Yahoo Finance contributor Jake Moore is a former hedge fund manager for Caxton, Tudor and Citadel. He now runs a new firm called Kaiseki Systems.
When investors look to systematize their process, most begin and end with the same technique: trend following. This is particularly true in FX markets, as currency movements can seem like a mystery for long periods of time and trend following has appealing characteristics. I argue for adding fundamental indicators. For longer term investors this is essential, but it makes tremendous sense even for those who take short-term positions.
For background, I have used a wide variety of quantitative approaches to currency investing for many years, from long-term signals to build discretionary portfolios to fully automated intra-day models. I implemented these for leading hedge funds with substantial resources and for myself as an individual investor. Currently, I am building a systematic investment business and I have spent the past couple of years redesigning and amalgamating all the approaches I have ever worked on for use in the markets as I see them today. Two things separate the models I have developed.
First, I strongly emphasize the importance of position sizing, portfolio composition and portfolio risk. I will discuss this in a future post but suffice it to say that these are as important as the directional models.
Second, I place strong emphasis on using fundamental factors. The simple fact is that trend following systems alone have a small expected pay-off because they are relatively easy to implement.
Choose trend following for the right reason
I am not bashing trend followers, especially because I strongly believe in the trending nature of markets. The question is, how best to capture the potential for trends. To me, a standalone trend following-only system has a number of limitations. My biggest problem with trend following is that small investors choose it for the wrong reasons. Small investors gravitate towards the plethora of online FX brokers that have tight trading spreads, but high rates to hold positions for longer periods. Many investors outside the U.S. use contracts for difference (CFDs) or spread bets, or other similar highly margined FX platforms. These lead people to a short-term momentum or trend following strategy. The problem with these instruments is that cost of carrying positions is very large compared with the underlying FX forwards, or futures. This cost is particularly large in relation to the underlying volatility of the market: The cost for the leverage the brokers provide and the volatility of the instrument decays with the square root of time. This means that the carry-to-volatility ratio is substantially against holding positions for any meaningful period of time for small investors.
A reason that larger investors use trend following systems is that they are easy to understand and theoretically easy to implement. There are off-the-shelf systems and packages built into retail brokerage platforms — these are good starting points. The expected Sharpe ratio of a diversified trend following system that uses say, a new-high or low breakout system filtered with a moving average crossover, is around 0.9 if the investor can access a sufficiently large range of instruments and understands risk well. This strategy has a low correlation with long-only strategies, therefore, it has great diversification benefits.
However, in theory these models are easy to implement but in practice, less so, for smaller investors. First, trading a sufficiently large range of instruments to get a Sharpe ratio above 0.5 is operationally difficult and requires substantial capital. Sizing of positions is critical to implementation and this is very hard to do in smaller portfolios. This means that investors fall back on only a few markets, say leveraged online FX trading, which leaves them highly concentrated and exposed to the problems I mention above. Second, the drawdowns experienced by even the best trend following strategies are so big that it forces most people to arbitrarily choose times to use, or not, use the system. Third, because the methods behind trend following are straightforward, the competition pool one is entering when choosing this strategy is both huge and has some very large fish in it. Fourth, if you are inclined to hold positions for short periods of time, I would recommend looking at strategies that look for mean-reversion. Markets spend more time in a mean-reverting state than in a trending state in short time periods. Mean reverting strategies are harder to programme and have fewer practitioners — this is a good thing!
To improve a trend following or momentum system, we add in fundamental variables. Official economic data is easy to access and, if well understood, can add great power to a trend following approach. To keep things simple, I recommend starting with the simplest fundamental idea for currency markets, purchasing power parity (PPP). This is a good start with for a few reasons. Consumer price and producer price inflation series are released in a timely fashion and are not subject to major revisions (unlike nearly all other data). And price data get to the heart of what determines long-term currency rates and provide a proven long-term methodology for valuation.
PPP works because it makes sense. FX is nothing more than the relative prices between two countries. The change in a FX rate is the change in relative prices and is one half of the return of currency investing. The other half is the carry, equivalent to holding the relative return on two zero-coupon bonds. I like PPP not for a trading signal approach, but for a filter of long-term valuation. Before you say, this is too long term, consider this: The half-life of deviations from PPP is only around two and a half years per currency pair and there are many currency pairs. It is not difficult to find currency pairs that are a very long way from their fair value, a factor that often leads to the emergence of big trends.
I find it very important to have a good idea of valuation of an asset. In equities there is price-to-earnings, for bonds there is inflation and policy rates. In both cases, the asset is worth the present value of all expected cash flows. For FX there is PPP.
Having a clear framework for long-term value can provide a trend follower with a great deal of conviction when various factors coincide. For example, when the AUD/USD exchange rate was above parity my valuation metrics were saying that it had to fall. Similarly, when GBP/USD fell to 1.20 after Brexit nearly every market participant was screaming for further depreciation when my models were screaming value. The Japanese yen has wild moves around fair value and all of the big trends in the yen in the past decade or more have started from a position of extreme over- or undervaluation.
Power of valuation
This valuation backdrop adds a lot of power to a trend following system. But if the exchange rate is away from extremes of its historical distribution, at a neutral level of valuation, it is ‘fair game’ for other approaches and I recommend adding a number of logical extensions. Without an extreme PPP reading, we look at growth, inflation and policy shocks, with a momentum overlay. In currencies, regimes are very important and the relationship between a regime and the currency’s performance is determined by the current account position. However, at extremes, things start to get more interesting and we will be looking for a substantial multi-month trend. We will prime all systems to be ready. This helps solve another really big problem of simple trend following: How to choose between different signals. Most systems simply have a ‘long-flat-short’ system, with no differentiation between signals, meaning there might be 60-70 signals and positions with an equal (low) conviction. My approach uses half a dozen fundamental variables, including PPP, to provide a guidepost to find instruments that are more likely to trend. Success in the most liquid macro market often comes down to fine margins.
To build this framework into your system, a good place to start is with the OECD database. There are some technical issues with it, such as the starting point for valuation, but as a beginners’ reference this dataset is excellent. In our case, we build our own PPP models and use econometric techniques such as cointegration to provide the valuation. But a measure of deviation of the exchange rate from the OECD’s estimates will show the extremes – and this is what we are looking for. The truth about FX is that opportunities from meaningful trends only come a few times a year. These big trends are what we should be looking for when using a trend following approach at any stage of the investment process. They often start from z position of extreme valuation.