__________________________________________________________________________________________________
This is the second part of an extended research whose first part, entitled Commodity Portfolio Management: Structuring Considerations, was previously published on the J.P. Morgan Global Commodities Applied Research Digest journal.
It is worth reminding that the main goal of the present research is to identify the most efficient way to structure and manage a commodity portfolio constructed using futures contracts. The first part of the study showed that commodity markets are highly idiosyncratic in nature hence an equity-style approach to investment would result in a sub-optimal solution. Even another quantitative research entitled Commodity Portfolio Management, which inspired the creation of the current one, pointed out the dangers of treating a commodity portfolio “equity-style” stating that the impact of volatility and seasonality on energy, metal and agricultural markets is so specific that has to be addressed separately.
Structuring a commodity portfolio is not easy. Each asset class has specific idiosyncrasies, which need to be addressed separately because, as mentioned in the first half of the research, an equity-style approach to portfolio construction, allocation and exposure to risk factors is often a sub-optimal solution when it comes to commodity markets. There are several reasons for that but among the most important ones there are certainly the diverse nature of commodity market participants as well as the higher implied and realized volatility that these fascinating markets experience.
__________________________________________________________________________________________________
The present research compares and contrasts 3 filtered monthly, median returns extracted from the following futures markets: Brent Crude, West Texas Intermediate (WTI) Crude, European Low Sulphur Gasoil (diesel), New York Reformulated Blendstock for Oxygenate Blending (RBOB) Gasoline, Dutch Title Transfer Facility (TTF) Natural Gas, US Sugar Number 11, White Sugar Europe, Gold, Silver and Copper. Hence, the time series that will be analyzed are the following:
- Monthly, median returns extracted from a dataset ranging between January 2010 and January 2020
- Monthly, median returns extracted from a dataset, ranging between January 2010 and January 2020, excluding the highest top 25% returns
- Monthly, median returns extracted from a dataset, ranging between January 2010 and January 2020, excluding the bottom 25% returns
The rationale behind this comparison is to understand if, how and how much excess returns (top 25% and bottom 25%) skew median returns and consequently how much they can positively or negatively influence the overall investment performance of a hypothetical commodity portfolio. The practical application of this study has a remarkable impact on risk management strategies to adopt to limit unwanted market risk.
Let’s get started by analyzing energy markets.
Brent futures median, positive returns, from a seasonality standpoint, seem to be the highest in the month of April (0.33%) while the 2nd highest returns are usually observed in September (0.25%). Conversely, Brent futures median, negative returns are usually observed only in 3 months (May, October and November) with October usually yielding the lowest, negative returns at minus 0.16%.
However, it is interesting to note that if we cut the top 25% returns from the original time series (see Brent below 75 columns), the median, monthly returns drop dramatically and become all negative with the month of June yielding minus 0.58%. Conversely, if we cut the bottom 25% returns (see Brent above 25 columns), all monthly, median returns become positive with the month of June yielding positive returns as high as 0.79%.
In practical terms, this means that the vast majority of investment yields in the Brent market come from positive, excess returns (top 25%) because the average, day-to-day returns are too low to offset losses. In fact, if we cut the top 25% returns from the series, all average returns become negative. Conversely, the bottom 25% returns are the real dead weight of a hypothetical Brent-only portfolio because once we filter them out from our analysis the entire time series of average returns turns positive meaning that an accurate risk management strategy, as far as downside risk is concerned, can make all the difference. If we adopt a buy-and-hold investment strategy, the bottom 25% negative returns need to be avoided with a risk management tactic focused to limit the impact of left-tail sigma-3, sigma-4 and sigma-5 movements. Long-term investors seek to minimize left-tail risk while maximizing their exposure to right-tail risk by staying invested in the long term as it is nearly impossible to predict with accuracy when a daily return will be high enough to rank in the top 25%. Conversely, a long/short investment strategy will seek to maximize the portfolio risk to both upside and downside risks and since top and bottom 25% returns yield higher profits, focusing on volatility regimes and volatility targeting strategies would probably be the best approach. It is interesting to note that in the Brent market, given the high demand for this type of commodity, the average, positive return in the top 25% category (0.50%) is higher than the average, negative return in the bottom 25% category (-0.38%) suggesting that down-trending volatility regimes could potentially be more profitable than up-trending ones.
