Margin of error accounts for random sampling variation only and says nothing about a bad sampling frame, leading questions, or non-response. A biased survey with a huge sample has a tiny margin of error and a worthless estimate — options claiming a bigger sample cures a flawed method test exactly this.