You own ten software stocks and two technology funds. The holdings list looks reassuringly long. Then big corporate customers announce cuts to software budgets, and nearly every holding falls.
Twelve holdings, one crowded bet.
Before adding another holding, ask what would hurt it. If the answer is the same spending slowdown that hurts the rest, you may be adding more of the same bet. Correlation puts a number on how closely the returns move together.
Count exposures, not ticker symbols
Your asset allocation sets the broad mix. An underlying exposure is something that can affect the holdings inside it: a particular company, an industry, or a shared source of demand. Ten software companies can sell different products while depending on the same corporate spending budgets. Different logos do not create different customers.
Diversification spreads risk across investments, but funds can overlap with each other and with stocks you own directly. Checking the holdings inside a fund reveals repeated names.
Correlation measures how closely two investments' returns vary together across a set of periods. It captures the strength and direction of a linear relationship: how closely the returns fit a straight-line pattern when paired up.
Compare total returns in the same currency over matching periods, such as monthly returns across the same years. A $20 stock and a $200 stock can have identical percentage returns. Their share prices alone tell you nothing about that relationship.
A one-day heatmap is a snapshot, not a correlation estimate.
Read the scale from −1 to +1
The scale runs from −1 to +1. Each value describes returns relative to each investment's own average:
- −1: Perfectly opposite. A's above-average returns pair with B's below-average returns in exact proportion, and vice versa.
- 0: No measured linear relationship. The paired returns show no overall straight-line pattern. Zero does not mean the investments are independent.
- +1: Perfectly aligned. Above-average returns line up with above-average returns, and below with below, in exact proportion.
That number is the correlation coefficient. It describes a partnership; it does not rate either partner's safety.
Even at +1, returns need not be equal or have matching signs. A loss of 5% beats an investment's average loss of 10%: above average can still mean losing money.
Keep the weights and each investment's swing size the same, and lower correlation means a smoother blend. That benefit starts before the number turns negative.
Put two return paths together
Take two made-up investments, A and B, with the annual returns below. Returns include income and are measured in US dollars, before costs, taxes and inflation.
Start every year with half the money in each investment to isolate how their returns move together. This reset keeps the arithmetic at 50/50; rebalancing rules come later.
In year 1: 0.5 × 10% + 0.5 × 10% = 10%.
| Year | Return A | Return B | 50/50 blend |
|---|---|---|---|
| 1 | +10% | +10% | +10% |
| 2 | −10% | +10% | 0% |
| 3 | +10% | −10% | 0% |
| 4 | −10% | −10% | −10% |
In year 2, A's −10% return carries half the weight: −5 percentage points. B contributes +5 points, bringing the blend to 0%.
Across these four years, the calculated correlation is zero. Each investment swings between +10% and −10%, but the blend has two flat years. Year 4 makes the limit clear: zero correlation still allows both investments to lose together.
Change only B's ordering and the result changes:
- Perfect alignment (+1). Give B A's path: +10%, −10%, +10%, −10%. The blend repeats it. Two holdings produce the same swings as one.
- Perfect opposition (−1). Give B −10%, +10%, −10%, +10%. The blend earns 0% every year. Here, equal weights and equally large opposite returns cancel; −1 alone would not flatten every 50/50 blend.
Correlation describes how the pieces fit together. Volatility measures the size of their return swings. An investment with much bigger swings can make your portfolio more volatile even when its correlation with your existing holdings is low. The weights matter too.
There is no magic number of stocks
No universal holding count guarantees adequate diversification. Twenty equally weighted stocks give each name 100% ÷ 20 = 5% of the stock portion of your portfolio.
That spreads the money evenly; it does not make the businesses different. All twenty can still react to the same shock. If a few names take most of the money, the count tells you even less. Position sizing tackles how much money sits in each holding.
Past correlations change with the sample and market conditions. They can rise in a crisis without every pair reaching +1. A relationship measured across calm years may offer less cushioning during a shared shock.
Inflation and interest-rate shocks can hurt stocks and bonds together, as 2022 demonstrated.
For the opening portfolio, the review starts inside the two funds: look for repeated stocks and how much of the money they represent. Then consider the companies with different names. If they sell to the same corporate buyers, a software-budget cut can reach them all.
Another software fund may add names without changing that spending bet. A grocery business draws on household food spending instead. That gives you a different exposure to investigate; correlation describes how differently the returns actually behaved. The useful question is what changes about the things your money depends on.
In short
- A long holdings list can still contain one crowded bet.
- Correlation measures how returns vary together, not how safe either investment is.
- Diversification can reduce swings without negative correlation.
- Holding counts and past relationships do not promise protection from losses.