Matstat Econometrics Matan - page 23

New comment
 
Is it correct to say that the existence of a cointegration of the two series is equivalent to a high correlation of their values?
 
secret:
Is it correct to say that the existence of a cointegration of the two series is equivalent to a high correlation of their values?
No
 
secret:
Is it correct to say that the presence of a cointegration of the two series is equivalent to a high correlation of their values?

More like incremental correlations.

 
Aleksey Nikolayev:

Correlation and selective correlation are very different things. For example, correlation may well be nonexistent, while sample correlation can be calculated for almost any sample.

The problem is a total misunderstanding of the simple fact that sample correlation is not the definition of correlation (but only an estimate of it, not always accurate).

And what does understanding this fact give us?

We are in a real non-stationary world, not a textbook theorist, spherical in a vacuum)
We are always dealing with a finite sample, and by "correlation" we always mean estimation. Why write the word "estimate" unnecessarily and clutter up the text?
Why do we need a "true, hospital average" correlation calculated from minus infinity to plus infinity? It doesn't happen in the real world, so we don't need it.
 
secret:
And what does understanding this fact give us?

We are in the real non-stationary world, not a textbook theorist, spherical in a vacuum)
We are always dealing with a finite sample, and always by "correlation" we mean estimation. Why write the word "estimate" unnecessarily and clutter up the text?
Why do we need a "true, hospital average" correlation calculated from minus infinity to plus infinity? It doesn't happen in the real world, so we don't need it.

It's just that many people forget that estimating a correlation doesn't mean having one at all.

2a identical processes can have a correlation of zero over the lifetime of the processes. And this should always be taken into account.

 
Valeriy Yastremskiy:


Two identical processes can have a correlation of zero over the lifetime of the processes. And this must always be taken into account.

And how is that?

 
Valeriy Yastremskiy:

Many people simply forget that estimating correlation does not mean that there is correlation at all.

Two identical processes can have a correlation of zero over the lifetime of the processes. And this must always be taken into account.

It is a rare case where the correlation of two assets is constant (and equal to zero, for example). Usually market assets change their "modes of operation", periods of high correlation are followed by periods of low correlation, etc.
It is a natural process, conditioned by life itself, by economic phenomena.
Therefore it does not make sense in most cases to count correlation (and any other metric) throughout life.
 
Dmytryi Nazarchuk:

How's that?

Sine and cosine)
 
secret:
It is an exceptionally rare case where the correlation between two assets is constant (and equal to zero, for example).

There is no such thing at all.

 
secret:
Sine and cosine)

No.

There are sections with positive and negative correlations.

1...161718192021222324252627282930...38
New comment