Friday, October 15th, 2021

Currency First is Second Best

For the last few months, we have had a significantly higher exposure to equities in our euro model than our dollar model. Our commentary has also noted that the equity exposure would been much higher if we just looked at US, not global, equities. We are now in a situation where our dollar model is underweight global equities, while the euro model is still overweight. The two results are obviously inconsistent, even though equity exposure in the euro-model is now moving towards neutral. So, what’s problem and what can we do about it?

The problem results from the two-stage process we use to construct our models. First, we use our normal process to create two preferred portfolios – one for equities and one for fixed income and then we combine these two portfolios into one. This means that the performance of US Equities can be diluted by the relative weakness of non-US equities, especially when the dollar is strong. In recent years, this has been a much bigger problem than it used to be. For instance, in 2003-07, this approach, we which call Asset-Class First, resulted in higher exposure to equities because of the strong performance of EM Equities. In 1995-1997, we had the same result, thanks to the Eurozone convergence rally.

However, this is all ancient history. Our choice of portfolio construction is giving us a problem now, so what can we do about it? There are two options. We could abandon the two stage-process and create One Big Matrix, like we do for our equity sector models, or we could switch to a Currency-First approach. This is a different two-stage process, combining a portfolio of US assets, comprising US Equities. US Treasuries, US Investment Grade and High Yield, with a non-US portfolio, comprising all the non-US equity regions and EM Bonds.

There are two main problems with One Big Matrix. First, it allows the model to make comparisons which simply aren’t used in the real world. The vast majority of investors never consider switching directly from Japanese Equities into US High Yield or from US Investment Grade directly into UK Equities. They don’t have the mandate or the expertise to consider these issues. Second, it doesn’t work, because it over prioritises fixed income over equity. Each equity region has to produce better risk-adjusted returns than every fixed income category in order to achieve an overweight, and the portfolio is not allowed to benefit from the natural hedges available to a global equity portfolio, rather than an individual region.

The Currency First approach is better and is actually used by some US-centric investors, who regard all non-US investment as entirely discretionary and justified only if it produces superior risk-adjusted returns. However, there are still issues with this approach. First, it is not really applicable in other countries / currencies. Even the Eurozone does not have a large enough equity market or enough variety in its fixed income markets to allow an investor to replicate this approach. It is simply unthinkable in the UK or Japan, though it may, in the future, be useful in China, depending on how its corporate bond market develops. We would always need to use an Asset-Class First approach outside the US, even if we used a Currency First approach within it.

The other issue with the Currency First approach is that it underperforms over the long-term. Since inception in January 1996, it has produced an annual return of 9.6% compared with 10.2% for the Asset-Class First approach. Every other important statistic is slightly worse. Annualised volatility is slightly higher; the return on risk is lower and the maximum drawdown is a lot worse (-20% vs -14%) as is the average time taken to regain previous highs.

The reason for the long-term outperformance is that Asset-Class model has more flexibility when it really needs it. In extreme circumstances (both good and bad), the different portfolio mixes generated by risk-on vs risk-off are more appropriate than those generated by “dollar-on vs dollar-off”. The Asset Class Model is allowed to be 100% US Treasuries, which was very useful in 2002 and 2008. It is also allowed to be 100% in equities in whatever region it likes, which is almost inconceivable in a Currency First Model.

That said, there have been three easily-identifiable periods when the Currency model has outperformed the Asset Class model, from March 2003 to January 2008, from March 2009 to June 2013, and from March 2020 to date. This regime needs two preconditions to outperform: a global bull market in risk assets and easy monetary policy from the FOMC. A bull market in risk assets on its own is not enough (see 1996-99 or 2017-18) nor is monetary policy (see late 2002 and late 2008).

This begs the question of whether a three-stage model which was allowed to switch between the Currency and Asset Class models would be useful. The answer, based on past performance, is not really. All the three-stage model does is to underperform the Asset Class Model by slightly less than the Currency Model. There is never an extended period of time, when the three-stage model performs better than both of the two-stage models. It can recognise which regime we are in – Asset Class or Currency – but it can’t identify the turning points in real time.

So, even though the Asset Class Model is currently underperforming the Currency Model, we are reluctant to switch away from it, because we have no means of identifying when to switch back, other than inspired guesswork. But if we believed in inspired guesswork, we wouldn’t have created these models in the first place.

Conclusion: At first sight, this note may appear to a bit theoretical and self-referential, but we believe it contains three important lessons. First, it is a coherent and systematic demonstration of why exchange rates are a second order issue when constructing globally-diversified portfolios. Second, it tells us that we are currently in a regime where it is sensible to pay more attention to currencies than normal. Third, this regime will only last as long as the Fed maintains its current monetary policy, so it may stop working as soon as the first half of 2022.

Leave a Reply

Full Blog Archive:


Other Recent Posts