Back to MyDD Weblog

Dick Morris has an article out that's worth reading, if only to spark the conversation: The perils of polling. I wrote a long article that included research on the accuracy of polls for the 2000 election a few months ago: Election 2000: The polls were wrong then, has anything changed?

First, as you can see from the chart included in the Election 2000 article, Morris is wrong to state Why were all the national polls (except Zogby) wrong in predicting that George Bush would get more votes than Al Gore in 2000? Harris called the race exactly dead-on. Harris is, without a doubt, the best polling group in the US over the past 25 years, and is the ultimate authority in my book. Morris again: Why were all the polls wrong again in predicting a close Senate race between Rep. Rick Lazio (R) and Hillary Clinton in New York that same year? Once again, the facts show Morris lacking. Though Zogby was surely off in predicting a 5% Lazio win, Quinnipiac polled Clinton over Lazio 51-39 in their final poll, November 1-5, 2000, which proved to be fairly accurate.

The Morris article does bring up some interesting points about Internet polling, and I wasn't a bit surprised to hear him state that internet surveys will become more accurate than telephone interviewing. Morris rightly points out that current poll errors are no doubt due to the skewed demographics of a telephone polls. CRAP polls like Survey USA and POA do nothing to avoid this outcome. If the populace being polled is homogenous, then it doesn't matter as much-- which is why POA did so wonderful in the Republican primaries during 2000. However, a contest like the NJ Senate race this year highlights the faults of CRAP polling. With Forrester having less than 50% name ID among overall voters, yet much higher name ID among conservatives who he targeted during the Republican primary he won, you can see how a poll finding Forrester with a 52-38 (moe 3.8%) takes shape.

It's just as simplistic to say polling doesn't work, as it is to change one's ratings with every poll that comes out. Election 2000 Wrap-up takes a good look at his results, and as you can see, there were some Zogby polls that were off, while others were right on.

One contest in particular, the WA Senate race in 2000, is a good example. The final result was a 49-49 near-tie, which lasted over a week while mailed-in ballots were counted. Zogby's final polling showed Cantwell with a 52-47 lead; in the early November days previously, Zogby polled a Cantwell lead of 50-43; and in a Zogby poll taken Oct. 27-29 found Gorton a 47 percent to 45 percent lead. Obviously, Zogby overshot the momentum that went toward Cantwell. This contest disproved assumptions regarding undecided voters breaking for the challenger, as the surge went for the incumbent Gorton. However, Zogby was well within the polls moe (margin of error) of 4.5 percent. This just goes to show, the MOE in polling cannot be underestimated.

evote.com has a couple of good web pages to follow polls, with their analysis of the contests. Hot Polls looks at some of the recent Senate polling done, and Recent Polls is updated about weekly, and has a good archive. The Progressive Review's Morning Line has the point spreads of all recent US Senate and Governor polling.

Back to MyDD Weblog