We would need a sample size of 560.
We first calculate the z-score associated. with this level of confidence:
Convert 95% to a decimal: 95% = 95/100 = 0.95
Subtract from 1: 1-0.95 = 0.05
Divide by 2: 0.05/2 = 0.025
Subtract from 1: 1-0.025 = 0.975
Using a z-table (http://www.z-table.com) we see that this is associated with a z-score of 1.96.
The margin of error, ME, is given by:

We want ME to be 4%; 4% = 4/100 = 0.04. Substituting this into our equation, as well as our proportion and z-score,