Detecting Opportunities in Commodities

Yes you can use valuations for commodities, here's how...

Learning Goals

• Understand valuation concepts for commodity markets

• Understand indicator design in non-traditional asset classes

• Apply principles to identify risk and opportunities in commodities

Concepts

Following on from my previous post introducing the concept of valuations for market timing and risk management, this post focuses on the application of value signals for navigating the cycles in commodity markets.

The same basic concept applies – we want to buy low (cheap) and sell high (expensive). It’s of course more nuanced in practice, but that’s the key objective and the core purpose of valuation indicators.

When it comes to commodities, figuring out a reliable and sensible valuation indicator can be quite challenging.  For equities it’s comparatively simple: e.g. just use some variation of the PE ratio (and maybe the equity risk premium, price to book ratio, relative PE ratios, etc), maybe with some transformations, but typically using well-established, conventional, mostly uncontroversial and publicly available data.

For commodities it is difficult because there is often no readily available fundamental anchor against which to gauge or standardize and compare price (across time).

So this post is as much about concepts and application, as it is about indicator design and input selection.

Firstly, one reason for even looking at commodities in the first place as an asset allocator is that commodity markets tend to follow repeating cycles – just as stocks and bonds do.  This makes them ripe for position taking, but as an added benefit they can act as a useful diversifier, sometimes even a hedge – and often run out of sync with bond/stock cycles.

The chart below shows this playing out in practice, specifically by looking at the cycles evident in valuation indicators for stocks, bonds, and commodities.

As you can probably tell from the chart commodities present another option for diversification.  And by the way, in my view – from a multi-asset portfolio management standpoint you should basically ignore historical correlations and look only at the current forward looking risk vs return menu.

Here’s why:

As I noted in the first post on valuations, bonds were expensive in 2021, that was a key reason why they were a source of risk rather than a diversifier of risk (in contrast to the long-term correlation stats which show bonds as a negatively correlated diversifier – not in 2022 though!).  Commodities meanwhile were cheap – the macro scenario that unfolded in 2022 of course wildly favored commodities over stocks and bonds, but the clues or conditions that lead to commodities outperforming and diversifying against the drop in stocks AND bonds were in the valuation signals.

This tells us about the utility of valuations, the dynamic nature of correlations, and the need to think rather than just rely on simple headline statistics. And by the way, to be clear, I see valuations as a tool to support thinking, not a substitute for thinking.

That example and the chart above also highlight the value of having more levers to pull (e.g. as opposed to just stocks vs bonds).  Indeed, a value-driven investor will want to have a wide universe of assets at their disposal, and a systematic process of identifying high/low extremes in valuations to maximize their longer-term chances of success (where success = higher returns + lower/avoided risk), and obtain an outright edge over those just allocating to stocks and bonds. So commodities are certainly worth looking at.

But back onto the issue of valuations for commodities, we need a quantitative anchor that tells us what we need to know in terms of the forward-looking risk vs opportunities…

The solution is to start from the principle that a valuation indicator should be high/expensive when the market is high/peaking, and low/cheap when the market is low/bottoming.  In that respect we need to be pragmatic as practitioners – to the extent that you might even readily reject an otherwise fundamentally/theoretically sound indicator if it does not give a reliable and sensible signal. Similarly, you adopt and integrate signals that may rank lower on economic logic (the reason why it works may be less clear, less easy to explain), but rank highly on the sensible + reliable front.

Let’s spell those characteristics out:

Sensible = it gives the correct signal (expensive at highs, cheap at lows)

Reliable = it acts more or less the same way through time

Explainable = makes intuitive or theoretical sense why it works

The visual below summarizes it – it has to start with sensibility.  If the signal doesn’t say sell (expensive) at the highs, and doesn’t say buy (cheap) at the lows, then it is quite simply useless and unhelpful.  But it also has to be reliable… it’s no good if the signal repeatedly fails to correctly identify market peaks and troughs, or if it breaks down entirely and stops working (this does happen). 

Finally, it helps if it’s explainable… partly because eventually you might be asked to explain it!  But the other thing is you’ll probably start from the explainable end most of the time during the indicator design process (e.g. think about what makes sense, and then go get the data and see if it actually works).

Overall, coming up with an indicator and using it really is the same basic process for stocks or any other asset.  But we do need to exercise a greater deal of pragmatism and creativity as an analyst in the case of relatively non-conventional assets like commodities – while at the same time deeply understanding the mechanics of your indicators, and the usual + any other limitations it might have.

