In this case: 

 UCSR0A = (1 << U2X0);

`1<<U2XO = 00000010`, `1` is shifted to left to U2XO position so when you give this value to UCSR0A the reister's content will be:

 UCSR0A = 00000010

This operation will set all bits to `0` but the U2X0, for example if RXC0 would have been set to `1`, then this operation would clear it to `0`. If you are not aware of this that could cause some headache later.


----------

In this one:

 UCSR0A = UCSR0A | (1 << U2X0);

which is equal to this:

 UCSR0A |= (1<<U2X0);

Here you set only the second bit to one and leave the rest unchanged which is preferable, because you won't change something accidentally. The content will be:

 // UCSRA0 = xxxxxxxx before operation something is in the register
 UCSR0A = UCSR0A | (1 << U2X0); // OR current with 00000010
 // UCSRA0 = xxxxxx1x something OR 1 will be 1

where `x` is the unchanged/previous value of the bit. If `x` is 1, **1 OR 0** will remain **1** and if `x` is 0, **0 OR 0** will remain **0**.


----------

So I suggest the second way, because that way it is easier to track which bit is set and which not and you won't change previously applied configuration by mistake.