If we assume that the SSN database internally only stores numbers today but could also store hexadecimal values without significant redesigns, I would assume that SSNs are stored as text already. So no matter if you put numbers, hex or text, 9 places will always use 9 bytes (assuming it’s ASCII only and doesn’t support UTF-8 etc.).

As the post implied that the current technical limit is 999,999,999. That very much sounds like a character data type to me. Otherwise, the limit is usually something like 2^x.

If SSNs are stored as numbers today, then hex and text would lead to quite some change. If you go for a re-design, you can as well just increase the length of the field.