Correlation risk is the danger that positions expected to diversify one another move together strongly enough to create a larger combined loss. It matters because a portfolio can contain many instruments and still depend on only a few underlying drivers.
Correlation is useful for describing how returns have moved together, but it is not a permanent property of two assets. Historical relationships can weaken, strengthen or reverse as market regimes and economic conditions change.
Correlation measures co-movement, not causation
A standard correlation coefficient ranges from -1 to +1. A value near +1 indicates that two return series have tended to move in the same direction together; a value near -1 indicates opposite movement; a value near zero indicates little linear relationship in the sample.
Correlation does not explain why the assets moved together. Two markets can be correlated because they share an economic driver, because one influences the other, or simply because the relationship appeared in the sample.
Zero correlation does not mean independence
Two variables can have little linear correlation and still share nonlinear or conditional relationships. They may also become strongly related only during certain regimes.
For risk management, the question is therefore broader than “What is the correlation coefficient?” It is “Under what conditions can these positions lose together?”
Hidden correlation often comes from shared factors
Different-looking positions can depend on the same underlying factor. Several equity positions may share broad market beta. Multiple currency trades may contain the same dollar exposure. Commodity producers and commodity futures may respond to a common price shock.
Grouping positions by economic driver can reveal concentration that is not obvious from ticker symbols alone.
Historical correlation can be unstable
A correlation estimate depends on the sample period, return interval and market regime. A relationship calculated over one year can differ from one calculated over one month, and both can differ from the relationship during a sudden market disruption.
This does not make correlation useless. It means the estimate should be treated as evidence with uncertainty rather than as a fixed law.
Correlation changes portfolio exposure
If two positions are strongly positively related, holding both may add less diversification than their separate names suggest. If they are negatively related, one may offset part of the other's movement, but that offset is not guaranteed to remain stable.
The Portfolio Exposure page owns the account-wide inventory of positions. Correlation analysis adds information about how those exposures may combine.
Pairwise correlation is not enough for a large portfolio
A portfolio with many positions creates a network of relationships. Looking at only one pair can miss the fact that several positions load on the same factor or that a hedge offsets one risk while adding another.
A useful review can combine pairwise correlations with exposure by asset class, currency, sector or factor. The objective is to identify clusters of positions that may behave like one trade under stress.
Correlation should affect risk budgets
If several positions are likely to lose together, treating each one as if it has an independent risk-per-trade budget can understate total account risk.
The Risk Budgeting framework therefore allocates risk across the portfolio rather than assuming every position can use the full standalone allowance at the same time.
Stress testing is the practical companion to correlation
Historical correlation describes what happened in a sample. Stress testing asks what happens if the relationship becomes worse than the sample suggests.
For example, a portfolio that assumes two assets provide diversification can be shocked under a scenario where both decline together. The Scenario & Stress Testing page owns that forward-looking exercise.
Rolling correlation can show regime changes
Instead of calculating one number over the entire history, rolling windows can show how the relationship evolved through time. A stable-looking full-sample correlation can hide periods of very different behaviour.
The window length changes the result: shorter windows react faster but are noisier, while longer windows are smoother but can respond slowly to a new regime. There is no universally correct window.
Do not build a hedge from correlation alone
A strong historical negative correlation does not guarantee that a hedge will offset losses at the required time or in the required amount. Position size, volatility, basis risk, liquidity and the economic mechanism behind the relationship still matter.
A hedge should therefore be evaluated in actual loss scenarios, not only from a correlation matrix.
Common correlation-risk mistakes
- assuming different tickers automatically provide diversification;
- treating historical correlation as permanent;
- confusing zero correlation with independence;
- ignoring common currency, sector or macro factors;
- using one full-sample coefficient without checking regime changes;
- assuming a historical hedge will remain effective during stress.
A practical correlation review
Identify the main economic drivers of each position, calculate historical relationships over more than one relevant window, inspect periods of stress, group positions that share factors, and test the account under scenarios where expected diversification weakens.
The purpose is not to predict the next correlation number. It is to avoid building a portfolio whose safety depends on one relationship remaining unchanged.
Correlation risk inside the MFXG framework
The parent Risk Management pillar treats correlation as a portfolio control because single-position rules cannot see joint losses. Correlation risk connects position sizing, exposure and stress testing into one account-level view.
Diversification is useful when risks are genuinely different. Correlation analysis helps test whether that assumption is supported rather than merely assumed.