It is generally good practice to keep internal correlation in your portfolio to a minimum – at least if it is intended for longer term ‘buy and hold’ investments. For short term positions, overall correlations are less important, particularly if you do not stay fully invested at all times and hold a significant portion in cash, but even then understanding how correlation can impact your overall portfolio is an essential part of understanding the markets. Correlation is not always obvious – during times of market stress, new correlations may appear, and all risk assets will tend to be correlated, potentially causing a nasty shock for the unprepared investor.
What is correlation?
When hedging efforts fail, it is often because the attempted hedge has a positive correlation with the hedged asset. That is to say, the price of your hedge moves in the same direction as the assets you are trying to hedge. Imagine you are hedging exposure to AUD/USD by opening an NZD/USD position. You hope that some of the risk of a downward price move in AUD would be offset by a stable or gaining NZD, only to find that the two assets are closely correlated, with both seeing losses, magnifying the overall damage to your portfolio.
This is an over-the-top example; most traders are aware that NZD and AUD are correlated, but it is also possible to accidentally discover correlations. Most risk-on assets: equities, emerging markets currencies, commodity-linked currencies, commodities themselves, high-yield debt and so on, are positively correlated during market turmoil, that is to say in a ‘risk-off environment’. During these periods traders rotate en masse out of risk-on assets and into safe havens, which may include gold, other precious metals, US Government bonds, the USD, CHF or perhaps EUR, and highly-rated corporate debt. It is this action that causes the correlation – so if there is a specific reason why one asset doesn’t follow the rule, the correlation may break down.
For example, during the 2012 eurozone crisis, the EUR, normally a safe haven currency, was considered a risk-on asset, a situation that continued for some years. This means it is not enough to simply be aware of each assets status as ‘risk-on’ or ‘risk-off’ as though it were a fixed property; you instead need to understand the dynamic conditions that cause markets to change, pushing assets first into one category and then another. Only a very limited number of assets never change: most equities will always be risk-on assets, although large state backed companies or extremely ‘boring’ stable dividend payers might sometimes be close to risk-off; and the USD and gold are fairly reliable risk-off assets in almost all market conditions. That said, it never pays to be complacent.
How does correlation normally work?
Perhaps you are happy with the difference between risk-on and risk-off assets, and want to understand more about the metric of correlation. A correlation of 1 means two assets increase together in perfect lockstep: the price of A increases from $10 to $12 and the price of B increases from $20 to $24 (both seeing an increase of +20%). A correlation of 0.5 would mean an asset increases in value by 50% for every 100% increase in the price of A; eg A moving from $10 – 12 and B moving $20 – 22. Negative correlations indicate an opposing price relationship: in the example above, if A and B had a correlation of -0.5, when A moves from 10 to 12 dollars B would decrease in value from 20 to 18, or 20 to 16 with a correlation of -1.
These correlations are of course only valid for a particular period, and are calculated using historic price data. If you extend or shrink the period you may find a lesser or even reversed relationship between the two assets, but when a high correlation is seen (anything above 0.5 is certainly high), it is unlikely that the relationship will break down completely. There are a few exceptions to this, however.
Correlation during market stress
In certain market conditions correlations can start to behave strangely. Fortunately as FX traders we do not need to worry about the various situations in which equities can start to behave as if they are 100% correlated (bubbles, mostly), but in market downturns new correlations often appear, both in the general pattern described above but also unexpected ones based on specific circumstances.
For example, the sudden repricing of the CHF following their central bank’s decision to abandon its peg saw significant demand for all other safe haven currencies – because one of them had failed, the others were now in greater demand and there was a corresponding flight to currencies such as EUR and USD. If the same situation had arisen with a risk-on asset, it is likely that the entire class would have lost value together instead. During these moments of tension in the market, correlations probably approach 1; with similar downward moves across the class.
After a period of market stress is finished, two safe haven currencies may return to a low correlation. The issue arises when traders load up on ‘uncorrelated assets’ as hedges when times are good, only to find that at the first sign of pressure all of their correlation -0.05 assets suddenly are correlated at 0.7 or worse. This brief spike may not be seen when calculating correlations over the very long term, as instances of market stress are rare (in some markets!).
Identifying uncorrelated assets
Finding uncorrelated assets is not as simple as it looks then: we need to be sure that two currencies maintain their lack of correlation – or negative correlation – in both normal market conditions and periods of stress. Historical backtesting can help us get some of the way, as you can check how the correlation performs over critical periods of a few months or a year (less than this swiftly becomes dubious – hour by hour correlations are irrelevant to all but the most microscopic algorithmic traders), and see how this related to the general correlation. Remember with a hedge the correlation in normal conditions actually limits your losses: a dream hedge would be a currency that maintains a low positive correlation during your wins that turns negative in losses. Of course such a thing is impossible, but its still nice to imagine.
What kind of strategies can be backtested?
Backtesting is best suited to technical strategies. It is possible to backtest fundamental strategies, but this requires far larger data sets. For example, if you use a fundamental strategy that goes long currencies with a positive balance of trade, and shorts those with a negative one, you will need a historical data set that includes balance of trade.
Many providers can help with this, but other trading styles are harder still. Lets say you are a news-driven trader, who tries to profit on overreactions to market data. To form a meaningful backtest you would need to download and analyse hundreds of old news reports, but do so in a way that is not coloured by your knowledge of what happens next. This is very difficult to do.
For this reason, most backtests involve simple technical strategies. The problem here is that technical strategies are the easiest to overfit. If changing the entry RSI level for a swing trading strategy by 0.5 improves performance, this is likely because of a few trades that were just missed out in the historical data set. It is not necessarily a good idea going forward. The advantage of fundamental strategies is they involve less fine tuning over levels and allow you to get the broad strokes of the strategy right first.
Better still is to combine technical and fundamental inputs. A strategy that filters currency pairs so that it only goes long on those with a positive international trade balance, then applies the 4 positive sessions test and exits on the first negative, is an ideal candidate for backtesting. Once running that strategy on a couple of historical backtests and finding satisfactory results, it is time to launch it in a demo account with a view to live trading straight after. Remember, no strategy ‘works’ until it is running live, and profitable.
Hedging correlation risk
Hedging a correlated portfolio can be tricky, and you should consider new positions in light of their correlation with your existing holdings. For example, if you go long (against the USD) in any currency that has stable inflation, rising GDP growth year on year, and which dips below 30 on the RSI, you could end up adding several technically sound but correlated positions in emerging markets currencies, which may perform well in a risk-on environment. However all of these positions will be vulnerable to a general economic slowdown, or any other circumstances leading to relative USD outperformance.
Hedging this concentrated correlation risk will require an understanding of different currency types; by adding some positions in safe haven assets, you will be able to minimise the effect of the concentration in emerging markets. This is a reliable correlation as the normal rotation away from emerging markets currencies would be into safe haven assets such as the CHF or EUR; therefore this trade will likely eat into returns in that period. It is worth trying to find an unusual safe haven currency that may have a positive case even during the risk on environment, and remember currency trades are two sided – you can simply short another currency against the USD to create a fairly reliable hedge to your short USD positions.
How Rakuten can help
Correlation is an important concept in trading risk management. If you want to test your new knowledge, why not open a free demo account and start trading today without risking any of your capital?