Java Database Connectivity (JDBC)

I need to be able to store unicode data in my sql server database. As I understand it, the unicode data types ntext and nvarchar and nchar will be twice as large as their non-unicode counterpats text, varchar and char. For anyone else that has had to do this, do you convert ALL your character columns to the unicode types? Or do you try to be efficient and convert ONLY those that NEED to be converted? (this latter option sounds like it might be confusing in the code to have some columns that have this and others that don't). Is it worth the time to leave a handful of columns as non-unicode column types? What is best practices here?

Thanks!

Michael