What you will learn
- Define what it means for two variables to be independent
- Understand that correlation only measures linear relationships
- See how two variables can be uncorrelated yet completely dependent
- Know that independence implies zero correlation, but zero correlation does not imply independence
This lesson clears up a common misunderstanding in statistics, and it is the kind of thing an ordinary statistics class skips over. People treat zero correlation as if it means two variables have nothing to do with each other. That is not true, and the gap between the two ideas has caught out serious investors. To see why, we first need a clean definition of independence.
What independence means
Two variables are independent when knowing the value of one tells you absolutely nothing about the other. Learning that it rained today tells you nothing about the outcome of a coin flip in another country, so those two are independent. Independence is a strong, complete statement about the relationship: there is no connection of any kind, linear or otherwise. Correlation, as you saw last lesson, is a much narrower measurement. It only captures whether the two move together in a straight-line pattern.
Independence and correlation (Carlos Fernandez-Granda, NYU)
A careful walk through why independence is stronger than zero correlation. Focus on the definition of independence and the distinction from correlation.
Correlation only sees straight lines
This is the heart of the matter. Correlation measures the strength of a linear relationship only. If two variables are related in a curved, non-straight-line way, correlation can completely miss it and report a value near zero, even when the two are tightly connected. Zero correlation means no linear relationship. It does not mean no relationship at all. So the rule is one-directional: if two variables are truly independent, their correlation must be zero, but a correlation of zero does not prove they are independent.
Uncorrelated but completely dependent
Let X take the values negative 2, negative 1, 0, 1, and 2, each equally likely. Let Y equal X squared, so Y is 4, 1, 0, 1, 4. Are X and Y related? Are they correlated?
- Are they related? Y is completely determined by X. If you know X, you know Y exactly. They are as dependent as two variables can possibly be.
- Now check the correlation. The average of X is 0. Because the values are symmetric, every positive X has a matching negative X that produces the same Y, so the ups and downs in the covariance cancel out perfectly.
- The covariance works out to exactly zero. And so the correlation is exactly zero.
Why it matters: The relationship here is a U-shape, a curve, not a straight line. Correlation only detects straight-line patterns, so it reports zero and misses the perfect dependence completely.
Sit with that example for a moment, because it is the whole lesson in one picture. Two variables can be perfectly, completely connected and still have a correlation of zero. Correlation was simply the wrong tool to see their curved relationship.
Uncorrelated but not independent (Pillai)
Reinforces the same point with a clear worked example. Watch how a symmetric, curved relationship produces zero correlation despite full dependence.
Which statement is correct?
Two variables have a correlation of exactly zero. What can you correctly conclude?
Zero correlation means there is no linear relationship. It does not rule out a nonlinear relationship, so you cannot conclude the variables are independent.Why this matters in finance
This is not just a mathematical curiosity. Investors often measure the correlation between two assets and, seeing a low number, assume the assets are unrelated and therefore safe to hold together for diversification. But two assets can have low ordinary correlation during calm periods and still be exposed to the very same hidden risk that only shows up in a crisis. Their relationship is nonlinear: quiet most of the time, then suddenly moving together violently when a shock hits. This is called tail dependence, and it is exactly why some portfolios that looked well diversified on paper collapsed together in 2008. A low correlation number gave false comfort.
Independence says the two have nothing to do with each other. Zero correlation only says they do not move together in a straight line. Confusing the two has cost people fortunes.
Match the claim to its truth
Explain the difference
In your own words, explain why zero correlation does not mean two variables are independent. Use the idea of linear versus nonlinear relationships.
Write an answer before comparing it with the model response.
Model answer
Correlation only measures whether two variables move together in a straight-line, linear pattern. If the relationship between them is curved or otherwise nonlinear, correlation cannot detect it and can report a value of zero even when the variables are tightly connected. So a correlation of zero only rules out a linear relationship. It does not rule out every relationship. Independence is a much stronger condition that says knowing one variable tells you nothing at all about the other. Independence forces correlation to be zero, but zero correlation does not force independence.
You have now mastered how two variables relate: covariance for direction, correlation for standardized strength, and the crucial caveat that zero correlation is not independence. Next we turn to the single most important distribution in all of statistics, the normal distribution.