What compiler are you using? With gcc 4.4 on 32-bit Linux netbook I get:

                         unpatched  union  shift
utf-16le  " "*10000         129      126    109
utf-16le  "\u263A"*10000    208      203    160
utf-16be  " "*10000         153      147    114
utf-16be  "\u263A"*10000    226      227    167

The difference is that for shift the compiler stores block in register, and for the union the compiler stores block in memory, so that it can get address. May be more recent compilers learned to do this more effectively?

Besides, shifts are more pronounced for CPython code, searching shows very few uses of union in the source code.