Well, without getting too technical (and if somebody out there wishes to get more technical, be my guest), I would characterize intuitively as what is going on as a more careful parsing of how multiple variables relate with each other through time. In that regard cointegration has a technical superiority over Granger "causality" in that it is better at ferreting out the multiple interactions over time of several variables. If somebody else wishes to improve on that, feel free to do so.
Again, however, if one is interested in "true causation" rather than "mere correlation" cointegration does not solve the problem for you. It merely is better at clarifying what is connected with what in time. But, as already noted, we do not have this implicit false philosophical claim that occurs with the Granger test because of its label.
There are competitors to both Granger and cointegration tests, but I think most on this list would not be interested in a full blow-by-blow account of this as this is a long and gory and ultimately very technical topic. Anybody really interested can go see James Hamilton's _Time Series Analysis_, Princeton University Press, which is a very long book. It is a graduate level textbook, excellent, and state-of-the-art, but not easy going.
It is, however, the case that cointegration is very current and viewed as about as technically correct and precise as anything out there for getting at such relationships in time between variables. Barkley Rosser -----Original Message----- From: Doug Henwood <dhenwood at panix.com> To: lbo-talk at lists.panix.com <lbo-talk at lists.panix.com> Date: Monday, March 22, 1999 8:28 PM Subject: Re: Junk Econometrics [was School vouchers]
>J. Barkley Rosser, Jr. wrote:
>
>> I would agree that cointegration is currently considered
>>to be somewhat more advanced and sophisticated than
>>Granger tests. I would say that they are subject to most of
>>the same limitations that Granger tests are. One thing that
>>they avoid is that they are not labeled with a philosophically
>>loaded and provocative name like "causality" which has
>>caused more trouble and fuss over the Granger tests than
>>they are worth.
>
>Ok, so what's the story on cointegration?
>
>Doug
>