Margin of error, error of margin, whatever

In this NYT/CBS poll has a big Obama lead, 53-39%, a whopping 14 point margin, and of course it might be true. Senator Obama is said to look plenty presidential these days.

However, there are several internals in the poll that raise questions about its accuracy. For example, the poll says that 17% of its respondents didn’t vote in 2004, when turnout was by far at an all time record, and that is curious indeed. That aspect of the current poll would appear to be suspect, unless there are an extraordinary number of new voters this November, which remains to be seen. Moreover, the D/R split was 36-28%, which one would think would generate a +8% for any Democrat from the get-go. We’re not saying the poll is wrong; we don’t know. But how could any Democrat not do very well in this poll, given the sample selected?

One Response to “Margin of error, error of margin, whatever”

  1. Tom Hoverman Says:

    Regarding “margin of error” in polls, can you tell me why when a poll shows Candidate A with a 5-point lead in a poll with a margin of error of 5 points that the media always refer to this as a “virtual tie?”

    Wouldn’t it be just as true to describe Candidate A as having a “virtual 10-point lead?” Why the bias toward the middle?

Leave a Reply