Correlation and diversification
Correlation measures how closely two assets move together, on a scale from −1 to +1. Diversification comes from low-correlation holdings, not from many holdings. Twenty technology stocks correlating around 0.8 are one risk written across twenty lines.
- Moving opposite
- One asset tends to rise when the other falls. Rare and valuable, but usually paid for with a lower return expectation.
- Largely independent
- This is where the real diversification effect lives. Total portfolio volatility falls noticeably below the average of the individual volatilities.
- Partly aligned
- Some spreading effect remains. Typical for equities across regions or sectors within the same market environment.
- Effectively one risk
- The positions move almost in lockstep. Adding lines to the portfolio buys no additional spread here.
How it is calculated
ρ(A,B) = covariance(A,B) / ( σ[A] × σ[B] )
Two portfolios, each with 16 % volatility, are blended half and half. How far total volatility falls depends entirely on correlation.
- At ρ = 1.0
- 16.0 %
- At ρ = 0.7
- 14.8 %
- At ρ = 0.3
- 12.9 %
- At ρ = 0.0
- 11.3 %
What the number does not tell you
- Correlations rise exactly when they should fall. In a panic, investors sell everything at once. Assets that sat at 0.3 for years jump to 0.8 in a crash. Diversification partly fails in precisely the phase it was bought for.
- It does not measure the size of the swings. Two assets can be perfectly correlated while one swings three times as hard. How forcefully a move lands is measured by beta.
- A single figure hides the structure. An average correlation of 0.4 can come from pairs all near 0.4, or from one tightly linked block plus a few outliers. Only the matrix shows which picture is true.
- It is no substitute for weighting. Low correlations help little when one position makes up 60 % of the portfolio. Spreading risk needs both: different movements and distributed weights.
Related metrics
How Evergrova calculates it
Correlation to the benchmark sits in the detail section of Performance & Risk alongside R-squared. Correlations between your positions feed the optimiser and the stress test.
Data basis: daily returns of portfolio and benchmark over the common window.
View the demo portfolioCommon questions
How many positions does a diversified portfolio need?
The count is the wrong question. A single world ETF spreads wider than thirty German single stocks. What matters is how differently the parts react to the same events, not how many lines the list has.
What does a negative correlation mean?
That the two assets tend to move in opposite directions. High-grade bonds held that property against equities for a long time, but not in the 2022 rate shock, when both fell together.
Correlation or beta, which do I need?
Correlation answers whether two things move the same way. Beta answers how strongly. For diversification questions correlation is the right figure; for market sensitivity, beta.
Last reviewed: 2026-08-08