It seems that the only change in the datafile is related to value labels.
Stata 9 now allows value labels to be up to 32,000 characters long and
in different languages. So if your Stata 9 dataset does not take advantage
of these new features, can it still be read by Stata 8 and by data transfer
software like DBMS/Copy and StatTransfer?
-svyset- has a lot more options now, and those get saved with the
dataset. So, perhaps that also creates compatibility problems across
versions? Apparently not, however, since it says at