What it means
The intuition is simpler than the formula suggests. For each period you ask how far above or below its own average each of the two variables sat, multiply those two gaps together, and then average the results.
When the two are usually above average together, the products are positive and the covariance comes out positive. The reason this matters in business is that risk is not just about how volatile individual items are; it is about whether they wobble together.
A holiday company whose flight revenue and hotel revenue both collapse in the same bad month faces far greater risk than one whose two lines offset each other, even if each line is equally volatile on its own. Covariance shows up most visibly in investment management, where a portfolio's overall risk depends on the covariance between every pair of holdings, not just on their individual volatility.
This is the mathematical heart of diversification: adding an asset that moves against the rest of the portfolio can reduce total risk even if that asset is individually risky. The main practical drawback is that covariance is measured in whatever units you fed into it, so the number itself is hard to interpret.
A covariance of 6.75 tells you the relationship is positive but says nothing about strength, which is why analysts normally divide it by the two standard deviations to get correlation, a tidy figure between -1 and +1. Outside finance, the same logic helps operations and commercial teams.
Working out the covariance between raw material prices and selling prices, or between headcount and support ticket volume, tells you which pairs of numbers can safely be planned independently and which need to be modelled together.
In practice
Real-world examples
Example 1
A pension trustee reviewing two equity funds finds a strongly positive covariance between them and realises the second fund adds cost without adding much protection. The trustee replaces it with a bond fund whose covariance against equities is close to zero.
Example 2
An airline analyses the covariance between jet fuel prices and passenger revenue and finds it mildly positive, because fuel spikes usually accompany strong economies. That partial offset informs how much of its fuel exposure it bothers to hedge.
Example 3
A retail group calculates the covariance between online and in-store sales across 24 months and finds it clearly negative, indicating channel switching rather than genuine growth. The insight reshapes how the board reads a strong online quarter.
Think of it
Covariance measures whether two things move in the same direction-and by how much.
Formula
Calculation
Sample covariance = sum of ((each x value - mean of x) x (each y value - mean of y)) / (n - 1)
Take five months of returns for two funds. Fund X returns 4%, -2%, 6%, 1%, 1%, which sum to 10% and give a mean of 2%. Fund Y returns 3%, -1%, 5%, 0%, 3%, which also sum to 10% and give a mean of 2%.
The deviations for X are 2, -4, 4, -1 and -1. The deviations for Y are 1, -3, 3, -2 and 1. Multiplying each pair gives 2, 12, 12, 2 and -1, which sum to 27.
With five observations, n - 1 = 4, so the sample covariance is 27 / 4 = 6.75. The positive sign confirms the two funds tend to have good and bad months together, which means holding both delivers less diversification benefit than an investor might hope.Case study
Seen in the real world
The following is an illustrative and fictional example. Cardrew Asset Partners, an invented boutique investment manager, marketed a fund built from eight separate strategies and described it to clients as widely diversified because no single strategy exceeded 15% of the portfolio.
An analyst joining the fictional firm ran the covariance figures across all pairs of strategies and found that six of the eight moved firmly together, because each relied on the same underlying assumption about credit spreads staying calm. Counting positions had disguised the fact that the fund was effectively one bet made six times over.
Cardrew rebuilt the portfolio around covariance rather than position count, cutting to five strategies with genuinely different drivers. Measured volatility fell by roughly a third even though the fund now held fewer names, which is precisely the result the mathematics predicts.
Watch out
Common mistakes
- Reading the size of a covariance figure as a measure of how strong a relationship is, when only the sign is directly interpretable without converting to correlation.
- Assuming a positive covariance proves one variable causes the other, when both may simply respond to a third factor such as the economic cycle.
- Calculating covariance over a short, calm period and treating the result as permanent, when relationships between assets often tighten sharply during a crisis.
Questions
People also ask
What is the difference between covariance and correlation?
Correlation is covariance divided by the two standard deviations, which strips out the units and puts the answer on a fixed -1 to +1 scale.
Should I divide by n or by n - 1?
Use n - 1 when your data is a sample of a larger population, which covers almost every practical business case; n is only for a complete population.
Can covariance be exactly zero?
Yes, and it means there is no linear relationship, though the two variables could still be connected in a curved or conditional way.