So let’s look at the data, and how it works in practice…

Data and Application

The chart on the left below shows the GSCI Commodities index and the composite valuation indicator for commodities.  On that note, just to clarify in this case we are looking at commodities at the aggregate or asset class level rather than individual commodities.

As you can easily see, the indicator hits extreme expensive (high) levels when the market is high/peaking, and extreme cheap (low) levels when the market is low/troughing.  The usual principles for using valuation indicators apply: the biggest risks and opportunities are found at extremes, and it’s important to be aware of the momentum information through the range (e.g. cheap can get cheaper, expensive can go higher).

But onto the more interesting part – the components. 

Firstly though, the reason why we use more than one input is we want to maximize signal vs noise, and enhanced the overall sensibility and reliability of the composite indicator.  The approach I use is to apply a simple unweighted average (giving each indicator an equal vote).

Into the detail, the first input is the commodities index deflated by the producer price index; this can be thought of as a fundamental-adjacent indicator: it has some explainability because all-else-equal commodity prices over the long-run shouldn’t wildly outpace the cost of production.  Thus on a real price basis, commodities should be, and are in practice: mean-reverting.

The second indicator looks at commodities vs stocks, this one ranks lower on explainability, but brings in useful signal shaping and helps strengthen the value signal for commodities vs stocks – as the next best risk-asset alternative.

The Log/Trend factor firstly takes the natural logarithm (to correct for the exponential growth of commodity prices over the longer-term), what’s left is a more uniformly trending series, and then we can just use a trend reversion model (i.e. (LN-price)/linear regression implied level based on the latest date).  One way to explain this one would be that because of supply constraints on earth and growing human population – over the long-term commodities should trend up… but there will be cycles around that trend due to demand (+speculation) cycles and longer-term supply investment cycles.  So any major deviation from underlying trend should be a signal to suppliers to adjust supply, and hence price will eventually move to reflect to that adjustment (with the usual principle of market overreaction applying).

The 3-year mean-reversion series is simply the latest price vs the 3-year rolling average.  This one works well in terms of sensibility and reliability – it also strengthens the composite signal (so there is basically a 4^th factor in composite indicator design which we could call signal amplification).  In terms of explainability, again it is less-obvious, but if we think about commodity prices as reflecting most of the fundamental information, then if prices deviate significantly up/down vs the past 3 years that’s unusual and contains information.  Perhaps most importantly, producers will probably be making decisions informed by that typical price level vs today’s price in terms of supply investment decisions, and consumers who are price sensitive may alter purchasing decisions – all of which will have an influence on the subsequent path of prices.

Basically on the explainability front it all comes back to the same principle for commodities, and we can distil much of this into a single visual.  Commodity prices go through fairly clear price cycles due to the underlying cycles in supply and demand.  Importantly, price sends signals to suppliers to either increase supply or pullback on supply growth – those actions then influence the future path of prices.  Or more simply, the cure to high prices is high prices (and vice versa).

Ultimately though, while we might be able to debate the explainability of each component, the composite indicator’s reliability and sensibility ranks high enough (it just works) that we can use this as an anchor to help set directional bias, identify risks and opportunities, and compare against other asset class valuations in managing portfolio level risk and return trade-offs.

Starting with that valuation anchor point we can then go and observe other factors like what’s happening with supply growth, how is demand tracking, what’s going on with monetary policy and currencies, and rounding that out with sentiment/positioning and technicals.

That in a nutshell is how to build maximum conviction and clarity in navigating risk vs opportunity in commodity markets.

Perspectives

In this section I turn to my mostly institutional investor following on LinkedIn to add some practitioner perspectives on the matter and as a prompt to consider some of the other practical issues in application.

Do you even invest in commodities?

A large proportion of those who responded noted that they do on a regular basis as a matter of course invest in commodities. Indeed anecdotally, a lot of asset allocators started making strategic allocations to commodities after/during the 2000’s super cycle – partly as a diversifier/hedge against inflation risk, but also due to attractive historical returns (it ticked a lot of boxes in the traditional mean-variance portfolio optimizations).

But aside from that, and I would say including that group, there is a large portion who I would say will opportunistically take risk in commodities (whether from the long or short side). And again that speaks to the opportunities (and risks) in commodities as they run through cycles.

