Lesson 6 of 9
Correlation is not cause
Return on equity and return on invested capital correlate at 0.74 across 2,001 US listings. Neither causes the other, and the reason is in the formulas.
By Javier Sanz
Founder of ValueScreener. Built Ninety Nine, a securities services provider, and sold it to a tier-one global bank; ran operations and internationalization at Alpaca, the brokerage API.
Who writes thisEditorial processLinkedIn (opens in a new tab)X (opens in a new tab)
In one minute
- Correlation measures whether two series move together, on a scale from minus one to one, and nothing more.
- Three explanations always compete with causation: the arrow points the other way, something drives both, or it is chance.
- Two ratios built from the same accounting line will correlate whatever the business does, which is neither a finding nor a coincidence.
Two numbers moving together is one of the weakest observations that can be made about them, and it is routinely presented as one of the strongest. The useful habit is to have the alternative explanations ready before the correlation arrives.
Two correlations from the same market, on the same day
- Return on equity against return on invested capital0.736
- Net margin against gross margin0.099
What the number is
A correlation coefficient runs from minus one to one. One means the two move together in perfect lockstep, minus one means perfectly opposite, and zero means no linear relationship at all. It measures only straight-line association: two series related by a curve can correlate at nearly zero while being completely determined by each other.
The strong one, explained without causation
Across 2,001 active US listings holding both figures, return on equity and return on invested capital correlate at 0.736. That is a strong relationship and it is entirely unmysterious.
Both ratios put a measure of profit on top. Return on equity divides it by shareholders' equity; return on invested capital divides it by equity plus debt. The numerators are close relatives and the denominators overlap. A company that earns well scores well on both, because both are largely measuring the same thing.
Nothing causes anything here. Two formulas share most of their inputs, so their outputs move together, and a correlation of 0.736 is what shared construction looks like.
The three rivals, named
The weak one, which is also instructive
The same day, net margin and gross margin correlate at 0.099 across 2,604 listings. Almost nothing. That is genuinely surprising: gross margin is a component of the arithmetic that produces net margin, so the naive expectation is a strong relationship.
The explanation is that everything between the two lines, operating costs, interest, tax and one-off charges, varies far more across companies than the gross margin does. A high gross margin does not travel down the income statement on its own. That is a real finding about US company accounts and it comes out of the correlation being low rather than high.
What to do with a correlation
Read it as a question, not an answer. A high one says two series move together and asks why. A low one says they do not, which is equally informative and is almost never published.
And never treat a correlation between two ratios as evidence about a business until you have checked whether the two ratios share a line item. That check takes a minute, and it dissolves a large share of the correlations quoted in this industry.
Check yourself
Two ratios correlate at 0.74. What is the first thing to check?
- Whether the sample is large enough
- Whether the two ratios share inputs in their formulas
- Whether the correlation is statistically significant
Show the answer
Answer: B.
Shared construction produces correlation mechanically. Checking the two formulas takes a minute and explains a large share of the correlations reported between accounting ratios.A correlation of exactly zero between two series establishes what?
- That the two are unrelated
- That there is no linear relationship between them
- That one causes the other with a lag
Show the answer
Answer: B.
Correlation measures straight-line association only. Two series bound together by a curve, such as a U shape, can correlate at nearly zero while one fully determines the other.Net margin and gross margin correlate at only 0.099 across 2,604 US listings. What does that suggest?
- That one of the two figures is computed wrongly
- That what happens between the two lines varies more across companies than gross margin does
- That gross margin is not part of net margin
Show the answer
Answer: B.
Gross margin is arithmetically upstream of net margin, so a weak correlation means the costs, interest, tax and charges in between dominate the variation. The low number is the finding.Sources
- Correlations computed across active US equity listings, read from this product's own store on 16 September 2026: return on equity against return on invested capital over 2,001 listings holding both within plus or minus 100 percent, and net margin against gross margin over 2,604 listings within their stated bounds.
What this leaves out
- Both correlations are computed on bounded ranges, chosen to stop a handful of extreme values dominating the coefficient. A different bound produces a different number, and the bound is stated rather than hidden.
- A correlation on one date across a cross-section is not a correlation through time. Two series can be related across companies and unrelated within one company over its history.
- These count listings, not deduplicated companies.
- Nothing here is advice about any security.