Two assets. One shared risk?
Owning several assets does not necessarily mean taking several independent risks. What matters is how their returns tend to move together—and how that relationship behaves when markets are stressed.
Updated 14 Nov 2026 · Original educational material · Hypothetical examples, not live market data
What does correlation mean?
Correlation is a way of describing whether two series of returns tend to move together. A correlation near +1 means strong positive linear co-movement in the sample. Near −1 means they tend to move in opposite linear directions. Near 0 indicates little measured linear association; nonlinear dependence can still exist.
The number is usually calculated from returns observed over a particular period and frequency—not simply from two upward-trending price charts. A historical correlation does not mean one asset causes the other to move, nor does it guarantee the same relationship tomorrow.
How can two portfolios experience different journeys?
Imagine two hypothetical assets, A and B. Over six periods, A's returns alternate between +10% and −10%. We deliberately create two possible behaviors for B. In the first, B copies every return of A. In the second, B has exactly the opposite return each period.
Suppose each portfolio starts at 100 and is rebalanced to 50% A / 50% B at the start of every period, with zero fees. When the two assets have the same return, the whole portfolio changes by that return. When the two returns are +10% and −10%, the equal-weight portfolio has 0% for that period.
Swipe sideways to view both portfolio paths →
How does correlation change portfolio volatility?
For a simple two-asset portfolio, risk can be modeled using each asset's return standard deviation, the portfolio weights and their correlation. The variance formula includes the covariance term: it is how the assets interact that makes portfolio volatility different from simply averaging their separate volatility numbers.
For a second, separate illustration, assume both assets have modeled one-period return volatility of 10% and portfolio weights of exactly 50/50. The following table gives the portfolio volatility implied by different hypothetical correlation inputs.
| Assumed correlation | Modeled portfolio volatility | Interpretation |
|---|---|---|
| -1 | 0.00% | Perfect cancellation in this ideal model |
| -0.5 | 5.00% | Partial variance reduction |
| 0 | 7.07% | Partial variance reduction |
| +0.5 | 8.66% | Partial variance reduction |
| +1 | 10.00% | No variance reduction |
At correlation zero, the implied volatility is about 7.07%, not 5%. At correlation +1 it is 10%; at −1 it is 0% in this symmetric, fixed-weight model. Those endpoints depend on unrealistically exact assumptions and omit costs, changing weights and regime shifts.
Different names do not guarantee independent exposures.
Imagine that an investor holds two assets, each represented as a 50% allocation. If both fall by 10% in the same period, the portfolio also falls by 10% before costs. If one rises 10% as the other falls 10%, a portfolio rebalanced to equal weights before that period returns 0% before costs.
In crypto, two different tokens can still share exposure to the same market-wide funding conditions, liquidity cycle, platform or sector. A low past correlation can increase during a broad selloff. These are practical reasons to investigate concentration and common drivers, not recommendations about which tokens to own.
What diversification cannot fix.
Diversification may reduce idiosyncratic risk from one holding, but it cannot remove risks shared by the entire market. Assets can become highly correlated under stress. Illiquidity can prevent rebalancing at modeled prices. Leverage can force sales before a portfolio recovers, and custody or operational failure can affect several holdings at once.
Measurement choices matter: weekly and daily correlations may differ, as may historical windows. Using an estimated covariance matrix to forecast a portfolio is an additional assumption. A single sample correlation is neither a future forecast nor proof of causality.
Three common mistakes.
“More tickers means safer.”
Several holdings can share the same underlying market, liquidity or operational risks.
“Correlation near zero means independence.”
Zero linear correlation does not rule out nonlinear relationships or joint extreme losses.
“Yesterday's hedge works tomorrow.”
Correlations and hedge ratios can change sharply as market regimes shift.
Frequently asked questions.
Does negative correlation always mean two investments hedge each other?
Not perfectly. The relationship is measured in a specific sample and timeframe and may change. Position weights, volatility magnitudes, costs and extreme events also affect the outcome.
Is zero correlation the same as zero risk?
No. Both assets can lose value, and a correlation estimate near zero only describes a limited linear relationship over the measured sample.
Why can correlations rise during a crash?
Investors may react to common funding, liquidity, macroeconomic and risk-aversion shocks. Relationships estimated in calm periods need not hold during stress.
Does the flat line in the chart mean guaranteed safety?
No. It is deliberately built from exact opposite periodic returns and ideal rebalancing with no trading costs. It is a mathematical illustration, not an achievable market claim.
Can Market History Lab optimize my real portfolio correlation?
Not currently. The site does not ingest actual asset-return histories or provide a correlation optimizer. Related risk tools illustrate conditional assumptions rather than a market portfolio covariance estimate.