IMHO, two reasons... 1) Stats will not and cannot be created against a table variable by StatMan. 2) Since the content of the table variable is not realized until run time (like a Temp Table) and they don't cause a recompile (which a temp table might), table variables are always evaluated as having only one row by the optimizer which also means that the optimizer may not (and frequently does not) make the best choice when many rows are present in the table variable.

As a side bar, they also make life a bit difficult when troubleshooting code because their scope is not only session sensitive but also run sensitive just like any variable is. With temp tables, you can run the code that populates it once and do continuous trouble shooting against it. With a table variable, you have to run the code that populates the table variable every time which may (usually :-D) require a bit of juggling in the code with commented out code.

There are places where table variables excel compared to temp tables (like where you don't want a rollback to affect the content) but, for the most part, the only reason why I'd ever use a table variable is because you can't use/reference temp tables in a function.

I also avoid ROLLBACKs like the plague (overcoming the "best" reason to use a table variable) because ROLLBACKs are comparatively VERY expensive... that's why you'll also not see much in the line of TRY/CATCH in my code... I make sure the "answer" is known and won't even start a transaction unless I'm absolutely sure (read that as "the code has prevalidated all data") that everything will successfully fly in the transaction.