Correlation

Correlation is the statistical measure of how two investments' returns move together, scaled from +1 for identical direction through 0 for no relationship to -1 for opposite. It is the number underneath diversification: a portfolio's risk falls below the weighted average of its parts only when the parts are less than perfectly correlated, and at +1 there is no diversification at all. The classic portfolio assumption is that stocks and bonds are negatively correlated, so bonds cushion stock drawdowns. That assumption held for roughly two decades and failed in 2022, when inflation pushed rates up and the Bloomberg US Aggregate fell 13% while stocks fell 18%.

Correlations are unstable, which is the part the averages hide. They converge upward in crises, exactly when the diversification is needed most: the old line is that everything correlates to 1 in a crash, and 2008 and March 2020 both behaved accordingly. They also drift with the rate regime, so gold's correlation to stocks has swung between negative 0.2 and positive 0.4 across decades, and international stocks correlate around 0.8 to US stocks. The number is a long-run average, not a guarantee for the year the portfolio actually needs it.

The arithmetic: correlation is computed from the co-movement of periodic returns over a window, normalized to run between -1 and +1. Worked once: a 60/40 portfolio with 16% stock volatility, 6% bond volatility, and a 0.2 correlation has a combined volatility near 9.9%, against a weighted average of 12%. The gap is the diversification benefit, and it exists entirely because 0.2 is below 1. At correlation 1 the weighted average is the portfolio, and the benefit is zero.

The stock-bond correlation is the load-bearing number in most US portfolios, and its history is the cautionary tale. It ran negative from the late 1990s through 2021, near negative 0.3 on five-year windows, which is the period that built the 60/40's reputation. In 2022 it flipped positive as inflation drove the Federal Reserve's rate hikes, bonds repriced directly, and stocks compressed on the discount rate. The Aggregate fell 13%, stocks fell 18%, and the 60/40 had its worst year in modern history with no cushion anywhere.

The instability cuts both ways. Crisis convergence means hedges sell with everything else: correlations between international stocks, high-yield bonds, and commodities all rose toward 1 in the 2008 drawdown. The window changes the number, with five-year and twenty-year calculations disagreeing by half on the same pair. The honest use is as a long-run planning input with a crisis haircut, not as a law.

Correlation is where diversification is actually decided, and most investors never measure it: they count funds instead. Five overlapping index funds can be one position with five fee lines, and a sector fund plus a broad fund can be two-thirds the same portfolio. The number of tickers says nothing; the correlation matrix says everything.

I watched a colleague named Brent (yes, that Brent) hold five funds he called diversified, and the pairwise correlations ran 0.9 and above across all of them. When 2022 arrived, all five fell together and the diversification he believed he owned did not exist. He had bought five tickets to the same ride, which is what the matrix would have told him for free.

The highest-leverage thing most Americans can do is check the correlation between what they own rather than the number of funds they own, and treat the stock-bond number as a regime-dependent input rather than a constant. Measure it with the correlation calculator and size the mix with the portfolio diversification calculator. Diversify by measurement, not by fund count.