Cointegration – A Superior Concept to Correlation?

Gunter Meissner

“If you can’t explain it in simple terms, you haven’t understood it very well”

– Albert Einstein

In the previous chapter we looked at the numerous limitations of the Pearson correlation model. In this chapter we discuss the concept of cointegration and evaluate if it is a better suited concept to analyse financial time series and their dependence.


Cointegration is a 2003 Nobel Prize-rewarded concept and goes back to Robert Engle and Clive Granger (1987). Before we analyse it, we have to understand some statistical properties, which are part of Cointegration.

What is a stationary process?

Definition: A stationary process is a process where mean and variance are constant in time (weak stationary) or all moments (strict stationary) are constant in time.

An example of a stationary process is a random drawing from a standard normal distribution εt, which we already encountered in Chapter 3. ε can be computed as =normsinv(rand()) in Excel or norminv(rand) in MATLAB. See for a spreadsheet deriving ε. Drawing 10,000 εt, we derive Figure 4.1.

However, stocks that grow are not stationary, since their mean increases in time. In

To continue reading...

You need to sign in to use this feature. If you don’t have a account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here: