Given the great inaccuracy of his 2004 state polls, one would think that John Zogby would retreat for a while, assess his methodology, and try to figure out why his polling was so off the mark.
Instead, Zogby's website links to two articles claiming that his polls were accurate! Suffice it to say, the two articles only make passing references to Zogby's last national poll (Bush 48%-Kerry 47%). One of the articles mentions Zogby's crazy prediction on Monday afternoon (that Kerry would win 311 electoral votes), but none mention the incredible inaccuracy of his state polls. Check back soon for a more in depth analysis of just how off his numbers were, but even in the states where Zogby predicted the right outcome, the polls' margins of victory significantly varied from actual vote totals. Zogby would surely note that many of these results were within the margin of error of the polls. However, whenever a pollster is not even close to being right in several different polls, the problem likely goes beyond statistical variance.
A quick glance shows how wrong Zogby's polls were: Zogby was off by 5% in Florida (Zogby forecast: Tie), 3% in Ohio (Zogby forecast: Bush +6), 5% in Colorado (Zogby forecast: Bush +2), 5.5% in Wisconsin (Zogby forecast: Kerry +6), and 6% in Iowa (Zogby forecast: Kerry +5).
Mason Dixon's polls for MSNBC and Knight Ridder seem to have been most accurate, forecasting the incorrect result in only Minnesota. In 2002, Mason Dixon correctly called 22 out of 23 races. Check out this wonderful chart on Mason Dixon's website, comparing the actual results to their poll results.
[To be fair to Zogby, other pollsters like Gallup also had wildly incorrect state polls, such as the Gallup poll showing Bush ahead by 8% points in Wisconsin a few days before the election.]
No comments:
Post a Comment