Agreed. And banks make a fortune selling correlation products - only because those correlations are not stable. But they are used for many strategies that simply do not work. I can think of just a few - usdcnh vs usdmxn, gold vs many fiat currencies (but even then timing was everything..)
What Dalio got wrong about correlation is that the analysis of it has stages, all of which require understanding of causality. It is not one or the other. Initially, asset prices correlate with macro economic themes, and if the macro is predicted, you profit from the first phase of correlation. Then predicting how & why correlation will begin to fade, breakdown, is the second phase of alpha to be extracted. And then finally, playing for reversals when over positioning on early correlation relationships occurs, becomes a third phase of analysis. Causality of the correlation is the analytical work, correlation merely defines the expected asset proxy move. It's not about stability. You have to predict correlation directionally just like any other variable.
This is a great one, Jens. Thanks for sharing!
Agreed. And banks make a fortune selling correlation products - only because those correlations are not stable. But they are used for many strategies that simply do not work. I can think of just a few - usdcnh vs usdmxn, gold vs many fiat currencies (but even then timing was everything..)
perhaps one thing to remember, though, is in a crisis, all correlations go to 1 (-1), which I have observed consistently over the past 40 years
What Dalio got wrong about correlation is that the analysis of it has stages, all of which require understanding of causality. It is not one or the other. Initially, asset prices correlate with macro economic themes, and if the macro is predicted, you profit from the first phase of correlation. Then predicting how & why correlation will begin to fade, breakdown, is the second phase of alpha to be extracted. And then finally, playing for reversals when over positioning on early correlation relationships occurs, becomes a third phase of analysis. Causality of the correlation is the analytical work, correlation merely defines the expected asset proxy move. It's not about stability. You have to predict correlation directionally just like any other variable.