Sponsored content from Bloomberg

Commodities have long been a staple of multi-asset investors. Traditionally used to diversify exposure to fixed income and equity holdings, they are more recently also a source of alternative risk premia. Whatever the use case, the desired feature of any commodities allocation is some combination of attractive performance, sufficient liquidity, and a transparent methodology. 

The literature covering optimal portfolio construction is extensive. Focused primarily on equities and fixed income, it covers topics such as balancing risk and combining alternative and asset premia. Commodities have been less well covered. Allocations to the asset class are relatively smaller and the variable liquidity of individual commodities can make it difficult to deviate significantly from standard benchmark weights. 

Risk premia strategies, also referred to as factor investing, seek a positive long-run expected return as compensation for taking on identified risks other than the traditional market beta. Initially constructed as long/short portfolios, they have increasingly been used to systematically enhance long-only portfolios in the form of tilts from market value weights. 

Among commodities investors, carry and trend following tend to be the most popular risk premia strategies. Trend following is an artefact of investors’ behaviour, which include the bandwagon and disposition effects and the participation of non-economic players. Signals are used to identify price patterns over different frequencies, based on which long/short positions are entered into. Trend strategies are characterised by modest long-term Sharpe ratios and a positive skewness of returns. The carry trade involves taking long/short positions in high/low yielding assets. The returns profile is similar to that of an insurance underwriter. Small and steady gains are accrued over extended periods, which are interspersed by sudden, sharp declines when a risk, such as market sentiment turning bearish, materialises. 

Using commodity sectors as building blocks, we outline a framework to construct liquid portfolios that both balance risk and incorporate alternative risk premia. A standard use case would be asset allocators seeking efficient sources of commodities returns.

Asset class characteristics

Market weights in commodities can be viewed as some combination of liquidity measures such as open interest and production measures. One such benchmark is the Bloomberg Commodities index (BCOM), which is comprised of 23 commodities. The weight for each commodity varies considerably: in July 2019, constituent weights ranged from 1% (KC Wheat) to 13% (Gold). It can be challenging for large investors to allocate assets to individual commodities based solely on trading signals, such as price trends, while not taking into account their weights in a broader benchmark index. 

Commodities are typically incorporated into a multi-asset portfolio to diversify equity and fixed income risk. Since the mid 1970s, the long-term correlation of commodities with US equities and fixed income has been 0.2 and –0.1, respectively. Commodities are also used to hedge inflation. Given that commodities are real assets and are included in consumer price indices, rising inflation is often reflected in higher commodity prices. This is especially the case for commodities in the energy sector. Regressing the returns of the BCOM and BCOM Energy indices on quarterly changes in CPI, we get slope coefficients (or ‘inflation beta’) of 2.4 and 5.5, respectively. Both slopes are statistically significant. 

sector statistics

Commodities can be classified into four broad sectors – energy, industrial metals, precious metals, and agriculture and livestock. The key drivers of returns are store of value (precious metals), industrial production and growth (energy and industrial metals), and harvest cycles and weather (agriculture and livestock). Correlations between sectors are low and stable – ranging from 0.15 to 0.3 over the past 30 years. Correlations within the sectors tend to be higher – ranging from 0.2 between agriculture and livestock to 0.7 between energy and precious metals. Both the intra- and inter-sector correlations suggest we can view sectors as reasonably homogeneous building blocks. Since sectors are comprised of market value-weighted commodities, they provide a solution to the liquidity problem posed by individual commodities. In turn, liquidity facilitates the construction of portfolios that deviate from market value weights. We use the BCOM sector indices as proxies for sector returns.

Financing and storage costs and the hedging behaviour of commodity producers determines the shape of the futures curve. When the curve is upward-sloping (the price of a distant future is higher than one closer to expiry) the curve is said to be in contango. When the opposite is true, the curve is in backwardation. The difference between the excess return and spot return is called the ‘roll return’. When the curve is in contango (backwardation), the roll returns are negative (positive). The differences in sector dynamics are evident through the roll returns (figure 1) and are important when constructing a carry strategy.

The systematic harvesting of roll returns is a carry strategy and there are two implementations. The curve trade takes long/short positions at different points of the curve for the same commodity to account for differences in local slopes. A backwardation strategy positions long/short in different commodities, at the same point in the curve.

