November means that the 65 financial markets included in the index were moving according to a similar pattern relative to each other 19% of the time over the 1-year calculation window of the index. One has to go back to 2006, prior to the subprime crisis, to find similar market correlation levels.
“During 2012 we reached almost 40 per cent, emphasises Jamet. This followed a first correlation spike of 36% in 2008 during the subprime crisis, and disappeared very quickly at the end of 2012, normalising during 2013 to around 20 per cent.”
“Back in 2006, 2007 when markets were de-correlated, taking a one-dimensional approach using different signals to decide whether to go long or short in a particular asset class was acceptable. However, when correlations are high, a trend follower needs to be more careful and adjust the risk allocation based on correlation expectations”, says Jamet
“Then, up to 2012, correlations were often not sufficiently taken into account in the asset allocation of trend following models. This typically ends up in higher non-rewarded risk taking since such risks can be diversified away. This is what we sought to improve with our risk allocation models in September 2012,” confirms Jamet. Jamet says that this is a key lesson he and his team learnt and is in fact something quite new for trend-following CTAs.
Revamping the statistical model
The Epsilon fund trades over 60 global futures markets. What Jamet and his team did in 2012 was update the allocation models to find a way to better account for, and effectively utilise, correlation dynamics between markets to construct the portfolio. This is easier said than done over such a large investment universe: given that the model trades 60 futures markets, there are literally thousands of risk factors that need to be considered in a 60×60 correlation matrix.
“This is the idea of diversification in a trend following strategy. Monitoring correlation is all about detecting which markets have the same risk factors and sizing your bets accordingly,” says Jamet. Markets that are driven by the same factor – as was seen in 2011 and 2012 – lead to increased correlation. The Epsilon model guards against this by sizing the risk allocation accordingly.
“In a high correlation regime ones positions in equities, bonds etc, behave the same; if they go wrong you’re going to be hit very hard. Just by knowing that is a benefit. You play the same game by allocating to equities and bonds but reduce the size of each bet. Maybe I find another signal and allocate risk to FX, to a number of commodities. The aim is to avoid having too much equity risk weighting and by default, too much correlation.
“We monitor ex post diversification not just ex ante diversification (before the event) to ensure that no single factor is consuming too much of the global risk of the portfolio. We presume this is a clear difference in our approach compared to other trend following strategies,” explains Jamet, who continues:
“Some markets will be driven by factors to a greater or lesser extent than other markets and it’s about dealing with that effectively in the portfolio allocation. What we don’t want to do is have a discretionary allocation to the model, a discretionary management of the parameters. When we built this model we wanted it to be as good as the previous model but make it work better when markets become more highly correlated.”
Given the Epsilon fund’s performance over the last twelve months at +35% as at the end of November 2014, the initial signs are that the model is working well. This means that the enhancement brought to the model did not prevent the strategy to perform well in an environment that is traditionally been favourable to trend following. The current correlation backdrop – at about 19% as illustrated by the Lyxor Epsilon Correlation Index – remains indeed very positive for trend-following trading.
“This level of diversification provides various sources of trading profit and makes CTAs less vulnerable to losses due to global market reversals” adds Jamet.
Yet, the next crisis is only ever around the corner. Severe market dislocations will continue to happen in future. The changes that Lyxor has made to its Epsilon fund, by taking on board the lessons learned in 2011 and 2012, have helped it to create a trend following strategy that is better equipped to handle today’s market conditions.
Bond markets and managed futures
Simply put, there is nothing uniquely special about the way bond markets behave that trend followers can use to their advantage. Just like any other asset class, once a trend emerges – and let’s face it, we’ve just witnessed a 30-year bond bull market – a CTA will look to ride that trend, be it up or down.
“By design there is nothing implicit in bonds that makes them more favourable. Regardless of the asset class, when trends start to disappear CTAs pull back on their positions; we saw that last May when the Fed starting to talk about the end of QE.
“Most funds responded the same way and it took one month to get out of US treasuries,” comments Jamet.
This is when the yield on US 10-year treasuries rose from 1.6 per cent to 3 per cent. Later in the year there was again a recovery on bonds, between Sep and Oct 2013 when yields fell back to 2.5 per cent “and Epsilon’s model went long bonds again”, concludes Jamet.