Correlation
Correlation measures how two assets move relative to each other, on a scale from +1, perfect lockstep, to -1, perfect mirror image, with 0 meaning no relationship at all. It is the raw material of diversification: owning many assets reduces risk only to the degree that their correlations fall short of +1.
The math
Take two hypothetical stocks, each with 15 percent volatility, split 50/50 in a $100,000 portfolio. At correlation +1 the portfolio’s volatility is simply 15 percent: a typical annual swing of $15,000, as if one stock were owned twice.
Lower the correlation and the turbulence shrinks.
| Correlation | Portfolio volatility | Typical annual swing |
|---|---|---|
| +1.0 | 15% | $15,000 |
| 0.5 | 13% | $13,000 |
| 0.0 | 10.6% | $10,600 |
Same two businesses, same weights: the $4,400 of reduced turbulence was manufactured entirely by how differently the two holdings move.
The trap
Trusting trailing correlations as if they were physical constants. They are averages over a lookback window, and they are least reliable at the worst moment: in a genuine crisis, correlations across risk assets lurch toward +1 as everyone sells everything, so the diversification measured during calm markets evaporates precisely when it was supposed to pay.
A portfolio built on a matrix of 0.3s can behave like one big position during the month that matters.
The move
Diversify on causes, not on statistics. Instead of trusting a historical correlation number, ask what would have to happen in the real economy for two holdings to fall together: shared customers, shared suppliers, shared interest-rate sensitivity, shared cycle.
Twelve stocks that all depend on advertising budgets are one bet wearing twelve tickers, whatever the trailing matrix says. A stock picker who spreads holdings across genuinely different demand drivers builds correlation protection that survives a crisis, because it rests on business logic rather than a lookback window.