Correlation and cointegration - You need to think about both when looking at connections across assets

When thinking about cross-market relationships the go-to statistic is correlation. Discuss diversification and the go-to again is correlation, but correlation is just one tool that is often overused especially when looking at time series and trading relationships across markets. A deeper yet critical concept for any portfolio tool is cointegration which focuses on commonality across prices not returns. Cointegration is not a new concept and is actively used by many quant firms, but the key concepts are not usually discussed by most portfolio managers. Markets can be correlated but not cointegrated and cointegrated markets may not be correlated. 

The best analogy used for cointegration is the "The Drunk and Her Dog". The wandering of the owner and dog on their way home may seem individually random, but the two are connected on their walk with corrections if one moves too far away from the other. Prices that are cointegrated will not be free to wander like random variables.  Cointegration measures the long-term co-movement in prices. The time series of two assets can have high correlation, but one series can have an upward trend drift and thus not be cointegrated. If you are expected that the prices will move together or mean-revert you will be disappointed.   

Information on cointegration augments any hedging analysis beyond correlation. Cointegration describes how asset prices are tied together, their tendency for mean reversion, and whether there is a common stochastic trend. 

Two asset price series are cointegrated if there is a linear combination between the two assets which is stationary. What is applicable for two series can also be applicable for a set of asset prices. There can be a cointegrating vector which defines common trends that will makes the system of markets stationary. Cointegration analysis can lead to some form of error correction modeling as a representation of the link across markets which is critical when looking at systems of asset prices. 

If someone starts talking about trading based on correlation of assets and how markets may move together and does not mention anything about cointegration raise your risk antenna. They may not be giving you the whole story, or they may not have done all of their homework. In either case, ask more questions.