In the subsequent sections, results are discussed for the period 1980–2019 and performance statistics are in total returns unless stated otherwise.

Balancing risk

In the absence of strong views on sector performance, allocating between commodity sectors to balance risk makes sense. There are several ways to construct such a portfolio – ranging from equal volatility contribution to incorporating the covariance matrix, such as in the risk parity portfolio. Given the stable, low-to-modest sector correlations, a straightforward approach is allocating based on (inverse) sector volatility. Under this approach, each month sector weights are assigned based on the inverse of their trailing historical volatility, calculated over the past year. 

Relative to the BCOM index, risk weighting sectors effectively redistributes weight from energy to industrial and precious metals (see figure 1). Performance statistics (figure 2) show the annual total return of the risk-balanced portfolio (RBP) is 0.3% pa lower than the BCOM index. However, the accompanying reduction in volatility results in the same risk-adjusted return (0.19). The portfolio drawdown is reduced while the rolling portfolio volatility reveals a more stable profile. For asset allocators, these features can help size the commodities exposure. Also important are the unchanged diversification properties of the asset class. Reducing the energy allocation leads to a more negative skew since energy tends to have large upside price shocks. The inflation beta is still statistically significant but declines to 1.8.

Introducing risk premia tilts

We can use carry and trend-following signals to introduce alternative risk premia to the risk-balanced portfolio. The four sectors are ranked based on historical returns (trend) and slope of the futures curve (carry). Starting from a baseline allocation, the constituent weights are then modified to reflect the relative factor scores. This results in a final portfolio which is long-only, fully invested and which combines asset class and alternative premia. We consider three cases: a trend signal, a carry signal and a combination of carry and trend (multifactor). The sector-based factor scores are a weighted sum of the individual commodity scores.

measuring performance

Case 1: Trend

Using the excess returns series, five individual trend signals measure the change in the price level (ie, –1/+1) over windows spanning 1-12 months. The aggregate signal for the sector is the average of the individual signals. Sector exposures in the RBP are then reweighted to reflect the sector trend signals. A higher dispersion in the signals translates to a larger deviation from risk weights. 

Relative to the RBP, returns and volatility increase (by 1.4% and 4.4% respectively), leading to a moderate improvement in risk adjusted returns (0.3). Asset class correlations remain unchanged and continue to provide diversification benefits for a cross-asset investor. The inflation beta remains significant (1.8) and, in keeping with one of the main characteristics of trend-following strategies, the skew of returns is positive (0.4).  

Case 2: Carry

Used as a proxy for expected roll return, the backwardation signal measures the slope of the futures curve using the price of the nearby contract versus the corresponding calendar contract 12 months away. As noted earlier, a larger positive (negative) slope implies a more negative (positive) roll return. Similar to the trend signals, we tilt the equal risk weights based on the slope; the tilting procedure overweights sectors that display a more negative slope. Figure 3 illustrates both the extent of the dispersion and the time-varying nature of the sector slopes.

Annual returns increase by 1.4% over the RBP, with a similar improvement in risk-adjusted returns as the trend case. Portfolio volatility is higher (14.2%), though not as much as with trend. The inflation beta and asset class correlations are virtually identical.  

Case 3: Multifactor

Is there a benefit to combining the carry and trend signals? Combining the individual signals results in a portfolio with the same risk-adjusted returns (0.3),  inflation beta, and asset class correlations. However, annualised returns are between the two stand-alone factor portfolios (4.6%), while the returns skew is positive and similar to the trend portfolio. 

mapping sector slopes

Exploiting the curve

Commodities curves spend a majority of time in contango. During these periods, the shape of the curve is concave, which means investing in a deferred contract can potentially reduce the negative roll return (ie, there is less negative roll down). A simple way to enhance returns is to invest in the futures contract nearest to the three-month point instead of the nearby contract. Starting in 1991, this increases the annualised returns for the RBP and tilted portfolios by approximately 2.5% per annum, with little or no increase in volatility (see figure 2).

The transparent framework discussed here to construct liquid commodities portfolios can be used by multi-asset investors to enhance returns. The results of incorporating risk premia factors are governed by requirements such as tracking error, turnover constraints, and regulatory frameworks. Using this approach, investors can customise portfolios to suit their needs.