A smaller portion say not now, not ever – anecdotally I know some won’t invest in commodities for philosophical reasons (e.g. ESG concerns), and/or perception reasons (don’t want to be seen to be punting risk in non-traditional assets).  But there’s also practical reasons e.g. lack of access to appropriate and cost effective exposure (commodities are different from stocks or bonds), or lack of proper analytical infrastructure and expertise.

Source: Callum Thomas (LinkedIn), Topdown Charts

Valuations for Commodities

Indeed, on that note, looking at the prevalence of those who use valuation signals for commodities – it’s quite small, only about a quarter.  From my discussions with clients, a big part of this is a lack of availability of established and sound indicators.  Few people have done the work on this, and I would say I’m one of the few to have worked and thunk about this extensively (and actually applied it over the past decade+).  This article is basically the culmination of that work.

There is also a view that you simply can’t value commodities, and it’s true you can’t use a discounted cash flow model or net-present value, but as I have shown – you can come up with indicators that give a consistent bounded signal as to whether the current price level is stretched in either direction.  And that’s the thing, when it comes to non-traditional assets, you need to adapt and adopt non-traditional analytical methods (based on sound principles that work across markets and time).

Source: Callum Thomas (LinkedIn), Topdown Charts

Reflections

When it comes to commodities, the core underlying principles for using valuation indicators is the same.  The objective of buy low (cheap), sell high (expensive) is the same when it comes to identifying opportunities and managing risk.  Meanwhile the pitfalls, limitations, and nuance are equally important to understand and factor in – just as with any indicator and strategy.

And while commodities are a unique and somewhat non-traditional asset class, the principles of indicator design are basically the same – you want an indicator that works (gives the right signal, and does so over and over again), and ideally you want an indicator that makes sense and has some economic logic to it.

Understanding these principles and objectives, and then understanding the unique drivers and characteristics of commodities, it’s entirely possible to come up with sound, sensible, and highly useful valuation indicators for this unique asset class.  And the benefits of doing so can be significant for those with the latitude and toolkit to venture into this space.

Lastly, in case you are wondering (since I have looked at this mostly at the asset class level), yes – you can apply the same principles and techniques from this article to individual commodities …which I have done (46 of them), albeit it does require nuance and attention to detail as some individual commodities require slightly different approaches (and have varying historical data availability).

But here’s what we can do once we have an indicator for all of them – breadth analysis!  Just as looking at the trend across components of a stock index can help bring in extra insight for the stock market, looking at the trend across individual commodity valuations can reinforce the asset class view.

The chart on the left shows the upper/lower quartile and median valuation across those 46 commodities over time.  The chart on the right shows the proportion of commodities showing up as either extreme (more than 1 Standard Deviation) cheap vs expensive. 

Using these further charts would have helped raise conviction and signal strength on valuations at some of the major peaks and troughs over the past few decades.  This is the principle of confirmation – to be clear, you want to be wary of “confirmation bias”, but as an analyst you need to build up and gather confirming evidence (and note any evidence that does not confirm, or even contradicts the core thesis).

Overall, the more tools and charts you can develop or acquire, the greater confidence, clarity, and conviction you can have in your conclusions.

Key Points

• The same basic concepts of valuation signals apply to commodities (there are cycles, the objective is to buy low (cheap), sell high (expensive)).

• Valuation signals can be used to identify risks and opportunities in commodities, and relative to other assets, across cycles.

• The key concepts for indicator design are: sensibility (does it work), reliability (does it work over time), and explainability (can you explain why it works).

• It is useful to combine indicators into a composite signal to maximize signal vs noise, and to improve the performance of the indicator (sensible + reliable).

• Taking a non-traditional approach to non-traditional asset classes, but informed by the same basic principles for traditional assets classes, can help generate valuable insights for market cycle analysis and investment decision making.

—

Best regards
Callum Thomas

Head of Research and Founder of Topdown Charts

Twitter: https://twitter.com/Callum_Thomas
LinkedIn: https://www.linkedin.com/in/callum-thomas-4990063/

---

Like what you see here? Check out our paid service for more insights and ideas

For more details on the service check out this post which highlights:
a. What you Get with the service;
b. the Performance of the service (results of ideas and TAA); and
c. What our Clients say about it.

---

Follow us on:
LinkedIn https://www.linkedin.com/company/topdown-charts
Twitter http://www.twitter.com/topdowncharts