Matthew Yglesias is irritated by journalists refusal to explain (learn?) statistics.

In a comment I said I was going to average the presidential horse race polls on the first page of polls of
http://www.pollingreport.com with Nader included pref for just kerry vs kerry/edwards. I forgot to add likely voters if availabe otherwise registered voters

I get a confidence interval for the average of 9 polls of about 1.3 %
averages
Kerry 47.6
Bush 44.9

the difference of 2.7 % is statistically significant even though almost all of the polls are "statistical ties" and Bush is ahead in one (see the gallup anomaly).

Without Nader Kerry does better. If all with registered voters Kerry does better.