When you hear polling numbers before an election, the margin of error is usually stated as +/- 3% based on a sample size of 1000 likely voters. The 3% (=.03) number is computed using the formula

k = margin of error = .03 = 1.96 * , where

p = .5, q = .5, and n =1000.

If we want to have a margin of error of about 4.5%, what should the sample size of likely voters be instead of n = 1000?