Strategic asset allocation defines the long-term proportion of individual assets that should be held in a portfolio to maximise risk-adjusted returns. While the ebb and flow of financial markets might dictate tactical deviations from the strategic allocation, it still provides a foundation for a portfolio, and is reflective of a reserve manager’s long term views and liquidity needs.
- In this section we present our analysis of whether adding gold to the reserve portfolio of a “typical” emerging or developing economy’s central bank can expand the efficient frontier and enhance risk adjusted returns.
- In order to conduct this analysis, we used New Frontier Advisor’s patented portfolio optimiser. The results of the optimisation exercise were unambiguously “Yes”.
- Choosing a conservative risk tolerance of 5 percent, the optimiser indicated that an optimal allocation to gold would be between 2.4 and 8.5 percent.
The following section describes our approach, the data and assumptions made, and the results in more detail.
(1) The data
The first step in any analysis is to define the “typical” portfolio. Currency composition of official foreign exchange reserves (COFER) data from the IMF provides a useful starting point, offering a breakdown of emerging and developing countries’ foreign reserves by currency. In the final quarter of 2009, 58% of the allocated reserves of those countries which reported were held in dollar denominated assets, 31% in euro-denominated assets, 6% in sterling-based assets, 2% in yen-based assets and 3% in other currencies.
After determining which currencies are commonly used in reserve portfolios, we identified the various asset classes within each currency that best matched the actual investment practices of developing and emerging market central banks. This was based on a variety of sources, including US Treasury data on holdings of US securities by foreign official institutions, the published investment guidelines of central banks and our discussions with industry experts.
Because we were using historical data as inputs for the expected real returns and risk profiles, we also needed a sufficiently long time series for each asset class, in order to avoid any period dependency effects.
In other words the period should not be unduly bearish or bullish for any one asset class. This required refining the list further to take account of historical data availability. The data sets we eventually chose, along with their real returns and standard deviations for the past fifteen years, are shown in table 1. All the data are displayed in dollars on an unhedged basis and all further calculations are made on this basis.
Selected Assets, Jan 1994-April 2010 - click to enlarge
At first glance the results do not seem to make an especially compelling case for gold as a reserve asset: gold has the highest annualised monthly return, but it also has the highest risk. As central banks are especially risk averse, the latter characteristic is likely to receive the most weight.
In addition, one could argue that the period under review – 1994 to 2010 - was particularly propitious for gold (the lack of a sufficiently long time series for some of the other data prevented us from beginning the analysis earlier). This is because the period encompassed the long gold bull market that began in 2001 (we address this point under “inputs and constraints”).
Returns and volatility are not the only considerations for risk-adjusted returns. The way the assets interact with one another, or the co-variance, also matters. On this basis, gold stands out as the most effective portfolio diversifier, displaying the lowest average correlation with each of the other assets. This reflects the unique drivers of the gold price.
(2) The optimiser
The next stage was to use a portfolio optimiser to assess whether gold’s low co-variance would offset the impact of its relatively high volatility on the portfolio’s performance. We used New Frontier Advisors’ patented portfolio optimiser which pioneered a technique called re-sampled efficiency optimisation (RE), to conduct the analysis. The RE optimiser has been acknowledged by Harry Markowitz, founder of Modern Portfolio Theory, to be more effective and robust than classical mean variance optimisation.
The mean-variance technique of portfolio optimisation pioneered by Markowitz is theoretically groundbreaking. It’s also limited in its practical use. If you are 100% certain of your risk return estimates, then classical mean-variance optimisation will provide the best solution. However, such input certainties do not exist in practice.
In RE optimisation, the optimal portfolio instead represents the average of a large number of simulated portfolios around the projected risk and return inputs. While the expected efficient frontier will tend to be slightly lower than estimated by classical mean variance, the level of confidence that one can have in the efficient frontier is improved. The resulting process is more stable and produces more reliable optimal portfolios.
Table 2: Correlations - click to enlarge
(3) Inputs and constraints
For the expected real returns and standard deviations we used the long-run historical averages shown in table 1, as is normal practice in this type of study. We made one exception. It was necessary to remain as conservative as possible with respect to gold - to avoid any criticism of the study having been conducted over a period that encompassed a long bull market. We therefore lowered gold’s expected real return to 4%, in keeping with its performance since 1974. We chose 1974 as the initial perturbations in the gold price following Nixon’s closure of the gold window in 1971 would have been negated by then.
The second constraint we imposed on the optimiser was to cap the allocation to US Agency bonds to 25%. We did this because preliminary optimisation runs had resulted in exceptionally large allocations to US Agencies relative to US Treasuries, based on the fact that the two assets had the same level of return, but Agencies had a lower standard deviation. However, this was inconsistent with what we knew about central bank’s actual allocations to these assets and the amount of each type of bond in issuance.
The Results
As we have explained, the results of this analysis were quite clear, that by including an allocation to gold in a reserve manager’s strategic asset allocation, the portfolio could enhance risk adjusted returns and push out the efficient frontier.
At all levels of Risk, an Optimal Portfolio includes Gold - click to enlarge
Selected Portfolios - click to enlarge
The optimiser then ran 1000 simulations around these inputs.
The efficient frontier and optimal set of portfolios for each level of risk are shown in figure 3. The coloured chart depicts the allocations of each asset class required to achieve the other optimal portfolios on the efficient frontier, which vary according to risk appetite. Some selected portfolios are shown in table 3.
The allocation to gold climbs steadily as risk tolerance increases. For example, at an expected annualised volatility of 8.3% the allocation to gold rises to 29%. However, as we assume central banks are typically far less risk tolerant, the actual strategic allocation to gold suggested by the optimiser for a central bank would be much lower.
For a risk tolerance of around 5%, the allocation to gold ranges between 2.4-8.5%. The RE optimiser also performs another function, which is to assess whether the allocation to gold is statistically significant. In this study, the optimiser calculated the allocation to be statistically significant at each and every risk level.
The results are not intended as a strategic asset allocation recommendation for any individual central bank. This is not least because central banks’ actual portfolios will differ from the one in the illustration, as will their risk and return expectations, and constraints. For example, a bank may wish to lower the proportion of dollar denominated assets due to concerns about future dollar weakness, or its allocation to Bunds due to the European sovereign debt crisis.
In addition, we have used a “base currency” approach, where the currency and individual asset compositions are determined simultaneously. Some central banks may pre-determine the amount of each currency they want to hold first, and then optimise each individual currency portfolio.
This illustration is simply intended as a starting point for discussions. In practise the allocations of emerging and developing countries – that already hold gold in their portfolios – vary widely. The appropriate allocation to gold for a central bank will depend on a combination of factors, including its investment policy objectives and guidelines, its existing asset mix, its risk tolerance, its tactical view on market trends and its liquidity requirements.