Or the survey he’s referring to is biased. Seems hard for it not to be… did they knock on doors all across the country? If it’s based on mail or telephone responses, are people who voted for Obama more likely to respond to those?
Or, he’s misquoting the survey. If you were testing the hypothesis that people misremember voting for the winner, wouldn’t you sample a smaller area than the whole country, and then compare your results with the vote count from that area? Why would an experiment like that ever get a number meant to be compared with the whole country’s votes?
Or the survey he’s referring to is biased. Seems hard for it not to be… did they knock on doors all across the country? If it’s based on mail or telephone responses, are people who voted for Obama more likely to respond to those?
Or, he’s misquoting the survey. If you were testing the hypothesis that people misremember voting for the winner, wouldn’t you sample a smaller area than the whole country, and then compare your results with the vote count from that area? Why would an experiment like that ever get a number meant to be compared with the whole country’s votes?