I was just thinking about this, having switched over from coffee to tea this
morning. It struck me as sort of weird. 

Let us suppose that one had a datatype defined for social security number, this
would essentially be a regular expression specifying how a SSN should be
written. 

now a ssn has more information implicit in it than just the SSN. There is
information that one could use to find out what state issued the number for
example, in the prefix.
Let us suppose then that we had a definition for a structure which splits the
SSN into two elements. The combined strings of both elements matches the
datatype, this is something I would like to be able to specify. 

Why not specify just a type that says a SSN type has two elements? 

because in some formats that would be overkill. 

Perhaps I am uninformed however, can anyone think of any particular schema
language one can do this in, and if you are the person who knows of such a
language can you give me an example if possible. (not that it's something I
need
to do, just something I thought would be extremely useful to be able to do at
some point)