EViews.com

I am running two different variables in EViews using the Cross-Correlogram function and i want to know how this function can tell that independent variable is a leading indicator of the dependent variable. If anyone knows how to read or translate the results from a cross-correlogram i would greatly appreciate the help.

By default, cross correlation displays correlograms from 0 to 36 leads-lags for a pair of series (say y and z). If, for example, z is a leading indicator of y, then you should observe the highest significant correlation at a lag greater than 0. In other words, correlation between y and z(-i) or y(+i) and z, where i>0, should be higher compared to other lags (leads). The reading changes with respect to ordering of variables.

Sorry, have the same question, and it is still not clear.

Let's say it in this way:
Grouping Y and Z, then, open cross-correlation at 5 lag.

Lets say we have the following result:
Date: 14/09/09 Time: 14:40
Sample: 2/01/1996 31/12/2007
Included observations: 3130
Correlations are asymptotically consistent approximations

Y,Z(-i) Y,Z(+i) i lag lead

|* | |* | 0 0.0948 0.0948
| | |*** |1 -0.0359 0.3185
| | |* | 2 -0.0069 0.0560
| | | | 3 0.0207 -0.0019
| | | | 4 -0.0204 0.0130
| | | | 5 -0.0105 0.0045

The question is: In Y,Z(-I) we observe the correlation at a given lags in the assumption where Y led by Z?
Or in Y,Z(-I) we have the assumption that Y lead Z?

file attached: Thank you.

Variable Y should be considered as the leading indicator of variable Z, since the highest correlation occurs at the first lead. In other words, you'll get the most explanatory power if you regress variable Z on the first lag of variable Y: or if you regress the first lead of variable Z on variable Y:

so, If I understand you correctly, column Y,Z(+i) refers to Y leads Z assumption.
And the highest correlation at the first lead is confirmation of causality at lag 1 from Y to Z.

Is it correct?

Thank you.

Your regression adding makes it clear.

But I think, headings in cross-correlation output is not clear. Also, user guide on this question is poor (just formulas, no explanation).

Thank you.

Best wishes,
Ruslan.

Yes, that is correct. However, please keep in mind that this analysis does not confirm the causality but rather determines the time lag between two variables.