Wednesday, September 12th, 2018

Almighty dollar

We have a problem. Our asset US allocation model currently recommends a very low exposure to US Equities and this feels like a mistake when the index has just hit a new all-time high and the US is outperforming global equities. We know why it’s happening. It is the direct result of the way we have structured the model. We divide our investment universe into equities and fixed income, then we try to find the equity and fixed income portfolios with the best risk-adjusted returns. These change from week to week, sometimes gradually, sometimes quickly. Only when we have determined the equity and fixed income portfolios separately do we combine them by calculating the probability that equities will beat fixed income on a risk-adjusted basis.

Every investor knows that the dollar-denominated returns of non-US equities have been disappointing this year and that EM Equities are close to a bear market (a drawdown of -18% on a total return basis). Even with a very large overweight in US Equities (we are close to maximum overweight), there is no way that this can compensate for the weak returns from international equities, especially when they are penalised by the extra volatility generated by the currency market. In other words, we are out of US Equities because the dollar is so strong. Perhaps we should review the logic of asset allocation model and see if there is a different way of doing things.

We could, of course, scrap the dichotomy between equities and fixed income and have one large matrix covering the nine asset classes we follow (Equity – US, UK, Eurozone, Japan, EM; Fixed income – US 7-10 Treasuries, US High Yield, US Investment Grade, EM Sovereign Bonds in US$). The problem with this is that we would require the model to express a view on combinations like Japanese Equities vs US High Yield or Eurozone Equities vs EM Sovereign Bonds. In twenty years of advising on investment strategy we have never been asked our opinion on either of these combinations and we would hate for our model to be affected by the answer. Every matrix of this sort has a series of cells which are automatically set to neutral in perpetuity, and our approach is really no different.

However, there is another way in which we can group our assets, one which conforms to the worldview of many US – particularly retail – investors: the US and everywhere else. Our US portfolio would consist of US Equities, US High Yield, Investment Grade and 7-10 year Treasuries, while the non-US portfolio would comprise equities from the UK, the Eurozone, Japan and Emerging Markets and also EM Sovereign Bonds in US$. We could use a broader index of international government bonds but we prefer to use the identical assets classes so we get a direct comparison of the two approaches. This dichotomy makes sense for a US investor because they operate in the world’s reserve currency and they benefit from a wide variety of liquid debt and equity markets. It would be much harder to create a model with the same level of diversification in say, sterling or yen. From then on, we use exactly the same process as for our equity/fixed income model: separately find the best US and non-US portfolios, then combine them using our normal probability-based approach.

The results over the last 23 years are illuminating. The good news is that the currency-based approach produces significantly better returns than the benchmark portfolio, which automatically rebalances to a pre-determined fixed weight. These returns are better in both risk-adjusted and absolute terms and as with the asset-class approach, there is a significantly lower maximum drawdown: -21% instead of -38%. The not-so-good news is that the currency-based approach produces a slightly lower total return than the asset-class approach: 9.7% vs 10.3%. The volatility is effectively the same which means that risk-adjusted returns are also lower while the maximum drawdown is higher than the -11% produced by the asset-class approach. If we wanted to use one approach only, we would have to go with asset-classes rather than currencies.

However, there are times when the currency-based approach does better than the asset-class approach. The two main periods run from May 2003 to August 2007 and from April 2009 to June 2013. These coincide with two of the most powerful bull markets in global equities. But there is no one-for-one relationship between currency-based outperformance and equity bull markets. The approach underperformed during the Clinton-boom from 1998-99 and again during the synchronised global growth spurt of late 2016 and 2017. The big risk, however, is that there are periods of traumatic underperformance which coincide with equity bear markets from May 2002 to March 2003 and from October 2007 to March 2009. The basic structure prevents the currency-based model from shedding risk as fast as the asset-class model, which is the main reason why it underperforms over the long term.

Our view is that it is too difficult ex ante to identify periods when the currency-based approach will outperform the asset-class approach. The risk/reward ratio is also unattractive. A perfect hindsight model which identified the two periods of outperformance during the last 23 years would have added only 0.8% to the annual return, whereas switching into the two major periods of underperformance would have subtracted -1.3% p.a.

Be that as it may, fans of the currency-based approach could argue that it has outperformed since May 2018 and that this is the start of an extended period of outperformance like 2009-2013. Note that this could just as easily be associated with a weak rather than a strong dollar. Using this model, US Equities would currently comprise 40% of our global portfolio as opposed to 2% using the asset-class approach, while the exposure to EM Sovereign bonds and US Treasuries would be materially lower. The problems are that this investor would have been switching into US Equities from May onwards at close to an all-time high and that the cumulative outperformance is just 60bps over three months for a materially higher risk-profile.

One reason why the asset-class model does not like global equities is that even US Equities have not produced an attractive risk-adjusted return during 2017, despite the all-time high. (Award yourself an extra-point if you already knew that US High Yield has the best year-to-date Sharpe ratio).

Our view is that it is too difficult ex ante to identify periods when the currency-based approach will outperform the asset-class approach. The risk/reward ratio is also unattractive. A perfect hindsight model which identified the two periods of outperformance during the last 23 years would have added only 0.8% to the annual return, whereas switching into the two major periods of underperformance would have subtracted -1.3% p.a.

Be that as it may, fans of the currency-based approach could argue that it has outperformed since May 2018 and that this is the start of an extended period of outperformance like 2009-2013. Note that this could just as easily be associated with a weak rather than a strong dollar. Using this model, US Equities would currently comprise 40% of our global portfolio as opposed to 2% using the asset-class approach, while the exposure to EM Sovereign bonds and US Treasuries would be materially lower. The problems are that this investor would have been switching into US Equities from May onwards at close to an all-time high and that the cumulative outperformance is just 60bps over three months for a materially higher risk-profile.

One reason why the asset-class model does not like global equities is that even US Equities have not produced an attractive risk-adjusted return during 2017, despite the all-time high. (Award yourself an extra-point if you already knew that US High Yield has the best year-to-date Sharpe ratio).

In conclusion, we have a lot of sympathy with clients who argue that we should have a higher exposure to US Equities. It feels like a mistake not to be taking more advantage of the strong dollar, but years of experience have taught us not to abandon our process because of a temporary disagreement with its recommendations. We also know that our asset-class approach outperforms a currency-based approach over the longer term. In particular, it offers better and faster protection against a falling market. Other investors may “know” that there isn’t going to be a near-term correction in the dollar or US Equities. We are not blessed with this foresight.

Leave a Reply

Full Blog Archive:


Other Recent Posts