That will reduce the random error, yes. But there's still the concern of systematic error, which has to be dealt with separately. If you're taking a biased sample to begin with, it doesn't matter how many times you repeat the test; the bias will still be there.