News & media Rethinking Commodity Diversification: Beyond Correlations and Linear Metrics
Conventional investment practice sees commodities as portfolio diversifiers and/or inflation hedges. The standard playbook relies on linear dependence metrics, such as standard deviation, correlation, and, ultimately, the Sharpe and Sortino ratios. But at least two issues undermine this view of commodity diversification: first, financialisation; and second, the overreliance on linear statistics (such as Pearson’s correlation) in portfolio construction. We propose a different angle on the commodity-diversification puzzle.
The Financialisation Effect
The financialisation of commodities has reshaped their co-movements with traditional financial assets, such as equities and bonds. Speculation, index trading, structural links to energy and oil markets, and safe-haven behaviour in precious metals have all strengthened the information shared between commodities and broad financial markets. This has been widely documented in the finance literature.[1]
Tail Events and Diversification Failure
Beyond financialisation, tail events act as another convergence force, especially for systemic risk. Tail events are rare realisations in the extremes of a probability distribution: unlikely, but with outsized impact. During broad economic crises and severe drawdowns, “low” correlations often vanish, and commodities can move in the same direction as equities. In other words, diversification may fail precisely when investors need it the most.
Understanding True Dependencies
There is another way to look at this. Information Theory studies how information is quantified, stored, and transmitted. Within this framework, Mutual Information measures how much information one variable, say, commodity prices, provides about another, such as equities. It is equivalent to the reduction in uncertainty about one variable when you observe the other. Crucially, it captures non-linear dependencies that standard correlation, Sharpe, and Sortino ratios miss.
When you build random portfolios with commodities, non-commodities, and blended mixes, you often find that most non-commodity and blended portfolios share similarly low information content, as measured by Mutual Information. In other words, a blended portfolio can convey the same information as a pure-equity portfolio. A counterintuitive result is that, in information-theoretic terms, a Pearson correlation of 0.80 (or −0.80) can be “closer” to 0 than to ±1. I developed this formally in a technical working paper (link: https://www.researchsquare.com/article/rs-8041591/latest).
Why This Matters for Investors
So why does this matter? Because an investor allocating to commodities (e.g., agriculture) with a diversification rationale based solely on correlation is likely using the wrong tool. That investor may cap returns on the false premise of diversification. Can agriculture diversify equity risk? Yes, but only on a case-by-case basis. This is why deal structure and on-the-ground expertise matter when investing. Whether the goal is diversification, return maximisation, or inflation hedging, the right analytical tools are needed before committing capital.
Data Meets Expertise
Cordiant has built these analytical capabilities in-house. Our team combines quantitative tools with sector-specific expertise in agriculture and real assets. This allows us to assess, case by case, whether a given commodity exposure truly diversifies equity risk, enhances returns, or improves inflation protection, rather than relying on headline correlations. In practice, this means structuring transactions grounded in data and on-the-ground fundamentals, rather than generic rules of thumb.
[1] See, for example, Tang, K. and Xiong, W. 2021. Index investment and the financialization of commodities. Financial Analysts J. 68(6).