How big of a random sample would have to be taken for the sample to predict with 99% confidence who would be elected president if everyone voted?