Who Should Use a Factor Risk Model?


The ARC Commodity Model is a powerful tool to help many constituencies in the financial industry, trading and real economy. Some of the applications of the model are very straightforward and follow directly, some uses of the model are more nuanced and require a bit of clarification. This short piece details both common and novel use cases of a factor model.

Use Case #1: The Risk Manager

The risk manager in today’s world faces risks that come from many sources. The language of risk is not consistent across asset classes. Equity risk managers use a model based approach to risk. They are accustomed to seeing risk enumerated in terms of style, sector and other stylized factors.

The current commodities models which exist use primarily the concept of a constant maturity structure to evaluate exposures. In this technique, a spline or parameterized curve is used to interpolate the futures term structure. Using the interpolation, a set of constant maturity prices are used to generate a time series of constant contracts. The drawback of this approach is the attempt to find constancy in a product which has a lifecycle. The measures of risk are a slave to the curve or interpolation methodology chosen. There is also all sorts of variation that these techniques inject into the risk numbers which are not driven by market dynamics. Instead, you get noise that stems from the fitting procedure. More problematic is the fact that the fundamental assumption of using a curve (continuity and smoothness) is often not a feature of the underlying market. Any of the shoulder months in the energy markets, new crop/old crop in ags and so on, exhibit behavior which does not look smooth and continuous.

Additionally, the futures world uses risk models based around shift, twist and butterfly paradigm. There is nothing wrong with this approach except that none of these effects are “in the data”. These 3 factors are determined by a post processing of principal component analysis (PCA) results. This method involves rotating the first three PCA factors. There is much in this approach that is arbitrary. Two analysts working with similar data can get different results in terms of exposure.

Whether using the constant maturity approach or the PCA approach, one overarching problem is how to summarize a heterogeneous book of futures. How does one make sense of a book containing corn, wheat, ethanol, metals and crude? How does one notice a commonality of position in such a mixed bag? It is not easy.

The ARC Commodity model begins with the concept that all futures are created equal and are first class financial instruments. The ARC Commodity model is nesting, beginning with a high level breakdown between three sectors (energy, agricultural and metals) and styles. It then nests a break out of the three sectors into a multitude of subsectors. It gives the risk manager the ability to classify risk into common factors across all sectors and exposures which are sector or subsector driven. This allows the risk manager to spot a set of trades which are accumulating a Momentum exposure (for example). A factor risk model starts with an a priori set of factors. One cannot hide in the noise created by the vagaries of curve fitting or PCA. Moreover, the ARC Commodity model allows the risk manager to view his or her risk at two levels of aggregation without affecting the style exposures. The ARC Commodity model is a clear modeling paradigm which explicitly deals with the futures as first class instruments thereby allowing for a clear and unique exposure measure and risk calculation. Unlike the PCA approach the model also clearly delineates idiosyncratic from systematic risk. Some of the factors such as basis are typical for futures. Finally, the factors of the model are very similar to those found in equity models. This means that the risk manager can take the tools and intuition from the equity markets into the commodity world.

Use Case #2: The Commodity Swap Structurer/RIA/ETF Manager

The financial markets have shown a large demand for commodity exposure which is only expected to increase with the post COVID financial and monetary stimulus. Unfortunately, many exchange traded funds, mutual funds, systematic players and commodity total return swaps investors are stuck in a world where one or two major benchmarks are tracked.

The situation in commodities is not terribly different from the early 1990s in equities-one or two major indices set the mood for the entire market. In our investigations of some of these indices, we find all sorts of stylistic exposures which make these indices a very risky proposition. Certainly, benchmarking to these indices without considering their fingerprints with respect to momentum, volatility or trading activity exposures is tacitly making bets on these style factors. An investor seeking gold exposure typically gets more than they bargain for by employing the popular ETFs that exist in the investment universe.

The ARC Commodity model can be used by a manager to choose a benchmark that reflects their investment style more faithfully. The structurer of a swap or etf can get a more focused exposure to the market aspect they target. For example, an investor who wants pure precious metals exposure would be getting some of that plus a momentum exposure among other things by using ETFs. The ARC model can be used to directly get “pure” exposures to the sector or subsector with minimal other leans. Finally, the structurer can always compare their portfolio’s exposure, attributed return and risk to that of their benchmark using the ARC Commodity Model.

Use Case #3: Allocations for Blended Funds

When a capital allocator or fund of funds manager makes an allocation to sub managers, a commodity factor model allows them several instruments with which to make an allocation. First, it gives an indication of realized idiosyncratic return and it also gives a measure of how the volatility of the funds cleaves into an idiosyncratic and systematic component. Second, it allows the allocator to combine the funds in such a way as to eliminate (potentially) stylistic or sector bets which are not consistent with the overall strategy. The model becomes a tool to judge manager performance and build a portfolio which (theoretically) exceeds the characteristics of the constituents.

Use Case #4: The Physical Consumer and Producer

The producers of physical commodities typically use very sophisticated models to keep track of their production or consumption schedules. Furthermore, many of these players have market impact systems which help them place or consume the commodities without upsetting the market price.

The ARC Commodity model would complement these approaches because it helps the physical buyer/seller understand the interconnectedness between their target commodities and the broader market. This opens up new avenues for hedging and spreading. Furthermore, the understanding of style exposure dynamics can help the user gain confidence in selling or buying.

For example, a producer of ethanol might observe that while the price action for this product has not been terribly strong, the momentum exposure has been steadily increasing. This would suggest that the ethanol market has more ability to absorb output than they may have thought. Furthermore, they may notice that the deferred months (which they had recently avoided) are showing exposures that are more similar to the front month. This change might allow them to push some of their exposure to the back months and alleviate oversupply in the prompt month.

As another example, a grains and softs consumer notices that one or two of its commodities have recently shown strong price movements. While the hedging scheme they have in place is successful, it might be possible to further expand the hedging program to commodities which have similar exposures (but which haven’t appreciated yet). A model of market interrelatedness (such as ARC’s model) opens up many more possibilities for risk management.

Use Case #5: Statistical Arbitrage/Alpha Generation

Factor models in the equity space have been used for decades to create strategies. In fact there is nothing peculiar to equities and the same can apply to the commodity world. A factor model explains the returns of a group of assets by a set of factors and an idiosyncratic part. In the simplest form a factor model can be used for statistical arbitrage. At a high level one can monitor the idiosyncratic return of an asset, pair of assets or basket of assets and use any “out of the ordinary” return to be a signal to trade.

Factor investing has developed massively in the last decade in the equity world. Portfolios are created with a tilt on a factor such as value, quality, low volatility, momentum. The same can be done with the ARC Commodity factor model. In order to demonstrate this, ARC creates and tracks 3 factor tilted portfolios. The can be found here. Please note that this is not trading advice and these portfolios are displayed as an example of application of the model. More information about the methodology is available on request in the Library section (“ARC Factor Investing”).


In the cases above, we highlight some easy to understand applications of ARC’s Commodity Model. There are many nuances in the real world. However, the model allows the user to organize the information logically in a way that does not depend on the commodity under analysis or arbitrary modeling choices which vary from commodity to commodity. The commodity world functions in one sense under a great deal of opacity. Using a model helps to pierce that veil of uncertainty in knowledge. Please contact us for a discussion around how our model can help you.

It is important that visitors to our page read our disclaimer notification.