Is the 10-asset strategy still winning?

Using data going back to 1999, it was shown that the mix of 10 different asset classes would have outperformed a two-asset portfolio split of 60:40 between FTSE 100 shares and gilts (UK government bonds). As well as a higher annualised rate of return, the 10-asset portfolio was less risky, with lower volatility and smaller peak-to-trough drawdown.

Thanks to the advent of low-cost exchange-traded funds (ETFs), broad asset allocation is now widely accessible, and one year ago a portfolio was created to implement the 10-asset system. UK-listed products from the IC’s 2014 Top 50 ETFs were used and, although some of these subsequently dropped out of the 2015 Top 50, the original funds for each asset class are maintained in the portfolios.

List of ETFs used in the 10 asset portfolio

The performance figures for the ETFs below are total returns (not adjusted for inflation) over the date ranges specified, to give an idea of how the products themselves have fared since the first 10 asset system article was published and since the first portfolio was selected. Other ETF suggestions, for each asset class, can be found on the IC website.

Note: In the actual 10 asset portfolio, all returns are adjusted for UK Retail Price Index inflation. Rebalancing the holdings also has considerable effect on the collective portfolio performance.

Source: Bloomberg

Keeping the 60:40 equity/bond portfolio as our benchmark, the ‘live’ 10-asset portfolio has outperformed, although the past 12 months have disappointed in absolute terms. Periodic volatility spikes in equity markets and a tough 2015 for commodities weighed on the 10 part system as it lost 3.8 per cent in real terms. This is, however, better than the 60:40 index, which is down nearly 4.25 per cent.

If the performance of the ETF portfolio is measured from when the 10 moves article first appeared in January 2014, the 10-asset selection makes an average annual real return of 1.64 per cent. The standard deviation was approximately 6.1 per cent and the worst drawdown experienced was minus 8.4 per cent during the market turbulence of August-September 2015. The 60:40 benchmark has been beaten over two years, as it has a lower real compound annual growth rate (CAGR) of 1.03 per cent. The benchmark generally had a higher exposure to risk, with portfolio standard deviation at 8.2 per cent. Conversely, though, the worst drawdown experienced by the equity and bonds split was less than the 10-asset system, as it only lost 7.8 per cent in the nervous spell late last summer.

Head to head: 10 asset annual rebalancing versus 60:40 portfolio*

*31.01.2014 to 31.03.2016 with rebalances on 31.01.2015 and 31.01.2016. Adjusted for UK RPI.

In the asset management industry, the Sharpe ratio is ubiquitous as a measure of risk-adjusted returns. It is calculated by subtracting a risk-free rate of return from that of the portfolio and then dividing the result by the volatility exposure taken. The return on three-month Treasury bills is commonly taken as the risk-free rate and, in real terms, this was (contradicting the term ‘risk-free’) slightly negative over the timeframe analysed. The Sharpe ratio of the 10-asset system since January 2014 was 0.38. Per unit of risk taken, that means it performed almost twice as well as the benchmark, which had a Sharpe ratio of 0.21 over the two years examined.

Can the system be improved?

Although representative of poor recent runs for some of the investments held, so far results of the tradable 10-asset system are consistent with academic research that strongly advocates the benefits of diversification. With overall results underwhelming, however, a natural question to ask is: could returns have been improved by getting out of the worst performers?

Making judgement calls in real time contradicts the principle of a rules-based strategy; so to avoid emotional portfolio decisions, a well-attested mechanical method for side-stepping big losses is to get out of an asset every time its price ends a month below the 10-month moving average. Once the index posts a monthly close that is higher than this rolling figure, it is time to buy back in. The method can be modified slightly to reduce turnover: positions being sold once they fall below 95 per cent of the moving average and bought back in when the monthly price performance posts above the full rolling average number. When assets are sold, holdings are not redistributed across other funds. Rather, the proceeds of the sales are parked in cash, to be reinvested in the same asset when there is the signal to buy again.

10 asset/10-month moving average system

Underlying indices tracked by ETFs in the 10-asset portfolio pre-date the products themselves and back-testing can be done to early 2005. This is still a short time frame, limiting scope for meaningful analysis, but the period covered does take in the 2007-09 financial crisis, giving an idea of how the 10-month moving average system works under extreme circumstances.

Annual rebalancing is dispensed with, as the moving average rule already provides a mechanism to cut losses. As back-testing shows, the method also benefits from running winners for longer; the 10-month average portfolio achieves a real CAGR of 5.3 per cent since 2005, versus 4.3 per cent for the annually rebalancing portfolio.

The most impressive aspect is the apparent reduction in risk. The 10-month portfolio had volatility of 6.8 per cent and the worst drawdown was 9.6 per cent. By contrast, the annually rebalancing system demonstrated 9.3 per cent volatility and lost more than 27 per cent in 2008-09.

Does this imply that straightforward rebalancing should be dispensed with in favour of the 10-month average rule? With just over a decade of index data (and the products that enable the strategies are even younger), there is not enough information to build a powerful case either way. Clearly, the moving average rule was very successful in retreating to cash and riding out the very worst of the financial crisis. The drawback, however, is that in shorter sell-offs the rules can trigger assets' disposal only for their repurchase to be signalled within a couple of months. In the meantime, gains are sacrificed and unnecessary trading costs incurred.

Tendency to whipsaw in and out of investments is particularly acute in more volatile asset classes such as emerging market shares and gold. If volatility spikes, as seen in August 2015 and January 2016, occur with greater frequency then conceivably such behaviour might be observed in blue-chip shares. Similarly, uncertainty around interest rates could be a precursor to trades in and out of bond ETFs. The danger for the 10-month average system is that it will lead to over-trading which erodes returns over time.

At the end of March, there would have been no holding in the FTSE 100, MSCI Emerging Markets or the broad commodity index, so a portfolio from this date would, including the primary cash allocation, have been 40 per cent out of the market. The only way to really test a system is with a live simulation going forward. So from the end of April we will select a 10-month moving average portfolio to run alongside the annual rebalancing version. Time will tell whether the methodology guides smart chess-like portfolio moves or if it proves to be more akin to snakes and ladders.