Correlation
How closely two investments' daily returns move together, on a scale from −1 to 1.
5 min readLesson 10 of 57Interactive
In short
1 means they rise and fall in lockstep, 0 means no link, and below 0 means they tend to move opposite ways. The lower the correlation between holdings, the more they smooth each other out, which is what makes diversification work.
Why it matters
Owning ten stocks only spreads risk if they don't all fall on the same days. Ten chip makers can drop together on one piece of industry news, so they behave more like one big holding than ten. Correlation is the number that tells you how much two holdings really move as one.
It's the quiet engine behind Diversification. Two investments with the same Volatility can either add their swings together or partly cancel them out, and correlation decides which. It also shows how closely a fund or strategy tracks a Benchmark, and it's half of what makes up Beta.
How it works
Take the daily returns of two investments over the same days. For each day, see how far each return was from its own average. On days when both are above average, or both below, they're moving together; when one is above and the other below, they're moving apart. Correlation averages that agreement and scales it so the answer always lands between −1 and 1.
At 1, the two move in perfect lockstep in direction and proportion. At 0, knowing one tells you nothing about the other. At −1, one rises exactly when the other falls. Most pairs of US stocks land somewhere between 0.2 and 0.7, because the whole market pulls them in the same direction on many days.
Correlation is about the direction of moves, not their size. A calm utility and a wild tech stock can be highly correlated if they tend to rise and fall on the same days, even though one moves far more. To get the size as well, combine it with volatility, which is exactly what beta does.
It matters for a mix because a portfolio's swings depend on how its holdings combine. Two holdings with a correlation of 1 add no diversification at all. Below 1, the mix swings less than the average of its parts, and the lower the correlation, the bigger that saving.
Correlation = covariance of A and B ÷ (standard deviation of A × standard deviation of B)
How much the two move together, divided by how much each moves on its own, so the answer runs from −1 to 1.
A worked example
Over five days, stock A returned +1%, −1%, +2%, −2% and 0%. Stock B returned +2%, 0%, +1%, −1% and −2%. Work out their correlation, then see what it does to a half-and-half mix.
- Averages: A's returns add up to 0, and so do B's, so both averages are 0% and each day's return is already its distance from the average.
- Multiply the two returns on each day and add them up: (1 × 2) + (−1 × 0) + (2 × 1) + (−2 × −1) + (0 × −2) = 2 + 0 + 2 + 2 + 0 = 6.
- Square each stock's returns and add them up: for A, 1 + 1 + 4 + 4 + 0 = 10; for B, 4 + 0 + 1 + 1 + 4 = 10.
- Correlation = 6 ÷ √(10 × 10) = 6 ÷ 10 = 0.6.
- Now suppose both stocks have a volatility of 20% a year. A half-and-half mix has a volatility of 20% × √((1 + correlation) ÷ 2). At a correlation of 0.6, that's 20% × √0.8 ≈ 17.9%.
- At a correlation of 1 the mix would still swing 20%; at 0 it would swing 20% × √0.5 ≈ 14.1%.
Result: The two stocks have a correlation of 0.6: they usually move the same way, but not always. That's enough to cut the mix's volatility from 20% to about 17.9%, without giving up any average return. Less related holdings would cut it further.
Try it
One stock, or many?
One stock swings with 30% volatility; an equal mix of 10 swings with 17%.
Volatility, one stock
30%
Volatility, mix of 10
17%
Worst fall, one vs mix
−24% / −14%
When stocks mostly move for their own reasons, their swings cancel out: a mix of 10–20 is far calmer than any one.
Generated prices for illustration, not market data.
How to read it
- Above 0.8
- The two move almost as one. Holding both adds little diversification; think of two funds that track the same index, or two big banks.
- 0.3 to 0.8
- The usual range for two US stocks, or a stock against the S&P 500. They share the market's big moves but have plenty of their own.
- Around 0
- Little link between their daily moves. Pairs like this smooth a portfolio the most, if each has a decent return of its own.
- Below 0
- They tend to move opposite ways. Rare between stocks; high-quality government bonds have often, but not always, played this role against stocks.
In the real world
- Fund managers look at a correlation matrix of their holdings, a grid with every pair's correlation, to spot clusters that would fall together. Pension funds use the same idea when they split money between stocks, bonds and other assets.
- Correlations change, and usually rise in a crash. In sharp sell-offs like those of 2008 and early 2020, many stocks fell together on the same days, so diversification helped less exactly when it was needed most. The correlation of US stocks with US government bonds has also flipped between positive and negative over the decades.
- Correlation with the market is how investors judge whether a fund is really doing something different. A fund with a correlation of 0.99 to the S&P 500 is close to an index fund, whatever its fees suggest.
Common mistakes
- Reading correlation as cause. Two stocks moving together may both be reacting to the same news, not to each other.
- Assuming a low correlation will hold in a crisis. Correlations measured in calm years tend to understate how much things fall together in a panic.
- Confusing correlation with similar returns. Two stocks can be highly correlated while one gains 30% a year and the other 5%; correlation compares the direction of daily moves, not how far each went in total.
- Trusting a correlation from a few weeks of data. Short windows are noisy; a year of daily returns is a more sensible minimum.
See it in MarketLab
Stock pages →
The Risk card shows each stock's correlation with the S&P 500 over the last year, next to its beta. Compare Apple with a steadier company like Coca-Cola (KO).
Portfolio risk analysis →
Under Risk analysis, the Risk tab's correlation map groups holdings that move together, names the most correlated pairs and counts how many independent bets they add up to.
Backtest →
Under the results, open More measures to see a strategy's correlation with its benchmark: close to 1 means it mostly moved like the thing it's compared with.
Test yourself
Optional: 4 quick questions on the words, using the idea, and how it plays out in real markets.
Loading your answers…