haxxiy on 17 November 2010
If I remember correctly, a thousand people survey is usually enough to be within 5% of reality if you choose spots well. So the problem here is with the sample being located in a single place, not it's size...