Let’s now move to the American oil benchamark.
WTI futures top, positive returns seem to be more probable in the month of April (0.29%) while August yields the lowest, negative returns (minus 0.13%). Nevertheless, it is interesting to note that WTI futures tend to yield negative, monthly returns more often than Brent. In fact, January, May, August, September and November are usually negative with the month of October usually yielding median returns equal to 0. This might be a consequence of the more speculative nature of the WTI futures market, which tends to be more heavily traded by financials while Brent futures have a more diverse distribution in terms of types of market participants.
Similarly to Brent, when the top 25% gets filtered out, all returns become negative with the month of September and January usually yielding the lowest: minus 0.68% and minus 0.58%, respectively. Conversely, when the bottom 25% are cut from the time series, all returns become positive with July yielding 0.71% and February, June, August and September all yielding monthly, median returns higher than 0.60%.
The WTI futures market, unsurprisingly, is similar to the Brent one when it comes to return distribution, hence, any investment strategy may be built in a similar fashion but the risk management of a buy-and-hold strategy should be constructed taking into account what observed in the previously explained analytics: excluding the top 25% of returns, the months of January and September tend to yield the lowest returns of the year implying that a higher hedging ratio covering 2-to-5-sigma left-tail events should be preferred as the risk of running in deeper losses is higher. Conversely, hedging ratios could be diminished in February, June and July when unfiltered, WTI returns are also positive. In any case, tail-hedging makes all the difference in a buy-and-hold strategy because market participants do not have the luxury to hop in and out the market but have to stay invested in order to bank on the top 25% returns and capture them when they happen. If we look at a long/short strategy, instead, volatility targeting continues to be a very good idea with the months of January, May, August, September and November being ideal to sell as they tend to be in a higher volatility regime while February, April, June, July and December seem to be better for a short gamma approach.
Let’s now expand our analytics to one of the most important natural gas benchmarks in the world.
Dutch TTF futures top, positive returns are observed in November (0.15%) while the lowest, negative returns tend to be clustered in the month of January (minus 0.26%) and December (minus 0.23%). Overall, it is worth pointing out that only 5 months (March, May, August, September and November) typically yield positive returns, which usually are much lower than those observed in Brent and WTI futures markets.
Dutch TTF returns, like observed in the previous markets, turn all negative when the top 25% get filtered out with the month of January, February, March and December yielding returns as low as minus 0.80%, minus 0.92%, minus 0.81% and minus 0.80%, respectively. Conversely, eliminating the bottom 25% skews all returns on the positive side with February yielding 0.73%.
Let’s start by stating that the futures term structure of Dutch TTF futures tend to be more affected by seasonality than the previously examined crude oil markets. This is predominantly given by the fact that both autumn and winter seasons tend to experience a higher demand of natural gas for heating purposes so monthly returns tend to be more impacted by the shape of the forward curve than crude markets. Please note that the time series has been deliberately not adjusted for seasonality in order to maintain the most “real-life, real-market” approach. In a hypothetical buy-and-hold strategy the months of March, May, August, September and November seem to have the highest chances of yielding excess, positive returns while downside hedging ratios should be increased more in January, February, April, June, July, October and December. It is important to point out that the seasonality effect not always plays a key role in determining where prices are going to go because average or in-line-with-consensus temperatures tend not to push prices dramatically higher and the higher amount of speculators now trading and investing in the Dutch TTF market smoothed out quite significantly the seasonal effect. Hence, in a long/short approach, instead, market participants would collect higher, negative returns in the months in which trend-followers need to increase their downside risk coverage while concentrating the buying activity for the months of March, May, August, September and November. All in all, even the Dutch TTF market presents the exact, same return feature observed in the crude markets: without top 25% returns, all average, monthly returns turn negative and without the bottom 25% returns, all average, monthly returns turn positive.
Gold futures top, positive returns happen usually in April (0.17%) while the lowest, negative ones happen in March (minus 0.13%). Gold is a particular market, in fact, the propensity to mean reversion is so high that 7 months out of 12 yield monthly, median returns very close to 0.
