Why a Correlation of 0.89 Can Mean Almost Nothing

If someone tells you a company's revenue correlates with the country's GDP at 0.89, it sounds like an iron law — when the economy moves, so does the firm. Yet that number often means far less than it appears. Between two quantities that both simply grow over time, a high correlation is nearly guaranteed — even when the link between them is weak.

This is not a theoretical example. In the very panel our Bulgarian-firm data comes from, the correlation between GDP and one company's revenue is 0.889 in levels — but it drops to 0.41–0.53 once you measure it on growth rates. This article explains why, and how not to be fooled.

Correlation in 30 seconds

Correlation is a number between −1 and +1 describing how much two quantities move together. +1 is perfect in one direction, −1 perfect in the opposite, 0 no link. On the common scale 0.89 is "very strong". And here is the trap: a strong correlation does not mean a strong causal link, nor even a meaningful one.

The trap: a shared trend

Picture two things that have only grown over the past 16 years — say Bulgaria's GDP and the revenue of almost any surviving firm. Both start from a low base and end high. Plot them against each other and the points will fall on an almost straight line — simply because both rise with time. The correlation will be near 1, even if their real link is weak.

This is what's called spurious correlation. The classic examples are absurd — cheese consumption and the number of drownings supposedly correlate because both grow with population. The mechanism is the same: a common third factor (time, growth) drives both.

Levels versus growth rates

The fix is to strip out the trend. Instead of comparing levels (how big revenue is, how big GDP is each year), you compare growth rates — the percentage change from the previous year. Now you ask the real question: "when the economy speeds up by 1% above its usual pace, does revenue speed up too?" — not "are both higher than they were 10 years ago?".

The effect on the numbers is dramatic:

Same data, different question — and the "iron law" of 0.89 shrinks to a moderate link. 0.4–0.5 still means something (the economy really does influence revenue), but it is a long way from destiny.

Three questions before you trust a correlation

• Levels or changes? If the number is on raw values that both grow with time, discount it. Ask for the correlation of the growth rates.
• Could both simply be trending? Anything that grows with inflation, population or time will correlate with anything else that grows. A shared trend imitates a link.
• Does it hold within the period? A strong link that falls apart when you split the period in two is probably an artefact, not a regularity.

Why it matters for an investor

"This stock correlates 0.9 with gold / oil / GDP" is one of the most common misleading phrases in investment marketing. It is almost always on levels. The same goes for backtests that look perfect because two rising assets drift up together.

When you read a financial statement or a metric like the Altman Z'' in the financial health calculator, remember that ratios and changes say more than absolute levels. And when you watch how revenue reacts to the cycle, watch it through the phase the economy is in via growth rates, not levels.

What this does NOT mean

• Correlation ≠ causation. Even a clean correlation of growth rates does not prove one causes the other.
• Not a recommendation. This article is about statistical literacy, not advice on what to buy.
• The numbers are a guide. 0.41–0.53 comes from a specific sample — it is not a universal constant.

Continue from here

• Financial health calculator — ratios, not raw levels.
• What happens to companies in a recession: half change phase — the cycle in action.
• How to read a financial statement — where the numbers come from.
• Company cycle profile — revenue and cash flows through the cycle.

Subscribe

Subscribe to the newsletter and get new calculators and articles straight to your inbox — no signals, no spam.