Why asset correlations work until they don’t. Structural vs. temporary.

Portfolio diversification, is probably the first (sometimes only) advice investors receive, to calibrate their thoughts towards risk mitigation.

The concept is sound (although not flawless), and relates to the chaotic behaviour of markets during different economic cycles, and how placing all your eggs in one basket can lock a portfolio in situations difficult to manage.

At the same time, when certain asset classes fall, others rise, allowing for cushions against drawdowns and a more balanced approach to investments.

A big part of diversification, falls in understanding how asset classes behave with one another, in different environments, to know what to expect. In other words, knowing the asset correlations.

What is correlation and what does it actually measure?

Correlation measures how similar, different assets move at any given time. Mathematically, they are displayed using binary numbering from -1 to 0 to 1.

Here’s how. A correlation of:

+1 means two variables move perfectly together
0 means no measurable linear relationship
–1 means they move in exact opposite directions

Two assets can show high correlation simply because they are reacting to the same external factor, or they are traded by the same participants, or they are funded the same way. The problem is that none of these reasons guarantee the relationship (correlation) will persist.

Therefore, correlation measures co-movement, but not causation. Not knowing what triggers the correlation, beats diversification principles at their own court because correlation is backward-looking, it is time-window dependent (more on this below), and most importantly it does not explain why assets move together.

Correlations are reliable until they aren’t. For that we need to understand the two correlation types. Structural and temporary. These are two concepts that are often commingled, but behave very differently under stress.

Structural correlations

These types of correlations are present by default. The “mechanics” behind them are anchoring the relationships allowing them to persist. Think of a stock index future and its underlying cash index. Think of a bond price and its yield, or a currency pair and its interest rate differential, or an ETF and the basket it must physically track.

There is no changing of the relationship between a bond price and its yield, because mechanically they are inversely connected. The bond has a fixed coupon (interest payment), so a lowering of the price will always raise its yield and vice versa. A cash index is meant to track the behaviour of its futures counterpart, so they will always move together because they were built to do so.

These designed relationships that persist across different economic environments, can only break if the structure (their design) changes, and even if there are periods of short-term dislocations, they will reassert themselves.

Temporary asset correlations

These types of asset correlations are not designed or structured in any way. They are organically caused by the prevailing conditions and sentiment, and they are present because they share these conditions. They are created and driven by risk on / risk off behaviour, central bank policies, positioning, leverage, narratives or economic environments (inflationary / deflationary).

Think of stocks and crypto moving together, or tech stocks and long duration bonds, or commodities and emerging markets equities, or different asset classes reacting the same way to monetary policies.

Everything mentioned above is situational and heavily depend on the current environment. They can appear strong and stable one day, and break abruptly the next. These are not embedded or engineered relationships like the structural ones.

Why temporary correlations feel so convincing

Temporary correlations can create a false sense of permanence. They often coincide with strong trends, clear narratives and high confidence environments that seem obvious even to the trained eye. Correctly so though, but the trained eye knows better than to take numbers -1, 0 and 1 for granted, without looking at the bigger picture.

The bigger picture involves economies that move in cycles, and these cycles have short-term debt cycles that are embedded in long-term debt cycles, external and internal partnerships / conflicts, all of which exist within developed or under-developed countries as well. A correlation number speaks, but not loud enough.

Temporary asset correlations may feel reliable because when many participants hold similar trades, price movements align, so they tend to reinforce their positioning. Also deleveraging, forces multiple assets to move together amplifying them through leverage. Lastly, narratives create coherence even when mechanics are absent, explaining the relationships by stories.

How asset correlations break

Asset correlations often increase during crises, then collapse afterward. Not set in stone, but there are two clearly defined reasonings to how asset correlations break.

An oversimplified example would be to think how markets move together under monetary policy expectations. Then inflation suddenly becomes the dominant constraint and assets once positively correlated begin diverging.

So the dominant force that drove the relationship, changed.

In the second reasoning we need to think stress scenarios where participants sell what they can, but not necessarily what they should. All investors try to reason and time their decisions, but emotional drivers are another factor of erratic behaviour. This shared activity spikes asset correlations toward 1 and structural differences are temporarily ignored.

So liquidity overrode logic, evidently changing the relationship.

The most overused case example: The relationship between stocks and bonds

For decades, stocks and bonds often showed negative correlation. How?

Growth weakens; bonds rally due to their safer nature
Growth strengthens; stocks rally due to their riskier but rewarding nature

This relationship felt structural, and gave rise to portfolio allocation percentages like the 60 / 40 (60% on stocks, 40% on bonds to maintain balance). But the relationship was anything but structural. It depended on stable inflation, central bank credibility through predictable policy reactions.

When inflation returned as a constraint in multiple occasions, both assets fell together and the correlation flipped. The structure of bonds did not change. The dominant macro variable did.

Asset correlations vs. dependency

Asset correlations do not capture the situation dependence. It captures linear movement, but markets often exhibit non-linear dependency. This is highly related to whether markets are calm or stressed. They might uncorrelated during calm times, then correlate during stressful times, and then decouple again. This is not a statistical error but a reflection of conditional behaviour.

Markets behave differently when volatility rises, when funding tightens, when margins increase, or when risk limits are hit. These important rolling windows are not captured in asset correlations that say more about recent conditions, than how strong and enduring the relationship is.

Structural relationships that survive stress

Structural asset correlations tend to reassert themselves even after breakdowns. As mentioned in their definition above, they are designed to behave in a mechanical way. Futures prices converge to meet spot at expiration. Irrespective of contango or backwardation conditions, the design is convergence. Other examples include arbitrage relationships enforced by balance sheets and hedged instruments tied by contractual obligations.

In contrast temporary asset correlations, often don’t return once broken.

This explains why diversification sometimes fails and a wider narrative should always accompany a narrow one. It explains why “safe” bets on relations reverse and why sometimes historical asset correlations disappoint. As investors we constantly interpret, we don’t predict.

How should we think of asset correlations?

Instead of asking whether two assets are correlated, a more informative question is what mechanism forces this relationship to exist in the first place.

Find the answer in between by examining the relationship’s design and if it follows the mechanical approach (by default, legal obligation, settlement, arbitrage), it makes it structural and hard to break.

If the answer is more abstract and relies on rolling windows (policy decisions, narratives, risk sentiment, positioning, economic environments), it makes it temporary and prone to break.

Final thought

Asset correlations is a description, not an explanation. Understanding why assets move together is more durable than measuring how much they have done so in the past.

*this article is provided for educational purposes only. It is not to be construed as investment advice and all visitors are encouraged to keep learning while understanding that trading is exciting but also carries unlimited risk. Lets all be responsible investors to make the most out of the experience.