Despite the gold market showing a return distribution different from the markets examined so far, if top 25% are excluded, all its returns turn negative with September and December being as low as minus 0.34% and minus 0.29%, respectively. The exact opposite happens when bottom 25% returns are taken out: all monthly returns become positive with January, March and April usually yielding more than 0.30%.
It is worth mentioning that gold is a peculiar type of asset class because its demand is only partially driven by fundamental supply/demand dynamics and this has a significant impact on futures prices fluctuations. In particular, the gold market is often traded by Central Banks that use gold as a global currency for their foreign exchange needs. In other words, if a particular type of currency is needed more than others, a Central Bank can always sell some of its gold reserves to the market and use the proceeds to purchase the required foreign currency. Besides, gold is one of the most important commodities used by asset managers to hedge their equity portfolios against inflation and gold acts as a “safe haven” asset class when risk assets plummet. Consequently, all these buy/sell signals, that are not directly correlated with demand/supply dynamics, significantly alter gold price developments.
Nevertheless, even if gold is a different asset class, the top and bottom 25% returns continue to be the most important ones and capturing or hedging against those makes all the difference. Basically, from a buy-and-hold perspective, all months are actually good with the exception of March, September and November when market participants should be prepared to increase their downside hedging ratios. As far as a long/short strategy is concerned, it looks like there are more opportunities on the buy side, hence, market participants might want to run a predominantly short gamma portfolio.
Silver futures, despite showing some similarities in return distribution with the gold market, usually have top, positive returns happening in January (0.21%) and December (0.19%). Negative returns, instead, are usually clustered in the months of March, September, October and November, with the latter yielding the lowest, negative returns (minus 0.14%).
All returns become negative if the top 25% ones are excluded (see Silver below 75), with March yielding minus 0.47%, while the elimination of the bottom 25% (see Silver above 25) turn all returns positive with January, April and December yielding more than 0.50%.
Silver, although being similar to gold under certain aspects, has a broader industrial use thanks to the fact that, of all the elements on Earth, silver is one the best heat and electricity conductor and reflector of light. Hence, this market tends to be much more impacted by general performances in the manufacturing sector than just speculation and inflation hedging. Generally speaking, buy-and-hold strategies tend to perform better in the months of January, February, April, June, July and December. Conversely, March, September, October and November seem to be best for shorting futures, in the case of a long/short strategy, or incrementing hedging ratios to preserve the yields earned on the short gamma exposure of a hypothetical buy-and-hold portfolio.
The copper market also shows an elevated propensity to mean reversion with limited extreme returns on both sides. Specifically, the months of April, July and December usually yield between 0.09% and 0.10% while the lowest, negative returns tend to happen in May (minus 0.14%).
Unsurprisingly, top and bottom 25% returns are very important in skewing the distribution in one direction or another. Cutting the top 25% causes all monthly, median returns to become negative with February and November yielding minus 0.48% and minus 0.51%, respectively. On the other hand, excluding the bottom 25%, make all returns go positive with January, March, April, May, June and November yielding more than 0.40%.
The copper market is highly indicative of industrial performances because this metal is highly used for electrical wiring, building batteries and it is often a key component for almost any electrical equipment as well as electronics (like semiconductors, etc). Hence, its demand is largely due to industrial needs but also speculators such as hedge funds and investment management companies look at this market to hedge inflation and as macroeconomic indicator.
On average, from a buy-and-hold standpoint, the last 4 months of the year seem yielding the best performance, at least on average while the first half of the year brings a higher level of volatility so downside hedging ratios would have to be increased during these months in order to avoid undesired losses. As far as long/short strategies are concerned shorts tend to be more profitable in the months of January, March, May and August while longs perform better, on average, in April, June, July, September, October and December.
ICE low sulfur gasoil futures are usually skewed on the positive side (7 out of 12 months) with the month of February usually yielding the highest, positive return (0.15%) but it is worth noting that, although positive, many median monthly returns are quite low. Negative returns, instead, are rather limited and do not go below the minus 0.03% observed in both October and December.
