Because the original design assumed that 16 bits are enough to encode everything worth encoding, hence UCS2 (not UTF-16, yet) being the easiest and most straightforward way to represent things.

Ah ok. Well even then, you end up spending 16 bits for every ASCII character.