With the rise of robo-advice, people are beginning to comprehend how the practicalities will play out. Currency markets are often considered the enemy of a quantitative multi-asset investor, as they are incredibly complex and can often move in unexpected ways. The biggest problem is that historical relationships can easily break and correlation analysis can provide the wrong messaging.
There are countless ways a robo-model can approach such a situation, although the most common is to focus on long-term correlations between assets and ‘optimise’ holdings based on these relationships. In the case of currency markets this has significant scope for error, which we will help uncover.
Brexit Causes Issues for Robo-Advice
We are going to explain this using two key markets; the U.K. and the wider European market. What we want to know is how a robo-model would consider currency exposure in these markets. Taking the U.K. as an example first, we can isolate the relationship between U.K. equities and the GBP/USD. A robo-model is trying to quantify this relationship, typically into a single figure that helps explain the future relationship and guides positioning.
It is clear that the relationship between sterling and equity markets has changed over time, making it extremely difficult to ‘quantify’. In the 2002 to 2011 period for example, there was a moderately positive relationship between sterling and U.K. equities, which becomes particularly obvious in a five-year correlation chart. However, as time has progressed, the relationship has turned negative and shows no signs of stopping.
In a portfolio context, this raises a serious issue. It means US dollar exposure historically offered greater diversification benefits for portfolios with a high allocation to U.K. equities, and indeed, equities in general, than it supposedly does based on correlations today. Therefore, as we assess drawdown risks, the question is whether a robo-adviser is equipped to understand the variety of potential outcomes and whether it would prefer sterling or US dollar exposure.
For instance, if U.K. equities crash, would a robo-model expect a positive or negative relationship to sterling? The answer inevitable depends on the assumption used when building the model. For example, a robo-model that uses a long-term average would end up with a very different portfolio composition than a model that assumes a blend of the one-year, three-year, five-year and 10-year average. This is what we would consider to be one of the biggest risks in the robo-advice model and reminds us of similar ‘quantitative’ problems that ended up causing Long-Term Capital Management and its Nobel Prize winning leaders to disastrously collapse in 1998.
What About European Markets?
For euro-denominated investors, there is a similar issue at play. While the relationship between the euro and US dollar has been very stable in the past 24 months, an investor needs to consider what might happen to the euro if equities crash.
The historical relationship from 2004-2013 saw a similar relationship to that of the U.K. and had a moderately stable and positive correlation between the euro and European equities; i.e. the euro tended to rise during periods of market optimism. Yet, this relationship deteriorated during the euro crisis before reverting once again.
This adds an additional layer of complexity. It is clear that there is no ‘perfect’ method of understanding correlations. Take the example below, which shows the 3- and 5-year correlations between the EUR/USD and European equities. What would a robo-model consider to be the ‘assumed’ correlation? 0.26 is the long-term average, 0.29 is the 5-year correlation and 0.74 is the 3-year correlation.
Are Qualitative Methods Better?
It is worth noting at this point that there are no guarantees a model with qualitative input would do any better. For example, logic may tell an investor that the US dollar should offer defensive properties against U.K. equities in a crisis, however recent correlation analysis may suggest the opposite. How a qualitative manager deals with these issues are subject to behavioural biases that may have further adverse effects against any quantitative model.
To get around these issues, we tend to advocate increased awareness on drawdown risks, the risk of a permanent loss of capital, rather than volatility as the primary tool. This means we will conduct ‘quantitative’ scenario testing depending on different relationships but ultimately decide on exposures using qualitative input that acknowledges the valuation-implied return expectations of each asset.
It is also important to recognise the speculative positioning, where we add a contrarian preference to our analysis. This supports our ambition to invest in unloved assets, as this is where the best returns lie. Based on current numbers, the euro and sterling remain unloved relative to the US dollar, likely reflecting the threat of a Marine Le Pen upset in the French elections and growing optimism about Donald Trump’s economic policies.
The last point to acknowledge is the inflationary environment and what we consider the ‘multiple deflators’ for a currency. This works on the concept of the Big Mac index - we use the more technical ‘purchasing price parity’ model or ‘PPP’, where the cost of goods in various countries provide insight into the long-term direction of currency movement. On this basis, we have seen some subtle shifts in our conviction and while we remain cognisant of the risks, we continue to see upside for the euro and sterling relative to the US dollar.
