Margin of error

A great note on Romenesko's letters page from John Maggs:

A good "On the Media" piece on how to understand the margin of error in political polling moves me to challenge what has become a widespread convention in news reporting: when a race is close enough to fall within the margin of error, news reporters say it is a "statistical dead heat" or even that the race is tied. This is wrong. A three point lead for Bush, even in a poll with a four point margin of error, means that there is a small probability that Kerry might actually be tied, or even ahead, but that it is far more likely that Bush is ahead (and most likely that he is ahead by three points.) Polls are an estimate of what millions think based on what a few hundred think, nothing more than a probability. If they have any use, then reporters ought to report a small advantage as an advantage.