More like comparing FG% as computed by a drunk guy sitting in the 300 level to FG% as computed by a stoner listening to the game on the radio.
But comparing two polls from the same pollster (but different dates) can be useful.
And, although one can criticize the methodology, 538.com is an attempt to predict the election outcome based on extrapolation from polling data.
barfo
Very good. Comparing the same pollster's polls does give you a trend. But you should probably still consider the methodology - having black callers ask questions vs. white ones may give you a different answer this week, or rewording the questions, and that kind of thing.
Speaking of polls, I'm no big fan of Ann Coulter or anything, but she writes about polls today, and the factual part is interesting:
http://www.anncoulter.com/cgi-local/printer_friendly.cgi?article=278
EIGHTY-FOUR PERCENT SAY THEY'D NEVER LIE TO A POLLSTER
by Ann Coulter
October 15, 2008
With an African-American running for president this year, there has been a lot of chatter about the "Bradley effect," allowing the media to wail about institutional racism in America.
Named after Tom Bradley, who lost his election for California governor in 1982 despite a substantial lead in the polls, the Bradley effect says that black candidates will poll much stronger than the actual election results.
First of all, if true, this is the opposite of racism: It is fear of being accused of racism. For most Americans, there is nothing more terrifying than the prospect of being called a racist. It's scarier than flood or famine, terrorist attacks or flesh-eating bacteria. To some, it's even scarier than "food insecurity."
Political correctness has taught people to lie to pollsters rather than be forced to explain why they're not voting for the African-American.
This is how two typical voters might answer a pollster's question: "Whom do you support for president?"
Average Obama voter: "Obama." (Name of average Obama voter: "Mickey Mouse.")
Average McCain voter: "I'm voting for McCain, but I swear it's just about the issues. It's not because Obama's black. If Barack Obama were a little more moderate -- hey, I'd vote for Colin Powell. But my convictions force me to vote for the candidate who just happens to be white. Say, do you know where I can get Patti LaBelle tickets?"
In addition to the social pressure to constantly prove you're not a racist, apparently there is massive social pressure to prove you're not a Republican. No one is lying about voting for McCain just to sound cool.
Reviewing the polls printed in The New York Times and The Washington Post in the last month of every presidential election since 1976, I found the polls were never wrong in a friendly way to Republicans. When the polls were wrong, which was often, they overestimated support for the Democrat, usually by about 6 to 10 points.
In 1976, Jimmy Carter narrowly beat Gerald Ford 50.1 percent to 48 percent. And yet, on Sept. 1, Carter led Ford by 15 points. Just weeks before the election, on Oct. 16, 1976, Carter led Ford in the Gallup Poll by 6 percentage points -- down from his 33-point Gallup Poll lead in August.
Reading newspaper coverage of presidential elections in 1980 and 1984, I found myself paralyzed by the fear that Reagan was going to lose.
In 1980, Ronald Reagan beat Carter by nearly 10 points, 51 percent to 41 percent. In a Gallup Poll released days before the election on Oct. 27, it was Carter who led Reagan 45 percent to 42 percent.
In 1984, Reagan walloped Walter Mondale 58.8 percent to 40 percent, -- the largest electoral landslide in U.S. history. But on Oct. 15, The New York Daily News published a poll showing Mondale with only a 4-point deficit to Reagan, 45 percent to 41 percent. A Harris Poll about the same time showed Reagan with only a 9-point lead. The Oct. 19 New York Times/CBS News Poll had Mr. Reagan ahead of Mondale by 13 points. All these polls underestimated Reagan's actual margin of victory by 6 to 15 points.
In 1988, George H.W. Bush beat Michael Dukakis by a whopping 53.4 percent to 45.6 percent. A New York Times/CBS News Poll on Oct. 5 had Bush leading the Greek homunculus by a statistically insignificant 2 points -- 45 percent to 43 percent. (For the kids out there: Before it became a clearinghouse for anti-Bush conspiracy theories, CBS News was considered a credible journalistic entity.)
A week later -- or one tank ride later, depending on who's telling the story -- on Oct. 13, Bush was leading Dukakis in The New York Times Poll by a mere 5 points.
Admittedly, a 3- to 6-point error is not as crazily wrong as the 6- to 15-point error in 1984. But it's striking that even small "margin of error" mistakes never seem to benefit Republicans.
In 1992, Bill Clinton beat the first President Bush 43 percent to 37.7 percent. (Ross Perot got 18.9 percent of Bush's voters that year.) On Oct. 18, a Newsweek Poll had Clinton winning 46 percent to 31 percent, and a CBS News Poll showed Clinton winning 47 percent to 35 percent.
So in 1992, the polls had Clinton 12 to 15 points ahead, but he won by only 5.3 points.
In 1996, Bill Clinton beat Bob Dole 49 percent to 40 percent. And yet on Oct. 22, 1996, The New York Times/CBS News Poll showed Clinton leading by a massive 22 points, 55 percent to 33 percent.
In 2000, which I seem to recall as being fairly close, the October polls accurately described the election as a virtual tie, with either Bush or Al Gore 1 or 2 points ahead in various polls. But in one of the latest polls to give either candidate a clear advantage, The New York Times/CBS News Poll on Oct. 3, 2000, showed Gore winning by 45 percent to 39 percent.
In the last presidential election the polls were surprisingly accurate -- not including the massively inaccurate Election Day exit poll. In the end, Bush beat John Kerry 50.7 percent to 48.3 percent in 2004. Most of the October polls showed the candidates in a dead-heat, with Bush 1 to 3 points ahead. So either pollsters got a whole lot better starting in 2004, or Democrats stole more votes in that election than we even realized.