ICE gasoil futures median returns turn all negative when the top 25% is taken away from the data with May yielding negative 0.70% while October and November return less than minus 0.42%. All median returns become positive, though, when the bottom 25% is cut off, with September returning 0.68% while February and April yield 0.53%.
ICE low sulfur gasoil futures, after crude oil, are some of the most liquid asset classes within the energy space (although only a small handful of speculators trade this commodity). The market composition and the price discovery process for this market is driven by multiple factors but refineries and hedge funds are definitely the biggest players in this market as large investment management companies tend to narrow their focus on Brent and WTI futures and don’t feel particularly at east trading fuel markets where liquidity is lower. Based on the outcome of the research, the months that provide best positive returns are February, April, July and September while short positions perform better, on average, in the months of May, October and December, which are also the months where market participants with buy-and-hold positions should increase their downside hedging ratios.
NYMEX RBOB futures median, positive returns usually happen in the months of February, March, April and May with February being the top performer (0.45%). Negative returns, instead, tend to be rather limited and most of the time range between minus 0.07% and minus 0.08%, except for the months of September (minus 0.25%) and January (minus 0.18%).
NYMEX RBOB futures behave like other markets when top and bottom 25% returns get filtered out. Specifically, if top 25% returns are cut, all returns become negative with the months of August, September and October yielding median, monthly returns equal to minus 0.80%, minus 1.04% and minus 0.72%, respectively. Conversely, if the bottom 25% gets cut, all returns become positive with the month of February yielding 1.02%.
RBOB futures, along to ICE low sulfur gasoil, are the most traded and liquid fuel futures. The composition of the market is quite similar to the one already mentioned for gasoil futures as investment managers, in general, prefer concentrating on more liquid markets (like crude oil) where their sizable positions have a lower impact and can be more easily “hidden”.
Buy-and-hold positions tend to return the best profits during the first months of the year with February and April yielding the best returns. On the other hand, if we look at long/short strategies, February, March, April and May are months where long positions perform best while shorts tend to be more profitable in January, June, July, August, September and October.
White sugar median returns tend to be negative 8 months out of 12 and the lowest, negative returns are normally observed in March (minus 0.26%) and August (minus 0.17%). Positive returns happen in April (0.01%), June (0.13%), September (0.19%) and October (0.01%).
White sugar futures returns become all negative when the top 25% are excluded with January, February, March, July, August and November all yielding less than 0.45%. Conversely, when the bottom 25% are cut off, median returns become all positive with February yielding 0.77%.
Buy-and-hold strategies are quite challenging in this market as June and September, despite yielding good returns, are the only months that seem to be having a consistent rise in price (perhaps October could be added to this list but volatility tends to be higher so its aggregate returns over time approximates 0). Conversely, the shorts tend to be more common in the white sugar market as January, February, March, July, August, November and December, on average, all yield negative returns.
Sugar 11 returns, similarly to the European sugar market, are negative 8 months out of 12 with top, positive returns coming in June (0.13%) and September (0.12%) while the lowest ones usually happen in August (minus 0.37%). Overall, the sugar 11 market tends to have positive and negative returns, which do not exceed +/- 0.15%.
Returns get all negative if the top 25% is filtered out with February, March, April and August all yielding less than minus 0.65% while they turn all positive if the bottom 25% get cut with April, May and September returning more than 0.57%.
The sugar 11 market resembles quite well the performances already discussed for the white sugar futures market. Overall, buy-and-hold strategies tend to be more challenging while shorts tend to be the preferred positions for 7 months (January, February, March, April, July, August and November).
All in all, the main takeaways from this segment are the following:
- Investment strategies are significantly impacted by returns in their top or bottom 25%
- When the top 25% returns are filtered out, all monthly median returns become negative
- When the bottom 25% returns are filtered out, all monthly median returns become positive
Conclusion
The present research is far from exhaustive but it highlights key points, which are crucial to efficiently structure an investment strategy for a commodity-based portfolio. The findings reinforce even more the need, already identified in the 1st part of the research called “Commodity Portfolio Management”, for a more specific investment technique. In fact, a deep commodity-focused investment approach is crucial to understand how changing portfolio weights on certain commodity markets can maximize portfolio stability or risk exposure depending on seasonality or volatility.