When Correlations Break: Diversification Failing in a Crisis

[Fairman]

When Correlations Break: Diversification Failing in a Crisis

Correlations between currency pairs, and between currencies and other assets, are not fixed. In crises, they often jump.

diagram.png (37.89 KiB) Viewed 34 times

During normal times, you might hold positions in several pairs and believe they diversify one another. Under stress, investors sell risk assets broadly and buy safe havens. Many positions then move together, as risk sentiment dominates individual fundamentals. The correlations rise toward one, and diversification shrinks exactly when you need it.

In March 2020, for example, the dollar surged against most currencies and stocks fell sharply, with many typical relationships distorted by a scramble for dollar liquidity.

Other patterns also reverse. A currency that was a safe haven in one episode may fail in another, depending on the cause of stress.

Implications for risk management are important. Do not rely on historical correlations to estimate worst-case risk. Stress-test your positions by assuming that correlated pairs move together, and that stops may slip.

Consider the total exposure by theme, such as risk-on versus risk-off, not only by currency.

Use position limits and total risk caps that remain valid when correlations go to one.

Hold more cash, or lower leverage, during periods of heightened uncertainty.

After events, review how your positions behaved, and update your assumptions.

Practical step: calculate your total loss if every open position hit its stop at once, and compare with your risk cap.

[Fairman]

The everything-hits-its-stop calculation is a good gut check. If that number scares you, your total risk is too high, no matter how diversified it looks. Add up the risk of every open position as if all stops trigger at once, then add a bit for slippage, maybe 20 percent extra since stops slip in fast markets. With five trades at 1 percent each, that's about 6 percent of the account in a bad hour. If losing that in one afternoon would make you change your plan or panic, reduce the number of open trades or the risk per trade. Diversification shows up most when you don't need it.

[Fairman]

Grouping positions by theme instead of by pair is a much better view. Long AUD, long NZD and short JPY are three tickets but one risk-on position. Common themes are risk-on versus risk-off, dollar strength, and commodity currencies. I write each open trade under a theme and cap each theme, say 2 percent of total risk at most. Then a fourth trade that looks like a new idea but sits in an already full theme gets skipped or reduced. During a crisis these themes can merge too. In March 2020 nearly everything became a dollar trade, so check the totals by currency as well.