Correlation and Diversification for Global Macro Economics

Ask what is moving together, and under which conditions

When do global assets move together, and when does that relationship change? Correlation summarizes how two measured series have co-moved over a chosen sample; it does not establish that one caused the other or that the same pattern will persist. Specify the assets, return or yield measure, currency, observation frequency, and dates before interpreting a coefficient.

A correlation calculated over a calm expansion may describe a different environment from one spanning an inflation shock, recession scare, or abrupt policy repricing. A single full-period average can conceal those shifts. Rolling windows reveal changes but introduce their own sensitivity to window length and exceptional observations.

Separate growth and inflation channels

Cross-asset relationships often reflect competing macro forces. Stronger expected activity may support earnings-sensitive assets while raising expected policy rates; an inflation surprise can lift nominal yields even as it changes growth expectations. In another episode, weakening demand and lower inflation pressure may dominate. The same pair can therefore show different signs across regimes without either calculation being erroneous.

Government bonds and equities, for example, are not dependable opposites. Their observed relationship depends partly on whether investors are responding chiefly to growth risk, inflation risk, or changing discount rates. Commodity prices also vary by product: energy, industrial inputs, and food have different supply structures and demand sensitivities.

• Compare nominal yields with real yields and inflation compensation where available.
• Distinguish broad equity indexes from sector or country exposures.
• Check whether currency translation changes the relationship seen by a local investor.

Correlation is not a complete measure of shared risk

A low historical correlation does not mean two positions cannot lose value together during a severe shock. Correlation describes linear co-movement in the selected sample; it may understate nonlinear exposures, tail losses, or common dependence on funding conditions. A portfolio can also be concentrated in one macro factor even when its holdings appear diversified by label.

Use covariance, volatility, scenario losses, and economic exposure maps alongside correlation. Ask whether different holdings rely on the same assumptions about real rates, credit availability, trade, or energy costs. A statistical relationship can help describe diversification, but it cannot replace a look-through assessment of what drives each exposure.

Make comparisons regime-aware

For a useful comparison, show the full-sample estimate beside selected subperiods and explain why those dates matter. Avoid choosing breakpoints only because they produce a preferred result. Note missing observations, different market holidays, data frequency, and whether returns are measured in local currency or a common base currency.

Treat a changing correlation as evidence to investigate rather than a forecast. The more robust conclusion is often conditional: these series have tended to move together under the specified conditions, but their shared drivers and sensitivity can change. This is educational information, not individualized investment, tax, legal, or accounting advice.