This strategy belongs to the broad class of “Risk Parity” portfolios which seek to equalise instruments in a portfolio by “risk” rather than by US Dollar amount.
The “All Weather” part of a risk parity portfolio is to choose portfolio constituents so that one or more instrument or asset class is likely to thrive whatever the investment climate. Bonds usually thrive when equities fail. Commodities usually benefit from an inflationary environment, bonds do not.
The best explanation is to consider first the famous 60/40 portfolio and to explain how a risk parity portfolio seeks to improve upon it.
Risk parity investment is largely about wide diversification using assets which, if not uncorrelated, then are at least not too highly correlated and which contribute equally to a portfolio in terms of risk.
The 60/40 Portfolio
The 60/40 portfolio invests 60% of the total assets of a portfolio (in US Dollar terms) into stocks and the remaining 40% in bonds. Bonds and stocks have tended, historically at least, to be relatively uncorrelated. Even better, at times of crisis in equity markets, investors tend to move their funds into the relative safety of high-quality bonds, causing bonds to move sharply upward. The equity portion of the portfolio falls in value at such times but, often as not, the bond allocation rises in value. This is only true of bonds with a high credit rating.
The portfolio is re-balanced at regular intervals to maintain the 60/40 split in US Dollar terms. Hence the portfolio breaks down as follows:
A naïve but understandable assumption would be that your risk is 60% stocks and 40% bonds. Most professionals however consider the risk of an investment to be its volatility – the percentage amount by which an instrument varies in price on a day to day basis.
Volatility is usually measured as the annualised standard deviation of monthly price returns. On this basis, the S&P 500 has a volatility of around 13.53% measured over the last ten years (using the SPDR S&P 500 ETF Trust as a proxy). 7 to 10-year US Government bonds (using the iShares 7-10 Year Treasury Bond ETF as a proxy) have a ten-year volatility of around 5.58%
Measured in terms of contribution to total volatility, the risk is 78% stocks and 22% bonds:
The HCA All Weather Optimised Volatility Portfolio takes the simplest route to redressing this imbalance by dollar weighting the portfolio in accordance with inverse volatility. By doing this, each instrument contributes equal volatility to the portfolio. Again, the portfolio is re-balanced periodically to ensure that such weighting persists.
The next question to raise, rightly, concerns performance. In the long run a 60/40 portfolio can be expected to outperform (in absolute terms) a portfolio consisting of only 29% stocks and a large allocation of 71% to bonds.
With a risk parity approach the allocation does not remain fixed throughout the period. On each re-allocation date, the volatility of each instrument is calculated according to a look back period and the portfolio is re-allocated accordingly. In the back tests set out below we have re-balanced annually and used a look back period of 12 months to calculate volatility.
In the following statistics and charts, we compare the results of back testing a 60/40 fixed ratio portfolio with a risk parity portfolio and the benchmark S&P 500. Bonds are represented by the iShares 7-10 Year Treasury Bond ETF and stocks by the S&P 500 Total Return Index.
Note that all prices are adjusted to include dividends and to account for splits or consolidations.
As can be seen from the statistics above, during this period (and as expected) both the 60/40 portfolio and the risk parity portfolio underperformed the benchmark in absolute terms.
In risk adjusted terms however, the risk parity portfolio was the clear winner. The risk adjusted return was calculated by multiplying the absolute return of the risk parity portfolio by the ratio of the benchmark’s recorded volatility and that of the risk parity portfolio: 7.2% * (14.33% / 3.86%) = 26.73%
What this also indicates, is that you could theoretically leverage the low volatility risk parity portfolio to achieve a higher absolute return for a similar volatility to that of the S&P 500.
Regarding other risk measures, the risk parity portfolio in this test achieves the highest MAR (Managed Account Ratio) of 1.11. Calculated by dividing the CAGR by the maximum peak to valley draw-down. A useful rule of thumb calculation.
As can be seen from the correlation matrices above, the 60/40 portfolio is highly correlated to its benchmark. Another indication that most of the risk from the 60/40 portfolio comes from stocks. By contrast, the risk parity portfolio is relatively lightly correlated to the benchmark.
All Weather Funds typically boost their returns by gearing the bond portion of the portfolio, using futures or funds borrowed from their prime broker.
A similar effect can be achieved by increasing the average maturity of the bond portfolio, which in general may give higher returns and higher volatility. Long term volatility on a long bond investment (such as iShares 20+ Year Treasury Bond ETF) reaches similar levels to that of the S&P 500, at 12.77% for the last 10 years. This compares to 5.53% for the shorter-term iShares 7-10 Year Treasury Bond ETF.
Long bonds have usually moved sharply upwards during an equity market crisis and are, for much of the time, relatively lowly correlated to stocks. While correlations may always change, All Weather Funds have long relied on adding leveraged bonds to a risk parity portfolio, thus increasing returns while still achieving a lower volatility and draw-down than is seen on a portfolio of pure stocks.
By way of example, the following back test uses a risk parity portfolio consisting of the iShares 20+ Year Treasury Bond ETF and the S&P 500 Total Return Index.
As can be seen, the results seem to provide quite an improvement, at least in terms of absolute return. While in relative terms, the risk parity portfolio containing 7 to 10 year bonds still outperforms the portfolio containing the US Long bond, this latter portfolio achieves an absolute return not so dissimilar to the S&P 500, with much reduced peak to valley draw-down and volatility.
It would be as well to consider what effect rising interest rates may have on a bond portfolio. 90% of the return from bonds is from the coupon, but a rapid, prolonged and steep rise in interest rates might be expected to have an adverse effect on bond portfolios (particularly on longer dated bonds) until such time as the higher coupon can compensate for the decline in prices.
Back testing is, unfortunately, not an exact science. Nor can it accurately predict future performance. The returns from stocks and bonds (or indeed any other investment) are unlikely to be the same over the next ten years as they were over the past ten years.
The best we can do is to try to ensure that on a fundamental basis we are convinced by an investment.
Does the risk parity portfolio make sense? Will it continue to outperform the 60/40 portfolio or straight stocks? Is the use of inverse volatility weighting (as used in the HCA All Weather Optimised Volatility Portfolio) a reasonable method of attempting to control the volatility of a portfolio? The decision is yours – our role is to provide educational software, not investment advice.
The back-testing practitioner should experiment with different portfolios. He might try adding commodities or very short-term bonds. He might try including REITS. He might try sector-based ETFs instead of broad market index trackers. Perhaps he should experiment using a portfolio of individual stocks, or foreign (non-US) ETFs. Whatever instruments he uses, he should bear in mind the correlation of those instruments. Low correlated assets generally lower volatility and possibly also drawdown, although correlations may well change over time. The relationship between asst classes may not remain stable.
As to the other parameters for the system, experiment with them all. Explore different lookbacks to calculate volatility: what period makes the best sense to you? 60 days? 240 days? Some other period?
How often should you rebalance? This system gives you the options to re-balance daily, monthly, quarterly, and annually. Consider the relative merits of each.
What commission do you expect to pay on your trades? What slippage do you expect to suffer? Add your expectations to your tests.
Above all remember that back testing is based on past data and cannot predict the future. The biggest mistake is to fit the system to past data, and this can be done by cherry picking a portfolio as well as a set of parameters. Disappointment in the future will be the inevitable result.
The object is to provide a robust and loose picture of how a system might have performed in the past in the hope that future performance will not disappoint.
Learn about algorithmic trading and strategies, through
zipline backed research, backtest, and live trading.
When live streamng, ask questions...get answers...live!