I wrote the following (the unitalicized part) to a friend regarding my obsession for political statistics, particularly as revealed by poll data. Polling is a little like the biblical line about "see[ing] through a glass [meaning "mirror" in the KJV of the Bible] darkly." To change the metaphor, polling is like a somewhat blurry snapshot, one that's taken ultimately a day or two before an event such as an election. Every hour thereafter, the picture changes slightly but doesn't get reflected in the polls because surveying has ended. So, prepare to stay up late to find out the true results.
It's amazing how many people have connections to Pennsylvania. Bill McAllister of KTUU-TV has links to Johnstown, home of (ugh) John Murtha. In PA politics, there is something called the "T," which refers to the fact that nearly every county other than Allegheny and Philadelphia vote Republican in statewide elections. The "T" reflects the final "shape" of the statewide vote.
The vote in Pittsburgh and Philly goes in a very lopsided way to Democrats (80% in Pgh., more than that in Philly). In 2004, the networks announced at 8 p.m. (polls close then) that Kerry had won 61 to 39 percent. By 2 a.m., the vote was 51-49 Kerry.
Jay Costa on the superb Horserace Blog (pure Internet) announced that Kerry had won PA by less than Bush had won Ohio, which was not yet "declared" by most of the networks, who were responding to Kerry campaign requests NOT to declare. The next morning Kerry conceded Ohio to Bush, and the election was over.
I hope Jay Costa (something of an Illinois version of Adam Brickley) does his wonderful blog in the next election. He is/was a graduate student in statistics at the University of Chicago. Jay almost went ballistic during the day when the exit polls were supposedly showing that Kerry had won the election.
He pointed out -- and I don't believe I'd ever heard it before -- that Democrats vote earlier than Republicans, partly because some of them (like my beloved folks in Pittsburgh and Philly) are not employed and have lots of time to cast ballots. That's the kind of stuff Karl Rove knows by heart. In short, it's probably a good idea to doubt the early exit polls.
One thing you''ll hear a lot about polls is the difference between registered voters and "likely" voters. Historically, the Gallup Poll has a lot to do with this distinction. Registered voters often don't . . . vote. If you get a driver's license in PA (and other states) you're registered. You may have a lot of interest in driving but none in voting. (Remember the famous old lady who said, "Vote? I never vote. It only encourages them."
On the "likely" voters, they're generally people who have voted in the past. However, in most elections, only 60% to 75% of likely voters will in fact . . . vote. Republicans tend to do better with "likely" voters because GOP stalwarts vote more frequently than Democrats. Gallup has tended to skew its totals slightly more to Republicans because of their tendency to vote more often.
In 2004, Gallup came under heavy criticism from Democrats for the practice mentioned. In a panicky response, it changed its statistical template so as to pare down the "skew" toward Republicans. That completely screwed up the poll data.
If you have a good memory, you may recall polls that said undecided voters were "breaking" for Kerry by a ratio of eight-to-one. One problem: that never happened. If anything, undecided voters (at least the relative few who voted) "broke" for Bush, who won the popular vote comfortably.
You'll hear a great deal about "undecided" voters. In fact, these are often people who remain undecided unto eternity. Apparently, the reason they stay "undecided" is that many of them are mostly uninterested in the election. Thus, undecided is a euphemism for bored-by-the-whole-thing. They're not really the kingmakers they appear to be.
In the days before the election, Gallup showed Kerry doing well in Florida, a state that went handily to Bush. Perhaps because of that, Kerry skipped re-visiting Florida -- he sent John Edwards -- and made a couple more trips to Cleveland.
Gallup also said in the last days before the election that Kerry was ahead by four points in Ohio and Bush ahead by four in Pennsylvania! My comment then was, "Perhaps they got the states mixed up?" (As expected, Kerry won Pennsylvania and Bush won Ohio.)
In Florida, Jay Costa showed on election eve that Bush was doing well in areas that has grown in population, while Kerry was winning in areas either with no growth or with population declines. If Kerry could have ramped up his support somewhat in Florida, he might have won that state and the election.
My Internet friend Larry Perrault, a fervent and articulate backer of Mike Huckabee, insists that the MSM (and the pollsters) are manipulating public opinion, and I always tend to disagree (slightly) with him.
What happens in very early polls, such the current ones, is that the surveyers are really reflecting name recognition. Larry's candidate, Mike Huckabee, doesn't have good name recognition -- especially outside the early primary states. So, the surveys are showing Mike with relatively low numbers. When the results get published, that tends to keep Mike's numbers down. (http://larryperrault.blogspot.com/)
For a less-well-known candidate like Mike, there's no easy answer to this dilemma. One partial answer is for Mike's supporters to continue what Larry calls "the clamor." That refers to the noise (advocacy) on the blogs, where Mike has many talented, committed advocates. If someone like Mike can move up slightly in the national polls, that could establish a level of recognition (and momentum) that would drive up his numbers even more. In short, some good numbers can lead to even better numbers.
I find all this stuff fascinating. On the key issue of whether polls ever intentionally "slant" their findings: I can't answer that question with certainty. However, I'm one of the people who have serious questions about the Zogby Poll, run by the brother of the president of the Arab-American Association. The Zogbys don't do a good job disguising their preference for Democratic candidates. Zogby also does Internet polls (as opposed to face-to-face or telephone), which seem to be chronically unreliable.
Here I've gone and written an entire column without mentioning Governor Sarah Heath Palin, Alaska's favorite daughter and my favorite candidate.
Stephen R. Maloney, Ambridge, PA
I'll be writing this week about John Edwards' decline in the polls (down to 8%), as well as the same thing happening to John McCain (down to 7%). The latter is losing support and a significant portion of it seems to be going to Rudy Giuliani. Fred Thompson may be looking hard at the "numbers" to determine if there's a plausible scenario where he could win the nomination. (I don't believe such a scenario exists.) I'm also going to write about the question of abortion. Mitt Romney appears to be adjusting his views (surprise!) on abortion by saying that the issue should be returned to the states. Surprisingly, pro-lifer Sam Brownback agrees with him up to a point. Many pro-life people seem to believe that overturning Roe v. Wade will somehow outlaw abortion (in most of its forms?). They are wrong. In the unlikely event Roe is overturned -- which it will not be if a Democrat is elected -- it would return the issue to the states. That's something Romney and Brownback (and we can throw in Giuiliani) do agree would happen. Brownback wants to go further and propose a constitutional amendment outlawing abortion (in all cases? in many cases?). The chances of such an amendment passing Congress by a big margin (and getting approval from the vast majority of states) seems very, very unlikely (read: impossible). Some red states (Utah, Idaho, Nevada, and perhaps some others (Arizona?) would oppose an amendment banning abortion in all cases, including rape, incest, and life of the mother.) At some point this week, I'll jump into this bear-pit.